Pharmacokinetic optimization of immunosuppressive therapy in thoracic transplantation: part I

Abstract

Although immunosuppressive treatments and therapeutic drug monitoring have significantly contributed to the increased success of thoracic transplantation, there is currently no consensus on the best immunosuppressive strategies. Maintenance therapy typically consists of a triple-drug regimen including corticosteroids, a calcineurin inhibitor (cyclosporine or tacrolimus) and either a purine synthesis antagonist (mycophenolate mofetil or azathioprine) or a mTOR inhibitor (sirolimus or everolimus). The incidence of acute and chronic rejection and of mortality after thoracic transplantation is still high compared to other types of solid organ transplantation. The high allogeneicity and immunogenicity of the lungs justify the use of higher doses of immunosuppressants, putting lung transplant recipients at a higher risk for drug-induced toxicities. All immunosuppressants are characterized by a large intra- and inter-individual variability of their pharmacokinetics and by a narrow therapeutic index. It is essential to know their pharmacokinetic properties and to use them for treatment individualization through therapeutic drug monitoring (TDM) in order to improve treatment outcome. Unlike the kidneys and the liver, the heart and the lungs are not directly involved in drug metabolism and elimination, which may be the cause of pharmacokinetic differences between patients from all these transplant groups.

TDM is mandatory for most immunosuppressants, and has become an integral part of immunosuppressive drug therapy. It is usually based on trough concentrations (C₀) monitoring, but other TDM tools include the area under the concentration-time curve over the dosing interval (AUC_0-12) or over the first 4 hours post-dose (AUC_0-4), as well as other single concentration-time points, such as the concentration 2 hours after dosing (C₂). Given the peculiarities of thoracic transplantation, a review of the pharmacokinetics and TDM of the main immunosuppressants used in thoracic transplantation is presented in this article. Even more so than in other solid organ transplant populations, their pharmacokinetics is characterized by wide inter- and intra-individual variability in thoracic transplant recipients. The pharmacokinetics of cyclosporine in heart and lung transplant recipients has been explored in a number of studies, but less is known about that of mycophenolate and tacrolimus in these populations, while there are also hardly any studies on the pharmacokinetics of sirolimus and everolimus. Given the increased use of these molecules in thoracic transplant recipients, their pharmacokinetics deserves to be explored more in depth. There is very little data, some of which is conflicting, on the practices and outcomes of the TDM of immunosuppressants after thoracic transplantation. The development of sophisticated TDM tools dedicated to thoracic transplantation are awaited, in order to evaluate accurately and precisely patients’ exposure to drugs in general and in particular, to immunosuppre ssants. Finally, large cohort TDM studies definitely need to be conducted in thoracic transplant patients, in order to identify the most predictive exposure indices, and their target values, and to validate the clinical usefulness of improved TDM in these conditions.

Introduction

Heart and lung transplantations are life-saving interventions for patients suffering from end-stage cardiac or pulmonary failure. Many fields have been improved since the first transplantations of these organs (in 1963 for the lung and 1967 for the heart): optimal patient selection, donor management, organ preservation, histocompatibility testing, surgical techniques, post-operative management and immunosuppressive treatment and monitoring have all significantly contributed to this progress.^[1] In the past 20 years (1986 to 2005), a large increase has been observed in the number of heart (from 2,158 to 3,095 per year) and lung (from 1 to 2,169 per year) transplantations in adults, as reported by the Registry of the International Society for Heart and Lung Transplantation (ISHLT).^[2,3] The survival of heart and lung transplant recipients has improved in parallel, with 1-year rates of 90 and 80%, respectively, and 5-year rates of approx. 70 and 50%, respectively.^[2,3]

Heart and lung transplantations are not comparable with the transplantation of other solid organs in terms of morbidity and mortality. First, contrary to graft failure in kidney transplant recipients which can often be treated by dialysis, graft loss in heart or lung transplant patients generally results in death. Secondly, the lung is a highly allogeneic and immunogenic organ because it is in direct contact with the environment (airborne organisms, polluting particles) through respiration, and the lung tissue is highly lymphoid.^[4] Thirdly, unlike the kidneys and the liver, the heart is not directly involved in drug metabolism and elimination, other than by providing blood supply to the other organs, while the lungs are rich in metabolism enzymes that participate in the elimination of certain drugs and are characterized by a very large contact surface area between the alveoli wall and the blood stream. These peculiarities may be the cause of pharmacokinetic (PK) differences between patients from all of these transplant groups.

The increasing success of thoracic transplantation is largely attributable to the development of effective immunosuppressive regimens.^[5] The strategies used in thoracic transplantation were initially derived from clinical trials performed in other solid organ transplant settings, such as kidney transplantation. However, there is no consensus on immunosuppressive strategies after heart^[2] or lung^[3] transplantation to date. Maintenance therapy typically consists of a triple-drug regimen including corticosteroids, a calcineurin inhibitor (CNI, i.e. cyclosporine or tacrolimus) and either a purine synthesis antagonist (mycophenolate mofetil –MMF– or azathioprine –AZA) or a mTOR inhibitor (sirolimus or everolimus). Between 2002 and 2006, the maintenance regimen of approximately 75% of heart or lung transplant recipients at both 1 and 5 years after transplantation was comprised of a calcineurin inhibitor plus a purine synthesis antagonist. More precisely, the combination of tacrolimus and MMF has now become the most widely used combination internationally.^[2,3]

Despite the improvement of immunosuppressive strategies, the incidence of acute and chronic rejection and of mortality after thoracic transplantation remains high compared to other types of solid organ transplantation. In 2007, 35 to 45% of heart^[2] and 40 to 50% of lung^[3] transplant recipients were treated for acute rejection during their first year post-transplantation. Moreover, many studies in lung transplantation have shown that acute rejection is a significant risk factor for the development of chronic rejection.^[6,7] Chronic allograft rejection (in the form of cardiac allograft vasculopathy in heart, and of bronchiolitis obliterans syndrome in lung transplantation) and infections are the main threats to long-term survival and quality of life.^[2,3,8–10] Therefore, further improvements in the treatment of heart and lung transplant recipients must be sought.

Optimal immunosuppression is essential to maintain a viable allograft. The well-known large intra- and inter-individual variability of the PK of the immunosuppressive agents as well as their narrow therapeutic index represent a challenge to the clinicians, who need to select the best treatment and the best dosage for a given patient. Knowing the PK properties of these drugs and using them for treatment individualization through therapeutic drug monitoring (TDM) is thus crucial to improve treatment outcome. Since the introduction of cyclosporine, TDM has become an integral part of immunosuppressive drug therapy, because of the high correlation between systemic exposure and clinical outcomes.^[11] The pre-dose concentration (so-called trough level or C₀) is routinely used for individualizing the dose of most immunosuppressive drugs. For cyclosporine, the area under the concentration-time curve over the dosing interval (AUC_0-12) or over the first 4 hours post-dose (AUC_0-4) were also identified as good, or even better predictors of outcome in renal transplant patients.^[12,13] TDM based on cyclosporine concentration 2 hours after dosing (C₂) has also been proposed, because it has been shown to be highly correlated with the AUC_0-4 and with the incidence of acute rejection in kidney and liver transplant recipients.^[14,15] Based on these observations, TDM has become mandatory for cyclosporine. It is also mandatory for tacrolimus, sirolimus and everolimus, for which it is usually based on C₀ monitoring. On the opposite, MMF is usually administered at a fixed dose, possibly adapted on clinical signs of inefficacy or toxicity. However, there is increasing evidence in renal transplantation that MMF TDM should also become standard clinical practice, particularly in the early post-transplantation period.^[16]

Heart and lung transplant patients display particularities when compared to the other solid organ transplant patients. First, lung transplant recipients usually require higher maintenance immunosuppression as compared to recipients of heart, liver or kidney grafts because of the high immunogenicity of the lung tissue, putting them at a higher risk for immunosuppressant-induced toxicities. Secondly, given the potential PK differences between populations, the PK of immunosuppressants may be different in heart and lung transplant recipients from that of liver or kidney transplant recipients and thus deserves to be specifically studied. For instance, heart and lung transplant patients often suffer from gastroparesis as a result of their initial pathology (e.g., cystic fibrosis) or of the transplantation procedure, which may modify the absorption and thus the PK profile of drugs.^[17,18] More generally, the lung transplant population is very heterogeneic as a consequence or the diverse indications for lung transplantation, which is also in favour of individual therapy and dose adjustment. Finally, close TDM in thoracic transplant patients may be justified given the high morbidity and mortality rates in these populations.

The aim of the present article was therefore to review the state of the art on the PK and TDM of the main immunosuppressants used in thoracic transplantation, and to determine what fields need to be further explored in future PK and TDM studies in these conditions in order to improve transplantation success and patients survival. All numerical results presented in this review are expressed as mean±SD, unless otherwise specified.

In preparing this review, the authors conducted a search of Medline (January 1975 to June 2008) for English-language articles using the following search terms: “cyclosporine”, “cyclosporin”, “tacrolimus”, “calcineurin inhibitor”, “mycophenolate”, “mycophelonic acid”, “MMF”, “MPA”, “sirolimus”, “rapamycin”, “everolimus”, “mTOR inhibitor”, “heart transplantation”, “lung transplantation”, “thoracic transplantation”, “pharmacokinetics”, “metabolism”, “drug interactions”, and “therapeutic drug monitoring”. This search was supplemented by a bibliographic review of all the relevant articles. Given the notably high number of articles on cyclosporine, only those considered as fair to high quality were kept for the review. Those rejected either concerned a very limited number of patients, or only reported mean C0 values in a given population, without connection with clinical data or other exposure indices. On the contrary, as there were only a few articles on tacrolimus, MMF and mTOR inhibitors, almost all were considered. Overall, around 80% of the original list of cited articles were included in this review.

1. Calcineurin inhibitors

According to the ISHLT database, calcineurin inhibitors represent the cornerstone of immunosuppression after thoracic transplantation.^[2,3]

1.1. Cyclosporine

Cyclosporine penetrates into the cells and forms complexes with intracytoplasmic proteins (cyclophilins), inhibiting calcineurin, hence IL-2 production and T-cell activation.

Cyclosporine has become an essential component of standard treatment after heart or lung transplantation, improving drastically patients’ survival and quality of life. In 2007, the ISHLT registry indicated that 21 and 26% of all lung transplant recipients received cyclosporine as part of their maintenance immunosuppressive regimen at 1 and 5 years post-transplantation, respectively.^[3]

1.1.1. Pharmacokinetics

Cyclosporine PK can be affected by drug formulation (Neoral^® vs Sandimmune^®), drug-drug interactions, pre-existing conditions (age, ethnic origin, pathology, diet, activity of metabolic enzymes and drug transporters), clinical status (time after transplantation, gastro-intestinal motility and bile flow, blood lipoprotein content, haematocrit, etc.), potentially altering blood concentrations and total exposure, and thus the degree of immunosuppression.

The vast majority of PK studies performed in thoracic transplantation were descriptive, using model-independent PK methods, where mean and standard deviation of exposure indices, PK parameters and, to a lesser extent, factors of variability in the evaluated population were calculated from individual data (Tables I and II).^{[4,6,19–28]} Some authors used the iterative two-stage (ITS) method, for which they either set up their own PK models or used already existing PK models in order to determine individual exposure indices and PK parameters and then to calculate mean PK parameters in the population.^[29–32] To our knowledge, only one population PK (popPK) study was performed in heart transplant recipients^[33] and one in heart-lung transplant recipients (Table III).^[34]

Table I. Cyclosporine exposure indices in cohort studies.

Results are expressed as mean±SD, unless otherwise specified

Ref	Number of patients	Post-transplant period	Cyclosporine dose		Co-administered immunosup-pressant	Assay	Measured AUC (h.μg/L)	C0 (μg/L)	C2 (μg/L)	Cmax (μg/L)	Tmax (h)
HEART TRANSPLANTATION
[23]	16 adults (out of 35 in the global trial)		NEO		N/R	FPIA TDx	Non-compartmental PK analysis; AUC0-12 Trap
		Week 1	200±58 mg				AUC/D: (h.μg/L)/mg cyclosporine	C0/D: (μg/L)/mg cyclosporine		Cmax/D: (μg/L)/mg cyclosporine
							6,521±1,940	329±119	--	1,136±315	1.9±0.6
							AUC0-12/D = 34.2±12.6	C0/D = 1.8±0.8		Cmax/D = 6.0±2.1
							r2 C0-AUC0-4 = 0.16; r2 C2-AUC0-4 = 0.6
		Week 12	159±40 mg				6,214±1,760	287±99	--	1,368±369	1.8±0.4
							AUC0-12/D = 39.4±7.3	C0/D = 1.9±0.6		Cmax/D = 8.8±1.8
							r2 C0-AUC0-4 = 0.4; r2 C2-AUC0-4 = 0.4
		Week 52	162±37 mg				6,111±1,627	284±74	--	1,256±379	2.2±0.6
							AUC0-12/D = 38.4±10.3	C0/D = 1.8±0.5		Cmax/D = 7.9±2.2
							r2 C2-AUC0-4 = 0.7
[29]	20 male adults	Stable 36.8±27.1 months	NEO and SAND 1.14±0.4 mg/kg TID		N/R No corticosteroids	FPIA TDx	Non-compartmental PK analysis; AUC0-8 Trap
	58.8±10.2 years		SAND				3,278±682	297±62	--	732±178	2.63±1.21
			NEO				3,798±757 (p < 0.001)	312±49 (ns)	--	935±250 (p < 0.001)	1.36±0.49 (p < 0.001)
[19]	47 adults	Stable	NEO + SAND(1)		N/R	EMIT	Non-compartmental PK analysis; AUC0-12 Trap
				D (mg/kg/day)
	54±12 years	49±19 months	None	3.90±1.03		SAND	3,655±1,120	167±77	--	827±204	2.27±0.47
			n = 11			NEO	4,911±935 (p < 0.05)	179±64 (p < 0.02)		1,147±307 (p < 0.01)	1.55±0.52 (p < 0.001)
	52±8 years	61±15 months	DILT	2.98±0.64		SAND	3,605±1,079	147±79	--	S: 816±183	2.09±0.54
			n = 11			NEO	4,747±923 (p < 0.05)	171±61 (p < 0.02)		1,288±415 (p < 0.01)	1.45±0.69 (p < 0.05)
	52±9 years	32±15 months	KETO	0.79±0.27		SAND	3,601±933	194±72	--	480±164	2.61±0.96
			n = 13			NEO	3,703±772 (ns)	189±54 (p < 0.02)		532±93 (ns)	2.00±0.57 (p < 0.05)
	47±12 years	28±5 months	DILT + KETO	0.90±0.54		SAND	4,070±1,090	201±68	--	574±192	2.58±0.90
			n = 12			NEO	4,369±993 (ns)	181±62 (p < 0.02)		652±204 (p < 0.01)	2.33±0.49 (ns)
[58]	47 adults	Stable	NEO(1)		N/R	EMIT	Non-compartmental PK analysis; AUC0-12 and AUC0-5 Trap
				D (mg/day)
			None	302±68			AUC0-12 = 4,912±935	179±64	1,088±165	1,147±307	1.5±0.5
			n = 11				AUC0-5 = 3,356±641			(CV = 27%)	(CV = 33%)
							r2 C0-AUC0-5 = 0.710 (p = 0.001); r2 C2-AUC0-5 = 0.197 (p = 0.17)
			DILT	243±55			AUC0-12 = 4748±924	171±61	1,022±237	1,222±501	1.4±0.7
			n = 11				AUC0-5 = 3,304±655			(CV: 41%)	(CV: 50%)
							r2 C0-AUC0-5 = 0.650 (p = 0.005); r2 C2-AUC0-5 = 0.650 (p = 0.005)
			KETO	66±35			AUC0-12 = 3,703±824	201±48	503±110	503±110	2.2±0.5
			n = 14				AUC0-5 = 1,980±414 (vs “none”, p < 0.05)			(CV: 22%)	(CV: 23%)
							r2 C0-AUC0-5 = 0.176 (p: N/R); r2 C2-AUC0-5 = 0.870 (p = 10−4)
			DILT + KETO	87±60			AUC0-12 = 4,386±1,082	179±68	627±209	659±224	2.4±0.5
			n = 12				AUC0-5 = 2,406±675 (vs “none”, p < 0.05)			(CV: 34%)	(CV: 21%)
							r2 C0-AUC0-5 = 0.022 (p: N/R); r2 C2-AUC0-5 = 0.898 (p = 10−4)
[21]	15 adults 48±10 years	Clinically stable 5.2±4.4 years	NEO 2.7±0.7 mg/kg/day D adjusted on Cav = 200–300 μg/L		ATG AZA Prednisolone	FPIA TDx	Non-compartmental PK analysis; AUC0-12 Trap
							1,285±367 Cav = 371±123	162±55	--	1,101±326	1.1±0.4
							r2 C0-AUC0-12 = 0.60; r2 C2-AUC0-12 = 0.94; r2 C4-AUC0-12 = 0.95
[22]	22 adults	Until Year 4	NEO(2)		ATG Prednisone	EMIT	Non-compartmental PK analysis; 2 consecutive periods
		Month 1	Group I D adjusted on C0		MMF D = 1g BID		--	197±98	1199±476	--	--
		Month 3					--	222±60	1202±587	--	--
		Month 6					--	218±83	999±467	--	--
		Month 12					--	--	664±203	--	--
		Month 24					--	--	593±208	--	--
		Month 36					--	--	561±147	--	--
		Month 1	Group II D adjusted on C2		MMF D = 1.5 g BID		--	--	809±160 (p = 0.02)	--	--
		Month 3					--	--	644±178 (p = 0.003)	--	--
		Month 6					--	--	665±169 (p = 0.02)	--	--
		Month 12					--	--	616±221 (ns)	--	--
		Month 24					--	--	464±234 (ns)	--	--
		Month 36					--	--	451±164 (ns)	--	--
[31]	14 patients de novo	3 PK profiles/patient	NEO D adjusted on C0		N/R	FPIA TDx	Exposure indices and PK parameters estimated using ITS method, Ciclo 1.3 software Absorption model based on a gamma distribution
							AUC0-12/D: (h.μg/L)/mg cyclosporine	C0/(100-mg D): (μg/L)/100 mg cyclosporine	C2/D: (μg/L)/mg cyclosporine	Cmax/D: (μg/L)/mg cyclosporine
		Week 1 (W1)	200±61.2 mg BID				6,395±1,921	310±134.6	C2/D = 5.47±2.33	1,128±299	1.78±0.46
							AUC0-12/D = 34±13	C0/(100-mg D) = 156±70		Cmax/D = 6±2
		Month 3 (M3)	152±36.0 mg BID (vs W1, p<0.05)				5,780±1,365	270±66.7	C2/D = 7.78±1.05 (vs W1, p < 0.05)	1,380±371	1.63±0.40
							AUC0-12/D = 37±8	C0/(100-mg D) = 163±34		Cmax/D = 9±2 (vs W1, p < 0.05)
		Year 1 (Y1)	164±40.1 mg BID				5,967±1,600	284±74.2	C2/D = 6.98±2.17 (vs W1, p < 0.05)	1,216±410	1.22±0.41
							AUC0-12/D = 38±10	C0 (100-mg D) = 163±48		Cmax/D = 8±2
							Bayesian estimator (AUC0-12), best sampling times: C0, C1, C3 for the three periods BE vs Trap: bias% = −0.20 to +3.06, ns; RMSE% = 1.61 to 18.59%
LUNG TRANSPLANTATION
[24]	11 adults All CF SL	> 6 months Medically stable			N/R	SRIA	Non-compartmental PK analysis; AUC0-12 Trap – Crossover study
			Capsules (1 week)				4,164±1,467	--	--	613±242	3.64±2.16
			NEO (1 week)				5,318±1,670	--	--	931±458	2.27±1.2 (p = 0.23)
			Ratio NEO/Capsules				1.48 (p = 0.047)	--	--	1.91 (p=0.045)	--
[4]	20 adults 1 CF	Stable, > 3 months	NEO, D (mean, range) = 330 mg/day (150-50) D adjusted on C0:		ATG (n = 6) MMF Steroids	FPIA Mc	Non-compartmental PK analysis; AUC0-4 Trap; data are presented as mean (range)
	8 SL, 12 DL	< 1 year	M1-M3	350–400			3,700 (2,400–5,700)	322 (248–492)	1,180 (890–1,700)	1,057 (303–2,337)	--
		n = 10	M4-M12	300–350
		1–3 years	M13-M24	N/A			2,440 (1,030–3,600)	222 (68–403)	780 (380–1,500)	929 (401–1,360)	--
		n = 8	M25-M36	250–300
		> 3 years	> M36	150–200			N/R (990–2,050)	N/R (134–143)	N/R (270–670)	N/R (333–804)	--
		n = 2
							r2 C0-AUC0-4 = 0,64; r2 C1-AUC0-4 = 0,44; r2 C2-AUC0-4 = 0,85; r2 C3-AUC0-4 = 0,67; r2 C4-AUC0-4 = 0,64
[66]	14 adults 8 SL, 6 DL	Stable	NEO Steady-state		N/R	FPIA TDx	Non-compartmental PK analysis; AUC0-12 Trap
							5,183±1,835	--	--	--	--
[6]	12 adults 53.2±7.4 years	Weeks 1 & 2	NEO D adjusted on C0 = 300–400 μg/L		Daclizumab Prednisolone MMF	EMIT	Non-compartmental PK analysis; AUC0-6 Trap (interpolation of C5 values)
							3,443±1,451	361±118	C1 = 481±231 C2 = 682±314 C3 = 715±347 C4 = 658±271 C6 = 571±260	--	--
							r2 C0-AUC0-6 = 0.31 (p = 0.001); r2 C1-AUC0-6 = 0.73 (p < 0.001); r2 C2-AUC0-6 = 0.88 (p < 0.001) r2 C3-AUC0-6 = 0.89 (p < 0.001); r2 C4-AUC0-6 = 0,84 (p < 0.001) ; r2 C6-AUC0-6 = 0.54 (p = 0.001)
HEART AND LUNG TRANSPLANTATION
[27]	22 adults(3)	Month 1	D adjusted on C and clinical data		AZA Prednisolone	Pc NSRIA	Non-compartmental PK analysis
	11 CF		16.7±7.2 mg/kg/day				--	892±385 CF vs non-CF: p = 0.58	--	--	--
	11 non-CF		8.2±1.9 mg/kg/day (p<0.01)
[26]	9 patients(4) All CF 12–32 years	Stable 10 months – 7.5 years	Switch from SAND to NEO		N/R	SRIA	Non-compartmental PK analysis; AUC0-12 Trap; data are presented as median (range) Crossover study; 9 PK profiles for each cyclosporine formulation
			D adjusted on C0 mean±SD median (range)				AUC/D: (h.μg/L)/mg cyclosporine	Cmin/D: (μg/L)/mg cyclosporine		Cmax/D: (μg/L)/mg cyclosporine
			NEO 12.6±6.7 mg/kg/day 450 mg/day (125–500)				6,032 (3,931–11,939)	Cmin = 224 (131–393)	--	1,360 (729–2,360)	1.0 (1.0–4.0)
							AUC0-12/D = 19,76 (8,73–47,97) (CV = 51%)	Cmin/D = 0.60 (0.29–2.37)		Cmax/D = 4.53 (1.62–7.69)
								CV on C0: Intra-ind. 36% Inter-ind. 79%
			SAND 8.5±3.6 mg/kg/day 450 mg/day (125–1000)				4,179 (3,377–7,833)	Cmin = 172 (85–443)	--	595 (402–1,065)	3.0 (0.0–6.0)
							AUC0-12/D = 9,29 (4,92–29,61) (CV = 63%)	Cmin/D = 0.43 (0.17–3.22)		Cmax/D = 1.32 (0.65–3.22)
								CV on C0: Intra-ind.45% Inter-ind.130%
			NEO/SAND				1.77 (0.56–4.35)	1.58 (0.41–9.71)	--	2.17 (0.78–8.51)	--
[28]	50 adults 9 CF	f/up: 12 months	D adjusted on C0 (μg/L) = M1-M2: 300–400 M3-M12: 200–300		RATG AZA Prednisolone	EMIT	Non-compartmental PK analysis ; 289 PK profiles ; AUC0-6 Trap AUC0-6 CF vs non-CF: ns AUC0-6/D: CF < non-CF (p<0.001)
	19 SL, 9 DL 22 HL	Week 2	NEO n = 28				3,582–6,822	--	--	--	--
		Week 4					4,601–9,536	--	--	--	--
		Month 6					2,955–6,478	--	--	--	--
		Month 9					3,698–6,365	--	--	--	--
							r2 C0-AUC0-6 = 0.671; r2 C2-AUC0-6 = 0.962; r2 C6-AUC0-6 = 0.651
		Week 2	SAND n = 22				735–5,526, p < 0.001	C0/D, NEO vs SAND: ns	C2/D: NEO > SAND (significant)	--	--
		Week 4					1,256–8,199, p < 0.001				--
		Month 6					550–7,839, p < 0.001		C6/D, NEO vs SAND: ns	--	--
		Month 9					586–9,400, p < 0.001			--	--
							r2 C0-AUC0-6 = 0.574; r2 C2-AUC0-6 = 0.890; r2 C6-AUC0-6 = 0.608
[25]	20 patients	Clinically stable Median:	NEO D adjusted on C0 = 250–350 μg/L		N/R	EMIT	Non-compartmental PK analysis; AUC0-12 Trap
	10 CF 6 DL, 4 HL	21 months					AUC0-12 = 7,639±3,353 to 8,388±2,859	267±96 to 269±85	1,219±802 to 1,603±614	1,664±824 to 1,874±539	1.40±0.39 to 2.11±1.39 (CF & non- CF)
							AUC0-4 = 3,876±2,876 to 4,481±1,311
							r2 C2-AUC0-12 = 0.76; r2 C3-AUC0-12 = 0.87 r2 C2-AUC0-4 = 0.90; r2 AUC0-4 – C3 = 0.81
	10 non-CF 5 DL, 5 HL	36.5 months					AUC0-12 = 7,202±1,610 to 7,447±1,870	300±392 to 346±112	1,372±337 to 1,545±530	1,710±482 to 1,862±572	1.40±0.39 to 2.11±1.39 (CF & non- CF)
							AUC0-4 = 4,081±1,135 to 4,348±1,262
							AUC/D: CF < non-CF
							r2 C2-AUC0-12 = 0.66; r2 C3-AUC0-12 = 0.82 r2 C2-AUC0-4 = 0.78; r2 C3-AUC0-4 = 0.63
[32]	19 adults	Stable (no evidence of rejection within previous 3 months)	NEO D adjusted on C0 = 250–350 μg/L		N/R	EMIT	PK parameters estimated using ITS method, CICLO 1.3; absorption described by a gamma distribution 3 consecutive PK profiles (P)/patient within 5 days Parameters standardized to a 100-mg dose
							AUC/D: (h.μg/L)/mg cyclosporine
	9 CF 6 L, 3 HL		250±76 mg BID (175–425)				AUC0-12	P1: 254±92	P1: 1641±638	P1: 1,952±732	P1: 1.47±0.24
							P1: 7.97±2.46	P2: 258±87	P2: 1544±621	P2: 1,977±479	P2: 1.33±0.34
							P2: 8.33±3.25	P3: 254±102	P3: 1492±678	P3: 1,966±708	P3: 1.33±0.34
							P3: 8.23±3.25
							CV%:	CV%:	CV%:	CV%:	CV%:
							Inter: 31.2; Intra: 10.6	Inter: 32.2 Intra: 13.3	Inter: 34.7 Intra: 14.7	Inter: 29.0 Intra: 11.1	Inter: 13.2 Intra: 15.9
							AUC0-12/D
							P1: 34.6±1.30
							P2: 35.6±13.0
							P3: 34.6±16.9
							CV%:
							Inter: 41.3; Intra: 10.6
							AUC0-4
							P1: 4.54±1.50
							P2: 4.63±1.31
							P3: 4.43±1.51
							CV%:
							Inter: 29.5; Intra: 9.5
							AUC0-4/D
							P1: 20.1±8.7
							P2: 20.1±7.0
							P3: 19.8±9.7
							CV%:
							Inter: 39.3; Intra: 9.7
	10 non-CF 5 L, 5 HL		175±52 mg BID (125–275) p<0.025				AUC0-12 (ns)	P1: 346±112	P1: 1545±530	P1: 1,752±427	P1: 1.62±0.31
							P1: 7.44±1.79	P2: 309±90	P2: 1500±426	P2: 1,959±576	P2: 1.34±0.19
							P2: 7.29±1.98	P3: 300±92	P3: 1372±337	P3: 1,802±587	P3: 1.55±0.46
							P3: 7.11±1.66	CF < non-CF	(ns)
							CV%:	CV%	CV%	CV%:	CV%:
							Inter: 23.8; Intra: 7.4	Inter: 28.9; Intra: 11.9	Inter: 26.5 Intra: 12.9	Inter: 26.9 Intra: 12.6	Inter: 17.6; Intra: 15.1
							AUC0-12/D (p < 0.005)
							P1: 41.6±7.7	C0/D:	C2/D:
							P2: 41.2±9.5	CF < non-CF	CF < non-CF
							P3: 40.6±8.5	(p<0.01)	(p<0.01)
							CV%:
							Inter: 19.7; Intra: 7.5
							AUC0-4 (ns)
							P1: 4.19±1.19
							P2: 4.30±1.27
							P3: 4.05±1.15
							CF vs non-CF :
							CV%:
							Inter: 27.3; Intra: 10.3
							AUC0-4/D (p < 0.005)
							P1: 23.4±3.9
							P2: 24.2±5.4
							P3: 22.7±4.9
							CV%:
							Inter: 18.2; Intra: 10.4
							Bayesian estimator (AUC0-12), best sampling times: C0, C1, C3 BE vs Trap on AUC0-12, AUC0-4, Cmax, tmax, C0: bias% < 5.3; RMSE%< 15

⁽¹⁾ Co-administration of enzyme inhibitors for minimum 2 years

⁽²⁾ In absence of ARE, dose decreased if SCr increased > 20% from baseline

⁽³⁾ Only adults except for 4 patients ≤ 16 years

⁽⁴⁾ 10 patients planned in the study: 9 L and 1 HL

ARE: Acute rejection episode – ATG: Anti-thymocyte globulin – AZA: Azathioprine – BE: Bayesian estimator – C: concentration – CF: Cystic fibrosis – D: Dose – DILT: Diltiazem – DL: Double lung transplantation – f/up: follow-up – HL: Heart-lung transplantation – Intra-ind.: Intra-individual – Inter-ind.: Inter-individual – KETO: Ketoconazole – MMF: Mycophenolate mofetil – NEO: Neoral – N/R: not reported – SAND: Sandimmune – SCr: Serum creatinine – SL: Single lung transplantation – Trap: Trapezes.

Table II. Cyclosporine pharmacokinetic parameters in cohort studies.

Results are expressed as mean±SD, unless otherwise specified

Ref	Number of patients	Post-transplant period	Cyclosporine dose		Co-administered immunosup-pressant	Assay	F, transfer k	Vd/F (unless otherwise specified)	CL/F (unless otherwise specified)	t1/2 (h) (unless otherwise specified)
HEART TRANSPLANTATION
[23]	16 adults (out of 35 in the global trial)		NEO		N/R	FPIA TDx	Non-compartmental PK analysis
		Week 1	200±58 mg				--	--	--	--
		Week 2	159±40 mg				F = 57±9%	--	--	--
		Week 52	162±37 mg				--	--	--	--
[29]	20 male adults 58.8±10.2 years	Stable 36.8±27.1 months	NEO and SAND 1.14±0.4 mg/kg TID		N/R No corticosteroids	FPIA TDx	2-compartment open model: 1-order absorption, 1-order elimination
							k: h−1	V1, V2: L/kg	CL: L.h−1/kg
			SAND				F = 66±16%	V1 = 1.16±0.77	CL = 0.22±0.04	17.27±3.23
							ka = 1.99±1.42	V2 = 5.38±1.37
							k10 = 0.23±0.07
							k12 = 0.37±0.14
							k21 = 0.12±0.04
			NEO				F = 75±19% (p<0.001)	V1 = 1.00±0.43 (ns)	CL = 0.21±0.04 (ns)	18.83±3.58 (ns)
							ka = 2.71±1.47 (p<0.05)	V2 = 5.84±1.37 (ns)
							k10 = 0.24±0.07 (ns)
							k12 = 0.41±0.16 (ns)
							k21 = 0.12±0.05 (ns)
			After IV infusion					V1 = 1.86±0.72 (ns)	CL = 0.20±0.04 (ns)	17.68±4.29 (ns)
								V2 = 5.58±1.57 (ns)
[19]	47 adults	Stable	NEO + SAND(1)		N/R	EMIT	Non-compartmental PK analysis
				D (mg/kg/day)			t1/2 abs: h		CL/F: L/h
	54±12 years	49±19 months	None	3.90±1.03		SAND	t1/2 abs = 2.72±1.45	--	46.7±21.1	--
			n = 11			NEO	t1/2 abs = 0.51±0.28 (p < 0.001)	--	31.8±9.5 (p<0.05)	--
	52±8 years	61±15 months	DILT	2.98±0.64		SAND	t1/2 abs = 1.44±1.16	--	36.7±14.1	--
			n = 11			NEO	t1/2 abs = 0.58±0.62 (p < 0.05)	--	26.8±9.0 (p<0.05)
	52±9 years	32±15 months	KETO	0.79±0.27		SAND	t1/2 abs = 1.65±1.22	--	9.7±4.3	--
			n = 13			NEO	t1/2 abs = 0.67±0.34 (p < 0.01)	--	9.1±3.8 (ns)	--
	47±12 years	28±5 months	DILT + KETO	0.90±0.54		SAND	t1/2 abs = 1.81±1.09	--	10.1±6.4	--
			n = 12			NEO	t1/2 abs = 1.07±0.57 (p < 0.05)	--	9.1±4.9 (ns)	--
[30]	47 adults	Stable (months)	NEO + SAND(1)		N/R	EMIT	2-compartment PK analysis using ABBOTTBASE pharmacokinetic system (PKS); Structural model and PK parameters selected from a study in stable heart transplant recipients by Baraldo et al[29]
				D (mg/kg/day)			k: h−1	V1: L/kg	CL: L.h−1/kg
		49±19	None	3.90±1.03			F = 75% (fixed)	V1 = 1	CL = 0.21	--
			n = 11				ka = 2.7 h−1 (fixed)	(CV = 43%)	(CV = 19%)
							k12 = 0.41 h−1 (CV = 39%)
		61±15	DILT	2.98±0.64			k21 = 0.12 h−1 (CV = 42%)
			n = 11
		32±15	KETO	0.79±0.27			With no covariate: r2 = 0.871 (p<0.001); bias% = 11.7; RMSE% = 13.4 Underestimation of AUC
			n = 14
		28±5	DILT + KETO	0.90±0.54			With “disposition factor”: (0.5 for CL if K or K+D; 0.76 for Vd if K+D) r2 = 0.698 (p<0.001); bias% = 2.2; RMSE% = 16.6 Overestimation of AUC in patients on diltiazem Bayesian model: best sampling strategy: C0, C1, C2
			n = 11
[21]	15 adults 48±10 years	Clinically stable 5.2±4.4 years	NEO 2.7±0.7 mg/kg/day D adjusted on Cav = 200–300 μg/L		ATG AZA Prednisolone	FPIA TDx	Non-compartmental PK analysis using non linear regression program WinNonLin
								Vd/F: mL/kg	CL/F: mL.min−1/kg
							--	2,421±945	5.7±1.7	5.0±1.3
[31]	14 patients de novo	3 PK profiles/patient	NEO D adjusted on C0		N/R	FPIA TDx	Population exposure indices and PK parameters estimated using ITS method, CICLO 1.3 Absorption model based on a gamma distribution
							MAT, SDAT: h	A, B: L−1		λ1, λ2: h−1
										t1/2(λ1), t1/2(λ2): h
		Week 1	200±61.2 mg BID				MAT = 1.25±0.40	A = 0.307±0.167	--	λ1 = 5.78±3.32
							SDAT = 0.48±0.20	B = 0.637±0.267		λ2 = 0.47±0.17
										t1/2(λ1) = 0.26±0.40
										t1/2(λ2) = 1.90±1.09
		Mont 3	152±36.0 mg BID (vs W1, p<0.05)				MAT = 1.27±0.40	A = 0.771±0.630	--	λ1 = 3.18±2.19
							SDAT = 0.45±0.19	B = 0.799±0.299		λ2 = 0.52±0.13
										t1/2(λ1) = 0.28±0.13
										t1/2(λ2) = 1.47±0.63
		Year 1	164±40.1 mg BID				MAT = 1.59±0.47	A = 0.690±0.445	--	λ1 = 2.595±1.368 (vs W1, p<0.05)
							SDAT = 0.66±0.32	B = 0.783±0.229		λ2 = 0.529±0.097
										t1/2(λ1) = 0.35±0.22
										t1/2(λ2) = 1.36±0.27
LUNG TRANSPLANTATION
[66]	14 adults 8 SL, 6 DL	Stable	NEO Steady-state		N/R	FPIA TDx	--	--	--	9.52±5.71
HEART AND LUNG TRANSPLANTATION
[27]	22 adults(2)	Month 1	D adjusted on C and clinical data		AZA + prednisolone	Pc NSRIA	PK parameters estimated using a Bayesian program for parameter estimation [48]
								Vd/F: L/kg	CL/F: mL.min−1/kg
	11 CF		16.7±7.2 mg/kg/day				--	7.2±5.3	3.5±1.3	--
	11 non-CF		8.2±1.9 mg/kg/day (p<0.01)				--	4.3±2.2 (ns)	1.7±0.5 (p=0.003)	--
[32]	19 adults	Stable (no evidence of rejection within previous 3 months)	NEO D adjusted on C0 = 250–350 μg/L		N/R	EMIT	PK parameters estimated using ITS method, CICLO 1.3; absorption described by a gamma distribution 3 consecutive PK profiles (P)/patient within 5 days Parameters standardized to a 100-mg dose
							MAT, SDAT: h	V1/F: L	CL/F: L/h	λ1, λ2: h−1
								F.A, F.B: L−1
	9 CF 6 L, 3 HL		250±76 mg BID (175–425)				F set to 1	V1/F = 88±41	50±20	λ1 = 4.14±3.01
							MAT = 1.00±0.21	F.A = 0.72±0.70		λ2 = 0.36±0.11
							SDAT = 0.30±0.09	F.B = 0.68±0.34
	10 non-CF 5 L, 5 HL		175±52 mg BID (125–275) p<0.025				F set to 1	V1/F = 74±44 (ns)	50±14 (ns)	λ1 = 2.16±1.75 (p < 0.01)
							MAT = 1.13±0.28 (ns)	F.A = 1.13±1.36 (ns)		λ2 = 0.49±0.12 (p < 0.01)
							SDAT = 0.37±0.15 (ns)	F.B = 0.77±0.30 (ns)

⁽¹⁾Co-administration of enzyme inhibitors for minimum 2 years

⁽²⁾Only adults except for 4 patients ≤ 16 years

ATG: Anti-thymocyte globulin – AZA: Azathioprine – C: concentration – CF: Cystic fibrosis – D: Dose – DILT: Diltiazem – DL: Double lung transplantation – F.A and F.B: estimated intravenous coefficients – f/up: follow-up – HL: Heart-lung transplantation – KETO: Ketoconazole – λ1, λ2: disposition rate constants – MAT: Mean absorption time – MMF: Mycophenolate mofetil – NEO: Neoral – N/R: not reported – SAND: Sandimmune – SDAT: Standard deviation of absorption time – SL: Single lung transplantation – V1: volume of central compartment – V2: volume of peripheral compartment.

Table III.

Cyclosporine population pharmacokinetic studies

Ref	Study	Subjects	PK parameters, exposure indices	ISV (%)	RRV	Covariates	Comments
HEART TRANSPLANTATION
[33]	Retrospective data from patients who underwent routine cyclosporine PK monitoring (HPLC-UV) NONMEM ADVAN2 TRANS2 One-compartment PK model First order absorption and elimination	2 groups of patients: Index n = 36; 51 years (18–63) Predictive performance n = 33; 50 years (25–65) Oral SAND for at least 3 months post-transplantation + AZA + prednisone	k = CL/V Fixed ka = 0.3 h−1 Fixed V = 4 x WT Initial analyses, F = 0.3 – then adjusted: F = 0.2 + 10 x POD-73/[(POD + 10) x 60] Cl = 0.281 x WT x (0.629 x DIL + 1 – DIL)	CL: 20.2	Add. 77.4 μg/L	POD Diltiazem	Over-prediction of concentrations (from150 to 400 μg/L), especially within the first two weeks post-transplantation Predictive performance on concentrations: When F = 0.3: ME = 42.1 μg/L [35.6–48.6] RMSE = 102.2 μg/L [61.0–143.4] When F adjusted: ME = 21.3 μg/L [15.0–27.6] RMSE = 92.5 μg/L [54.0–131.0]
HEART AND LUNG TRANSPLANTATION
[34]	Retrospective PK analysis of study performed by Trull et al. TDM 1999 (NEO vs SAND) – EMIT NONMEM ADVAN2 TRANS2 One-compartment PK model First order absorption and elimination	48 adults (8 CF) 18 SL, 9 DL, 21 HL NEO, n = 27 – SAND, n = 21 (D = 417±218 mg/day) + ATG, AZA, prednisolone Follow-up: Week 1 to week 52	1,004 samples collected CL/F (L/h) = 22.1 (19.5–24.7) Vd/F (L) = 147 (130–164)	CL/F: 17.1	Prop. 44.0% Add. 76.4 μg/L	On CL/F Itraconazole,: CF, WT On F: SAND, time post-transplantation	Comments of Saint-Marcoux[37] Underestimation of C 2 values(95%CI = +5 to +115 μg/L) Overestimation of C6 values(95%CI = −127 to −69 μg/L) Final model not validated, robustness not tested ISV of Vd/F not estimated Influence of individual factors tested only against CL/F

Add.: Additive – AZA: Azathioprine – CF: Cystic fibrosis – ISV: Inter-subject variability – ME: Mean error – NEO: Neoral – POD: post-operative day – Prop.: Proportional – RMSE: Root mean square error – RRV: Residual random variability – SAND: Sandimmune – WT: Weight.

In most cases, PK studies were performed in contexts where cyclosporine monitoring was based on C₀ or on C₂.

1.1.1.1. Absorption

Cyclosporine is a highly lipophilic molecule. Early PK studies with the first oil-based formulation Sandimmune^® showed a poor and highly variable relative oral bioavailability. The microemulsion formulation (Neoral^®) designed in a second time to increase the solubility of cyclosporine in the small bowel improved bioavailablity and decreased absorption variability, with a higher maximum concentration (C_max), shorter time to maximum concentration (t_max), higher AUC, and lower intra-patient variability on absorption than with Sandimmune^®.[35]

An increased and more consistent exposure to cyclosporine after Neoral^® than Sandimmune^® administration was also reported in heart^[29] and heart-lung^[28] transplant recipients. As a consequence, Neoral^® became the core maintenance immunosuppressant chosen by most thoracic transplantation centers worldwide.^[1] However, the high intra- and inter-patient variability of Neoral^® absorption was confirmed by Johnston et al in a study performed in 15 de novo heart transplant recipients, in which three PK profiles were collected during the first year post-transplantation. Intra-patient and inter-patient variability of cyclosporine concentrations was the highest within the first two hours post-dose.^[36]

Cyclosporine absorption typically lasts 4 hours, as illustrated by many studies in thoracic transplantation where C_max usually occurred between 1 and 3 hours post-dose.^{[4,19,23–26,29,32]} In a randomized study comparing Neoral^® to Sandimmune^® in 35 heart transplant recipients, absolute oral bioavailability based on AUC_0-12 with respect to IV administration was 57±9% and 47±12%, respectively (p = 0.02), at 12 weeks post-transplantation.^[23] A significant difference of absolute bioavailability between Neoral^® and Sandimmune^® was also reported by Baraldo et al in 20 stable heart transplant recipients on cyclosporine three times daily, 8 of whom were on diltiazem (75±19% for Neoral^® vs 66±16% for Sandimmune; p < 0.001).^[29]. However, these two studies employed a non-specific cyclosporine assay (FPIA for TDx), which cross-reacts with cyclosporine metabolites. Oral bioavailability might have been overestimated as blood concentrations of cyclosporine metabolites are higher after oral than after IV administration due to the first-pass hepatic and intestinal effects, which are particularly important for substrates of CYP3A isoforms such as cyclosporine.

The PK models developed for cyclosporine by three groups^[29,33,34] described absorption as following a first-order process, however with inconsistent mean values for the absorption rate constant k_a. The first study^[33] was a popPK study performed in 69 heart transplant patients on Sandimmune^® during the first three months post-transplantation. Unfortunately, the fixed k_a value (0.3 h⁻¹) implemented in the model did not take into account the important variability of cyclosporine absorption, leading to an overestimation of cyclosporine concentrations by 150 to 400 μg/L. The second study^[34] was a retrospective popPK analysis of data collected from 48 heart-lung transplant recipients, followed throughout their first year post-transplantation. Unfortunately, this study focused on apparent clearance (CL/F) and volume of distribution (V_d/F) and fixed k_a to values estimated by the same group in one of their previous works (0.25 and 1.35 h⁻¹ for Sandimmune^® and Neoral^®, respectively).^[19] The model chosen led to a large residual random variability of 44.0 % (proportional) + 76.4 μg/L (additive). As pointed out by Saint-Marcoux et al, this model systematically underestimated C₂ values, showing the limits of a linear model for cyclosporine absorption.^[37] In the third study,^[29] the PK of cyclosporine was investigated using the ITS-method in 20 stable heart transplant males grafted for more than 3 years, switched from Sandimmune^® to Neoral^®. The absorption rate constant was k_a = 1.99±1.42 h⁻¹ and 2.71±1.47 h⁻¹, respectively (p < 0.05). It is therefore noticeable that k_a differed almost 8-fold between studies for Sandimmune^®, and 2-fold for Neoral^®.

Considering that model-independent PK methods or first-order absorption could not accurately describe the erratic and variable absorption of cyclosporine, our group used an original absorption model based on a Gamma distribution, initially developed for renal transplant recipients.^[38] In this model, the absorption rate is described as an asymmetrical peak and the absorbed fraction of the dose as a function of time as an S-shaped curve. This model was evaluated in a subpopulation of 14 de novo heart transplant recipients on Neoral^® enrolled in study OLN351, in whom a PK profile of cyclosporine was drawn at three post-transplantation periods: the first week (W1), the third month (M3) and the end of the first year (Y1).^[32] Population exposure indices and PK parameters were estimated using the standard ITS-method. The PK model fitted well with the concentration-time data at all periods, with an excellent correlation between observed and modelled concentration values. For each period, the mean relative bias on concentrations was less than 1% and precision was good (> 88%). This model was also tested successfully on a population of 19 stable lung and heart-lung transplant recipients, with or without cystic fibrosis (CF). Three consecutive profiles were collected in each patient within 5 days. Good correlation and non significant differences were observed between measured and modeled concentration values, in non-CF (r² = 0.986) as well as in CF patients (r² = 0.976), confirming the robustness of the Gamma model (Table I). The mean absorption time (MAT) was not different between the two groups, with corresponding CV% of 21% and 25% in CF and non-CF patients, respectively.

1.1.1.2. Distribution and elimination

Cyclosporine is mainly metabolized by the cytochrome P450 (CYP) 3A enzymes in the liver and the gut. It is also a substrate of P-glycoprotein (P-gp), particularly in the intestinal mucosa. Hence, it may interact with many of the substrates, inducers and inhibitors of both of these systems. V_d/F, CL/F and elimination half-life (T_1/2) of cyclosporine in thoracic transplantation were estimated in a few studies using non-compartmental analysis,^[19,21] the ITS-method^[29,32] or population PK analysis.^[34]

Formal comparison of V_d/F between studies is not presented here because only four of them addressed this parameter, unfortunately with different models and units. Two studies used a 2-compartment model, one of which expressed the apparent volume of the central compartment (V₁/F) in liters,^[32] the other in L/kg.^[29] In the two remaining studies, V_d/F was expressed in L/kg.^[21,27] The same issue was faced for clearance, which was reported in studies using two-, single- or non-compartment models and expressing clearance or CL/F in L.h⁻¹ or in L.h⁻¹/kg. In 15 stable heart transplant recipients on Neoral^®, V_d/F was 2.42±0.95 L/kg, CL/F was 5.7±1.7 mL.min⁻¹/kg (0.342±0.102 L.h⁻¹/kg) and T_1/2 was 5.0±1.3 h.^[21] In 47 stable heart transplant adults switched from Sandimmune^® to Neoral^®, CL/F (as dose/AUC_0-12) was significantly lower with Neoral^® in the two patient subgroups who were not on ketoconazole, but there was no significant difference in the other two subgroups who received ketoconazole.^[19] This is obviously due to improved bioavailability, as illustrated earlier. In 20 stable heart transplant patients who consecutively received Sandimmune^®, IV cyclosporine then Neoral^®, CL/F and the volume of distribution of the central compartment V₁/F were 0.20±0.04 L.h⁻¹/kg and 1.86±0.72 L/kg after IV infusion, 0.21±0.04 L.h⁻¹/kg and 1.00±0.43 L/kg after Neoral^®. T_1/2 was longer than usually reported, ranging between 17.27±3.23 (Sandimmune^®) and 18.83±3.58 h (Neoral^®).^[29] Finally, using the ITS method in 19 lung and heart-lung transplant patients (of whom 9 with CF) on Neoral^®, CL/F was 50±20 and 50±14 L/h and V₁/F = 88±41 and 74±44 L in CF and non-CF patients, respectively (ns).^[32]

In the popPK study in heart-lung (21) and lung transplant recipients (27) performed by Rosenbaum et al, using a one-compartment PK model, the results were expressed as pooled Sandimmune^® and Neoral^® data. V_d/F was 147 L (range, 130–164 L) and CL/F was 22.1 L/h (range, 19.5–24.7 L/h) on average.^[34] The inter-patient variability on modelled clearance was 17.1%, but no intra-patient variability was reported for this PK parameter. The following factors were reported to influence CL/F: co-administration of itraconazole, weight, and CF. No inter- or intra-patient variability of, and no covariates on V_d/F were reported. In a popPK study performed in heart transplant recipients, Parke et al proposed an estimation of clearance based on body weight and the co-administration of diltiazem.^[33] The inter-patient variability on clearance was 20.2%.

1.1.1.3. Drug-drug interactions

Lipid-lowering therapy is commonly administered to transplant patients to overcome hyperlipidemia and prevent coronary-artery disease.^[39] Considering the affinity of cyclosporine for plasma lipoproteins, it was hypothesized that the modification of lipoprotein concentration may influence cyclosporine disposition and PK. Akhlaghi et al evaluated the effect of simvastatin on cyclosporine PK in heart transplant patients with a population approach using the P-PHARM program, though only to estimate cyclosporine elimination based on C₀, which might have limited the power of this study.^[40] The authors found a moderate increase in cyclosporine CL/F in patients on lipid lowering agents. However, the difference in clearance before and after simvastatin administration did not reach statistical significance, whether patients received ketoconazole or not.

Diltiazem is another enzyme and transporter inhibitor often given early after heart transplantation to prevent coronary-artery narrowing and coronary artery disease on the transplant. Lung transplant patients often need azole antifungals such as ketoconazole, itraconazole or voriconazole as prophylactic or curative treatments, because they risk developing invasive fungal infections of the graft. To our knowledge, the impact of enzyme or transporter inhibitors on the PK of cyclosporine after thoracic transplantation was evaluated in only two studies.^[19,41] One hundred and eighty two patients receiving Sandimmune^® as part of their immunosuppressive regimen after cardiothoracic transplantation were recruited in the first study.^[41] The authors defined a linear relationship between cyclosporine dose rate (DR) and trough concentrations at steady-state (“Css_trough”). The relationship was described as follows: DR = θ x Css_trough, where θ was the dose rate-steady-state trough concentration ratio, assuming that the PK of cyclosporine is linear. Posterior Bayesian estimation using P-PHARM indicated that the concomitant administration of metabolic inhibitors influenced significantly θ, probably because of a change in CL/F (although a modification of cyclosporine distribution could not be excluded). A dramatic reduction (−82%) of θ was caused by the co-administration of ketoconazole and diltiazem, while co-administration of either ketoconazole, itraconazole or diltiazem resulted in 75%, 40% and 23% reduction, respectively. Consistently, patients who had been administered ketoconazole had a 75% reduction in cyclosporine dose and those on diltiazem a 23% decrease.

The 47 stable heart transplant recipients included in the second study were switched from Sandimmune^® to Neoral^® and were divided into four groups depending on the long-term (beyond two years) co-administration of enzyme inhibitors (diltiazem and/or ketoconazole or none).^[19] Cyclosporine AUC was significantly higher after administration of Neoral^® compared to Sandimmune^® only in patients who were either receiving no metabolic inhibitors (4,911±935 vs 3,655±1,120 h.μg/L, p < 0.05) or diltiazem alone (4,747±923 vs 3,605±1,079 h.μg/L, p < 0.05), but not in those taking concomitant ketoconazole or a combination of ketoconazole and diltiazem. Similarly, cyclosporine CL/F was significantly lower with Neoral^® formulation, but only for patients on diltiazem (26.8±9.0 vs 36.7±14.1 L/h, p < 0.05) or on no metabolic inhibitor (31.8±9.5 vs 46.7±21.1 L/h, p < 0.05). The authors hypothesized that in patients on ketoconazole, maximal cyclosporine absorption was already achieved, so that no further increase was possible by switching to Neoral^®. The mechanism of metabolic inhibition by ketoconazole is not completely known, but it seems that it inhibits the metabolism and enhances the absorption of cyclosporine in the gastrointestinal mucosa by a complete blockage of prehepatic metabolism and inhibition of intestinal P-gp. On the other hand, diltiazem appeared here to have very little, if any, effect on cyclosporine PK.^[19]

1.1.1.4. Special populations

1.1.1.4.1. Pediatric transplantation

According to the ISHLT registry, between January 1998 and June 2006, 32.8% of heart and heart-lung transplant recipients were aged less than 19 years at the time of transplantation.^[3] Since 2001, there has been a slow rise in the total number of transplant procedures performed in children: 77 were reported in 2005 versus 59 in 2001. CF is the main indication for thoracic transplantation in children (69% of patients aged 12 to 17 years who underwent transplantation between January 1991 and June 2006).^[42]

The PK of cyclosporine in paediatric patients is known to be highly variable,^[43] and different from that of adults in terms of gastro-intestinal absorption, volume of distribution and systemic clearance (hepatic metabolism).^[44,45]

To our knowledge, no study has been specifically designed to evaluate the PK of cyclosporine in children after thoracic transplantation. One study performed in both adults and children, in different types of transplantation (renal, cardiac, hepatic and bone marrow transplantation) reported PK parameters of cyclosporine in 12 heart-transplant patients (apparently mainly children).^[46] Their mean V_d/F and mean residence time were 7.4±1.8 L/kg and 10.7±4.4 h respectively, with no significant difference compared to other types of transplantation.

Given the high proportion of children among patients benefiting from thoracic transplantation, more studies are definitely needed to better characterize cyclosporine PK in this population.

1.1.1.4.2. Cystic fibrosis

CF is a major indication for lung and heart-lung transplantation. Despite the administration of exogeneous pancreatic enzymes, patients with CF are poor absorbers (gastro-intestinal disorders and malabsorption of lipids owing to pancreatic insufficiency). They exhibit higher variability on absorption compared with non-CF patients, with erratic PK profiles and a delayed t_max, and consequenly require different dosage regimens.^{[8,25–27,47]}

A study in 31 heart and lung transplant candidates (mainly adults) reported lower cyclosporine bioavailability, higher apparent oral clearance and shorter mean residence time in subjects with CF as compared to those with Eisenmenger’s syndrome. The authors suggested that the difference on cyclosporine CL/F may be due to the lower bioavailability in CF patients.^[47] In a retrospective case-control study, the same authors compared the PK of cyclosporine in CF patients (n = 11), vs non-CF patients (n = 11), following heart-lung transplantation. Patients’ daily dose was 16.7±7.2 mg/kg vs 8.2±19 mg/kg, respectively (p < 0.01). Cyclosporine PK parameters were estimated using a Bayesian approach.^[48] A statistically significant difference on cyclosporine CL/F was found between CF and non-CF patients (3.5±1.3 vs 1.7±0.5 mL.min⁻¹/kg, respectively; p = 0.003), but not on V_d/F (7.2±5.3 vs 4.3±2.2 L/kg, respectively; ns).^[27] The authors proposed that patients with CF be given 1.5 to 2 times higher oral doses of cyclosporine than patients without CF.^[27,47]. Higher drug doses were required in 9 CF vs 41 non-CF patients (p < 0.001) on either Sandimmune^® or Neoral^®, resulting in lower dose-normalized AUC_0-6 (p < 0.001).^[28]

Several studies in limited numbers of mixed, adult and paediatric CF patients addressed the potential benefit of improved cyclosporine oral absorption after Neoral^® rather than Sandimmune^® administration. Three studies reported that Neoral^® allowed a substantial increase in AUC_0-12 (by 2.63; 1.44 and 1.48-fold, respectively) and C_max (by 2.28; 2.28 and 1.91, respectively) when compared to Sandimmune^®, with inconsistent findings regarding C₀.^[24,26,49] The intra-individual and inter-individual coefficient of variation on C₀ and on AUC_0-12 were less with Neoral^® than Sandimmune^®, confirming improved PK consistency.^[26] Increased exposure was attributed to higher bioavailability, with CL/F significantly and V_d/F numerically higher in CF than non-CF patients.^[50]

The PK of cyclosporine was also studied in 10 CF and 10 non-CF clinically stable lung and heart-lung transplant recipients on Neoral^® in an open-label trial, with a non-compartmental approach.^[25] Cyclosporine exposure was similar in patients with and without CF, but exposure-per-milligram-dose was approx. 25% lower in CF patients, due to lower bioavailability. Rousseau et al conducted a retrospective study of the data collected from 19/20 patients included in this trial,^[32] confirming that although CF patients received higher doses (250±76 mg BID vs 175±52 mg BID, respectively, p < 0.025), no statistically significant difference was found for measured C₂ values between the 2 groups, resulting in lower C₂/dose ratios in CF patients (p < 0.001). Similarly, lower AUC_0-12/dose and AUC_0-4/dose ratios were found in CF patients (p < 0.005). No significant difference was found between CF and non-CF patients, in V₁/F (88±41 and 74±44 L, respectively) and CL/F (50±20 and 50±14 L/h, respectively), maybe due to the small number of patients. The two elimination constants were significantly different between CF and non-CF patients, the first higher and the second lower, which is consistent with the similar CL/F values.^[32]

1.1.2. Therapeutic drug monitoring

Cyclosporine is characterized by a marked inter- and intra-patient variability in its PK parameters and a narrow therapeutic index: subtherapeutic exposure may be associated with rejection and supratherapeutic exposure with drug-induced toxicity (e.g., nephrotoxicity: end-stage renal failure requiring dialysis occurs in up to 6.5% of heart transplant recipients^[51]).

TDM of cyclosporine has been recognized as an essential tool in the management of allograft transplant recipients and has been performed routinely in transplantation for almost 20 years.^[52,53] However, there is still no consensus about the optimal method for monitoring cyclosporine, and very little data is available on the practices and outcomes of cyclosporine monitoring after thoracic transplantation. Much of the available information has been extrapolated from data obtained in other types of organ transplantation. Table IV presents a summary of TDM methods together with their advantages and limitations, derived from reviews and tables by Fernandez de Gatta et al^[43] and Dumont et al.^[53]

Table IV.

Pharmacokinetic methods for therapeutic drug monitoring of cyclosporine after thoracic transplantation (derived from Fernandez de Gatta et al. Clin Pharmacokinet 2002 and Dumont RJ et al. Clin Pharmacokinet 2000).^[43,53]

Method	Advantages	Limitations
Single concentration
C0	Standard method: simple and practical (inpatients and outpatients)	Does not reflect absorption, drug exposure, or elimination Does not give information on other PK parameters Prediction of clinical outcome: conflicting data
C2	Surrogate marker of absorption (Neoral®) Higher sensitivity as indicator of AUC and clinical effect than C0	Rigid collection time No distinction between poor absorbers and slow absorbers Prediction of clinical outcome: conflicting data
Other time points (C3, C4, C6)	C0 better correlated with AUC	Rigid collection time No distinction between poor absorbers and slow absorbers Hardly any data on their relationship with clinical outcomes
AUC
Full	The best indicator of drug exposure, absorption profile and clinical outcome Characterization of abnormal absorption patterns on concentration-time profile Allows the calculation of oral pharmacokinetic parameters Reduces analytical variability	Need for multiple blood samples: impractical for routine clinical use Expensive Inconvenient for patients (outpatients ++) Optimal targets to be defined for most immunosuppressants
Abbreviated (ex: 4 h post-dose)	Good predictor of absorption profile and full AUC	Further studies required Optimal targets to be defined
Sparse sampling strategies	Balance between precision and practicality Acceptable predictor of AUC	Multi-linear regression not based on a PK model. Equations result from correlations and cross-correlations between sampling times and AUC Rigid collection times Analytical method- and centre-specific, often not validated in independent populations
Bayesian forecasting
	Flexibility in sampling times Limited number of samples needed Can easily be integrated in clinical practice Simultaneous estimation of individual PK parameters and exposure indices Identification of absorption or elimination problems (e.g., gastroparesis)	Only a few studies in small populations Predictive performance not tested Requires sophisticated PK models and software

1.1.2.1. Single concentrations monitoring

The traditional method for cyclosporine dose adjustment is trough concentration monitoring, and it is widely used in most clinical settings. However, many studies in thoracic transplantation have evidenced a poor correlation between C₀ and exposure to cyclosporine, expressed either by the full AUC over the dosing interval (AUC_0-12) or by an abbreviated AUC (AUC_0-4, AUC_0-6).

Correlation coefficients between C₀ and AUC_0-12 or AUC_0-4 reported in adult heart transplantation, ranged from 0.06 to 0.6, and from 0.16 to 0.4, respectively.^{[21,23,54–56]} Similar results were reported after lung or heart-lung transplantation: in a study of 50 patients on cyclosporine after heart-lung transplantation, Trull et al found that C₀ was well correlated with AUC_0-6 estimated from three time-points (C₀, C₂, C₆) using the linear-log trapezoidal rule (r² = 0.574 and 0.671 for Sandimmune^® and Neoral^® respectively).^[28] Surprisingly, using the same database, apparently limited however to 48 patients, Akhlaghi et al reported a poor correlation between C₀ and AUC_0-6 (r² = 0.267).^[20] This may be due to the fact that Sandimmune^® and Neoral^® data were pooled and the linear rather than the linear-log trapezoidal rule employed. Jaksch et al^[4] reported a correlation coefficient of 0.64 between C₀ and AUC_0-4 in 20 lung transplant recipients, and Hangler et al^[6] a correlation coefficient of 0.31 between C₀ and AUC_0-6 in 12 lung transplant recipients.

As it was recognized that C₀ values were rather poorly correlated with the AUC,^[52] monitoring, another single concentration was proposed as an alternative to C₀ monitoring. The immunosuppressive effect of cyclosporine is maximal and most consistent around t_max, which is usually within 2 hours post-dose. C₂ has been demonstrated as an excellent predictor of AUC_0-4 in renal transplantation.^[14,57] In thoracic transplantation, a series of studies have shown that C₂ quite reliably predicts AUC_0-12 and abbreviated AUC: coefficients of determination between C₂ and AUC_0-12 ranged from 0.94 to 0.99 in heart transplant recipients;^[21,54,55] lower coefficients of determination between C₂ and AUC_0-12 were reported in one study in heart-lung transplantation (0.76 and 0.66 in CF and non-CF patients, respectively);^[25] coefficients of determination reported between C₂ and AUC_0-4 were comprised between 0.78 and 0.98 in all types of transplantation.^[4,25,54,56]

Some authors also reported a good correlation between C₂ and AUC_0-6 (r² > 0.80).^[6,20,28] However, in two of them,^[20,28] performed on the same database, AUC_0-6 was estimated using only three time-points (C₀, C₂, C₆) and the trapezoidal rule, which necessarily implies a good correlation with each of the three points, particularly with the middle one which is used twice for the trapezes calculation.

In another paper analysing the same database of 47 heart transplant patients as Akhlaghi et al^[19] (though with inconsistencies in patients’ numbers and classification, as well as AUC values despite the use of identical samples, analytical method and method of calculation), Ray et al considered AUC_0-5 and concluded that in general, it was poorly predicted by both C₂ (r² = 0.197 to 0.898) and C₀ (r² = 0.022 to 0.710), but that C₀ did better than C₂ in patients on cyclosporine alone (r² = 0.710, p = 0.001, vs 0.197), while C₂ did better than C₀ in patients co-administered ketoconazole (± ditiazem) (r² = 0.870, p < 10⁻⁴, vs 0.176 and 0.898, p < 10⁻⁴, vs 0.022) and both were weak predictors in patients on cyclosporine and ditiazem (but no ketoconazole) (r² = 0.65 for both).^[58] However, these conclusions were drawn from very small numbers of patients in each group.

In conclusion, C₂ seems to be a better single-time-point surrogate marker than C₀ for cyclosporine AUC_0-12, and is now strongly encouraged. It seems to be of particular interest in CF patients and in patients in whom cyclosporine-induced toxicity is suspected, despite acceptable C₀.^[9] Unfortunately, the optimal concentration range for cyclosporine C₂ has not been established for thoracic transplant recipients. Moreover, rigid collection timing is a drawback of C₂ monitoring, and the pertinence of C₂ may be compromised by all the factors of variability of cyclosporine absorption or exposure, including CF (see above).^[30,58]

Alternatives to C₀ or C₂ monitoring are other single concentrations measured at specific times post-dose. Several time-points have been proposed, based on the results of linear regression analyses intended for developing sparse sampling strategies.^[53] C₆ was proposed for heart transplant recipients, based on efficacy, toxicity and cost in 20 patients.^[59] A good correlation was found between C₄ and AUC_0-12 (r² = 0.95) in 15 heart transplant patients.^[21] In heart-lung transplant recipients, the best correlation was between C₃ and AUC_0-12, in CF as well as in non-CF patients (r² = 0.87 and 0.82, respectively).^[25]

Single concentration monitoring (whether based on C₀, C₂, C₃ or C₆) is widely used in clinical routine. However, as further discussed in paragraph 1.1.2.3.1 (“C₀ monitoring”), robust data on the correlation between such concentrations and overall exposure to cyclosporine and target values is still lacking, especially in populations such as children or CF patients. Further exploration is still needed, and a consensus on such tools, with the definition of target concentrations, is awaited.

1.1.2.2. AUC monitoring

1.1.2.2.1. Full AUC

The inter-dose AUC over the dosing interval (AUC_0-τ) is believed to be a better surrogate marker of cyclosporine efficacy and toxicity than single concentrations, as it gives a complete picture of its absorption and elimination and is proportional to the mean concentration over the dosing interval (AUC = C_ss × τ). Indeed, in the absence of any evidence that cyclosporine effects are linked to its maximal or minimal blood levels, the mean concentration is supposed to better reflect its activity on calcineurin.^[60,61] As a consequence, AUC monitoring has become the gold standard for the TDM of cyclosporine.^[62] However, proper calculation of AUC requires repeated blood sampling, followed by the calculation of AUC from cyclosporine blood concentrations using the trapezoidal method.^[53] As a consequence of the intensive sampling strategy, it is uncomfortable for the patients, inconvenient for the clinical team and costly to the health-care provider, which explains why it is only performed infrequently.

1.1.2.2.2. Abbreviated AUC

As full AUC_0-12 is impractical to obtain, abbreviated AUCs have been tested. The 4-hour AUC (AUC_0-4) has been validated as a reliable estimate of AUC_0-12.^[57,63] Indeed, the largest part of cyclosporine AUC_0-12 variability occurs in the first 4 hours following administration. The AUC_0-4 reflects the highly variable absorption profile of cyclosporine, suggesting it is a more accurate measure of exposure than C₀.^[36]

A retrospective study of 156 de novo renal graft recipients showed that cyclosporine AUC_0-4 (sampling at 0, 1, 2, 3, 4h) measured in the first 2–4 days post-transplantation (but not C₀) was lower in patients who experienced acute rejection at 3 months than those who did not.^[12] In addition, AUC_0-4 was reported to be a sensitive marker of nephrotoxicity (contrary to C₀). In a second time, the same authors performed a prospective, non comparative study in 59 de novo renal transplant patients, with cyclosporine dose adjustment based on AUC_0-4 (target range: 4,400 – 5,500 h.μg/L).^[64] They found that the incidence of acute rejection was 3% when AUC_0-4 at day 3 was > 4,400 h.μg/L (n = 33) vs 45% when it was < 4,400 h.μg/L (n = 22) (p = 2.10⁻⁴). Again, they did not note any nephrotoxicity episodes when AUC was < 5,500 h.μg/L.

Unfortunately, no equivalent or better designed studies were conducted in thoracic transplant patients.

1.1.2.2.3. Sparse sampling strategies

Rather than replacing AUC_0-12 monitoring by the determination of the abbreviated AUC_0-4, some authors sought for other clinically applicable methods to accurately estimate AUC_0-12. This led to the development of sparse sampling strategies, in which specific algorithms allow AUC estimation from a limited number of blood samples collected at precisely defined times. A sparse sampling strategy is developed using full AUC values in a sample population. Stepwise multiple regression (MLR) analysis is then performed on the concentration-time points sampled. The points that do not correlate well with AUC are removed until a regression equation consisting of 2 or 3 concentration-time points is left. The AUC estimate obtained using the retained equation must have a high coefficient of correlation with the full AUC, calculated using a reference method. A good predictive performance (bias and precision) of the sparse sampling strategy is also mandatory.

To our knowledge only a few studies have developed such sparse sampling strategies and MLR equations for thoracic transplant recipients (Table V). Sparse sampling strategies simplify AUC monitoring but there are some limitations that need to be pointed out. First, the sampling strategies that were originally developed in renal transplantation proposed impracticable collection times in an outpatient setting.^[65] To be clinically applicable, such strategies should be limited to the first 4 hours post dose. This prerequisite was fulfilled by most sparse sampling strategies developed in thoracic transplantation.^[6,21,66] Secondly, sparse sampling strategies developed using Sandimmune^® are no longer applicable because of the PK differences between Neoral^® and Sandimmune^®.[53] Fortunately, all the sparse sampling strategies developed in thoracic transplantation were developed using Neoral^®. Thirdly, no study has been conducted to date to demonstrate the superiority in terms of outcome of sparse sampling strategies over C₀ or C₂ monitoring. As a consequence, there are no evidence-based data justifying the use of a sparse sampling strategy which requires 2 or 3 blood samples, rather than a single concentration. In addition, although few blood samples are needed, multiple linear regression requires a strict respect of sampling times, which is not easily feasible in a clinical setting. Moreover, in most cases, sparse sampling strategies were chosen solely on the basis of high coefficients of correlation^[6,21] rather than looking for other criteria such as bias or prediction error.^[53,66,67] In addition, to be applicable routinely, sparse sampling strategies should be validated in external populations, which is seldom the case.^[6,21] Finally, the algorithm proposed should allow AUC calculation over an appropriate dosing interval (AUC_0-12 or AUC_0-8 for patients receiving cyclosporine two or three times daily, respectively); unfortunately, the actual dosing interval was only reported in 2/4 studies.^[21,68] The algorithm proposed by Hangler et al^[6] allows the determination of AUC_0-6 which is of a limited interest, as the clinical pertinence of AUC_0-6 in patients dosed twice daily has never been demonstrated, nor have any targets been proposed. The algorithm proposed by Dumont et al^[66] also presents major drawbacks rendering it useless: first of all, there is no indication on the type of lung transplant patients (CF or not) in whom the strategy was developed; secondly, the algorithm allows the calculation of “AUC_0-τ” without any precision on the value of τ.

Table V.

Summary of selected sparse sampling strategies for cyclosporine monitoring after Neoral^® administration

Ref	Patient group	Number of patients	AUC interval (h)	Equation	r2	%PE
[21]	HTx Stable	15	12	${\text{AUC}}_{0-12}=531+0.64\ast {\text{C}}_1+8.12\ast \text{C}4$	0.973	N/R
[215]	HTx	20 Validation: 17[215] then 14[31]	12	${\text{AUC}}_{0-12}=330.841+2.235\ast {\text{C}}_2+7.970\ast {\text{C}}_6$
	Week 1				0.951	−8.68 to +6.24
	Month 3				0.918	−7.3 to +8.58
	Year 1				0.949	−5.33 to +6.49
[6]	LTx < Week 2	12	N/R	${\text{AUC}}_{0-6}=0.496\ast {\text{C}}_2+0.517\ast {\text{C}}_3+776.9$	0.94	NR
[66]	LTx Stable	8 Validation: 6	τ, N/R	${\text{AUC}}_{0-\tau }=185.62+1.75\ast {\text{C}}_1+4.91\ast {\text{C}}_3$	0.904	−6.96 to +13.05
				${\text{AUC}}_{0-\tau }=53.54-5.47\ast {\text{C}}_0+2.11\ast {\text{C}}_1+6.44\ast {\text{C}}_3$	0.908	−10.31 to 22.43

HTx: Heart transplantation – LTx: Lung transplantation – N/R: not reported – PE: Prediction error.

In brief, only one published sparse sampling strategy seems to be clinically applicable in heart transplantation (i.e., with a maximum of 3 samples collected within the first 4 hours post-dose, actually C₁ and C₄), although it still needs external validation.^[21] There is currently no applicable sparse sampling algorithm for lung transplant patients.

1.1.2.2.4. Bayesian forecasting

Maximum a posteriori (MAP) Bayesian forecasting based on a population model allows simultaneous estimation of individual PK parameters and exposure indices, which helps define the dose and administration schedule needed to achieve the desired exposure in a given patient, whatever the targeted exposure index (single concentration value, full AUC or abbreviated AUC).^[32] Bayesian forecasting of cyclosoporine AUC_0-12 has been proposed as a TDM tool in renal transplantation,^[69–72] but seldom in thoracic transplantation.

In heart transplantation, the first such study intended to develop a Bayesian estimator based on a routinely applicable sparse sampling strategy to accurately estimate cyclosporine AUC_0-12.^[31] Concentration-time data was obtained from 14 heart transplant recipients at three periods during the first year post-transplantation. The popPK parameters were evaluated using the standard two-stage method. The accuracy of the model was evaluated using the Jackknife method. A MAP-Bayesian estimator was built using the distribution of the population parameters, and all sparse sampling strategies combining a maximum of three sampling times within 4 hours post-dose were tested. The best sampling time combination, with respect to its predictive performance for AUC_0-12 estimation and to the D-optimality criterion (a statistical test) applied to Bayesian estimation, was found to be T₀, T_1h and T_3h, whatever the period post-transplantation, yielding excellent correlation with the “reference” AUC_0-12 calculated using the linear trapezoidal rule: r² = 0.887 to 0.998, bias = −0.20 to +3.06%.

In a second study performed in 31 heart transplant recipients, cyclosporine AUC_0-12 was calculated with Bayesian estimation and multiple linear regression in parallel, using previously reported estimator and equations.^[54] The Bayesian estimator used for this purpose was originally designed for renal transplant recipients.^[69] The authors investigated the correlation between AUC_0-12 Bayesian estimates and measured C₀ (r² = 0.43) or C₂ (r² = 0.85) values, confirming that C₂ was a better surrogate marker of AUC_0-12. Unfortunately, as they did not compare the values of AUC_0-12 obtained using the different methods, Bayesian estimation could not be validated, neither vs multiple linear regression nor vs the linear trapezoidal rule.^[54]

The third study^[30] was performed on a database of cyclosporine concentrations previously collected from 47 adult heart transplant recipients (although presenting discrepancies with the original description of the population, notably on patient numbers and classification as already mentioned).^[19] Patients were divided into four groups based on the long-term concomitant administration of enzyme inhibitors (no enzyme inhibitor, diltiazem and/or ketoconazole). The structural PK model was selected from the literature,^[73] and parameters for model building were selected from a previously published PK study in stable heart transplant recipients on Neoral^®.[29] Of note, the authors added a “disposition factor” of 0.76 on V_d for patients on ketoconazole + diltiazem and of 0.5 on CL for patients on ketoconazole with or without diltiazem, thus improving the model and AUC estimation.^[30] The Bayesian models tested included one, two or three concentrations out of four (C₀, C₁, C₂ and C₇). Unfortunately, no other sampling times were tested, although C_0.5, C₃ and C₅ were available. AUC_0-12 Bayesian estimates were compared to the “reference” AUC_0-12 calculated with the linear trapezoidal rule, based on linear regression analysis, bias and precision RMSE%. The best prediction of AUC_0-12 was obtained with samples collected at 0, 1 and 2 hours after cyclosporine administration in the four groups of patients. In patients who did not receive concomitant enzyme inhibitors, a good correlation was observed (r² = 0.871), with a bias of 11.7% and a RMSE% of 13.4%. However, the predictive performance was not so good for patients on enzyme inhibitors, with a bias ranging between −14.2% for patients on ketoconazole and diltiazem and +19.0% for patients on diltiazem only. In these groups, squared correlation coefficients were 0.818 and 0.791 and RMSE% 16.9 and 22%, respectively.^[30]

In heart-lung transplantation, Bayesian estimators for cyclosporine AUC_0-12 forecasting and dose adjustment using a limited number of blood samples were also developed by our team,^[32] using data from a previous trial where repeated cyclosporine profiles were collected from 19 stable transplant patients (9 CF patients).^[25] Three PK models and Bayesian estimators were designed using a similar method as that presented for heart transplant patients above. The best sampling strategy was also 0, 1 and 3 hours in both groups of patients (CF and non-CF), allowing accurate estimation of AUC_0-12 in all patients. The comparison between AUC_0-12 calculated by non-linear regression (PK modelling using all available time points) and AUC_0-12 calculated using the Bayesian method found bias values of +3.0% (ns) and +5.3% (p < 0.05) in CF and non-CF patients, respectively, with a good precision in both groups (RMSE% = 8.7% and 10.0%, respectively). In non-CF patients (where bias was statistically significant, though small), the examination of individual AUC_0-12 showed relative estimation errors in the range of −20.2 to +23.7%, of which only 3 were outside the ±20% interval.^[32]

In conclusion, blood sampling schedules such as T₀, T_1h and T_2h or T_0h, T_1h and T_3h can easily be integrated into clinical practice because they do not require a prolonged hospital stay. Moreover, as opposed to multiple linear regression and single concentrations monitoring such as C₂, Bayesian forecasting is characterized by its flexibility in sampling times, as long as the true sampling times are reported to the pharmacologists. Another advantage is that such a method can estimate several PK parameters and exposure indices simultaneously. Therefore, it could help to identify absorption and elimination problems in diabetic or CF patients, for instance.

1.1.2.3. Impact of cyclosporine TDM on patient outcome

Studies on Bayesian forecasting for cyclosporine after thoracic transplantation have not evaluated patient and graft survival, and only a few have looked for relationships between global exposure as expressed by the AUC and surrogate markers of efficacy or toxicity. A number of studies describing other TDM tools, particularly C₀ and C₂, collected indicators of efficacy (e.g., acute rejection, graft loss, left ventricular ejection fraction (LVEF), graft atherosclerosis, mortality, forced expiratory volume in one second (FEV1), depending on the type of graft) (Table VI) or toxicity (renal dysfunction, infections) (Table VII).

Table VI. Cyclosporine exposure – efficacy studies.

Results are presented as mean±SD, unless otherwise specified

Ref	Number of patients	Immonsuppressants		Sampling periods	Analytical method, exposure			Event (number of episodes)	Comments
HEART TRANSPLANTATION
[54]	31 adults OHT 52±10 years	Cyclosporine D adjusted on C0:		244±178 days Median (range): 223 (22–582)	HPLC-U Calculation of AUC:			Routine surveillance EMB (ISHLT)	Trend for a significant relationship between C2 values and the incidence of rejection.
		< M6	200–275 μg/L		AUC0-12: MAP-BE[1] or sparse-sample algorithms[2–4]
		M6-M12	150–250 μg/L		AUC0-4: sparse-sample algorithm[5]
		> M12	100–200 μg/L
		D = 194±57 mg BID (4.8±1.4 mg/kg BID)			C0	215±68 μg/L		Rejection (≥ grade 2) n = 3 patients; all had C2 < 1250 μg/L	Number of patients experiencing rejection too small: To reach statistical significance based on C0 or C2 thresholds; To perform ROC curve.
					C2	949±204 μg/L
					AUC0-12	4,875±956 to 5,897±1,457 h.μg/L
					AUC0-4	2958±579 h.μg/L
					C0	242±62 μg/L (ns)		No rejection (< grade 2)
					C2	1,359±474 μg/L (ns)
					AUC0-12	6,723±2,119 to 7,445±1,871 h.μg/L (ns)
					AUC0-4	3970±1150 h.μg/L vs. 2958±579 (ns)
[93]	60 adults	Cyclosporine D adjusted on C0:		3–60 months	FPIA-AxSYM 4 AUC/patient = 240 AUC, calculated from C0, C2, C4, C6			At least one ARE grade 3A (ISHLT) in 20/60 patients	No difference between “rejection” and “no rejection” groups on mean C2.
		Y1	300–400 μg/L
		Y2	200–250 μg/L
		Y3	150–180 μg/L		Constant absorbers, AUC CV < 15% (n = 21)			Acute rejection	Variability on absorption = risk factor for rejection.
		> Y3	> 150 μg/L		AUC = 6,521 h.μg/L			n = 4 patients (19%)
		ATG Corticosteroids until M12 AZA or MMF (n = 10)			Inconstant absorbers, AUC CV > 15% (n = 39)			Acute rejection
					AUC = 6,751 h.μg/L (ns)			n = 16 patients (41%), p < 0.05
[55]	10 children out of 50 in the global trial	SAND or NEO, D adjusted on C0 (and blood cell count):		De novo ≥ 2 months	HPLC/MS			Rejection: IMEG, echocardiography, TDE, BPAR (≥ grade 2 – ISHLT)	Prospective blinded analysis Small number of patients with rejection: study not powerful enough? Impossible to define different C2 target levels in relation to adjunctive immunosuppressants
		≤ M1	200–300 μg/L					ROC analysis: best sensitivity and specificity for C2 < 600 μg/L – Se = 100%, Sp = 83% C2 < 600: rejection 14, no rejection 0 C2> 600: rejection 6, no rejection 30
		> M1	150–200 μg/L
		(D: median, range) ATG AZA or MMF
	8.1±6.3 years	D = 8.6 mg/kg (4–21)		2.2±1.6 years	C2	345±163 μg/L	< Y1: 456±164 μg/L	Rejection (BPAR: n = 14)	Impossible to assess chronic rejection
							Y1-Y2: 428±196 μg/L	n = 5 patients
							Y2-Y5: 244±147 μg/L
					Cmin	213±29 μg/L
					Cmax	467±38 μg/L
					AUC0-12	3,615±508 h.μg/L
					AUC0-4	1,498±132 h.μg/L
	8.9±5.6 years	D = 7.9 mg/kg (3.7–24), ns		2.4±1.8 years	C0: ns			No rejection
					C2 = 952±310 μg/L (p<0.001)		< Y1: 1,134±356μg/L	n = 5 patients
							Y1-Y2: 996±266 μg/L
							Y2-Y5: 859±274 μg/L
					Cmin	219±32 μg/L (ns)
					Cmax	966±231 μg/L (p<0.001)
					AUC0-12	5,530±889 h.μg/L (p < 0.001)
					AUC0-4	2,713±536 h.μg/L (p<0.001)
HEART AND LUNG TRANSPLANTATION
[80]	31 adults (12 CF) HL	Cyclosporine median D (mg/kg/day) starting: 2.7 maintenance: 10.3 (CF patients: TID) ATG, AZA, Prednisone		< 3 months	SRIA			Treated ARE (66% BPAR)	Variability on C0: risk factor for subsequent rejection; no such relationship for C0 itself
					Observed C0 (median):	< W3: 347.5 μg/L		If C0 CV > 40%:
						W3-M3: 445.5 μg/L		RR = 1.51 (95%CI = 1.01–2.27)
[28]	50 patients (9 CF) 19 SL, 9 DL, 22 HL	NEO (n = 28) vs SAND (n = 22), D adjusted on C0		12 months	EMIT 289 PK profiles				Study designed to compare the PK of NEO vs SAND, not to establish a relationship between drug exposure and efficacy
		M1-M2	300–400 μg/L
		M3-M12	200–300 μg/L		C0 = 394 μg/L			Patients with at least one treated ARE
		ATG, AZA, prednisolone			C0 = 449 μg/L (p = 0.042)			Patients with no ARE	Relationship C2, C6 or AUC0-6 – incidence of treated AR: ns
[20]	48 patients SL, DL, HL	NEO (n = 27) vs SAND (n = 21), D adjusted on C0:		Weeks 1, 2, 3, 4, 13, 26, 39, 52	EMIT 341 AUC0-6
		M1-M2	300–400 μg/L					ARE frequency similar in the 3 C2 groups
		M3-M12	200–300 μg/L		3 groups based on C2 (μg/L) at M1: low, < 1000 (18); intermediate, 1000–1500 (16); high >1500 (14)			AR-free periods: Intermediate C2 (226 days) ≫ low C2 (197 days) and high C2 (191 days) (p<0.0001)
		(CF patients: TID) ATG, AZA, prednisolone			C2	875±546 μg/L		Median number of ARE > 2
					AUC0-6	4,036±1,904 h.μg/L
					C2	1,114±633 (p = 0.01)		Median number of ARE ≤ 2
					AUC0-6	4,870±2,182 (p = 0.01)
					C0 and C6: ns

ARE: Acute rejection episode – ATG: Antithymocyte globulin – AZA: Azathioprine – BPAR: Biopsy-proven acute rejection – CF: Cystic fibrosis – D: Dose – DL: Double lung transplantation – EMB: Endomyocardial biopsy – HL: Heart-lung transplantation – IMEG: Intramyocardial electrocardiogram – MAP-BE: Maximum a posteriori Bayesian estimation – MMF: mycophenolate mofetil – NEO: Neoral – OHT: Orthotopic heart transplantation – SAND: Sandimmune – SL: Single lung transplantation – TDE: Tissue Doppler echocardiography.

Table VII. Cyclosporine exposure – toxicity studies.

Results are presented as mean±SD, unless otherwise specified

Ref	Number of patients	Immunosuppressants		Sampling periods	Analytical method, exposure		Event (number of episodes)		Comments
HEART TRANSPLANTATION
[54]	31 adults OHT 52±10 years	Cyclosporine D adjusted on C0:		244±178 days 223 (22–582) Median (range):	HPLC-UV Calculation of AUC: AUC0-12: MAP-BE[1] or sparse-sample algorithms[2–4] AUC0-4: sparse-sample algorithm[5] Rejection vs no rejection:		Impairment of renal function: If SCr > 2.5 mg/dL n = 5 patients
		< M6	200–275 μg/L
		M6-M12	150–250 μg/L
		> M12	100–200 μg/L				No relationship between cyclosporine C0 or C2 and impairment of renal function.
		D = 194±57 mg BID (4.8±1.4 mg/kg BID)			C0	215±68 μg/L
					C2	949±204 μg/L
					AUC0-12	4,875±956 to 5,897±1,457 h.μg/L
					AUC0-4	2958±579 h.μg/L
					C0	242±62 μg/L (ns)
					C2	1,359±474 μg/L (ns)
					AUC0-12	6,723±2,119 to 7,445±1,871 h.μg/L (ns)
					AUC0-4	3970±1150 h.μg/L vs. 2958±579 (ns)
[93]	60 adults	Cyclosporine D adjusted on C0:		3–60 months	FPIA-AxSYM		Mean SCr rise of 50% from baseline in both groups (ns)
		Y1	300–400 μg/L		4 AUC/patient = 240 AUC, calculated from C0, C2, C4, C6		% patients:
		Y2	200–250 μg/L
		Y3	150–180 μg/L		Constant absorbers, AUC CV < 15% (n = 21)		Hypertension	48%
		> Y3	> 150 μg/L		AUC = 6,521 h.μg/L		Neurological events	24%
		ATG Corticosteroids until M12 AZA or MMF (n = 10)					Gum hyperplasia	14%
							New onset hyperlipidemia treated with statins	48%
					Inconstant absorbers, AUC CV > 15% (n = 39)		Hypertension	56%
					AUC = 6,751 h.μg/L (ns)		Neurological events	41%
							Gum hyperplasia	18%
							New onset hyperlipidemia treated with statins	51%
LUNG TRANSPLANTATION
[79]	15 adults (0 CF) 3 SL, 12 DL	Cyclosporine D adjusted on C0 = 200–300 μg/L AZA, methylprednisolone		42–359 days	EMIT				C0 and C2: same sensitivity (84.6%) Better specificity of C0 (66.7%) vs C2 (57.1%)
					C0 = 289±93 μg/L		CMV infection or disease
					C2 = 1,583±421 μg/L		n = 9 patients (13 episodes)
					C0 = 222±82 μg/L (p<0.05)		No CMV infection or disease
					C2 = 1,074±423 μg/L (p<0.05)
HEART AND LUNG TRANSPLANTATION
[77]	32 patients 24 HTx, 28 HL	Cyclosporine D adjusted on C0 (target: N/R) ATG, AZA, steroids		< 3 months	SRIA First 5 postoperative days, 88% C0 < 200 μg/L		Renal function (624 complete sets of dose, C0 and SCr) No correlation between dose and change in renal function Correlation between log C0 and 1/SCr: Simultaneous measurements: r2 = −0.33 (95%CI = −0.52;−0.10) In the subsequent 5-day period: r2=−0.69 (95%CI = −0.69;−0.15)		Exclusion of all data up to the first nadir in plasma creatinine following transplantation (mean time: 6.7±3.6 days)
[28]	50 patients (9 CF) 19 SL, 9 DL, 22 HL	NEO (n = 28) vs SAND (n = 22), D adjusted on C0:		12 months	EMIT 289 PK profiles		Renal function: None of C0, C2, C6 or AUC0-6 correlated with any of the markers of cyclosporine nephrotoxicity		Study designed to compare the PK of NEO vs SAND, not to establish a relationship between drug exposure and efficacy
		M1-M2	300–400 μg/L
		M3-M12	200–300 μg/L		Patients with at least one treated ARE C0 = 394 μg/L
		ATG, AZA, prednisolone			Patients with no ARE C0 = 449 μg/L (p = 0.042)
[20]	48 patients SL, DL, HL	NEO (n = 27) vs SAND (n = 21), D adjusted on C0:		Weeks 1, 2, 3, 4, 13, 26, 39, 52	EMIT 341 AUC0-6 3 groups based on C2 (μg/L) at M1: low, < 1000 (18); intermediate, 1000–1500 (16); high >1500 (14)		Infections (71 episodes): Patients with infection episodes > 2 vs ≤ 2: C0, C2, C6, AUC ns Number of infection episodes ns among patients stratified on C2
		M1-M2	300–400 μg/L
		M3-M12	200–300 μg/L
		(CF patients: TID)
		ATG, AZA, prednisolone			Median number of ARE > 2		Renal function (1,693 observations): Difference between groups: ns
					C2 = 875±546 μg/L
					AUC0-6 = 4,036±1,904 h.μg/L
					Median number of ARE ≤ 2		Blood pressure (1,891 measurements): Linear trend ↗ BP – C2 strata (p < 0.001) No correlation between BP and exposure indices
					C2 = 1,114±633 (p = 0.01)
					AUC0-6 = 4,870±2,182 (p = 0.01)
					C0 and C6: ns

ARE: Acute rejection episode – ATG: Antithymocyte globulin – AZA: Azathioprine – BP: Blood pressure – CF: Cystic fibrosis – D: Dose – DL: Double lung transplantation – HL: Heart-lung transplantation – HTx: Heart transplantation – MAP-BE: Maximum a posteriori Bayesian estimation – MMF: Mycophenolate mofetil – NEO: Neoral – N/R: not reported – OHT: Orthotopic heart transplantation – SAND: Sandimmune – SCr: Serum creatinine – SL: Single lung transplantation.

1.1.2.3.1. C₀ monitoring

Various observational studies sought for a relationship between C₀ and clinical outcome, with contrasted conclusions (Tables VI to VIII).

Table VIII.

Cyclosporine TDM or concentration-controlled studies

Ref	Study	Patients	Time post-transplant	Immunosuppressants	Analytical method, exposure index and target		Evaluated endpoints and results	Comments
HEART TRANSPLANTATION
[59]	Prospective study Consecutive patients, randomized in 2 groups	20 adults	< 1 year	SAND	EMIT		BPAR: ISHLT classification
		Group I 53±8 years n = 10	11±2 months	Thymoglobuline AZA Prednisone	D adjusted on C0 = 150–250 μg/L		Efficacy: BPAR: 50% (grade I: 70/172 biopsies: 40.7%)
							Others: Cyclosporine D at end of f/up : 3.5±1 mg/kg/day Mean cyclosporine D adjustment/patient: 1/month Total cyclosporine cost: 5,106±1,045 CDN $
		Group II 58±4 years n = 10	10±3 months		D adjusted on C6 = 150–250 μg/L		Efficacy: BPAR: 50% (grade I: 47/111 biopsies: 29.7%, p=0.04) Left VEF and graft atherosclerosis similar in both groups
							Toxicity: Renal function similar in both groups (SCr, GFR)
							Others: Cyclosporine D at end of f/up : 2.6±0.6 mg/kg/day (p=0.002) Mean cyclosporine D adjustment/patient: 1/month Total cyclosporine cost: 3,589±1,116 CDN $ (p=0.005)
[56]	Prospective study Randomization to C0 or C2 monitoring Follow-up: 6 months	30 adults	Stable, ≥ 1 year No BPAR in previous 6 months SCr < 250 μmol/L	NEO (Week 0: conversion from SAND to NEO) Thymoglobuline AZA Prednisone	EMIT		Routine biopsies and after each dose reduction No instance of BPAR ≥ grade 2 (ISHLT classification)	No comparison of outcome between two groups based on monitored exposure index: this study rather looked for a correlation between a single time-point concentration and AUC0-4 and AUC0-12
		Group I 59±7 years n = 15			Week 2: D adjusted on C0 = 100–200 μg/L ≥ Week 6: D adjusted on C2 = 300–600 μg/L		D increase in 8 patients after the first visit D reduction in all patients after the second visit C0 = 147±39 μg/L C1 = 762±385 μg/L C2 = 835±320 μg/L C4 = 379±138 μg/L AUC0-4 (Trap) = 2,467±804 h.μg/L
		Group II 56±9 years n = 15			Week 2: D adjusted on C2 = 200–400 μg/L ≥ Week 6: D adjusted on C2 = 300–600 μg/L		D reduction after the first visit C0 = 102±35 μg/L (p = 0.0001) C1 = 621±402 μg/L (ns) C2 = 555±271 μg/L (p = 0.01) C4 = 218±83 μg/L (p = 0.001) AUC0-4 (Trap) = 1,723±808 h.μg/L (p = 0.02)
[10]	Longitudinal follow-up of all patients followed at heart transplantation clinic 2-phase study	114 adults 57±9 years	Stable, ≥ 1 year	NEO ATG (most patients) Cyclosporine monotherapy (7.9%) or + AZA + prednisone (40.3%) or + AZA (36%) or + prednisone (15.8%)	EMIT		Primary endpoint: clinical benefit = composite measure Positive cardiac outcomes (no mortality, no acute rejection, no decrease in LVEF > 10%), and Positive renal outcomes (absence of increase in SCr > 10%)	Limitation: sequential longitudinal assessment of two strategies for NEO D adjustment in the same patients
				Phase 1 (10±4 months)	D adjusted on C2 = 300–600 μg/L		Primary endpoint: Clinical benefit: 69.3%
				Initial			Secondary endpoints: Incidence of acute rejection: 0.87% Incidence of mortality: 7.9% Incidence of increase in SCr > 10%: 18.1%
				D = 2.6±0.8 mg/kg/day	C0 = 136±42 μg/L
					C2 = 776±316 μg/L
				Final
				D = 1.9±0.6 mg/kg/day	C0 = 87±36 μg/L
					C2 = 422±153 μg/L
				Phase 2 (10±2 months)	Dose adjusted on		Primary endpoint: Clinical benefit: 43.3% RR = 1.6 (p = 10−5)
					C0 = 100–200 μg/L
				Initial
				D = 1.9±0.6 mg/kg/day	C0 = 86±35 μg/L		Secondary endpoints: Incidence of acute rejection: 0.96% (ns) Incidence of mortality: 9.6% (ns) Incidence of increase in SCr > 10%: 48.9% RR = 0.37 (p<10−4) LVEF: stable across both time periods
					C2 = 428±159 μg/L
				Final
				D = 2.4±0.6 mg/kg/day	C0 = 134±46 μg/L
					C2 = 592±156 μg/L
[22]	Prospective analysis of 2 consecutive periods		De novo f/up: 4 years	NEO In absence of ARE, D decreased if SCr increased > 20% from baseline ATG, MMF, prednisone	EMIT		Efficacy Routine EMB: ARE ≥ grade 3A (ISHLT classification) Echocardiography Coronary angiography Toxicity Renal function: SCr, ClCr (Cockroft-Gault), GFR	MMF D: group 1 ≠ 2 D/kg different at M1
		13 adults 54±9 years		Group I (MMF D: 1 g BID)	< M9: D adjusted on C0 < M3: 200–300 μg/L M4-M6: 150–250 μg/L M6-M9: 100–200 μg/L > M9: D adjusted on C2 =400–600 μg/L		Efficacy Incidence of ARE ≥ grade 3A Month 6: 38.5% Month 12: 38.5% Month 24: 46% Month 36: 54% Time to acute rejection: 363±462 days 3 deaths LVEF: 58±5 (M3) to 60±5% (M12) Incidence of graft atherosclerosis at M12: 11%
		9 adults 53±12 years		Group II (MMF D: 1.5 g BID)	D adjusted on C2 < M3: 600–800 μg/L M4-M6: 500–700 μg/L > M6: 400–600 μg/L		Efficacy Incidence of ARE ≥ grade 3A (ns) Month 6: 11% Month 12: 44% Month 24: 56% Month 36: 56% Time to acute rejection: 344±237 days (ns) No death LVEF: 59±2 (M36) to 61±2% (M24) (ns) Incidence of graft atherosclerosis at M12: 0% (ns) Toxicity: Renal function, incidence of infections: ns
[86]	2-phase prospective study Paired determination of C0 and C2 Investigators blinded to exposure index	58 adults 56±11 years	2.03±1.28 years (0.3–3.6)	Cyclosporine ATG or basiliximab Prednisone AZA or MMF	FPIA-TDx		Toxicity: Incidence of life-threatening infections Renal dysfunction (SCr, ClCr) Efficacy BPAR ≥ grade 2 (ISHLT)	Short follow-up in both phases. C0 or C2 values in the whole population for each phase: N/R. No monitoring of MPA levels.
					Phase I (6 months): D adjusted on C0 M3-M6: 300–350 μg/L M6-M12: 250–300 μg/L M12-M24: 200–250 μg/L >M24: 150–200 μg/L		Toxicity 8 infection episodes Renal function deterioration in 2 patients Efficacy: ARE ≥ grade 3, 7/58 patients Rejection vs no rejection: C0 = 195±121 vs 197±100, ns C2 = 777±326 vs 1,015±422, p=0.022
					Phase II (6 months): D adjusted on C2 associated with no rejection during phase I M3-M6: 1,403±285 M6-M12: 1,175±215 M12-M24: 947±170 >M24: 824±120		Toxicity 6 infection episodes (ns) Renal function deterioration in 1 patient; no difference in SCr and calculated ClCr Efficacy: ARE ≥ grade 3, 6/56 patients (ns) Rejection vs no rejection: C0 = 204±85 vs 209±138, ns C2 = 765±297 vs 967±470, p=0.03
[51]	Prospective randomized controlled study f/up: 6 months	125 adults	Stable, > 1 year	Cyclosporine	CEDIA			Short follow-up duration. Low incidence of AR after the first year post-transplant. Study not powered to detect a small possible effect of C2 vs C0 monitoring on AR.
		Group I n = 62 55±10 years	6.4±2.8 years	± Prednisone ± MMF or AZA	D adjusted on C0 = 80–120 μg/L		Primary endpoint Decrease of cyclosporine D: 11 mg/day Secondary endpoints Deaths: 1 patient ARE: no suspicion Infections: 9.7% ClCr: + 0.54 mL/min
		Group II n = 63 55±9 years	6.4±2.5 years		D adjusted on C2 = 300–600 μg/L		Primary endpoint Decrease of cyclosporine D: 26 mg/day (p = 0.0025) – C0 in target range Secondary enpoints Deaths: 1 patient ARE: no suspicion Infections: 15.9% (p = 0.14) ClCr: −0.16 mL/min (p = 0.61)
HEART AND LUNG TRANSPLANTATION
[88]	Single center study using historic controls Intent-to-treat analysis		De novo	Cyclosporine	EMIT			Larger proportion of DL (p < 0.01) and of CF patients (p < 0.01) in C2 vs C0 group.
	Transplantation between 1989 and 2000	338 patients 124 SL, 150 DL, 64 HL		ATG(1) AZA Prednisolone	C0 group Target: N/R		Primary endpoint : efficacy (ISHLT classification) Freedom from AR: M3: 51% – M6: 45% – M12: 41% Freedom from BOS: Y1: 87% – Y3: 53% – Y5: 36% Actuarial survival: 30-days: 93% – Y1: 83% – Y3: 66% – Y5 : 55% Secondary endpoint: toxicity Chronic renal failure within 5 years: 18 patients
	Transplantation between 2001 and 2002	50 patients 3 SL, 44 DL, 3 HL	f/up: 16–1,790 days (1,185±426)	AZA (n = 15) or MMF (n = 35) Prednisolone	C2 group		Primary endpoint : efficacy (ISHLT classification) 15 patients with ARE (23 episodes/171 TBB) 18/23 episodes within 2 weeks of subtherapeutic level 12 patients with ARE/25 with C2 subtherapeutic levels Freedom from AR (p = 0.001) M3: 74% – M6: 69% – M12: 69% Freedom from BOS (p = 0.002) Y1: 98% – Y3: 79% – Y5: 59% Actuarial survival (p = 0.029) 30-days: 98% – Y1: 94% – Y3: 82% – Y5 : 77% Secondary endpoint: toxicity Chronic renal failure in one patient, with SCr > 250 mg/L (needed dialysis and renal transplantation); (p > 0.10) SCr = 122±64 mg/L
					Day 0–2	> 800
					Day 2–7	> 1,200
					Day 8–30	1,200–1,700
					Day 30–60	1,200–1,500
					Day 61–90	800–1,200
					Day 91–180	700–1,000
					Day 181–365	600–900
					Day >365	600–800
LUNG TRANSPLANTATION
[92]	Sequential groups	36 adults (17 CF) DL	De novo f/up: 3 months	Cyclosporine Prednisolone	EMIT			C0 group older than C2 group Baseline SCr lower in C0 group Associated immunosuppressants different between the 2 groups
		n = 18		AZA (n = 14) or MMF (n = 4)	D adjusted on C0 Week 1: 450 μg/L Month 3: 250 μg/L		Efficacy FEV1 at M3: no difference AR (A and B): similar rates in both groups (numbers: N/R) Survival at M3: 100%
		n = 18		AZA (n = 2) or MMF (n = 16)	D adjusted on C2 Week 1: 1,200 μg/L Month 3: 800 μg/L		Toxicity SCr: greater increase from baseline in C0 group Infections: similar rates in both groups (1.85 vs. 1.79 events per 100 patients-days)
[87]	Uncontrolled single-center pilot study	15 adults with renal dysfunction (5 CF) 5 SL, 10 DL	3.5±2.7 years (0.2–9.0)	Cyclosporine AZA (n = 10) or MMF (n = 5) Prednisolone	EMIT Switch from C0 to C2 monitoring C2 = 300–600 μg/L		Primary endpoint Renal function improvement at 3 and 12 months: SCr = 0.20±0.07 mmol/L at baseline to 0.15±0.05 mmol/L at M3 (p < 0.001) Secondary endpoint Stable lung function except for 1 ARE Cyclosporine D (mg/kg/day) divided by two within 3 months:6.4±7.3 to 3.9±3.7 at M3 (p = 0.031) to 3.1±2.7 at M12 (p = 0.041)

⁽¹⁾ Patients transplanted before 1995

AR: Acute rejection – ARE: Acute rejection episode – ATG: Antithymocyte globulin – AZA: Azathioprine – BOS: Bronchiolitis obliterans syndrome – BPAR: Biopsy proven acute rejection – ClCr: Creatinine clearance – D: dose – EMB: Endomyocardial biopsy – f/up: Follow-up – GFR: Glomerular filtration rate – LVEF: Left ventricular ejection fraction – M: Month – MMF: Mycophenolate mofetil – NEO: Neoral – N/R: not reported – SAND: Sandimmune – SCr: Serum creatinine – TBB: Transbronchial biopsy – Y: Year.

A significant relationship between C₀ and clinical outcome was established in a series of studies.

In a study in 48 adult heart transplant recipients, on a steroid-free regimen containing Sandimmune^® and azathioprine, who were at least 8 months beyond transplantation, an inverse relationship was found between cyclosporine C₀ and the probability of cellular rejection (p < 0.03):^[74] C₀ = 211 μg/L was associated with a 2.5% probability of rejection ≥ grade 1B and a 1.1% probability of rejection ≥ grade 3A. The lowest probability of acute cellular rejection was obtained with C₀ > 300 μg/L. On the other hand, no correlation was found between cyclosporine C₀ and simultaneously obtained serum creatinine measurements, but this may well be due to the partly cumulative mechanism of cyclosporine nephrotoxicity.

Results from a retrospective analysis of 1,407 endomyocardial biopsies (EMB) and cyclosporine concentrations in 105 heart transplant adults (time post transplantation not reported) showed significantly lower C₀ in patients with grade 3A cellular rejection compared to patients with no rejection (173 [95%CI = 148–197] vs 206 [95%CI = 182–230] μg/L, p = 0.005).^[75] Again, no correlation was found between cyclosporine C₀ and serum creatinine levels.

In a randomized clinical trial performed in 50 lung or heart-lung transplant recipients, the mean C₀ observed in patients who did not experience acute rejection was 449 vs 394 μg/L in patients who had at least one rejection episode (p = 0.042).^[28] Once again, C₀ did not correlate with any of the markers of nephrotoxicity. Based on those results, the authors recommended the use of C₀ rather than C₂ as a monitoring strategy. However, this study was not designed to establish a potential relationship between exposure parameters and effects but to compare clinical outcomes between patients on Sandimmune^® and patients on Neoral^®. In a retrospective analysis of 32 heart or heart-lung transplant recipients in their first three months post-transplantation, the same authors had previously reported a relationship between log(C₀) and renal function (1/Cr) evaluated at the same time or in the subsequent 5-day period, showing r² = −0.33 and −0.69, respectively.^[76,77]

In a recent study in 70 de novo heart transplant adults, mean C₀ levels over the first week post-transplantation were significantly lower in patients with vs without acute rejection during the first year post-transplantation (126±56 vs 169±48 μg/L, p = 0.003).^[78] ROC curve analysis showed that patients whose mean C₀ during the first week was > 150 μg/L had a significantly lower incidence of acute rejection (30.3 vs 64.9%, p = 0.009).

In a small study performed in 15 adult lung transplant recipients on cyclosporine (dose-adjusted on C₀, same target whatever the period) during the first year post-transplantation, significantly higher C₀ were reported in patients who developed CMV infection or disease than in patients who did not (289±93 vs 222±82 μg/L, p < 0.05).^[79] C₀ and C₂ displayed the same sensitivity (84.6%) but C₀ yielded better specificity (66.7%) than C₂ (57.1%).

Cyclosporine C₀ may not always be an appropriate TDM tool, however, as shown by a series of studies in which there was no correlation between C₀ and clinical outcomes.

In 31 heart-lung transplanted adults followed during the first 3 months after transplantation, no relationship was found between C₀ and subsequent rejection. The authors put more emphasis on the evidence of C₀ variability as a risk factor for subsequent rejection, with intra-individual CV > 40% resulting in a relative risk of 1.51 (95%CI = 1.01–2.27) for rejection.^[80]

In a more recent study in 48 patients who underwent lung or heart-lung transplantation, treated by either Sandimmune^® or Neoral^® and followed for one year after transplantation, no statistically significant difference on C₀ was found between patients who experienced 2 acute graft rejection episodes or more, and patients who experienced less than 2 episodes. No relationship was found either between C₀ and infections or the deterioration of renal function.^[20]

The same observations were made in another study performed in 31 heart transplant recipients between 3 weeks and 2 years post-transplantation, cyclosporine dose-adjusted on C₀, and in whom no relationship was found between C₀ and acute graft rejection or impairment of renal function.^[54]

1.1.2.3.2. C₂ monitoring

A significant number of observational studies demonstrated that C₂ monitoring is an efficient TDM tool in solid organ transplant patients receiving cyclosporine: it allows the reduction of incidence of acute and chronic rejection and of hypertension after transplantation. Unfortunately, most of these studies were performed in hepatic or renal transplantation.^[81–83] As a consequence, although C₂ targets were validated in various other solid organ transplant settings,^[81–83] there is no published consensus to date for C₂ target levels after heart or lung transplantation.

In thoracic transplantation, a number of authors succeeded in establishing a relationship between C₂ and clinical outcomes and thus advocated C₂ monitoring.^{[10,56,84–88]}

C₂ was an excellent tool to predict AUC_0-12 and AUC_0-4 in CF and non-CF patients in small single-center studies.^{[4,6,21,25,54]} A longitudinal follow-up was conducted in 114 adults who had undergone heart transplantation more than one year prior to the study.^[10] Follow-up was performed in two phases: cyclosporine dose was adjusted on C₂ during the first phase (target: 300 to 600 μg/L) then adjusted on C₀ during the second phase (target: 100 to 200 μg/L). The primary endpoint was the clinical benefit, defined by the authors as a composite criterion comprised of positive cardiac outcomes (no mortality, no acute rejection, no decrease in LVEF > 10%) and positive renal outcomes (absence of increase in serum creatinine > 10%). There was no difference between the two phases in terms of incidence of acute rejection or mortality, but there was a significant clinical benefit of C₂ monitoring vs C₀ monitoring, with a relative “risk” of positive outcome of 1.6 (p = 10⁻⁵). Moreover, there was a lower incidence of serum creatinine increase > 10% during phase 1, with a relative risk of 0.37 (p < 10⁻⁴). However, one limitation of this study is that it was a non-randomized sequential assessment of 2 strategies for Neoral^® dose adjustment: as this concerns a chronic, progressive pathology, there is no certainty that patients were comparable between phases I and II in terms of both cardiac and renal functions.

Another prospective study was performed in 30 stable heart transplant adults who were randomized to either C₀ or C₂ monitoring.^[56] None of the patients experienced biopsy-proven acute rejection ≥ grade 2 according to the ISHLT classification. As a consequence, this study could not report any relationship between any monitoring tool and the clinical outcome, and was rather a study on the correlation between single concentration time points and AUC_0-4.

In a similar prospective study, Delgado et al performed parallel determination of C₀ and C₂ levels in 58 adults with orthotopic heart transplantation.^[86] Follow-up consisted of two phases: during phase 1 (6-months follow-up on C₀), the authors found no significant difference on C₀ between patients who experienced rejection and those who did not (195±121 vs 197±100 μg/L, p = 0.96) but C₂ was lower in patients with rejection than in patients without (777±326 vs 1,015±422 μg/L, p = 0.022). During phase 2 (6-months follow-up on C₂), no significant difference was found between C₀ levels (204±85 vs 209±138 μg/L, p = 0.88), whereas rejection was associated with lower C₂ levels (765±297 vs 967±470 μg/L, p = 0.03). There was no difference on serum creatinine, creatinine clearance or infections between the two phases. As a consequence, the authors concluded that higher C₂ levels were associated with significantly fewer rejection episodes and that C₂ monitoring was safe in terms of preservation of renal function and infection rates.

In lung transplantation, Glanville et al conducted a single-arm, single-center pilot study on 15 stable transplanted adults with renal dysfunction.^[87] Cyclosporine monitoring was switched from C₀ to C₂ (C₂ target: 300 to 600 μg/L), resulting in cyclosporine dosage being divided by two within three months, with a significant improvement in renal function. Lung function remained stable in all patients, except for one acute rejection episode. Based on these results, the authors concluded that C₂ monitoring allows safe dose reductions in patients with altered renal function, in cases where AUC monitoring is not feasible.

The same team performed a study in 50 de novo lung or heart-lung transplant patients, 20 of whom with CF, using 338 (73 CF) historic controls.^[88] The authors concluded that C₂ monitoring and dose adjustment on C₂ brought a clinical benefit compared to C₀ monitoring in terms of survival, incidence of acute rejection, occurrence of BOS and renal dysfunction. Unfortunately, the comparison with historic controls is of limited interest given the significant time bias and the lower proportion of patients with CF or bilateral lung transplant in the control group.

In another study, Caforio et al retrospectively compared C₀ vs C₂ as predictors of rejection and renal dysfunction in 269 adults after 1 year post-heart transplantation.^[85] Dose adjustment was based on C₀ (target: 100 to 250 μg/L). Patients with higher C₂ levels (> 740 μg/L) had higher severe rejection score at 2 years than patients with lower C₂ (p = 0.02), whereas no significant association was found with C₀. However, based on the rejection history of the patients, the authors used higher cyclosporine dose and C₀ targets even in the long term in patients considered high rejectors vs the low rejectors, which probably accounts for this quite unexpected result.

Finally, Baraldo et al compared the distribution and variability of C₂ values after Neoral^® twice daily (BID) or three times daily (TID) between 2 groups of 25 stable heart transplant recipients followed for one year.^[84] Cyclosporine dose adjustment was based on C₀ (target: 250 to 350 μg/L). Significantly higher C₂ variability was observed in the BID than in the TID regimen (CV = 25.3 vs 15.1%, p < 0.001), but it is not quite clear whether the authors referred to intra-patient or inter-patient variability or a mixture of the two (overall C₂ variability). There was no significant difference in rejection incidence or cardiac function (LVEF) between the 2 groups, but the TID regimen led to a lower incidence of nephrotoxicity. Based on these results, the authors suggested that in stable heart transplant patients, cyclosporine TID would lead to similar (or better, considering a lesser variability on cyclosporine concentrations) cardiac outcomes than BID, with decreased nephrotoxicity (decreased C₂ and decreased variability on C₂), with a target range for C₂ of 650 to 850 μg/L. However, these statements are speculative and have never been confirmed by a prospective trial.

To our knowledge, only one prospective study compared C₀ to C₂ monitoring, in 50 de novo pediatric heart transplant recipients.^[55] No significant difference was found on C₀ between patients with or without acute graft rejection, but C₂ was significantly higher in children who did not experience acute rejection, whatever the post-transplantation period. The ROC curve analysis for C₀ and C₂ found better sensitivity and specificity for C₂, with respective values of 100% and 83% for a C₂ target of 600 μg/L.

On the other hand, many observational studies failed to evidence a relationship between C₂ and clinical outcome. In 7 consecutive male adults on Neoral^® followed for one year after heart transplantation, no statistically significant difference in C₂ was found between patients with and without acute rejection, which might be due to the fact that cyclosporine dose was adjusted in all the patients according to C₂ (target range: 300–600 μg/L)^[89] and, above all, that too few patients were included in order to perform a powerful analysis.

In another retrospective study, Cantarovich et al analyzed the results of 517 EMB from 39 adult heart transplant recipients during the first year post-transplantation.^[90] When EMB were split into 2 groups based on the degree of acute rejection (≤ 2 or ≥ 3A according to the ISHLT classification), no statistically significant difference in C₂ values was found. However, cyclosporine monitoring was not homogeneous in all patients, being based on C₀ in 13 of them and on C₂ in the 26 others, and cyclosporine was assayed using different analytical methods throughout the study (EMIT or CEDIA), which may have introduced confusing factors in the analysis. In another study, a prospective analysis of 2 consecutive periods was performed in 22 de novo heart transplant patients.^[22] In group I, follow-up was based on C₀ whereas in group II, follow-up was based on C₂ (targets depending on time post-transplantation). Of note, all patients received MMF but the dose was different between the groups (1 g twice daily in group I vs 1.5 twice daily in group II). No significant difference was found between the 2 groups on the following: the incidence of acute rejection episodes ≥ grade 3A according to the ISHLT classification, time to acute rejection, death, LVEF, incidence of graft atherosclerosis, nephrotoxicity or infections.

More recently, Cantarovich et al performed a multi-centre randomized open-label study in 87 de novo adult heart transplant recipients, with the objective of determining the minimal effective exposure to cyclosporine with C₂ monitoring.^[91] Patients were stratified into 2 cohorts based on renal function, and three C₂ ranges were defined (high, intermediate and low C₂), which varied with time post-transplantation. Again, the study failed to show any statistically significant difference between groups, in terms of acute rejection incidence or renal function. The authors finally concluded that it was safe to monitor Neoral^® with a “low” C₂ range (< M1: 1,200–1,400 μg/L; M2-M3: 1,000–1,200 μg/L; M4-M5: 800–1,000 μg/L; M6-M12: 700–900), preserving renal function without increasing the risk of acute rejection.

The studies performed by Trull et al^[28] and Akhlaghi et al^[20] in 50 lung or heart-lung transplant recipients during the first year post-transplantation did not show any relationship between C₂ and the incidence of treated acute rejection episodes, and C₂ did not correlate with any of the markers of cyclosporine nephrotoxicity.

A prospective randomized controlled study, the aim of which was to compare C₀ (target: 80 to 120 μg/L) to C₂ (target: 300 to 600 μg/L) monitoring in terms of clinical outcomes, was performed in 125 heart transplant recipients after 1 year post-transplantation, followed-up for 6 months.^[51] The study failed to detect a possible effect of C₂ vs C₀ monitoring on acute rejection, infections or nephrotoxicity, probably because it was not powered to detect a small effect. However, a significant difference between the two groups was found for the primary endpoint: cyclosporine daily dose decreased by 26 mg over the follow-up period in patients monitored on C₂, vs 11 mg in patients monitored on C₀ (p = 0.0025).

In lung transplantation, a study was performed on 2 sequential groups of 18 de novo bilateral lung transplant patients, 17 of whom had CF.^[92] In the first group, cyclosporine dose was adjusted on C₀ (target at W1: 450 μg/L; at M3: 250 μg/L) whereas in the second group, dose was adjusted on C₂ (target at W1: 1200 μg/L; at M3: 800 μg/L). Acute rejection rates were similar in both groups and there was no statistically significant difference on FEV1 and on the incidence of infections between the two groups of patients. A greater increase of serum creatinine from baseline was observed in the C₀ group. However, these two groups were not comparable as patients in the C₀ group were older and their baseline serum creatinine was lower (no mention was made of creatinine clearance). Moreover, the associated immunosuppressive treatment was different, as 14 of the 18 patients in the C₀ group were on azathioprine whereas 16 of the 18 patients in the C₂ group were on MMF.

1.1.2.3.3. C₆ monitoring

One study in heart transplantation investigated the use of C₆ as monitoring tool. In this prospective study, 20 adults were randomized into 2 groups: in the first group, monitoring was based on C₀ (target: 150 to 250 μg/L) whereas in the second group, cyclosporine dose was adjusted on C₆ (same target). The two groups were similar in terms of efficacy and toxicity, but patients of group II received a significantly lower dose of cyclosporine (2.6±0.6 vs 3.5±1 mg/kg/day, p = 0.002), with a significantly lower total cyclosporine cost (3,589±1,116 vs 5,106±1,045 CDN $, p = 0.005).^[59] The result is however evident and directly depends on the protocol, but in no way guarantees that C₆ is as safe and as efficient as C₀ monitoring on the long term.

1.1.2.3.4. AUC monitoring

Our literature search allowed us to identify only three studies in which a potential relationship between AUC and clinical outcomes was evaluated (Tables VI & VII).

Thirty-one patients were included in the first study, in which the main objective was to determine the clinical significance of C₂ compared with C₀ following heart transplantation.^[54] No significant difference on AUC_0-12 and AUC_0-4 was found between patients who experienced acute rejection and those who did not (whether these AUCs were calculated by Bayesian estimation or using sparse-sampling algorithms).

The aim of the second study, performed in 50 children, was to assess the relative value of C₀ and C₂ monitoring to prevent acute rejection in pediatric de novo heart transplant patients.^[55] A subgroup analysis was performed in 10 children, 5 with and the other 5 without acute rejection. A statistically significant difference was found between the two groups on AUC_0-12 (rejectors vs non-rejectors: 3,615±508 vs 5,530±889, p < 0.001) and on AUC_0-4 (rejectors vs non-rejectors: 1,498±132 vs 2,713±536, p < 0.001), both calculated using the linear trapezoidal rule. However, this is only a case-control study in a very small number of patients, which does not permit definitive conclusions to be drawn, even more so in that the first study (also conducted in a small population) did not confirm this finding.

The third study enrolled 60 heart transplant patients, who were divided in 2 groups: “constant absorbers” (group A, CV of mean individual AUC < 15%); and “inconstant absorbers” (group B, CV of mean individual AUC > 15%).^[93] Mean AUC levels were not different between the 2 groups (5,585 vs 5,772 h.μg/L), but patients with more variable absorption profiles were at higher risk for rejection (rejection rate of 19% vs 41% in groups A and B, respectively; p < 0.05).

In conclusion, the vast majority of these studies may have been of sufficient size to study the correlation between C₀ or C₂ and AUC, but underpowered for studying the relationship between drug concentrations and clinical outcome. Also, we did not find any study designed to evaluate the relationship between exposure measured by the AUC and efficacy (acute graft rejection, graft loss) or cyclosporine toxicity. Despite more than 20 years of use, no prospective trial compared cyclosporine dose adjustment based on full AUC and C₀ or C₂ in any type of allograft transplantation, let alone thoracic transplantation.^[53]

Available data has to be expanded to propose recommendations for optimal cyclosporine TDM in thoracic transplant recipients, including the identification and validation of the best exposure index and the definition of optimized target alues. Further investigations in special populations such as CF patients with gastroparesis are also awaited.

1.2. Tacrolimus

Tacrolimus (FK506, Prograf^®) is a 10–100 times more potent immunosuppressant than cyclosporine in vitro.^[8,9,94] Despite a chemical structure different from that of cyclosporine, tacrolimus suppresses the immune system by similar mechanisms: it inhibits the synthesis of IL-2 through the binding of a specific immunophilin, FK506 binding protein (FKBP12) and inhibition of calcineurin by this complex.^[9]

Tacrolimus was proposed as an alternative to cyclosporine in combination with AZA or MMF after thoracic transplantation, leading to lower rates of acute rejection, similar infection rates and slightly higher incidence of new onset diabetes mellitus compared to cyclosporine-based therapy.^[95–97] There has been a clear trend over the past decade to use tacrolimus instead of cyclosporine as part of maintenance immunosuppressive regimens after thoracic transplantation.^[3] Tacrolimus is now the main calcineurin inhibitor given to adult and paediatric thoracic transplant recipients (approx. 60% of the patients).^[2,3,42]

Some guidelines have been proposed for the use of tacrolimus after thoracic transplantation, with a recommended initial oral tacrolimus dose of 0.05 to 0.15 mg/kg/day in heart transplant patients.^[95] and of 0.1 to 0.15 mg/day in lung transplant recipients.^[98] Administration of tacrolimus sublingually or via a naso-gastric tube may be useful in some specific situations, such as in patients with severe gastro-intestinal troubles. In the early post-transplantation period, IV tacrolimus may be used when necessary^[99] (for instance, during the very first days post-transplantation, or to treat resistant rejection), via continuous infusion over 24 hours, with a dose ranging between 0.015 and 0.10 mg/kg/day. Nevertheless, there are some concerns regarding neurotoxicity and nephrotoxicity when it is administered by this route.^[9]

1.2.1. Pharmacokinetics

The PK of tacrolimus has been previously described in healthy volunteers and in diverse solid organ transplant settings (kidney, liver).^[100] As with cyclosporine, it is characterized by a large inter-individual and intra-individual variability.^[101,102] However, PK and variability data on tacrolimus used after thoracic transplantation is relatively scarce and inconsistent (Tables IX & X). In many published studies, exposure indices and PK parameters were considered after the first dose.^{[100,103–105]} Except for two studies^[100,103] in which tacrolimus was assayed by HPLC-MS/MS, the PK studies we identified used immunoassays (mostly MEIA) to measure tacrolimus concentrations. Such assays (also including ELISA, EMIT, ACMIA, CEDIA and CMIA technologies) all lack specificity because of cross-reactivity of the assay antibody with metabolites, hence overestimating the actual blood levels (as assessed using the reference technique HPLC-MS/MS). Moreover, immunoassays may interact with haemoglobin: some authors observed a negative correlation between haematocrit and tacrolimus concentrations determined by MEIA.^[106] This observation was confirmed in subsequent publications, which reported that both low haematocrit values and low albumin concentrations are likely to show artificially high concentrations of tacrolimus when measured by MEIA.^[107,108] Finally, the majority of the PK studies performed in thoracic transplantation were descriptive in nature and used model-independent PK methods, where mean and standard deviation of exposure indices and PK parameters in the evaluated population were calculated from individual data.^{[100,102,103,105,109–111]} One used the iterative two-stage (ITS) method, for which a specific PK model was set up in order to determine individual exposure indices and PK parameters and thus to calculate mean PK parameters in the population.^[112] We could not find any popPK study of tacrolimus in thoracic transplantation.

Table IX. Tacrolimus exposure indices in cohort studies.

Results are expressed as mean±SD (range), unless otherwise specified

Ref	Number of patients	Post-transplantperiod	Tacrolimus dose	Co-administered Immunosup-pressant	Assay	Measured AUC	C0 (μg/L)	Cmax (μg/L)	tmax (h)	Others
HEART TRANSPLANTATION
[102]	14 adults 55.5 years (23–61)	< 6 months	First D = 0.03–0.40 mg/kg (median: 0.052 mg/kg)	ATG AZA Prednisone	MEIA – IMx	Non-compartmental PK analysis using Pharm-NCA computer program; AUC Trap; median (range) PK profiles after the first dose of tacrolimus
						AUC0-12 = 155.6 (103.5–728.7) AUC0-∞ = 285.3 (121.8–1540.0)	--	23.0 (10.7–84.4)	2 (1–4)	Cmin = 8.3 (< 5–54.1)
						Best correlation: C12-AUC: C12 = 0.0371 x AUC0-∞ – 2.1; r2 = 0.98 (p<0.001)
[100]	11 adults OHT		First D up to 0.15 mg/kg BID D adjusted on C0 < 15 μg/L	N/R		Non-compartmental PK analysis using TOPFIT v2.0; tmax: median (range)
		Profile 1 (day 1) NG tube	D = 0.053±0.031 mg/kg (0.026–0.143) CV = 59%		ELISA	AUC0-∞ = 191.8±137.1 (32.9–422.2)	--	23.6±22.4 (1.3–76.6)	2.84 (2–9.5)	--
						CV = 72% – n = 7		CV = 95% – n = 10	n = 10
					HPLC- MS/MS	AUC0-∞ = 141.9±93.6 (41.6–254.9)	--	20.8±15.9 (2.1–49.4)	2.75 (1.5–9.5)	--
						CV = 66% – n = 5		CV = 77% – n = 10	n = 10
		Profile 2 (last day) Hard capsules	D = 0.076±0.069 mg/kg (0.013–0.250) CV = 91%		ELISA	AUC0-∞ = 257.4±76.3 (144.8–442.7)	--	34.2±18.2 (19.0–84.7)	1.96 (0–4)	--
						CV = 30% – n = 10		CV = 53% – n = 10	n = 10
					HPLC- MS/MS	AUC0-∞ = 211.1±77.2 (104.2–362.9)	--	33.1±14.8 (13.3–66.6)	1.5 (1–8)	--
						CV = 37% – n = 10		CV = 45% – n = 10	n = 10
[103]	25 adults 47±10 years	Immediate post-transplantation period		Induction: ATG/OKT3	HPLC-MS/MS	Non-compartmental analysis using computer program PC Modfit v/6; AUC0-12 Trap
	n: N/R		D = 0.075 mg/kg/day
			First dose			78.1±59.0	--	14.7±7.8	2.1±1.7	--
			Steady-state			186±68.7		26.6±9.5	1.9±1.2
	n: N/R		D = 0.15 mg/kg/day
			First dose			142±116	--	24.5±13.7	1.9±1.2	--
			Steady-state			198±73.1		29.9±14.5	1.8±0.6
						r2 Cmin – AUC0-12: first D = 0.86; steady-state = 0.79
[115]	19 adults 46 years (28–48) OHT		D = 0.06 mg/kg/day (0.04–0.08) D adjusted on C0 = 5–20 μg/L	ATG AZA Prednisone	MEIA – IMx	2-compartment open model using P-Pharm popPK software
		PK1 (day 10)				--	12.5	--	--	--
		PK2 (month 2)				--	15.0	--	--	--
[111]	8 adults 45–60 years	4 to 16 months (mean: 10.6)	D = 0.02–0.1 mg/kg/day D adjusted on C0 = 5–12 μg/L	MMF Prednisone	MEIA II– IMx	Non-compartmental PK analysis using PK Calc software; AUC Trap; Mean (range)
						AUC0-12 = 151.7 (129.2–174.2) AUC0-∞ = 317.5 (189.2–445.7)	7.9 (6.0–9.7)	22.9 (18.3–27.5)	2.2 (1.5–2.9)	C12 = 7.6 (6.2–9.0) (vs C0: ns)
						r2 AUC0-12: C0, C1, C2, C3 = 0.09 to 0.26 (ns) C4, C5, C6, C7, C8 = 0.53 to 0.64 (p < 0.05) C9, C10, C11, C12 = 0.70 to 0.80 (p < 0.01)
[109]	22 adults 55±8 years (36–64) OHT	< Year 1	D = 0.3 mg/kg/day D adjusted on C0 = 10–20 μg/L Steady-state	MMF or AZA Steroids	MEIA – IMx	Non-compartmental PK analysis using MOMENT software (n = 25 PK profiles); AUC Trap
						AUC0-12 = 236.3±88.6 (81.5–419.8) AUC0-4 = 93.5±40.8 (35.9–180.9)	14.8±5.7 (3.9–25.9)	30.5±13.8 (10.3–57.8)	2.3±1.5 (0.5–6.0)	Cav = 19.7±7.4 (6.8–35.0)
						r2 AUC0-12: C0 = 0.80 C1 = 0.40 C2 = 0.81 C4 = 0.90 C12 = 0.89	r2 AUC0-4: C0, C1 = 0.70 C2 = 0.96 C4 = 0.73	Sparse sampling algorithms for AUC0-12: 3.75 + 5.52*C0 + 6.97*C4 – r2 = 0.95 0.98 + 4.17*C0 + 2.29*C2 + 5.3*C4 – r2 = 0.97
[104]	13 patients		D adjusted on C0 = 10–15 μg/L	ATG	N/R	Non-compartmental PK analysis performed using WinNonlin; AUC Trap
						AUC/D: (μg.h/L)/mg tacrolimus
		First dose	D = 0.077±0.011 mg/kg BID			AUC0-∞ = 186.7±96.6 CV = 52% AUC0-∞/D = 46.0±24.3 r2 AUC0-∞: C0.5 = 0.020; p = 0.66 C1 = 0.143; p = 0.226 C2, C3, C4, C6, C8, C10 = 0.764 to 0.88; p < 0.05 C12 = 0.895; p < 0.05	--	--	--	--
		First month Steady-state	D = 0.068±0.047 mg/kg BID			AUC0-12 = 202.6±61.3 CV = 30% AUC0-12/D = 69.0±43.9 r2 AUC0-12: C0 = 0.841; p < 0.05 C1 = 0.469; p = 0.014 C0.5, C2, C3, C10, C12 = 0.640 to 0.822; p < 0.05 C4, C6, C8 = 0.902 to 0.976; p < 0.05	--	--	--	--
[105]	23 patients 38 years (18–56)		D = 0.10 mg/kg/day D adjusted on C0 = 10–20 μg/L	N/R	MEIA – IMx	Non-compartmental PK analysis
		First dose				AUC0-12 = 116.3±58.9 Best r2 C4-AUC = 0.93	--	18.8±7.7	2.1±1.4	C12 = 5.8±2.7
		Day 3				AUC0-12 = 149.7±63.2 Best r2 C4-AUC = 0.92	--	23.7±10.4	2.2±1.3	C12 = 10.6±3.8
		Day 7				AUC0-12 = 216.6±89. 4 Best r2 C4-AUC = 0.95	--	32.4±11.5	1.9±0.8	C12 = 17.7±6.4
LUNG TRANSPLANTATION
[117]	20 patients		D adjusted on C0 = 7–12 μg/L	Patients converted from cyclospsorine to tacrolimus (rescue therapy)	N/R	Non-compartmental PK analysis
	8 CF	326 days (120–689)	7.9±2.8 mg/day (0.16±0.06 mg/kg/day)			--	10.3±2.6	--	--	--
	12 non-CF	436 days (155–714)	6.4±3.7 mg/day (p = 0.14) 0.11±0.06 mg/kg/day (p = 0.08)			--	10.7±1.6 (ns)	--	--	--
[114]	16 adults (5 CF) 45.5±3.4 years	98±19 wks	D = 3.6±2.5 mg BID (0.25–7.0) D adjusted on C0 = 5–15 μg/L	N/R	HPLC- MS/MS	Non-compartmental PK analysis (n = 22 profiles)
						AUC0-8 = 115.8±45.2 (41.0–212.3)	10.5±5.7 (3.4–30.4)	21.6±11.8 (5.9–50.0)	2.3±1.3 (0.0–5.0)	C2 = 19.1±11.7 (4.6–50.0)
						AUC0-4 = 67.9±31.1 (20.7–146.9)
						r2 C2-AUC0-8 = 0.50; r2 C4-AUC0-8 = 0.53
HEART AND LUNG TRANSPLANTATION
[110]	22 adults		D adjusted on C0 = 8–12 μg/L Steady-state	Low-dose corticoids	MEIA – IMx	Non-compartmental PK analysis – 3 full PK profiles/patient within 1 week; AUC trap; Median (range)
						AUC/D: (h.μg/L)/mg tacrolimus	C0/D: (μg/L)/mg tacrolimus	Cmax/D: (μg/L)/mg tacrolimus
	11 CF 30 years (21–43)	36.1 months	D = 12 mg/day (5–25)	AZA (n = 5) or MMF (n =2)		AUC0-12	C0	Cmax	tmax	C3 V1: 16.2 (11.4–30.5) V2: 21.6 (10.1–33.0) V3: 16.4 (6.5–26.0)
						V1: 170 (140–260)	V1: 8.2 (6.0–14.0)	V1: 23.4 (16.6–64.8)	V1: 1.5 (0.5–6.0)
						V2: 180 (110–310)	V2: 9.2 (4.7–13.2)	V2: 25.6 (15.6–65.7)	V2: 1.5 (1.0–8.0)
						V3: 160 (100–270)	V3: 8.5 (5.2–13.2)	V3: 23.4 (10.6–76.8)	V3: 1.5 (0.5–5.0)
						AUC0-12/D	C0/D	Cmax/D
						V1: 14 (8–34)	V1: 0.7 (0.5–2.3)	V1: 2.6 (1.2–4.7)
						V2: 17 (8–33)	V2: 0.7 (0.4–1.7)	V2: 3.2 (1.1–3.9)
						V3: 12 (9–32)	V3: 0.7 (0.3–1.7)	V3: 2.2 (1.2–3.4)
						AUC0-4				C3/D V1: 1.5 (0.7–3.3) V2: 1.7 (0.6–3.6) V3: 1.4 (0.9–3.3)
						V1: 80 (60–150)
						V2: 100 (50–170)
						V3: 70 (40–180)
						AUC0-4/D
						V1: 8 (4–6)
						V2: 10 (3–13)
						V3: 7 (4–14)
						AUC/D smaller by about 50% in CF patients: AUC0-12/D, p < 0.004; AUC0-4/D, p < 0.001
						r2 AUC0-12: C0.5 = 0.35 C0, C1, C1.5 = 0.53 to 0.58 C2, C2.5, C4 = 0.71 to 0.80 C3 = 0.86		r2 AUC0-4: C0 = 0.34 C0.5, C4 = 0.46–0.53 C1, C1.5, C3 = 0.79 to 0.83 C2, C2.5 = 0.89–0.90
	11 non-CF 51 years (31–56) p < 0.01	32.6 months p = 0.97	D = 6.1 mg/day (3–10) p = 0.008	AZA (n = 5)		AUC0-12	C0	Cmax:	tmax	C3 V1: 19.4 (12.5–25.0) V2: 16.3 (9.4–26.9) V3: 16.9 (11.0–26.4)
						V1: 210 (120–270)	V1: 9.9 (6.4–14)	V1: 32.8 (15.5–57.8)	V1: 1.5 (1.0–4.0)
						V2: 180 (80–270)	V2: 10.9 (5.0–17)	V2: 29.1 (11.1–68.0)	V2: 1.0 (0.5–5.0)
						V3: 170 (80–240)	V3: 9.8 (5.4–13.7)	V3: 30.6 (11.6–43.5)	V3: 1.5 (1.0–3.0)
						AUC0-12/D	C0/D	Cmax/D:
						V1: 31 (17–71)	V1: 1.8 (1.0–2.4)	V1: 5.5 (2.4–15.2)
						V2: 33 (12–72)	V2: 1.9 (0.7–3.1)	V2: 6.5 (1.4–14.1)
						V3: 30 (12.69)	V3: 2.0 (0.8–3.4)	V3: 5.6 (1.6–4.5)
						AUC0-4				C3/D V1: 3.5 (1.9–8.3) V2: 3.6 (1.2–7.2) V3: 2.8 (1.6–6.3)
						V1: 110 (60–140)
						V2: 80 (40–150)
						V3: 90 (40–130)
						AUC0-4/D
						V1: 15 (6–43)
						V2: 15 (6–42)
						V3: 16 (6–39)
						r2 AUC0-12: C0, C0.5 = 0.44–0.58 C1, C1.5, C2 = 0.72 to 0.76 C2.5, C4 = 0.84–0.87 C3 = 0.92		r2 AUC0-4: C0, C0.5 = 0.44–0.48 C1, C1.5, C2, C2.5, C3, C4 = 0.83 to 0.88
[112]	22 adults				MEIA- IMx	PK analysis using a one-compartment model with first-order elimination convoluted with a double gamma absorption phase – PK software Ciclo®2.3 3 full PK profiles/patient within 5 days (V1, V2, V3)
						AUC/D: (h.μg/L)/mg tacrolimus	C0/D: (μg/L)/mg tacrolimus	Cmac/D: (μg/L)/mg tacrolimus
	11 CF 30 years (21–43)	3–116 months	D = 0.23 mg/kg/day (0.09–0.47)	AZA (n = 5) or MMF (n =2)		AUC0-12	C0	Cmax	tmax	--
						V1: 182.3±48.2	V1: 9.05±2.29	V1: 31.9±16.9	V1: 1.5±0.6
						V2: 187.4±66.0	V2: 8.91±3.20	V2: 32.4±18.4	V2 : 2.1±1.8
						V3: 169.7±60.5	V3: 8.52±3.10	V3: 26.2±17.7	V3: 1.7±1.2
						AUC0-12/D	C0/D	Cmax/D
						V1: 18.8±9.1	V1 : 0.98±0.61	V1: 3.0±1.3
						V2: 18.5±8.2	V2: 0.89±0.45	V2: 2.9±1.2
						V3: 16.5±6.3	V3: 8.52±3.10	V3: 2.2±0.7
						AUC0-4
						V1: 81.3±39.0
						V2: 80.2±39.3
						V3: 71.0±35.0
						AUC0-4/dose
						V1: 8.1±3.9
						V2: 7.6±3.5
						V3: 6.3±2.7
						Bayesian estimator (AUC0-12): best sampling times: C0, C1.5, C4; r2 = 0.91, mean bias −1.1% (−18.6;16.7%) BE vs Trap on AUC0-12, AUC0-4, Cmax, tmax, C0: bias% < 7.3%; RMSE% < 15%
	11 non-CF 51 years (31–56) p < 0.01	9–86 months	D = 0.1 mg/kg/day (0.06–0.19)	AZA (n = 6)		AUC0-12	C0	Cmax	tmax	--
						V1: 184.0±55.1	V1: 9.62±2.54	V1: 33.1±15.0	V1: 1.4±0.7
						V2: 180.2±59.6	V2: 9.87±2.85	V2: 31.1±17.4	V2: 1.6±1.4
						V3: 183.4±47.0	V3: 9.98±2.49	V3: 30.1±10.8	V3: 1.5±0.8
						AUC0-12/D	C0/D	Cmax/Dose :
						V1: 33.5±14.7	V1: 1.71±0.49	V1: 5.9±3.4
						V2: 33.9±16.7	V2: 1.82±0.72	V2: 6.0±3.8
						V3: 35.8±18.3	V3: 1.90±0.81	V3: 5.7±3.5
						AUC0-4
						V1: 83.7±28.0
						V2: 80.3±33.7
						V3: 80.1±25.9
						AUC0-4/D
						V1: 15.3±7.8
						V2: 15.1±8.5
						V3: 15.9±9.3
						Bayesian estimator (AUC0-12): best sampling times: C0, C1, C3; r2 = 0.96, mean bias −3.6% (−17.4;6.8%) BE vs Trap on AUC0-12, AUC0-4, Cmax, tmax, C0: bias% < 4.6; RMSE% < 15.6

A_IV: intravenous coefficient – ATG: Anti-thymocyte globulin – AZA: Azathioprine – BE: Bayesian estimator – D: Dose – CF: Cystic fibrosis – MMF: Mycophenolate mofetil – NG: Nasogastric – N/R: not reported – Trap: Trapezes.

Table X. Tacrolimus pharmacokinetic parameters in cohort studies.

Results are expressed as mean±SD (range), unless otherwise specified

Ref	Number of patients	Post-transplant period	Tacrolimus dose	Co-administered immunosuppressant	Assay	F, transfer k	Vd/F (unless otherwise specified)	CL/F (unless otherwise specified)	T1/2 (h)
HEART TRANSPLANTATION
[102]	14 adults 55.5 years (23–61)	< 6 months	First D = 0.03–0.40 mg/kg (median: 0.052 mg/kg)	ATG AZA Prednisone	MEIA – IMx	PK parameters after the first dose of tacrolimus
						ka, kel: h−1 Lag-time: h	Vd/F: L/kg	CL/F: L.h−1/kg
						Non-compartmental PK analysis using Pharm-NCA computer program
						--	2.4±0.79	0.21±0.08	8.7±3.5 kel = .091±0.040
						One-compartment open model with first-order elimination using P-Pharm computer program
						ka = 1.26±0.61	2.0±0.4 (1.3–3.9)	0.23±0.08 (0.098–0.328)	6.5±2.4
						Lag-time = 0.22±0.13	CV = 33%	CV = 36.3%	kel = 0.116±0.034
[100]	11 adults OHT		First D up to 0.15 mg/kg BID D adjusted on C0 < 15 μg/L	N/R		Non-compartmental PK analysis using TOPFIT v2.0 software
		Profile 1 (day 1)	D = 0.053±0.031 mg/kg (0.026–0.143)		ELISA	--	--	--	7.1±4.3 (2.4–15.9)
		NG tube	CV = 59%						CV = 60% – n = 7
					HPLC-MS/MS				4.0±1.0 (2.8–5.3) CV = 24% – n = 5
		Profile 2 (last day)	D = 0.076±0.069 mg/kg (0.013–0.250)		ELISA	--	--	--	14.1±6.3 (7.1–19.5)
		Hard capsules	CV = 91%						CV = 32% – n = 10
					HPLC-MS/MS	--	--	--	10.7±5.3 (1.8–18.5) CV = 50% – n = 10
[103]		Immediate post-transplantation period		N/R	HPLC-MS/MS	Non-compartmental PK analysis using PC Modfit v/6 software CL: L/h
	10 adults 57±9 years		IV, continuous infusion D = 0.01 to 0.02 mg/kg/day	Induction: N/R		--	--	CL = 2.4	--
	25 adults 47±10 years		Oral	Induction: ATG/OKT3					After the first dose:
	n: N/R		D = 0.075 mg/kg/day			--	--	--	10.9±7.0
			D = 0.15 mg/kg/day			--	--	--	9.8±5.2
[115]	19 adults 46 years (28–48) OHT		D = 0.06 mg/kg/day (0.04–0.08) D adjusted on C0 = 5-20μg/L	ATG AZA Prednisone	MEIA –IMx	2-compartment open model – PK analysis using P-Pharm popPK software
						ka: h−1 Lag-time: h	V1/F: L/kg	CL/F: L.h−1/kg
		PK1 (day 10)				ka = 1.1±0.5 (0.3–1.6)	V1/F = 1.7±0.6 (0.7–3.5)	0.19±0.08 (0.09–0.38)	--
						CV = 42%	CV = 36%	CV = 42%
						Lag-time = 0.22±0.09 (0.1–0.45) CV = 41%
		PK2 (month 2)				ka = 1.0±0.5 (0.1–1.8)	V1/F = 1.1±0.4 (0.3–1.8) – p < 0.01	0.23±0.15 (0.04–0.62)	--
						CV = 48%	CV = 34%	CV = 65%
						Lag-time = 0.44±0.05 (0.29–0.50) – p ≤ 0.05 CV = 11%
[111]	8 adults 45–60 years	4 to 16 months (mean: 10.6)	D = 0.02–0.1 mg/kg/day D adjusted on C0 = 5–12 μg/L	MMF Prednisone	MEIA II – IMx	Non-compartmental PK analysis using PK Calc software
						--	--	--	Mean = 13.8 (95% CI = 7.0–20.7)
[109]	22 adults 55±8 years (36–64) OHT	First year	D = 0.3 mg/kg/day D adjusted on C0 = 10–20μg/L Steady-state	MMF or AZA Steroids	MEIA –IMx	Non-compartmental PK analysis using MOMENT software (n = 25 PK profiles) CL/F: L/h
						--	--	11.6±5.5 (3.8–23.5)	--
[105]	23 adults 38 years (18–56)		D = 0.10 mg/kgd D adjusted on C0 = 10–20 μg/L	N/R	MEIA –IMx	Non-compartmental PK analysis
		First dose				--	--	--	11.1±6.3
		Day 3				--	--	--	10.6±5.8
		Day 7				--	--	--	9.8±5.5
LUNG TRANSPLANTATION
[114]	16 adults (5 CF) 45.5±3.4 years	98±19 wks	D = 3.6±2.5 mg BID (0.25–7.0) D adjusted on C0 = 5–15 μg/L	N/R	HPLC-MS/MS	Non-compartmental PK analysis (n = 22 profiles)
						--	--	--	7.9±2.3 (4.0–12.0)
HEART-LUNG TRANSPLANTATION
[112]	22 adults				MEIA –IMx	One-compartment model with first-order elimination convoluted with a double gamma absorption phase – PK analysis using Ciclo®2.3 software
						MAT, SDAT: h AIV: L−1	V1/F: L	CL/F: L/h; λ: h−1
	11 CF 30 years (21–43)	3–116 months	D = 0.23 mg/kg/day (0.09–0.47)			MAT1 = 1.10±0.68	V1/F = 2,011±1,740	68.22±29.80	λ = 0.64±0.33
						SDAT1 = 0.27±0.15
						MAT2 = 5.14±2.11
						SDAT2 = 1.96±0.68
						AIV = 4.03±1.67
						r = 0.58±0.22
	11 non-CF 51 years (31–56) p < 0.01	9–86 months	D = 0.1 mg/kg/day (0.06–0.19)	AZA (n = 6)		MAT1 = 0.92±0.43 (ns)	V1/F = 444±326 (p < 0.01)	36.49±18.98 (p < 0.05)	λ = 0.80±0.31 (p < 0.01)
						SDAT1 = 0.30±0.10 (ns)
						MAT2 = 5.47±2.30 (ns)
						SDAT2 = 1.89±0.71 (ns)
						AIV = 9.36±4.98 (p < 0.01)
						r = 0.62±0.12 (ns)

ATG: Antithymocyte globulin – AZA: Azathioprine – CF: Cystic fibrosis – D: Dose – NG: Nasogastric – N/R: not reported – OHT: orthotopic heart transplantation – r: fraction of the dose absorbed by the faster phase.

The intra- and inter-individual variability of tacrolimus PK is related to factors such as erratic bioavailability, time post-transplantation, concomitant pathologies and drug-drug interactions, which unfortunately were rarely explored.

1.2.1.1. Absorption

Generally, the bioavailability of tacrolimus after oral administration is poor, averaging 25%, but is also characterized by a large inter-individual and intra-individual variability, ranging from 5% to 93% depending on the reports.^{[8,52,94,100,101]} This low and variable bioavailability is caused by an extensive presystemic metabolism by CYP3A isoenzymes and by active transport by the trans-membrane efflux pump P-gp, all localized in the gut wall in addition to the liver.^[8,101] Poor aqueous solubility and alterations in gut motility (e.g. in CF) may also be responsible for poor and erratic drug uptake.^[101] Tacrolimus bioavailability can be reduced when the molecule is administered with low-fat food.^[8,101] On the contrary, some authors reported an increased bioavailability in patients with persisting diarrhea, resulting in an increase in tacrolimus trough levels.^[113] However, in contrast to cyclosporine, tacrolimus is absorbed in a completely bile-independent manner.^[8,94,101]

In most subjects, tacrolimus is rapidly absorbed following oral administration, t_max occuring usually around 2 hours after single administration^{[102,103,105]} and between 0.5 and 1.5 hours at steady-state (Table IX).^{[52,100,101,103,111]} However, drug uptake can also last longer, resulting in a flat absorption profile, an extended lag-time or secondary peaks, with maximum blood concentrations occurring up to 8 hours after dosing.^{[8,94,100,101,105,109,110,112]}

With oral dosing, steady-state blood concentrations are usually achieved within approximately 48 to 72 hours.^[95] In heart transplant patients, C_max was characterized by a coefficient of variation ranging between 41 and 56% after single oral administration,^[103,105] and between 36 and 49% at steady-state^{[100,103,105,109]} and between 53 and 68% in lung and heart-lung transplant patients.^[112,114]

The rate of absorption of tacrolimus has been reported to be highly variable.^[101] In heart transplant patients, absorption was characterized by a rate constant of absorption k_a ranging from 0.13 to 1.76 h⁻¹ (CV > 40%) and a lag-time comprised between 0.1 and 0.5 hours.^[115] In stable lung transplant recipients (with or without CF), tacrolimus absorption was best fitted using two Gamma distributions (double, sigmoid absorption profile with two different velocities), describing an early peak and a possible secondary peak of absorption.^[112] This resulted in the calculation of two mean absorption times (MAT₁ and MAT₂), with similar values in CF and non-CF patients (MAT₁ = 1.10±0.68 h and 0.92±0.43 h; MAT₂ = 5.14±2.11 h and 5.47±2.30 h, respectively).

1.2.1.2. Distribution

Tacrolimus is highly lipophilic and binds extensively to erythrocytes, with higher blood than plasma concentrations. In plasma, it binds strongly to proteins (99%), mainly α₁-acid glycoprotein, lipoproteins, globulins and albumin.^[101] After IV administration, tacrolimus concentration undergoes a rapid initial decline due to distribution, which slows down over the next 24 hours after reaching distribution equilibrium.^[94] To our knowledge, tacrolimus V_d/F in thoracic transplant recipients was reported in only three studies, in two of which tacrolimus PK was fitted with a one-compartment model. In the first one,^[112] V_d/F was determined at steady-state in 22 heart-lung transplant patients, with a significant difference between patients with and those without CF (2,011±1,740 L vs 444±326 L, respectively; p < 0.01), whereas in the second one^[102] V_d/F ranged between 1.3 and 3.9 L/kg after single dose administration in 14 heart transplant recipients. In the third study,^[115] performed in 19 heart transplant recipients, the authors used a 2-compartment model and reported a mean V₁/F = 1.69±0.61 L/kg and 1.13±0.38 L/kg at ten days and two months after transplantation, respectively (p < 0.01). These rather inconsistent results might be partly explained by the different study conditions, the obviously large inter-patient variability (in the 3 studies, V_d/F coefficients of variation ranged from 33 to 86%) as well as by the generally suboptimal estimation of V_d/F by classical PK models, which usually estimate clearance more accurately than volume of distribution.

1.2.1.3. Metabolism and elimination

Like cyclosporine, tacrolimus is extensively metabolized (mainly by CYP3A4 and CYP3A5 isoenzymes in the liver and intestinal wall). More than 95% of tacrolimus metabolites are excreted in the bile, and less than 0.5% of tacrolimus dose appears unchanged in the urine or faeces.^[52,94,101] As a consequence, its potential for drug-drug interactions seems similar to that of cyclosporine. One exception is that tacrolimus does not seem to interact with MMF PK like cyclosporine does, leading to greater MPA exposure when tacrolimus and MMF are used in combination.^[52]

A few studies reported tacrolimus clearance values in thoracic transplant patients. In heart transplant patients, tacrolimus clearance after continuous IV infusion was 2.4 L/h.^[103] After tacrolimus oral administration, CL/F was 0.23±0.08 L.h⁻¹/kg after a single dose,^[102] 0.19±0.08 L.h⁻¹/kg ten days and 0.23±0.15 L.h⁻¹/kg two months after transplantation,^[115] with high coefficients of variation (42–65%). In a cohort of 12 lung and heart-lung transplant patients, CL/F = was 68.2±29.8 L/h in CF patients and 36.49±18.98 L/h in non-CF patients (p < 0.05),^[112] which is higher than the above mentioned values (considering a mean bodyweight of 70 kg) and than those reported by Aumente-Rubio et al^[109] in heart transplant patients (11.6±5.5 L/h); this may be explained by the difference in patients (heart-lung transplant vs heart recipients) and by the difference in the PK approach (one-compartment PK model vs non-compartmental PK analysis).

T_1/2 was determined mainly using non-compartmental PK analysis and was also found to be highly variable (CV = 30–64%).^{[100,102,103,105,114]} When tacrolimus was assayed using a HPLC-MS/MS method, T_1/2 at steady-state was 10.7±5.3 h in heart transplant patients^[100] and 7.9±2.3 h in lung transplant recipients.^[114] The elimination half-life determined from tacrolimus concentrations measured by immunoassays may be overestimated due to the lack of specificity and the cross-reactions reported between the assay antibody and tacrolimus metabolites. Indeed, in a cohort of 8 heart transplant patients, T_1/2 reached 33.3 h (mean: 13.8 h, 95%CI = 7.0–20.7 h). ^[111]

1.2.1.4. Drug-drug interactions

We did not find any formal PK studies investigating tacrolimus drug interactions in thoracic transplant recipients. Shitrit et al reported an interaction between itraconazole and tacrolimus in 40 patients with lung transplantation followed for more than one year.^[116] All patients were on tacrolimus and received prophylactic itraconazole for the first 6 months post-transplantation. In those patients, the mean dose of tacrolimus required rose by 76% after itraconazole withdrawal (3.26±2.1 mg/day during the first 6 months compared to 5.74±2.9 mg/day during the following 6 months, p < 10⁻⁴).

In other solid organs transplantation, various lists of drugs that may increase or decrease whole blood concentrations have been compiled.^[101] Since tacrolimus is a substrate for CYP3A and P-gp like cyclosporine, the expected drug interactions are essentially the same^[95] and any drug known to induce or inhibit these systems may increase or decrease its blood concentrations (erythromycin, calcium-channel blockers, imidazole anti-fungal agents).

1.2.1.5. Special populations

The PK of tacrolimus changes in time after transplantation and is affected by many factors, including liver function, patient age (pediatric transplant recipients require 2- to 4-fold higher doses of tacrolimus than adults to maintain similar C₀^[101]), time of administration (diurnal variations, co-administration of food), associated drugs, and modifications in the GI tract motility and secretions observed in CF patients.^[101,109]

Lower bioavailability of tacrolimus has been reported in CF patients, probably partly because of fat malabsorption due to pancreatic insufficiency.^[110,112]

Walker et al reported that heart-lung and double lung transplant recipients with CF required a 39% higher dose of tacrolimus than patients without CF in order to reach similar C₀.^[117] In a study performed in 22 heart-lung transplant patients (11 CF), tacrolimus CL/F after oral administration was higher in CF patients (68.22±29.80 vs 36.49±18.98 L/h, p < 0.05),^[112] who required significantly higher doses to reach comparable exposure. In those patients, AUC_0-12/dose and AUC_0-4/dose were approx. half of those of non-CF patients (p < 0.004 and p < 0.001, respectively).^[110] Comparatively to non-CF patients, exposure indices in CF patients (C₀, C₃, C_max, AUC_0-12, AUC_0-4) were characterized by higher inter-individual variability (CV% 24 to 46% vs 19 to 30%) and comparable intra-individual variability (CV% 16 to 40% vs 17 to 37%).^[110]

1.2.2. Therapeutic drug monitoring

Issues relating to the TDM of tacrolimus have been addressed by numerous authors and recently reviewed by Staatz & Tett.^[101] Like cyclosporine, tacrolimus is characterized by a narrow therapeutic index and significant systemic side effects.^[8,52,115] As stressed above, tacrolimus absorption is erratic and incomplete with a variable metabolism mediated by intestinal and hepatic CYP3A isoenzymes, resulting in high intrapatient and interpatient PK variability. As a result, the correlation between dose and blood concentrations is poor,^[52,115] confirming that the administered dose cannot reliably predict systemic exposure to tacrolimus, rendering it difficult to propose a suitable a priori dosing regimen for heart and lung transplant patients.^[100] This was the rationale for monitoring tacrolimus blood concentrations to optimize immunosuppression.^[52,94] However, the poor sensitivity and specificity of the first commercial assays used was an obstacle in the exploration of tacrolimus PK for optimizing TDM.^[52]

1.2.2.1. Single concentrations monitoring

Most results of clinical studies in kidney and liver transplantation suggest that C₀ levels of tacrolimus are, to some extent, predictive of efficacy and toxicity.^[52] Accordingly, tacrolimus dose after thoracic transplantation is mainly adapted to C₀ levels. The following target C₀ were proposed: in lung transplantation, 10–25 μg/L for the first 2 weeks, 10–20 μg/L for the next 6 to 10 weeks, and 10–15 μg/L thereafter;^[98] after heart transplantation, 15–20 μg/L for the first 2 months, 10–15 μg/L from the third to the sixth month, and 8–10 μg/L after 6 months.^[95] However, the authors did not mention on which assay these target levels were based, making it difficult to correctly interpret them, and to implement them in routine clinical practice.

Studies of the correlation between tacrolimus C₀ and AUC_0-12 have reported a wide range of results, from r² = 0.20^[111] to around 0.80 (Table IX).^{[103,104,109]} Some studies have shown that the correlation between blood concentration and AUC can be improved by using different time points: a good correlation was found between AUC_0-12 and C₃^[110] or C₄.^{[104,105,109]}

In other transplant types (mainly kidney), the reliance on C₀ monitoring has been questioned due to the observations that stable function, rejection or toxicity may be associated with the same range of tacrolimus concentrations.^[101,109] Unfortunately, no study in thoracic transplantation has been performed with the intent of exploring the relationship between single concentrations (C₀ or other concentrations) and efficacy or toxicity.

1.2.2.2. AUC monitoring

As with cyclosporine, tacrolimus AUC rather than single concentrations may be used for TDM, as it is a better marker of patient’s exposure to tacrolimus.

In studies in thoracic transplantation, tacrolimus AUC_0-12 was mainly calculated using the linear trapezoidal rule.^{[100,102–105,109–111]} As with cyclosporine, AUC_0-4 was considered as a surrogate exposure index and AUC_0-4 values reported in a few studies in heart^[109] and in heart-lung^[110,112] transplant recipients. However, AUC_0-4 still has to be validated as a reliable estimate of AUC_0-12 and as a reliable marker of clinical outcomes in heart or lung transplant patients.

As another way to avoid the numerous blood samples needed for the determination of AUC_0-12, some authors investigated sparse sampling strategies. In heart transplantation, Aumente Rubio et al proposed two algorithms, one with 3 sampling times (0, 2 hours and 4 hours, r² = 0.97 with trapezoidal AUC_0-12) and one with only 2 sampling times (0 and 2 hours, r² = 0.95).^[109] In lung transplantation, Morton et al failed to find an efficient sparse sampling strategy,^[114] but Saint-Marcoux et al developed and validated Bayesian estimators for the determination of tacrolimus AUC_0-12 in CF and non-CF patients, and proposed the following sampling strategy: 0, 1 hour and 3 hours for non-CF patients and 0, 1.5 hours and 4 hours for CF patients.^[112]

In a study of 25 primary heart transplant recipients, the authors found a significant difference in the AUC_0-12 after the first oral dose between patients who experienced acute rejection and those who did not (71 vs 168 h.μg/L, p = 0.012).^[103] However, this paper does not provide any detail about patient’s follow-up, so there is no information on when acute rejection occurred.

So far, no prospective studies have been conducted in thoracic transplantation to investigate the relationships between tacrolimus inter-dose AUC and clinical outcome. A fortiori, the use of AUC as a monitoring tool still remains to be validated in prospective controlled studies, and optimal AUC targets to be defined in heart and lung transplantation.

2. Mycophenolate, purine synthesis antagonist

Two compounds containing mycophenolate are currently available, mycophenolate mofetil (MMF, Cellcept^®) and mycophenolate sodium (EC-MPS, Myfortic^®). As mycophenolate sodium has been released later and its use in thoracic transplantation more rarely reported, this review will focus on MMF only.

Following oral administration, MMF, the 2,4-morpholino-ester of the active moiety mycophenolic acid (MPA), is rapidly hydrolyzed to MPA by esterases in the stomach, small intestine, blood, liver and tissues.^[118,119] MPA exerts its immunosuppressive effects by reversibly inhibiting inosine 5′-monophosphate deshydrogenase (IMPDH),^[120] a key enzyme in the de novo pathway of purine biosynthesis,^[121,122] essential for DNA replication when cells proliferate. Through IMPDH inhibition, MPA specifically blocks the proliferation and clonal expansion of T and B lymphocytes, providing effective immunosuppression in transplant patients.^[121–124]

The use of MMF has dramatically increased in the past decade, MMF-based regimens representing now approx. 80% and 50% of maintenance strategies in adult heart and lung transplant recipients, respectively, and 50% in pediatric lung transplant recipients.^[2,3,42,125]

MMF is available in oral (capsules, tablets and powder for oral suspension) and IV formulations.^[122] It is generally combined with a calcineurin inhibitor and corticosteroids for rejection prophylaxis after solid organ transplantation.^[124,126] Reduced rejection rates, improved graft and patient survival with MMF compared with AZA have lead it to progressively replace AZA as the antiproliferative agent of choice after thoracic transplantation, although it has not been approved by most of the national health agencies in lung transplantation.^{[2,5,127,128]} There has been a growing interest in MMF-based as part of calcineurin inhibitor- or steroid-sparing strategies, as MMF is free of toxicities adversely affecting long-term patient and graft survival (nephrotoxicity, hypertension, metabolic perturbations).^[120]

2.1. Pharmacokinetics

As opposed to other immunosuppressants, MMF dose in adults is not based on patient body size or weight. MMF is usually administered at a fixed oral dose of 0.75–1.5 g twice daily, depending on the associated calcineurin inhibitor.^{[119,120,124,129]} Oral suspension can be administered via a nasogastric tube, if necessary.^[118]

Little is known about the PK of MPA after thoracic transplantation.^[119,124] Most PK studies have been conducted in renal transplant patients, in whom MMF exhibits time-dependant PK (concentrations increase with time following transplantation over the first 1–3 months).^[124] However, data derived from renal transplant recipients cannot be readily extrapolated to thoracic organ transplant patients, as neither the heart nor the lungs are involved in the elimination of mycophenolate. Although blood pressure may have some impact on MPA glomerular filtration rate, cardiac and pulmonary functions do not impact so much MPA plasma concentrations.^[119,124] In a PK study of MMF in 50 stable thoracic (23 heart and 27 lung) transplant recipients (0.2–19.7 years post-transplantation), no significant variations in MPA exposure indices or PK parameters was observed with post-transplantation time (Tables XI & XII).^[130] On the contrary, an increase of MPA C₀ throughout the follow-up period of 40.0±20.5 months was reported (p = 0.0162) in 57 “stable” lung transplant recipients (all non-CF patients), but the time elapsed between transplantation and enrolment was not reported and the authors did not give any hypothesis on the cause of this increase.^[131] Consequently, currently available data on thoracic transplantation does not allow any reliable conclusion to be drawn on the evolution of MPA PK parameters with post-transplantation time, even in the early periods.

Table XI. Mycophenolate exposure indices in cohort studies.

Results are expressed as mean±SD unless otherwise specified

Ref	Number of patients	Post-transplant period	MMF dose	Co-administered immunosuppressant	Analytical method, molecule	Measured AUC (h.mg/L)	C0 (mg/L)	Cmax (mg/L)	tmax (h)	Others
HEART TRANSPLANTATION
[122]	9 patients 55±11 years (33–71)	First 10 days Day 3		Cyclosporine (same dose throughout the period) Prednisolone	HPLC-UV	Non-compartmental PK analysis; AUC Trap; data are presented as median (range)
			IV, D = 1.5 g BID Infusion over 3 hours (day 0 – day 5)			AUC0-6
					MMF	7.5 (4.2–37.2)	--	4.7 (2.2–18.4)	2.0 (1.0–3.0)	--
						AUC0-12
					MPA	34.2 (22.3–52.1)	0.65 (0.24–1.35)	10.2 (7.6–16.7)	2.0 (1.0–3.0)
					MPAG	1,030 (537–2,049)	69 (34–153)	110 (65–187)	3.0 (2.5–3.5)
					AcMPAG	8.5 (5.6–11.9)	0.36 (0.09–0.69)	1.31 (0.99–1.68)	1.3 (1.0–1.7)
		Day 5				AUC0-6
					MMF	6.5 (5.0–26.3)	--	3.7 (2.3–10.3)	2.0 (1.0–2.0)	--
						AUC0-12
					MPA	33.8 (23.7–43.1)	0.96 (0.26–1.11)	9.8 (6.1–11.2)	2.0 (2.0–3.0)
					MPAG	881 (456–1,584)	63 (34–135)	109 (51–166)	3.2 (3.0–4.0)
					AcMPAG	5.4 (1.8–9.3)	0.25 (<0.05–0.59)	0.86 (0.38–1.22)	3.0 (3.0–4.0)
		Day 6	Oral, D = 1.5 g BID (day 5 – day 10)			AUC0-12
					MPA	29.7 (22.3–37.6)	0.56 (0.24–1.12)	6.0 (4.4–10.5)	2.0 (0.33–6.0)	--
					MPAG	959 (471–1,216)	56 (26–82)	96 (51–111)	4.0 (0.67–8.0)
					AcMPAG	5.1 (3.2–5.9)	0.14 (<0.05–0.4)	0.77 (0.48–1.30)	2.0 (0.67–6.0)
		Day 10				AUC0-12
					MPA	33.8 (26.6–40.3)	0.95 (0.48–1.39)	9.0 (3.6–10.8)	1.25 (0.67–1.25)	--
					MPAG	1,032 (573–1,498)	79 (28–122)	108 (59–179)	2.5 (1.25–4.0)
					AcMPAG	6.4 (4.5–10.7)	0.32 (0.19–0.85)	0.90 (0.44–1.36)	2.0 (1.25–4.0)
[121]	9 patients Mean: 59 years	Weeks 1, 2, 4, 8, 12	D = 1 g or 1.5 g BID	Cyclosporine Prednisone	HPLC-UV	Non-compartmental PK analysis (n = 44 full MPA profiles); AUC0-12 Trap
						45.9±15.4 (13.4–91.7)	--	10.4±6.6	1.25	--
						r2 C2-AUC0-12 = 0.610; r2 C12-AUC0-12 = 0.003 Recommended sparse sampling algorithms: AUC0-12 = 5.568 + 0.902*C1.25 + 2.022*C2 + 4.594*C6; r2 = 0.926 AUC0-12 = 3.800 + 1.015*C1.25 + 1.819*C2 + 1.566*C4 + 3.479*C6; r2 = 0.948
[137]	14 patients 57±13 years	54±42 months	D = 35±7 mg/kg/day	Cyclosporine	N/R	Method for PK analysis: N/R
					MPA	AUC0-12 = 63±11	--	--	1–2	Cav = 5.2±1.0
						r2 AUC0-12 C0 = 0.48 C1, C2 = 0.08–0.09 C3, C4, C6, C8, C12 = 0.23 to 0.60
[132]	38 patients OHT 53±10 years	310±278 days	Steady-state D = 1.1±0.4 g BID (13.6±4.9 mg/kg BID)	No induction Cyclosporine Prednisone	HPLC-UV	MPA AUC0-12 calculated from a 2-hour abbreviated AUC developed in renal transplant patients[216]
						AUC0-12
					MPA	44.5±16.1	1.2±0.6	--	--	f = 1.9±0.4%
					Free MPA	0.83±0.30
						r2 MPA C0-AUC0-12 = 0.40 (p = 0.01) r2 MPA C0-MPA free fraction : ns
[134]	62 patients Mean: 59 years	Stable, > 1 year	Steady-state		EMIT	Non-compartmental PK analysis; AUC0-12 estimated from C0, C30, C120 using a sparse sampling algorithm developed in renal transplant patients[120]
						AUC/D: (h.mg/L)/g MMF	C0/D: (mg/L)/g MMF	Cmax/D: (mg/L)/g MMF
	n = 47 57±9 years	3.6±4.0 years	D = 2.9±0.8 g/day D ≥ 3 g/d, n = 32	Cyclopsorine Corticosteroids (45% patients)		AUC0-12	C0/D = 1.41±0.95	Cmax/D = 18.9	Med.: 40 min	--
					MPA	41.9±14.1 (19.7–81.8)			Mean: 75 min
						< 40 in 50% pts with dose ≥ 3 g/day
						AUC0-12/D = 31.9±16.1 (13.4–82.3)
						r2 AUC0-12 C0 = 0.36, p < 0.01 C40, C75, C120 = 0.62–0.64, p < 0.01
	n = 15 64±10 years	8.5±3.6 years (p < 0.01)	D = 1.9±0.7 g/day p < 0.001 D = 1–3 g/day, n = 15	Sirolimus Corticosteroids (13% patients; p < 0.01)		AUC0-12	C0/D = 5.10±3.41	Cmax/D = 21.8 (ns)	Med.: 40 min	--
					MPA	51.1±15.8 (34.4–87.6); p < 0.03	p < 0.001		Mean: 75 min
						< 40 in 27% pts with dose = 1–3 g/day
						AUC0-12/D = 61.0±27.4 (23.7–131.5); p < 0.001
						r2 AUC0-12 C0, C75, C120 = 0.56–0.61, p < 0.01 C40 = 0.82, p < 0.01
	56 years	3.4 years	Mean D Before the switch: 2.9±1.0 g/day	Switch group (cyclosporine to sirolimus), n = 9
						AUC0-12	C0/D = 4.85±4.39 (0.44–5.20)	--	--	--
					MPA	49.9±12.4 (34.3–56.9)
						AUC0-12/D = 37.5±19.9 (20.6–73.6)
			After the switch: 2.0±0.7 g/day (p < 0.01)			37.5±20.0 (34.6–70.1) (p < 0.001)	C0/D = 2.00±1.68 (0.33–14.6)	--	--	--
							p < 0.001
						AUC0-12/D = 58.3±32.5 (27.7–131.5); p < 0.001
[136]	26 patients 15±10 years (1 month –33 years)	50% < 1 year 50% > 1 year	Steady-state D = 37.9±12.5 mg/kg = 1.207±0.302 g/m2	Cyclosporine or tacrolimus Corticosteroids	HPLC	Non-compartmental PK analysis; 120 samples
					MPA	--	2.2±2.0 C0 ≥ 1 mg/L in50% of patients	--	--	--
					MPAG	--	48±4	--	--	--
	16 children		1.1±0.3 g/m2/day	Cyclosporine, n = 8 40 samples	MPA	--	1.6±1.5	--	--	--
					MPAG	--	49±38	--	--	--
					MPAG/MPA	--	25.2±21.7	--	--	--
			D: N/R	Tacrolimus, n = 8 42 samples	MPA	--	3.0±2.2 (vs children on CsA, p = 0.04)	--	--	--
					MPAG	--	36±17 (vs children on CsA, p = 0.04)	--	--	--
	10 adults		1.3±0.4 g/m2/day (vs. children on cyclosporine, ns)	Cyclosporine, n = 10 37 samples	MPA	--	2.3 ±2.2 (vs children on CsA, ns)	--	--	--
					MPAG	--	98±47 (vs children on CsA, p < 0.0001)	--	--	--
					MPAG/MPA	--	37.7±40.2 (vs children on CsA, p = 0.016)	--	--	--
[143]	7 patients 21–63 years	6–23 days	Steady-state D = 1 or 1.5 g BID	Cyclosporine Corticosteroids	HPLC-UV	Non-compartmental PK analysis
						--	--	--	--	fMPA = 3.6±3.9%
										fMPA – total MPA
										r2 = 0.20
										fMPAG = 26±8%
										fMPAG – total MPAG
										r2 = 0.87
										fMPA – fMPAG :
										r2 = 0.83
[135]	23 adults n = 14 57±13 years	Stable, > 1 year	Steady-state 34.6±6.0 mg/kg/day	Group 1 Cyclosporine	EMIT	Non-compartmental analysis – AUC0-12 Trap
						62.8±11.5	2.9±1.3	--	1-2	--
						r2 AUC0-12 C30, C1, C2 = 0.06–0.09 C0, C3, C4, C8, C12 = 0.23–0.69 C6 = 0.60
	n = 9 54±12 years		24.3±11.6 mg/kg/day	Group 2 Tacrolimus		55±24	2.4±1.2	--	30 min–1 h	--
						r2 AUC0-12 C30 = 0.04 C0, C1, C2, C3, C12 = 0.38–0.72 C4, C6, C8 = 0.85–0.93
[167]	28 patients 54 years (19–67)	Stable, > 1 year 2.5±3 years	1.5 g/day (0.5–4.0) D adjusted on C0 = 1.5–4 mg/L	Tacrolimus	HPLC-UV	Non-compartmental PK analysis –AUC0-12 Trap
						42.0±18.3 (12.2–79.8)	2.0±1.2	--	--	C12h = 2.0±1.7
						r2 AUC0-12 C0, C3 = 0.14–0.17 C2, C10, C12 = 0.25–0.39 C0.5, C1, C6, C8 = 0.48–0.57 Recommended sparse sampling algorithms: AUC0-12 = 1.11*C0.5 + 1.16*C1 + 3.72*C2; r2 = 0.84 AUC0-12 = 3.37*C0 + 0.97*C0.5 + 1.20*C1 + 2.70*C2; r2 = 0.87
HEART TRANSPLANT RECIPIENTS + LUNG TRANSPLANT RECPIENTS
[124]	9 patients(5 HTx, 4 LTx)		Steady-state 2–3 g/day	ATG in HTx patients Cyclosporine (n = 7) or tacrolimus (n = 2) Prednisone	HPLC-UV Total and unbound MPA	Non-compartmental PK analysis using WinNonlin in the 7 patients on cyclosporine (4 HTx, 3 LTx) AUC0-τ, τ = N/R
	53±11 years	Period 1 15±13 days (n = 10)	2.4±0.5 g/day			AUC = 30.1±18.2	3.79±4.04	Cmax = 9.34±4.24	4.0±4.2	Cmin = 0.33±0.58
						DN-AUC = 26.0±15.0		DN-Cmax = 8.00±4.15		f = 6.1±2.8 %
						fAUC = 2.22±2.32
		Period 2
		56±33 days (n = 46)	2.4±0.5 g/day			AUC = 19.8±11.0	2.21±2.63	Cmax = 8.30±4.19	0.7±0.4	Cmin = 0.28±0.50
						DN-AUC = 18.2±12.0		DN-Cmax = 7.20±3.81		f = 4.3±2.4 %
						fAUC = 1.01±0.81
		Period 3
		125±73 days (n = 114)	2.2±0.4 g/day			AUC = 36.8±29.7	2.20±2.29	Cmax = 10.32±7.86	1.7±1.3	Cmin = 1.21±1.78
						DN-AUC = 32.5±23.7		DN-Cmax = 9.49±7.78		f = 4.4±3.0 %
						fAUC = 1.50±2.16
[130]	50 patients 56.5 years (20.7–77.6)		Steady-state		HPLC-UV	Non-compartmental PK analysis using WinNonlin
	27 LTx (6 CF) 49.9 years (20.7–70.5)	2.1 years (0.2–14.0)	1.5 g BID (0.50–1.50) 0.034 g/kg/day (0.013–0.054)	Prednisone (n = 27) Cyclosporine (n = 11)
						DN-AUC0-12		DN-Cmax		MPA DN-Cmin
					MPA	18.6 (3.4–35.1)	--	5.66 (0.64–15.53)	1.0 (0.3–6.0)	0.44 (ND-1.05)
					MPAG	387 (152–909)	--	--	--
					AcMPAG	8.7 (ND-147)	--	--	--	fMPA: N/R
					MPA	DN-AUC0-6
					MPAG	14.3 (2.5–28.2)
					AcMPAG	226 (69–495)
						5.8 (ND-76)
					MPA	DN-AUC6-12
					MPAG	6.5 (0.9–16.8)
					AcMPAG	162 (61–427)
						3.0 (ND-70)
					MPA	AUC6-12/AUC0-12 ratio
						0.27 (0.12–0.55)
						fAUC: N/R
						MPAG/MPA = 28.5 (9.8–55.2)
						AcMPAG/MPA = 0.4 (ND-12.3)
				Tacrolimus (n = 16)		DN-AUC0-12		DN-Cmax		MPA DN-Cmin
					MPA	41.4 (8.3–115.31) (p = 0.022)	--	8.24 (1.81–37.11)	1.1 (0.3–10.0)	1.13 (0.17–3.64) (p = 0.002)
					MPAG	471 (72–928)	--	--	--
					AcMPAG	12.7 (0.7–160)	--	--	--	fMPA = 1.9% (0.7–3.8)
						DN-AUC0-6
					MPA	25.4 (5.5–96.1) (p = 0.046)
					MPAG	257 (43–603)
					AcMPAG	4.6 (0.5–101)
						DN-AUC6-12
					MPA	14.8 (2.8–26.4) (p = 0.007)
					MPAG	166 (29–389)
					AcMPAG	7.40 (ND-59)
						AUC6-12/AUC0-12 ratio
					MPA	0.33 (0.17–0.56)
						fAUC = 1.73 (0.51–2.43)
						MPAG/MPA = 12.6 (2.4–25.8) (p = 0.002)
						AcMPAGMPA = 0.3 (0.05–2.4)
	21 HTx(and heart + kidney)	4.0 years (0.4–19.7)	0.75 g BID (0.25–1.50) (p < 0.05, lung vs heart)	Prednisone (n = 1) Cyclosporine (n = 14)
						DN-AUC0-12		DN-Cmax		MPA DN-Cmin
					MPA	50.9 (16.9–218.7)	--	16.8 (3.6–47.3)	1.5 (0.3–12.0)	1.45 (0.34–3.02)
	59.8 years (23.3–77.6) (p < 0.05)		0.019 g/kg/day (0.008–0.038) (p < 10−4, lung vs heart)		MPAG	219 (50–1869)	--	--	--
					AcMPAG	17.9 (2.2–178)	--	--	--	fMPA = 2.2% (0.2-15)
						DN-AUC0-6
					MPA	34.9 (10.7–134.0)
					MPAG	408 (35–1142)
					AcMPAG	7.7 (1.2–101)
						DN-AUC6-12
					MPA	18.9 (5.3–102.2)
					MPAG	287 (14–726)
					AcMPAG	6.6 (0.8–77)
						AUC6-12/AUC0-12 ratio
					MPA	0.33 (0.16–0.49)
						fAUC =1.38 (0.05–12.26)
						MPAG/MPA = 11.9 (0.9–28.6)
						AcMPAG/MPA = 0.2 (0.07–2.03)
				Tacrolimus (n = 9)		DN-AUC0-12		DN-Cmax		MPA DN-Cmin
					MPA	106.8 (40.9–180.5)	--	21.8 (9.9–44.5)	1.0 (0.4–10.0)	3.28 (1.29–8.40) (p = 0.009)
					MPAG	717 (113–1723)	--	--	--
					AcMPAG	20.1 (2.6–109)	--	--	--	fMPA: 3.5% (0.5–14)
						DN-AUC0-6
					MPA	58.1 (23.9–90.1)
					MPAG	419 (40–940)
					AcMPAG	13.4 (1.7–80.3)
						DN-AUC6-12
					MPA	29.2 (9.9–89.8)
MPAG					298 (60–829)
AcMPAG					9.4 (0.9–62.6)
					AUC6-12/AUC0-12 ratio
MPA					0.42 (0.23–0.61)
					fAUC = 3.54 (0.85–18.89)
					MPAG/MPA = 8.7 (2.3–13.4)
					AcMPAG/MPA = 0.3 (0.06–0.87)
LUNG TRANSPLANTATION
[119]	7 patients 50±10 years	4.4±3.9 years (0.3–11.5)	Steady-state, 1–3 g/day 2.4±0.9 g/day (35.5±14.1 mg/kg/day)	Cyclosporine Prednisone	HPLC-UV Total Unbound	Non-compartmental analysis using WinNonlin AUC0-τ, τ = N/R
						AUC = 45.8±18.4	--	Cmax = 17.37±7.69	1.2±0.4	Cmin = 3.12±1.41
						DN-AUC = 23.6±15.8		DN-Cmax = 9.30±9.66		f = 2.9±0.6%
						fAUC = 1.29±0.50
[138]	19 adults (4 CF) 48±15 years	Stable 4.2±3.7 years	Steady-state 2.5±0.5 g/day	Prednisone Cyclosporine (n = 9) or tacrolimus (n = 10)	HPLC-UV	Non-compartmental analysis using WinNonlin
						29.4±17.9	--	--	--	--
						r2 C0-AUC0-12 = 0.714; r2 C2-AUC0-12 = 0.663; r2 C8-AUC0-12 = 0.884 ; otherwise, r2 = 0.176–0.732 Recommended sparse sampling algorithms: log(AUC0-12) = 1.140 + 0.241*log(C0) + 0.406*log(C2); r2 = 0.828, ME = −5.8%, RMSE = 6.0% log(AUC0-12) = 1.09 + 0.202*log(C0) + 0.411*log(C1.5); r2 = 0.791, ME = −5.7%, RMSE = 7.0%

D: Dose – DN: Normalized to a 1000 mg MMF dose – f: Free fraction – HTx: Heart transplantation – LTx: Lung transplantation – N/R: not reported – OHT: Orthotopic heart transplantation.

Table XII. Mycophenolate pharmacokinetic parameters in cohort studies.

Results are expressed as mean±SD, unless otherwise specified

Ref	Number of patients	Post-transplant period	MMF dose	Co-administered immunosuppressant	Analytical method, molecule	F, transfer k	Vd/F (unless otherwise specified)	CL/F (unless otherwise specified)	T1/2 (h)
HEART TRANSPLANTATION
[122]	9 patients 55±11 years (33–71)	First 10 days		Cyclosporine (same dose throughout the period)	HPLC-UV MPA	Non-compartmental analysis – Data presented as medians (range) IV/PO AUC and Cmax ratios were calculated using log-transformed data and are presented as mean [90%CI]
		Day 3	IV (infusion over 3 hours) D = 1.5 g BID (days 0–5)	Prednisolone		--	--	--	--
		Day 5				--	--	--	--
		Day 6	Oral D = 1.5 g BID (days 5 – 10)			Day 6/Day 3: F = 82.9% [71.4–96.3] Cmax PO/IV = 63% [48–82]	--	--	--
						Day 6/Day 5: F = 91.6% [79–106] Cmax PO/IV = 74% [57–95]
		Day 10				Day 10/Day 3: F = 97.5% [83–114] Cmax PO/IV = 77% [60–99]	--	--	--
						Day 10/Day 5: F = 108% [93–125] Cmax PO/IV = 90% [71–115]
[137]	14 patients (57±13 yrs)	54±42 months	D = 35±7 mg/kg/day	Cyclosporine	N/R	Method for PK analysis: N/R
						--	--	23.3±6.5 L/min	--
HEART TRANSPLANT RECIPIENTS + LUNG TRANSPLANT RECPIENTS
[130]	50 patients 56.5 years (20.7–77.6)	2.7 years (0.2–19.7)	D = 1 g BID (0.25–1.50) 0.028 g/kg/d (0.008–0.054)		HPLC-UV	Non-compartmental PK analysis using WinNonlin
			Steady-state		MPA MPAG AcMPAG		Vd/F: L	CL/F: L
	LTx (n = 27) (6 CF) 49.9 years (20.7–70.5)	2.1 years (0.2–14.0)	1.5 g BID = (0.50–1.50)	Prednisone (27)
			0.034 g/kg/d (0.013–0.054)	Cyclosporine (11)		--	248.1 (54.1–644.6)	53.9 (28.5–294.7)	--
				Tacrolimus (16)		--	124.8 (29.8–607.1)	20.8 (8.7–120.5) (p = 0.026)	--
	HTx (n = 23) (5 HTx + KTx) 59.8 years (23.3–77.6) (p < 0.05)	4.0 years (0.4–19.7)	D = 0.75 g BID (0.25–1.50) (p < 0.05)	Prednisone (1)
				Cyclosporine (14)		--	101.5 (36.6–1141.1)	21.2 (4.6–59.2)	--
			0.019 g/kg/d (0.008–0.038) (p < 10−4)	Tacrolimus (9)		--	111.0 (53.2–261.3)	11.0 (5.5–24.4)	--

CF: Cystic fibrosis – D: Dose – HTx: Heart transplantation – KTx: Kidney transplantation – LTx: Lung transplantation – N/R: not reported.

Like calcineurine inhibitors, the PK of MMF is complex and somewhat erratic,^[118] with large intra- and inter-individual variability.^{[8,9,52,120,124,126,132,133]} All studies we identified in thoracic transplant patients were performed using a model-independent PK approach, reporting mean and standard deviation, or range of exposure indices and PK parameters derived from individual data (Tables XI & XII).^{[119,121,122,124,130,132,134–137]} We could not find any study using either the iterative two-stage (ITS) method or popPK analysis. Exposure indices of MPAG and AcMPAG were reported in 3 studies.^{[122,130,136]} Finally, except for four studies in which MPA was assayed by EMIT^{[123,126,134,135]} and one in which the analytical method was not reported,^[137] all the PK studies we identified measured MPA, MPAG and AcMPAG concentrations with HPLV-UV.

2.1.1. Absorption, bioavailability

The PK of MPA after oral and IV MMF administration is characterized by an important variability. In a study involving 9 heart transplant patients who received 1.5 g BID MMF by IV infusion over three hours, immediately after transplantation and for 5 days, C_max displayed large inter-individual variations, occurring between 1 and 3 hours and ranging from 2.2 to 18.4 mg/L.^[122] Following oral administration, MMF is rapidly absorbed (within 5 minutes of ingestion) in the upper gastro-intestinal tract because of its high solubility at a low pH.^[8] Hydrolysis to MPA is responsible for the rapid disparition of MMF from the plasma.^[118] MPA C_max after oral dosing is also highly variable, with a coefficient of variation ranging from approximately 30 to 80% depending on the studies,^{[119,121,124]} and usually occurs within 1 to 2 hours following oral MMF administration (range 20 minutes-12 hours).^{[119,121,122,130,134,135,137]} Delayed absorption has been reported when the drug is taken with food.^[131]

In a review by Shaw et al, the bioavailability of MPA was reported to be incomplete, suggesting metabolism in the gastro-intestinal tract. This pre-systemic metabolism may take a part in the high overall PK variability of MPA.^[133] On the contrary, the above mentioned study performed in 9 heart transplant patients reported excellent, high and consistent MPA bioavailability after MMF oral dosing compared with the IV formulation (oral F = 71 to 125%).^[122]

Secondary peaks of MPA plasma concentration have been observed between 6 and 12 hours after drug administration and have been attributed to enterohepatic circulation.^{[8,124,131,138,139]} This phenomenon is believed to contribute 10–60% to MPA exposure.^{[118,130,140]} However, these secondary peaks are sometimes higher than the first peaks and in such cases, cannot be accounted for by enterohepatic cycling.^[141] Further exploration of the absorption of MPA would thus be of major interest.

2.1.2. Distribution

In whole blood, MPA is poorly distributed into cellular fractions, with over 95% found in plasma.^[118] In patients with normal renal and liver function, MPA is extensively and tightly bound to human serum albumin (97–99%).^{[118–120,133,142]} The variations in MPA PK and pharmacodynamics (PD) may be partly explained by this affinity for serum albumin. Indeed, as only unbound MPA is pharmacologically active, variations in the free fraction may be responsible for variations of IMPDH inhibition and thus of the therapeutic response.^{[118,119,133,142]} Moreover, in clinical situations with decreased serum albumin levels, such as severe hepatic impairment, or decreased serum albumin binding capacity, such as severe renal impairment, the increased proportion of free MPA may result in a higher clearance.^[119,133] It has been suggested that in such situations with variations of plasma serum albumin, free MPA concentration might be worth monitoring,^[119,120] though no paper reported that free MPA was better correlated to effects than total MPA plasma levels. Moreover, measuring free MPA is out of reach of traditional immunoassays.

Data on MPA protein binding in heart and lung transplantation was reported in only a few papers. In a cohort of 38 heart transplant patients (time post-transplantation: 310±278 days), the free MPA fraction (f) was 1.9±0.4%, with a free MPA AUC_0-12 (fAUC_0-12) = 0.83±0.30 h.mg/L,^[132] which is consistent with those reported in 7 lung transplant patients (time post-transplantation: 4.4±3.9 years), in whom f = 2.9±0.6%, and fAUC_0-τ = 1.29±0.50 h.mg/L;^[119] as well as with the free fraction reported in 7 heart transplant recipients (time post-transplantation: 6–23 days), in whom f = 3.6±3.9%.^[143] In another group of 7 heart (n = 3) and lung (n = 4) transplant patients, f was higher, decreasing from 6.1±2.8% immediately after transplantation (15±13 days) to 4.4±3.0% later on (125±73 days), with corresponding fAUC_0-τ of 2.22±2.32 h.mg/L and 1.50±2.16 h.mg/L, respectively.^[124] However, these results are to be interpreted cautiously, as f was determined in each patient after pooling all the plasma samples of a given profile. Indeed, averaging fractions is mathematically different from dividing the mean free by the mean total concentration and is simply not coherent. Finally, a recent study performed in a cohort of 23 heart (and heart-kidney) and 27 lung transplant recipients (time post-transplantation: 0.2–19.7 years) in whom MMF was associated to cyclosporine or tacrolimus, reported f values ranging between 0.2 and 15%, with fAUC_0-12 normalized to a 1 g-MMF dose of 0.05 to 19.0 h.mg/L.^[130] Although they were discussed, the factors potentially linked with these high variations in free fraction were not investigated (nor reported).

Once absorbed, MPA is rapidly distributed in the tissues, within approx. 2 hours after t_max.^[133] Values for MPA V_d/F were reported in only one study, conducted in 27 lung and 23 heart (and heart-kidney) transplant recipients who received MMF associated with either cyclosporine or tacrolimus.^[130] In the lung transplant group, mean V_d/F = 248 L (range: 54–645) and 125 L (range: 30–607) in the patients on cyclosporine (n = 11) and on tacrolimus (n = 16), respectively. In the heart transplant group, mean V_d/F = 102 L (range: 37–1,141) and 111 L (range: 53–261) in the patients on cyclosporine (n = 14) and on tacrolimus (n = 9), respectively. A statistically significant difference on V_d/F was found neither between heart and lung transplant recipients, nor between patients on cyclosporine and those on tacrolimus.

2.1.3. Metabolism and elimination

MPA is primarily metabolized by uridine diphosphate glucuronyl transferases (UGT, mainly UGT1A9, 1A7 and 1A8) into its pharmacologically inactive metabolite, 7-O-mycophenolic acid glucuronide (MPAG).^[144] MPAG accounts for more than 90% of all MPA metabolites, but at least 4 other metabolites have been identified in plasma from renal, liver and cardiac transplant recipients,^[144–147] including the acyl glucuronide of MPA (AcMPAG), which inhibits IMPDH in vitro.^[122,148] The primary site of conversion is the liver, but MPA may also be glucuronidated in the gastro-intestinal tract mucosa and the kidney.^[8,94,118] As for MPA, concentrations of MPAG and AcMPAG are highly variable throughout the dose interval. MPAG plasma concentrations are said to be usually 20 to 100 times higher than MPA concentrations,^[118] but this seems to be variable across sampling times and studies. In 9 heart transplant patients, MPAG C₀ was on average 65–106 times higher than MPA C₀ after IV or oral MMF administration, whereas MPAG C_max was only 10 to 16 times higher than MPA C_max.^[122] This study was the only one we identified reporting MPAG C_max (51 to 187 mg/L) and t_max (40 minutes to 8 hours). In another set of 26 heart transplant recipients, MPAG C₀ was only 22–30 times higher than MPA C₀.^[136] The overall difference in exposure between MPAG and MPA may be more precisely evaluated by comparing their AUC. The AUC of MPAG was reported in two studies in thoracic transplantation, and was on average 9 to 32 times higher than that of MPA.^[122,130] In a study conducted in 7 heart transplant recipients,^[143] MPAG free fraction was estimated to be 26±8%, which is consistent with literature reports in renal transplant recipients (MPAG free fraction of 18–37% in 27 “stable” patients; time post-transplantation not reported).^[149] MPAG is excreted into the bile by a mechanism probably involving the multidrug resistance protein 2 (MRP2). Upon delivery to the GI tact, through the action of β-glucuronidases shed by gastro-intestinal tract bacteria, part of MPAG is deconjugated back to MPA, which then undergoes enterohepatic recirculation.^{[8,94,118,133]}

As opposed to MPAG, AcMPAG was once thought to display a pharmacological activity similar to that of MPA as it did in vitro with the recombinant enzyme,^[146] but a more recent in vitro study suggested that, due to a lower intrinsic inhibitory activity, a lower ability to cross cell membranes and lower plasma concentrations, AcMPAG probably exhibits a negligible IMPDH inhibition activity in vivo. (Gensburger O et al, unpublished observations) AcMPAG t_max varies between 40 minutes and 6 hours, AcMPAG C₀ and C_max are 2 to 4 times lower and 8 to 11 times lower than MPA C₀ and C_max, respectively.^[122] In two studies, AcMPAG AUC was 2.5 to 6 times lower than MPA AUC ^[122,130], which seems to be higher than the average of 10.3% reported in renal transplantation – though in this study AcMPAG was determined as MPAG equivalent due to the absence of a standard compound for quantification.^[150]

We found two studies reporting MPA CL/F, with diverging results, showing the difficulty of estimating this PK parameter using a non-compartmental approach. The first study, performed in 14 heart transplant recipients who were all on MMF-cyclosporine, reported CL/F = 1,397±389 L/h.^[137] The second study included 27 lung and 23 heart transplant recipients who received MMF associated with either cyclosporine (n = 25) or tacrolimus (n = 25).^[130] In the lung transplant group, mean CL/F was 54 L/h (range: 29–295 L/h) and 21 L/h (9–121) in patients on cyclosporine and tacrolimus, respectively (p = 0.026). In the heart transplant group, mean CL/F was 21 L/h (5–59) and 11 L/h (6–24) in patients on cyclosporine and tacrolimus, respectively (ns). CL/F in the lung transplant group was significantly higher than in the heart transplant group, regardless of the associated calcineurin inhibitor (36 vs 13 L, p = 5.10⁻⁴). One of the hypotheses made by the authors to explain this difference between heart and lung transplant recipients is that the higher albumin levels observed in the heart transplant group may have contributed to an increase in MPA protein binding and thus a decrease in total CL/F of MPA.

Apparent half-life of MPA estimated from apparent distribution volume and clearance reported in the abovementioned study^[130] was comprised between 3.2 and 7.0 hours, which is shorter compared to 18 hours reported in a review published in 1997.^[94]

2.1.4. Drug-drug interactions

The variability in MPA concentrations may be partly explained by drug-drug interactions affecting the MPA concentrations achieved for a given MMF dose. Indeed, the PK of MPA is known to be altered by concomitant medications.^[142] First, it is strongly affected by the nature of the associated calcineurin inhibitor. Such observations were initially made in renal transplant patients on tacrolimus, who displayed significantly higher MPA C₀ and AUC than patients on cyclosporine.^[151] In heart transplantation, a randomized study in 60 patients found significantly lower MMF dose requirements for patients on tacrolimus (n = 30) than on cyclosporine (n = 30) to achieve MPA C₀ targets of 0.45 to 4.0 mg/L (1.6±0.8 vs 2.8±1.0 g/day at one year post-transplantation, p = 4.10⁻⁵).^[152] A prospective, randomized controlled study in 50 heart transplant patients compared the MMF – tacrolimus (n = 25) and the MMF – cyclosporine (n = 25) combinations in terms of efficacy, toxicity and PK.^[153] Both groups displayed similar MPA C₀ (2.3±0.5 μg/L vs 2.7±0.6 μg/L), although the patients on cyclosporine received higher doses of MMF (3.3±0.5 g/day vs 2.1±0.7 g/day, p = 0.04). In 30 adult stable lung transplant recipients (7 CF) over the first 2 years post-transplantation, MPA C₀ were 50% lower in patients on cyclosporine (n = 16, C₀ estimated at approx. 1.3 mg/L) than in patients on tacrolimus (n = 14, C₀ estimated at approx. 2.7 mg/L, p < 10⁻⁵) despite substantially higher MMF doses (estimated at approx. 25 vs 22 mg/kg/day).^[142] In a study in 26 heart transplant patients aged 1 month to 33 years, lower MPA concentrations and higher MPAG/MPA C₀ ratios were reported in a subset of 8 children on cyclosporine compared to 8 others on tacrolimus (C₀ = 1.6±1.5 vs 3.0±2.2 mg/L, p = 0.04 and average MPAG/MPA C₀ ratio = 25.2 vs 11.9, p = 0.005).^[136] Unfortunately, the MMF dose was only reported in children on cyclosporine, allowing no dose comparison.

Two other studies compared MPA C₀ (assayed by EMIT) in patients on cyclosporine vs tacrolimus,^[123,126] but both pooled data from kidney and lung transplant recipients. The first one included 120 kidney and 20 lung transplant patients and reported higher MPA C₀ in patients on tacrolimus than in patients on cyclosporine (3.63±2.63 vs 2.14±1.22 mg/L, p < 0.01) despite similar MMF doses (2 g/day).^[126] The second one included 49 kidney and 11 lung transplant recipients and reported higher median dose-to-concentration (D/C) ratios in patients on cyclosporine than in patients on tacrolimus (p < 10⁻⁴),^[123] which is an atypical presentation of results similar to those of the previous studies.

This was confirmed by a later study where 27 lung transplant patients were included together with 23 heart (heart-kidney) transplant recipients,^[130] where the difference between dose-normalized AUC (DN-AUC) of lung transplant recipients on cyclosporine (n = 11) or tacrolimus (n = 16) was significant (DN-AUC = 18.6 h.mg/L, range: 3.4–35.1, vs 41.4 h.mg/L, range: 8.3–115.3, p = 0.022). Consistently, MPAG/MPA AUC ratios were higher in patients on cyclosporine (28.5, range: 9.8–55.2) compared to patients on tacrolimus (12.6, range: 2.4–25.8, p = 0.002). On the contrary, no significant difference on DN-AUC was found between the 14 heart transplant recipients on cyclosporine (50.9 h.mg/L, range: 16.9–218.7) and the 9 on tacrolimus (106.8 h.mg/L, range: 40.9–180.5, p = 0.14). Neither was the difference between MPAG/MPA AUC ratios of these 2 groups of heart transplant recipients significant (11.9, range: 0.9–28.6 in patients on cyclosporine vs 8.7, range: 2.3–13.4 in patients on tacrolimus, p = 0.10).

In summary, dose-adjusted MPA concentrations are significantly higher and MPAG/MPA concentrations ratios lower when MMF is associated with tacrolimus than cyclosporine.^[136,142] The mechanisms underlying the interactions between MPA and calcineurin inhibitors are still being investigated. At first, some authors postulated that the increase of MPA concentrations in patients on tacrolimus was caused by the inhibition of UGTs and thus of MPA glucuronidation by tacrolimus.^[154] However, most authors now agree that these differences in MPA concentrations are due to a diminution in exposure caused by cyclosporine, rather than an increase caused by tacrolimus.^[133] Indeed, it has been suggested that cyclosporine decreases MPA exposure by inhibiting the biliary excretion of MPAG into the gut, reducing back-transformation of MPAG into MPA through enterohepatic circulation.^{[126,133,142,155,156]} It has also been suggested that cyclosporine might inhibit the deglucuronidation of MPAG to MPA by inactivating the β-glucuronidases in the gastro-intestinal tract.^[157]

Consistently, in a recent study in 62 heart transplant recipients, exposure to MPA was shown to be higher in patients on sirolimus compared to those on cyclosporine for similar doses of MMF, as evidenced by significantly higher dose-adjusted AUC_0-12 in patients on sirolimus (31.9±16.1 vs 61.0±27.4 h.mg/L/1000 mg MMF, p < 0.001),^[134] which is in favour of cyclosporine being the interacting drug rather than tacrolimus and sirolimus.

In conclusion, the associated immunosuppressive drugs (cyclosporine, tacrolimus, sirolimus) need to be taken into account when defining MMF dose, as the under-immunosuppression potentially resulting from decreased MPA exposure in patients on cyclosporine may strongly influence clinical outcome.^[142] Also, the influence of corticoids, which are known to induce the expression of the UGTs and have been shown to induce MPA metabolism in renal transplant patients,^[158] might be worth investigating in thoracic transplantation.

2.1.5. Special populations

The PK of mycophenolate can be affected by several other factors such as pre-existing conditions (age, pathology, activity of UGTs and drug transporter MRP-2), and clinical status (time post-transplantation, gastro-intestinal motility, hypoalbuminemia, liver disease, renal dysfunction). These factors can alter plasma MPA, MPAG and AcMPAG total and free concentrations and global exposure, hence the degree of immunosuppression.

Pediatric thoracic transplantation will be briefly presented before focussing on CF.

2.1.5.1. Paediatrics

Children display different PK profiles from those of adults’ ^[44,45] and different MMF disposition rates are expected in children vs adults, based on the ontogeny of UGTs.^[159] Very little information is available on appropriate dosing or on PK of MMF in children. Similarly to adults, MMF in children is rapidly absorbed and hydrolyzed to MPA, with a first peak occurring approximately 45 minutes after intake and a potential second peak at a variable time. However, there is no information on the correlation between MPA exposure and clinical outcome in this population.^[160] The recommended MMF starting dose is 1800 mg/m²/day in two divided doses in children < 2 years and 1200 mg/m²/day in older children and adolescents.^[120,159] We identified two studies performed in heart transplant children. The first one was a retrospective review of MPA C₀ in 44 children (aged 7 days to 18 years; associated calcineurin inhibitor not reported).^[160] MMF doses (in mg/kg and mg/m²) required to reach the target C₀ < 3 mg/L tended to decrease with increasing age (2.2±0.7 g/m² between 0 and 1 year, 2.2±0.9 g/m² between 1 and 5 year, 1.8±0.9 g/m² between 5 and 10 years, 1.6±0.8 between 10 and 16 years, and 1.6±0.6 beyond 16 years). However, the authors emphasized the limitations of their study, which included the retrospective design and more importantly the diversity of the population in terms of indication for MMF treatment (acute rejection prophylaxis or treatment) and time post-transplantation (7 days to 10 years). In the second study, a comparison was made between 8 children (with no more precision) and 10 adults who received MMF in association with cyclosporine.^[136] Adults and children received comparable MMF daily doses (1,280±360 vs 1,180±260 mg/m², ns) and displayed comparable MPA C₀ (2.3±2.2 vs 2.2±2.0 mg/L, ns), but MPAG C₀ and MPAG/MPA C₀ ratios were significantly higher in adults (98±47 vs 49±38, p < 10⁻⁴ and 37.7±40.2 vs 25.2±21.7, p = 0.016). The authors hypothesized that this might be explained by the existence of an alternate pathway in children with enhanced MPAG clearance compared to adults, or by higher glomerular filtration and tubular secretion rates in children, or even that cyclosporine-mediated inhibition of MPAG excretion may also be a pathway specific to adults, not active (or to a lesser extent) in younger patients. In our opinion, not all these hypotheses have the same credibility, and we would rather favour the higher renal excretion of MPAG in children, which can also be put forward for CF patients.

2.1.5.2. Cystic fibrosis

In CF patients, chronic hepatic and gastro-intestinal disorders affect drug absorption, and lower serum albumin levels may result in an increased plasma clearance of protein-bound drugs, resulting in potentially higher risks for under-immunosuppression and therapeutic failure.^[142] Unfortunately, little data is available on the PK of MPA in this population.^[124] In a study of lung transplant recipients, patients with CF received on average 30% higher MMF doses than weight-matched patients without CF, all being given tacrolimus (doses estimated at approx. 32 vs 22 mg/kg/day, p < 10⁻⁵), to reach MPA C₀ of approx. 2.5 vs 1.5 mg/L (ns).^[142] Among the 21 patients recruited in another study, 5 had CF and received similar MMF doses compared to the remaining 16 patients (data not shown).^[161] In CF patients, the metabolic AUC ratio of MPAG/MPA was significantly lower than in non-CF patients (14.3 vs 25.1, p < 0.01), which according to us could result from increased renal MPAG excretion, decreased hepatic metabolism or increased biliary excretion in CF patients, in order of decreasing probability. However, due to the small number of patients, these results would need to be confirmed and further investigations on the metabolism of MPA in CF patients would be of major interest.

2.2. Therapeutic drug monitoring

The advantages and limitations of TDM of MMF in transplantation were discussed recently in two reviews.^[120,162] While TDM is mandatory for calcineurin inhibitors, it is not for MPA and MMF is usually administered at a fixed oral dose (adults: 2–3 g/day in combination with cyclosporine and 1.5–2 g/day in combination with tacrolimus; children, at least in renal transplantation: 600 mg/kg body weight twice daily, up to a maximum daily dose of 2 g). As a consequence, MPA levels have not been routinely monitored despite PK and PD (variations of IMPDH inhibition) variability comparable to that of calcineurin inhibitors.^[8,124,163] However, a significant relationship between MPA AUC and clinical outcomes and the clinical benefits of MMF monitoring for dose individualization have been demonstrated in a growing number of studies, including prospective concentration-controlled clinical trials, notably in renal transplantation and, to some extent, in heart transplantation.^{[16,124,132,164–166]} Controlled exposure to MPA is essential to maximize its immunosuppressive effects and minimize its toxicity.^[120] Despite this, MPA monitoring is still debated and no consensus has been reported on the best TDM tool for the optimal management of graft rejection and adverse effects (diarrhea, leucopenia, infections).^{[118,120,135]} TDM guidelines for MMF in heart and lung transplant patients are even more limited.^[120]

2.2.1. Single concentrations monitoring

When performed, MMF monitoring is sometimes based on MPA C₀ concentrations. In heart transplantation, the current recommendation is an MPA C₀ target of 1.2–3.5 mg/L when MPA is assayed by HPLC (C₀ ≥ 2 mg/L when MPA is assayed by EMIT).^[120]

Studies of the relationship between MPA C₀ and AUC_0-12 in heart transplantation have reported poor correlation, with coefficients of determination (r²) ranging from 0.36 to 0.69.^{[132,134,135,137]} A better correlation was found with different time points, such as C_40min (r² = 0.82),^[134] C₄ (r² = 0.86), C₆ (r² = 0.85) or C₈ (r² = 0. 93) (Table XI).^[135] However, AUC_0-12 was not always measured and was sometimes estimated using multiple linear regression equations developed in other populations or transplant types.^[134]

It is noteworthy that C₀ may not be the minimal concentration during the dosing interval, because of secondary peaks that occur between 6 and 12 hours after drug administration and of more or less delayed absorption, which translates into decreasing blood levels in the 10–60 min post-dose.^[141] These secondary peaks may also explain the poor correlation of MPA C₀ with AUC_0-12.^[124,163]

2.2.2. AUC monitoring

2.2.2.1. Full AUC

Mycophenolate is characterized by a poor correlation between blood concentrations and dose. In a study performed in 120 kidney and 20 lung transplant recipients on MMF and cyclosporine (n = 107) or tacrolimus (n = 33), no correlation was found between MMF dose and MPA C₀ assayed by EMIT.^[126]

As with calcineurin inhibitors, AUC_0-12 is likely to be the most useful TDM tool rather than single concentrations. The recommended therapeutic range for MPA AUC_0-12 after renal transplantation is 30–60 h.mg/L.^[120] In thoracic transplantation, MPA AUC_0-12 was calculated using the linear trapezoidal rule in almost all studies we identified. In two studies in heart transplant patients co-administered cyclosporine or sirolimus,^[132,134] AUC_0-12 was estimated using the EMIT assay and sparse sampling algorithms previously developed using HPLC in adult kidney transplant recipients treated with MMF-cyclosporine. The estimation precision is thus questionable because, as previously discussed sparse sampling strategies are usually specific of the transplant population, analytical method and associated immunosuppressant, unless validated in another population of interest. Fortunately, in these two studies the AUC estimates were not used for dose adjustment.

2.2.2.2. Sparse sampling strategies

As discussed previously, the determination of full AUC_0-12 is impractical compared to other strategies such as C₀ or abbreviated sampling estimation of the drug’s exposure. No abbreviated AUC has been proposed as a surrogate exposure index in thoracic transplantation. Experience of sparse sampling strategies is limited to three studies, two in heart and one in lung transplantation (Table XI). In heart transplant patients, the proposed sampling strategies included three or four MPA concentrations: at 1.25h, 2h, 6h and possibly 4h after MMF administration in the first study^[121]; and at 0.5h, 1h, 2h and possibly the trough in the second.^[167] In lung transplantation, an algorithm was proposed for the estimation of log(AUC_0-12), based on the log(concentration) measured at pre-dose and either 1.5h or 2h post-dose.^[138] Of note, both algorithms were developed using concentration-time profiles from a small number of patients. Clinical acceptability of sparse sampling strategies is better when the samples are drawn within 4 hours post-dose, which is not the case of the former. On the other hand, samples drawn before 6 to 8 hours post-dose are likely to miss MPA enterohepatic recycling,^[118] unless Bayesian estimators, able to compensate for this, are used.^[168]

2.2.2.3. Bayesian forecasting

To our knowledge, MAP-Bayesian procedures using popPK models have been developed for the estimation of AUC_0-12 in renal transplant patients,^[168,169] but not in thoracic transplantation. Much still has to be done in order to refine TDM strategies for MMF after thoracic transplantation.

2.2.3. Impact of MMF TDM on patient outcome

Only a few studies have evaluated the relationship between exposure to MPA and clinical outcomes in heart transplantation,^{[129,132,136,139,163,170]} and one in lung transplantation,^[142] most of which were retrospective and dealt only with C₀ (Table XIII).

Table XIII. MMF exposure – effect studies in heart transplantation.

Results are expressed as mean±SD, unless otherwise specified

Ref	Immunosuppressants	Number of patients	Sampling periods	Exposure (analytical method)		Event (Number of episodes)		Comments
[132]	MMF, steady-state 1.1±0.4 g BID (13.6±4.9 mg/kg BID)	38 patients OHT 53±10 years	310±278 days	MPA AUC, fAUC, C0 (HPLC-UV) MPA AUC0-12 calculated from a 2-hour abbreviated AUC developed in renal transplant patients[216]		3 groups (ISHLT classification)		No significant differences in cyclosporine exposure
				AUC	42.8±14	Grade 0		Prednisone dose varying among the patients
				fAUC	0.81±0.25	n = 22 patients
				C0	1.20±0.58
				AUC	51.7±17.5	Grade 1		Small sample size (3 patients in the grade2/3 rejection group)
				fAUC	0.95±0.34	n = 13 patients
				C0	1.24±0.72
				AUC	26.1±6.6 (p < 0.05 vs grade 0 and 1)	Grade 2/3		Patients sampled at varying times throughout their post-transplantation course
				fAUC	0.49±011 (p < 0.05, vs grades 0 and 1)	n = 3 patients
				C0	0.65±0.15 (ns vs grades 0 and 1)
[163]	MMF, steady-state Increased to the maximal tolerated dose D = 2.1±0.9 g/day (1.0–4.0) Corticosteroids (50%)	26 patients 54±14 years (22–72)	3.0±1.7 years (1.0–7.8)	MPA and MPAG C0 (HPLC-UV) “Therapeutic” C0 levels: MPA > 2 mg/L Cyclosporine ≥ 175 μg/L Tacrolimus ≥ 10 μg/L		48 routine EMB at time of blood sampling (ISHLT) Grade 0, 46%; grade 1A/1B, 33%; grade 2, 8%; grade 3A, 4% MPA C0 vs overall incidence of rejection: ns
	Cyclosporine (42%)			n = 19 samples		Overall rejection ≈ 20%
	MMF D = 2.4±0.9 g/day			MPA C0 = 1.65±0.97		Cyclosporine C0 < 175 μg/L in 100% of patients
				MPAG C0 = 128.9±55.6
	Tacrolimus (58%)			n = 29 samples		Overall rejection ≈ 60% (significantly higher, p: N/R)
	MMF D = 1.8±0.8 g/day (p = 0.01)			MPA C0 = 2.86±2.07 (p = 0.02)		Tacrolimus C0 < 10 μg/L in 52% of patients
				MPAG C0 = 101.1±84.9 (ns)
[136]	MMF, steady-state D = 37.9±12.5 mg/kg = 1.207±0.302 g/m2	26 patients (16 children, 10 adults)	50% < 1 year 50% > 1 year	MPA and MPAG C0 (HPLC)		78 EMB (ISHLT classification)
				MPA C0 = 2.5±2.3		Grade < 2 (n = N/R)
	Cyclosporine or tacrolimus Corticosteroids	15±10 years (1 month – 33 years)		MPA C0 = 1.2±0.9 (p = 0.02)		Grade ≥ 2, (n = N/R) More frequent with MPA C0 ≥ 2.5 mg/L (p = 0.03)
[170]	MMF, 1.5 g BID (decreased based on clinical symptoms of toxicity)	20 patients	< 1 year (mean: 10.1 months)	MPA C0 (EMIT) Median: 1.51 (DF50 = 0.96–2.23)		147 EMB (ISHLT)
	Cyclosporine (possibly converted to tacrolimus)			Difference in cyclosporine and MPA C0 between categories of biopsies: ns		Grade 0, n = 54; grade 1A, n = 61; grade 2, n = 16, grade 3A, n = 16
	Prednisone			Median MPA C0: 1.76 (range: 0.496–7.65)		Patients without AR, n = 9
				Median MPA C0: 1.36 (range: 0.26–6.13); p = 0.015		Patients with AR, n = 11
[139]	MMF, 2 g/day D adjusted on C0: 2–4 mg/L	215 patients 36±14 yrs		MPA C0 (EMIT); 892 plasma samples (3 groups based on C0: < 2, 2–4, > 4, or 2 groups based on C0 < 2 and C0 ≥ 2, to compare the incidence of rejection)		Scheduled EMB Acute rejection if grade ≥ 3A (ISHLT)		50 patients had samples taken on more than one occasion
	Cyclosporine (89%) or tacrolimus (11%) D adjusted on C0 Prednisone (all)
	Rejectors: 2.8±0.8 g/day	Period I n = 104	< 6 months	No difference in mean C0 between patients with (C0 = 2.2±2.0 mg/L) and without (C0 = 2.3±1.7 mg/L) AR		No difference in AR incidence between the 3 C0 groups (p = 0.1): 7.8 to 14.9%
	Non rejectors: 2.8±0.7 g/day (ns)			Higher proportion of C0 < 2 in rejectors (n = 34/54) vs non rejectors (n = 194/401), p = 0.05
				Difference in cyclosporine or tacrolimus C0 between MPA C0 groups: ns Mean daily MMF dose to reach MPA C0 > 2, cyclosporine group vs tacrolimus group: ns When MPA C0 < 2, no difference on AR incidence in relation to therapeutic or subtherapeutic cyclosporine or tacrolimus C0 (14.4 vs 13.9%, ns) When MPA C0 > 2, AR incidence is significantly reduced if cyclosporine or tacrolimus C0 are therapeutic vs subtherapeutic (3.6 vs 15.4%, p = 0.002)
	Rejectors: 2.5±1.0 g/day	Period II n = 90	6–12 months	No difference in mean C0 between patients with (C0 = 2.0±1.5 mg/L) and without (C0 = 2.7±2.0 mg/L) AR		No difference in AR incidence between the 3 C0 groups (p = 0.15): 4.0 to 11.3%
	Non rejectors: 2.6±0.7 g/day (ns)			Higher proportion of C0 < 2 in rejectors (n = 9/14) vs non rejectors (n = 71/188), p = 0.05
	2.3±0.8 g/day 2.3±0.8 g/day (ns)	Period III n = 71	> 12 months	No difference in mean C0 between patients with (C0 = 2.4±2.0 mg/L) and without (C0 = 2.6±2.2 mg/L) AR		No difference in AR incidence between the 3 C0 groups (p = 0.13): 2.6 to 15.1%
				Comparable proportion of C0 < 2 in rejectors (n = 13/27) vs non rejectors (n = 98/208), p = 0.92
[129]	MMF Tacrolimus Prednisone (weaning at 6 months)	45 adults Primary OHT		MPA C0 (EMIT) Tacrolimus C0 (IMx)		EFFICACY Patient survival EMB (ARE, ISHLT classification)	TOXICITY GI toxicity Infections
	MMF D = 2 g/d (fixed) Tacrolimus D adjusted on C0	Phase I n = 15 51±11 years	696±62 days (606–790)	Mean MPA levels (Months 0–6)		Patient survival: 100%	GI toxicity: n = 6 (40%)
						Tacrolimus C0, patients with vs without AR: ns	Infections:
				3.6±0.4		0 ARE/patient, n = 5	Bacterial, n = 10 (66.7%) Fungal, n = 8 (53.3%) Viral, n = 9 (60.0%)
				2.2±0.4		1–2 ARE/patient n = 7 patients
				1.4±0.2		3 ARE/patient, n = 3
	MMF and tacrolimus D adjusted on C0	Phase II n = 30 54±9 yeras	436±88 days (175–562)	MPA C0 target: 2.5–4.5		27 patients: rejection-free 3 patients with AR had confounding factors	GI toxicity: n = 9 (30%) Infections: Bacterial, n = 17 (56.7%) Fungal, n = 5 (16.7%) Viral, n = 7 (23.3%)

AR: Acute rejection – ARE: Acute rejection episode – CF: Cystic fibrosis – D: Dose – f: Free fraction – EMB: Endomyocardial biopsy – GI: Gastro-intestinal – N/R: not reported.

2.2.3.1. C₀ monitoring

In a retrospective study conducted in 215 heart transplant patients who received MMF associated with either cyclosporine (n = 191) or tacrolimus (n = 24), the incidence of acute rejection was higher when patients had MPA C₀ < 2 mg/L (EMIT) compared to patients with MPA C₀ > 2 mg/L (14.9 vs 8.8% between 0 and 6 months post-transplant and 11.3 vs 4.2% between 6 and 12 months post-transplant, p = 0.05 for both).^[139] However, this difference was not further observed in patients beyond the first year post-transplantation. More importantly, patients who were within the therapeutic range for the associated calcineurin inhibitor in addition to MPA C₀ > 2 mg/L had lower rejection rates compared to patients who had only MPA C₀ > 2 mg/L (3.6 vs 15.5%, p = 0.002).

Forty-eight EMB were performed at the time of blood sampling in 26 heart transplant patients on MMF associated with cyclosporine or tacrolimus.^[163] In the cyclosporine group, the overall acute rejection incidence was approximately 20% with MPA C₀ values of 1.65±0.97 mg/L (cyclosporine C₀ < 175 μg/L in all samples), whereas in the tacrolimus group, the overall acute rejection incidence was approximately 60% (p = 0.02) with MPA C₀ values of 2.86±2.07 mg/L, (tacrolimus C₀ < 10 μg/L in 52% of samples). These results are obviously very surprising and might be due to the dispersion of the individual exposure to both MPA and calcineurin inhibitor, even though the difference in MPA levels between rejectors and non-rejectors was not statistically significant.

In another retrospective study of 147 EMB from 20 patients in the first year after transplantation, the mean MPA C₀ (assayed by EMIT) was significantly lower in patients with acute rejection (n = 11, C₀ = 1.36 mg/L) than in those without (n = 9, C₀ = 1.76 mg/L, p = 0.015).^[170]

In 26 heart transplant adults and children aged 1 month to 33 years who received MMF associated with either cyclosporine (n = 8 children + 10 adults) or tacrolimus (n = 8 children), lower MPA C₀ were associated with EMB grade ≥ 2 (ISHLT classification) compared to grade < 2 (1.2±0.9 vs 2.5±2.3 mg/L, p = 0.02).^[136] and no patient experienced grade 2 or 3A rejection when MPA C₀ was > 2.5 mg/L. Of note, issues regarding the heterogeneity of the population in this study have already been addressed in the PK section.

A statistically lower MPA C₀ (assayed by EMIT) was also reported in lung transplant patients with (n = 14, 28 acute rejection episodes) compared to patients without (n = 16) acute rejection (2.11±0.08 vs 2.71±0.11 mg/L, p < 10⁻⁵).^[142]

A prospective study conducted in 45 patients was organized in 2 phases:^[129] during the first phase, 15 patients received fixed-dose of MMF (2 g/day, resulting in 5/15 patients without acute rejection; during the second phase, 30 patients received MMF at doses adjusted on C₀ with a target of 2.5 to 4.5 mg/L, resulting in 27/30 rejection-free patients, which, even if not statistically significant and despite the sequential design of the study, advocates for further trials.

2.2.3.2. AUC monitoring

The relationship between MPA AUC_0-12 and acute rejection was demonstrated in renal transplant recipients^[120] and confirmed by a recent study which demonstrated that TDM of MMF based on MPA AUC_0-12 (estimated using Bayesian forecasting) compared to fixed-dose MMF resulted in a significant decrease of the acute rejection incidence:^[16] patients who benefited from MPA TDM had fewer acute rejection episodes compared to patients of the fixed dose group (12% vs 30%, p = 0.01). Interestingly, a more than 10-fold inter-individual variability of MPA AUC values was reported in heart transplant patients receiving a fixed dose of MMF.^[133]

In a cohort study, thirty-eight cardiac transplant recipients were divided into 3 groups based on the rejection severity on their biopsy (grade 0, n = 22; grade 1, n = 13; grade 2/3, n = 3).^[132] Patients with grade 2 or 3 rejection had a lower mean total MPA AUC_0-12 (26.1±6.6 vs 42.8±14.0 h.mg/L, p < 0.08) and unbound MPA AUC_0-12 (0.49±0.11 vs 0.81±0.25 h.mg/L, p < 0.05) compared with patients without rejection (EMIT assay). However, due to the very limited number of patients with grade 2/3 rejection in this study and the already mentioned methodological limitations (estimation algorithm developed for renal transplantation and HPLC assay), this finding is to be used with caution.

Some authors hypothesized that high concentrations of MPA and AcMPAG may account for adverse effects such as gastro-intestinal disturbances and bone marrow suppression.^{[150,164,171–173]} According to other authors, the occurrence of GI toxicity is probably not related to systemic exposure to MPA or MPAG, but rather to high concentrations of MPA and/or its metabolites in the gut.^[94,133] No study except for one^[129] addressed the relationship between exposure to MPA and toxicity, more particularly gastro-intestinal toxicity, in heart or lung transplantation. In this study performed in 45 heart transplant recipients, the incidence of gastro-intestinal toxicity was 40% in patients who received fixed dose of MMF (2 g/day), and 30% in patients who benefited from MMF dose adjustment on C₀ (target: 2.5–4.5 mg/L). Unfortunately, no formal statistical comparison was made between the 2 groups, preventing us from drawing any conclusion from this observation.

To conclude, though prospective clinical validation is still missing, there is much indirect evidence that optimized exposure to MPA is crucial for successful thoracic transplantation and that MMF therapy can be improved by TDM.

3. mTOR inhibitors

Sirolimus and everolimus display a distinctive mechanism of action from that of calcineurin inhibitors. They form a complex with FKBP12, which inhibits the mTOR (mammalian target of rapamycine) involved in the proliferation of lymphocytes, blocking DNA and protein synthesis and consequently the proliferation of IL2 activated B and T cells. mTOR inhibitors reduce T-cell activation in cell cycle stages later than calcineurin inhibitors: their mechanism of action is downstream of and complementary to that of calcineurin inhibitors. ^{[1,9,174,175]}

3.1. Sirolimus

Sirolimus (rapamycin, Rapamune^®) has been used mainly in renal transplantation, usually in combination with a calcineurin inhibitor or MMF and a corticosteroid, or as a calcineurin-sparing agent (because it allows the associated calcineurin inhibitor to be discontinued or its dose to be decreased). In thoracic transplantation, although used off-label, sirolimus has been of great interest as a rescue therapy when other immunosuppressants are inefficient or contra-indicated and in patients with impaired renal function.^[9,175] Sirolimus was demonstrated to display a protective effect against the development of bronchiolitis obliterans in stable lung transplant patients.^[1] However, it has been associated with interstitial pneumonia^[176] and bronchial anastomotic dehiscence.^[177] Consequently, in lung transplantation, it has been recommended not to introduce sirolimus before the third month post-transplantation. In heart transplantation, the safety of early CNI withdrawal has been questioned because of an associated increase in acute rejection rate, an observation that led to the discontinuation of a randomized trial of early CNI replacement by sirolimus. According to Zuckermann et al, CNI withdrawal should not be attempted until after the first year post-transplantation.^[178]

These limitations and side-effects may explain why the use of sirolimus in thoracic transplantation has so far been limited.

3.1.1. Pharmacokinetics

The majority of PK studies of sirolimus were performed in healthy volunteers and kidney transplant recipients,^[94] and we identified only one in heart transplantation.^[179] This brief review will thus be mainly based on the PK and TDM of sirolimus in renal transplantation, as already reviewed in more detail in two articles.^[175,180]

The standard sirolimus dosing strategy in renal transplantation consists of a loading dose of 6 mg/day, followed by a maintenance dose of 2 mg/day, with dose adjustment based on target C₀ values. In patients with high rejection risks and in heart transplant recipients, the loading dose is 15 mg/day and the maintenance dose 5 mg/day.^{[175,181–183]}

In kidney transplant recipients, sirolimus exhibits wide inter- and intra-individual PK variability, like calcineurin inhibitors do.^[52,180]

3.1.1.1. Absorption

Sirolimus is rapidly absorbed (t_max ranging between 0.7 and 3 hours), with an apparent oral bioavailability in association with cyclosporine of approx. 15% and a variable C_max ranging between 14 and 355 μg/L.^{[52,94,175,180]} Sirolimus displays an intensive intestinal and hepatic metabolism by CYP3A4/CYP3A5 and is a substrate of P-gp, accounting for its low bioavailability and high PK variability.^[94,144,175]

In a study conducted in 22 healthy adult volunteers, high fat meal decreased sirolimus AUC by 35%.^[184] The drug should thus be administered consistently with or without food in order to reduce fluctuations in drug exposure.^[180]

3.1.1.2. Distribution

In whole blood, sirolimus is extensively distributed in cellular elements (much more so than calcineurin inhibitors), mainly in erythrocytes (95%), lymphocytes (1%) and granulocytes (1%). The remaining 3% are found in plasma, where only 4% are bound to soluble plasma proteins, and 40% to lipoproteins, although the latter ratio goes up with increasing sirolimus concentrations.^[94,180] Its high hydrophobic properties result in a wide distribution in lipid membranes of body tissues and confer sirolimus a large V_d/F (5.6–16.7 L/kg).^[180]

In a popPK study performed in 31 de novo heart transplant recipients using a one-compartment PK model with first-order absorption and elimination, the estimated V_d/F ranged approximately between 1,200 and 1,600 L, which is slightly higher than values reported in renal transplant recipients.^[179]

3.1.1.3. Metabolism

Sirolimus, like cyclosporine and tacrolimus, is metabolized in the liver by the CYP3A4, and to a lesser extent CYP3A5 isoenzyme.^[52,94] The large inter-individual variability of sirolimus PK may be related to the variability in its metabolism,^[180] which consists mainly of O-demethylation and/or hydroxylation.^[175] Over 16 metabolites have been identified in the bile of rats receiving sirolimus.^[180] In patients’ whole blood, concentrations of metabolites exceed those of the parent drug,^[94] but there is no evidence that they display significant biological activity (< 10% of that of sirolimus at most).^[180] An in vitro study with human liver microsomes showed that CYP3A5 exhibited a lower activity than CYP3A4 for sirolimus metabolism, but as CYP3A4 was more inhibited by cyclosporine, cyclosporine could unveil the influence of CYP3A5 and of its *1/*3 polymorphism on sirolimus disposition.^[144] However, the CYP3A5*1/*3 polymorphism was shown to have a dramatic impact on the sirolimus CL/F and C₀ actually achieved over the first 1–3 months post-transplantation despite TDM in renal transplantation, probably due to its higher impact on sirolimus intestinal metabolism and first pass effect.^[185]

3.1.1.4. Elimination

Sirolimus is characterized by a long elimination half-life, varying between 44 and 87 hours, which allows once daily administration.^[94,180] Steady-state is thus normally reached on average 6 days after the initiation of therapy, which justifies the use of a loading dose to achieve steady-state concentrations rapidly. The primary route for the clearance of sirolimus is biliary, over 90% of the metabolites being recovered in the faeces.^[180] A large intra-patient (CV = 86%) and inter-patient variability (CV = 90%) on CL/F has been reported in stable renal transplant recipients, with 4- to 8-fold differences.^[175,180]

The only published study on the PK of sirolimus in heart transplantation aimed to build a popPK model to estimate sirolimus CL/F and its intra- and inter-patient variability.^[179] Thirty-one de novo, adult heart transplant recipients between 4 days and 5 years post-transplantation, on sirolimus, cyclosporine and oral prednisone were included. Sirolimus dose was adjusted on C₀ (target: 8–18 μg/L, assay: LC-MS/MS), and PK modelling was performed using NONMEM. As the majority of blood samples were collected towards the end of the dosing interval with a median of 20 hours, no data was available during the absorption phase. Consequently, a one compartment PK model with first-order absorption and elimination was used, and the rate constant of absorption k_a was fixed at a value of 0.752 L/h. In this population, the maximum a posteriori Bayesian estimate of CL/F was 7.1 L/h (95%CI = 6.75–7.41 L/h), which is similar to values reported in renal transplantation. A 21% increase in sirolimus CL/F was observed for every 100 mg increase in the cyclosporine daily dose. Patients with primary diagnosis of ischemic heart disease had a 62% higher CL/F compared to others, and patients with hypertriglyceridemia (TG ≥ 2 mmol/L) had a 38% lower CL/F compared to patients with TG < 2 mmol/L, though the meaning of these findings is not quite clear.

3.1.1.5. Drug-drug interactions

Concomitant administration of sirolimus with cyclosporine microemulsion (but not the oil-based formulation) results in a 50–80% increase in sirolimus AUC because of the inhibition of CYP3A and P-gp by cyclosporine, with no modification of cyclosporine AUC. Moreover, a synergistic PD interaction between cyclosporine and sirolimus has been widely reported.^[52,180] These are the reasons why sirolimus has been largely used as a cyclosporine-sparing agent.

As a substrate of CYP3A4 and P-gp, sirolimus has a potential for drug interactions similar to that of cyclosporine.^[52] In renal transplantation, the association of sirolimus with diltiazem and ketoconazole resulted in a 60 and 990% increase of sirolimus AUC respectively, whereas the association with rifampicin resulted in an 82% decrease of sirolimus AUC.^[180,186] On the contrary, in a study in 31 de novo heart transplant patients, sirolimus C₀ was lower in patients during diltiazem administration (14.1 μg/L, range: 9.7–21.9) than at other times (18.9 μg/L, range: 12.6–26.0; p < 10⁻⁵),^[179] which is not consistent with the known inhibitory effect of diltiazem on P-gp and hepatic metabolism, nor with the calculated sirolimus clearance in the same paper, which was lower under diltiazem. This might be due to an error in C₀ reporting in this paper.

Combination of cerivastatin with cyclosporine + sirolimus may result in a significant increase in cerivastatin exposure, with possible HMG-CoA reductase inhibitor-associated toxicity.^[175]

3.1.2. Therapeutic drug monitoring

A wide intra-individual variability of sirolimus PK has been reported in renal transplant patients, with CV = 42–82% on observed and dose-adjusted C_max, C₀ and AUC values.^[180,185] High variability was similarly observed in 31 de novo heart transplant patients, with the C₀ ranging between 0.2 and over 70 μg/L, despite targeting 8–18 μg/L,^[179] and a poor correlation between the dose and C₀ (r² = 0.162, n = 524, p < 0.001). The wide range of C₀ in patients receiving the same dose is due to the variability in drug absorption and clearance.^[180]

On the other hand, better though variable correlation was reported in renal transplant patients between steady-state C₀ and AUC (assay: LC-MS/MS), with r² = 0.85,^[52,180] or r² = 0.43^[185].

The relationship between sirolimus C₀ and efficacy has been evaluated in renal transplantation.^[175] Current TDM strategies are based on C₀, but appropriate C₀ for the prevention of acute rejection depends on the concomitant immunosuppressive regimen and would thus probably need to be refined. In renal transplant patients treated with sirolimus, cyclosporine and prednisone, C₀ < 5 μg/L (assays: LC/MS, LC/UV) was associated with an increased risk for acute rejection episodes, whereas C₀ > 15 μg/L was associated with a higher risk of adverse events (anemia, leukopenia, decreased platelet count, hypertriglyceridemia).^[180] In association with cyclosporine, the following target values were proposed for sirolimus C₀: 5 to 10 μg/L for 30% reduction in usual cyclosporine exposure, 10 to 15 μg/L for 60% reduction in usual cyclosporine exposure, and ≥ 20 μg/L for no cyclosporine exposure.^[52] In CNI-free studies in renal transplantation, where the C₀ target was 30 μg/L before M2 and 15 μg/L after M2, 41% and 27.5% acute rejection episodes were reported in patients on sirolimus associated with azathioprine and prednisone and in those on sirolimus associated with MMF and prednisone, respectively.^[180]

As sirolimus has a long elimination half-life, the following schedule for C₀ monitoring was proposed: initially not more than once every 4 days, then once a week during the first month and once every two weeks during the second month. Moreover, sirolimus dose adjustments should be based on C₀ obtained more than 5–7 days after a dosage change or initiation of therapy.^[175]

Sirolimus monitoring is a regulatory requirement in Europe. However, in a recent review, routine TDM of sirolimus was not deemed necessary in all patients after the first 2 months of dose titration, but only in case of suspected toxicity or non-compliance, and in some high-risk categories, identified as: young children (5–11 years), patients with hepatic impairment, patients on enzyme inhibitors or inducers, patients at high risk of rejection, patients on a CNI-sparing strategy.^[175]

In conclusion, sirolimus pharmacogenetic-PK studies are required in heart or lung transplantation, in order for this drug to be used safely in these populations. Moreover, although there appears to be a relationship between sirolimus C₀ and clinical outcomes, the value of sirolimus TDM needs to be tested in the clinical setting,^[52] and the demonstration of the superiority of sirolimus TDM over fixed dose regimens requires controlled trials.^[175]

3.2. Everolimus

Everolimus (SDZ RAD, RAD, Certican^®) was derived from sirolimus in order to improve its PK, and in particular to increase its oral bioavailability,^[174,187] and decrease its half-life. It is currently the only mTOR inhibitor with regulatory approval in heart transplant recipients, as it has been shown to reduce significantly the incidence of acute rejection, CAV, graft loss, and death.^{[9,178,187,188]} To date, there is limited evidence in favour of the use of everolimus in lung transplantation. Given the fibroproliferative process underlying the BOS and everolimus general antiproliferative effect, everolimus may prevent the onset of BOS after lung transplantation.^[174] Significant clinical benefit of everolimus vs AZA on pulmonary function and BOS was evidenced in stable lung transplant recipients.^[189]

3.2.1. Pharmacokinetics

The few PK studies of everolimus performed to date have been conducted in healthy volunteers and kidney and liver transplant recipients. Only exposure indices have been reported in heart and heart-lung transplantation.

3.2.1.1. Absorption

Everolimus is characterized by a low oral bioavailability of around 16%, with t_max ranging between 0.5 and 4 h depending on the type of patients (healthy volunteers, solid organ transplant patients).^[187,190] Everolimus, like sirolimus, is a P-gp substrate, which probably affects its absorption.^[175,190] As high-fat meals seem to influence its PK, it was recommended to take everolimus consistently with or without food, in order to reduce fluctuations in drug exposure.^[190]

The dispersible tablets tested during a pediatric development program in 19 renal transplant children (aged 2–16 years), showed a 10% lower bioavailability relative to the conventional ones.^[191]

3.2.1.2. Distribution

In whole blood, over 75% of everolimus is partitioned into erythrocytes.^[187,190] In a popPK analysis of everolimus (one-compartment model) in 673 kidney transplant recipients, V_d/F was estimated to be 110±5 L.^[192]

3.2.1.3. Metabolism

Everolimus is metabolized extensively in the gastro-intestinal tract and liver by CYP3A4 (mainly), 3A5 and 2C8, so the large inter-individual variability in everolimus bioavailability and biotransformation is related to the variations in activity of these enzymes and P-gp.^{[175,187,190]} The major metabolic pathways are thought to be hydroxylation and demethylation, leading to the production of at least 11 metabolites, whose immunosuppressive or toxic activities are unknown.^[190] Patients with hepatic impairment were shown to have significantly lower everolimus CL/F, with longer T_1/2 (by 84%) and higher AUC (by 115% higher) compared to healthy subjects. As a consequence, it was recommended to reduce the everolimus dose by half in such patients.^[190]

3.2.1.4. Elimination

About 98% of everolimus is excreted in bile as metabolites and only 2% is eliminated in urine.^[190] Everolimus displays a shorter half-life than sirolimus, ranging between 18 and 35 hours in kidney and liver transplant patients, which allows twice daily administration and a steady-state to be achieved more quickly.^[187,190] In healthy volunteers, CL/F and T_1/2 were 19.7±5.4 L/h and 32.2±6.1 h, respectively.^[190] As already mentioned, no clearance or half-life values have been reported so far in heart or lung transplant recipients.

3.2.1.5. Drug-drug interactions

Everolimus and cyclosporine are given as a combination therapy because of their synergistic interaction.^[9,190] As they are both substrates for CYP3A and P-gp (cyclosporine inhibits whereas everolimus induces P-gp), they may interfere with each other’s absorption or elimination.^[193] Any alteration in P-gp and inhibition of CYP3A4 metabolism may result in the accumulation of everolimus and/or cyclosporine.^[187] In a PK study conducted in 634 heart transplant patients on everolimus (n = 420) or AZA (n = 214) associated with cyclosporine and corticosteroids, similar cyclosporine C₀ were achieved with lower doses of cyclosporine in patients on everolimus vs those on AZA (p = 10⁻⁴), resulting in higher dose-normalized cyclosporine C₀ in patients on everolimus vs those on AZA (p = 10⁻⁴).^[193]

Similar results were reported in a PK study conducted in 213 patients on everolimus (n = 101) or AZA (n = 112) associated with cyclosporine and corticosteroids, with cyclosporine C₀ during the first month of 220±102 μg/L in patients on everolimus vs 176±76 μg/L in patients on AZA (p < 0.02).^[194]

A PK conversion study (switch from AZA to everolimus or from cyclosporine to tacrolimus) was recently performed in 12 adult heart transplant patients (2 groups of 6 patients, aged 18–65 years) over 3 months post-transplantation.^[195] Two full PK profiles per patient were collected, before and 6–8 weeks after conversion. Everolimus, cyclosporine and tacrolimus were assayed by LC-MS/MS. The first group received tacrolimus adjusted on C₀ (target: 6–10 μg/L), and was converted from AZA to everolimus adjusted on C₀ (target: 3–8 μg/L), showing no significant difference in the PK of tacrolimus before and after the conversion. The second group received everolimus adjusted on C₀ (target: 3–8 μg/L) and was switched from low-dose cyclosporine (target C₀: 60–80 μg/L) to tacrolimus (target C₀: 6–10 μg/L). A significant decrease (p < 0.05) of everolimus exposure before and after conversion from cyclosporine to tacrolimus was found, on C₀ (4.2±1.3 vs 2.3±1.2 μg/L), C_max (9.1±1.9 vs 5.9±1.1 μg/L) and AUC_0-12 calculated using the linear trapezoidal rule (64.2±6.8 vs 33.7±9.7 h.μg/L). In conclusion, no significant effect of everolimus on the PK of tacrolimus was evidenced, but the PK interaction between everolimus and cyclosporine led to 1.9-fold and 8-fold higher everolimus AUC_0-12 and C₀ on concomitant cyclosporine compared with tacrolimus.

In healthy volunteers, rifampicin was reported to increase significantly everolimus CL/F (+172%), and shorten T_1/2 from 32 to 24 hours (p = 10⁻⁴), with a significant decrease of its AUC (−63%, p = 10⁻⁴).^[196] In de novo renal transplant patients, erythromycin and azithromycin resulted in a significant decrease in everolimus CL/F (−22 and −18%, respectively). Fluconazole had no significant influence, but itraconazole induced a 74% reduction in everolimus CL/F.^[192]

Analysis of PK profiles of everolimus in 13 lung and heart-lung transplant recipients found 2.3-fold higher everolimus C₀ in patients on azole antifungal drugs compared to others (p < 10⁻⁴).^[194] Fluconazole appeared to have minimal influence on everolimus exposure whereas itraconazole and ketoconazole had a stronger effect.

3.2.1.6. Special populations

3.2.1.6.1. Paediatric transplantation

We found no published study in heart or lung pediatric patients. A review published in 2004 analyzed two studies performed in pediatric renal transplant patients on everolimus associated with cyclosporine and prednisone.^[190]

In the first study (n = 19 stable transplant recipients), V_d/F and CL/F were positively correlated with age, weight and body surface area (BSA) (p < 0.001) and were significantly lower to those of adults, whereas elimination half-life was similar, regardless of age, and comparable to that of adults. The authors recommended adjustment of everolimus doses to the bodyweight in children.^[197]

In the second study (n = 19 de novo transplant recipients followed for 6 months after transplantation), a positive correlation was also found between CL/F and age, weight and BSA. Intra- and inter-patient variability on AUC was 29 and 35%, respectively. The authors recommended adjustment of everolimus doses to the body surface area in children.^[198]

3.2.1.6.2. Cystic fibrosis

One study reported the PK of single-dose everolimus in patients with CF, using a non-compartmental approach.^[199] This was a phase I multicenter, randomized, double-blind, crossover study in 19 lung and heart-lung transplant recipients (7 CF, 12 non-CF), beyond 3 months post-transplantation treated with cyclosporine, azathioprine and prednisone. Each patient received everolimus at a low-dose (0.035 mg/kg without exceeding 2.5 mg) and at a high-dose (0.10 mg/kg without exceeding 7.5 mg), in random order and separated by a wash-out period. Everolimus was assayed using LC-MS/MS, and AUC_0-τ was calculated using the trapezoidal rule. Dose-normalized C_max was significantly lower in CF patients vs non-CF patients (low-dose everolimus: 12.4±3.9 vs 13.8±3.1 μg.L⁻¹; high-dose everolimus group: 10.2±2.3 vs 12.8±2.5 μ.L⁻¹; p = 0.03 for patients pooled across everolimus dose levels), whereas everolimus AUC/dose (which is the inverse of CL/F) did not differ significantly between the patients with and those without CF. T_1/2 was similar in all groups, ranging from 25 to 29 hours.

Similarly, the analysis of everolimus PK profiles in 13 lung and heart-lung transplant recipients showed no significant difference in exposure between CF and non-CF patients.^[194]

3.2.2. Therapeutic drug monitoring

Everolimus is characterized by a narrow therapeutic index.^[187,190] Moreover, data from various solid organ transplant populations (adult renal, cardiac and liver transplantation and pediatric renal transplantation) suggests that during the first year post-transplantation, there is a good relationship between everolimus concentrations and the clinical outcomes (efficacy, thrombocytopenia).^[190,200] Consequently, TDM has been recommended for everolimus.^[187]

In heart transplantation, two PK studies were conducted in the patients of study B253,^[188] a prospective randomized double-blind efficacy trial in which 634 heart transplant patients (aged 52±11 years) on cyclosporine and corticosteroids were randomized to receive either 0.75 or 1.5 mg of everolimus BID (n = 209 and 211), or azathioprine (n = 214). Based on exposure-efficacy and exposure-safety analyses, everolimus C₀ of 3 to 8 μg/L was shown to ensure an optimal benefit/risk ratio.^[178]

The first study focused on the PK of everolimus (assay: ELISA) during the first 6 months post-transplantation.^[193] A measurement of C₀ was performed in all patients and AUC_0-12 estimated in 55 of them, using the trapezoidal rule based on 5 samples drawn over 8 hours (C₁₂ was extrapolated from C₅ and C₈ by log-linear regression). No statistically significant difference in dose-normalized C_max or AUC was found between the 0.75-mg BID and the 1.5-mg BID dose groups, showing dose proportionality. A good correlation was found between C₀ and AUC (r² = 0.81, p < 0.001). Intra-individual variability on C_min, C_max and AUC was 38%, 30% and 30%, and inter-individual variability 40%, 33% and 37%, respectively. Sixty-nine and 80% of patients with C₀ comprised between 3.6 and 5.3 μg/L and between 5.4 and 7.3 μg/L respectively, remained rejection-free vs 65% of patients with C₀ ≤ 3.5 μg/L, and cox proportional hazard analysis showed that the relative risk for biopsy-proven acute rejection was 2.4 for C₀ < 3 μg/L vs C₀ in the 3–7 μg/L range (p < 10⁻⁵). However, ROC analysis showed that C₀ = 3.4 μg/L was the best one, able to distinguish rejectors from non-rejectors, yielding 75% sensitivity and 47% specificity. These variable thresholds and targets, apparently chosen for statistical reasons are puzzling and probably weaken the conclusions of this study. No significant relationship was evidenced between exposure and the level of triglycerides or serum creatinine, but there was a significant correlation between C₀ and thrombocytopenia (r² = 0.85, p = 0.03).

The second study evaluated, throughout the first year post-transplantation, the relationship between everolimus C₀ and cyclosporine C₀ on the one hand and efficacy (biopsy-proven acute rejection, BPAR) and safety (renal function, serum lipids, platelet counts) on the other.^[201] Everolimus C₀ were stable in the first year post-transplantation and averaged 5.2±3.8 and 9.4±6.3 μg/L in patients treated with 1.5 and 3 mg/day, respectively. The same correlation as in the previous study was found between C₀ and AUC (r² = 0.81). The incidence of biopsy-proven acute rejection (BPAR ≥ grade 3A based on ISHT classification) was 44% in patients with C₀ < 3 μg/L, 24% in patients with C₀ = 3–8 μg/L, and 17% in patients with C₀ ≥ 8 μg/L, leading to a relative risk of BPAR of 2.5 (p = 10⁻⁴) when C₀ < 3 μg/L vs C₀ = 3–8 μg/L. The relative risk of BPAR when C₀ = 3–8 μg/L vs C₀ ≥ 8 μg/L was not significant. No effect of everolimus exposure on renal events (serum creatinine > 200 μmol/L) was found, but hypercholesterolemia, hypertriglyceridemia and thrombocytopenia tended to increase with higher everolimus C₀.

A PK study of everolimus was performed in lung and heart-lung transplant recipients (aged 46±14 years, 3–36 months after transplantation at enrollment, 23% CF) enrolled in a 3-year randomized, multi-center, double-blind, parallel group study comparing everolimus (n = 89) and azathioprine (n = 100) as part of a cyclosporine-based immunosuppressive regimen.^[194] Patients received an initial everolimus dose of 1.5 mg BID, stabilized at 1.2±0.4 mg BID between M3 and M12. Everolimus concentration time profiles were obtained in 13 patients (assay: LC-MS). AUC_0-12 was calculated (method: N/R) using 5 blood samples drawn between 0 and 8 hours and considering C₁₂ equal to C₀ at steady-state. T_max ranged between 1 and 5 hours, and dose-normalized C₀ were regarded as stable throughout the observation period, with intra- and inter-individual CV of 35 and 37%, respectively (C₀: 7.3±2.6 μg/L at M2, 9.2±4.7 μg/L at M6, 7.6±3.1 μg/L at M12). Everolimus AUC was 130±31 h.μg/L at M2 (n = 13), 144±52 h.μg/L at M6 (n = 9) and 138±32 h.μg/L at M12 (n = 7). Over the range of everolimus C₀ measured throughout the study (2–36 μg/L), the exposure-efficacy relationship (efficacy criteria: FEV1, BOS and BPAR) appeared to be in the near-maximal plateau region. A significant relationship was found between everolimus C₀ over the first 2 months and the incidence of cholesterol > 6.5 mmol/L (r² = 0.95, p = 0.025) as well as the incidence of triglycerides > 2.9 mmol/L (r² = 0.96, p = 0.022). A significant relationship was also found between everolimus C₀ during the first month and platelet count < 100 G/L (r² = 0.93, p = 0.038). A tendency was found between renal impairment and the increase in everolimus C₀ (p = 0.051), which might be attributed to the synergistic nephrotoxic effect of cyclosporine associated with everolimus, as already described for sirolimus in vitro.^[202,203] From data obtained in this population, the everolimus C₀ range of 3 to 12 μg/L was considered as safe and well tolerated. Unfortunately, the number of AUC values available in this study was too small to be able to study the AUC – effect relationships.

In conclusion, clinical evidence suggests that TDM of everolimus is probably beneficial in thoracic transplantation, C₀ as measured by LC-MS/MS being a good predictor of clinical efficacy. The recommended minimum effective concentration for everolimus when used in conjunction with a cyclosporine-corticosteroid immunosuppressive regimen in heart and lung transplant recipients is 3 μg/L. The upper bond has not been formally established because patients have tolerated levels up to 15 μg/L without damage. These levels are usually achieved within 1 to 2 weeks, and everolimus dose adjustment should always be based on blood trough levels obtained at least 4–5 days after the previous dose change. It is noticeable that correlation coefficients between C₀ and AUC are suboptimal, which suggests that AUC monitoring might represent further improvement in everolimus dose personalisation, but almost everything remains to be done, particularly AUC-effects observational studies and C₀/AUC comparative trials.

In heart transplant patients, cyclosporine dose reduction coupled with everolimus TDM could optimize immunosuppressive efficacy while reducing treatment-related toxicity.^{[178,187,194,200]} The use of everolimus in CNI-free regimens, maybe with higher C₀ (or AUC) targets to maintain immunosuppressive efficacy has not been studied extensively.^[178]

4. Discussion

The increasing success of thoracic transplantation is largely attributable to the development of effective immunosuppressive strategies. However, there is still a high rate of mortality on the mid- and long-term, as the percentage of 5-year survival is still only 70% after heart transplantation and 50% after lung transplantation.^[2,3] Substantial improvement of the long-term survival after thoracic transplantation may be obtained by adapting to some extent the immunosuppression to the recipient’s rejection risk factors. In other solid organ transplantations, TDM derived from PK studies has been demonstrated to be crucial to improve patients’ outcome, by targeting individualized doses of different immunosuppressants.^[11] An overview of the PK and the clinical evidence in favour of the TDM of immunosuppressants in thoracic transplantation has been lacking.

4.1. Pharmacokinetics

Like in other solid organ transplant populations, the pharmacokinetics of the immunosuppressive agents in thoracic transplant recipients is characterized by wide inter- and intra-individual variability. A number of studies have been performed in order to compare the PK of Neoral^® to that of Sandimmune^® in heart and, to a lesser extent, in lung transplantation. Indeed, the old oil-based cyclosporine formulation Sandimmune^®, which was characterized by erratic PK and consequently random efficacy was replaced in the mid 90’s by the microemulsion Neoral^®, characterized by less variability and consequently a more constant efficacy. Little is known about the pharmacokinetics of MPA and tacrolimus in heart and lung transplant recipients, even though both have been available for around 15 years. Their introduction into the immunosuppressive armamentarium occurred progressively in both heart and lung transplantation, based on clinical experience, though in most countries tacrolimus and MMF have still not been approved by health authorities for lung transplantation. Given their increased use in thoracic transplant recipients, their PK deserves to be explored more in depth. Moreover, with the recent approval of another formulation of tacrolimus (Advagraf^®) by the EMEA, further investigation of its PK will also be necessary. Finally, there are hardly any studies on the pharmacokinetics of the mTOR inhibitors sirolimus and everolimus after thoracic transplantation, probably because they have been developed and released more recently. Sirolimus use in heart and lung transplantation has been questioned due to a poor benefit/risk ratio with the prescription schemes proposed. To our knowledge, the vast majority of studies on the PK of sirolimus were performed in healthy volunteers and in kidney transplant recipients, while its PK in thoracic transplantation was studied by only one team.^[179] There are only few studies of the PK of everolimus in thoracic transplantation, as it is the drug most recently developed and introduced in the maintenance immunosuppressive strategy. Its anti-proliferative activity slowing down the process of cardiac vessels intima thickening, a major mechanism of cardiac allograft vasculopathy, makes it a very promising molecule.^[204–207] Moreover, its synergistic immunosuppressive activity with cyclosporine allows the use of low doses of cyclosporine and makes it a key drug in calcineurin inhibitor-sparing strategies, which is of major interest in patients with renal dysfunction, but its potential synergistic effect on cyclosporine nephrotoxicity reinforce the need for combined pharmacokinetic studies in heart and lung transplant recipients.

4.2. Therapeutic drug monitoring

4.2.1. Evidence-based TDM

TDM comprises the measurement and interpretation of drug exposure and aims at assisting physicians in the choice of drug dose for an individual patient. The state of the art of evidence-based TDM should be consistent comparative, randomized clinical trials showing the superiority of TDM vs fixed dose, on hard clinical outcomes (morbidity, mortality, graft function). Despite the paucity of such trials (althought there might be a publication bias by which small studies with negative outcome are less likely to be published than studies of similar size but with a positive outcome), the clinical benefits of TDM have been validated by usage and are now well recognized in most situations. In clinical practice, TDM is even mandatory for calcineurin inhibitors and mTOR inhibitors due to the poor relationship between the dose given and the systemic concentrations achieved. Consequently, one may question the actual necessity of imposing upon the medical staff and patients similar clinical trials with different graft types, when a clear effect has already been demonstrated, for instance in renal transplantation. Indeed, once the clinical usefulness of TDM has been demonstrated in a given graft type, it might be acceptable to extrapolate the results to the other graft types, provided a case-control or a cohort study (regarded as of a lower evidence level) confirms a good exposure-outcome relationship. A good example for that would be MMF, for which no clear correlation has been established between dose and MPA plasma levels, much like with calcineurin inhibitors and mTOR inhibitors, and which probably deserves to be monitored as much as the other molecules.

The gold-standard exposure index is in most cases the full AUC, but when it is not easy to measure, a surrogate marker (abbreviated AUC, AUC calculated using sparse sampling algorithms or Bayesian estimators, single concentrations) can be proposed based on tight correlation with the former. These correlation studies are indeed of the lowest evidence level but can be useful to evaluate the pertinence of a simplified TDM strategy.

4.2.2. Is there clinical evidence for TDM of the immunosuppressants after thoracic transplantation?

TDM of cyclosporine has been recognized as an essential tool in the management of allograft transplant recipients and has been performed routinely in transplantation for almost 20 years.^[52,53] However, much of the information available has been obtained in renal and liver transplantation, where there is still no clear consensus about the optimal monitoring method. We found very little data on the practices and outcomes of cyclosporine monitoring after thoracic transplantation. Various observational studies sought for a relationship between C₀ and clinical outcome, with sometimes positive^[28,74–79] and sometimes negative^[20,54,80] results. Similarly, a number of authors found a relationship between C₂ and clinical outcomes and thus advocated for C₂ monitoring,^{[10,56,84–88]} while such a relationship could not be evidenced in a number of other studies.^{[20,22,28,51,89–92]} The superiority of C₂ over C₀ monitoring in terms of clinical outcome has been demonstrated in few studies, either in terms of efficacy, or in terms of toxicity (nephrotoxicity, infections).^{[10,55,79,86]} There is no published consensus to date for C₂ target levels after heart or lung transplantation, although tentative targets have been suggested based upon retrospective or single-center experience on small numbers of adult patients.^{[10,23,56,89]} Our literature search allowed us to identify only two studies evaluating the potential relationship between AUC and clinical outcomes in heart transplant recipients, with diverging results,^[54,55] and none in lung transplant recipients. A few methods have been proposed to estimate the AUC using a limited number of samples, three of which seem clinically applicable, one based on blood samples taken at 1 and 4 hours and multilinear regression,^[21] the other two requiring 3 samples (0, 1 and 3 hours) and Bayesian estimation.^[31,32] However, to date, the cyclosporine AUC target values have not been established and no studies in heart or lung transplantation have demonstrated the superiority of AUC management over C₀ or C₂ monitoring in terms of clinical outcomes.

Tacrolimus dosage after thoracic transplantation is generally adapted to C₀ levels, mainly based on the results of clinical studies in kidney or liver transplantation. Studies of the relationship between C₀ and AUC_0-12 in thoracic transplant recipients have reported a wide range of correlation coefficient values.^{[103,104,109,111]} Some authors proposed to monitor tacrolimus C₃^[110] or C₄^{[104,105,109]} levels because of their good (and sometimes better) correlation with AUC_0-12. AUC_0-4 was proposed as a surrogate exposure index for heart^[109] and heart-lung^[110,112] transplant patients. Sparse sampling strategies have been proposed for the determination of the tacrolimus inter-dose AUC, based either on multilinear regression and 2 blood samples (C₀ and C₂) or 3 blood samples (C₀, C₂ and C₄),^[109] or on Bayesian estimators and 3 blood samples taken at 0, 1 and 3 hours and 0, 1.5 and 4 hours for non-CF and CF patients, respectively.^[112] Unfortunately, up until now, no prospective study has been conducted in thoracic transplantation to investigate the relationships between tacrolimus AUC_0-12, AUC_0-4 or single concentration-time points and clinical outcomes (acute rejection or toxicity), compare them and propose target levels.

Unlike calcineurin inhibitors, MPA TDM is not mandatory and MMF is usually administered at a fixed oral dose. Consequently, MPA levels have not been routinely monitored, despite variability in its pharmacokinetics and its pharmacodynamics comparable to that of other immunosuppressants,^[8,124,163] and TDM guidelines for MMF in heart and lung transplantation are very limited. When performed, despite poor correlation between MPA C₀ and AUC_0-12, the TDM of MMF is often based on C₀,^{[132,134,135,137]} with a recommended target range in heart transplantation of 1.2–3.5 mg/L (MPA assayed by HPLC).^[120] The relationship between MPA C₀ and efficacy was evaluated in a few studies in heart transplantation,^{[129,132,136,139,163,170]} and in a single one in lung transplantation,^[142] showing that a higher rejection incidence was related to lower C₀. No abbreviated AUC has been proposed as a surrogate for inter-dose AUC in thoracic transplantation, and the experience of sparse sampling strategies using multilinear regression equations is limited to two studies with small numbers of patients.^[121,138] Even though it was clearly demonstrated in renal transplantation,^[120] the relationship between MPA AUC_0-12 and rejection, or between any exposure index and toxicity (such as gastro-intestinal side-effects, the mechanism of which is still debated) has not been demonstrated in thoracic transplantation yet.^[94,133,161] To date, no study comparing the different exposure indices in thoracic transplant recipients has been performed and no AUC target established and MMF dose individualization has not been validated vs. fixed-dose regimens. However, there is increasing evidence in favour of the clinical benefit of MPA AUC_0-12 monitoring, particularly in renal transplantation.^[16,120]

The current TDM strategies for both sirolimus and everolimus are based on C₀. There appears to be a good relationship between C₀ and clinical outcomes for both molecules. Most studies on sirolimus were performed in renal transplantation, and we found none evaluating the clinical benefits of sirolimus C₀ monitoring in thoracic transplantation. Everolimus was evaluated in various solid organ transplant populations, including heart transplantation. A target everolimus C₀ range of 3 to 8 μg/L was proposed for heart transplant recipients, based on the results of only 2 studies where a limited number of patients were enrolled.^[193,201] The only study performed in lung transplantation failed to show a relationship between everolimus C₀ and efficacy, but a significant relationship was found between C₀ and toxicity.^[194] Based on these results, an everolimus C₀ target range of 3 to 12 μg/L was proposed for lung transplant recipients. We found no study in thoracic transplant patients evaluating the relationship between sirolimus or everolimus AUC and effect or looking for the best TDM tools in this population.

Because appropriate exposure for the prevention of acute rejection depends on the concomitant immunosuppressive regimen, the use of sirolimus or everolimus in CNI-free regimens in stable patients, maybe with higher C₀ (or AUC) targets to maintain immunosuppressive efficacy, also needs to be studied extensively.

Induction therapy with a polyclonal anti-lymphocyte or anti-thymocyte globulin is now used in around 50% of thoracic transplant recipients, with a significant decrease of the prevalence of acute graft rejection episodes.^[2,3] Induction therapy allows a quicker and stronger immunosuppression peri-operatively, and one could hypothesize that the level of maintenance immunosuppression should be driven by the presence or not of induction. The therapeutic benefit of a higher level of immunosuppression, in terms of rejection, might be however counterbalanced by a higher risk of toxicity, in particular infections and lymphoproliferative disorders. Anyway, the TDM of immunosuppressants used as maintenance therapy may be as relevant in patients with induction as in those without, but specific targets for CNI, MPA or mTOR inhibitors levels should then be defined in these patients in order to avoid overimmunosuppression and optimize the benefit/risk ratio.

4.3. What are the needs to improve the therapeutic management of thoracic transplant patients?

4.3.1. Development of sophisticated TDM tools

Heart and lung transplantations are not comparable with kidney transplantation. Firstly, the incidence of acute rejection is higher. Secondly, where graft failure in kidney transplant recipients can be handled by dialysis, graft loss in heart or lung transplant patients generally leads to death. Thirdly, patients with thoracic transplantation are more often administered drugs that intensively interact with the PK of the immunosuppressants (azole antifungals, calcium channel blockers, HMG-CoA reductase inhibitors), implying that the immunosuppressive treatment has to be closely monitored to avoid under- or over-immunosuppression. Therefore, it is legitimate to expect the development of sophisticated TDM tools dedicated to thoracic transplantation, in order to evaluate accurately and precisely patients’ exposure to the drugs at large, and to immunosuppressants in particular. Bayesian forecasting presents two major advantages compared to AUC monitoring based on the measurement of full concentration profiles or AUC estimation with multiple linear regression algorithms or from a single concentration: (i) it is characterized by a high flexibility in sampling times, as long as the true sampling times are known; (ii) in addition to the AUC, it can estimate several PK parameters and exposure indices simultaneously. Therefore, it can help identify absorption and elimination problems in specific populations such as for instance, diabetic or CF patients.

The use of Bayesian estimators for cyclosporine and MMF monitoring has been validated in renal transplantation by clinical studies.^[16,208] In thoracic transplantation, although Bayesian estimators have been set up for the calculation of the AUC of cyclosporine^[30–32] and of tacrolimus,^[112] they remain to be validated as TDM tools. Indeed, the implementation in routine clinical practice of efficient Bayesian estimators requires prior validation in large, collaborative, prospective clinical trials. Bayesian estimators should also be developed for the TDM of MMF, sirolimus and everolimus in thoracic transplant patients. Furthermore, as multiple immunosuppressive strategies now arise and as these patients often receive interacting drugs, Bayesian estimators should be built, and ideally validated, for the different immunosuppressive associations, and for the association of the immunosuppressant of interest with other types of interacting drugs.

4.3.2. Clinical trials dedicated to TDM and dose individualization

Once relationships have been established between clinical outcome and different markers of drug exposure, the steps necessary towards optimal TDM and dose adjustment are: (i) to identify the most predictive exposure index, either retrospectively from the exposure-effects relationships, or prospectively (which is seldom the case) by comparing the clinical outcome of different groups monitored using different exposure indices and their appropriate targets; (ii) to determine the target values for this exposure index, i.e. those associated with the best benefit/risk ratio, which again can be done retrospectively (such as with ROC curve analyses) or prospectively (concentration-controlled studies); and (iii) to validate the clinical usefulness of TDM using this marker and these target values in comparative trials vs. the standard of care at the time of the trial. In conclusion, large cohort TDM studies definitely need to be conducted in thoracic transplant patients.

One difficulty with the immunosuppressants is that they are always used in combination regimens, making the establishment of therapeutic ranges for individual drugs complex, as the exposure – effect relationship for a given drug may not be the same depending on the nature or even the dose of the molecules it is associated with.

Another difficulty is that patients’ follow-up in clinical trials is most of the time limited to the first one to three years post-transplantation, focusing on the incidence of acute rejection as the primary efficacy endpoint and on primary toxicity endpoints related with over-immunosuppression (infections, cancer, post-transplant lymphoproliferative disorders) or on molecule-specific side effects (nephrotoxicity, hypertension, hypertrichosis, gingivitis, diabetes for calcineurin inhibitors; gastro-intestinal and haematological toxicity for MMF; hyperlipidemia for sirolimus; nephrotoxicity, hyperlipidemia, thrombocytopenia for everolimus). Such short follow-up durations consequently limit the conclusions to the short-term outcomes and lead to underestimate the incidence and/or severity of cumulative toxicities (which can be particularly pertinent in lung transplantation where the doses given are higher than in other types of transplantation). If one wants to validate the benefits of therapeutic or TDM strategies in these populations and improve the evaluation of their risk, longer follow-up periods using hard clinical outcomes (morbidity, mortality, graft function) and other long-term outcomes (cardiac allograft vasculopathy, bronchiolitis obliterans syndrome, long-term cumulative toxicity such as renal insufficiency, metabolic disorders, etc.) should be considered. Long-term clinical trials are often expensive and complicated to conduct. One option could thus be to switch from clinical trials to large epidemiologic or cohort studies collecting all the information on therapeutic and monitoring strategies, compliance, side effects and quality of life, together with clinical outcomes. This would allow comparing the different therapeutic strategies and exposure levels, thus helping delineate the best therapeutic regimen and define precise exposure targets for each clinical condition and in each situation in the different transplant (sub)populations.

4.3.3. Perspectives

Increased knowledge and a better use of older drugs would also be of great interest. Even though they probably are the oldest and most widely used immunosuppressive drugs in medicine, there is almost no published experience regarding the application of TDM to corticosteroid therapy. Despite the emergence of steroid-sparing strategies, prednisolone therapy is still largely used, sometimes at high doses, in paediatric and adult thoracic transplant patients.^[3,42] High doses of corticosteroids are believed to induce the expression of a number of enzymes, including UGTs, which may interfere with the disposition of MPA.^[157] Moreover, inter-individual variability of the pharmacokinetics of prednisolone has been reported in lung^[209] and in heart transplant recipients,^[210] suggesting that optimization of corticosteroid therapy using TDM might also improve the management of thoracic transplant patients.

It has been suggested that pharmacodynamic monitoring of the cellular targets of immunosuppressant drugs may better reflect clinical outcomes than drug level monitoring. For instance, IMPDH pretransplant activity has been demonstrated to be associated with clinical outcome after renal transplantation.^[211] Thus, pharmacodynamic monitoring of calcineurin activity, IMPDH activity or IL2 levels may address some of the limitations of PK monitoring, either alone or associated with PK monitoring. However, the assays available for the pharamcodynamic monitoring are either non-existant or technically challenging, expensive and time-consuming, which could limit their use in routine monitoring and dose adjustment. Moreover, pharmacodynamic monitoring has not yet reached the level of evidence PK monitoring has. Although promising, its complementarity with, or superiority over PK monitoring still has to be proven.^[162]

Finally, in the last decade, single nucleotide polymorphisms (SNPs) have been intensively investigated and found on the genes encoding P-glycoprotein (P-gp), the cytochromes P450 (CYP450) and the uridine-diphosphate glucuronyl-transferases (UGT), all of which are involved in the disposition of most immunosuppressants. Some of the SNPs were found to be associated with altered protein expression or function, and with drug PK variability. A few SNPs have also been reported for IMPDH and IL2,^[212–214] involved in the immunosuppressive response, some of which are potentially associated with pharmacodynamic variability. The implications of these findings are important for transplant patients’ care, as the efficacy and toxicity of a given drug or association of drugs may be different depending on the genotype. Moreover, the combination of multiple substrates for P-gp, CYP450’s and UGTs can cause competitive inhibition of the proteins or up-regulate their function. Therefore, the addition of such agents to a transplant patient’s drug regimen may be accompanied by modifications in drug disposition or effect, which may be different depending on the genotype of the patient. Therefore, pharmacogenetic characterization of transplant patients (such as MDR1 and CYP3A5 genotypes for calcineurin inhibitors and mTOR inhibitors, and UGT1A9 and MRP2 for mycophenolate) may have the potential to optimize immunosuppressive therapy, in addition or, less likely as a replacement to the TDM approach. However, the level of evidence is low once again. Further investigations are thus needed in thoracic transplant patients in order to verify the clinical significance of these SNPs, before being able to conduct prospective, comparative trials investigating the impact of treatment individualisation based on this approach. If confirmed, a priori genotyping may become a new useful tool to help select the appropriate drugs and optimal starting doses for an individual patient, and thus improve clinical outcomes and participate in thoracic transplantation progress.

List of abbreviations

AcMPAG

acyl glucuronide of MPA

AUC_0-τ

area under the concentration-time curve over the dosing interval

AZA

azathioprine

BOS

bronchiolitis obliterans syndrome

BPAR

biopsy-proven acute rejection

CAV

cardiac allograft vasculopathy

CF

cystic fibrosis

CL/F

apparent clearance

C_max

maximum concentration

CNI

calcineurin inhibitor

CYP

cytochrome P450

D

dose

EMB

endomyocardial biopsy

FEV1

forced expiratory volume in one second

LVEF

left ventricular ejection fraction

MAT

mean absorption time

MMF

mycophenolate mofetil

MPA

mycophenolic acid

MPAG

7-O-mycophenolic acid glucuronide

P-gp

P-glycoprotein

PD

pharmacodynamics

PK

pharmacokinetics

TDM

therapeutic drug monitoring

ISHLT

International Society for Heart and Lung Transplantation

T_1/2

elimination half-life

t_max

time to maximum concentration

UGT

uridine diphosphate glucuronyl transferase

V₁/F

apparent volume of the central compartment

V_d/F

apparent volume of distribution