Etiologies and Outcomes Associated with Tacrolimus Levels out of Typical Range that Lead to High Variability in Kidney Transplant Recipients

Abstract

BACKGROUND:

High tacrolimus intrapatient variability (tac IPV) is associated with poor outcomes in kidney transplantation, including rejection, donor-specific antibodies, and graft loss. A common cause of high tac IPV is related to patient nonadherence, but this is yet to be conclusively demonstrated.

METHODS:

This was a longitudinal cohort study comprising adult kidney recipients, who received transplants between 2015 and 2017, with follow-ups through February 2020. The goal of this study was to identify the most common etiologies of tac levels outside the typical range, which lead to high tac IPV, and assess the etiology-specific associations between high tac IPV and graft outcomes. Multivariate Cox regression was used to assess time-to-event analyses.

RESULTS:

In total, 537 adult kidney recipients were included; 145 (27%) were identified as having a high tac IPV (>40%) 3–12 months post-transplant. Common etiologies of tac levels significantly outside the standard goal range (6–12 ng/mL) leading to high tac IPV included patient non-adherence (20%), infections (19%), tac-related toxicities (17%), and undocumented issues (27%). In multivariable Cox modeling, those with high tac IPV because of nonadherence had a 3.5 times higher risk of late acute rejection (p = 0.019) and 2.2 times higher risk of late graft loss (p = 0.044). No other etiologies in the typical tac level range were significantly associated with either acute rejection or graft loss.

CONCLUSIONS:

Although high tac IPV has many causes, only high tac IPV caused by nonadherence is consistently associated with poor allograft outcomes.

Tacrolimus (tac) is the cornerstone immunosuppressant therapy utilized in 95% of kidney transplant recipients.¹ In the early the 2000s, this agent became the gold-standard maintenance immunosuppressive therapy utilized in transplantation, leading to very low 1-year acute rejection rates and moderate improvements in long-term graft survival.² Although it is highly effective at preventing rejection, it has several notable limitations, including chronic adverse effects, such as nephrotoxicity, neurotoxicity, and cardiometabolic abnormalities.^3,4 Tac also has high intra- and inter-patient pharmacokinetic (PK) variability, predominantly related to variations in drug metabolism.^5,6

Because of its high PK variability profile, tac is universally monitored using whole blood trough levels.^7,8 In more than a dozen studies to date, the intrapatient variability (IPV) of tac levels, measured using the coefficient of variation (CV), hereafter called tac IPV, has been established as a significant and sensitive predictor of graft outcomes, including acute rejection, development of donor-specific antibodies, graft fibrosis, graft function, and graft loss.^5,6 Such observational studies have provided strong association evidence, but are not sufficient for full causal inference.⁵ In most of these studies, investigators inferred that high tac IPV was often caused by non-adherence. In support of this, a recent prospective tac PK cross-over study demonstrated that high adherence to tac regimens was correlated with low tac IPV.⁹ In a randomized controlled trial (RCT), our research group demonstrated that improved adherence to tac regimens lead to improved tac IPV.¹⁰ These studies provide a moderate level of evidence that non-adherence is a significant determinant of high tac IPV, but large studies elucidating the etiology-specific causes of high tac IPV are lacking in kidney transplantation cases.^5,6

Because tac levels are currently the only direct measure of adherence that can feasibly be monitored in real-world clinical practice, it is essential that there is greater evidence establishing the predominant issues leading to high tac IPV and how these levels are associated with clinical outcomes.¹¹ Documentation of the primary etiologies of levels that are significantly outside the standard tac goal range leading to high tac IPV, and which of these are significant risk factors for poor graft outcomes would provide insights that may aid in the development of interventions aimed at improving long-term graft outcomes.^12–15 Therefore, the objective of this study was to conduct detailed assessments of the primary factors associated with non-typical range tac levels leading to high tac IPV, which are significantly associated with graft outcomes in kidney transplant recipients.

Methods

STUDY DESIGN AND PATIENTS

This was a single-center retrospective longitudinal cohort study, including kidney recipients transplanted between January 1, 2015, and December 31, 2017; 2015 was chosen as the starting date to ensure adequate and consistent review of all tac levels with documentation of the likely etiology and action in our electronic medical record (EMR). Patient identification for the cohort ended in 2017 to allow sufficient follow-up time to capture long-term events occurring after the first year post-transplant. Patients were screened for inclusion if they received a kidney transplant at our transplant center during this time period. Pediatric patients (<18 years of age at the time of transplant), those who received non-renal transplants, and those with graft loss or insufficient tac levels to estimate IPV within a year of transplant were excluded. Patients were followed longitudinally from the time of transplant until graft loss, death, loss of follow-up, or the end of the study (February 20, 2020). This study was approved by the local institutional review board.

DATA CAPTURE, PRIMARY EXPOSURE, AND COVARIATE DEFINITIONS

Patient-level data were gathered using both electronic and manual chart abstractions. The center-specific Standard Transplant Analysis and Research (STAR) files were utilized to identify the cohort and obtain baseline recipient, donor, and transplant characteristics, as well as dates of graft loss and death. The United Network of Organ Sharing Standard Transplant Analytic Research (UNOS STAR) data dictionary was used to define all baseline covariates. These data were augmented with electronic data obtained from the EMR, which included all internal and external labs (tac levels), as well as transplant biopsy pathology reports.¹⁶ Manual chart abstraction was utilized to identify the specific etiologies of tac levels outside the standard range and to document the actions taken by clinicians based on these laboratory results.

To determine tac IPV, all tac levels drawn between 3 and 12-months post-transplant (exposure period) were reviewed. This time frame was chosen because it has been documented to be the optimal time to assess high tac IPV associated with poor long-term outcomes.⁵ From this, the tac IPV was calculated for all patients using the CV equation ([standard deviation / mean] × 100). A cut-off of >40% was established to identify patients with high tac IPV based on our previous work, which demonstrated that this threshold had the strongest association with outcomes, including acute rejection and graft loss. We recognized that other studies used a cut-off point of 30%; however, our higher threshold may been caused by our large African-American population, which can significantly influence tac IPV.¹⁷ From this subset of patients, tac levels significantly outside the goal range, defined as those <6 or >12 ng/mL, were identified and allowed for a buffer around our goal range of 8–10 ng/mL. Each of these 1,499 levels was reviewed in the EMR. Clinicians are required to document the likely etiology of levels that are out of the typical range and the action(s) taken as the standard of care. The location where the level was monitored (our outpatient lab, our inpatient lab, or an external lab) was also documented. After a review of a subset of these levels, the investigators defined seven potential etiologies in cases out of the typical range: patient-induced (non-adherence, patient was late to the laboratory, or the patient took the dose prior to the blood draw), purposeful reduction because of infection, purposeful reduction because of other toxicities (AKI neurotoxicity), purposeful reduction because of concomitant mammalian target of rapamycin (mTOR) therapy, drug-drug interactions, GI issue(s), or undocumented etiology. A minority of levels had more than one documented etiology, which were also classified as multifactorial. Actions were classified into three categories: ignored or no documentation, repeat levels, or tac dose change. There were multiple actions, such as dose changes and repeat levels. All assessments for etiologies of tac levels that were out of the typical range were performed in a blinded manner, such that the assessor did not know the clinical outcomes of these patients.

IMMUNOSUPPRESSION

The standard protocol at our institution is to utilize rabbit antithymocyte (rATG) or basiliximab induction, tacrolimus, mycophenolate, and prednisone. All patients received this regimen de novo. Several patients were converted from mycophenolate to everolimus or sirolimus because of toxicity. Prednisone was tapered to 5 mg daily by 6 weeks post-transplantation. Mycophenolate was administered at 1000 mg twice daily, with doses reduced for toxicities (infections, diarrhea illness or leukopenia). Tacrolimus was started at 2 mg twice daily and adjusted to achieve a level of 8–10 ng/mL.

OUTCOMES

The primary outcomes included biopsy-proven acute rejection, graft loss, and death, all occurring one year after transplantation. One-year post-transplant was set as our landmark date as all tac level assessments occurred between and 3–12 months post-transplant and this prevented immortal time bias, since those with graft loss within one year of transplantation were excluded.¹⁸ Event dates were recorded for time-t o-event analysis. Banff classification was used to determine the rejection episodes. These were all for cause biopsies, and treated borderline rejections were counted in our acute rejection outcome.

STATISTICAL ANALYSES

All data were captured and merged in Microsoft Excel CSV files (Microsoft, Redmond, WA, USA), including the OPTN/UNOS STAR file, EMR electronic abstracted data, and manual chart abstraction. Descriptive statistics were reported using proportions (%), means, and standard deviations (SD). Univariate comparisons were made using the Student’s t-test, Mann Whitney U test, Chi-square test, or Fisher’s exact test, where appropriate, based on data type and normality assessment. Cox regression was used for multivariate time-to-event analyses, adjusted for potential confounders (age, sex, race, donor type, HLA mismatches, panel reactive antibody levels, kidney donor profile index (UNOS definition [https://optn.transplant.hrsa.gov/resources/guidance/kidney-donor-profile-index-kdpi-guide-for-clinicians]), delayed graft function (dialysis within 7 d of transplant), number of tac levels from 3 to 12 months post-transplant, and <1-year acute rejection). All statistical analyses were conducted using SPSS version v23.0 (IBM Corp, Armonk, NY, USA).

Results

STUDY POPULATION AND BASELINE CHARACTERISTICS

Between January 2015 and December 2017, 690 kidney transplants were performed at our transplant center and screened for inclusion in this study. Of these, 23 (3.3%) were excluded because of age < 18 years at the time of transplant, 67 (9.7%) were excluded for receiving a non-renal transplant, 27 (4.5%) were excluded for having graft loss within 1 year of transplant, and 36 (5.2%) were excluded for not being able to estimate tac IPV (≤1 tac level during 3–12 months post-transplant, usually because of alterations in immunosuppression [with changes from tac to cyclosporine or belatacept], or lost follow-ups), leaving 537 in the final analysis (see Supplemental Figure 1 for the Consort diagram); 392 patients (73.0%) were deemed to have a mean tac IPV ≤ 40% during months 3–12 post-transplant, whereas 145 (27.0%) had a high mean tac IPV (>40%) during this time. The baseline characteristics of the cohort compared based on the tac IPV are displayed in Table 1. Our typical patient was an African-American male in their early to mid-50s with diabetes or hypertension as the predominant cause of kidney failure. Patients were similar in terms of age, sex, race, comorbidities, and transplant characteristics. Those with high IPV tended to have slightly older donors and to have received an African-American donor kidney, as compared to tac IPV ≤40% patients, but these differences were not statistically significant.

Table 1 –

Baseline characteristics compared between those with normal and high (>40%) tacrolimus (tac) intrapatient variability (IPV) 3–12 months post-transplant

Baseline Characteristics	Tac ≤40% (N=392)	Tac >40% (N=145)	p-Value
Age (years)	52.3±13.2	51.7±14.4	0.649
Female	38.5%	38.6%	0.983
African-American	61.5%	60.7%	0.868
History of Diabetes	38.5%	38.6%	0.983
Previous Transplant	8.4%	8.3%	0.958
HLA Mismatches	4.1±1.5	4.0±1.5	0.499
cPRA at Transplant	47.8±37.6	50.5±38.1	0.463
Deceased Donor	86.2%	86.9%	0.840
Donor Female	39.5%	37.2%	0.628
Donor African-American	20.9%	26.9%	0.141
Donor Age (years)	35.3±14.9	36.8±14.3	0.285
Kidney Donor Profile Index (KDPI)	0.32±0.26	0.36±0.27	0.176
Delayed Graft Function	18.1%	17.9%	0.961

ASSESSMENT OF TAC VARIABILITY, ETIOLOGIES, AND ACTIONS REGARDING OUT OF TYPICAL RANGE TAC LEVELS

A total of 9,759 tac levels were assessed for the 537 patients included in this study (3,117 were in the high tac IPV cohort and 6,642 were in the normal IPV cohort); 24.9% of the levels were out of range in the normal IPV cohort, whereas 48.1% of levels were out of range in the high IPV cohort. In the 145 patients with mean tac IPV > 40% for 3–12 months post-transplant, 1,499 tac levels that were substantially outside our goal range (8–10 ng/mL; these levels were either <6 or >12 ng/mL) were assessed to determine the documented etiologies of the abnormal levels leading to tac IPV > 40%. These were grouped into one of seven categories: patient-induced (non-adherence, patient was late in visiting the laboratory, or the level was drawn after the patient took the tac dose), infection-related (provider purposely lowered tac exposure while treating an infection), GI-induced (diarrhea or other GI-related toxicity), other toxicity (provider lowered tac exposure to treat other tac-related toxicity, usually nephrotoxicity or neurotoxicity), drug-drug interaction, tac used in combination with mTOR therapy (goal tac level is 2–5 ng/mL in this combination), or an undocumented etiology. The most common etiologies were patient-inducted-, infection-related, and unknown or undocumented (see Figure 1). Of the 1,499 abnormal tac levels assessed, 24.7% were determined to be multifactorial. The mean tac levels and proportions that were low or high based on etiology are shown in Supplemental Table 1. Non-adherent causes led to levels <6 ng/mL in 90% of the cases, with a mean of 4.5 ± 3.3 ng/mL, whereas those out of range because of the patient taking the dose prior to visiting the laboratory, led to levels >12 ng/mL 85% of the time, with a mean of 18.6 ± 10.2 ng/mL. In general, most causes of out-of-range levels were more likely to be low (<6 ng/mL) than high (>12 ng/mL), except for patients taking the dose prior to visiting the laboratory. Documented actions based on the out of typical range tac levels did not differ by etiology (p = 0.461, Table 2); analysis was only for those with high tac IPV [n = 145]); 12% of levels were ignored or disregarded, 49% were repeated, 24% led to tac dose changes, and in 16% no documented action was taken. In both univariate and multivariable analyses, the location the tac was measured (internal vs. external lab) had no bearing on outcomes.

Figure 1 –

Documented causes of out-of-typical-range tacrolimus (tac) levels in those with tac IPV >40%.

Table 2 –

Documented actions for tacrolimus (tac) trough concentrations for the specific-etiologies of the out of typical range levels in those with high tacrolimus (tac) intrapatient variability (IPV)

Reasons for Out of Typical Range Tac Levels	Action Based on Out-of-Range Level
	No Documentation	Level Ignored	Level Repeated	Dose Changed
Patient-Induced	20%	11%	50%	19%
Infection-Induced	16%	11%	45%	29%
GI-Induced	20%	7%	46%	26%
Tac Toxicity	13%	13%	49%	25%
Drug-drug interaction	9%	15%	55%	21%
mTOR-Based Regimen	18%	13%	47%	21%
Unknown/Not Documented	15%	12%	48%	25%

CLINICAL OUTCOMES

Late clinical outcomes (>1-year post-transplant) were assessed for all patients with tac IPV >40% vs. ≤40%, as well as specifically for six of the seven etiology domains (drug-interactions could not be assessed independently because of very low numbers of patients in this domain). Overall, multivariable Cox regression demonstrated that those with tac IPV >40% at 3–12 months post-transplant had higher rates of late acute rejection (HR 9.1 [0.2–337], p = 0.233) and late graft loss (HR 1.51 [0.7–3.1], p = 0.261), which were not statistically significant. The Cox regression model results for the etiology of high tac IPV are shown in Table 3. Those with tac IPV >40% because of patient-induced issues (non-adherence, late labs, or taking drugs prior to level measurement) had 3.5 times higher risk of late acute rejection (p = 0.019) and 2.2 times higher risk of late graft loss (p = 0.044, see Figure 2 for adjusted survival curves). Outcomes were further assessed by subdividing patient-induced etiologies into either non-adherence (missing doses, running out of medications, late labs) or tac dose taken prior to blood draw. The non-adherent patients had 5.4 times the risk of late acute rejection (HR 5.36 [95% CI 1.7–16.6], p = 0.004) and 3.1 times the risk of late graft loss (HR 3.09 [1.1–8.8], p = 0.034). Patients with high tac IPV because of taking the tac dose prior to labs exhibited no difference in late acute rejection (HR 1.27 [0.15–10.5], p = 0.827) or late graft loss (HR 1.08 [0.3–3.7]; p = 0.903). Gi-induced, treatment of tac toxicity and mTOR-based etiologies all failed to demonstrate an impact on outcomes. Infection-induced etiologies of high tac IPV were associated with a higher risk of late acute rejection with no effect on late graft loss, whereas undocumented etiologies were also associated with higher rates of late acute rejection with no impact on late graft loss.

Table 3 –

Association between etiologic-specific reasons for out of typical range tacrolimus (tac) levels leading to high tac IPV (>40%) and late acute rejection or graft loss (>1-year post-transplant)

Documented Reasons for Out of Typical Range Tac Levels	Late Acute Rejection		Late Graft Loss
	HR* (95% CI)	p-Value	HR* (95% CI)	p-Value
Patient-Induced	3.53 (1.2–10.1)	0.019	2.21 (1.1–4.8)	0.044
Infection-Induced	4.40 (1.4–13.5)	0.009	1.16 (0.5–2.8)	0.749
GI-Induced	1.97 (0.2–17.0)	0.537	1.48 (0.3–6.4)	0.599
Tac Toxicity	1.55 (0.4–5.7)	0.509	1.71 (0.7–4.0)	0.222
mTOR-Based Regimen	0.92 (0.1–7.2)	0.935	1.42 (0.4–4.8)	0.571
Unknown/Not Documented	3.00 (1.1–8.4)	0.032	1.38 (0.6–3.0)	0.413

HR: hazard-ratio; Tac: tacrolimus

^*Models adjusted for age, sex, race, donor type, HLA mismatches, panel reactive antibody levels, kidney donor profile index, delayed graft function, acute rejection within 1-year of transplant for rejection models, and # of tacrolimus levels from 3–12 months post-transplant

Figure 2 –

Association between high tacrolimus (tac) IPV because of patient issues and acute rejection (top panel) and graft loss (bottom panel).

Discussion

The results of this study demonstrated that high tac IPV (>40%) because of patient non-adherence occurring between 3–12 months post-transplant was a significant risk factor for late acute rejection and late graft loss. Other common etiologies of high tac IPV, including deliberate provider reductions in goal levels because of infection, toxicity, or concomitant mTOR therapy, had limited impact on the risk of developing late acute rejection and were not associated with graft loss. These results provide novel insights into the etiology-specific reasons for high tac IPV and their differing levels of influence on long-term outcomes. These findings add to the growing body of evidence indicating that tac IPV is a significant predictor of long-term outcomes in kidney transplantation, especially when the reason for the high tac IPV is related to patient non-adherence.^5,6,19

Several large-scale epidemiologic studies have demonstrated a significant association between high tac IPV, acute rejection, and graft loss. Shuker et al. conducted a cohort study with 808 patients and demonstrated that high tac IPV was an independent predictor of adverse allograft outcomes, including late rejection, decreased graft function, and graft loss. In Cox regression modeling, the risk of reaching a composite endpoint was significantly higher in the high tac IPV group (HR 1.41, 95% CI 1.06–1.89, p = 0.019).²⁰ Rozen-Zvi et al. conducted a cohort study in 803 patients with a median follow-up of 3.7 years. In survival analysis, patients in the highest tac IPV tertile had significantly reduced patient and graft survival compared to lower tertiles (HR 1.74, 95% CI 1.14–2.63, p = 0.01).²¹ Rahamimov et al. conducted a follow-up study with 878 patients, demonstrating that a tac IPV >25% was associated with more than 3 times the risk of graft loss (HR 3.66, 95% CI 2.3–5.8, p < 0.001).²² However, these studies did not determine the reasons for the high tac IPV. Our findings are similar with regard to outcomes, but with the added insight that only high tac IPV because of patient non-adherence was strongly associated with these late events.^5,19

This study had several implications that have the potential to impact clinical practice as it moves forward. Tac IPV can be feasibly monitored at most transplant centers, because the laboratory results are discrete values and the calculation is straightforward.¹¹ Identifying patients with high tac IPV may aid in triaging patients at the highest risk for deleterious graft outcomes, including rejection and graft loss.^8,23 These results suggest that the measurement should focus on trying to identify high variability because of patient non-adherence. This can be facilitated by focusing on levels 3–12 months post-transplant, not including levels drawn during hospitalization or levels that were clearly drawn after the tac dose was taken (levels that are substantially elevated and drawn after the typical morning dose, such as after 9 or 10 AM), and eliminating patients with purposeful reductions in goal levels to manage infections, tac toxicities, or with mTOR-based regimens.^5,6 Although this adds several layers of complexity to a triage system, it is feasible and would limit the number of false positives, reduce provider workload, and has the potential to identify better those patients who truly need interventions aimed at improving adherence. Clearly, future studies are warranted to determine the feasibility and effectiveness of such a triage system.

Beyond non-adherence, there were several other reasons for high tac IPV that were of interest. High tac IPV because of levels drawn after the patient took the dose were not associated with long-term outcomes, a finding that should reassure providers that these out of typical range levels can be safely ignored. Undocumented reasons for out of typical range tac levels leading to high tac IPV were common and were significantly associated with late acute rejection, but not graft loss. There are several potential explanations for this finding. Patients may not outwardly admit to being non-adherent or missing doses of tac or providers may not directly ask the patient about this (unidentified non-adherence). Other factors could also explain these out of typical range tac levels, including unidentified alterations in drug metabolism because of drug-drug or drug-food interactions, natural variation in intrapatient tac PK, or even genetic polymorphisms.^3,24 Nevertheless, future studies should focus on minimizing the number of undocumented reasons for abnormal tac levels, since this is clearly a limitation of this retrospective study.

This study had several limitations that should be discussed. Because of its retrospective design, there were a significant number of out of typical range tac levels (27%) where the etiology was not documented or unknown to providers. It is likely that if this was a prospective study, the number of out of typical range levels without clear etiology would have been markedly reduced. Assessments were completely dependent on detailed provider documentation in the EMR, which is a limitation of most retrospective chart abstraction studies. Pediatric patients and those who lost their graft within 1 year of transplantation were excluded; thus, these results are not applicable to those <18 years of age or externally valid for early graft events. Late events were the focus of this study as non-adherence tends to be more common and more debilitating after 1-year post-transplant. An important limitation is the likelihood of unmeasured or residual confounding factors. This was minimized using several comprehensive data sources and conducting detailed manual chart abstraction. Nonetheless, there are likely to be potential confounding factors that were not measured in this study, including health literacy, education, income, caregiver support, and other social determinants of health. There were a significant number of out-of-range levels that were not well-documented in the electronic medical records; thus, this is likely an influential issue that could possibly impact our inferences. Finally, we also did not capture tac doses in a comprehensive manner; thus, we could not estimate tac clearance in this analysis.

In summary, these data demonstrated that high tac IPV is common because of patient non-adherence and is a significant risk factor for poor outcomes, including acute rejection and graft loss; on the other hand, high tac IPV resulting from taking the dose prior to labs, or deliberate reductions in goal levels because of infection or other toxicities has limited impact on graft outcomes. If high tac IPV is to be used as an indicator for patients who are at risk for deleterious outcomes, efforts should be made to identify the reason for the high variability, and when non-adherence is identified, aggressive corrective action should be instituted.