Thrombotic microangiopathy associated with sirolimus levels following allogeneic hematopoietic cell transplantation with tacrolimus/sirolimus-based GVHD prophylaxis

Abstract

Post-transplant thrombotic microangiopathy (TMA) is a multi-factorial complication of allogeneic hematopoietic cell transplantation (allo-HCT) whose incidence is increased when using a sirolimus plus tacrolimus regimen for acute graft-versus-host disease (aGVHD) prophylaxis. We evaluated the incidence and possible risk factors for TMA in a case series of 177 patients who received allo-HCT using SIR/TAC-based GVHD prophylaxis. Donors were either sibling (n=82) or matched unrelated (n=95). Within the first 100 days post-HCT, 30 (17%) patients were diagnosed with TMA, and an additional nine patients (5%) were classified as probable TMA cases. The median time to TMA onset was 4.6 weeks (range: 1.6-10.6). Thirty-four patients developed both TMA and aGVHD, with the majority of patients (81%) developing aGVHD first. By multivariable analysis, the following factors were found to be associated with increased risk of TMA: day 14 serum sirolimus: ≥9.9 ng/ml (HR: 2.19, 95% CI: 1.13-4.27, p=0.02), presence of prior aGVHD grades II-IV (HR: 3.04, 95% CI: 1.38-6.71, p<0.01), and fully myeloablative conditioning (HR: 3.47, 95% CI: 1.60-7.53, p<0.01). The risk factors for TMA suggest that when using sirolimus/tacrolimus for GVHD prophylaxis, careful monitoring and adjustment of sirolimus dosages is critical, particularly in patients with active aGVHD.

Introduction

Post-transplant thrombotic microangiopathy (TMA) is a serious complication of hematopoietic stem cell transplantation (HCT). This condition is characterized by the presence of arteriolar thrombi associated with damaged vessel walls that subsequently leads to intravascular platelet activation and formation of platelet-rich thrombi within the microcirculation, thereby causing platelet consumption, mechanical damage to blood cells, or organ damage [1]. Post-HCT TMA has been attributed to the vascular endothelial damage caused by radiation therapy, high-dose chemotherapy, immunosuppressive agents, graft-versus-host disease (GVHD), or infections [2]. Unlike most patients with familial or acquired thrombotic thrombocytopenic purpura (TTP), bone marrow transplant-associated thrombotic microangiopathy is usually not associated with severely reduced or absent plasma metalloprotease ADAMTS13 activity [3]. With a few exceptions, plasma exchange is ineffective in treatment of this disorder [3].

Diagnosis of TMA has proven very challenging in patients following HCT due to the presence of multiple contributing factors such as opportunistic infections, delayed engraftment, medications, regimen-related toxicity, presence of acute graft-versus-host disease (aGVHD), and other disorders associated with the HCT process. In 2005, the toxicity committee of the Blood and Marrow Transplant Clinical Trials Network (BMT CTN) proposed an operational definition of TMA (Table 1). Subsequently an International Working Group was formed to develop a consensus set of criteria for diagnosing TMA [4] (Table 1).

Table 1. Diagnostic Criteria for TMA.

Criteria	COH*	CIBMTR	BMT CTN	IWG
Hemolysis	Presence of schistocytes, persistent nucleated RBC’s	Microangiopathic hemolysis	RBC fragmentation and ≥2 schistocytes/ HPF on peripheral smear, Negative direct and indirect Coombs test	(>4%) of schistocytes in the blood, Decrease in haptoglobin or increase in RBC transfusion requirement
Thrombocytopenia	Prolonged or progressive thrombocytopenia (platelets <50 ×109/L or ≥50% decrease from previous counts)	Thrombocytopenia (platelets <50 ×109/L)		Prolonged or progressive thrombocytopenia (platelets <50 ×109/L or ≥50% decrease from previous counts)
Liver Function	LDH >2x upper limit of normal	LDH >2x upper limit normal, Bilirubin >2x upper limit normal	Concurrent increase in LDH above institutional baseline	Sudden, persistent increase in LDH
Renal Function	SCr >1.5 × baseline	SCr >2mg/dL or >50% over baseline	Concurrent renal and/or neurological dysfunction without other explanations
Neurological		Neurological changes

^*The presence of all 4 COH criteria listed defines definite TMA. Probable TMA is defined as concomitant presence of ¾ of the above criteria in presence of clinical diagnosis by an independent attending physician.

COH – City of Hope, CIBMTR – Center for International Blood and Marrow Transplant Research, BMT CTN – Bone Marrow Transplant Clinical Trials Network, IWG – International Working Group, RBC – Red blood cell, HPF – high power field, LDH – lactose dehydrogenase, SCr – serum creatinine

Calcineurin inhibitors, such as cyclosporine and tacrolimus may contribute to development of TMA by increasing production of thromboxane A₂ and decreasing production of prostacyclin (PGI₂), as well as by damaging renal endothelial cells directly. These agents can also cause neurotoxicity, another feature of TMA [1, 2, 5]. Recent studies have demonstrated that addition of sirolimus to calcineurin inhibitors increases the risk of TMA after HCT [6, 7]. The etiology of sirolimus-induced TMA is not clear, but may be attributable to enhanced platelet activation and aggregation, leading to endothelial damage. Another theory involves the pharmacokinetic interaction between sirolimus and calcineurin inhibitors, which may potentially lead to increased serum and kidney levels of these agents by sirolimus.

In this retrospective analysis, we sought to extend our earlier observations [7] and determine the incidence and clinical characteristics of post-HCT TMA associated with sirolimus/tacrolimus (SIR/TAC)-based GVHD prophylaxis. In addition we extensively evaluated the association between TMA and SIR/TAC serum levels.

Materials and methods

Study Design

From January 2005 to August 2007, a consecutive case-series of 177 patients underwent allogeneic HCT (alloHCT) at City of Hope (COH) using a SIR/TAC-based GVHD prophylactic regimen and hematopoietic stem cells from either a matched related sibling or matched unrelated donor (MUD). Patients were identified and selected for retrospective analysis from a prospective observational research database. The COH Institutional Review Board approved the analysis of these data. Patients with matched related donors were previously reported as part of a phase I/II study [7].

Conditioning and GVHD Prophylaxis Regimens

Conditioning regimens consisted of fludarabine/melphalan (n=106, 59.9%), fractionated total body irradiation (FTBI)/ cyclophosphamide (n=11, 6.2%), FTBI/etoposide (n=46, 26.0%) and busulfan/cyclophosphamide (n=14, 7.9%). Of these regimens only fludarabine/melphalan was classified as reduced intensity. Patients were selected for a reduced intensity regimen based on advanced age or other comorbidity per institutional standard operating procedures. The GVHD prophylaxis was administered according to published reports [7, 8] as follows: sirolimus 12 mg by mouth on day −3 (loading dose), followed by 4 mg orally per day, with subsequent dose adjustments to maintain serum levels between 3-12 ng/ml. Tacrolimus was initially dosed at 0.02 mg/kg intravenously per day, starting on day −3, then switched to an equivalent oral dose when oral intake was adequate, to maintain target serum levels of 5-10 ng/ml. Tacrolimus and sirolimus levels were measured at least weekly until day 100 with dose adjustments made for target levels and/or clinical toxicity. For MUD HCT with <10/10 matched donor, additional methotrexate was administered at a dose of 5mg/m² on days +1, +3, and +6. One patient undergoing MUD-HCT also received antithymocyte globulin at a dose of 0.5 mg/kg on day −3, 1.5 mg/kg on day −2 and 2.5 mg/kg on day −1.

Supportive Care

Supportive care, including prophylactic antibiotics, antifungal therapy, total parenteral nutrition, hematopoietic growth factors, immune globulin replacement and treatment of mucositis and neutropenic fever, was administered in accordance with institutional standard operating procedures. All patients received acyclovir 250 mg/m² starting on day −1. Antifungal prophylaxis consisted of low-dose lipid complex amphotericin B (Abelcet®) (1 mg/kg), caspofungin or micafungin starting on day 0 or day +1. In addition, oral azole antifungals including posaconazole, voriconazole, fluconazole, and itraconazole were later initiated in select patients for prophylaxis or treatment of fungal infections. Sinusoidal obstructive syndrome (SOS) prophylaxis was provided using low-dose heparin (100 units/kg/day) or ursodiol 300 mg twice daily.

TMA Definition

We evaluated patients for thrombotic microangiopathy based on the criteria used at City of Hope. Our criteria are compared to other commonly-used criteria in Table 1. Patients were diagnosed with ‘definite’ thrombotic microangiopathy if they met the following diagnostic criteria: serum creatinine (sCr) increase of ≥ 50% above baseline, lactate dehydrogenase (LDH) > 2x upper normal limit, the presence of schistocytes or persistent presence of nucleated RBCs, and prolonged or progressive thrombocytopenia (<50 × 10⁹/L or ≥ 50% decrease). Patients were considered to have ‘probable’ thrombotic microangiopathy if they experienced 3 out of 4 of the above-mentioned criteria. Patients who met the characteristics of TMA secondary to disease relapse or progression were not classified as TMA cases in this study.

GVHD and Response to Intervention Definitions

Acute GVHD was defined and staged according to Glucksberg et al. [9]. Chronic GVHD was defined and staged by the limted/extensive classification [10]. TMA response to intervention is defined as follows. Complete response (CR) is defined as resolution of all TMA criteria to patient baseline. Partial response (PR) is defined as improvement of individual TMA criteria such that the patient no longer meets the definition of TMA, but has not returned to baseline levels.

Statistical Analysis

Demographic, disease and treatment characteristics were summarized using descriptive statistics. Survival estimates were calculated based on the Kaplan-Meier product-limit method [11], 95% confidence intervals were calculated using the logit transformation and the Greenwood variance estimate [12]. Differences between Kaplan-Meier curves were assessed by the log-rank test. Patients who were alive at the time of analysis were censored at the last contact date. Overall survival (OS) was measured from transplant to death from any cause. Event-free survival (EFS) was defined as time from transplant to recurrence, progression or death. The relapse/progression incidence was defined as time from transplant to disease recurrence or progression. The cumulative incidence of relapse/progression was computed treating a non-relapse death event as a competing risk using the method described by Gooley et al [13].. Non-relapse mortality (NRM) was measured from transplant to death from any cause other than disease relapse or disease progression. The cumulative incidence of NRM was calculated using relapse/progression incidence as a competing risk. The cumulative incidence of acute GVHD and hazard ratio were estimated after taking into account the competing risk of engraftment failure and early death/relapse. Time to onset of cGVHD was estimated accounting for the competing risks of early death or second transplant. Similarly the cumulative incidence of TMA was estimated after accounting for the competing risk of death and relapse. Possible differences between cumulative incidence curves in the presence of a competing risk were tested using the Gray method [14]. The significance of disease and treatment features on TMA risk and on NRM risk was assessed using the Cox proportional hazards regression analysis competing risks analogue [15].

In addition, a joint model [16] of tacrolimus/sirolimus serum levels and time to TMA was examined to consider the shared evolution of the repeated drug level measurements and event times, when a possible association between the two processes exists. Square-root transformed sirolimus and tacrolimus serum levels were used to reduce the variability of individual measurements and improve normality of the data. A desirable feature of the joint modeling approach is that when an association is absent, the joint analysis will produce the same results as would be obtained by performing separate analyses for each process.

An exploratory ‘drug level’ analysis was also performed using the calculated median values for both tacrolimus and sirolimus levels over set durations (7, 14, and 30 days) post-HCT for each patient. For example, the day 14 sirolimus level was the median value for the patient over the 14 days post-HCT (twice weekly levels, so 4 levels). Based on our published SIR/TAC experience, as part of a phase II trial utilizing sibling donors, we hypothesized that SIR/TAC serum levels in the early period post-alloHCT (first 30 days) would be the most predictive of TMA outcome, even for those patients who maintain serum levels considered therapeutic/non-toxic. We therefore conducted serum level evaluations for modeling at days 7, 14 and 30 days, among the upper quartiles of each of the distributions (≥ median values) for both tacrolimus and sirolimus.

TMA endpoints (e.g., definitive, definitive/probable) were modeled as a function of prognostic variables that were determined by a literature review identifying factors associated with development of TMA in patients treated with alloHCT. Factors evaluated for association with outcome included: patient age at transplant (<46, ≥46 years), patient/donor gender combinations (female donor to male patient, others), disease risk-status at transplant (Low, Intermediate, or High: defined in footnote of Table 2), patient/donor CMV status, conditioning regimen (Flu/Mel, FTBI/VP16 or Cy, Bu/Cy; and reduced intensity, myeloablative), donor type (related, unrelated) and acute GVHD grade (0-I, II-IV). Acute GVHD (grade II-IV) was treated as a time-dependent covariate in the risk-factor analysis for TMA. All calculations were performed using SAS v9.2 (SAS Institute, Cary, NC) and R (version 2.11.1; http://www.r-project.org). Statistical significance was set at the P <0.05 level; all P values were two-sided. The data were locked for analysis on July 31, 2010 (analytic date).

Table 2. Summary of Patient, Disease, Transplant Characteristics (n=177).

Variable	Number (%) or Median (Range)
Patient Gender
Female	79 (44.6)
Male	98 (55.4)
Patient/Donor Gender
Male/Female	40 (22.6)
Others	137 (77.4)
Age at Transplant (years)	46 (10 – 70)
Disease Risk-Status at Transplant*
Low Risk	69 (39.0)
Intermediate Risk	33 (18.6)
High Risk	75 (42.4)
Diagnosis
AML	63 (35.6)
ALL	39 (22.0)
Non-Hodgkin Lymphoma	25 (14.1)
Myelodysplastic Syndrome	17 (9.6)
CML	9 (5.1)
Myeloproliferative Disorder	9 (5.1)
Hodgkin Lymphoma	8 (4.5)
Multiple Myeloma	4 (2.2)
CLL	3 (1.7)
Donor
Sibling	82 (46.3)
Unrelated	95 (53.7)
Stem Cell Source
Bone marrow	23 (13.0)
Peripheral Blood	154 (87.0)
CMV Serostatus
Patient+/Donor+	75 (42.4)
Patient+/Donor−	51 (28.8)
Patient−/Donor+	25 (14.2)
Patient−/Donor−	26 (14.7)
Conditioning Regimen^
Flu / Mel	106 (59.9)
FTBI / Cy	11 (6.2)
FTBI / VP-16	46 (26.0)
Busulfan / Cy	14 (7.9)

^*Disease Risk-Status categories include: Low Risk – all diseases 1s complete remission, myeloma partial remission, CML 1^st chronic phase, refractory anemia, and refractory anemia with ringed sideroblasts; Intermediate Risk – lymphoma/leukemia 2^nd complete remission or partial remission, CML 2^nd chronic phase or accelerated phase, myeloproliferative disorders; High Risk – all diseases relapse, induction failure or progressive disease, chronic myeloid leukemia blast crisis, refractory anemia with excess blasts

^{^}For conditioning regimens, only Flu/Mel was reduced intensity and all others were fully myeloablative.

AML – acute myeloid leukemia, ALL – acute lymphoblastic leukemia, CML – chronic myeloid leukemia, CLL – chronic lymphocytic leukemia, CMV – cytomegalovirus, Flu – fludarabine, Mel – melphalan, FTBI – fractionated total body irradiation, Cy – cyclophosphamide, VP-16 – etoposide

Results

This study is comprised of a consecutive case series of 177 patients who underwent alloHCT at City of Hope between January 2005 and August 2007. Eighty-two patients (46%) received stem cells from a sibling donor and ninety-five patients underwent a matched unrelated HCT. The median patient age was 46 years (range 10-70). A summary of the patient, disease and transplant characteristics is provided in Table 2.

Overall HCT outcomes

With a median follow-up of 50 months, 89 (49.5%) patients are alive. Of the 88 patient deaths, 42 were due to disease relapse/progression, and the remaining 46 patients died of non-relapse causes. The median time of neutrophil engraftment was 16 days; four patients failed to engraft. The overall survival (OS), event-free survival (EFS), and cumulative incidence of relapse/progression at 2 years were 59.9% (95% CI: 55.4-64.1%), 51.7% (95% CI: 47.8-55.5%), and 31.1% (95% CI: 25.0-38.7%), respectively. Non-relapse mortality was 7.4% (95% CI: 4.4-12.4%) at 100 days, and 15.8% (95% CI: 11.3-22.2%) at 2 years. The cumulative incidence of acute GVHD grade II-IV was 50.5% (95% CI: 43.2-59.1%), with a median onset of 23 days (Figure 1A). The cumulative incidence of grade III-IV aGVHD was 17.9% (10.6-30.2%). The cumulative incidence of chronic GVHD (time to cGVHD with competing risks of early death or second transplant) was 62.5% (95% CI: 55.6-70.3%) at 1 year and 71.3% (95% CI: 64.6-78.7%) at 2 years.

Figure 1. Cumulative Incidence of aGVHD and TMA (First 100 Days).

Panel A. Time to onset of grade II-IV aGVHD.

Panel B. Time to onset of definitive TMA with death prior to 100 days calculated as a competing risk.

Incidence and Timing of TMA

TMA data are summarized in Table 3. Of the 177 patients studied, 30 (17%) were diagnosed with definite TMA based on the institutional diagnostic criteria. In addition, there were 9 patients who did not meet the definitive diagnostic criteria due to a missing test but were clinically diagnosed with TMA by independent attending physicians (probable TMA). If those patients with probable TMA are included, the overall incidence of TMA in this study would be 22% (39/177) (Figure 1B). Seven of these 39 patients (17.5%) met the criteria for definitive/probable TMA in the setting of ongoing multi-organ failure. The median time from HCT to onset of definite TMA was 4.6 weeks (1.6-10.6 weeks). TMA outcomes are described in Table 3.

Table 3. TMA Outcomes.

Variable (N=30 for definite TMA)	Percent (number) or Median (range)
TMA incidence	17% (30 of 177patients )
Median time to TMA onset (weeks)	4.6 (1.6-10.6)
TMA with aGVHD (any time)	90% (27 of 30 TMA cases)
TMA with prior aGVHD	81% (22 of 27 cases with aGVHD)
TMA response to treatment
Response	73% (22 of 30 TMA cases)
Median time to response	4.5 weeks (range: 1-27 weeks)
Complete resolution	70% (21 of 30 TMA cases)
Median time to complete resolution	5.5 weeks (range: 1-27 weeks)
Max LDH	1128-62817
Median LDH	2230
Max creatinine	0.6-4.5
Median creatinine	1.5

TMA – thrombotic microangiopathy, aGVHD – acute graft-versus-host disease

Management of TMA and Outcomes

TMA interventions and responses are detailed in Table 4. Initial treatment for patients diagnosed with definite TMA (n=30) consisted of discontinuation of sirolimus (n=9), tacrolimus (n=8), or both (n=6) and immunosuppressant dose adjustments for the remaining 7 patients. At the time of resolution, tacrolimus and sirolimus were both discontinued in 12 patients and only 4 patients remained on the combination. Sirolimus alone was continued for 11 patients whereas tacrolimus alone was continued in only 3 patients. Out of 30 patients diagnosed with TMA, 8 were treated with plasma exchange at the discretion of the treating physician, and 4 patients responded. Twenty-two patients (73%) in total responded to treatment, with a complete resolution rate of 70%. Median time from diagnosis of TMA to response and resolution of symptoms were 4.5 and 5.5 weeks respectively (range: 1-27 weeks). Twenty-five patients had both tacrolimus and sirolimus discontinued and were managed with mycophenolate mofetil (MMF) plus prednisone (n=15), prednisone alone (n=9), or MMF alone (1). One patient was discontinued from sirolimus but remained on tacrolimus alone.

Table 4. Response of TMA to tacrolimus/sirolimus interventions.

Intervention	Initial Interventions	Final Intervention
D/C Sirolimus responded/treated	9/9	11/11
D/C Tacrolimus* responded/treated	5/8	3/3
D/C Tacro & Siro** responded/treated	2/6	5/12
Dose Modified* responded/treated	6/7	3/4

The number of patients who responded to a treatment is given over the number who were treated. The initial intervention shows first treatment responses in each category and final shows the results of the last treatment attempted. D/C – discontinued

^*One patient was treated with dose reduction of sirolimus, discontinuation of tacrolimus and plasma exchange, and responded to the overall therapy.

^**Seven patients were treated with discontinuation of both tacrolimus and sirolimus as well as plasma exchange, and three patients responded to the overall therapy.

We evaluated the survival outcomes of patients with or without TMA. The cumulative incidence of NRM at 2 years was 33.3% in those who developed TMA compared with 12.25% in patients without TMA (p=0.004) (Figure 2). The results of the multivariable model for NRM showed that after adjusting for conditioning regimen, acute GVHD and sirolimus drug level, TMA was independently associated with an increase in NRM hazard [HR: 2.76 (1.29-5.92), p=0.009]. Multivariable results for NRM are summarized in Table 5a.

Figure 2. NRM stratified by presence of TMA.

Time to death from non-relapse causes calculated with relapse, progression or second transplant as competing risks. The dashed line represents patients with definitive TMA (by laboratory criteria).

Table 5. Hazard Risks for NRM and TMA NRM.

A. NRM Risk
Parameter	Value	N	# of events	Hazard Risk Ratio (95% CI)	p value
Sirolimus	< 9.9	132	21	Baseline
14-day levels	≥9.9	44	12	1.67 (0.79-3.51)	0.18
TMA	No TMA	146	22	Baseline
	Definitive TMA	30	11	2.76 (1.29-5.92)	0.009
B. TMA Risk
Parameter	Value	N	# of events	Hazard Risk Ratio (95% CI)	p value
Conditioning	Reduced intensity	106	9	Baseline
	Fully myeloablative	70	21	3.47 (1.60-7.53)	0.002
Acute GVHD	Grade 0-1	95	8	Baseline
	Grade 2-4	81	22	3.04 (1.38-6.71)	0.006
Sirolimus	< 9.9	131	18	Baseline
14-day levels	≥9.9	45	12	2.19 (1.13-4.27)	0.02

NRM – non-relapse mortality, TMA – thrombotic microangiopathy, GVHD graft-versus-host disease, Reduced intensityconditioning includes only the Flu/Mel regimen – all others were fully myeloablative

Acute GVHD and azoles in TMA patients

The overall cumulative incidence of aGVHD (grade II-IV) was 50.5% (95% CI: 43.2-59.1%). Among those 30 patients who developed definite TMA, 27 (90%) also developed aGVHD, with the majority of patients (81%) developing TMA after a diagnosis of aGVHD had been made.

For the patients with definitive TMA, 14 of 30 patients did not receive azoles, 15 received azoles (3 posaconazole, 5 itraconazole, 7 voriconazole) prior to their TMA diagnosis and 1 patient received voriconazole at the time of the TMA diagnosis.

High sirolimus levels associated with TMA

Factors found to be significantly associated with TMA at the p=0.10 level (conditioning regimen, presence of aGVHD, sirolimus day-14 levels) by univariate Cox regression analysis were included in the multivariable model. Results of the multivariable analysis are displayed in Table 5b. The highest quartile of serum sirolimus exposure at day 14 (cutoff: 9.9 ng/ml) was found to be independently associated with an increased risk of definitive TMA (HR: 2.19, 95% CI: 1.13-4.27, p=0.02). Other significant risk factors were prior aGVHD grades II-IV and use of a fully myeloablative conditioning regimen (Table 5b). Tacrolimus levels were not significantly associated with TMA. Median drug levels over the first 30 days post alloHCT were not found to be predictive of TMA overall.

Because sirolimus binds to the same family of intracellular FK-506 (tacrolimus)-binding proteins at a site distinct from tacrolimus, these two drugs may interact or when used in combination produce a total effect greater than the sum of their individual effects. We further evaluated the possible impact of tacrolimus and sirolimus levels on TMA risk by constructing joint models of the repeated drug level measurements over time, and time-to-event data (TMA). The joint models included both additive (model 1) and multiplicative (model 2) parameters for sirolimus and tacrolimus (siro + tacro, siro × tacro), with and without other covariates (e.g., age, and conditioning regimen type). The additive joint model showed marginal significance of a drug level effect (association p-value 0.06 for definite TMA, see Supplementary Table 1), suggesting that TMA risk may be further explained by the additive effect of the two drugs beyond the individual effect of sirolimus as seen by multivariable analysis. This marginal significance remained after adjusting for other covariates, age and conditioning regimen (p-value 0.07 for definite TMA, see Supplementary Table 2). The results of the joint multiplicative model were not significant.

Discussion

Our data extend earlier observations, including those from City of Hope and Dana Farber Cancer Institute, that SIR/TAC-based GVHD prophylaxis is associated with an increased risk of TMA. With increased numbers of patients and detailed drug level data/analysis, our data corroborate that sirolimus level is one of the risk factors associated with TMA.

Diagnosis of TMA is difficult in patients who have undergone HCT, since they often develop thrombocytopenia, anemia, renal dysfunction, and fragmented red blood cells post-transplant [17]. Schistocytes occur in a variety of conditions such as chronic renal failure, pre-eclampsia, and prosthetic heart valves. Patients receiving marrow allografts or autografts for a variety of indications have also been reported to display 0-4 percent schistocytes, in the absence of other TMA hallmarks, placing them at risk for mis-diagnosis of TMA [3]. Cho et al. [18] validates criteria for TMA proposed by the Blood and Marrow Transplants Clinical Trials Network (CTN) [17] and the International Working Group (IWG) [4], in an analysis of 672 allogeneic HCT patients. In their study the incidence of TMA by CTN-defined criteria is 6.1%, while the incidence of TMA by the IWG definition is only 2.5%. Their overall cumulative incidence of TMA is 12.7%, a concept Cho defines as including both definite-TMA cases (by CTN criteria) and probable-TMA cases (meeting CTN criteria without renal or neurologic dysfunction). Sixty-six percent of the TMA patients by CTN criteria did not have any degree of schistocytes, which is required by IWG criteria, and 18% of TMA by IWG criteria did not have renal or neurological dysfunction.

In the current study, we used institutional criteria of TMA (detailed in Table 1). We did not set a quantitative criterion for schistocytes, since precise quantification is difficult and the presence of any schistocytes in the setting of hemolysis is considered enough to suspect TMA. In order to capture a broad range of TMA or TMA-like pathology, we included nRBC as part of TMA assessment, as it was our impression that nRBC tended to show up earlier than schistocytes, or were easier detect in routine CBC reports. Ninety-three percent of our TMA cases presented with both nucleated RBCs and schistocytes. In our analysis, we also considered probable cases in which there was one missing test/criterion and/or the treating physicians made a clinical diagnosis of TMA and treated accordingly. Although this study reports results only for the definite TMA cases, all analyses were also conducted on the probable plus definite cases and results were qualitatively similar. The reason for examining probable cases is that we believe TMA is often underestimated in the literature. For example, at our institution among the probable cases, many were “called” as TMA by the treating physician prior to meeting all 4 criteria in order to allow an earlier start of treatment, a sort of pre-emptive strike.

Recent studies, including our own, showed that TMA incidence is notably high after allo-HCT using sirolimus-containing GVHD prophylaxis [6-8]. This study confirms and extends those data and further associates the incidence of TMA with higher serum levels of sirolimus, even those believed to be therapeutic and non-toxic. Importantly, significant associations with sirolimus serum levels above 9.9 ng/ml were observed for the early time points (post-transplant day 14). Based on this data we have modified our upper range of therapeutic sirolimus levels from the original 12 ng/ml to 10 ng/ml. There was no direct impact of tacrolimus on TMA in the current analysis, but when the tacrolimus levels were computed together with sirolimus using a square root additive joint model, there was a trend for an association with TMA (p=0.07 by multivariable analysis). In other words, tacrolimus when given in combination with sirolimus may enhance the risk of TMA beyond that associated with sirolimus alone. This question will require additional study with a larger number of patients. Serum levels of both tacrolimus and sirolimus are also influenced by genes affecting drug absorption, distribution, metabolism, and excretion (ADME). Allelic variants in ADME genes can determine the pharmacokinetic variability of medications and are shown to influence the outcomes of medical treatments using these drugs for kidney transplant. We are currently investigating the effects of ADME variants in HCT recipients with the goal of better tailoring the initial doses of these drugs, potentially avoiding toxicities including TMA.

We also identified other TMA risk factors including fully myeloablative conditioning regimens and prior occurrence of aGVHD (grades II-IV). Concurrent GVHD has been reported as a risk factor by the Dana Farber Cancer Institute [6] and others [19]. Bu/Cy conditioning has been previously associated with TMA and SOS, but in our study there was no independent association of Bu/Cy with incidence of TMA. Rather, fully myeloablative conditioning was independently significantly associated with TMA, even if the Bu/Cy patients were not included. Addition of methotrexate to sirolimus has previously been shown to increase the risk of SOS [20]. In our study, all patients in the MUD group used methotrexate, yet there was no significant difference in the incidence of TMA between sibling donors and MUD donors in this cohort of patients. For those patients who develop TMA, there is an independently significant increase in the risk of non-relapse mortality, highlighting the need to better understand and prevent this transplant complication.

A potential risk factor that we would like to pursue in a prospective study is the role of prior azole anti-fungal agents. Due to the retrospective nature of this study, detailed data on azole use were not collected for everyone, but we were able to obtain it for the TMA cases. Among patients with definitive TMA 50% had prior exposure, which is generally high in our practice., since the primary prophylaxis uses non-azole antifungal meds. Since azole is commonly used for patients with GVHD requiring high-dose steroids, an independent effect of azoles will be difficult to analyze. Thus we opted not to include this factor in our Cox model.

There have been no standardized treatments for post-HCT TMA. Since plasma exchange has not been shown to provide a significant benefit in treatment of patients with TMA and is associated with major complications [21], most patients were not treated with this modality. Discontinuation of tacrolimus and/or sirolimus was the primary means of treatment for TMA and three-quarters of the patients improved in this study. However, survival outcomes were significantly different between those who developed TMA and those who did not (Figure 2). The majority of patients who died without relapse after experiencing TMA had causes of death associated with GVHD/infections, many of which met the TMA definition in the context of multi-organ failure. Therefore one cannot conclude that the occurrence of TMA is an independent predictor of poor survival based on the current study.

Lastly our data should be interpreted with caution. The observed associations between sirolimus drug levels and TMA may be due to hidden confounding factors in this heterogeneous cohort. Although sirolimus levels were checked at trough levels and in a consistent manner in our program, there is inherent variability in the exact timing of the blood draws and patients’ compliance to instructions in an outpatient setting. To minimize this variability, we focused our multivariable analysis on the early post-HCT period (within 30 days), during which most drug levels were obtained in the inpatient setting. Our data indicate a dose-response relationship between use of sirolimus and the occurrence of TMA, and suggest that concurrent use of tacrolimus and sirolimus may exacerbate this effect. The currently accepted sirolimus therapeutic ranges may be too high, contributing to unacceptable toxicity in some patients, particularly those with concurrent GVHD or a genetic predisposition based on ADME genes.

Recommendations

Our institutional guidelines have been modified to maintain both sirolimus and tacrolimus levels in the 5-10 ng/ml range whenever they are used in combination. Previously, we allowed 3-12 ng/ml for sirolimus but have amended this range based on data from this study and Rodriguez et al. [7]. Additionally, in the event of fungal infection, if azoles are used, we recommend dose reduction of tacrolimus and sirolimus to prevent supra-therapeutic levels secondary to drug-drug interaction. For patients with active GVHD who develop TMA, based on the severity of TMA, we recommend dose adjustment or discontinuation of sirolimus or of both sirolimus and tacrolimus. Other immunosuppressive agents such as steroids and mycophenolate mofetil (MMF) are typically used in the event of discontinuation.