Biol Blood Marrow Transplant 25 (2019) 10221030 Biology of Blood and Marrow Transplantation journal homepage: www.bbmt.org Effect of Sirolimus Exposure on the Need for Preemptive Antiviral Therapy for Cytomeglovirus Infection after Allogeneic Hematopoietic Stem Cell Transplantation Beatriz Guglieri-Lopez1, Alejandro Perez-Pitarch2, Irene Garcia-Cadenas3, Estela Gimenez4, Pere Barba5, Nuria Rabella7, Juan Carlos Hernandez-Boluda6, Laura Fox5, David Valcarcel5, Albert Esquirol3, Rafael Ferriols-Lisart2, Jorge Sierra3, Carlos Solano6,8, David Navarro4,9, Rodrigo Martino3, Jose Luis Pin ~ ana10,11,* 1 Pharmacy Department, Hospital Universitario Doctor Peset, Valencia, Spain 2 Pharmacy Department, Hospital Clínico Universitario, Valencia, Spain 3 Department of Hematology, Hospital de la Santa Creu i Sant Pau, Barcelona, Spain 4 Microbiology Service, Hospital Clínico Universitario, Valencia, Spain 5 Department of Hematology, Hospital Universitari Vall D'Hebron, Barcelona. Spain 6 Department of Hematology, INCLIVA Research Institute, Hospital Clínico Universitario, Valencia. Spain 7 Department of Microbiology, Hospital de la Santa Creu i Sant Pau, Universitat Auto noma de Barcelona, Spain 8 Department of Medicine, School of Medicine, University of Valencia, Valencia, Spain 9 Department of Microbiology, School of Medicine, University of Valencia, Valencia, Spain 10 Department of Hematology, Hospital Universitari i Politecnic la Fe, Valencia. Spain 11 n Biomedica en Red Ca Centro de Investigacio ncer, Instituto Carlos III, Madrid, Spain Article history: A B S T R A C T Received 9 October 2018 The current study evaluates the clinical effect of sirolimus exposure on the occurrence of cytomegalovirus (CMV) Accepted 4 January 2019 DNAemia necessitating preemptive antiviral therapy. A total of 167 consecutive recipients of reduced-intensity conditioning (RIC) allogeneic hematopoietic stem cell transplantation (allo-HSCT) who received sirolimus- and Key Words: tacrolimus-based graft-versus-host disease (GVHD) prophylaxis and whose CMV serostatus was positive for Cytomegalovirus DNAemia donors and/or recipients were included in this multicenter retrospective study. A parametric model with consecu- Cytomegalovirus infection tive sirolimus blood levels describing the time to CMV DNAemia-RAT was developed using NONMEM version 7.4. Cytomegalovirus disease Overall, 122 of 167 patients (73%) were allografted from an unrelated donor, and the donor CMV-serostatus was Preemptive antiviral therapy negative in 51 cases (31%). Fifty-six recipients (34%) developed CMV DNAemia necessitating preemptive therapy, Sirolimus with a cumulative incidence of 36% at a median follow-up of 25 months. Time to CMV DNAemia necessitating Mechanistic target of rapamycin inhibitor preemptive therapy was best described using a Gompertz function. CMV DNAemia necessitating preemptive ther- Allogeneic hematopoietic stem apy-predicting factors were antithymocyte globulin-based conditioning regimen (hazard ratio [HR], 2.2; 95% con- cells transplantation fidence interval [CI], 1.1 to 4.1; P < .01) and sirolimus concentration (HR, .94; 95% CI, .87 to .99; P < .01). The risk Quantitative PCR of CMV DNAemia-RAT decreased by 6% for each 1 ng/mL increase in sirolimus trough concentration. In conclusion, Sirolimus exposure we provide evidence on the association between sirolimus blood concentration and incidence of CMV DNAemia Time-to-event analysis necessitating preemptive therapy in allo-HSCT recipients. Moreover, this study presents the first predictive model PK/PD describing the time to CMV DNAemia necessitating preemptive antiviral therapy as a function of sirolimus drug concentration. © 2019 American Society for Blood and Marrow Transplantation. INTRODUCTION disease (CMV-D) in organ transplantation (OT) in several ran- Sirolimus has been associated with a significant protective domized controlled clinical trials [1-4]. In contrast, this benefi- effect against active cytomegalovirus (CMV) infection and cial effect remains controversial in the allogeneic hematopoietic stem cell transplantation (allo-HSCT) setting [5-9]. OT and allo- HSCT procedures differ in many aspects, including lymphode- Financial disclosure: See Acknowledgments on page 1029. pleting conditioning chemotherapy, thymus dysfunction, and * Correspondence and reprint requests: Jose Luis Pin ~ ana, PhD, Division of Clinical Hematology, Hospital Universitario la Fe de Valencia, Avda Fernando dysregulation from the conditioning regimen. In addition, char- Abril Martorell, 106 CP, 46026 Valencia, Spain. acteristics related to sirolimus exposure and treatment duration, E-mail address: jlpinana@gmail.com (J.L. Pin~ ana). such as lower targeted immunosuppressant drug concentration, https://doi.org/10.1016/j.bbmt.2019.01.012 1083-8791/© 2019 American Society for Blood and Marrow Transplantation. B. Guglieri-Lopez et al. / Biol Blood Marrow Transplant 25 (2019) 10221030 1023 common concentration fluctuations (eg, variability in absorp- 100 days and monthly until 1 month after the withdrawal of immunosup- tion, due in part to frequent episodes of diarrhea; drug-drug pression. Patients who developed acute or chronic GVHD were monitored at each outpatient visit. At HCUV, CMV DNA load monitoring in plasma was per- interactions), and early tapering in allo-HSCT recipients, also formed using 2 PCR platforms: a Qiagen Real-Time PCR Kit (Qiagen, Hilde, vary widely [10]. In combination, these latter factors may con- Germany) from October 2008 to May 2012 and an Abbott Real-Time CMV tribute to reduced overall exposure to sirolimus, mitigating its PCR (Abbott Molecular, Des Plaines, IL) thereafter [7,14]. At HVH, CMV DNA potential anti-CMV effect in allo-HSCT compared with OT. load monitoring in plasma was performed using a RealStar CMV PCR system (Altona Diagnostics, Hamburg, Germany). At HSCSP, CMV DNA load monitor- We previously demonstrated that higher sirolimus expo- ing in plasma was done with the Affigene CMV trender (Cepheid, Solna, Swe- sure significantly reduced the risk of detectable CMV DNAemia den) from October 2008 to December 2012 and with the RealStar CMV PCR at any level after allo-HSCT [11]. Our next step was to explore Kit 1.2 (Altona Diagnostics) thereafter. whether higher sirolimus concentrations could also have the clinical benefit of reducing the number of CMV DNAemia Preemptive Anti-CMV Therapy and Definitions At HCUV and HVH, preemptive therapy was initiated on detection of CMV events necessitating preemptive antiviral therapy in allo-HSCT DNA levels >1500 IU/mL irrespective of the PCR assay used. Beginning in 2014 at recipients. With this aim, we developed a fully parametric HCUV, preemptive antiviral therapy was started either when CMV DNA load dou- time-to-event analysis taking into account the dependency bling time was 2 days or when CMV DNAemia reached >1500 IU/mL, which- and association between longitudinal data (ie, trough blood ever occurred first [15]. At HSCSP, preemptive therapy against CMV was started when the CMV DNAemia viral load reached >1000 UI/mL, whereas 3 consecutive concentration of sirolimus) and time-to-event data (ie, occur- samples with CMV DNAemia values ranging from 500 to 1000 UI/mL were rence of CMV DNAemia necessitating preemptive therapy). We required before to prompt the initiation of preemptive therapy. examined the pharmacokinetic (PK)-pharmacodynamic (PD) CMV DNAemia necessitating preemptive therapy was defined when the relationship between CMV DNAemia necessitating preemptive CMV DNAemia load exceeded the established trigger levels for preemptive anti- therapy and sirolimus exposure using consecutive trough viral therapy at each center. Preemptive anti-CMV therapy was based on ganci- clovir 5 mg/kg i.v. or foscarnet 90 mg/kg i.v. every 12 hours during the admission blood concentrations of sirolimus. Here we report the clinical (reserving foscarnet for patients with neutropenia) or with oral valganciclovir at effect of sirolimus exposure over time on the risk of developing a usual dose of 900 mg twice daily in the outpatient clinic. CMV-D was divided a first CMV DNAemia episode necessitating preemptive ther- into pneumonia, gastrointestinal disease, and retinitis based on consensus defini- apy in a multicenter series of consecutive reduced-intensity tions [16]. Acute GVHD (aGVHD) and chronic GVHD (cGVHD) were defined and graded based on preestablished criteria [17]. Transplantation-associated micro- conditioning (RIC) allo-HSCT recipients who received a siroli- angiopathy (TAM) was classified as confirmed or probable according to previ- mus and tacrolimus combination as graft-versus-host disease ously defined international criteria [18-20]. (GVHD) prophylaxis. Statistical Methods METHODS The primary objective of this study was to identify risk factors for a first Patients CMV DNAemia necessitating preemptive therapy episode, with special This multicenter study included 167 consecutive patients who under- emphasis on the potential effect of estimated exposure to sirolimus on this went RIC-allo-HSCT between October 23, 2008, and October 29, 2015, and complication. Secondary objectives were to analyze CMV DNAemia necessi- received sirolimus- and tacrolimus-based GVHD prophylaxis and whose CMV tating preemptive therapy and/or CMV-D characteristics and to evaluate the serostatus was positive for donors and/or recipients. The Institutional Review effect of CMV DNAemia necessitating preemptive therapy in nonrelapse mor- Board of each participating institution approved the study, and written tality (NRM) and overall survival. informed consent was obtained from all patients in accordance with the Dec- laration of Helsinki. The study was registered by the Spanish Agency of Medi- Nonparametric and Semiparametric Analyses cines and Health Products (reference code PIN-SIR-2016-01). The probabilities of post-transplantation events were estimated by the cumulative incidence method. Univariate analysis of the association between Conditioning Regimen and GVHD Prophylaxis clinical risk factors and post-transplantation outcomes were calculated taking Three RIC regimens were used in this study, as described previously [12]: competing events into account by using the Gray test [21,22]. Time-dependent fludarabine combined with melphalan (for lymphoid malignancies), fludara- covariates were analyzed using univariate and multivariate Cox regression mod- bine combined with oral (p.o.) or i.v. busulfan (for myeloid malignancies), els. Tests of significance were 2-sided, with statistical significance at P  .05. All and fludarabine with i.v. busulfan and thiotepa (for both myeloid and lym- nonparametric and semiparametric statistical analyses were performed using phoid malignancies). Antithymocyte globulin (ATG) was administered to SPSS version 20 (IBM, Armonk, NY) and R version 2.12.2 (R Project for Statistical transplant recipients with 1 HLA-mismatched (considering high-resolution Computing, Vienna, Austria) with survival v2.36-10, Design 2.3-0, prodlim v1.2.1, typing of HLA-A, -B, -C, and -DRB1) with either a sibling donor or an unre- and cmprsk v2.2-221packages. lated donor (URD) at doses of 6 mg/kg divided over 3 days. GVHD prophylaxis consisted of sirolimus at a dose of 6 mg/day p.o. on Parametric Analysis day -6, followed by 4 mg/day p.o. starting on day -5; tacrolimus was started A parametric survival model describing time to CMV DNAemia necessitat- on day -3 at a dose of .02 mg/kg/day as a continuous i.v. infusion or equiva- ing preemptive therapy after allo-HSCT was developed by means of nonlinear lent oral doses. In the absence of GVHD, sirolimus and tacrolimus doses were mixed-effects modeling using NONMEM version 7.4 (ICON Development Solu- tapered as described previously [10]. tions, Dublin, Ireland) [12,23]. The model was developed in 2 steps: a baseline model without any explanatory factors was developed, and the impact of the Sirolimus and Tacrolimus Blood Concentration Monitoring, Management, study variables was explored and included in the baseline model. and Technical Considerations Sirolimus and tacrolimus trough blood concentrations were monitored at least twice weekly for the first 4 weeks after transplantation or until dis- Development of the Baseline Model charge, then once a week until day +100 and thereafter at each outpatient The time to CMV DNAemia-RAT was calculated using the following parametric survival function: visit. Sirolimus levels were measured by liquid chromatography for separa- tion and tandem mass spectrometry at the Hospital Vall d'Hebron of Barce- Rt  hðt Þdt lona (HVH), by the semiautomated microparticle enzyme immunoassay s ðt Þ ¼ e 0 ; method (IMx sirolimus assay; Abbott Laboratories, Abbott Park, IL) at the where h(t) is the hazard and survival, S(t), is a function of the cumulative hazard Hospital de la Santa Creu i Sant Pau (HSCSP), and by chemiluminescent mag- within the time interval 0 and the time t describing the probability of not netic microparticle immunoassay (Architect i-System; Abbott Laboratories) experiencing CMV DNAemia necessitating preemptive therapy within this inter- at the Hospital Clínico Universitario of Valencia (HCUV). A cross-validation val. The baseline model was developed by exploring different functions for the study comparing sirolimus levels measured by the 2 laboratory methods hazard h(t), starting from a simple time-independent constant hazard and gradu- demonstrated a significant correlation (r .91) [13]. Doses were titrated to ally progressing to more complex functions, including Weibull and Gompertz. achieve the target blood concentrations of 5 to 12 ng/mL for sirolimus and For nested models, decisions about the best choice of model were made based on 5 to 10 ng/mL for tacrolimus. the likelihood ratio test principle. The difference in objective function value (DOFV) between a full and reduced model follows approximately a chi-square CMV DNA Monitoring and Load Quantitation distribution, with the number of degrees of freedom corresponding to the differ- At 3 centers, all patients at risk were routinely monitored at least twice ence in number of parameters between the 2 competing models. For a statistical weekly during hospital admission, and once weekly thereafter during the first significance level of P = .05 and 1 degree of freedom, a DOFV of -3.84 is required. 1024 B. Guglieri-Lopez et al. / Biol Blood Marrow Transplant 25 (2019) 10221030 Development of the Covariate Model Table 1 Potential covariates explored were patient, donor and transplanta- Patient Characteristics and Post-Transplantation Outcomes (n = 167) tion characteristics, transplantation center, sirolimus and tacrolimus trough concentrations, and post-transplantation events tested as time- Characteristic Value dependent covariates (ie, several combinations of GVHD; aGVHD grade I- Recipients age, yr, median (range) 59 (23-72) IV, grade II-IV, and grade III-IV; cGVHD; and a combination of moderate- Male recipient, n (%) 100 (60) to-severe GVHD) taking competitive events into account. Sirolimus and Previous lines of therapy, n (range) 3 (0-8) tacrolimus levels were censored at the time of either occurrence of CMV Previous allo-HSCT, n (%) 58 (35) DNAemia necessitating preemptive therapy or relapse of the baseline dis- HCT-CI, n (%) ease. Initially, the covariates were tested in a univariate manner; that is, 0 33 (20) each covariate relationship was evaluated on the baseline hazard sepa- 1-2 71 (42) rately. Based on the results, covariate inclusion was carried out using a 3 63 (38) stepwise covariate inclusion procedure. The significance level was set to Diagnosis, n (%) .05 for the forward inclusion approach and .01 for the backward deletion Acute leukemia/myelodysplastic syndrome 60 (36)/21 (13) approach. Covariates were incorporated into the model as shown Non-Hodgkin lymphoma/Hodgkin disease 42 (25)/16 (10) in the following equation: Multiple myeloma/chronic lymphocytic leukemia 10 (6)/7 (5) hðt Þj ¼ hðt Þ ¢ euCov ¢ Cov ; Other 11 (7) Disease status at allo-HSCT, n (%) where h(t)j represents the instantaneous hazard for an individual with covari- CR 103 (62) ate value Cov, h(t) represents the hazard function of patients with values PR 37 (22) equal to 0 (for continuous covariates) or in the reference groups (for categori- PROG 27 (16) cal covariates), and u represents the regression coefficient describing the Donor/recipient CMV serostatus influence of the covariate on the instantaneous hazard. HR can be calculated +/+ 106 (63) as euCov ¢ Cov . Drug concentration effect on hazard was explored using linear +/- 10 (6) and Emax models: -/+ 51 (31) hðt Þj ¼ hðt Þ ¢ expCdrug ¢ Eff Female donor to male recipient, n (%) 36 (22) Donor type and Related 45 (27) Emax ¢ Cdrug Unrelated 122 (73) hðt Þj ¼ hðt Þ ¢ exp C50 þCdrug ; HLA mismatch, n (%) 32 (19) Conditioning regimen, n (%) where Cdrug represents immunosuppressant drug concentration, Eff repre- Flu-Bu 50 (30) sents the effect of drug concentration on instantaneous hazard, Emax repre- Flu-Mel 73 (44) sents the maximum drug effect, and C50 represents the concentration that Flu-Bu-TT 44 (26) obtains one-half the maximum effect. ATG as a part of conditioning, n (%) 31 (19) To ensure that the models described the data adequately, internal valida- tion of the model was performed by visual predictive check (VPC) Kaplan- Peripheral blood source, n (%) 164 (98) Meier plots. The plots were based on simulations of 100 replicates of the CD34+ cell count, £ 106/kg, median (range) 6 (.5-13.8) study dataset. For the final model, a nonparametric bootstrap analysis of the Transplantation outcomes data with 1000 resampled datasets was performed, and the uncertainty (95% Days to myeloid recovery, median (range) confidence interval [CI]) of the parameter estimates was obtained from the Neutrophils >.5 £ 109/L 15 (8-27) distribution of the bootstrap estimates (2.5th and 97.5th percentiles). The Platelets >20 £ 109/L 12 (0-386) final model was then used to perform deterministic simulations of CMV aGVHD DNAemia necessitating preemptive therapy-free survival curves for increas- Cumulative incidence of aGVHD II-IV, 37 (30-45) ing sirolimus trough concentrations. % (95% CI) Day of onset, median (range) 65 (5-275) Cumulative incidence of aGVHD III-IV, 17 (12-23) % (95% CI) RESULTS Day of onset, median (range) 53 (6-202) Patient Characteristics cGVHD in evaluable patients Patients’ disease and transplantation characteristics and Cumulative incidence of cGHVD at 2 yr, 75 (66-84) % (95% CI) outcomes are summarized in Table 1. The median age of Day of onset, median (range) 188 (77-982) recipients was 59 years (range, 23 to 72 years), and 60% were Cumulative incidence of extensive cGHVD, 39 (20-50) male. The median duration of follow-up was 25 months % (95% CI) (range, 4 to 85 months). Most recipients (73%) were allog- Cumulative incidence of NRM, % (95% CI) At day +100 9 (4-13) rafted from an URD. Thirty-two patients (19%) had an HLA At day +180 12 (8-18) mismatch with the donor and received ATG as a part of condi- At 2 yr 21 (15-28) tioning. In 51 of 167 cases (31%), CMV serostatus was positive Cumulative incidence of TAM (95% CI) for the recipient and negative for the donor. For the whole Proven 9 (3-14) cohort, the cumulative incidence was 37% for aGVHD grade II- Probable-proven 17 (8-21) Cumulative incidence of SOS (95% CI) 4 (2-8) IV, 17% for aGVHD grade III-IV, 75% for overall cGVHD, and CMV 39% for extensive cGVHD. Cumulative incidence of CMV DNAemia-RAT at 33 (25-40) 24 wk, % (95% CI) Cumulative incidence of DNAemia-RAT at 2 yr, 36 (29-44) % (95% CI) CMV-DNAemia-RAT and CMV Disease Characteristics Cumulative incidence of CMV-D at 2 yr, 2 (0-5) Fifty-six of 167 patients (34%) developed CMV DNAemia % (95% CI) necessitating preemptive therapy at a median of 56 days Cumulative incidence of relapse at 2 yr, % (95% CI) 18 (12-25) (range, 12 to 535 days) after stem cell infusion. The 24-week DFS at 2 yr, % (95% CI) 61 (58-65) OS at 2 yr, % (95% CI) 68 (64-72) and 2-year cumulative incidences of CMV DNAemia necessitat- Follow-up for survivors, mo, median (range) 25 (4-85) ing preemptive therapy were 33% (95% CI, 25% to 40%) and 36% (95% CI, 28% to 44%), respectively. The 2-year cumulative inci- HCT-CI indicates hematopoietic cell transplantation-specific comorbidity index; CR, first complete remission; PR, partial remission; PROG, nonre- dence of CMV-D was 2% (95% CI, 0 to 5%). Three patients (2%) sponder or progression before RIC-allo-HSCT; Flu, fludarabine; Bu, busulfan; developed CMV-D at 120, 200, and 201 days after stem cell Mel, melphalan; TT, thiotepa; SOS, sinusoidal obstruction syndrome of the infusion. liver: CMV-D, CMV disease: DFS, disease-free survival. B. Guglieri-Lopez et al. / Biol Blood Marrow Transplant 25 (2019) 10221030 1025 Risk Factors for CMV DNAemia Necessitating Preemptive preemptive therapy model included the use of ATG and siro- Antiviral Therapy limus exposure as statistically significant predictors of CMV Nonparametric and Semiparametric Analyses DNAemia necessitating preemptive therapy. The risk of Table 2 shows the univariate and multivariate risk factor experiencing CMV DNAemia necessitating preemptive ther- analyses for CMV DNAemia necessitating preemptive therapy. apy was twice the baseline risk in patients who received ATG In multivariate Cox model analysis, ATG use, presence of a lym- as part of conditioning regimen (HR, 2.2; 95% CI, 1.1 to 4.1). phoid malignancy, and undergoing allo-HSCT at the HSCSP With regard to sirolimus exposure, the risk of experiencing transplantation center were associated with a risk of CMV CMV DNAemia necessitating preemptive therapy decreased DNAemia necessitating preemptive therapy. linearly by 6% per each 1 ng/mL increase in sirolimus trough concentrations (HR, .94; 95% CI, .87 to .99). Tacrolimus expo- Parametric Analyses sure had no influence on the risk of CMV DNAemia necessi- For the parametric time-to-event model, a total of 3289 tating preemptive therapy. sirolimus blood samples throughout a median of 140 days Regarding model validation, the VPC Kaplan-Meier plots (range, 7 to 535 days) after transplantation were available for (Figure 2) proved an adequate performance of the model analysis. Sirolimus withdrawal was done in 98 recipients (58%) given that the observed survival curve was situated inside the at a median of 217 days (range, 1 to 2088 days), whereas tacro- simulation prediction intervals. In addition, the bootstrap limus was withdrawal in 148 patients (88%) at median of results reinforced the robustness of the model because final 100 days (range, 1 to 1237 days) after stem cells infusion. Dif- model parameter estimates were inside the CIs obtained from ferences over time in median and range values of sirolimus bootstrap reestimations, as shown in Table 3. Furthermore, and tacrolimus levels according to transplantation center and the statistical significance of the effect of sirolimus exposure baseline disease are illustrated in Figure 1A and B, respectively. on CMV DNAemia necessitating preemptive therapy-free sur- The OFV for the main structural models tested in the non- vival was reinforced because the 95% CI of the reestimated linear model building procedure were 860.7 for the exponen- values for this parameter did not include the 0 value (-.135 to tial survival function with 1 parameter, 837.9 for the Weilbull -.001). This was also true for the parameter representing ATG function with 2 parameters, and 809.1 for the Gompertz influence. The influence of sirolimus exposure on the proba- function with 2 parameters. Finally, the time to CMV DNAe- bility of not experiencing CMV DNAemia necessitating pre- mia necessitating preemptive therapy was best described by emptive therapy is represented through deterministic a Gompertz function with regard to OFV and Kaplan-Meier simulations in Figure 3. A subgroup analysis through nonlin- VPC plots. After the stepwise covariate inclusion procedure, ear mixed-effects modeling, fitting the final model to data the retained final time to CMV DNAemia necessitating from each center separately and checking the estimated effect Table 2 Univariate and Multivariate Risk Factor Analyses of CMV DNAemia Necessitating Preemptive Antiviral Therapy at a Median Follow-Up of 25 Months P Variable CMV DNAemia-RAT Multivariate Analysis, Cox Regression Cumulative Incidence, % (95% CI) P Value HR (95% CI) P Value Disease Myeloid 26 (16-36) .01 Lymphoid 45 (34-56) 1.8 (1.01-3.1) .05 ATG Yes 51 (33-70) .01 2.2 (1.2-3.99) .009 No 33 (24-41) Conditioning regimen* Flu-Bu 24 (12-36) Flu-Mel 43 (31-55) .10 NS TBF 37 (22-52) Donor/recipient sex mismatch Female donor to female recipient 14 (0-29) .02 NS Others 40 (31-48) CMV serostatus R+/D- 45 (31-59) R+/D+ 35 (26-45) .03 NS R-/D+ 0 Previous lines of therapy <3 31 (21-41) .10 NS 3 42 (31-54) Previous aGVHD grade II-IVy 1.36 (0.91-2) .13 Previous aGVHD grade III-IVy 1.37 (0.93-1.9) .10 NS Transplantation center HCUV 49 (35-63) 1 HSCSP 25 (14-35) .03 0.45 (0.23-0.92) .03 HVH 39 (23-55) 0.8 (0.4-1.6) .50 D indicates donor; R, receptor; NS, not significant; TBF, fludarabine with i.v. busulfan and thiotepa. Other tested covariates that did not achieve statistical significance included donor type (related versus unrelated), recipient age, CD34+ cell count, HCT-CI, and previ- ous autologous HSCT. * Conditioning was not entered into the multivariate analysis because it showed collinearity with the transplantation center variable, because all recipients at HVH were conditioned with TBF, whereas Flu-Bu was used at HCUV and Flu-Mel was used at HSCSP. y Analyzed as time-dependent covariates. Results are expressed as HR from a univariate Cox regression model. 1026 B. Guglieri-Lopez et al. / Biol Blood Marrow Transplant 25 (2019) 10221030 of sirolimus on the risk of CMV reactivation was performed sample size, but indicated a numerical benefit (risk reduc- (data not shown). All 3 subgroup analyses demonstrated a tion). In addition, a subgroup analysis was performed in numerical benefit (risk reduction) for higher exposures to donor-negative/recipient-positive CMV serostatus pairs that sirolimus, and 2 of them estimated this parameter with suffi- also demonstrated a numerical benefit (risk reduction) for cient precision to prove statistical significance. Only 1 sub- greater exposures to sirolimus. However, the results of this group (that with the smallest number of patients) did not subgroup analysis were not statistically significant, likely due prove statistical significance, most likely due to reduced to the small sample size. Figure 1. Sirolimus and tacrolimus levels by transplantation center (A) and baseline disease (B). A total of 3289 sirolimus blood samples at a median of 140 days (range, 7 to 535 days) after transplantation were available. Of these, 1033 sirolimus trough concentrations were available after 100 days, corresponding to 107 of the 167 patients included in the analysis. B. Guglieri-Lopez et al. / Biol Blood Marrow Transplant 25 (2019) 10221030 1027 Figure 1 Continued. NRM, Relapse, and Survival the development of CMV DNAemia necessitating preemptive Thirty-one patients (19%) died due to NRM at a median therapy was not associated with NRM (HR, 1.5; 95% CI, .8 time of 124 days (range, 13 to 652 days). The +100-day, +180- to 3.2; P = .2) or with lower relapse incidence (HR, .5; 95% CI, day, and 2-year cumulative incidence of NRM for the whole .2 to 1.2; P = .10). Patients who developed CMV DNAemia group was 9% (95% CI, 4% to 13%), 12% (95% CI, 8% to 18%), and necessitating preemptive therapy had a 2-year overall survival 21% (95% CI, 15% to 28%), respectively. In univariate analyses, of 65%, compared with 70% for those who did not (P = .20). 1028 B. Guglieri-Lopez et al. / Biol Blood Marrow Transplant 25 (2019) 10221030 Figure 2. Kaplan-Meier visual predictive check plot. The blue line represents observed CMV DNAemia-RAT-free survival data, and the green area represents the 95% CI of 200 simulated CMV DNAemia-RAT-free survival curves based on the final model. Table 3 The fact that sirolimus exposure could be a key element Final Model Parameter Estimates and Bootstrap Results for anti-CMV activity merits several considerations. First, it Parameter Value RSE, % Bootstrap Results may partially help to explain the inconsistencies in the anti- CMV effect observed between OT and allo-HSCT studies [1-9]. Median 95% CI (Percentile) In the allo-HSCT setting, sirolimus exposure is lower in terms λ0 .006 31 .006 .003-.011 of targeted blood concentrations and duration. Second, in the b -.009 19 -.009 (-.013)-(-.006) context of narrow therapeutic levels, we observed that vari- uATG .793 39 .805 .110-1.418 ability of sirolimus exposure among centers, among diseases uSIRO -.063 50 -.067 (-.135)-(-.001) (Figure 1), and likely among studies could be common. This RSE indicates relative standard error; λ0, baseline hazard; b, coefficient from fact may have distorted the ability to perceive an anti-CMV Gompertz function; uATG, regression coefficient for ATG; uSIRO, regression coef- ficient for sirolimus concentration effect. effect of sirolimus in allo-HSCT studies [5-9]. In this regard, exposure-response analyses emerge as a valuable tool that could deal with discrepant observations analyzing the effect DISCUSSION of exposure variability on the outcome. Variability in expo- This study found a 2-year incidence of CMV DNAemia necessi- sure could result in different anti-CMV activity patterns, as tating preemptive therapy of 36% in a multicenter cohort of RIC- has occurred with other anti-CMV drugs [30-32]. Third, our allo-HSCT recipients at very high risk of CMV infection (73% of data provide a rationale for exploring higher sirolimus tar- URDs and 31% of CMV-seronegative donors) given sirolimus-based geted levels, similar to those used in OT. Although sirolimus GVHD prophylaxis. By means of a fully parametric survival data blood concentrations in allo-HSCT should be increased with analysis using a detailed drug concentration dataset, we found evi- caution due to increased safety risks, knowledge from OT pro- dence of a significant inverse linear relationship between sirolimus spective studies comparing calcineurin inhibitors and siroli- concentration over time and the risk of developing CMV DNAemia mus have shown comparable safety profiles with targeted necessitating preemptive therapy, even when targeting a narrow levels of 15 to 20 ng/mL [33]. To explore the safety of greater therapeutic range (5 to 12 ng/mL). We also identified ATG use as a sirolimus exposure with regard to the development of TAM, risk factor, associated with a 2-fold increase in the baseline risk of in the present study we also performed an exploratory non- CMV DNAemia necessitating preemptive therapy. linear mixed-effects analysis of the impact of sirolimus and We previously observed that higher sirolimus concentrations tacrolimus concentrations on the incidence of TAM. This had an inhibitory effect on CMV replication by reducing the proba- exploratory analysis has not found a statistically significant bility of detectable CMV DNAemia at any level after RIC-allo-HSCT relationship between sirolimus and tacrolimus concentra- [11]. The present study confirms that higher sirolimus exposure tions and the incidence of overall and proven TAM (data not also translates into a clinical benefit by reducing the probability of shown). However, adverse events are common and should be preemptive anti-CMV therapy in a larger multicenter cohort of managed with increased awareness and close monitoring of RIC-allo-HSCT recipients. Our findings are supported by previous trough blood concentrations [34]. The lack of prospective in vitro models demonstrating the role of mechanistic target of PK/PD efficacy and safety studies comparing different tar- rapamycin signaling pathways in CMV protein synthesis [24-29] geted doses of sirolimus in terms of GVHD prophylaxis and and the potent inhibition on CMV replication exerted by sirolimus CMV infection limits the potential to optimize its use in the [24,27] in an inverse concentration-dependent manner [27]. allo-HSCT setting. B. Guglieri-Lopez et al. / Biol Blood Marrow Transplant 25 (2019) 10221030 1029 Figure 3. Simulated CMV DNAemia-RAT-free survival curves at various sirolimus exposure levels. Simulations were based on patients who did not receive an ATG- based conditioning regimen and were exposed to sirolimus trough concentrations ranging between 4 and 20 ng/mL. The mixed-effects model was applied across the data as a whole. We also performed a subanalysis for each individual center that confirmed the findings of the primary analysis. Another relevant issue that arises from our study is the optimization of sirolimus exposure (setting the lower limit of discrepant results observed in the 2 multivariate statistical mod- the therapeutic range at 8 ng/mL) could facilitate clinical els. Although ATG use as part of a conditioning regimen was research on CMV prophylaxis by exploring potential synergisms consistently identified as a risk factor for CMV DNAemia necessi- between sirolimus and the novel anti-CMV drugs in allo-HSCT. tating preemptive therapy in both statistical methods (paramet- Finally, we acknowledge that our study has several limita- ric and semiparametric), lymphoid malignancies and tions, including its retrospective nature, the use of different transplantation center were identified as risk factors only in the PCR tests among centers (all of which were calibrated to the multivariate Cox regression model. ATG is a strong and well- first World Health Organization international standard), differ- known risk factor for CMV reactivation because it produces a sig- ences in CMV DNA loads prompting the initiation of preemp- nificant delay in T cell immune reconstitution [35,36]. Thus, it is tive anti-CMV therapy (although they all used relatively high not surprising that regardless of sirolimus exposure, ATG had a CMV DNA level cutoffs triggering antiviral inception) and dif- strong effect on the development of CMV DNAemia necessitating ferent laboratory methods to estimate sirolimus levels. None- preemptive therapy. In contrast, lymphoid malignancies and theless, this study has some strengths that merit transplantation center showed different sirolimus exposure consideration. First, our findings confirm, in a larger and multi- (Figure 1). These data suggest that, irrespective of the preemptive center cohort, our previous observation in a smaller and anti-CMV therapy strategy and transplantation characteristics, homogenous (in terms of PCR test and preemptive anti-CMV sirolimus exposure may outweigh pretransplantation conditions therapy strategy) series of RIC-allo-HSCT, that sirolimus expo- that could have a marginal effect on CMV DNAemia necessitating sure matters to perceive an anti-CMV effect. In addition, the preemptive therapy in conventional analyses. Although there are implementation of fully parametric PK/PD analyses with a no well-defined cutoff and range of sirolimus levels in terms of large dataset and the homogenous GVHD prophylaxis offer anti-CMV activity, we previously reported in a sensitivity analysis certain advantages, such as more efficient parameter estimates that sirolimus blood concentration >8 ng/mL significantly compared with the Cox regression model [38]. reduced the risk of CMV DNAemia, whereas blood concentration In conclusion, our pharmacometric analysis supports the <8 ng/mL were not significantly protective [11]. Interestingly, at idea that sirolimus may have a clinically relevant exposure- HSCSP, the median trough blood concentration of sirolimus was dependent anti-CMV effect in the allo-HSCT setting. The 8 ng/mL during most of the study period (13 out of 20 weeks; identification of sirolimus concentration as a key element of 65%). In contrast, sirolimus blood concentrations were less fre- anti-CMV activity should facilitate further prospective PK/ quently 8 ng/mL in the other 2 centers (HCUV, 7 out of 20 PD studies exploring the value of increasing sirolimus blood weeks [35%]; HVH, 5 out of 20 weeks [25%]). concentration levels in allo-HSCT. A recent in vitro cell culture model also showed that rapa- mycin at concentration of 8 nM inhibited CMV replication by up ACKNOWLEDGMENTS to 60%. Remarkably, a much greater inhibition (>90% reduction) Financial disclosure: P.B. was supported by grants from the was observed when combined with maribavir, making this Instituto de Salud Carlos III (FIS16/01433) and PERIS 2018- combination appealing for CMV prophylaxis [37]. The 2020 from Generalitat de Catalunya (BDNS357800). 1030 B. Guglieri-Lopez et al. / Biol Blood Marrow Transplant 25 (2019) 10221030 Conflict of interest statement: There are no conflicts of interest 16. Ljungman P, Boeckh M, Hirsch HH, et al. Definitions of cytomegalovirus to report. infection and disease in transplant patients for use in clinical trials. Clin Infect Dis. 2017;64:87–91. Authorship statement: J.L.P., B.G.-L., A.P.-P., I.G.-C., P.B., J.C.H.- 17. Glucksberg H, Storb R, Fefer A, et al. Clinical manifestations of graft-versus- B., A.E., M.L.F., D.N., and R.M. were involved in study concep- host disease in human recipients of marrow from HL-A-matched sibling tion and design; J.L.P., B.G.-L., A.P.-P., E.G., N.R., R.F.L., J.S., C.S., donors. Transplantation. 1974;18:295–304. 18. Ruutu T, Barosi G, Benjamin RJ, et al. Diagnostic criteria for hematopoi- D.N., and R.M. were involved in data analysis and interpreta- etic stem cell transplant-associated microangiopathy: results of a con- tion; J.L.P., B.G.-L., A.P.-P., D.N., and R.M. drafted the manu- sensus process by an International Working Group. Haematologica. 2007; script; and J.L.P., B.G.-L., A.P.-P., I.G.-C., P.B., J.C.H.-B., A.E., M.L.F., 92:95–100. 19. Ho VT, Cutler C, Carter S, et al. Blood and marrow transplant clinical trials E.G., N.R., R.F.L., J.S., C.S., D.N., and R.M. performed a critical network toxicity committee consensus summary: thrombotic microangi- review for important intellectual content and approved the opathy after hematopoietic stem cell transplantation. Biol Blood Marrow final manuscript. Transplant. 2005;11:571–575. 20. Cho BS, Yahng SA, Lee SE, et al. Validation of recently proposed consensus criteria for thrombotic microangiopathy after allogeneic hematopoietic REFERENCES stem-cell transplantation. Transplantation. 2010;90:918–926. 1. Nashan B, Gaston R, Emery V, et al. Review of cytomegalovirus infection 21. Gooley T, Leisenring W, Crowley J, Storer BE. Estimation of failure proba- findings with mammalian target of rapamycin inhibitor-based immuno- bilities in the presence of competing risks: new representations of old suppressive therapy in de novo renal transplant recipients. Transplantation. estimators. Stat Med. 1999;18:665–706. 2012;93:1075–1085. 22. Fine JP, Gray RJ. A proportional hazards model for subdistribution of a 2. Ekberg H, Tedesco-Silva H, Demirbas A, et al. Reduced exposure to cal- competing risk. J Am Stat Assoc. 1999;94:496–509. cineurin inhibitors in renal transplantation. N Engl J Med. 2007; 23. Beal SL, Sheiner LB. NONMEM User's Guide. 3rd ed. University of California, 357:2562–2575. San Francisco: NONMEM Project Group; 1989. 3. Webster AC, Lee VW, Chapman JR, Craig JC. Target of rapamycin inhibitors 24. Kudchodkar SB, Yu Y, Maguire TG, Alwine JC. Human cytomegalovirus (sirolimus and everolimus) for primary immunosuppression of kidney infection induces rapamycin-insensitive phosphorylation of downstream transplant recipients: a systematic review and meta-analysis of random- effectors of mTOR kinase. J Virol. 2004;78:11030–11039. ized trials. Transplantation. 2006;81:1234–1248. 25. Kudchodkar SB, Yu Y, Maguire TG, Alwine JC. Human cytomegalovirus 4. Demopoulos L, Polinsky M, Steele G, et al. Reduced risk of cytomegalovirus infection alters the substrate specificities and rapamycin sensitivities of infection in solid organ transplant recipients treated with sirolimus: a raptor- and rictor-containing complexes. Proc Natl Acad Sci U S A. pooled analysis of clinical trials. Transplant Proc. 2008;40:1407–1410. 2006;103:14182–14187. 5. Marty FM, Bryar J, Browne SK, et al. Sirolimus-based graft-versus-host dis- 26. Moorman NJ, Shenk T. Rapamycin-resistant mTORC1 kinase activity is ease prophylaxis protects against cytomegalovirus reactivation after allo- required for herpesvirus replication. J Virol. 2010;84:5260–5269. geneic hematopoietic stem cell transplantation: a cohort analysis. Blood. 27. Poglitsch M, Weichhart T, Hecking M, et al. CMV late phase-induced mTOR 2007;110:490–500. activation is essential for efficient virus replication in polarized human 6. Cutler C, Kim HT, Hochberg E, et al. Sirolimus and tacrolimus without macrophages. Am J Transplant. 2012;12:1458–1468. methotrexate as graft-versus-host disease prophylaxis after matched 28. Moorman NJ, Cristea IM, Terhune SS, Rout MP, Chait BT, Shenk T. Human related donor peripheral blood stem cell transplantation. Biol Blood Mar- cytomegalovirus protein UL38 inhibits host cell stress responses by antag- row Transplant. 2004;10:328–336. onizing the tuberous sclerosis protein complex. Cell Host Microbe. 2008; 7. Solano C, Gime nez E, Pin ~ ana JL, et al. Impact of cytomegalovirus DNAemia 3:253–262. on overall and non-relapse mortality in allogeneic stem cell transplant 29. Terhune S, Torigoi E, Moorman N, et al. Human cytomegalovirus UL38 recipients. Transpl Infect Dis. 2017;19:e12717. protein blocks apoptosis. J Virol. 2007;81:3109–3123. 8. To€rlen J, Ringde n O, Garming-Legert K, et al. A prospective randomized 30. Marty FM, Winston DJ, Rowley SD. CMX001-201 Clinical Study Group. trial comparing cyclosporine/methotrexate and tacrolimus/sirolimus as CMX001 to prevent cytomegalovirus disease in hematopoietic-cell trans- graft-versus-host disease prophylaxis after allogeneic hematopoietic stem plantation. N Engl J Med. 2013;369:1227–1236. cell transplantation. Haematologica. 2016;101:1417–1425. 31. Chemaly RF, Ullmann AJ, Stoelben S, et al. Letermovir for cytomegalovirus 9. Cutler C, Logan B, Nakamura R, et al. Tacrolimus/sirolimus vs tacrolimus/ prophylaxis in hematopoietic-cell transplantation. N Engl J Med. 2014; methotrexate as GVHD prophylaxis after matched, related donor alloge- 370:1781–1789. neic HCT. Blood. 2014;124:1372–1377. 32. Winston DJ, Young JA, Pullarkat V, et al. Maribavir prophylaxis for prevention 10. Perez-Simo  n JA, Martino R, Parody R, et al. The combination of sirolimus of cytomegalovirus infection in allogeneic stem cell transplant recipients: a plus tacrolimus improves outcome after reduced-intensity conditioning, multicenter, randomized, double-blind, placebo-controlled, dose-ranging unrelated donor hematopoietic stem cell transplantation compared with study. Blood. 2008;111:5403–5410. cyclosporine plus mycofenolate. Haematologica. 2013;98:526–532. 33. Larson TS, Dean PG, Stegall MD, et al. Complete avoidance of calcineurin 11. Pin~ ana JL, PerezPitarch A, GuglieriLopez B, et al. Sirolimus exposure inhibitors in renal transplantation: a randomized trial comparing siroli- and the occurrence of cytomegalovirus DNAemia after allogeneic hemato- mus and tacrolimus. Am J Transplant. 2006;6:514–522. poietic stem cell transplantation. Am J Transplant. 2018;18:2885–2894. 34. Ventura-Aguiar P, Campistol JM, Diekmann F. Safety of mTOR inhibitors 12. Pin~ ana JL, Perez-Pitarch A, Garcia-Cadenas I, et al. A time-to-event model in adult solid organ transplantation. Expert Opin Drug Saf. 2016;15:303– for acute kidney injury after reduced-intensity conditioning stem cell 319. transplantation using a tacrolimus- and sirolimus-based graft-versus-host 35. Thomson BG, Robertson KA, Gowan D, et al. Analysis of engraftment, disease prophylaxis. Biol Blood Marrow Transplant. 2017;23:1177–1185. graft-versus-host disease, and immune recovery following unrelated 13. Schmid RW, Lotz J, Schweigert R, et al. Multi-site analytical evaluation of a donor cord blood transplantation. Blood. 2000;96:2703–2711. chemiluminescent magnetic microparticle immunoassay (CMIA) for siroli- 36. Mohty M, Bay JO, Faucher C, et al. Graft-versus-host disease following allo- mus on the Abbott ARCHITECT analyzer. Clin Biochem. 2009;42:1543–1548. geneic transplantation from HLA-identical sibling with antithymocyte 14. Gime nez E, Mun ~ oz-Cobo B, Solano C, Amat P, Navarro D. Early kinetics of globulin-based reduced-intensity preparative regimen. Blood. 2003; plasma cytomegalovirus DNA load in allogeneic stem cell transplant recip- 102:470–476. ients in the era of highly sensitive real-time PCR assays: does it have any 37. Chou S, Ercolani RJ, Derakhchan K. Antiviral activity of maribavir in combi- clinical value? J Clin Microbiol. 2014;52:654–656. nation with other drugs active against human cytomegalovirus. Antiviral 15. Solano C, Gime nez E, Pin~ ana JL, et al. Preemptive antiviral therapy for CMV Res. 2018;157:128–133. infection in allogeneic stem cell transplant recipients guided by the viral 38. Ibrahim JG, Chu H, Chen LM. Basic concepts and methods for joint models doubling time in the blood. Bone Marrow Transplant. 2016;51:718–721. of longitudinal and survival data. J Clin Oncol. 2010;28:2796–2801.