Cytomegalovirus Breakthrough and Resistance During Letermovir Prophylaxis

Abstract

Letermovir is a relatively new antiviral for prophylaxis against cytomegalovirus (CMV) after allogeneic hematopoietic cell transplantation (HCT). CMV-seropositive HCT recipients who received letermovir prophylaxis from 2018–2020 at our center were evaluated for letermovir resistance and breakthrough CMV reactivation. Two-hundred twenty-six letermovir recipients were identified and 7/15 (47%) with CMV DNAemia ≥ 200 IU/mL were successfully genotyped for UL56 resistance. A single C325Y resistance mutation was identified in an umbilical cord blood recipient. Ninety-five (42%), 43 (19%), and 15 (7%) patients had breakthrough CMV at any level, ≥ 150 IU/mL, and ≥ 500 IU/mL, respectively. Risk factors for breakthrough CMV reactivation at each viral threshold were examined. Cumulative steroid exposure was the strongest risk factor for CMV at all evaluated viral thresholds. Graft-versus-host disease prophylaxis with post-transplantation cyclophosphamide (aHR 2.34, 95% CI 1.28–4.28, p=0.001) or calcineurin inhibitors plus mycophenolate (aHR 2.24, 95% CI 1.30–3.86, p=0.004) were also associated with an increased risk of CMV reactivation at any level. De novo letermovir resistance is rare and can be successfully treated using other antivirals. Letermovir effectively prevents clinically significant CMV, however, subclinical CMV reactivation occurs frequently at our center.

Introduction

Letermovir is the first antiviral in decades to be approved for prophylaxis against clinically significant CMV after HCT based on the results of a phase III, randomized, double-blind, placebo-controlled trial.(1) Letermovir has a favorable side-effect profile compared to older antivirals (i.e., ganciclovir, foscarnet),(2) including the absence of associated cytopenias or renal injury.(1, 3) Letermovir prevents CMV replication by binding to the UL56 component of CMV terminase, a hetero-oligomer complex that cleaves concatemeric viral DNA into unit-length genomes as it is packaged into viral capsids.(3–8) Following FDA approval in the United States, letermovir became standard prophylaxis for all adult, CMV-seropositive allogeneic HCT recipients at the Fred Hutchinson Cancer Center (Fred Hutch) in 2018.

A potential challenge that healthcare providers may face is the emergence of CMV letermovir resistance. Letermovir resistance is characterized by nonsynonymous changes in CMV UL56 in contrast to mutations in UL54 and UL97 that confer ganciclovir resistance (Supplemental Figure 1). Specifically, UL56 mutations at amino acid (aa) positions 236, 257, 325, and 329 confer >3,000-fold antiviral resistance, and C325Y can confer >5,000-fold antiviral resistance.(9) Mutations in UL51 and UL89 genes also confer letermovir resistance when combined with mutations in CMV UL56.(9, 10). Among patients from the original phase III trial with evaluable genotypic data, 5/8 (63%) patients were found to have mutations in UL56 conferring letermovir resistance(11, 12) and current real-world data highlight the need for rapid CMV UL56 genotyping.(13–15)

To determine the incidence of de novo letermovir resistance at our center, we performed a retrospective analysis of all CMV-seropositive allogeneic HCT recipients who received letermovir prophylaxis following its adoption as standard CMV prophylaxis. Patients with breakthrough CMV reactivation ≥ 200 IU/mL were sequenced using a Sanger-based CMV UL56 resistance assay developed at the University of Washington (UW) Virology Clinical Laboratory. Finally, we aimed to characterize possible clinical and transplantation risk factors for breakthrough CMV infection during letermovir prophylaxis.

Materials and Methods

Study population and patient samples.

All adult CMV-seropositive allogeneic HCT recipients who received letermovir prophylaxis at our center from October 2018 through October 2020 were considered for the current study. Letermovir prophylaxis was initiated per institutional standard practice at post-HCT day 8 after “high-risk” non-cord HCT, Day 1 after “high-risk” cord HCT, or at engraftment after “low-risk” HCT (Supplemental Figure 2).(2, 16). Letermovir prophylaxis initiation is delayed or given intravenously if patients have poor oral intake or are experiencing severe diarrhea. In addition, letermovir is dose adjusted or held altogether if significant drug-drug interactions are present. All patients had a documented negative CMV DNA PCR test prior to letermovir initiation.

Baseline clinical and transplantation data were extracted from the electronic medical record. Weight-based, prednisone-equivalent steroid use in the first 100 days post-HCT were recorded on days when steroids were started/stopped, or when the daily dose was increased/decreased. Cumulative steroid exposure was calculated as the area under the curve (AUC) of weight-based, prednisone-equivalent steroid dose (mg/kg) using the trapezoid rule.(16) For regression analyses, only cumulative steroid AUC up to the day of CMV reactivation was considered.

Patients were tested for CMV UL56 resistance using stored plasma from the Fred Hutch Vaccine and Infectious Diseases Division (VIDD)/Infectious Diseases Sciences (IDS) sample repository. The study was approved by our Institutional Review Board; all participants provided informed consent.

CMV surveillance and management.

Before November 2019, CMV-seropositive allogeneic HCT recipients underwent weekly virologic surveillance in the first 100 days post-HCT using a CMV DNA PCR assay developed at the UW Virology Clinical Laboratory with a limit of detection (LoD) and limit of quantitation (LoQ) of 7.5 IU/mL and 25 IU/mL, respectively.(17) After November 2019, CMV monitoring was performed using the Abbott RealTime Assay with a LoD of 31.2 IU/mL and LoQ of 50 IU/mL.(18) Inter-assay variability for both assays was previously shown uninfluenced by high quantitative levels of CMV DNA and less variation was noted when reported by the same laboratory (intra-laboratory variability).(17) We have compared both assays and observed a difference of ≤ 0.5 log₁₀ across a dynamic range of CMV DNA PCR levels.(19)

Letermovir recipients with clinically significant CMV in the first 100 days post-HCT are treated preemptively with ganciclovir, valganciclovir, or foscarnet according to institutional treatment thresholds in Supplemental Figure 2.(2) Clinically significant CMV in the first 100 days post-HCT for “high-risk” and “low-risk” HCT is defined as CMV DNA PCR of ≥ 150 IU/mL and ≥ 500 IU/mL, respectively. Patients are not routinely restarted on letermovir prophylaxis following completion of preemptive antiviral therapy.

CMV UL56 Resistance Assay Sensitivity and Optimization.

Details of our CMV UL56 resistance assay including sensitivity and optimization analyses are described in the supplemental methods.

For the current study, HCT recipients with CMV DNA PCR ≥ 200 IU/mL during letermovir prophylaxis had letermovir resistance testing. Samples were selected for sequencing upon reaching a threshold of CMV DNA PCR ≥ 200 IU/mL and not necessarily at the onset of CMV reactivation.

Statistical Analysis.

CMV terminase complex components (i.e., UL51, UL56, UL89) were visualized using I-TASSER software which generates and ranks most likely protein configurations.(20) Protein configurations with the highest C-score were selected for UL51, UL56, and UL89 and amino acid accessions UTO68723.1, ABB71143.1, and APA46110.1 were used, respectively. The weekly proportions and cumulative incidence frequencies of CMV reactivation in the first 100 days post-HCT were calculated for any CMV DNAemia, ≥ 150 IU/mL, and ≥ 500 IU/mL. For any CMV DNAemia, a positive CMV event was defined as any CMV DNA PCR level greater than the LoD the CMV DNA PCR assay used. Weekly proportions were calculated as the total weekly incidence of CMV at each viral threshold divided by the total number of patients alive that week. Cumulative incidence plots were created using the ‘cmprsk’ package in the R statistical computing environment, version 3.5.0. Univariable Cox regression was performed to identify associations of a priori selected variables with CMV at each viral threshold. All tests were 2-sided, and variables with p-values ≤ 0.05 were considered statistically significant and subsequently used to construct multivariable models. Death was treated as a competing risk in all models. All statistical analyses were performed using SAS 9.4 for Windows (SAS Institute, Cary, NC).

Data Sharing Statement.

For original data, please contact dzamora2@fredhutch.org

Results

Study population.

A total of 226/266 (85%) allogeneic HCT recipients were identified as having received letermovir prophylaxis at our center from October 2018 through October 2020 and their baseline characteristics are shown (Table 1). Four patients underwent at least two separate HCT within the period of interest, but only HCT where patients received letermovir prophylaxis were considered for analyses. Baseline characteristics according to HCT risk are listed in Supplemental Table 2.

Table 1 –

Patient/Transplant Characteristics

Variables	Categories	Total (N=226)
Age	Median (IQR)	55.9 (41.0–63.3)
Age (categorical)	≤20	6 (3%)
	21–60	144 (64%)
	>60	76 (34%)
Sex	Male	116 (51%)
	Female	110 (49%)
Race	Caucasian	177 (78%)
	Non-Caucasian	49 (22%)
Ethnicity	Hispanic or Latino	21 (9%)
	Not Hispanic or Latino	196 (87%)
	Unknown/Not reported	9 (4%)
Underlying disease 1	Acute leukemia	127 (56%)
	Chronic leukemia	6 (3%)
	MDS/MPN	70 (31%)
	Other	23 (10%)
Cell source	BM	12 (5%)
	CORD	30 (13%)
	PBSC	184 (81%)
HLA matching	Matched related	45 (20%)
	Matched unrelated	120 (53%)
	Haploidentical	16 (7%)
	Mismatched	15 (7%)
	CORD	30 (13%)
Donor relationship	Not related	165 (73%)
	Related	61 (27%)
Donor CMV serostatus	−	126 (56%)
	+	100 (44%)
Conditioning regimen	Myeloablative	101 (45%)
	Reduced intensity	95 (42%)
	Non-myeloablative	30 (13%)
T Cell depletion 2	No	213 (94%)
	Yes	13 (6%)
Peak acute GVHD Grade	Grade 0–1	72 (32%)
	Grade 2–4	154 (68%)
Acute GVHD Onset (days post-HCT)	Median (IQR)	29.5 (21.0–41.0)
Cumulative steroid dose by post-HCT day 100 (mg/kg of cumulative steroid dose AUC)	Median (IQR)	11.7 (0.0–29.4)
GVHD prophylaxis 3,4	MTX+CNI	81 (36%)
	MMF+CNI	56 (25%)
	PTCy	41 (18%)
	Siro-based	48 (21%)
Transplant number	1	211 (93%)
	2	15 (7%)

BM= bone marrow, CNI= calcineurin-inhibitor; CMV= cytomegalovirus, CORD= umbilical cord stem cell; HLA=human leukocyte antigen; IQR=interquartile range; GVHD=graft-versus-host disease; MDS=myelodysplastic syndrome; MMF=mycophenolate; MPN=myeloproliferative neoplasm; MTX= methotrexate; PBSC=peripheral blood stem cell; PTCy= post-transplantation cyclophosphamide; siro=sirolimus.

¹“Other” underlying disease includes: aplastic anemia, immune deficiency disorder, myelofibrosis, myelodysplastic syndrome, myeloproliferative disorder (not otherwise specified), monoclonal gammopathy, multiple myeloma, paroxysmal nocturnal hemoglobinuria, plasma cell leukemia, prolymphocytic leukemia, refractory anemia or other refractory cytopenias, and sideroblastic anemia.

²T-cell depletion was performed using selective depletion of CD45RA+ T-Cells in all cases.

³”PTCy-based” GVHD prophylaxis includes: PTCy+CNI and PTCy+MMF+tacrolimus.

⁴”Siro-based” GVHD prophylaxis includes: triple immunosuppressive therapy with CNI+MMF+sirolimus.

The median day of neutrophil engraftment was post-HCT day 17 (interquartile range [IQR] post-HCT day 15–21). The median cumulative steroid AUC up to 100 days post-HCT among all letermovir recipients was 11.7 mg/kg of cumulative steroid dose AUC (IQR 0–29.4 mg/kg*days) and cumulative steroid use by GVHD prophylaxis-type is shown in Supplemental Figure 4. The median days of letermovir initiation and discontinuation were day 19 (IQR days 8–25) and day 98 (IQR days 91–98), respectively. The median duration of total letermovir exposure was 78 days (IQR 65–86 days). Forty-six (20%) patients discontinued letermovir treatment before day 90 and reasons for discontinuation are shown in Supplemental Table 3.

CMV Reactivation and UL56 Resistance during Letermovir Prophylaxis.

Cumulative incidence frequencies and weekly proportions of CMV reactivation at multiple levels during letermovir prophylaxis are shown in Figures 1 and 2, respectively. Ninety-five (42%) patients reactivated CMV at any level (i.e., any positive CMV PCR test above the assay LoD) in the first 100 days post-HCT, whereas 43 (19%) and 15 (7%) reactivated CMV at ≥ 150 IU/mL and ≥ 500 IU/mL, respectively. One patient was diagnosed with CMV gastrointestinal disease during the first 100 days, and the patient had preceding CMV DNAemia before diagnosis. Cumulative incidence and weekly proportions of CMV reactivation according to HCT risk are shown in Supplemental Figures 5 and 6, respectively. High-risk HCT recipients had an increased frequency and weekly proportion of CMV reactivation compared to low-risk HCT recipients at each evaluated viral threshold. The median day of onset of CMV reactivation was not different between groups.

Figure 1. – Cumulative incidence of CMV reactivation during the first 100 days after HCT.

The cumulative incidence of CMV infection from days 0 through 100 after HCT in patients treated with letermovir. A positive viral event was defined by CMV PCR DNAemia at any level (i.e., any positive CMV PCR test above the level of detection for the assay used; blue), ≥ 150 IU/mL (red), or ≥ 500 IU/mL (green). Death was treated as a competing risk.

Figure 2. – Weekly proportion of CMV reactivation during the first 15 weeks after HCT by CMV DNAemia level.

The weekly proportion (i.e., prevalence at any time during each week) of CMV reactivation during the first 15 weeks after HCT in letermovir-treated recipients defined by CMV PCR DNAemia at any level (i.e., any positive CMV PCR test above the level of detection for the assay used; blue), ≥ 150 IU/mL (red), or ≥ 500 IU/mL (green). Weekly proportions were calculated by the total number of incidences of each week divided by the total number of patients alive that week.

The cumulative incidence of CMV reactivation was evaluated in alternative iteration with initiation of letermovir as day zero (Figure 3). Most CMV reactivation occurred in the first month of letermovir prophylaxis. We also stratified analyses by HCT risk (Supplemental Figure 7), and findings were similar to the entire cohort except with an additional increase of any CMV reactivation ~50–60 days after letermovir initation in high-risk HCT recipients. In total, 7/127 (6%) low-risk HCT recipients and 24/99 (24%) high-risk HCT recipients developed clinically significant CMV during letermovir prophylaxis that required initiation of preemptive antiviral therapy.

Figure 3. – Cumulative incidence of CMV reactivation during the first 100 days after letermovir initation.

The cumulative incidence of CMV infection from days 0 through 100 after letermovir initiation. A positive viral event was defined by CMV PCR DNAemia at any level (i.e., any positive CMV PCR test above the level of detection for the assay used; blue), ≥ 150 IU/mL (red), or ≥ 500 IU/mL (green). Death and early discontinuation of letermovir were treated as competing risks.

Fifteen of 226 patients (6.6%) had CMV reactivation ≥ 200 IU/mL during letermovir prophylaxis (Supplemental Table 4). All 15 patients had minimum required plasma for CMV UL56 resistance testing, however, CMV DNA PCR amplification was successful in only 7 (47%) patients, including in 1/7 (14%) low-risk HCT recipient and 6/8 (75%) high-risk HCT recipients. The lowest level CMV DNAemia successfully sequenced was 210 IU/mL. One sample was successfully sequenced after pooling from two rounds of CMV DNA PCR amplification. The only known resistance mutation detected was C325Y in an umbilical cord blood recipient and their clinical and CMV antiviral course is summarized in Figure 4. Only 5/8 (63%) non-sequenced patients received preemptive antiviral therapy a median duration of 21 days. One haploidentical HCT recipient had persistent low-level CMV DNAemia (i.e., <100 IU/mL) that became undetectable after 51 days of antiviral therapy.

Figure 4. – Clinical course in patient with a mutation in CMV UL56 confering letermovir resistance.

Summary of post-transplant clinical course in a single umbilical cord blood transplant recipient identified to have CMV UL56 mutation (C325Y) confering letermovir resistance. CMV DNA PCR levels (log10 IU/mL; red circles) and prednisone-equivalent weight-based daily steroid dose (mg/kg; black squares) during the first 100 days post-transplant are shown. The patient received a total of 56 days of letermovir beforebefore treatment for clinically significant CMV (purple shaded area). The patient was initially treated with foscarnet (PFA) due to concerns for cytopenias, however, treatment was subsequently changed to ganciclovir (GCV) followed by valganciclovir (VGCV).

Risk factors for CMV reactivation during letermovir prophylaxis administration.

To assess the associations of clinical and transplantation variables with CMV reactivation during letermovir prophylaxis, we used univariable Cox regression at multiple CMV DNA PCR thresholds (Supplemental Table 5).

Multivariable Cox regression modeling was also used (Figure 5). Cumulative corticosteroid use before CMV reactivation was associated with significantly increased risk of CMV reactivation at any level (i.e., any positive CMV PCR test above the assay LoD) during letermovir prophylaxis (adjusted hazard ratio [aHR] 10.3 per log₁₀ mg/kg of cumulative steroid dose AUC, 95% CI 5.48–19.5, p< 0.001). GVHD prophylaxis with post-transplantation cyclophosphamide (PTCy; [aHR 2.34, 95% CI 1.28–4.28, p=0.001]) or with calcineurin inhibitors plus mycophenolate (CNI+MMF; [aHR 2.24, 95% CI 1.30–3.86, p=0.004]) were also associated with significantly increased risk of any CMV reactivation. Underlying myelodysplastic syndrome (MDS)/myeloproliferative neoplasm (MPN) was associated with a decreased risk of any CMV reactivation compared to acute leukemia (aHR 0.52, 95% CI 0.30–0.92, p=0.03). Non-Caucasian patients were at significantly increased risk of any CMV reactivation during letermovir prophylaxis (aHR 1.72, 95% CI 1.10–2.71, p=0.02). Only GVHD prophylaxis with PTCy and cumulative steroid use were associated with a statistically significant increased risk of CMV reactivation ≥ 150 IU/mL during letermovir prophylaxis. The number of CMV reactivations ≥ 500 IU/mL during letermovir prophylaxis was too small to be analyzed by multivariable Cox regression.

Figure 5. – Multivariable Cox regression of clinical risk factors on time to CMV reactivation during letermovir prophylaxis.

Multivariable Cox regression of the association of clinical risk factors with time to CMV reactivation at (A) any level (i.e., any positive CMV PCR test above the level of detection for the assay used) or (B) ≥ 150 IU/mL during the first 100 days post-HCT during letermovir prophylaxis. Models were adjusted for underlying disease, graft-versus-host disease (GVHD) prophylaxis, cumulative steroid exposure, or race, which were selected according to univariate Cox regression results. For underlying disease: “Other” group includes aplastic anemia, immune deficiency disorder, monocloncal gammopathy, multiple myeloma, paroxsymal nocturnal hemoglobinuria, plasma cell leukemia, prolymphocytic leukemia, refractory anemia or other refractory cytopenias, and sideroblastic anemia. For GVHD prophylaxis: CNI=calcineurin inhibitor; MMF=mycophenolate mofetil; MTX=methotrexate; PTCy=post-transplantation cyclophosphamide; and siro=sirolimus. Cumulative steroid exposure was calculated as the area under the curve (AUC) of daily weight-based, prednisone-equivalent steroid dose (mg/kg). For each 1 Log10 mg/kg increase in cumulative steroid dose AUC per patient, there was an 11.8 increased hazard of CMV reactivation by day 100 post-transplant. Multivariable Cox regression on time to CMV reactivation at ≥ 500 IU/mL could not be performed due to the limited number of events.

Because patients with alternative donor grafts (i.e., mismatched, haploidentical, cord) are likely to receive GVHD prophylaxis with CNI+MMF or PTCy at our center, multivariate Cox regression was repeated in alternative iterations adjusting for donor HLA-type to identify possible interactions between covariates (Supplemental Figure 8). Our observations persisted following adjustment for donor HLA-type, except PTCy was no longer associated with a statistically significant increased risk of CMV reactivation ≥ 150 IU/mL. Finally, we constructed CMV cumulative incidence curves to identify possible differences in patients who received PTCy with or without MMF (Supplemental Figure 9). Patients who received PTCy with MMF had a higher incidence of CMV reactivation than patients who received PTCy without MMF, however, the differences were less pronounced at higher viral thresholds.

Discussion

At our center, development of de novo CMV UL56 resistance during letermovir prophylaxis after HCT appeared infrequent and clinically significant CMV reactivation during letermovir prophylaxis was successfully managed using other antiviral therapies. Although letermovir decreased the risk of clinically significant CMV reactivation after allogeneic HCT, low-level CMV reactivation still occurred at an increased and at a persistent frequency throughout the study period. Finally, multiple clinical and transplantation risk factors for breakthrough CMV reactivation during letermovir prophylaxis were identified including GVHD prophylaxis with PTCy or CNI+MMF. However, cumulative steroid exposure before CMV reactivation was the strongest risk factor at all evaluated viral thresholds.

Historically, mutations in CMV UL97 or UL54 that confer ganciclovir resistance required clinically challenging and complex CMV management.(21) We systematically evaluated patients with CMV reactivation for CMV UL56 resistance and successfully genotyped CMV DNAemia as low as 210 IU/mL. A recent study was performed that examined the incidence of UL56 mutations in 1165 clinical specimens submitted to a reference laboratory using Sanger sequencing and known UL56 mutations that confer letermovir resistance were identified in 130 (11%) samples.(22) However, unlike the current study, no matching treatment history was available for analysis. Reassuringly, de novo CMV UL56 letermovir resistance was infrequent at our center and our results are in alignment with other studies that observed decreased rates of letermovir resistance.(11, 12)

The ~40% incidence of subclinical CMV reactivation and low incidence of clinically significant CMV infection during letermovir prophylaxis observed in our study is comparable to other transplant centers.(14, 23–31) Allowing for a relatively permissive level of CMV DNAemia during antiviral prophylaxis prior to initiating preemptive antiviral therapy may offer an immunologic advantage(32, 33) by promoting CMV-specific immune reconstitution after HCT. We previously showed that controlled CMV replication (including levels < 150 IU/mL) during letermovir prophylaxis may enhance CMV-specific immunity after HCT.(16) Whether CMV DNAemia ≥ 150 IU/mL during letermovir prophylaxis necessarily requires preemptive antiviral therapy is unclear given that some patients will likely not progress to higher levels following repeat testing, especially in situations of low to moderate immunosuppression.

The association of traditional risk factors for CMV reactivation after HCT is poorly defined following the introduction of letermovir prophylaxis.(14, 25–28, 31) Cumulative steroid exposure was the strongest risk factor for CMV reactivation at all viral thresholds; whereas GVHD did not enter the multivariate model after accounting for steroid use, yet the collinearity between these risk factors makes it impossible to determine if GVHD contributes independently to the risk of CMV. In 2013, our center began utilizing lower prednisone dosing (0.5–1.0 mg/kg/d) for initial treatment of GVHD based on results from a local randomized clinical trial.(34) It is possible that increased steroid exposure at other transplant centers may lead to an increased risk of CMV reactivation during letermovir prophylaxis.

Traditionally, GVHD prophylaxis regimens have primarily been CNI-based in combination with MTX or MMF.(35) These drugs are known to impair T-cell function and delay immune reconstitution after HCT(36), but their effects on clinical and infectious outcomes in the era of letermovir prophylaxis are only beginning to be explored.(28, 31) At our center, patients with alternative donor grafts (i.e., mismatched, haploidentical, cord) are more likely to receive GVHD prophylaxis with CNI+MMF or PTCy; whereas HLA-matched related/unrelated HCT recipients typically receive CNI+MTX if they have received myeloablative or reduced-intensity conditioning. Interestingly, we observed a significantly increased risk of CMV reactivation in the setting of GVHD prophylaxis with CNI+MMF and with PTCy even after adjusting for donor HLA-type. The association with MMF was unexpected because several trials previously showed no difference in the incidence of CMV infection with CNI+MTX vs CNI+MMF for post-HCT immunosuppression.(37, 38) In contrast, the association of PTCy and increased risk of CMV reactivation has been reported in multiple studies(25, 28, 39, 40) and an increase in non-CMV herpesvirus infections with PTCy suggests a possible drug effect.(41) CMV antigen exposure through the use of PTCy for GVHD prophylaxis might offer an immunologic “sweet spot” or balance in the protection against GVHD while concomitantly improving longer-term virus-specific immune reconstitution.(32, 33) The significant reduction in chronic GVHD risk and subsequent immunosuppressive therapy with PTCy may also improve long-term immune reconstitution.(42–44) Finally, some trials previously noted a decreased risk of CMV reactivation with sirolimus-based GVHD prophylaxis.(45, 46) We did not observe this association possibly due to small number of sirolimus recipients as well as the diverse conditioning regimens and donor-types included in our study.

Strengths of our study include the size of our cohort as well as our in-depth analyses on the weekly proportion of CMV infection and risk factors for breakthrough CMV reactivation at multiple viral thresholds. Because of the retrospective and observational design of our study, we could not perform therapeutic drug monitoring which can detect differences in trough concentrations in patients with concomitant interacting drugs or with abnormal gastrointestinal absorption.(47) The inability to amplify DNA in all samples from patients with CMV DNAemia >200 IU/mL is another limitation due primarily to insufficient leftover clinical sample for retesting. Thus, it is possible that the 8 non-sequenced patients could have had genotypic resistance. In addition, previous studies have shown that CMV DNA PCR positivity (specifically at low levels) in the setting of letermovir prophylaxis may not always be indicative of active viral replication given that letermovir does not actually inhibit CMV DNA replication per se.(48, 49) Furthermore, CMV-specific cellular immunity was not measured though we did adjust for absolute lymphocyte counts as a measure of global immunosuppression. It is also possible that our method for calculating cumulative steroid exposure may slightly underestimate steroid exposure during periods of tapering doses. Finally, the association of patient race with CMV reactivation during letermovir prophylaxis could not be analyzed in more depth and there was a paucity of umbilical cord blood recipients and recipients with HLA-matched related donors in our study.

In conclusion, mutations in CMV UL56 and development of de novo letermovir resistance was rare in our study. Although clinically significant CMV during letermovir prophylaxis is uncommon, subclinical CMV reactivation occurred frequently. Cumulative steroid exposure increases the risk of breakthrough CMV infection during letermovir prophylaxis and may potentially contribute to the emergence of resistance. Future studies should address whether decisions to begin preemptive antiviral therapy during letermovir prophylaxis should be influenced by risk factors for persistent CMV replication. Monitoring for CMV UL56 resistance may be important in settings where known risk factors are prevalent via institutional standards.