Late cytomegalovirus disease after hematopoietic cell transplantation: significance of novel transplantation techniques

In the past 2 decades, significant progress has been made in the prevention of cytomegalovirus (CMV) end-organ disease occurring during the first 100 days after allogeneic hematopoietic cell transplantation (HCT). Widespread adoption of preemptive therapy (PET) strategies and, more recently, the introduction of letermovir for antiviral prophylaxis have decreased the incidence of these once common entities to only 3% to 5% of patients at risk in the early posttransplant period.^1,2 However, letermovir may only be used for selected patients at high risk, and access to letermovir is not universal. Thus, preemptive therapy remains a widely used approach for CMV disease prevention worldwide.³ Late-onset CMV disease is well described after PET strategies,^4,5 and recent data indicate that rebound clinically significant CMV infection, including disease, is more likely to occur within the first year after withdrawal of letermovir prophylaxis.^1,6-9 

The prevailing strategy for prevention of late CMV disease in patients at high risk is continued surveillance and PET,¹⁰ however, little is known about the real-world effectiveness of this strategy in the prevention of late CMV disease in other at-risk populations. Furthermore, since the first comprehensive description of late CMV disease >20 years ago,^11,12 optimization of early CMV preventive strategies as well as the development of novel transplantation techniques have improved overall survival rates and may have changed the risk and clinical manifestations of late CMV disease.^5,13,14 

The purpose of this study was to analyze the key characteristics of late CMV disease in a large cohort of allogeneic HCT survivors. Specifically, we examined the clinical manifestations, timing, risk factors for initial and recurrent incidence of late CMV disease as well as the impact of novel graft-versus-host disease (GVHD) prevention strategies, risk stratification algorithm for late surveillance, letermovir use, and outcome of late CMV disease in the first 2 years after transplantation.

This was a retrospective cohort study that included all CMV seropositive recipients of allogeneic HCT who received transplantation at Fred Hutchinson Cancer Center (Fred Hutch) between 2001 and 2017 (PET cohort) and 2018 and 2021 (letermovir; LET cohort) who survived at least 100 days after HCT. At ∼100 days after HCT, patients were discharged from the center to their referring provider. Additional data were obtained by record review (supplemental Materials).

The standard practice at Fred Hutch throughout the study period was to recommend ongoing CMV surveillance by plasma polymerase chain reaction (PCR) weekly until 1 year after transplantation for patients who received transplantation from cord blood, recipients of antithymocyte globulin or alemtuzumab, those who either received treatment for CMV reactivation or disease before day 100, or who were receiving corticosteroids for treatment of chronic GVHD (cGVHD). In non–cord blood HCT recipients, surveillance frequency was changed to every other week if steroid dose was decreased to <0.5 mg/kg per day or doses of other immunosuppressive agents were tapered and the patient had 3 consecutive negative surveillance PCR tests. Surveillance was discontinued entirely after 2 additional negative PCR tests if tapering of immunosuppression continued. If immunosuppression increased before 1 year after HCT, it was recommended to resume weekly testing. There were no formalized guidelines for late CMV management for the second year after transplantation or thereafter. Detailed CMV surveillance data were collected for all patients diagnosed with CMV disease within the first year after HCT.

Management of CMV reactivation before day 100 varied slightly over time as was previously described.^5,15-17 Since 2018, all CMV seropositive HCT recipients received letermovir prophylaxis for up to 100 days after HCT according to institutional guidelines. After day 100 posttransplantation, treatment for CMV viremia was recommended if the viral load increased to >5 times the baseline value within one month or if was ≥250 IU/ml (PET cohort) or ≥500 IU/ml in LET cohort. Since 2006, a total of 3 different antiviral prevention strategies that consisted of high-dose valacyclovir and valganciclovir prophylaxis were used for cord blood transplant recipients, as previously described.¹⁷ 

CMV disease was defined by standard international definitions,^18,19 and a detailed description is available in the supplemental Materials. Early and late CMV disease were defined as CMV disease diagnosed before or after day 100 post-HCT, respectively. Recurrent CMV disease was defined as a subsequent episode of CMV disease that occurred at least 4 weeks after the discontinuation of antiviral therapy for previously diagnosed CMV disease, with resolution of clinical or radiological symptoms observed after initial treatment.

cGVHD was diagnosed based on the institutional protocol until 2005 and later according to the most recent National Institutes of Health criteria with an organ-specific scoring system.^20,21 

The probability of late CMV disease was estimated by cumulative incidence curves, treating death and morphologic relapse (additional model) as competing risks and subsequent transplant as a censoring event. Gray test was used to compare the cumulative incidence probabilities between categories. Kaplan-Meier estimation method was used to estimate the incidence of overall mortality. Cox proportional hazards models were used to estimate the associations between potential risk factors and the development of late CMV disease and overall mortality. The Fine and Gray version of the Cox model was used if applicable. Risk factors with P values <.05 in the univariable analysis were included in the final multivariable model. A sensitivity analysis was performed to verify the role of posttransplant cyclophosphamide (PT-Cy) in study population, including years with at least 8 cases of PT-Cy per year (since 2008). Landmark analyses were performed among day 100 and 1-year survivors. Demographic and clinical factors evaluated as potential risk factors for late CMV disease are listed in the supplemental Materials. Statistical significance was defined as 2-sided P value <.05. All analyses were performed using SAS 9.4 (TS1M6; SAS Institute Inc, Cary, NC).

The study was approved by the institutional review board, and all patients signed informed consent before participation in the study.

Of the 2625 CMV seropositive allograft recipients who received transplantation during the study period, 2155 of 2298 survived up to day 100 after HCT in the PET cohort, and 314 of 327 did in the LET cohort (Table 1; supplemental Table 1). By day 100, any level of CMV reactivation was observed in 1555 (73%) and 168 (53%) in the PET and LET groups, respectively. Letermovir prophylaxis was typically discontinued 98 days (range, 10-272) after HCT, with an overall median drug exposure of 78 days (range, 1-224). Letermovir was paused or discontinued early in 37 patients (11.7%) because of the development of clinically significant CMV infection; however, 5 patients resumed letermovir prophylaxis after antiviral treatment. Among day 100 survivors, CMV disease before day 100 was diagnosed in 119 (6%) and 2 patients (0.6%) in the PET and LET cohorts, respectively.

There were 203 episodes of late CMV disease in the entire study population, with an estimated cumulative incidence of first late CMV disease of 7.2% (95% confidence interval [CI], 6.2-8.3). In the PET group, there were 1182 episodes of late CMV disease among 160 patients by 2 years after transplantation, whereas 21 episodes of late CMV disease in 17 patients were observed among letermovir recipients. The cumulative incidences of first late CMV disease in these cohorts were 7.4% (95% CI, 6.4-8.6) and 5.4% (95% CI, 3.2-8.3), respectively (Figure 1A). In 9 patients (5.2%), systemic antineoplastic treatment due to morphologic relapse of the underlying disease occurred between day 100 and the first late CMV disease diagnosis. The median time between morphologic relapse and CMV disease episode was 88 days (range, 38-242), and the median time of CMV disease onset in this population was 265 days after HCT (range, 215-542).

Among the late CMV disease cases, 142 (78%) occurred by 1 year after transplantation, and 40 (22%) occurred in the second year (2.2%; 95% CI, 1.5-3). The most common manifestation of late CMV disease was gastrointestinal (GI) tract involvement (4.0%; 95% CI, 3.2-4.9) followed by pneumonia (3.0%; 95% CI, 2.3-3.8). First late disease episodes were diagnosed at a median time of 185 days (range, 104-675) after HCT for GI disease and 211 days (range, 105-544) for pneumonia (Figures 1C and 2A). In 3 patients, both organs were involved at the time of CMV diagnosis. Other manifestations included retinitis, central nervous system, and bone marrow involvement with a cumulative incidence of 0.3% (95% CI, 0.1-0.6).

Recurrent CMV disease occurred in 34 patients with late CMV disease (probability of 13.6% by 1 year; 95% CI, 9.6-18.1; Figures 1B and 2B). Most recurrences involved the same anatomical site as the first episode, with a median time of 142 days (range, 49-444) after the initial event. Five patients developed a third episode, with a median of 145 days (range, 85-350) after their second episode.

Among patients with late CMV GI disease, the diagnosis of intestinal GVHD preceded GI CMV disease in 52 (59%). Between days 100 and 180 after HCT, among 42 cases of GI disease, grades 1 to 2 and 3 to 4 GI GVHD were diagnosed in 23 (55%) and 12 (28%) patients, respectively; 7 (17%) patients did not experience concurrent GI GVHD (Figure 2C). In the first year after HCT, the median time between GVHD and late GI CMV disease diagnosis was 46 days (range, 0-215). Ten (11%) patients had concomitant biopsy-proven GVHD and CMV disease at the time of diagnosis.

Most CMV disease diagnoses (n = 15 [71%]) occurred within the first year after HCT, and the main manifestation was GI disease (3.8%; 95% CI, 2.0-6.3), followed by pneumonia (1.3%; 95% CI, 0.4-3.1) and 1 episode of retinitis (0.3%; 95% CI, 0.03-1.6; Figures 1D and 2D). The median time of first late GI disease diagnosis was 184 days (range, 124-334) after HCT and 363 days for pneumonia (range, 140-443). Recurrent late disease was observed in 3 patients (18%), with the recurrence present in the originally involved organ with a median time of 217 days (range, 148-274) after the initial event (Figure 2E). Among patients with first late CMV GI disease, the diagnosis of intestinal GVHD preceded GI CMV disease in 7 patients (58%; Figure 2F); the median time between GVHD and late GI CMV disease diagnosis was 43 days (range, 0-104).

In a multivariable Cox regression model including the entire cohort, the greatest risk was associated with corticosteroid treatment after post-HCT day 100 at dose ≥1 mg/kg and high-risk viremia detected before day 100. Other independent factors included grade 3 to 4 acute GVHD, cGVHD, lymphopenia, donor type, early CMV disease, and time of HCT. PT-Cy was not significant in the combined cohort but reached significance in the sensitivity analysis in PET era. Results are summarized in Figure 3A-B.

PT-Cy showed a trend towards an increased risk of late CMV disease by 2 years in the absence of CMV prophylaxis, whereas the other risk profile was otherwise similar to the overall cohort (Figure 3B), with high-dose steroid treatment as time-dependent variable and high-level viremia before day 100 showing the greatest risk of developing late CMV disease. In a sensitivity analysis restricted to patients who underwent transplantation between 2008 and 2017 when PT-Cy was more regularly given (at least 8 cases of PT-Cy per year), PT-Cy–based GVHD prophylaxis was a significant independent factor for first-onset late CMV disease by second year after HCT (adjusted HR, 2.36; 95% CI, 1.22-4.59), regardless of donor type.

In a landmark analysis of patients who survived 1 year after HCT, the risk of CMV disease during the second year was associated with CMV disease during the first year after HCT (adjusted HR, 3.75; 95% CI, 1.82-7.75). and with cGVHD (adjusted HR, 1.96; 95% CI, 0.94-4.06).

Among patients diagnosed with CMV disease (early and late episodes), who survived at least 100 days after HCT, factors associated with an increased risk of recurrent of CMV disease included transplantation from HLA-mismatched donors (adjusted HR, 2.91; 95% CI, 1.03-8.17) and treatment with high-dose steroids after day 100 (as time-dependent variable; adjusted HR, 9.22; 95% CI, 4.18-20.03).

The only factors associated with the increased risk of first late CMV disease in univariable model were lymphopenia <300 cells per mm³, steroid treatment at the dose >1mg/kg as time-dependent variable, cord blood transplantation, viremia before day 100 (in all cases <1000 IU/mL), and cGVHD (supplemental Table 2). There was no impact of letermovir exposure or clinically significant CMV infection while on letermovir prophylaxis on the risk of late CMV disease. In several multivariable models adjusted for letermovir exposure, corticosteroid treatment was the strongest risk factor for CMV disease (supplemental Figure 1).

In univariable models, there were differences between patients who developed CMV pneumonia or GI disease including both pretransplant and posttransplant factors (supplemental Table 3). In multivariable models, CMV disease before day 100 increased the risk of late GI disease (adjusted HR, 3.24; 95% CI, 1,8-5.82), and patients receiving PT-Cy and mycophenolate mofetil–based GVHD prophylaxis were at the highest risk for GI disease; however, this association was not observed for late CMV pneumonia (Figure 4A-B). High-dose steroid treatment after day 100 had the strongest impact on the incidence of CMV disease in both groups, with the greatest effect on CMV pneumonia (Figure 4), followed by CMV viremia before day 100, which also remained significant for both pneumonia and GI disease (Figure 4).

Among day 100 survivors in the PET cohort, prolonged CMV surveillance was recommended for 1787 patients (83%). Based on day 100 risk stratification, cumulative incidence of late CMV disease was 8.4% (95% CI, 7.1-9.7) and 3.0% (95% CI, 1.6-5.1) in the high- and low-risk groups, respectively. All patients in the LET cohort met the criteria for high-risk group.

The main cause of surveillance failure leading to CMV disease during the first year after HCT in the PET cohort was viremia level below the recommended treatment threshold (38 episodes [27%]; 14% with GI disease and 13% with pneumonia). CMV disease despite preemptive antiviral treatment for >7 days before the onset of disease and in situations in which no monitoring or inadequate frequency of surveillance were done led to CMV disease in 22% and 18%, respectively. Breakthrough disease (undetectable CMV with adequate monitoring) was diagnosed in 13%, and situations in which the treatment threshold was reached only within 7 days before disease onset occurred in 12% (Figure 5A).

In the LET cohort, the main cause of surveillance failure leading to CMV disease during the first year after HCT was viremia level below the recommended treatment threshold (4 cases [27%]). CMV disease despite preemptive antiviral treatment for >7 days before the onset of disease and in situations in which monitoring or inadequate frequency of surveillance was done led to CMV disease in 16% and 13%, respectively. Breakthrough disease or situations in which the treatment threshold was reached only within 7 days before disease onset occurred in 20% and 20%, respectively, Figure 5B.

Of 177 patients with late CMV disease, 35 (19%) died within 6 weeks of diagnosis of the first late CMV episode (Figure 6), and 31 (88%) of the deaths were related to multiple infectious complications; among these, 13 (42%) had progressive GVHD; in 7 cases (23%), GVHD was active but under control; 3 (10%) were in a process of reducing immunosuppression; and in 8 (25%), no active GVHD was observed. In multivariable Cox regression models assessing overall mortality by 2 years after transplantation, both early- and late-onset CMV disease were significantly associated with death in the PET cohort, whereas late CMV disease was an independent risk factor in the LET cohort. CMV-related mortality was driven primarily by CMV pneumonia (PET cohort: adjusted HR, 4.02; 95% CI, 2.55-6.32; LET cohort: adjusted HR, 16.1; 95% CI, 3.77-68.5), whereas GI disease only slightly increased the risk of mortality in the PET cohort (adjusted HR, 1.68; 95% CI, 1.1-2.56) but more pronounced in the LET cohort (adjusted HR, 3.26; 95% CI, 1.21-8.77). Other factors affecting survival in the entire cohorts are shown in Supplemental Table 4.

In this large retrospective cohort study, we demonstrate that late CMV disease after HCT occurred in ∼7.4% of patients who survived the 3 months after HCT in patients receiving PET. A similar late CMV disease risk was observed in patients receiving letermovir prophylaxis. Almost half of the cases were CMV pneumonia, which carried a fourfold risk of mortality by 2 years after transplantation, whereas the risk was only slightly increased with CMV GI disease. Other manifestations, such as retinitis, encephalitis, and late marrow failure, were rare. CMV reactivation during the first 3 months after HCT and the use of high-dose systemic corticosteroids after day 100 were risk factors for late CMV disease regardless of letermovir prophylaxis. PT-Cy–based GVHD prophylaxis increased the risk of late CMV disease, but the effect was diminished by letermovir prophylaxis. The risk of GI CMV disease remained significantly increased in the entire population. Risk stratification at day 100 to assign patients into an intensive surveillance protocol identified most patients who developed late CMV risk, however, the recommended thresholds for PET appeared to be a key reason for breakthrough CMV disease.

The late CMV disease rate in this study is improved compared with that reported from an earlier Fred Hutch cohort without PET in which late disease occurred in 18% of patients with CMV seropositivity.¹¹ Our results are in alignment with much smaller cohorts that analyzed the effectiveness of late preemptive strategies^22-26 but less favorable than results seen in prospective clinical trials published in the past decade in which PET programs were variably used.^1,4,27,28 Although the incidence of early CMV disease continues to decline, delayed-onset of CMV infection is an emerging problem because many new transplantation techniques and drugs may affect the CMV-specific viral responses.^7,9,16,29 Our data show that late CMV disease continues to be a clinical problem with the most recent innovations in transplantation techniques and CMV drugs.

Interestingly, in the later years of our study, there was a noticeable decline in pneumonia (supplemental Figure 2), which continued in the letermovir era. One possible explanation is that there has been a shift to limit invasive procedures, including bronchoscopies and bronchoalveolar lavages, and to initiate treatment based on results from noninvasive tests and clinical presentation. This practice is well documented for respiratory virus lower respiratory tract infections and invasive fungal disease.^30-33 Such “possible” CMV disease cases would not have been captured by our review. Future studies should examine whether there is an increased trend to forgo bronchoscopic evaluation of suspected CMV lung involvement. There are other possible factors that may explain the decline of documented proven CMV pneumonia, including an increased use of sirolimus in immunosuppressive protocols or other unknown factors. However, the change in diagnostic practices appears to be the most plausible explanation.

To our knowledge, this is the first and most comprehensive study in the present era to specifically analyze risk factors and clinical manifestations of late CMV disease including PET and LET populations. The rate of late CMV disease in the LET group appears to be higher than that reported in a recent meta-analysis,³⁴ likely due to prolonged follow-up. The finding that PT-Cy was a risk factor for late CMV disease is novel and important because PT-Cy is increasingly used as GVHD prophylaxis. Our results expand upon previous study of PT-Cy and CMV disease risk, which largely targeted early CMV disease events and were restricted to HLA-matched transplants.³⁵ Letermovir prophylaxis appears to reduce the unfavorable effect of PT-Cy prophylaxis on CMV disease risk, likely due to protection against high levels of CMV viremia before post-HCT day 100, which was one on the main factors increasing the risk of late CMV disease. However, it seems that different transplantation and CMV prevention protocols may influence specific organ involvement of CMV disease, as indicated by the higher incidence of CMV GI disease seen in PT-Cy and mycophenolate mofetil recipients, whereas such association was not observed for patients with pneumonia. This may be related to direct GI toxicity of both drugs as well as the overall higher incidence of viral infections seen with these regimens.^35,36 Late CMV GI disease was the most common manifestation and primarily affected patients with a history of acute GI GVHD; this observation is consistent with previously published data that showed severe grade 3 to 4 acute GVHD increased the risk of any GI CMV disease.^37-40 

We also examined, for the first time, risk factors for recurrent CMV disease and disease risk in the second year after HCT. Administration of high-dose steroid treatment appeared to be the most important factor for prediction of second episode and disease in the second year after HCT. This suggests that prolonged or reinstitution of PCR monitoring is advisable in these situations.

Our data indicate that risk stratification at the time of discharge from the transplant center is insufficient. Dynamic risk stratification and restart of PCR monitoring when immunosuppression increases should be applied to all patients regardless of day 100 stratification, because the high-dose steroid treatment as a time-varying variable was the strongest risk factor for late CMV disease. Community providers caring for patients after transplantation need to remain vigilant in their surveillance for CMV viremia in patients at risk, particularly those who receive novel immunosuppressive therapy prophylactic protocols and also high-dose corticosteroids for treatment of cGVHD.

Improving adherence to surveillance testing and PET will likely decrease the incidence of late CMV disease, as suggested by a multicenter randomized trial for prevention of late CMV disease,^4,27 in which the near-perfect adherence to weekly plasma PCR tests and rapid initiation of PET resulted in a CMV disease incidence of only 2% by day 270 after transplant, approximately one-third the rate observed in our cohort study. Due to insufficient data, current guidelines are silent about viral load thresholds for the treatment of viremia after day 100. The most common reason for PCR surveillance failure was viral load below the threshold for treatment, and recurrent low-level viremia was observed in both patients with GI disease and those with pneumonia. Using a lower treatment threshold than that was used in this study may allow for earlier initiation of PET, and a population with persistent low-level viremia could potentially benefit from antiviral treatment despite not reaching the established cutoff level.

Strengths of this study include the large sample size, a detailed review of end points and variables in all patients via record review, the use of uniform management guidelines throughout the study period, and the inclusion and analysis of recent transplantation practices including letermovir prophylaxis. However, the study also has limitations that should be noted, including the retrospective nature of this study that may have resulted in an underestimation of late CMV disease, especially among ∼80% of patients who were followed outside the Fred Hutch system after day 100. Furthermore, adherence to the prescribed surveillance frequency was only obtained in patients with late CMV disease, which limits the data on adherence for the entire cohort.

In conclusion, in this contemporary cohort of >2400 allogeneic HCT recipients with CMV seropositivity, late CMV disease was a persistent problem and was associated with significant mortality. Although the risk stratification at day 100 that we used was consistent with professional society recommendations,^41-43 improved risk stratification algorithms, possibly with the inclusion of CMV-specific immunity testing are needed.⁴⁴ Our observation suggests that initiation of PET in patients with persistent viremia is one potentially modifiable risk factor for late CMV disease, particularly in patients receiving systemic corticosteroid treatment. Moreover, improving access and adherence to PCR monitoring through home-based testing methods such as dried blood spot testing⁴⁵ or other new devices are being investigated for viral PCR monitoring.^46-48 Future studies could examine alternative options to decrease the frequency of late-onset CMV disease including extended antiviral prophylaxis⁹ and strategies aimed at decreasing systemic immunosuppression to treat cGVHD as well as improving CMV-specific immunity with vaccines or cellular therapy.

The authors thank Chris Davis and Ryan Basom for database services and Louise Kimball, Rachel Blazevic, Nina Ozbek, Ashley Patajo, Jocelyn Meyer, and the clinical research team for medical records review.

This study was supported by grants from the Polish National Agency for Academic Exchange, the Medical University of Gdańsk, the Joel D. Meyers Scholarship Endowment at the Fred Hutchinson Cancer Center, Seattle, WA (A.S.-K.), and from the National Institutes of Health (K23AI097234 [National Institute of Allergy and Infectious Diseases], K24HL093294 [National Heart, Lung and Blood Institute], CA15704 [National Cancer Institute], and National Institute of Allergy and Infectious Diseases Broad Agency Announcement [272201600015C] [M.B.], and K23AI163343 [D.Z.]).

Contribution: A.S.-K., S.Ö., M.L.G., and M.B. contributed to conception and design; J.S., A.S.-K., S.Ö., and S.S. were responsible for provision of study materials and data collection; A.S.-K., M.B., H.X., W.L., S.Ö., M.L.G., and D.Z. conducted data analysis and interpretation; A.S.-K., S.Ö., M.L.G., and M.B. prepared the initial draft of the manuscript; and all authors provided a critical review and final approval of the manuscript and are accountable for all aspects of the work.

Conflict-of-interest disclosure: M.B. reports research support from Merck and consulting fees from AlloVir, Symbio, Helocyte, Evrys Bio, and Moderna. The remaining authors declare no competing financial interests.

The current affiliation of S.Ö. is an independent researcher.

Correspondence: Michael Boeckh, 1100 Fairview Ave North, E4-100, Seattle, WA 98109; email: mboeckh@fredhutch.org; and Alicja Sadowska-Klasa, 1100 Fairview Ave North, E4-100, Seattle, WA 98109; email: asadowsk@fredhutch.org.