Antiviral prophylaxis for cytomegalovirus infection in allogeneic hematopoietic cell transplantation

Human cytomegalovirus (CMV) is an enveloped double-stranded DNA virus that belongs to the herpesviridae family. Infection with CMV is common, with seroprevalences ranging from 50% to over 90% depending on age, geographical location, and socioeconomic factors.¹  CMV establishes latency in human epithelial tissue, polymorphonuclear cells, myeloid progenitors, and T lymphocytes and is normally controlled by the host’s immune system.^2,3  Immunosuppression after allogeneic hematopoietic cell transplantation (HCT) frequently leads to CMV reactivation, which is associated with increased morbidity and mortality in this patient population.^4-6  Primary CMV infection and reactivation increase the risk of CMV disease after allogeneic HCT, which can manifest clinically in diverse ways, including colitis, pneumonitis, retinitis, and hepatitis.^7,8  Recent studies have also shown that despite use of preemptive therapy, CMV reactivation may be associated with an increased risk of invasive fungal disease⁸  and is also an independent risk factor for nonrelapse mortality.⁹ 

The evolution of effective antiviral agents against CMV has resulted in the emergence of 2 distinctive strategies to prevent CMV-related outcomes among patients undergoing HCT: universal prophylaxis and preemptive therapy. The latter is defined as antiviral treatment triggered by early detection of active CMV infection, before clinical disease occurs. Specifically, patients undergo blood CMV surveillance with viral DNA or antigen detection, and antiviral therapy is initiated above a certain detection threshold.^10,11  However, any level of CMV viremia has been associated with increased nonrelapse mortality after allogeneic HCT, despite use of highly sensitive diagnostic assays to detect low-level CMV viremia.¹²  This disadvantage highlights the need for safe and effective antiviral agents to be used in prophylactic strategies.

The quest for successful prophylactic strategies against CMV for allogeneic HCT patients started in the 1980s. Although universal prophylaxis was effective in preventing CMV primary infection and reactivation after transplantation in some trials, the overall benefit of prophylactic agents has been difficult to assess. Universal prophylaxis has been associated with toxicities particularly detrimental after HCT, including clinically significant myelosuppression associated with ganciclovir and valganciclovir use, which may increase nonrelapse mortality.¹³ 

We conducted a systematic review of all antivirals that have been studied for universal prophylaxis to reduce risk of CMV infection among patients undergoing allogeneic HCT. This review assesses the overall efficacy of antiviral prophylaxis in view of novel antiviral therapies and increasingly sensitive diagnostic tests and puts these into perspective with letermovir, which was recently approved for CMV prophylaxis in this patient population.

This manuscript was prepared and reported using PRISMA guidelines and registered in PROSPERO in 2016 as #CRD42016052180.¹⁴  PubMed electronic databases were queried from database inception to 31 December 2017. Search terms combined MeSH terms, text words, and exploding terms, including cytomegalovirus, CMV, allogeneic, stem-cell transplant, hematopoietic cell transplant, bone marrow transplant, and prophylaxis. The complete strategy and search terms are listed in supplemental Table 1 of the supplemental Material. The search was limited to articles published in English. Additional studies were identified from references from relevant articles.

We included randomized clinical trials involving IV or oral antiviral prophylaxis where CMV infection or CMV disease was a measured outcome. Nonrandomized trials, nonprophylactic trials involving preemptive therapy, and nonantiviral therapies were excluded. Studies of patients who had undergone allogeneic HCT irrespective of age, CMV serostatus, transplantation conditioning regimen, or HLA matching were included. Any intervention that compared an antiviral agent with either placebo or a different antiviral early after HCT (before day +100) with the intent of preventing initial episodes of CMV infection or disease posttransplantation was considered. Authors of selected papers were contacted for additional data if key outcomes were not reported. Study quality was assessed using a standardized tool in Review Manager 5.3 (Cochrane Collaboration, Copenhagen, Denmark) based on the Cochrane handbook, where selection, performance, detection, attrition, and reporting biases were assessed.¹⁵ 

To evaluate the efficacy of CMV prophylaxis, 2 primary outcomes were assessed. The first was to determine the effect of antiviral prophylaxis on incident CMV primary infection or reactivation. The second was to determine all-cause mortality through follow-up. Secondary end points included rate of preemptive therapy for CMV reactivation while receiving prophylaxis, incidence of CMV disease, and antiviral drug–related toxicity. The primary and secondary outcomes were obtained by referring to end points prespecified in the individual studies, but outcomes up to a year were captured if data were available.

Two independent reviewers (K.C. and M.P.C.) first assessed the titles and abstracts of the search results for eligibility. The full text of the eligible studies was then reviewed for inclusion. The reviewers then designed a data collection form to document authors, year of publication, definition of CMV infection, CMV infection rate, all-cause mortality rate, preemptive therapy rate, antiviral drug used, dosing information and timeline, days of treatment, days of follow-up, type and frequency of CMV measurement, and predefined end points for each study. A formal metaanalysis was not performed because of the heterogeneity in study design and diagnostic methods over time.

Data summary and figures were generated in Review Manager 5.3 (Cochrane Collaboration).¹⁵  Dichotomous data were analyzed and presented using forest plots, which summarized treatment effect on all-cause mortality, CMV disease, CMV incidence (as defined by study parameters), and preemptive therapy rate if available. Secondary outcomes, such as drug toxicity and study characteristics, were summarized in tables for qualitative analysis.

The electronic searches retrieved 225 unique results. After reading the titles and abstracts, 11 trials were selected for full review. Of these, 1 was not a prophylaxis trial, and 1 was not randomized. Ten additional studies were selected through identification of relevant references. A trial of late CMV prophylaxis comparing valganciclovir with placebo beginning after day +100 that included patients with prior episodes of CMV infection was not considered further.¹⁶  Nineteen trials were included in this review (Figure 1). Fourteen studies compared an antiviral prophylaxis with placebo, whereas 5 studies compared one antiviral with another.

A total of 4173 patients were randomly assigned in prospective trials and analyzed for the primary outcomes analyses. Of these patients, 58% were male, with an estimated median age of 41 years (range, 1-78 years). The most common underlying malignancy was acute leukemia (46%), followed by myelodysplastic syndrome (12%), lymphoma (12%), and chronic myeloid leukemia (11%). Description of key study characteristics, trial intervention, and standard-of-care CMV management of each study can be found in Table 1. Antiviral drugs used included acyclovir, valacyclovir, ganciclovir, maribavir, brincidofovir, and letermovir. Follow-up times after antiviral prophylaxis ranged from 0 to 265 days, with 14 of 19 studies having at least 28 days of follow-up. Study design outlines, including length of prophylactic treatment and follow-up period, are shown in Figure 2.

Four studies^17-20  were judged to be at high risk of reporting bias because of incomplete reporting of outcome measures. The full summary and graphs for risk of bias can be found in supplemental Figures 1 and 2. The overall quality of the studies was good. Allocation bias was the most common risk identified, because most studies did not report their allocation or randomization methods.^20-30  Four studies randomly assigned patients using a form of interactive voice or Web response system.^31-35  One study randomly assigned patients using a computer-generated table of random numbers.¹⁹  Fourteen of 19 studies were double blinded,^{18-20,22,23,25,26,29-35}  whereas 5 studies did not mention blinding and were assumed to be open label.^{17,21,24,27,28}  Because adverse event reporting can be biased in open-label studies, these studies should be judged carefully in their reporting of drug-related toxicity. A minority of studies had incomplete outcome data.^17,18,27  One study did not have results on CMV infection rate.²²  Three studies lacked results on CMV disease in treatment or control arm as well as adverse event or toxicity data^17,18,20  (Tables 1 and 2).

Because of changing CMV detection standards, the approach to CMV management has evolved over time. Before 2002, a majority of CMV primary infection, reactivation, and disease diagnoses were based on CMV culture. Around 2004, trial diagnostic methods shifted toward culture-independent techniques, including antigenemia and polymerase chain reaction (supplemental Table 3). Supplemental Table 2 summarizes each CMV detection technique used by each study. For phase 2 dose-ranging studies, each dose level is presented separately. Forest plots summarizing all-cause mortality, CMV disease, CMV infection, and preemptive therapy incidences to the predefined end point periods of the individual trials that compared antivirals with placebo are presented in Figure 3. Overall, the 14 trials that compared antiviral prophylaxis with placebo demonstrated effectiveness in reducing incident CMV infection (odds ratio [OR], 0.49; 95% confidence interval [CI], 0.42-0.58), CMV disease (OR, 0.56; 95% CI, 0.40-0.80), and use of preemptive therapy (OR, 0.51; 95% CI, 0.42-0.62; 6 trials) but not all-cause mortality (OR, 0.96; 95% CI, 0.78-1.18). The most relevant toxicity and adverse event data are shown in Table 2.

The first antiviral studied for universal prophylaxis against CMV was acyclovir, with a total of 8 studies published from 1981 to 2006; some studies did not report enrollment periods.^17-24  Five studies were placebo controlled, and 1 study each compared acyclovir with IV ganciclovir, IV acyclovir, and oral valacyclovir. These studies had a total of 1347 participants and initiated prophylaxis between days −8 to −3 before transplantation. Treatment continued for a median time of 102 days (range, 18-216 days), and follow-up continued for a median time of 69 days (range, 0-335 days). The overall results were mixed (Figure 3) and suggested that acyclovir was associated with low toxicity after allogeneic HCT (Table 2). However, although a delay in the onset of CMV reactivation was demonstrated, acyclovir showed nonsignificant efficacy in preventing CMV disease (Figure 3; supplemental Figure 3).

The next antiviral agent studied was ganciclovir, with 5 reported studies spanning from 1990 to 1998.^25-28,30  Three studies compared IV ganciclovir against placebo, 1 compared IV ganciclovir against valacyclovir, and 1 compared oral ganciclovir against IV ganciclovir (supplemental Figure 4). A total of 647 participants were included in these studies. Median treatment time was 101 days (range, 84-128 days), and median follow-up time was 80 days after treatment (range, 0-265 days). These studies suggested that ganciclovir was effective in reducing incidence of CMV infection and CMV disease after allogeneic HCT but was not associated with a reduction in all-cause mortality, likely secondary to drug discontinuation related to clinically significant myelosuppression (Figure 3; supplemental Figure 4; Table 2).

In the last 10 years, 3 novel antivirals have emerged as potential CMV prophylactic candidates and have been studied in an era of highly sensitive molecular testing. Two studies involved maribavir, which is a UL97 viral protein kinase inhibitor that prevents nuclear egress of CMV virions.^36,37  The studies included 792 HCT participants recruited from 2004 to 2008.^29,34  Median treatment time was 84 days (range, 1-92 days), and median follow-up time after treatment was 147 days (range, 56-238 days). Although a dose-escalation phase 2 trial had demonstrated antiviral activity at doses ranging from 100 to 400 mg twice daily,²⁹  the results of the phase 3 trial demonstrated that maribavir at 100 mg twice daily started after engraftment had no significant effect on incidence of CMV disease, CMV reactivation, or preemptive therapy for CMV after allogeneic HCT compared with placebo by HCT week 24 and had no statistically significant effect on mortality (Figure 3).^34,38  Maribavir was largely well tolerated; the proportion of patients with adverse events leading to study drug discontinuation and serious adverse events was largely the same between maribavir and placebo arms. However, patients receiving 400 mg of maribavir twice daily in the phase 2 trial experienced increased rates of nausea and taste disturbance (Table 2).²⁹ 

Brincidofovir (CMX001) is an oral lipid conjugate formulation of cidofovir and was recently evaluated in 2 randomized placebo-controlled studies for prevention of CMV infection in HCT recipients from 2009 to 2015.^32,33  In total, 682 participants were treated for a median duration of 66.5 days (range, 1-99 days) and followed for a median time of 71 days (range, 70-72 days) after treatment. Although a phase 2 dose-ranging trial demonstrated significantly lower CMV events with brincidofovir at a dose of 100 mg twice weekly and a treatment completion rate of 60% when started after engraftment through week 13 post-HCT,³³  brincidofovir did not improve CMV-related outcomes in the phase 3 trial that evaluated treatment at 100 mg twice weekly against placebo beginning a median of 15 days post-HCT (Figure 3), with a low completion rate (38%).³²  Furthermore, brincidofovir was associated with increased rates of diarrhea, acute GVHD with gastrointestinal involvement, other gastrointestinal adverse events, and a nonsignificant increased risk of death when compared with placebo (Table 2).^32,33 

Letermovir is an antiviral agent with a novel mechanism of action involving inhibition of the human CMV terminase complex.^39-43  It was studied in 2 randomized placebo-controlled studies for CMV prophylaxis from 2010 to 2016 with a total of 686 HCT participants.^31,35  In these studies, median treatment time was 77.5 days (range, 1-113 days), and median follow-up time was 122.5 days (range, 7-238 days). Among patients undergoing allogeneic HCT, letermovir at a dose of 480 mg per day (or 240 mg per day when administered concomitantly with cyclosporine) was found to significantly reduce CMV reactivation, use of preemptive anti-CMV therapy, and all-cause mortality by week 24 posttransplantation (Figure 3). Letermovir had a favorable adverse event profile and high treatment completion rate (71%) despite being started preengraftment in a majority of patients (Table 2). Reduction in CMV reactivation and mortality was prominent in patients at higher risk of CMV reactivation and CMV disease, including those undergoing haploidentical HCT or mismatched-donor HCT and those receiving antithymocyte globulin. All-cause mortality was nonsignificantly lower in patients who received letermovir compared with placebo by week 48. The results of this trial led to regulatory approvals by the US Food and Drug Administration, European Medicines Agency, and Health Canada in late 2017.^44-46 

Letermovir is not myelosuppressive and is available in oral and IV formulations, allowing treatment to start a median of 9 days after HCT in the phase 3 trial. Letermovir is excreted by the liver and does not require dose adjustments based on renal or hepatic function except in patients with advanced cirrhosis (Child-Pugh class C).^44,47  Although pharmacokinetic studies have found that letermovir increased exposure to certain drugs,⁴⁴  including atorvastatin, tacrolimus, sirolimus, midazolam, and other medications that may require dosing adjustments, letermovir itself only required dose adjustment (50% reduction) when administered with cyclosporine.⁴⁸  Detailed letermovir characteristics are presented in Table 3.

It is important to note that the last 2 phase 3 trials of CMV prophylaxis conducted in patients undergoing HCT^31,32  imputed premature trial discontinuations for any reason (eg, withdrawal of consent, death, or loss to follow-up) as primary end point events, a conservative approach requested by regulatory agencies, which helps inform the clinical benefit and drug tolerability of the overall strategy. However, for the letermovir phase 3 trial, the proportion of patients with CMV-specific end points was 25 (7.7%) of 325 at the end of the treatment period (week 14, day +100). Of these events, 12 (3.7%) occurred after patients had ended letermovir treatment for a median of 43 days (range, 14-75 days), and 1 patient began preemptive therapy for CMV within the week-14 study window after stopping letermovir. Another 12 events (3.7%) of preemptive therapy occurred while patients were receiving letermovir, but 10 of these events had nonquantifiable (<137 IU/mL) CMV DNA in the central laboratory; only 2 patients (0.6%) had quantifiable CMV viral loads at the time of preemptive therapy, and in 1 (0.3%) of these 2 patients, a mutation (UL56 V236M) that confers letermovir resistance was documented.³¹  Furthermore, no mutations associated resistance were found in patients who experienced CMV reactivation after discontinuation of letermovir.⁴⁹  An additional mutation (UL56 C325W) was identified in 1 of 48 patients who began letermovir treatment with detectable CMV DNA (not part of the primary efficacy population); the patient developed breakthrough CMV viremia a few weeks into treatment.^43,44  Therefore, the on-treatment efficacy of letermovir when used in patients without CMV viremia at the start of prophylaxis was high.

CMV infection has been an obstacle to improved outcomes for patients who undergo allogeneic HCT and are CMV seropositive. Several studies in the past 3 decades have evaluated different antiviral agents in an attempt to find a safe and effective agent to be used as universal prophylaxis. The scope of the treatments covered in this review highlights the longstanding search for suitable CMV prophylaxis stretching from 1981 to present day.

The randomized trials reviewed demonstrate that among the 6 antiviral therapies studied, ganciclovir and letermovir were the most effective in reducing incidence of CMV reactivation when used as universal prophylaxis agents. Furthermore, CMV disease rates have decreased over the study period (Figure 3), in part because of the introduction of more sensitive molecular methods for CMV surveillance and use of preemptive therapy during this time.^50,51  Given the known disadvantages of preemptive therapy, such as treatment with drugs that have frequent toxic effects and an increased overall risk of mortality associated with CMV reactivation, the results presented suggest that patients undergoing allogeneic HCT would significantly benefit from universal prophylaxis with an agent that is tolerable after HCT. The data suggest that although effective at reducing CMV reactivation and disease, ganciclovir use cannot be recommended as a universal prophylaxis agent because of an increased risk of myelosuppression and subsequent drug discontinuation.

In contrast, the data suggest that letermovir has an excellent safety profile, and its use should be considered for this indication in patients at risk. Letermovir was associated with a decrease in CMV-related outcomes and all-cause mortality through 24 weeks after HCT. These benefits are likely due in part to its tolerability, which allowed patients to continue treatment through week 14 posttransplantation, and the possibility of administering IV treatment in patients who were acutely ill or could not take oral medications. Although there were several cases of CMV reactivation in the letermovir arm in the phase 3 trial, a majority of these occurred after the period of drug administration or in patients who discontinued letermovir therapy prematurely.³¹  Given these data, weekly surveillance for CMV reactivation during administration of letermovir may not be necessary for a majority of patients; targeted testing when CMV reactivation is clinically suspected may be a reasonable approach, including the evaluation of fever, cytopenias, or clinical syndromes that could be due to CMV disease. CMV monitoring after discontinuation of letermovir prophylaxis is advisable in patients who remain at higher risk of CMV infection, especially those with GVHD.

Risk of CMV reactivation remains a concern among high-risk patients undergoing HCT, including those undergoing haploidentical HCT, cord-blood recipients, ex vivo T cell–depleted graft recipients, antithymocyte globulin recipients, and patients with grade ≥2 GVHD requiring systemic glucocorticoids for treatment. As such, letermovir use may be preferentially considered in this patient population to prevent CMV reactivation. However, because letermovir does not have any activity against other human herpesviruses, concomitant acyclovir, valacyclovir, or famciclovir should be prescribed to reduce the risk of herpes simplex and varicella zoster clinical events.

On the basis of the criteria for study inclusion at the onset of this review, the trials presented are compelling in their quality. This review prioritized the selection of prospective, randomized studies, most of them double blinded and placebo controlled, which allow more direct comparisons. Through randomization, these studies minimized selection bias and addressed the most important outcomes in thoroughly understanding the feasibility of an antiviral CMV prophylaxis: the impact of the intervention on CMV infection incidence and all-cause mortality. The large number of overall participants also strengthens the qualitative conclusions reached by this review.

A key limitation of the evidence is that many of the studies identified had a small sample size. Only 3 studies had established dosing regimens and large sample sizes (N > 300 patients). Another 3 studies evaluated varying drug doses, affecting the confidence of the overall outcome. Most of the trials included in this review had wide CIs, making it difficult to measure the true efficacy of these interventions. Heterogeneity among studies resulting from vastly differing interventional methods, outcome measures, and study designs over >30 years further limited their comparability, so a formal metaanalysis was not pursued. Three earlier studies also lacked details about primary study end points and experimental design.^17,18,20  Changes in diagnostic sensitivity is another confounding factor when analyzing the results. Because culture-based methods are less sensitive than molecular methods for CMV detection,⁵¹  antiviral interventions during the era of culture-based testing may have seemed more favorable than they were in actuality.

The results from this review reflect the current clinical consensus that most antiviral prophylaxis options to date have been inadequate in overall efficacy, and those that are efficacious against CMV reactivation introduce undesirable toxicities that limit their use. Although the decision to pursue CMV prophylaxis in post-HCT patients has historically been nuanced to balance drug-related toxicity with CMV-related outcomes, the results of recent studies have changed this landscape.

Patients at increased risk for primary CMV reactivation and CMV disease are most likely to benefit from anti-CMV prophylaxis. Additional research is warranted to further refine which particular HCT populations would benefit most from anti-CMV prophylaxis in a rapidly evolving landscape. Further research is also warranted to study the impact of CMV surveillance after the prophylactic period, the optimal threshold at which to initiate preemptive therapy after prophylaxis, and the role of CMV-specific immune monitoring for guiding prophylactic and preemptive CMV strategies.^49,52-55  These parameters will remain fluid and are likely to change in the future with the incorporation of CMV immunotherapies^56-61  and CMV vaccines.^62-66 

The authors thank professor Robert H. Rubin for his enduring mentorship in the study of CMV in transplantation and for his discussion of ideas presented here.

M.P.C. receives salary support from the Detweiler Travelling Fellowship, provided by the Royal College of Physicians and Surgeons of Canada.

Contribution: F.M.M. and K.C. conceptualized this systematic review; K.C., M.P.C., and F.M.M. performed data collection and analyses; K.C. wrote the first draft of the manuscript; M.P.C., S.P.H., H.E., and F.M.M. revised the manuscript and provided intellectual content; and all authors reviewed the manuscript and agreed to its submission in its current form.

Conflict-of-interest disclosure: S.P.H. has received institutional research support from Merck. H.E. has participated in advisory boards for Chimerix, Clinicgene, and Merck. F.M.M. has received institutional research support from Astellas, Chimerix, Merck, and Shire and consulting honoraria from Alexion, Chimerix, Fate Therapeutics, GlaxoSmithKline, LFB, Merck, Roche Molecular Diagnostics, and Shire. The remaining authors declare no competing financial interests.

Correspondence: Francisco M. Marty, Division of Infectious Diseases, Dana-Farber Cancer Institute and Brigham & Women’s Hospital, 75 Francis St, Boston, MA 02115; e-mail: fmarty@bwh.harvard.edu.