Introduction

Since the early years of transplant, a positive CMV serology in the recipient and CMV reactivation have been associated with inferior outcomes after hematopoietic cell transplantation (HCT), reflecting an increase in non-relapse mortality (NRM).1 However, the development and adoption of effective monitoring of CMV either by PCR or pp65 antigenemia (pp65-AG) in the blood, and the use of preemptive antiviral therapy in the late 1990s/early 2000, has had a favorable impact on reducing the incidence of CMV disease and resulted in a decline in CMV-associated mortality.2

Recently, some studies have reported an association of positive CMV serology or early (before day 100) CMV reactivation with decreased incidence of relapse after HCT.3, 4, 5, 6, 7, 8 This protective effect against relapse appears restricted to AML, and, in some single-center studies, is associated with an improvement in overall survival (OS).3, 4, 5, 8 One hypothesis is that CMV reactivation results in expansion of natural killer (NK) cells with a mature phenotype (CD56dim, NKG2C+ and CD57+) that produce INFγ.9 Mature NK cell expansion is hypothesized to enhance antitumor responses, providing a biological rationale for reduced relapse occurring after CMV reactivation. T cells may be also involved as some γδ T cells recognize CMV peptides that are cross-reactive against leukemia cells.10 Consistent with this hypothesis, use of antithymocyte globulin or alemtuzumab (serotherapy) has been reported to abrogate the benefit of CMV reactivation on relapse.11, 12, 13, 14, 15 Published literature is conflicting, as several studies have failed to show a favorable association between CMV serology or reactivation and leukemia relapse or survival.4, 5, 7, 8, 14, 16, 17, 18

None of these prior studies focused on the impact of CMV-recipient serostatus or CMV reactivation in cord blood (CB) recipients. It is critical to look at CMV reactivation outcomes in CB HCT, specifically as CB HCT is unique. CB T cells are naive and there is no transfer of protective memory T cells;19 therefore, CB is considered inherently seronegative and CB transplant is associated with a high incidence of CMV reactivation in the seropositive recipient. CB HCT also tends to be associated with less acute and chronic GvHD, given the degree of mismatch, yet a powerful GvL effect remains.20, 21, 22

Given the disparity in the medical literature for the influence of CMV on leukemia relapse, and lack of sufficient CB data, the primary aim of this study was to use the large multi-institutional database of the Center for International Blood and Marrow Transplant Research (CIBMTR) to more completely analyze the impact of CB-recipient serostatus and early CMV reactivation primarily on leukemia relapse; secondary outcomes included the impact on OS, and NRM, in the era of preemptive therapy.

Patients and methods

Data source

The CIBMTR is a working group of more than 500 transplant centers worldwide that provide detailed patient, disease, transplant characteristics and outcomes of consecutive transplantations. Data are collected at the Statistical Center at the Medical College of Wisconsin or at the Data-Coordinating Center of the National Marrow Donor Program, where computerized checks for discrepancies, physicians' review of submitted data and on-site audits of participating centers ensure data quality. The CIBMTR collects both Transplant Essential Data and Comprehensive Report Form data before transplantation, 100 days (D100) and 6 months after transplantation, and annually thereafter. All subjects whose data were included in this study provided institutional review board-approved consent to participate in the CIBMTR Research Database and have their data included in observational research studies. The Institutional Review Boards of the Medical College of Wisconsin and the National Marrow Donor Program approved this study.

Data collection and criteria for selection

All patients reported to the CIBMTR, and receiving first CB HCT between 2003 and 2010, for AML and ALL using any conditioning regimen, were included. Patients with only former National Marrow Donor Program forms were excluded, given lack of data on CMV reactivation (n=298). Patients from centers with less than 30% completeness of follow-up index were also excluded (n=64). Other exclusion criteria were as follows: no signed informed consent (n=35), lack of day 100 follow-up forms capturing CMV reactivation data (n=70), CMV serostatus match missing and death before transplantation (n=4). A total of 1684 CB recipients with AML and ALL are included in this analysis. The database was locked on 31 August 2013. The completeness index was very good, 99% at 1 year and 97% at 5 years of follow-up.

Data regarding CMV reactivation, surveillance and treatment

The CIMBTR day 100 follow-up forms collect information on the date of onset of CMV reactivation and the site where CMV was identified. No information is collected regarding diagnostic method, level of virus detected or on therapy provided.

Study design and statistical analysis

Patients with AML (n=925) and ALL (n=759) were analyzed separately. For univariate outcomes and the initial multivariable analysis, patients were categorized as D−/R+ or D−/R−, as CB is considered inherently CMV-seronegative. A second multivariable analysis was performed based on the presence or absence of CMV reactivation post transplant as a time-dependent covariate. Patients receiving single (n=944) and double (n=740) cord were analyzed together as randomized trials demonstrate no difference in transplant outcomes of relapse and survival with single versus double cord transplant.23

Variables included time to CMV reactivation, CMV (D/R) serology, recipient age, gender, race, Karnofsky score at HCT, time from diagnosis to HCT, disease risk category based on ASBMT RFI 2014 classification,24 year of transplant (2003–2006 versus 2007–2010), conditioning intensity, use of total body irradiation, use of serotherapy, GvHD prophylaxis regimen and development of acute and/or chronic GvHD post transplant. For multivariable analyses, the main effect variable was either D/R CMV serology (D–/R– as reference versus D–/R+) or CMV reactivation as a time-dependent covariate (yes versus no). Additional variables analyzed in the models included age (⩽10 years versus 10–30 versus >30 years), disease risk category, conditioning intensity, serotherapy (Yes versus No), GvHD prophylaxis (Tacrolimus/Cyclosporine+Methotrexate±Others versus Tacrolimus/Cyclosporine+Others versus Others) and acute GvHD (aGvHD) occurring before CMV reactivation as a time-dependent covariate.

Patient, disease and transplant-related factors were compared between groups using the Pearson X2-test for discrete variables and the Kruskal–Wallis test for continuous variables. Probabilities of disease-free survival (DFS) and OS were calculated using the Kaplan–Meier estimator. Values for other end points were generated using cumulative incidence estimates to account for competing risks. OS was defined as the time to death from any cause with surviving patients censored at time of last follow-up. DFS was defined as the time to relapse or death from any cause. NRM was defined as death without evidence of disease with relapse as a competing risk. Relapse was recurrence/progression of acute leukemia with death as the competing risk. For aGvHD grades II–IV and chronic GvHD (cGvHD) of any severity, death was the competing risk and patients were censored at time of relapse. In both multivariable analyses of CMV serology and CMV reactivation, the proportional hazard assumption was examined. If violated, it was included as a time-dependent covariate. A stepwise selection procedure was used. Interactions between the main effect and significant covariates were examined. A P-value <0.05 was considered significant. SAS v9.3 (Cary, NC, USA) was used for statistical analysis.

Results

Distribution of CMV serology and CMV reactivation

Table 1 provides basic information on the patients included in the analysis. A total of 1011 CB recipients were CMV-serostatus-positive, with 606 (66%) in the AML cohort and 405 (53%) in the ALL cohort reflecting the older median age for patients with AML [(ALL, median 12 (<1–68) years versus AML, 28 (<1–79) years; P<0.001). The median time to CMV reactivation for the entire cohort was 34 days (range, 2–287) after HCT. It was similar for the AML cohort regardless of the D/R serostatus; however, for ALL patients, the 4% of D−/R− who developed reactivation occurred significantly later after transplant. Nearly all CMV reactivations occurred in the first 3 months after transplant.

Table 1 Characteristics of patients undergoing cord blood transplant between 2003 and 2010 for AML and ALL
Full size table

Relapse

For patients with AML, relapse by 1 year occurred in 27% (95% confidence interval (CI), 23–31%) of seropositive recipients and 26% (95% CI, 21–31%) of seronegative recipients (P 0.77; Table 2 and Figure 1a). Relapse at 1 year was slightly lower in patients with ALL, but there was no effect of recipient CMV serostatus (D−/R+, 20% (95% CI, 16–24%); D−/R−, 18% (95% CI, 14–22%), P=0.49; Table 2 and Figure 1b).

Table 2 Univariate analysis for 3-year outcomes according to CMV serologic status of the donor and recipient
Full size table
Figure 1
figure 1

Cumulative incidence curves of relapse for patients with AML (a) and ALL (b) based upon the CMV serostatus of the recipient.

Full size image

In multivariable analysis adjusted for other known risk factors for relapse, CMV-seropositive recipients with AML had a higher risk of relapse (RR 1.39, P=0.022) compared with seronegative recipients (Table 3). For ALL patients, the risk of relapse was slightly higher for the CMV-seropositive recipients, albeit not statistically significantly (Table 4). CMV reactivation as a time-dependent event did not favorably have an impact on relapse for patients with either AML or ALL [(AML: RR 0.8 (95% CI, 0.62–1.04), P=0.097; ALL: RR 1.01 (95% CI, 0.70–1.46), P=0.95). As expected, AML and ALL patients with high-risk disease or recipients of reduced intensity conditioning had a higher risk of relapse. ALL patients developing acute GvHD were less likely to have disease relapse (RR of 0.7 (95% CI, 0.52–0.94), P<0.02).

Table 3 Multivariable outcomes for patients with AML receiving CB transplant
Full size table
Table 4 Multivariable outcomes for patients with ALL receiving CB transplant
Full size table

GvHD

Acute GvHD grades II–IV by day 100 were similar regardless of disease indication for transplant or CMV-recipient serostatus (AML D−/R+: 43% (95% CI, 39–47%); AML D−/R−: 40% (95% CI, 34–45%); ALL D−/R+: 39% (95% CI, 34–44%); ALL D−/R−: 40% (95% CI, 35–45%)). cGvHD by 3 years did not differ (AML D−/R+: 30% (95% CI, 26–33%); AML D−/R−: 34% (95% CI, 28– 39%); ALL D−/R+: 31% (95% CI, 27–36%); ALL D−/R−: 35% (95% CI, 30–40%)).

In multivariable analysis, positive CMV serology had no impact on the incidence of aGvHD or cGvHD in AML or ALL. CMV reactivation occurring before the onset of aGvHD demonstrated a slightly increased but not statistically different risk of aGvHD (AML: RR 1.31 (95% CI, 0.96–1.79), P=0.083; ALL: RR 1.38 (95% CI, 0.96–1.98), P=0.084). CMV reactivation had no impact on the development of cGvHD when adjusted for other factors.

Other factors having an impact on development of aGvHD grades II–IV included the use of serotherapy for either disease; although for patients with AML, this effect became attenuated after 1 month post transplant (Table 3). There was also a decreased risk of aGvHD for patients with AML receiving a reduced intensity conditioning regimen, but this only occurred in the first month from transplant. As expected, cGvHD was decreased in patients receiving serotherapy, regardless of underlying malignancy (Table 3). In addition, for patients with ALL, older age and prior aGvHD increased the risk for development of cGvHD (Table 4).

Non-relapse mortality

Recipient CMV serology and CMV reactivation were both associated with increased NRM. In univariate analysis, patients with AML who were D−/R+ had an increased risk of NRM at 1 and 3 years compared with D−/R− pairs (1 year, D−/R+: 26% (95% CI, 23–30%) versus D−/R−: 18% (95% CI, 14–22%), P=0.003; Table 2 and Figure 2a). This was similar in patients with ALL as well (Table 2 and Figure 2b). This effect remained significant for both AML and ALL when adjusted for other factors in multivariable analysis (Table 3 and Table 4). Similarly, CMV reactivation increased NRM in multivariable analysis such that patients with AML had a 1.4-fold greater risk (95% CI, 1.07–1.85; P <0.02) and those with ALL were about 1.6 times as likely to die of NRM (95% CI, 1.14–2.23; P>0.007).

Figure 2
figure 2

Cumulative incidence curves of NRM for patients with AML (a) and ALL (b) based upon the CMV serostatus of the recipient.

Full size image

For patients with AML, older age was also associated with increased NRM. For ALL, increased age, use of serotherapy and the development of aGvHD were all associated with increased NRM.

Survival

As shown in Table 2, both DFS and OS were inferior for patients with acute leukemia who were CMV seropositive (Figure 3a, b). When adjusting for other factors, this effect was no longer significant in AML patients (Table 3). However, for ALL patients, CMV-seropositive recipients had a higher risk of death from any cause (Table 4). Use of serotherapy had an adverse impact on DFS. For ALL patients not receiving serotherapy, DFS was similar to CMV-seronegative recipients; however, the use of serotherapy was associated with significantly inferior DFS in CMV-seropositive recipients.

Figure 3
figure 3

Overall survival for patients with AML (a) and ALL (b) based upon CMV serostatus of the recipient.

Full size image

When adjusting for CMV reactivation, there was no impact on DFS or OS for patients with AML (DFS, RR 1.05 (95% CI, 0.87–1.27), P=0.61; OS: RR 1.07 (95% CI, 0.88–1.30), P=0.48); however, patients with ALL and a CMV reactivation had inferior DFS (RR 1.31 (95% CI, 1.02–1.67), P=0.03) and higher risk of death (RR 1.44 (95% CI, 1.12–1.84), P<0.005).

Regardless of type of acute leukemia, patients with more advanced disease and older age had a higher risk of death and impaired DFS. For patients with ALL, patients receiving reduced intensity conditioning had inferior DFS and OS. Specific causes of death for all patients by the D/R serostatus are shown in Table 5.

Table 5 Causes of death for patients with AML or ALL undergoing CB HCT based on D/R CMV serostatus
Full size table

Discussion

Several studies have examined the impact of CMV reactivation specifically in AML after HCT. However, none of these studies examined outcomes in the CB setting. This CIBMTR analysis of 1684 patients was specifically designed to look at the influence of CMV serology and early CMV reactivation on relapse and survival following CB HCT for either AML or ALL. Our data suggest that CMV reactivation does not prevent relapse of AML or ALL in patients receiving CB HCT. Furthermore, our study highlights a persistent negative impact for CB recipients who are CMV seropositive and for patients experiencing CMV reactivation after CB HCT.

The lack of benefit of CMV seropositivity and CMV reactivation seen in our analysis contrasts with other studies that have described a protective effect for positive CMV serology or CMV reactivation in preventing AML relapse.4, 5, 7, 8, 17, 25 However, other studies have reported a lack of benefit from positive CMV serology and a negative effect of early CMV reactivation with regards to the risk of leukemia relapse.18, 26 A recent analysis from the European Blood and Marrow Transplant Group demonstrated a higher risk of leukemia relapse and poorer OS associated with positive CMV serology.17, 27

CMV reactivation and positive CMV serology (D−/R+ versus D−/R−) increased NRM in our cohort. Similar to the European Blood and Marrow Transplant Group study, we found that CMV-seropositive recipients were more likely to have infection reported as a primary cause of death compared with CMV-seronegative CB recipients (R+, 21%; R−, 14%). Notably, there was a roughly similar increase in recurrence/persistent disease as the primary cause of death in CMV-seronegative recipients (R+, 37%; R−, 48%). Despite these findings, there was no protection from relapse seen in either univariate or multivariable analysis. Since an indirect effect of CMV serology on the risk of infections with other pathogens has been reported, it is possible that any small decrease in relapse risk is counterbalanced by an increase in fatal infections.28

For patients with ALL who received either ATG or alemtuzumab serotherapy, CMV reactivation and a positive CMV-recipient serology was detrimental to LFS. Influence of serotherapy on LFS may be explained by the higher likelihood of CMV reactivation in these patients.12 CMV-positive serology (P=0.0007) and CMV reactivation (P=0.0042) had a negative impact on OS in ALL. This negative impact in ALL is not surprising, given the negative impact on both NRM and LFS in ALL CB recipients. Notably, CMV status (positive serology or CMV reactivation) had no impact on LFS or OS in AML. It is interesting that, despite its negative impact on NRM, the CMV status did not have a negative impact on OS in AML CB. The differences in outcomes between AML and ALL cannot be solely attributed to disease category as non-disease-related factors, such as age and prior chemotherapy, may have had a role as well. Within the ALL group, predominantly younger patients, recipient CMV seropositivity was a stronger predictor of NRM compared with AML patients, which may account for the lack of impact of the CMV status on LFS and OS for AML patients.

Although multicenter registry studies such as ours have the advantage of a large sample size, the caveat is the lack of detailed information. Data with regards to the viral monitoring test used to diagnose CMV reactivation, frequency of testing or preemptive versus prophylactic policies applied by each center were not available. In addition, information regarding antiviral therapy itself, trigger for initiation, timing of initiation and total duration of treatment for CMV reactivation were not captured. Moreover, CMV reactivation may be under-reported in multicenter registry studies. This is likely explained by a trend for reporting to registries only CMV reactivations that are considered to have a clinical importance because of their level or duration. The lack of detailed data may introduce bias diluting the effect of CMV serostatus and CMV reactivation; however, the large number of patients and centers should counteract this bias somewhat.

The implications of our CB study showing the persistent negative impact of the CMV serostatus and reactivation on NRM, despite the existence of effective antiviral pharmacotherapy, are far-reaching. Although antiviral pharmacotherapy, diligent surveillance and early preemptive antiviral therapy have significantly reduced the progression to CMV disease and mortality observed before this era, there remains significant limitations to current therapy including myelosuppression (ganciclovir) and electrolyte imbalances/renal impairment (foscarnet). Our data justify the need for development of novel antiviral drugs, with superior efficacy and toxicity profiles.29, 30 For example, a prophylactic rather than preemptive approach may lead to superior outcomes in CB stem cell transplants. Some centers have reported using IV ganciclovir during conditioning regimen (days −8 to −2) and high-dose valacyclovir prophylaxis (2 gm every 8 h) in the post-transplant setting in their CMV-seropositive CB recipients. An alternative approach to antiviral pharmacotherapy, especially in CB HCT, is adoptive cellular immunotherapy for CMV and other viral diseases.31, 32 Adoptive cellular immunotherapy, after having amassed an excellent record for efficacy and safety, is advancing from therapy of refractory disease to prophylaxis. Boosting immune reconstitution and antiviral immunity post (CB) HCT rather than focusing on pharmacotherapy has the potential to be cost effective.33, 34, 35 Regardless of the approach, our data confirm that there is a great scope for further improving the management of CMV after CB HCT.

In conclusion, in this only multicenter data set of CB recipients, CMV reactivation is associated with decreased DFS and OS in ALL but not in AML. Positive recipient CMV serology and CMV reactivation continue to contribute to increased NRM even in the era of preemptive treatment. If CMV reactivation has an effect on decreasing AML relapse, our study suggests that the benefit is small and likely balanced by the increased NRM, particularly in the setting of CB HCT.