Abstract
Cytomegalovirus (CMV) infection post-hematopoietic cell transplantation (HCT) remains a significant cause of morbidity and mortality. While letermovir prophylaxis is available for CMV-seropositive recipients, optimal donor selection for CMV-seronegative recipients remains unclear, with donor age often prioritized over CMV serostatus. We investigated the relative impact of donor age and CMV serostatus in CMV-seronegative recipients (n=1013) with either CMV-seropositive (n=318) or CMV-seronegative donors (n=695), who underwent HCT with HLA-matched sibling donors with calcineurin inhibitor-based or post-transplant cyclophosphamide (PTCy)-based graft-versus-host disease prophylaxis, or haploidentical donors with PTCy. Multiple conventional and machine-learning approaches were employed to account for confounding factors. Across all analyses, CMV-seropositive donor status was consistently associated with significantly inferior overall survival (OS) and disease-free survival, primarily driven by increased non-relapse mortality (NRM). Older donor age showed a statistically significant association with OS and DFS in some but not all models, and its effect size was small (about 1% increased hazard per year in Cox proportional hazard model) compared to the substantial impact of CMV seropositive donors (about 27-33% increased hazards for worse OS, approximately 50-60% higher NRM in Cox proportional hazard model). However, our machine learning model revealed a more nuanced and complex non-linear effect of donor age, further demonstrating the adverse impact of donor CMV serostatus. Our findings suggest that prioritizing a CMV-seronegative donor may be associated with improved outcomes in CMV-seronegative recipients. Further research is needed to validate these findings and explore underlying mechanisms.
Keywords: Donor age, Cytomegalovirus, CMV, HLA matched sibling donor, Haploidentical, PTCy, post-transplantation cyclophosphamide, CNI, calcineurin inhibitor
INTRODUCTION
Cytomegalovirus (CMV) is one of the most common causes of viral infections after an allogeneic hematopoietic stem cell transplant (HCT), and it is associated with increased non-relapse mortality (NRM) and poorer overall survival (OS).1–4 Letermovir is approved by the United States Food and Drug Administration for preventing CMV infection, but only in CMV-seropositive patients. While CMV-seronegative recipients have a low risk of CMV reactivation, especially when the donor is also seronegative, the risk increases substantially with a seropositive donor.1–3 Therefore, a CMV-seronegative donor is preferred for seronegative recipients. However, donor age is typically prioritized over other non-HLA factors during donor selection, as it has been associated with improved OS in numerous studies in the unrelated donor setting.5–7 It remains unclear whether this principle applies to CMV-seronegative recipients, where the risk of CMV reactivation varies significantly based on donor serostatus. We aimed to investigate whether donor age or CMV serostatus has a greater impact on OS in CMV-seronegative recipients, to inform donor selection decisions in this specific patient population.
SUBJECTS AND METHODS
Study population
We used publicly available datasets8 from the Center for International Blood and Marrow Transplant Research (CIBMTR) from the referenced publications.9,10 These datasets included patients who underwent related donor HCT with either HLA-matched sibling donors (MSD) using calcineurin inhibitor (CNI)-based or post-transplant cyclophosphamide (PTCy)-based graft-versus-host disease (GVHD) prophylaxis, [Goldsmith et al10] or HLA-haploidentical donor with PTCy [Fuchs et al9]. To expand the sample size and avoid overlap, the haploidentical group from the Goldsmith dataset10 (n=757) was replaced with the larger haploidentical group from the Fuchs dataset9 (n=1434). We included this seemingly diverse population because, as shown in the original CIBMTR analysis,10 although PTCy use was associated with an increased risk of CMV infection in both PTCy groups (haploidentical, MSD-PTCy, and MSD-CNI) compared to the MSD-CNI group, this effect was only observed in CMV-seropositive recipients. Notably, there were no significant differences in the risk of CMV infection by donor type (haploidentical or MSD) or by GVHD prophylaxis (PTCy or CNI) in the CMV-seronegative recipients. Only CMV-seronegative recipients were included in our study, with the cohort stratified by donor CMV serostatus: CMV- seronegative donors (n=695) and CMV-seropositive donors (n=318).
Statistical approaches
We observed significant differences in baseline characteristics between the two donor CMV groups, particularly in donor age, a key variable of interest (Table 1). Therefore, to account for biases associated with these differences, we employed several complementary analytical procedures. First, we fitted a Cox proportional hazards (PH) model to determine the association of donor age and CMV serostatus, among all other covariates listed in Table 1, with OS. Second, we used inverse probability of treatment weighting (IPTW) to estimate the marginal effects of donor age and donor CMV serostatus on OS. IPTW is a statistical technique that can be used to adjust for confounding by balancing covariates between treatment groups. By assigning weights to each observation based on the probability of receiving the treatment (being CMV seropositive donor) given their observed covariates, IPTW creates a pseudo-population in which the distribution of covariates is similar across the groups. This allows for a potentially more accurate estimation of the marginal effect of donor CMV serostatus on OS by reducing the impact of confounding variables. A standardized mean difference (SMD) plot illustrating covariate balance between the two donor CMV groups before (red dots) and after (blue dots) IPTW is shown in Figure S1. Nearly all covariates exhibit excellent balance, with absolute SMD values very close to zero.
Table 1:
Baseline Characteristics
| Overall population | Matched-pair cohort | |||||
|---|---|---|---|---|---|---|
| CMV Neg Donor | CMV Pos Donor | p | CMV Neg Donor | CMV Pos Donor | p | |
| (n=695) | (n=318) | (n=555) | (n=318) | |||
| Recipient age, years (median [IQR]) | 55.0 [39.4, 63.6] | 55.3 [37.8, 64.0] | .92 | 55.4 [40.9, 64.1] | 55.3 [37.8, 64.0] | .51 |
| Donor age, years (median [IQR]) | 43.7 [28.9, 56.9] | 50.6 [35.4, 61.2] | <.001 | 49.0 [32.9, 58.8] | 50.6 [35.4, 61.2] | .06 |
| Donor type, n (%) | .37 | .92 | ||||
| MSD-CNI | 325 (46.8) | 163 (51.3) | 281 (50.6) | 163 (51.3) | ||
| Haplo-PTCy | 291 (41.9) | 119 (37.4) | 206 (37.1) | 119 (37.4) | ||
| MSD-PTCy | 79 (11.4) | 36 (11.3) | 68 (12.3) | 36 (11.3) | ||
| Disease, n (%) | .63 | .89 | ||||
| AML | 424 (61.0) | 185 (58.2) | 332 (59.8) | 185 (58.2) | ||
| MDS | 192 (27.6) | 97 (30.5) | 162 (29.2) | 97 (30.5) | ||
| ALL | 79 (11.4) | 36 (11.3) | 61 (11.0) | 36 (11.3) | ||
| Disease stage*, n (%) | ||||||
| Early-Intermediate | 485 (69.8) | 227 (71.4) | .66 | 383 (69.0) | 227 (71.4) | .51 |
| Advanced | 210 (30.2) | 91 (28.6) | 172 (31.0) | 91 (28.6) | ||
| Conditioning intensity, n (%) | 1.00 | .87 | ||||
| MAC | 340 (48.9) | 155 (48.7) | 275 (49.5) | 155 (48.7) | ||
| RIC/NMA | 355 (51.1) | 163 (51.3) | 280 (50.5) | 163 (51.3) | ||
| Graft type, n (%) | .09 | .27 | ||||
| Bone Marrow | 215 (30.9) | 81 (25.5) | 162 (29.2) | 81 (25.5) | ||
| Peripheral blood | 480 (69.1) | 237 (74.5) | 393 (70.8) | 237 (74.5) | ||
| HCT-CI, n (%) | .08 | .13 | ||||
| 0-2 | 343 (49.4) | 177 (55.7) | 277 (50) | 177 (55.7) | ||
| >=3 | 351 (50.6) | 141 (44.3) | 277 (50) | 141 (44.3) | ||
| KPS, n (%) | .004 | .04 | ||||
| <90 | 188 (27.1) | 59 (18.6) | 138 (24.9) | 59 (18.6) | ||
| >=90 | 507 (72.9) | 259 (81.4) | 417 (75.1) | 259 (81.4) | ||
| D/R Sex, n (%) | <.001 | <.001 | ||||
| Male/Male | 291 (41.9) | 96 (30.2) | 221 (39.8) | 96 (30.2) | ||
| Female/Female | 97 (14.0) | 61 (19.2) | 86 (15.5) | 61 (19.2) | ||
| Female/Male | 154 (22.2) | 108 (34.0) | 128 (23.1) | 108 (34.0) | ||
| Male/Female | 153 (22.0) | 53 (16.7) | 120 (21.6) | 53 (16.7) | ||
| Race, n (%) | <.001 | <.001 | ||||
| White | 630 (90.6) | 257 (80.8) | 502 (90.5) | 257 (80.8) | ||
| Others (%) | 65 (9.4) | 61 (19.2) | 53 (9.5) | 61 (19.2) | ||
| Time to HCT, n (%) | .08 | .04 | ||||
| <6 months | 357 (51.4) | 143 (45.0) | 291 (52.4) | 143 (45.0) | ||
| 6-12 months | 179 (25.8) | 83 (26.1) | 143 (25.8) | 83 (26.1) | ||
| >12 months | 159 (22.9) | 92 (28.9) | 121 (21.8) | 92 (28.9) | ||
| Year of HCT, n (%) | .17 | .48 | ||||
| <2014 | 251 (36.1) | 130 (40.9) | 212 (38.2) | 130 (40.9) | ||
| = >2014 | 444 (63.9) | 188 (59.1) | 343 (61.8) | 188 (59.1) | ||
| Follow-up among survivors, months (median [IQR]) | 35.8 [24.0-48.3] | 36.0 [24.1-48.4] | .68 | 35.5 [23.7-48.4] | 36.0 [24.1-48.4] | .48 |
Defined as early, intermediate or advanced based on CIBMTR defined criteria
Abbreviations: ALL, Acute Lymphoblastic Leukemia; AML, Acute Myeloid Leukemia; CMV, Cytomegalovirus; CNI, Calcineurin Inhibitor; D/R, Donor/Recipient; Haplo, Haploidentical; HCT, Hematopoietic Cell Transplantation; HCT-CI, Hematopoietic Cell Transplantation Comorbidity Index; IQR, Interquartile Range; KPS, Karnofsky Performance Status; MAC, Myeloablative Conditioning; MDS, Myelodysplastic Syndromes; MSD, Matched Sibling Donor; Neg, Negative; Pos, Positive; PTCy, Post-Transplant Cyclophosphamide; NMA, Non-Myeloablative Conditioning; RIC, Reduced-Intensity Conditioning.
Next, we performed a 1:2 propensity score matching [matched for recipient age, donor age, donor type (haploidentical vs MSD), disease, disease stage, graft type and conditioning intensity] of the two donor CMV groups to enhance comparability between donor groups and fitted Cox PH models on the matched dataset. Not all CMV-seropositive donors could be matched with CMV-seronegative donors, resulting in a matched-pair cohort that included CMV-seronegative donors (n=555) and CMV-seropositive donors (n=318). Density plots showing donor age and recipient age by donor CMV groups in the matched-pair cohort are shown in Figure S2, showing that this approach nicely balanced donor and recipient age across the two cohorts. While the matched dataset helps reduce confounding and improve internal validity, it can also potentially miss true associations by significantly reducing the sample size. Since all models have unique advantages and disadvantages, their findings should be considered complementary.
As donor age was another key variable of interest, we retained it as a continuous variable to preserve maximum information. We did not find any strong evidence of non-linearity for either donor age or recipient age. We tested the linearity assumption using likelihood ratio tests to evaluate the fit of models with donor and recipient ages as continuous variables, natural splines, B-splines, and polynomial terms (squared and cubed). More complex age transformations did not significantly improve model fit over models using recipient and donor ages as continuous variables, which provided the best fit and simplest interpretation [Table S1].
We noted a moderately-strong correlation between donor age and recipient age (r = 0.61, p < .0001). To tackle this and other potential complex interactions between several patient and donor-related factors, we then employed a machine learning technique - Random Survival Forest (RSF) for two main survival outcomes [OS and disease-free survival (DFS)], to complement other models described above. RSF constructs an ensemble of decision trees, each trained on a random subset of the data and predictor variables. This approach enables RSF to capture complex relationships and variable interactions without strict statistical assumptions, and it is less sensitive to multicollinearity. The details of RSF, IPTW and Matched Pair analyses are provided in supplemental file.
Outcomes
Our primary outcomes of interest were OS and DFS (survival without disease recurrence). Secondary outcomes included relapse, NRM (mortality from causes other than relapse), grade II-IV acute GVHD, and chronic GVHD. Baseline characteristics between the groups (CMV-seropositive vs CMV-seronegative donors) were summarized using descriptive statistics with median and interquartile range for continuous variables and numbers with percentages for categorical variables. These were compared between the groups using the χ2 test for categorical variables, and the Mann-Whitney U test for continuous variables. In univariate analysis, the Kaplan-Meier method was used to estimate OS and DFS probabilities. The cumulative incidence method accounting for competing risks was used to estimate the incidences of relapse, NRM, and GVHD. Competing risks considered were death before relapse (NRM) for relapse; disease relapse for NRM; and death for GVHD. All variables listed in Table 1 were considered in the univariate analysis. Multivariable analysis was performed using Cox PH models. Cause-specific hazards are reported for all outcomes with competing events. Factors that were significant (p<.10) in the univariate analysis were introduced in the multivariate model. Then, a stepwise selection model was used to identify all significant variables (p≤.05) that were retained in the final multivariate model. Donor age and donor CMV serostatus were kept in all final models regardless of their significance. We tested the PH assumption for all variables graphically and by including interaction terms with time in the model. Models were stratified by variables that violated PH assumptions. Interaction effects between the main effects (donor CMV serostatus and donor age) and other statistically significant covariables were tested in multivariable regression analyses and accounted for by stratification when identified. RSF analyses were performed using Python with the scikit-survival, lifelines, Optuna, and scikit-learn libraries. All other analyses were done using R Statistical Software (version 4.3.1; R Core Team 2023).
RESULTS
Table 1 presents the baseline characteristics of the entire study cohort as well as matched-pair cohort stratified by donor CMV serostatus. Recipients of CMV-seronegative donors had significantly younger donors (median age 43.7 years vs. 50.6 years, p < .001) in the overall cohort but not in the matched-pair cohort (median age 49.0 years vs. 50.6 years, p=.06). Median recipient age was similar in both groups (approximately 55 years). Recipients in the CMV-seronegative donor group were more likely to have a Karnofsky Performance Status (KPS) < 90 (27% vs. 19%, p < .001), were less likely to be female-to-male donor-recipient pairs (22% vs 34%, p < .001), and were more likely to be White (91% vs 81%, p < .001). These differences in KPS, race, and donor-recipient sex persisted in the matched-pair cohorts but were well-balanced in the IPTW cohort (Figure S1). Donor type, disease, disease status, conditioning intensity, graft source, HCT-CI, and year of transplantation were comparable between the two groups. Median follow-up among survivors was approximately 36 months in both the groups.
Overall survival
In Cox PH multivariate analysis (MVA), both older donor age [hazard ratio (HR) 1.01, 95% confidence interval (CI) 1.002-1.02, p = .02] and CMV-seropositive donor status (HR 1.31, 95% CI 1.07-1.61, p = .009) were independently associated with inferior overall survival (OS) [Table 2]. No significant interactions were observed between donor age and CMV serostatus (p = .75), donor age and recipient age (p = .92), or between these variables and other covariates in the model. Inverse probability of treatment weighting and matched-pair analyses, however, did not reveal a significant effect of donor age (IPTW: HR = 1.01, p =.23; matched-pair: HR = 1.005, p=.20), but maintained a significant association between CMV-seropositive donor status and reduced OS (IPTW: HR 1.33, p=.01; matched-pair: HR 1.27, p=.02). Donor type was not a significant predictor of OS. Kaplan-Meier OS curves stratified by donor CMV serostatus in the entire cohort (Figure 1a) and separately within each donor type are shown (Figure S3a–c).
Table 2:
Multivariate Analysis of outcomes
| Cox PH (Entire cohort) | IPTW Analysi (Entire cohort) | Cox PH (Matched-Pair Cohort) | |||||||
|---|---|---|---|---|---|---|---|---|---|
| OS* | HR | 95% | P | HR | 95% | p | HR | 95% | p |
| Donor Age | 1.01 | 1.002-1.02 | .02 | 1.01 | 0.996-1.01 | .23 | 1.005 | 0.997-1.01 | .20 |
| Donor: CMV Seropositive | 1.31 | 1.07-1.61 | .009 | 1.33 | 1.07-1.65 | .01 | 1.27 | 1.03-1.56 | .02 |
| Recipient Age | 1.01 | 1.0004-1.02 | .04 | 1.01 | 1.002-1.02 | .01 | 1.01 | 1.001-1.02 | .02 |
| HCT-CI >=3 | 1.41 | 1.16-1.72 | .0006 | 1.49 | 1.20-1.85 | .0003 | 1.39 | 1.14-1.72 | .002 |
| KPS<90 | 1.32 | 1.06-1.65 | .01 | 1.29 | 1.01-1.64 | .04 | 1.38 | 1.10-1.74 | .005 |
| DFS ** | |||||||||
| Donor Age | 1.009 | 1.001-1.016 | .03 | 0.99 | 0.99-1.01 | .30 | 1.009 | 0.99-1.01 | .15 |
| Donor: CMV Seropositive | 1.24 | 1.03-1.49 | .02 | 1.25 | 1.04-1.51 | .02 | 1.24 | 1.03-1.49 | .03 |
| Recipient Age | 1.006 | 0.99-1.01 | .08 | 1.01 | 1.003-1.02 | .01 | 1.006 | 1.00-1.02 | .04 |
| HCT-CI >=3 | 1.43 | 1.19-1.70 | <.001 | 1.54 | 1.27-1.87 | <.001 | 1.37 | 1.14-1.65 | <.001 |
| KPS <90 | 1.39 | 1.14-1.70 | <.001 | 1.35 | 1.10-1.67 | <.001 | 1.45 | 1.18-1.79 | <.001 |
| Year HCT >=2014 | 0.86 | 0.72-1.04 | .12 | 0.84 | 0.69-1.02 | .08 | 0.82 | 0.68-1.00 | .05 |
| Relapse ** | |||||||||
| Donor Age | 1.01 | 1.00-1.01 | .07 | 1.00 | 1.00-1.01 | .26 | 1.01 | 1.00-1.01 | .18 |
| Donor: CMV Seropositive | 1.12 | 0.90-1.39 | .30 | 1.11 | 0.89-1.38 | .34 | 1.12 | 0.90-1.40 | .30 |
| HCT-CI >=3 | 1.26 | 1.03-1.55 | .03 | 0.79 | 0.63-1.00 | .05 | 1.22 | 0.98-1.52 | 07 |
| KPS<90 | 1.48 | 1.17-1.86 | <.001 | 1.35 | 1.07-1.70 | .01 | 1.57 | 1.23-2.01 | <.001 |
| Year HCT >=2014 | 0.84 | 0.68-1.04 | .11 | 1.38 | 1.06-1.79 | .02 | 0.79 | 0.63-0.98 | .03 |
| RIC/NMA | 1.25 | 1.02-1.55 | .03 | 1.40 | 1.11-1.78 | .01 | 1.25 | 1.00-1.56 | .05 |
| NRM *** | |||||||||
| Donor Age | 1.008 | 0.99-1.02 | .30 | 1.006 | 0.99-1.02 | .48 | 1.005 | 0.99-1.02 | .54 |
| Donor: CMV Seropositive | 1.49 | 1.07-2.08 | .02 | 1.58 | 1.13-2.26 | .009 | 1.47 | 1.05-2.05 | .02 |
| Recipient Age | 1.02 | 1.006-1.03 | .004 | 1.03 | 1.01-1.04 | .001 | 1.02 | 1.007-1.04 | .004 |
| Early-Intermediate disease | 0.77 | 0.55-1.07 | .12 | 0.68 | 0.47-0.98 | .04 | 0.75 | 0.54-1.06 | .10 |
| Peripheral blood graft | 1.39 | 0.92-2.11 | .12 | 1.49 | 0.94-2.35 | .09 | 1.52 | 0.99-2.35 | .06 |
| HCT-CI >=3 | 1.87 | 1.33-2.60 | <.001 | 1.95 | 1.35-2.82 | <.001 | 1.74 | 1.24-2.45 | <.001 |
| Acute GVHD, grade II-IV **** | |||||||||
| Donor Age | 1.01 | 1.00-1.01 | .15 | 1.00 | 1.00-1.01 | .26 | 1.00 | 1.00-1.01 | .30 |
| Donor: CMV Seropositive | 0.94 | 0.74-1.20 | .61 | 1.00 | 0.78-1.28 | .99 | 0.95 | 0.74-1.21 | .67 |
| Donor type | |||||||||
| Haplo-PTCy | 1.58 | 1.22-2.06 | <.001 | 1.63 | 1.22-2.17 | <.001 | 1.59 | 1.20-2.10 | <.001 |
| MSD-PTCy | 1.45 | 1.02-2.07 | .04 | 1.68 | 1.12-2.50 | .01 | 1.49 | 1.03-2.16 | .03 |
| Peripheral blood graft | 1.64 | 1.24-2.17 | <.001 | 1.60 | 1.17-2.19 | <.001 | 1.61 | 1.19-2.17 | <.001 |
| Chronic GVHD | |||||||||
| Donor Age | 1.00 | 0.99-1.00 | .40 | 1.00 | 0.99-1.00 | .20 | 1.00 | 0.99-1.00 | .28 |
| Donor: CMV Seropositive | 0.91 | 0.73-1.13 | .39 | 0.92 | 0.73-1.15 | .45 | 0.88 | 0.70-1.10 | .26 |
| Peripheral blood graft | 2.10 | 1.59-2.76 | <.001 | 2.02 | 1.47-2.77 | <.001 | 2.04 | 1.51-2.75 | <.001 |
| D/R Sex | |||||||||
| Female/Female | 1.32 | 0.98-1.77 | .07 | 1.52 | 1.11-2.07 | .01 | 1.29 | 0.94-1.77 | .11 |
| Female/Male | 1.38 | 1.07-1.77 | .01 | 1.45 | 1.10-1.91 | .01 | 1.45 | 1.11-1.90 | .01 |
| Male/Female | 1.07 | 0.81-1.41 | .62 | 1.18 | 0.86-1.62 | .30 | 1.00 | 0.74-1.34 | .98 |
Stratified by disease stage
Stratified by disease stage and donor type
Stratified by donor type
Stratified by conditioning intensity and race
Abbreviations: Cox PH, Cox proportional hazard; CMV, Cytomegalovirus; D/R, Donor/Recipient; GVHD, graft-versus-host disease; Haplo, Haploidentical; HCT, Hematopoietic Cell Transplantation; HCT-CI, Hematopoietic Cell Transplantation Comorbidity Index; IPTW, Inverse Probability of Treatment Weighting; KPS, Karnofsky Performance Status; MSD, Matched Sibling Donor; NRM, Non-Relapse Mortality; OS, Overall Survival; DFS, Disease-free Survival; PTCy, Post-transplant Cyclophosphamide; NMA, Non-Myeloablative Conditioning; RIC, Reduced-Intensity Conditioning.
Fig 1.
Kaplan-Meier OS by Donor CMV serostatus
Kaplan Meier estimate of DFS by Donor CMV Serostatus
The findings from RSF model also generally aligned with these results, reinforcing the importance of donor CMV serostatus as a key predictor of OS in CMV-seronegative recipients. The partial dependence plot (Figure 2a) shows the relationship between donor age, donor CMV status, and average survival probability at 24 months post-HCT. The best survival probability was noted in the donor age group of about 20-30 years and was lower with donors outside of this range. Although survival probability declined with older donors, the effect of donor age appeared to be rather marginal up to approximately 60 years. On the other hand, CMV-seronegative donors exhibited higher average survival probabilities across all donor age groups, although the difference became narrower in donors above the age of approximately 60 years.
Fig 2.
Partial dependence plot illustrating the relationship between donor age and the average predicted 24-month overall survival probability, by donor CMV serostatus
Partial dependence plot illustrating the relationship between donor age and the average predicted 24-month PFS probability, by donor CMV serostatus
Disease-free survival
The Cox PH model indicated a small yet statistically significant effect of donor age (HR = 1.009, p = .03), whereas IPTW and matched-pair analyses showed no significant effect of donor age (HR = 0.99, p = .30 and HR = 1.009, p = .15, respectively). Conversely, donor CMV serostatus consistently exhibited a significant impact across all models, with CMV seropositive donors associated with higher hazards of disease progression or death (HR ranging from 1.24 to 1.25, p = .02-.03). No statistically significant interactions were observed between donor age and CMV serostatus (p=.83), donor age and recipient age (p=.42), or between these variables and other covariates in the model. Kaplan-Meier DFS curves stratified by donor CMV serostatus shown in Figure 1b.
The partial dependence plot (Figure 2b) shows the relationship between donor age, donor CMV status, and average DFS probability at 24 months post-HCT, providing additional context. It indicated that CMV-seronegative donors consistently had higher DFS probabilities across all donor age groups when compared to CMV-seropositive donors, with narrower differences in older donors. A general trend was noted, with younger donor age being associated with higher DFS probabilities for both CMV-negative and CMV-positive donors compared to older donors. However, the effect of donor age appeared marginal in middle-aged donors (mid-20s to late 50s), with more pronounced changes occurring outside this range.
These machine learning results highlight the complexity of interpreting donor age effects, suggesting potential non-linear relationships or interactions not fully captured by traditional statistical models. The pronounced drop in DFS probability around the mid-20s and its stabilization at lower levels after around 60 years indicate that while donor age may show trends in certain analytical contexts, the significant impact of donor CMV serostatus was a consistent finding across all methods.
Secondary outcomes
The survival differences between the groups were not explained by relapse, but by NRM. Neither donor age nor donor CMV serostatus was associated with relapse risk in Cox PH, IPTW or matched-pair analyses [Table 2]. In contrast, all models consistently showed a significantly higher risk of NRM with CMV seropositive donors (HR ranging from 1.47 to 1.58, p = .009-.02), but no impact of donor age. The reason for a higher NRM with CMV seropositive donors might be related to a higher risk of infections, as the risk of acute or chronic GVHD did not differ significantly between the groups in any of the models. The cumulative incidence plots for relapse, NRM, acute and chronic GVHD are shown in Figure S4–S7. We were unable to precisely assess the cumulative incidences of CMV DNAemia in this combined dataset, as these data were only available in the dataset by Goldsmith et al.10 In that cohort, it was 13.1% (95% CI 8.9-17.3) in the CMV-seropositive donor group versus 1.6% (95% CI .5-2.6) in the CMV-seronegative donor group (Figure S8). Unavailable in the Fuchs et al. dataset, the exact incidence in our entire population could not be determined, but is expected to follow a similar trend. Also, infection was the most common cause of NRM in that dataset; however, the specific infections leading to death were unclear.
DISCUSSION
Our study shows that among CMV-seronegative recipients undergoing related-donor HCT, donor CMV serostatus had a significant impact on outcomes, which was consistently noted across different analytic approaches, while the effect of donor age was marginal and variable. A CMV-seropositive donor was associated with a significantly worse OS, primarily driven by NRM, even after adjusting for donor age and other relevant clinical factors. Although donor age showed a statistically significant association with OS and DFS in the Cox PH models, this effect was small (1% increased hazard per year) compared to the substantial impact of donor CMV serostatus (34% increased hazard). Other statistical methods did not consistently demonstrate a significant association between donor age and survival. This suggests that while donor age may play a role, the influence of donor CMV serostatus might be more impactful. Thus, prioritizing a CMV-seronegative donor over focusing primarily on younger age when selecting a donor for CMV-seronegative patients could be considered. Also, the donor age difference required to offset the detrimental effect of a CMV-seropositive donor would likely need to be substantial, making donor age a less practical consideration in donor selection compared to CMV serostatus in this population.
We included an apparent ‘diverse’ cohort of CMV-seronegative recipients receiving transplants from HLA-MSD (with CNI-based or PTCy-based GVHD prophylaxis) and haploidentical donors with PTCy prophylaxis. While some studies have linked PTCy use to a higher risk of CMV reactivation,11 this effect is primarily observed in CMV-seropositive, but not seronegative, recipients.10 A previous CIBMTR study demonstrated no significant difference in CMV reactivation risk based on donor type or GVHD prophylaxis among CMV-seronegative recipients.10 Based on these observations, we combined these seemingly heterogeneous groups, focusing specifically on the impact of donor CMV serostatus in a broader population of CMV-seronegative recipients. Our analysis focused only on recipients of related-donor transplants due to data availability, as comparable datasets for HLA-matched unrelated donors were not accessible. Further research incorporating unrelated donor data would be valuable to assess the generalizability of our findings to a broader transplant population.
Our study is limited by a retrospective analysis of registry data. Donor selection biases are difficult to control, as the reasons for choosing haploidentical donors over other options or the rationale for using PTCy or CNI-based GVHD prophylaxis are not known. The lack of CMV surveillance and management data, including preemptive treatment approaches, hinders definitive conclusions about CMV's impact on outcomes. Based on the trajectory of CMV reactivation, which occurred early post HCT, it is less likely to have accounted for increased NRM risk. Whether this was attributable to other infections or secondary to the toxicities of CMV treatment, as noted in some prior studies is unclear.12 The generalizability of these findings to other hematological malignancies and unrelated donor HCT remains uncertain. Although letermovir is approved only for CMV-seropositive recipients, the possibility that some CMV-seronegative recipients, particularly those with CMV-seropositive donors, may have received letermovir cannot be excluded. If this occurred, the observed negative impact of CMV serostatus on outcomes might be underestimated in our study. The possibility of false-negative or false-positive CMV serology cannot be ruled out. Finally, the potential influence of specific HLA factors on outcomes in this haploidentical donor HCT population remains unexplored.
Beyond the direct adverse effects of CMV infection or treatment, CMV-seropositive donor cells may exert significant immunomodulatory effects on the recipient. Even without overt CMV reactivation, CMV-seropositive grafts may induce an increase in CD57+/CD27− CD4+ T cells, potentially reducing T cell diversity and antigen-presenting cells expressing MHC class II13, and thus impairing immune reconstitution. This impairment may result from weakened T-cell signaling, senescence, impaired proliferation, and reduced IL-2 production.14–16 These findings elucidate a biological rationale for the potential benefit of CMV-seronegative donors in CMV-seronegative recipients.
Our study has several strengths, including consistent results across diverse conventional as well as machine-learning analytical strategies (Cox PH, IPTW, propensity score matching, and RFS models). This convergence reinforces the robustness of our findings and reduces the likelihood that the observed association between donor CMV serostatus and outcomes is due solely to methodological biases or chance. Further research is needed to confirm these findings in larger studies and to investigate the specific mechanisms driving the increased NRM observed in recipients of CMV-seropositive donors. Understanding these mechanisms is crucial for developing targeted interventions to mitigate risks in this population. For example, if CMV reactivation is the primary driver, the role of prophylactic letermovir in CMV-seronegative recipients with CMV-seropositive donors warrants investigation, while carefully considering the risks and benefits of prophylaxis.
CONCLUSION
In CMV-seronegative related-donor HCT recipients, donor CMV serostatus appears to be a more influential determinant of outcomes than donor age. Our findings suggest that prioritizing a CMV-seronegative donor may be more beneficial than selecting a younger, CMV-seropositive donor, except where age differences are substantial. These results warrant validation in larger datasets and may advocate for a reassessment of current prioritization strategies that universally emphasize donor age above other factors.
Supplementary Material
Key points:
Donor CMV serostatus may have a more dominant impact on survival than donor age in CMV-seronegative HCT recipients.
CMV-seropositive donors vs negative donors have higher non-relapse mortality and no difference in relapse in CMV-seronegative recipients.
ACKNOWLEDGMENT:
This dataset was provided by the Center for International Blood and Marrow Transplant Research (CIBMTR) which is supported primarily by the Public Health Service U24CA076518 from the National Cancer Institute; the National Heart, Lung, and Blood Institute; the National Institute of Allergy and Infectious Diseases; 75R60222C00011 from the Health Resources and Services Administration; N00014-21-1-2954 and N00014-23-1-2057 from the Office of Naval Research; the National Marrow Donor Program; and the Medical College of Wisconsin. We acknowledge that European Union patient-level data is redacted. The authors are grateful to the CIBMTR staff for providing the datasets publicly.
Funding source:
These datasets were collected by the Center for International Blood and Marrow Transplant Research (CIBMTR) which is supported primarily by the Public Health Service U24CA076518 from the National Cancer Institute; the National Heart, Lung, and Blood Institute; the National Institute of Allergy and Infectious Diseases; 75R60222C00011 from the Health Resources and Services Administration; N00014-21-1-2954 and N00014-23-1-2057 from the Office of Naval Research; the National Marrow Donor Program; and the Medical College of Wisconsin.
Footnotes
Disclosure of Conflicts of Interest: No relevant COI to declare.
Data Sharing and Availability:
The data are publicly available and accessible at https://cibmtr.org/CIBMTR/Resources/Publicly-Available-Datasets.
REFERENCES
- 1.Chorao P, Henriques M, Villalba M, Montoro J, Balaguer-Rosello A, Gonzalez EM, et al. Cytomegalovirus Reactivations in Allogeneic Hematopoietic Stem Cell Transplantation from HLA-Matched and Haploidentical Donors with Post-Transplantation Cyclophosphamide. Transplant Cell Ther. May 2024;30(5):538 e1–538 e10. doi: 10.1016/j.jtct.2024.01.082 [DOI] [PubMed] [Google Scholar]
- 2.Mariotti J, Legrand F, Furst S, Giordano L, Magri F, Richiardi L, et al. Risk Factors for Early Cytomegalovirus Reactivation and Impact of Early Cytomegalovirus Reactivation on Clinical Outcomes after T Cell-Replete Haploidentical Transplantation with Post-Transplantation Cyclophosphamide. Transplant Cell Ther. Mar 2022;28(3):169 e1–169 e9. doi: 10.1016/j.jtct.2021.12.014 [DOI] [PubMed] [Google Scholar]
- 3.Marty FM, Ljungman P, Chemaly RF, Maertens J, Dadwal SS, Duarte RF, et al. Letermovir Prophylaxis for Cytomegalovirus in Hematopoietic-Cell Transplantation. N Engl J Med. Dec 21 2017;377(25):2433–2444. doi: 10.1056/NEJMoa1706640 [DOI] [PubMed] [Google Scholar]
- 4.Teira P, Battiwalla M, Ramanathan M, Barrett AJ, Ahn KW, Chen M, et al. Early cytomegalovirus reactivation remains associated with increased transplant-related mortality in the current era: a CIBMTR analysis. Blood. May 19 2016;127(20):2427–38. doi: 10.1182/blood-2015-11-679639 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Logan BR, Maiers MJ, Sparapani RA, Laud PW, Spellman SR, McCulloch RE, et al. Optimal Donor Selection for Hematopoietic Cell Transplantation Using Bayesian Machine Learning. JCO Clin Cancer Inform. May 2021;5:494–507. doi: 10.1200/CCI.20.00185 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Shaw BE, Logan BR, Spellman SR, Marsh SGE, Robinson J, Pidala J, et al. Development of an Unrelated Donor Selection Score Predictive of Survival after HCT: Donor Age Matters Most. Biol Blood Marrow Transplant. May 2018;24(5):1049–1056. doi: 10.1016/j.bbmt.2018.02.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Kollman C, Spellman SR, Zhang MJ, Hassebroek A, Anasetti C, Antin JH, et al. The effect of donor characteristics on survival after unrelated donor transplantation for hematologic malignancy. Blood. Jan 14 2016;127(2):260–7. doi: 10.1182/blood-2015-08-663823 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.CIBMTR. https://cibmtr.org/CIBMTR/Resources/Publicly-Available-Datasets#, (accessed 9/6/2023).
- 9.Fuchs EJ, McCurdy SR, Solomon SR, Wang T, Herr MM, Modi D, et al. HLA informs risk predictions after haploidentical stem cell transplantation with posttransplantation cyclophosphamide. Blood. Mar 10 2022;139(10):1452–1468. doi: 10.1182/blood.2021013443 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Goldsmith SR, Abid MB, Auletta JJ, Bashey A, Beitinjaneh A, Castillo P, et al. Posttransplant cyclophosphamide is associated with increased cytomegalovirus infection: a CIBMTR analysis. Blood. Jun 10 2021;137(23):3291–3305. doi: 10.1182/blood.2020009362 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Ueda Oshima M, Xie H, Zamora D, Flowers ME, Hill GR, Mielcarek MB, et al. Impact of GVHD prophylaxis on CMV reactivation and disease after HLA-matched peripheral blood stem cell transplantation. Blood Adv. Apr 25 2023;7(8):1394–1403. doi: 10.1182/bloodadvances.2022009112 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Nichols WG, Corey L, Gooley T, Davis C, Boeckh M. High risk of death due to bacterial and fungal infection among cytomegalovirus (CMV)-seronegative recipients of stem cell transplants from seropositive donors: evidence for indirect effects of primary CMV infection. J Infect Dis. Feb 1 2002;185(3):273–82. doi: 10.1086/338624 [DOI] [PubMed] [Google Scholar]
- 13.Yeh AC, Varelias A, Reddy A, Barone SM, Olver SD, Chilson K, et al. CMV exposure drives long-term CD57+ CD4 memory T-cell inflation following allogeneic stem cell transplant. Blood. Dec 30 2021;138(26):2874–2885. doi: 10.1182/blood.2020009492 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Brenchley JM, Karandikar NJ, Betts MR, Ambrozak DR, Hill BJ, Crotty LE, et al. Expression of CD57 defines replicative senescence and antigen-induced apoptotic death of CD8+ T cells. Blood. Apr 1 2003;101(7):2711–20. doi: 10.1182/blood-2002-07-2103 [DOI] [PubMed] [Google Scholar]
- 15.Kim TK, St John LS, Wieder ED, Khalili J, Ma Q, Komanduri KV. Human late memory CD8+ T cells have a distinct cytokine signature characterized by CC chemokine production without IL-2 production. J Immunol. Nov 15 2009;183(10):6167–74. doi: 10.4049/jimmunol.0902068 [DOI] [PubMed] [Google Scholar]
- 16.Shindo T, Kim TK, Benjamin CL, Wieder ED, Levy RB, Komanduri KV. MEK inhibitors selectively suppress alloreactivity and graft-versus-host disease in a memory stage-dependent manner. Blood. Jun 6 2013;121(23):4617–26. doi: 10.1182/blood-2012-12-476218 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data are publicly available and accessible at https://cibmtr.org/CIBMTR/Resources/Publicly-Available-Datasets.