|Home | About | Journals | Submit | Contact Us | Français|
We read with interest the letter from Travi and colleagues  reporting results from their own institution that differed from those we recently published in the Journal . The letter inquired whether differing distributions of demographics or conditioning regimens might explain why the similar, contemporary samples yielded contrasting findings on whether donor/recipient CMV serostatus affects the incidence of leukemic relapse after pediatric HCT. Here we reply to the question and offer a potential explanation of our own.
After stating the traditional view, established before the current era of CMV preemptive therapy, that “seronegative recipients with a seronegative donor (D−/R−) have shown an improved outcome” after HCT, Travi et al. reported that, among pediatric recipients at their institution, D−/R− grafts were not associated with improved outcome, either in relapse or non-relapse mortality (NRM). In contrast, our study of a similar pediatric sample  yielded the novel finding that D−/R− grafts were significantly associated with poorer outcome (higher incidence of relapse, inferior relapse-free survival) than other grafts. Because relapse was rarely detected by Travi et al. after 2 years post-transplant, we compared 2-year incidence by institution and serostatus, using Fig 1B from Travi et al.  and Fig 2 from our paper . At 2 years post-HCT, the two samples had similar cumulative incidence of NRM (which did not differ by D−/R− serostatus) and similar cumulative incidence of relapse in the D−/R− subgroup, but at our institution only, recipients of seropositive grafts (D+ and/or R+) had significantly reduced incidence of relapse. We interpret this finding as evidence that certain conditions at our institution promoted the graft-versus-leukemia (GVL) effect of HCT in our pediatric sample, specifically in those recipients who received CMV-seropositive grafts.
What conditions might have differed between our 2 institutions and accounted for the contrasting findings? Travi et al.  suggested that the explanation might be found in the distribution of ethnicity (Caucasian vs other), total body irradiation (TBI)-based conditioning, or age in our 2 samples. To address this suggestion, we examined whether the samples differed on demographics or TBI use. We then investigated whether the association between D−/R− serostatus and relapse in our sample became nonsignificant after adjustment for age, TBI, or ethnicity or was significantly modified by interaction with these characteristics.
Our sample had a higher proportion of Caucasians (86.4% vs 74.3%, p<0.01), but TBI use was similar (85.0% vs 81.9%) in the 2 samples. Whether they differed in age could not be determined, because Travi et al. did not report an age distribution. In our published model of relapse , D−/R− serostatus remained significant after adjustment for age and other covariates; the effect of serostatus did not vary with age. Further analysis determined that TBI use was not associated with relapse and did not modify the effect of serostatus. Non-Caucasian ethnicity was not associated with relapse but appeared to enhance the effect of serostatus: the cumulative incidence-based hazards ratio of relapse with D−/R− graft was 2.81 (95% CI 1.25-6.28) in Caucasians and 12.3 (2.94-51.4) in non-Caucasians, compared to recipients of seropositive grafts of any ethnicity. We regard the interaction between serostatus and ethnicity with caution, however, because our sample had few non-Caucasian patients (19/140), a biological rationale for the interaction is lacking, and the model's fit was not appreciably improved after stratifying seronegative graft by ethnicity. In summary, differences in demographics or TBI use are unlikely to explain why an association between D−/R− graft and relapse was present in our sample but not in that of Travi et al.
We believe that a more compelling explanation for why the two institutions obtained different results may be found in their different methods of CMV surveillance. At our institution, surveillance at the time was primarily by viral culture, a less sensitive method than pp65 antigenemia or PCR, used historically at the other institution. Moreover, the sensitivity of viral culture in our patients may have been further reduced if the volume of blood cultured was less than that typically used for adults. Less sensitive surveillance may have delayed or reduced detection of CMV activation, which in turn may have delayed or reduced initiation of preemptive therapy. As a result, exposure to subclinical CMV viremia may have been prolonged, and ganciclovir may have been initiated later or less often, among our recipients of seropositive grafts than among similar patients at the other institution. Recipients of seronegative grafts, in contrast, would have been little affected by differences in sensitivity of CMV detection, because viremia is far less common in that group. It is possible that prolonged early exposure to subclinical CMV had a stimulatory effect on NK and T cells [4-7]. It is also possible that delayed or reduced initiation of ganciclovir, a drug that can suppress T cell proliferation  and delay T cell recovery , benefitted repopulating T cells. Either possibility could have enhanced the GVL effect in our recipients of seropositive grafts.
We agree with Travi et al. that a definitive investigation of the effect of graft serostatus on the outcome of pediatric HCT will require a large, multicenter study. In fact, we formally proposed such a study to the CIBMTR in November 2008. It must be recognized, however, that results from a multicenter study will be conclusive only if the analysis controls for differences between centers, and also within centers over time, in the methods of CMV surveillance and CMV management (preemptive therapy, prophylaxis, or both). In addition, the choice of eligibility criteria will be important. For conclusive results, not only should all subjects be pediatric recipients of primary, myeloablative, allogeneic HCT for acute leukemia, but the primary analysis should exclude T-cell depleted or cord blood grafts, because in those grafts, the effect of serostatus on incidence of relapse might be abolished.
This work was supported by NIH grant CA30206 and U.S. Public Health Service grant AI58148.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.