In vivo T-Cell Depletion with Antithymocyte Globulins Improves Overall Survival after Myeloablative Allogeneic Stem Cell Transplantation in Patients with Hematologic Disorders

Hematopoietic stem cell transplantation (HSCT) is increasingly used worldwide as curative therapy for malignant and nonmalignant hematologic disorders. Graft-versus-host disease (GvHD) still remains the leading cause of mortality and morbidity after allogeneic HSCT using the current standard regimen for GvHD prophylaxis, which usually contains various combinations and doses of a calcineurin inhibitor (cyclosporine or tacrolimus), methotrexate and mycophenolate mofetil [1, 2, 3, 4]. Donor T cells have been recognized as main effectors of acute GvHD [5]. Based on these experimental data, T-cell depletion (TCD) has been proposed to lower GvHD incidence and severity. Most early trials documented that ex vivo TCD could substantially limit GvHD. However, this reduction in GvHD did not translate into improved overall survival (OS) because of unexpectedly high rates of graft failure and rejection after TCD bone marrow transplantation [6, 7]. Another major problems with ex vivo TCD is the risk of an increase in treatment-related mortality (TRM) and relapse mortality. In subsequent trials, different anti-T-cell globulin (ATG) regimens have been tested as part of conditioning regimens to achieve in vivo TCD in order to prevent GvHD. In a recent large-scale retrospective cohort of reduced-intensity allogeneic (RIC) HSCT, a total of 1,676 patients received standard GvHD prophylaxis with or without antithymocyte globulin, which suggests that the use of ATG with RIC regimens reduces GvHD but significantly increases the likelihood of disease relapse, which negatively affects disease-free survival and OS [8]. Since results are encouraging but also conflicting at the same time in the myeloablative setting, the benefit of prophylactic treatment with ATG in GvHD remains to be determined.

As a result, whether ATG should be used upfront is uncertain. We did this analysis to assess the impact of ATG therapy on OS, TRM, relapse mortality, progression-free survival (PFS), and acute and chronic GvHD in the myeloablative setting.

Medline, Embase, the Cochrane Controlled Trials Register and the Science Citation Index were searched for randomized, controlled or cohort studies (RCTs) using the medical subject headings ‘thymoglobulin’, ‘antithymocyte globulin’, ‘anti-T-cell globulin’, ‘stem cell transplantation’, ‘bone marrow transplantation’ and ‘myeloablative transplant’. Reference lists from studies selected for this review and from other published systematic reviews and practice guidelines were also manually searched.

Studies were eligible for inclusion in the meta-analysis if they met all the following criteria: (1) They were published up to December 2011 and written in English. (2) They dealt only with hematologic disease, such as acute or chronic leukemia, myelodysplastic syndrome, aplastic anemia, malignant lymphoma and multiple myeloma. (3) They provided sufficient information to allow the calculation of crude hazard ratios (HRs) for OS, TRM, relapse mortality, PFS, GvHD in patients undergoing transplantation treated with ATG compared with those treated with a similar regimen without ATG. (4) All participants were intended for myeloablative transplant and standard GvHD prophylaxis. Multiple reports of a single study were considered as one publication, and only the most recent or complete article was investigated. All potentially relevant articles were reviewed by two independent investigators (J.Y. and Z.S.).

In addition to the number of patients, studies were categorized into randomized control or cohort studies, and cohort studies were further divided into prospective or retrospective studies. Prospective cohort studies were defined as those that identified a group of participants and followed them up through time, whereas retrospective cohort studies were defined as those that identified a group of participants and used existing records to evaluate their clinical characteristics and course.

The primary outcome for this review was OS; secondary outcomes were TRM, relapse mortality, PFS, and acute and chronic GvHD.

Two reviewers (J.X. and Z.S.) independently assessed the quality of selected randomized, controlled studies using the following criteria: (1) generation of allocation sequence; (2) description of dropouts; (3) masking of randomization, intervention and outcome assessment, and (4) intention-to-treat analyses. Each criterion was rated as yes, no or unclear. The Newcastle-Ottawa Scale [9] was used to assess the quality of each cohort study. This measure assesses aspects of methodology in observational studies related to study quality, including case selection, comparability of populations and ascertainment of exposure to risks. In all cases, disagreements among raters were resolved through discussion so that a consensus was reached. Analyses of the relationship between study quality assessed by Newcastle-Ottawa Scale items to study outcomes were completed via meta-regression analyses.

Time-to-event data (e.g. OS, TRM and PFS) were pooled and reported as HR, using 95% confidence intervals (CIs). To assess outcome, we have used two different models: the fixed-effect model if no evidence was found and the random-effect model if significant evidence was detected. For each study included, for the purpose of analysis, we calculated the log-rank HRs and its standard error to perform this meta-analysis. When not available from the trial reports, they were estimated with the methods proposed by Parmar et al. [10] and described elsewhere [11].

Data were subjected to meta-analyses stratified by study type (randomized controlled vs. cohort study) to obtain composite estimates of HRs separately for each study type and for all studies combined. Sensitivity analyses included comparisons of studies of randomized controlled compared with cohort studies. Between-study and between-subgroup variations were calculated using the χ² test. Publication bias was evaluated by a funnel plot that related standard errors of studies to their effect sizes. If the funnel plot shows asymmetry, it suggests that studies that might have reported negative results may not have been published [12].

Furthermore, we used the ‘trim-and-fill’ method proposed by Duval and Tweedie [13]to calculate the estimated outcome rates of all studies included after correction for potential confounders, such as various ATG preparations (ATG-H, ATG-S and ATG-F), timing of ATG (before or after transplantation), donor type (sibling or unrelated donor) and study type (randomized or cohort studies) to assess the effects of those potential covariates or confounders.

We defined a value of p < 0.05 as statistically significant for all estimated outcomes and of p < 0.15 as statistically significant for heterogeneity, publication bias and regression analysis. Due to the small number of studies included, publication bias was not formally assessed.

All meta-analyses were completed using Review Manager (version 5.1; Cochrane Collaboration, Oxford, UK) and Stata software (version 11.0; Stata, College Station, Tex., USA).

A comprehensive search of Medline, Embase, the Cochrane Controlled Trial Register and the Science Citation Index yielded 938 articles, of which 11 studies met the predetermined inclusion criteria, covering a total of 1,549 patients. Four randomized-controlled trials [14, 15, 16, 17, 18, 19], six retrospective cohort studies [20, 21, 22, 23, 24, 25] and one prospective cohort study were included [26].

Details on these studies are listed in table 1. All RCTs reported final analyses. All randomized studies reported intention-to-treat analyses, but dropouts and blind randomization were recorded in only one trial [16, 19]. Quality assessments of these retrospective/prospective cohort studies are shown in Appendix 1 and Appendix 2. We evaluated the relationship between study quality and the observed risk through meta-regression of study characteristic ratings assessed by the Newcastle-Ottawa Scale and log HRs. In the regression model for cohort studies, none of the individual study characteristics was independently significantly related to the HRs of any estimated outcome, and no publication bias was found.

Before combining studies in the meta-analysis, we evaluated the presence of heterogeneity in OS analysis. No evidence for the presence of heterogeneity between studies or between subgroups was found, suggesting the study results included were consistent. The summary HRs for OS were 0.84 (95% CI 0.63–1.12; p = 0.23) for randomized studies, 0.70 (95% CI 0.57–0.88; p = 0.002) for cohort studies and 0.75 (95% CI 0.63–0.89; p = 0.001) for all studies combined, suggesting that ATG might confer a survival benefit in patients receiving myeloablative transplant relative to standard GvHD prophylaxis (fig. 1). In the regression model for cohort studies, none of the individual study characteristics was independently significantly related to HRs.

As shown in figure 2, a similar favorable trend was detected in the ATG arm for TRM analysis. The summary HRs for TRM were 0.81 (95% CI 0.54–1.22; p = 0.32) for randomized studies, 0.70 (95% CI 0.49–0.99; p = 0.05) for cohort studies and 0.74 (95% CI 0.57–0.95; p = 0.02) for all studies combined, suggesting that ATG might confer a TRM benefit in patients receiving myeloablative transplant relative to standard GvHD prophylaxis. Evidence for the presence of heterogeneity between studies and subgroups was also noted.

As shown in figure 3, the summary HRs for relapse mortality were 1.18 (95% CI 0.69–2.02; p = 0.55) for randomized studies, 1.02 (95% CI 0.65–1.61; p = 0.93) for cohort studies and 1.05 (95% CI 0.74–1.50; p = 0.79) for all studies combined, suggesting that ATG did not have a negative effect on relapse mortality, and evidence for the presence of heterogeneity between cohort studies was found.

Additionally, our pooled analysis has shown that adding ATG to standard GvHD prophylaxis leads to a lower incidence and severity of acute and chronic GvHD, without having a negative effect on PFS (table 2).

To evaluate associations of potential covariates and confounding variables, such as various ATG preparations (ATG-H, ATG-S and ATG-F), timing of ATG, donor type and study type to study effects, meta-regression was conducted to estimate the size of pooled study effects after correction for the variables of interest. Even after correction for those variables, neither the survival data nor other estimated outcomes, such as TRM, relapse mortality, PFS and GvHD, substantially changed (table 3).

Despite significant progress in the past 20 years, GvHD remains a significant cause of morbidity and mortality following allogeneic HSCT. Most early trials documented that ex vivo TCD could substantially limit acute and chronic GvHD. However, this reduction in GvHD did not translate into improved OS due to unexpectedly high rates of graft failure and disease recurrence after ex vivo TCD for bone marrow transplantation [27]. In subsequent trials, the efficacy of ATG as in vivo TCD regimen was tested to prevent GvHD. In the RIC setting, Soiffer et al. [8] compared in vivo TCD with an ATG preparation in 1,676 patients with hematologic malignancies in a large retrospective cohort trial; they strongly suggested that in RIC recipients ATG lowers the risk of chronic GvHD at the expense of increased relapse rates, resulting in inferior disease-free survival and OS. Because the cytoreductive effects of RIC are usually insufficient to eradicate malignancy, it is possible that the cytotoxic effects of myeloablative conditioning blunted the impact of in vivo TCD on relapse. In the standard myeloablative transplantation setting, several randomized trials have documented that ATG could diminish the incidence rate and severity of GvHD [14, 15, 16, 19]. However, as discussed elsewhere, the most stringent proof of the benefit of ATG regarding improved survival compared with standard GvHD prophylaxis is still missing. Our meta-analysis helps to clarify the impact of ATG on the survival of patients with standard myeloablative transplantation. Of the eleven trials included, OS data were available for nine [14, 15, 16, 17, 18, 19, 20, 21, 22, 24, 25]. Pooling these survival data enabled us to increase the power of the survival analysis and confirmed a significant and consistent relative survival benefit with the addition of ATG to myeloablative HSCT for patients with hematologic malignancies compared to standard GvHD prophylaxis alone. A significant survival benefit was also detected in the subgroup of patients pooled from five retrospective cohorts, but not in the subgroup of patients pooled from four randomized trials. Of note, a major shortcoming of those retrospective studies must be addressed: a retrospective cohort may be biased by a greater recall of history of patients treated with ATG, who often undergo more careful evaluation than do control subjects. However, in our opinion, the survival benefit of cohort studies in the ATG arms was reliable, because little heterogeneity was found between cohort and randomized studies suggesting that the two study types have consistent findings. The randomized studies seem not powered enough to detect differences in survival among the two groups (fig. 1, the weight of cohort vs. randomized studies: 63.5 vs. 36.5%). Additionally, other impressive result also emerged from our analysis: adding ATG to standard GvHD prophylaxis leads to a lower incidence of acute and chronic GvHD and improved TRM, without having a negative effect on relapse mortality. The protective effect of ATG on chronic GvHD included the limited form, but more so the extensive form.

To evaluate the influence of potential covariates and confounding variables (such as various ATG preparations, timing of ATG, and donor and study types) to estimated outcomes, we conducted meta-regression to estimate the size of pooled study effects after correction for these variables. Even after correction for those variables, the survival benefit continued to be significant, and other estimated outcomes (such as TRM, relapse mortality, PFS and GvHD) did not substantially change (table 3).

However, several other caveats need to be considered when interpreting our findings: The first and major problem is that we had no access to primary data and only used abstracted data, while an individual patient data-based meta-analysis would have provided a more robust estimate of the efficacy of the addition of ATG in the myeloablative transplant setting. Secondly, the sample size of randomized trials included for survival analysis was too small to detect a more reliable estimate. Thirdly, as is often the case with meta-analysis, the effect of heterogeneity needs to be considered, and the different doses of ATG preparations might have a great impact on effects and side effects and thus might negatively affect the comparability of the data.

Despite the caveats of our analysis, we believe that it still makes an important contribution to the allogeneic HSCT field, as it suggest that the addition of ATG to standard GvHD prophylaxis leads to a lower incidence of acute and chronic GvHD and lower TRM without having a negative effect on relapse mortality, resulting in a statistically significantly longer OS.

Appendix 1. Check list for quality assessment and scoring of nonrandomized studies

Appendix 2. Assessment of quality of nonrandomized studies (see check list in Appendix 1)

The authors declare no competing financial interests, and they did not receive any financial support.

Z.S., H.M., W.P., and S.N. contributed equally to this work.