We systematically searched PubMed, Embase, Web of Science, and the Cochrane library to select relevant articles up to April 17, 2022. We applied no language restrictions. We used the following combined text and MeSH terms: “ATG,” “anti-thymocyte globulin,” “GVHD,” “graft-versus-host disease,” “PTCy,” “posttransplant cyclophosphamide,” and “allogeneic transplantation.” The complete search used for PubMed is provided in Supplementary Data 1 . We considered all potentially eligible studies for analysis, irrespective of the primary outcome or language.

After excluding duplications, two independent investigators reviewed the study titles and abstracts. All studies that satisfied the inclusion criteria were retrieved for full-text assessment. Two investigators analyzed trials selected for detailed analysis and data extraction with an agreement value (κ) of 96.5%; a third investigator resolved disagreements.

The criteria for eligible articles were as follows: (1) original human study; (2) patients diagnosed with hematological malignancies receiving allo-HCT; (3) receive MUD or MMUD; (4) comparison between PTCy and ATG; and (5) at least one of the following outcomes: incidence of grade II-IV and III-IV aGVHD, cGVHD, OS, RI, NRM, CMV reactivation, BKV reactivation, and EBV reactivation.

Criteria for exclusion: (1) duplicate studies, conference reports, letters, or other articles without available full text; (2) cell or animal studies; (3) receive other donor types: haploidentical donor or matched sibling donors; and (4) studies using PTCy and ATG in the same group.

Data extracted from all eligible articles included the first author, year of publication, journal, sample size, number of patients in the PTCy and ATG groups, patient age, patient sex, graft source, disease type, donor type, conditioning regimen, and follow-up time, and all these characteristics are provided in Supplementary Table 1 .

In addition, primary clinical outcomes (grade II-IV and III-IV aGVHD, cGVHD, overall survival (OS), relapse incidence (RI), non-relapse mortality (NRM)), and infections (cytomegalovirus (CMV) reactivation and CMV disease, Epstein−Barr virus (EBV)-related posttransplant lymphoproliferative disease (PTLD), BKV-related hemorrhagic cystitis (HC)) were independently extracted from each article for aggregation and analysis. Two authors worked separately on a quality assessment by the Newcastle−Ottawa Scale (NOS) according to the NOS system’s three components (selection, comparability, and outcome). The range of scores is from 0 to 9, and a study with a score of 5 or higher is considered high quality ( Supplementary Table 2 ).

This meta-analysis was performed with Review Manager (RevMan) software version 5.4. A two-sided P value ≤ 0.05 was considered statistically significant. All outcomes were counted as dichotomous variables, and events and total numbers were aggregated to compare the efficacy between the PTCy and ATG groups. This method was also adopted in subgroup analysis. We graphically displayed the results by forest plot. Heterogeneity was checked by the chi-square-based Q test and I ² statistic (I ² > 50%, P ≤ 0.1 indicated high heterogeneity). If heterogeneity was considered not obvious, a fixed-effect model was employed for calculation; otherwise, a random-effect model was adopted.

We identified 1091 articles, 206 from PubMed, 247 from Embase, 595 from Web of Science, and 43 from the Cochrane Library. After eliminating 189 duplications electronically, 902 potentially relevant publications were screened. Of these records, 855 citations were removed by reviewing the titles and abstracts. Full text or further details were retrieved for the remaining 47 citations. Finally, six eligible studies were included in the meta-analysis (12, 15–19). The flow diagram of the search and selection process is shown in Figure 1 .

The six studies were published between 2016 and 2021 and were of relatively high quality (NOS scores higher than 5). All were retrospective studies, with a total of 2379 patients, of which 472 received PTCy and 1907 were treated with ATG.

The average age of patients in the six included studies ranged from 31 to 62. The median follow-up time varied substantially between groups, namely, 12 to 26 months in the PTCy group and 17 to 78 months in the ATG group. Acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) are the most common diseases. Among these six studies, there were three studies whose number of male subjects outweighed the number of female ones, only one study has more female patients, while another two did not mention the female-to-male ratio. The PTCy group had 472 unrelated donors, including 297 (62.9%) matched unrelated donors (MUDs) and 175 (37.1%) mismatched unrelated donors (MMUDs); while the ATG group included 1607 (84.3%) MUDs and 300 (15.7%) MMUDs. Of all the patients, 1093 (45.9%) subjects received myeloablative conditioning (MAC), including 201 (18.4%) in the PTCy group and 892 (81.6%) in the ATG group. The remaining 1286 (54.1%) patients were treated with nonmyeloablative or reduced-intensity conditioning (RIC), with 271 (21.1%) in the PTCy group and 1015 (78.9%) in the ATG group. For the graft source, 43 (9.1%) in the PTCy group were transplanted with bone marrow (BM) and the rest 429 (90.9%) received peripheral blood (PB) stem cells. In the ATG group, the number of patients who chose BM or PB as the graft source was 165 (8.7%) and 1742 (91.3%), respectively.

Six studies with 2379 patients all reported the incidence of grade II-IV aGVHD. We calculated each study’s events and total number as a dichotomous variable. Pooling the data of these studies showed that the PTCy has an advantage over ATG in grade II-IV aGVHD prevention. (RR=0.68, 95% CI 0.50-0.93, P=0.010, I ^{2 =} 67%; Figure 2 ). Considering the apparent heterogeneity, a sensitivity analysis and subgroup analysis were conducted. The studies were divided into two groups based on the sample size. Two studies were included into big sample group while the other four were in a small sample group. Patients in small sample size reported a lower incidence of grade II-IV aGVHD (RR=0.53, 95% CI 0.38-0.74, P=0.23, I ^{2 =} 31%; Figure 2 ) than large sample studies (RR=0.97, 95% CI 0.79-1.20, P=0.86, I ^{2 =} 0%; Figure 2 ), implying that the sample size of investigations might be the source of heterogeneity.

Regarding grade III-IV aGVHD, all the included studies reported this outcome. The results demonstrated that prophylaxis based on PTCy was associated with a lower risk of developing grade III-IV aGVHD (RR=0.32, 95% CI 0.14-0.76, P=0.001, I ^{2 =} 75%; Figure 3 ). There was still heterogeneity, which could not be neglected. However, a sensitivity analysis showed that any separate study did not influence the heterogeneity. Therefore, subgroup analysis according to age was carried out. We took age 50 as the categorizing criterion. If the median age of both PTCy and ATG groups were lower than 50, the study would be categorized as younger subgroup; if not, the article would be placed in elder subgroup. The pooled results showed that the PTCy group tended to have a lower risk of III-IV aGVHD in the younger subgroup (RR=0.11, 95% CI 0.04-0.26, P=0.64, I ^{2 =} 0%; Figure 3 ) than elder subgroup (RR=0.64, 95% CI 0.37-1.12, P=0.22, I ^{2 =} 33%; Figure 3 ).

Six studies were available for the analysis for moderate to severe cGVHD. The aggregated results showed no difference between PTCy and ATG regimens in cGVHD (RR=0.66, 95% CI 0.35-1.26, P<0.00001, I ^{2 =} 86%; Figure 4 ). A sensitivity analysis revealed that the obvious heterogeneity did not come from any included article. And subgroup analysis according to median age or the published year could not lower the heterogeneity.

As for OS, according to the pooled RR, the use of PTCy was associated with a higher overall survival (RR=1.29, 95% CI 1.03-1.62, P=0.0001, I ^{2 =} 80%; Figure 5 ). Considering the obvious heterogeneity, a sensitivity analysis was applied to find the source of it. But ruling out any separate study did not significantly influence the result. However, subgroup analysis based on published year found that studies published before 2020 reported higher overall survival in PTCy group (RR=1.59, 95% CI 1.32-1.91, P=0.41, I ^{2 =} 0%; Figure 5 ).

Among the six included studies, five of them reported the outcome of RI, while the rest one did not choose the relapse rate as primary outcome. There was no obvious heterogeneity among the included studies, therefore a fixed-effect model was used. The pooling data of RI showed no difference between the PTCy and ATG groups (RR=0.95, 95% CI 0.78-1.16, P=0.37, I ^{2 =} 7%; Figure 6 ).

For NRM, we gathered data from 2379 patients in all studies and adopted a fixed-effect model, which showed that, compared with ATG group, patients treated with PTCy is associated with lower NRM (RR=0.67, 95% CI 0.53-0.84, P=0.17, I ^{2 =} 36%; Figure 6 ). And the result was statistically significant as well.

The related data on infectious complications can be found in four articles. Four studies reported incidences of CMV reactivation and CMV disease. However, only three of them reported EBV-related PTLD and BKV-related HC. According to the aggregated RR value, we found that there was no significant difference in CMV reactivation and CMV disease between the PTCy group and ATG group (RR=0.89, 95% CI 0.63-1.24, P=0.07, I ^{2 =} 57%; Figure 7 ). Although apparent heterogeneity was observed, we did not conduct any sensitivity analysis or subgroup analysis due to the limited study data.

For BKV-related HC, we adopted the fixed-effect model because I ²<50%. The aggregated RR indicated that PTCy-based protocol and ATG-based prophylaxis were comparable in developing BKV-related HC. The heterogeneity was not significant (RR=0.88, 95% CI 0.76-1.03, P=0.44, I ^{2 =} 0%; Figure 7 ).

EBV-related PTLD was only reported in three articles, and further analysis showed that the heterogeneity was not apparent; therefore, a fixed-effect model was used to assess the outcome. The results indicated that adopting the PTCy regimen was associated with a lower chance of developing EBV-related PTLD disease (RR=0.23, 95% CI 0.09-0.58, P=0.85, I ^{2 =} 0%; Figure 7 ), and the result was statistically significant.