Included studies
We gained 880 citations from the electronic database and manual screen and 27 potentially related citations were retrieved as full-text or were checked for more detailed investigation (Fig. 1). Five reviews and three abstracts were excluded, two studies were excluded for insufficient patient number, and two were excluded for potentially repeated reports. Ultimately, 15 studies with 5336 patients met the predefined selection criteria (Table 1).
Characteristics of the included studies
All studies reported the outcomes of AA patients treated with first-line allo-HSCT or IST [11-25]. The allo-HSCT group was divided into two subgroups: BMT-MRD and BMT-MUD. In one study, the BMT-matched unrelated donor (MUD) subgroup was deleted for having an insufficient patient number (n = 5) [12]. The non-transplant group consisted of two subgroups, cyclosporin alone and androgen. In one study, the androgen subgroup was deleted for receiving non-IST treatment [15]. Those studies were published between 2000 and 2019, and all were retrospective studies. The case collection period ranged from 1976 to 2016. Sample sizes ranged from 31 to 2479 (15 to 1567 in the allo-HSCT group and 16 to 912 in the IST group). The study population was children in 7 studies, adults in 3 studies, and both children and adults in 5 studies. Most studies included only SAA patients, but 3 studies included SAA, VSAA, and NSAA patients. The median age was 8–28 years in the allo-HSCT group and 6–55 years in the IST group. Only one study reported mean age, and one study only reported the median age of the allo-HSCT group. One study reported neither the median age nor the sex ratio. For the allo-HSCT group, the donor was an MRD in 9 studies, an HID in 3 studies, and an MUD in 1 study. One study used mainly MRD donors together with mismatched related donor (MMRD) or MUD. Most of the studies adopted a cyclophosphamide-based regimen as a conditioning program. Prophylaxis against GVHD mainly consisted of CsA and methotrexate (MTX). Stem cell sources consisted of bone marrow (BM), peripheral blood stem cells (PBSCs), and a few patients with CB. The IST drugs mainly consisted of ATG or ALG combined with CsA. The quality of the studies in the analyses was high, with a mean overall NOS assessment score of 7.2 (range, 7–8).
Primary outcomes
All studies reported OS. The pooled HR for OS was 0.4 (95% CI 0.074–0.733, P = 0.016, I2 = 58.8%) (Fig. 2). These data indicate that first-line allo-HSCT is significantly superior to IST for patients with AA. However, we found marked heterogeneity in the pooled HR for OS, and a subsequent sensitivity analysis revealed four studies had caused significant heterogeneity. After excluding these four studies [13, 16, 17, 23], the pooled HR for OS from the remaining 11 studies with 1875 patients was 0.955 (95% CI 0.443–1.468, P = 0.000, I2 = 34.7%) (Additional file: Figure S1), which still indicates the superiority of first-line allo-HSCT over IST for patients with AA. When OS was analyzed according to the publication year, there was a trend towards longer survival among patients undergoing first-line allo-HSCT compared to IST between 2010 and 2019 (HR = 0.286, 95% CI -0.008–0.58, P = 0.057, I2 = 0%). Data analysis prior to 2010 was not performed due to the extreme heterogeneity of the results. When OS was analyzed according to the study population, there was a trend towards longer survival among adult patients undergoing first-line allo-HSCT compared to those undergoing IST (HR = 0.801, 95% CI -0.056–1.658, P = 0.067, I2 = 0%). Although large heterogeneity was detected, first-line allo-HSCT was significantly superior to first-line IST for children with AA (HR = 1.068, 95% CI 0.358–1.779, P = 0.003, I2 = 54%). We then analyzed OS according to the disease severity, and there was significantly longer survival among SAA patients undergoing first-line allo-HSCT compared to first-line IST (HR = 0.506, 95% CI 0.13–0.881, P = 0.008, I2 = 38.8%). We further analyzed OS according to the donor type and found no difference in survival between patients undergoing first-line haploidentical-HSCT and IST (HR = 0.563, 95% CI -0.315–1.441, P = 0.209, I2 = 0%). However, significantly longer survival with large heterogeneity was observed in patients undergoing first-line MRD-HSCT compared to IST (HR = 0.711, 95% CI 0.053–1.37, P = 0.034, I2 = 73.4%).
Five studies including 802 patients reported FFS in the meta-analysis. The pooled HR for FFS was 1.962 (95% CI 1.43–2.493, P = 0.000, I2 = 0%) (Fig. 3), which indicates that first-line allo-HSCT was significantly superior to IST for patients with AA in regard to FFS.
Secondary outcomes
Three studies including 314 patients reported comparable ORR in the meta-analysis. The pooled RR for ORR was 1.691 (95% CI 1.433–1.996, P = 0.000, I2 = 11.6%). These data indicate that first-line allo-HSCT caused a higher treatment response compared to first-line IST for patients with AA. Four studies reported comparable CRs but they were not included in the data analysis due to their extreme heterogeneity. Only two studies including 238 patients reported comparable TRM. There was a significantly higher TRM among patients undergoing first-line allo-HSCT compared to IST (pooled RR 3.98, 95% CI 1.911–8.29, P = 0.000, I2 = 0%).
Ten studies including 3339 patients reported all-cause mortality. For these studies, the pooled RR was 0.851 (95% CI 0.618–1.174, P = 0.327, I2 = 57.1%) (Fig. 4). No difference in all-cause mortality was observed between patients who received first-line allo-HSCT and those who received IST. Subsequent analysis revealed no difference in mortality resulting from hemorrhage (pooled RR = 0.491, 95% CI 0.199–1.208, P = 0.122, I2 = 37%) but there was significantly higher mortality resulting from infection among patients who received first-line IST compared to first-line allo-HSCT (pooled RR = 1.378, 95% CI 1.081–1.757, P = 0.01, I2 = 0%). The median rate of engraftment was 96% (range 80–100%) and the median rate of graft failure was 5% (range 1–13%) for patients who received first-line allo-HSCT. Acute GVHD developed in 42.5% (range 23–100%) of patients, 25% (range 4–48%) of which were grade II–IV. Chronic GVHD developed in 30% (range 7–61%) of patients, 28% (range 6–39%) of patients had limited cGVHD, 6.5% (range 3–30%) of patients had extensive cGVHD, and 7% (range 2–27%) died of GVHD. Of the patients that received first-line IST, 4% (range 1–19%) developed MDS/AML and 15% (range 9.5–45%) experienced relapse. Only two studies reported the incidence of paroxysmal nocturnal hemoglobinuria (PNH) after first-line allo-HSCT or IST (6% and 12%, respectively).