3.1 Study selection procedure
The study selection processes are illustrated in Figure 1. First, our initial search yielded 3438 citations from PubMed, Embase, Cochrane. 309 of which were removed due to duplication. By scanning their titles and abstracts, 2970 were excluded due to the following reasons: 2430 studies were not relevant to the subject; 72 studies were in vitro, animal, and other non-clinical studies; 386 studies were other drugs; 23 studies were other diseases; 51 studies were case reports; 35 studies were reviews. Of the 159 full-text studies, 144 studies were deleted because they didn’t report the primary outcomes of interest. Therefore, 15 studies were included eventually containing 6 without control[23, 37-41] and 9 studies with control comprised of 3 prospective studies [24, 42, 43]and 6 retrospective studies[22, 44-47].
3.2 Study Characters
The characteristics of included studies were presented in Supplemental Table 1. 13 studies [22-24, 37, 38, 40, 41, 43-48] used sorafenib, one[39] used quizartinib and one [42]used midostaurin. Four studies[20, 24, 42, 47] focus on patients with FLT3-ITD AML in first CR. While other studies included patients with FLT3-ITD positive AML in first, second, and more hematologic remission (CHR). The dose of sorafenib medication was 200 and 400mg twice daily for negative and positive MRD separately[49]. While the midostaurin [42, 50]was given 50 mg twice daily in twelve 4-week cycles. And the quizartinib [39] was given 40 and 60 mg daily in the first and second dose levels separately.
3.3 Prognosis of AML patients post HSCT
Among the 15 studies, the end points analysis including OS, LFS, CIR, and NRM were conducted for evaluating the clinical effect. The HR of OS extracted from nine articles (eight sorafenib[22, 24, 43-48] and one midostaurin [42] were analyzed with fixed-effect models due to no significant heterogeneity between them (P=0.63, I2=0%), and the results were shown in a forest plot (HR=0.38, 95% CI=0.29–0.49, P<0.00001; Figure2.1.1). The pooled HR of LFS from seven articles [22, 24, 42-44, 46, 50] was performed using a fixed-effect model (HR=0.35, 95% CI=0.27–0.47, P<0.00001), with no heterogeneity observed (P=0.97 I2=0%; Figure 2.1.2). We also calculated the HR of CIR from six articles[22, 24, 43, 44, 46, 48] with no significant heterogeneity (P=0.96, I2=0% Figure 2.1.3) and discovered the result (HR=0.32, 95% CI=0.23–0.45, P<0.00001) with fixed-effect models. The pooled HR of NRM from five articles[22, 24, 43, 44, 46] was evaluated using fixed-effect models (HR=0.74, 95% CI=0.46–1.19, P=0.21), without heterogeneity between studies (P=0.86 I2=0%; Figure 2.1.4).
3.4 Adverse effects
No significant difference was found for hematological toxic effects, infection, and other non-hematological toxic effects among whether to use sorafenib (Figure 3). Hematological toxic effects were common adverse events including anemia, thrombocytopenia, and neutropenia. The pooled OR of hematological toxic effects from four articles[24, 42-44] was 1.47(95% CI=0.43–5.09, P=0.54), evaluated by random-effect models with heterogeneity between studies (P=0.07 I2=58%; Figure 3.1.1). And the OR for gastrointestinal、cardiovascular、cutaneous and infectious toxicity was 1.20(95%CI 0.52-2.81, P=0.67; Figure 3.1.1), 1.75(95%CI 0.50-6.10, P=0.38; Figure 3.1.2), 2.52(95%CI 0.65-9.71, P=0.18; Figure 3.1.3), and 1.15(95%CI 0.65-2.04, P=0.63; Figure 3.1.4) separately.
3.5 Graft-versus-host disease (GVHD)
The pooled OR of aGVHD from five articles [24, 42-44, 46] was 1.00 (95% CI=0.64–1.54, P=0.99) evaluated by fixed-effect models with none significant differences in heterogeneity. (P=0.44, I2=0%; Figure 4.1.1). As to the pooled OR of cGVHD, we acquired the result (OR=1.30, 95% CI=0.86–1.96, P=0.21) from four articles[24, 42-44, 46] with no significant heterogeneity (P=0.49, I2=0%; Figure 4.1.2).
3.6 MRD status before and after HSCT on LFS
A China study[43] and a Germany study[44] reported the effect of MRD status before and after HSCT on LFS, so that a subgroup analysis was conducted to research it. The HRs of LFS for patients with MRD-positive before HSCT, MRD-negative before HSCT, MRD-positive after HSCT, and MRD- negative after HSCT were (HR=0.48, 95% CI=0.25–0.95, P=0.04), (HR=0.23, 95% CI=0.10–0.52, P=0.004), (HR=0.19, 95% CI=0.07–0.54, P=0.002), (HR=0.45, 95% CI=0.19-1.03, P=0.06) (Figure 5).
3.7 Single-arm meta-analysis
These results were very consistent with those obtained from early researches without control[23, 37-41], in which decreased recurrence and prolonged survival was discovered in our single-arm meta-analysis. As additional supporting evidence, the single-arm meta-analysis was conducted (supplemental Fig. 1-3, available as supplemental material). The pooled rates of 2-year OS, 2-year LFS, 2-year relapse after HSCT with FLT3 inhibitors maintenance were 66% (95% CI, 45–85%, I2 = 84.15%; Fig.S1.1), 72% (95% CI, 62–80%, I2 = 0%; Fig.S1.2), 21% (95% CI, 3–45%, I2 = 69%; Fig.S1.3), separately. The pooled rates of hematological toxic effects, gastrointestinal、cardiovascular、cutaneous and biochemical toxicity were 46% (95% CI, 22–74%, I2 = 88.02%; Fig.S2.1), 24% (95% CI, 15–34%, I2 = 28.03%; Fig.S2.2), 12% (95% CI, 3–25%, I2 = 67.34%; Fig.S2.3), 22% (95% CI, 6–42%, I2 = 80.66%; Fig.S2.4), 12% (95% CI, 1–28%, I2 = 77.56%; Fig.S2.5), separately. The pooled rates of aGVHD and cGVHD were 29% (95% CI, 9–53%, I2 = 85.50%; Fig.S3.1), 20% (95% CI, 2–48%, I2 = 89.99%; Fig.S3.2), separately.