Data extraction

The basic characteristics recorded for each study included the first author, year of publication, HMA, treatment schedule, trial number, phase or type of clinical trial, diagnostic criteria for LR-MDS, ethnicity of patients, and response criteria. Additionally, we recorded the baseline characteristics of the study participants including mean age, number of males, number of transfusion-dependent patients, number of 5q− patients, number of BM blast cells, prior treatment, karyotype, gene mutations, median EPO level, median ANC, median Hb, median platelet, median ferritin level, and use of ESA. We also extracted treatment-related data including overall response and TI rates, OS, and AEs if available.

The response rate was defined as the rate of different combinations of CR, mCR, PR, and HI according to IWG 2000 or 2006 criteria. For LR-MDS, TI rate was calculated as the number of TI patients at the time of evaluation relative to the total number of transfusion-dependent patients at baseline [42]. As response and TI rates showed a binomial distribution, they were subjected to arcsine transformation to normalize the proportions. For studies reporting OS, data were extracted either directly from the text or from Kaplan–Meier survival curves using Engauge Digitizer v11.1 for Windows software (https://markummitchell.github.io/engauge-digitizer/). Predicted OS rates at 1 and 2 years were estimated by pooling the OS. If the selected study included both high-risk and LR-MDS data, we used only the latter. We excluded studies that did no report risk categories. A subgroup analysis was conducted to compare the efficacy and safety of AZA and DAC in relation to patients’ baseline characteristics.