Fifty-six studies including 3067 patients proved eligible [10,11,12,13,14, 24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75] (Fig. 1; Supplementary Table 2). The studies had a median sample size of 53 participants (range 9–190). The mean study age ranged between 32 and 75.1 years (median 47.4 years). The mean study baseline creatinine ranged from 0.74 to 2.7 mg/dL (median 1.18 mg/dL) and the mean study 24-h urine protein excretion ranged between 1.64 and 12.8 g/day (median 7.86 g/day). Follow-up for clinical outcomes ranged from 6 months to 10 years (median 23.4 months). Of the 56 studies included in the analysis, 51 had 2 arms and 5 had three arms in the study design.

Cyclophosphamide was compared with steroid monotherapy (3 trials, 137 participants) [27, 41, 57], calcineurin inhibitor (4 trials, 299 participants) [34, 44, 64, 72], rituximab (1 trial, 74 participants) [14], calcineurin inhibitor plus rituximab (1 trial, 86 participants) [12], chlorambucil (3 trials, 145 participants) [29, 59, 65], mycophenolate mofetil (1 trial, 22 participants) [68], mizoribine (1 trial, 55 participants) [71], ACTH (1 trial, 32 participants) [60], non-immunosuppressive therapy (4 trials, 209 participants) [38, 47, 54, 70] and Chinese herbal medicine (1 trial, 190 participants) [35]. There were 3 trials with 59 participants, which compared different protocols of cyclophosphamide (early versus late initiation or intravenous versus oral) [37, 45]. Chlorambucil was compared with steroid monotherapy (1 trial, 20 participants) [26], mycophenolate mofetil (1 trial, 20 participants) [33] and non-immunosuppressive therapy (2 trials, 130 participants) [61, 62]. Calcineurin inhibitors were compared with rituximab (1 trial, 130 participant) [13], non-immunosuppressive therapy (3 trials, 95 participants) [32, 63, 74], steroid monotherapy (1 trial, 51 participants) [10] and azathioprine (1 trial, 23 participants) [55]. There were 7 trials with 290 participants, which compared calcineurin inhibitors against calcineurin inhibitor (such as different doses or cyclosporin versus tacrolimus) [48, 51, 52, 56, 67, 69, 75]. Steroid monotherapy was compared with placebo (2 trials, 203 participants) [31, 36] and non-immunosuppressive therapy (1 trial, 40 participants) [30]. Mycophenolate mofetil was also compared with non-immunosuppressive therapy (2 trials, 77 participants) [39, 40]. Mycophenolate mofetil was compared with a combination of calcineurin inhibitor and mycophenolate mofetil (1 trial, 20 participants) [73]. Rituximab was compared with non-immunosuppressive therapy (1 trial, 75 participants) [11]. Mizoribine was compared with steroid monotherapy (1 trial, 36 participants) [43] and also against different doses of mizoribine (1 trial, 37 participants) [66]. Azathioprine was compared with non-immunosuppressive therapy (2 trials, 23 participants) [24, 25]. Pentoxifylline was compared with placebo (1 trial, 18 participants) [28].

Of the 5 studies with 3 arms in the study design, one compared chlorambucil with steroid monotherapy and non-immunosuppressive therapy (60 participants) [42], one compared cyclophosphamide with calcineurin inhibitor and non-immunosuppressive therapy (28 participants) [50], one compared chlorambucil with calcineurin inhibitor and non-immunosuppressive therapy (108 participants) [46], one compared cyclophosphamide with mycophenolate mofetil and calcineurin inhibitor (90 participants) [58] and one compared cyclophosphamide with leflunomide and a combination of cyclophosphamide plus leflunomide (72 participants) [53].

In the cyclophosphamide group, all studies except one [38] used steroids along with cyclophosphamide. The definition of complete and partial remission was not uniform across all the studies (Supplementary Table 3).

Seventeen (29.3%) studies were adjudicated as being at low risk of bias in methods used to generate the random sequence and 12 (20.6%) studies were at low risk of bias in methods of allocation concealment (Supplementary Table 4). Only two (3.7%) studies reported blinding of participants and investigators. None of the studies reported blinding of outcome assessment. Twelve (20.6%) studies were adjudicated as being at low risk of incomplete outcomes and selective reporting.

Figure 2 and Supplementary Fig. 1 (1.1–1.9) show the formed evidence network for each outcome. There was no strong evidence for global network inconsistency (Supplementary Table 7). Direct and indirect estimates were generally coherent (Supplementary Table 5). Tables 1, 2 and Supplementary Table 8 present the network estimates for each strategy comparison for all outcomes.

Network plots for effects of immunosuppression on disease complete and partial remission in idiopathic membranous nephropathy. The size of each node is proportional to the sample size and the width of the lines represents the number of each pairwise comparison. ACTH adrenocorticotropic hormone

Twenty-six studies including 1475 patients reported complete remission (Fig. 2) [10, 12, 14, 26, 31, 33, 36, 39, 40, 42, 47, 50, 53,54,55, 57, 58, 60,