Results of the search

• The initial review in 2001 included 18 studies (18 reports) with 828 children (Durkan 2001a).

• The 2005 update (Durkan 2005) included 20 studies (24 reports) with 923 children.

• The 2008 update (Hodson 2008) included 26 studies (32 reports) with 1173 children.

• The 2013 update (Pravitsitthikul 2013) included 31 studies (46 reports) with 1443 children.

For the 2020 update we identified 71 new reports. The results of the literature search are shown in Figure 1.

Figure 1

---

---

Study flow diagram

The literature search identified 12 new included studies. Of these, there were nine new studies (APN 2006; ATLANTIS 2018; Iijima 2014; NEPHRUTIX 2018; Ravani 2015; RITURNS 2018; Sinha 2019; Uddin 2016; Zhang 2014). Three studies (Ahn 2018; Gruppen 2015; Iijima 2011) were previously listed as awaiting assessment or ongoing studies. APN 2006 was transferred from another Cochrane review (Hahn 2015) because it evaluated a non‐corticosteroid agent (cyclosporin). A total of 43 studies (91 reports) with 2541 children enrolled and 2428 children analysed were included.

Nine new studies were excluded (Basu 2015; Khemani 2016; Liu 2016c; Lou 2004; NCT02390362; Sasinka 1976; Wu 2015; Zedan 2016; Zhu 2013).

There are 11 new ongoing studies (Hama 2018; Horinouchi 2018; INTENT 2018; JSKDC 10 2019; LEARNS 2019; NCT02818738; NCT02972346; Ravani 2017; RITURNS II 2019; RITUXIVIG 2018; Sinha 2019b) (Characteristics of ongoing studies) and three recently completed studies (NCT01092962; NCT01895894; Sawires 2019) (Characteristics of studies awaiting classification) which will be assessed in a future update of this review.

The language used in this review has been updated to reflect uncertainty about findings per the Cochrane Consumer and Communication Guide (Glenton 2010).

Included studies

The characteristics of the 43 completed studies are shown in Characteristics of included studies. One study enrolled children with their initial episode of nephrotic syndrome (APN 2006) while the other studies enrolled children with relapsing nephrotic syndrome.

• Rituximab

  • Rituximab versus placebo or standard care: 6 studies (326 children) (Ahn 2018; RITURNS 2018; Iijima 2011; NEPHRUTIX 2018; Ravani 2011; Ravani 2015).

• Mycophenolate mofetil

  • MMF versus levamisole: one study (149 children) (Sinha 2019)

  • MMF versus cyclosporin: three studies (144 children) (Dorresteijn 2008; Gellermann 2013; Uddin 2016).

• Levamisole

  • Levamisole versus steroids or placebo or both: eight studies (476 children) (Abeyagunawardena 2006a; Al‐Saran 2006; BAPN 1991; Dayal 1994; Gruppen 2015; Rashid 1996; Sural 2001 (groups 1 and 3); Weiss 1993)

  • Levamisole versus cyclophosphamide: two studies (97 children) (Donia 2005; Sural 2001).

• Calcineurin inhibitors

  • Cyclosporin and prednisone versus prednisone alone: one study (106 children) (APN 2006). This study used cyclosporin for eight weeks, in addition to steroids, for patients with an initial presentation of nephrotic syndrome.

  • Different dosing regimens of cyclosporin: two studies (141 children) (Ishikura 2008; Iijima 2014)

  • Alkylating agents versus cyclosporin: two studies (112 children) (Niaudet 1992; Edefonti 1988)

• Alkylating agents (cyclophosphamide, chlorambucil)

  • Alkylating agents versus steroids or placebo: six studies (202 children) (Alatas 1978; Barratt 1970; Chiu 1973; Grupe 1976; ISKDC 1974; Sural 2001 (group 1 and 2))

  • Different cyclophosphamide or chlorambucil regimens: six studies (223 children) (Abeyagunawardena 2006b; Baluarte 1978; Barratt 1973; McCrory 1973; Prasad 2004; Ueda 1990)

  • Cyclophosphamide versus chlorambucil: one study (50 children) (APN 1982)

  • Cyclophosphamide versus vincristine: one study (39 children) (Abeyagunawardena 2007)

• Other agents

  • Mizoribine versus placebo: one study (197 children) (Yoshioka 2000)

  • Azithromycin versus steroids: one study (190 children) (Zhang 2014)

  • Azathioprine versus steroids: two studies (60 children) (ISKDC 1970; Barratt 1977)

  • ACTH versus placebo: one study (31 children) (ATLANTIS 2018)

  • Fusidic acid versus steroids: one study (18 children) (Cerkauskiene 2005).

• Ongoing studies (see Characteristics of ongoing studies)

  • MMF versus prednisone in the initial episode of nephrotic syndrome (INTENT 2018)

  • MMF and rituximab versus rituximab alone (Horinouchi 2018)

  • Daily prednisolone during URTI compared with ongoing levamisole (Sinha 2019b)

  • Levamisole/prednisolone versus placebo/prednisolone in the initial episode of nephrotic syndrome (NCT02818738)

  • Levamisole versus placebo on alternate days from week 4 for those in remission (LEARNS 2019)

  • Mizoribine/prednisolone versus prednisolone (Hama 2018)

  • ACTH with routine therapy versus routine therapy (NCT02972346)

  • Ofatumumab versus rituximab (Ravani 2017)

  • Different doses of rituximab versus placebo (JSKDC 10 2019)

  • Rituximab versus rituximab plus five immunoglobulin injections (RITURNS II 2019; RITUXIVIG 2018)

Prednisolone or prednisone was used in all studies either in combination with the study medication or to treat relapses. No studies comparing mizoribine or azathioprine with other non‐corticosteroid medications or comparing tacrolimus with steroids with or without placebo or any other corticosteroid‐sparing medication were identified.

Excluded studies

Previously 47 studies (49 reports) were excluded. For this update these studies were reviewed and we excluded non‐RCTs and studies with clearly ineligible interventions or populations leaving 26 excluded studies. Nine new studies were excluded: five enrolled the wrong population (Khemani 2016; Liu 2016c; Lou 2004; Sasinka 1976; Wu 2015); two did not use a non‐corticosteroid immunosuppressive (Zedan 2016; Zhu 2013); one was withdrawn from publication (Basu 2015); and one study was discontinued due to inadequate enrolment (NCT02390362).

See Characteristics of excluded studies.

Allocation

Blinding

Incomplete outcome data

More than 20% of participants did not complete follow‐up in five studies which were assessed as being at high risk of attrition bias (Alatas 1978; APN 2006; Dorresteijn 2008; Edefonti 1988; Uddin 2016). Completeness of follow‐up was unclear in one study (Rashid 1996) and the remaining 37 studies were considered to be at low risk of attrition bias.

Selective reporting

Selective reporting was considered to be present if studies did not report a measure of relapse that could be meta‐analysed or did not report adverse effects of study drugs. Overall, 11 studies were assessed at high risk of reporting bias. Specific issues identified were failure to document adverse effects or expression in percentages (Alatas 1978; APN 2006; Barratt 1970; Barratt 1973; Barratt 1977; NEPHRUTIX 2018; Weiss 1993; Zhang 2014); only combined data for both groups were available in a cross‐over study (Cerkauskiene 2005); or data were available as hazard ratios (Ishikura 2008; Yoshioka 2000). Risk of reporting bias was unclear in four studies that were available only as abstracts (Abeyagunawardena 2006a; Abeyagunawardena 2006b; Abeyagunawardena 2007; Uddin 2016) and one study (Ravani 2015) with limited reporting of adverse effects. The 27 remaining studies were considered to be at low risk of reporting bias.

Other potential sources of bias

Five studies (ATLANTIS 2018; BAPN 1991; Dorresteijn 2008; Gellermann 2013; Weiss 1993) received funding from pharmaceutical companies and were considered to be at high risk of bias. Sixteen studies reported funding from governments, universities or not‐for‐profit groups and were classified at low risk of bias (Ahn 2018; Alatas 1978; Al‐Saran 2006; APN 1982; Baluarte 1978; Barratt 1970; Barratt 1973; Dayal 1994; Gruppen 2015; Ishikura 2008; ISKDC 1974; McCrory 1973; NEPHRUTIX 2018; Ravani 2011; Sinha 2019; Zhang 2014). The remaining 22 studies did not report if they received external funding.

Rituximab versus placebo or control

• Rituximab used alone or with prednisone and CNI versus placebo, prednisone or CNI and prednisone probably reduces the number of children experiencing relapse at three months (Analysis 1.1.1 (3 studies, 132 children): RR 0.32, 95% CI 0.14 to 0.70; I² = 26%), six months (Analysis 1.1.2 (5 studies, 269 children): RR 0.23, 95% CI 0.12 to 0.43; I² = 31%) and 12 months (Analysis 1.1.3 (3 studies, 198 children): RR 0.63, 95% CI 0.42 to 0.93; I² = 53%) (moderate certainty evidence).

• Rituximab may increase the number of children with severe infusion reactions (Analysis 1.2.1 (4 studies, 252 children): RR 5.83, 95% CI 1.34 to 25.29; I² = 0%) but not severe infections (Analysis 1.2.2 (3 studies, 222 children): RR 0.90, 95% CI 0.26 to 3.15) (low certainty evidence) or arthropathy (Analysis 1.2.3 (2 studies, 84 children): RR 3.92, 95% CI 0.45 to 33.98; I² = 0%) (low certainty evidence).

The evidence was downgraded due to imprecision, risk of bias and heterogeneity (summary of findings Table for the main comparison).

Mycophenolate mofetil versus levamisole

• Levamisole when compared with MMF may make little or no difference to the number of children in relapse at 12 months (Analysis 2.1.1 (1 study, 149 children): RR 0.90, 95% CI 0.70 to 1.16) (low certainty evidence).

• It is unclear whether levamisole decreased abdominal pain (low certainty evidence).

The evidence was downgraded because of risk of bias and imprecision (summary of findings Table 2).

Mycophenolate mofetil versus cyclosporin

• MMF versus cyclosporin may make little or no difference to the number of patients relapsing within 12 months (Analysis 3.1 (2 studies, 82 children): RR 1.90, 95% CI 0.66 to 5.46; I² = 30%) or to the relapse rate/year Analysis 3.2 (3 studies, 142 children): MD 0.83, 95% CI 0.33 to 1.33; I² = 0%) (low certainty evidence).

• MMF versus cyclosporin probably reduces the number of patients with hypertrichosis (Analysis 3.3.2 (3 studies, 140 children): RR 0.23, 95% CI 0.10 to 0.50; I² = 0%) and gum hypertrophy (Analysis 3.3.4 (3 studies, 144 children): RR 0.09, 95% CI 0.02 to 0.47; I² = 0%) (low certainty evidence), but may make little or no difference to the number of children with other adverse effects (hypertension, lymphopenia, reduced GFR, pneumonia and diarrhoea) (Analysis 3.3) (low certainty evidence).

The evidence was downgraded due to risk of bias and imprecision (summary of findings Table 3).

Levamisole versus steroids or placebo or both, or no treatment

Levamisole was administered for four months (BAPN 1991), six months (Rashid 1996; Sural 2001; Weiss 1993) and 12 months (Abeyagunawardena 2006a; Al‐Saran 2006; Dayal 1994; Gruppen 2015).

• Levamisole compared with steroids or placebo may reduce the number of children with relapse during treatment (Analysis 4.1 (8 studies, 474 children): RR 0.52, 95% CI 0.33 to 0.82; I² = 89%) and at 6 to 12 months (end of treatment) (Analysis 4.2 (8 studies, 462 children): RR 0.65, 95% CI 0.48 to 0.88; I² = 87%). (low certainty evidence).

• It is uncertain whether levamisole compared with placebo or no treatment alters the mean relapse rate because the certainty of the evidence is very low (Analysis 4.3).

• Levamisole compared with steroids or placebo may make little or no difference to the number of children developing leucopenia or arthritis (Analysis 4.5) (low certainty evidence).

The evidence was downgraded because of risk of bias and imprecision (summary of findings Table 4).

The heterogeneity between studies was investigated in sensitivity analyses in which studies at low risk of bias for allocation concealment and blinding were separated from those at unclear or high risk of bias (Figure 4). This identifies that studies at unclear or high risk of bias resulted in a greater benefit of levamisole (Analysis 4.4.2) compared with those studies at low risk of bias (Analysis 4.4.1). These data are consistent with empirical evidence that lack of blinding is associated with an increase in apparent treatment effect (Schulz 1995). Among the three studies at low risk of bias, heterogeneity was eliminated when Weiss 1993 was excluded. This study showed no apparent benefit of levamisole, but levamisole was given on two consecutive days out of seven rather than on alternate days as in the other studies.

Figure 4

---

---

Forest plot of comparison: 6 Levamisole versus steroids or placebo or both, or no treatment, outcome: 6.4 Relapse during treatment according to risk of bias.

Levamisole versus cyclophosphamide

• Levamisole when compared with cyclophosphamide may make little or no difference to the numbers of children at six to nine months (Analysis 5.1.2 (2 studies, 97 children): RR 1.17, 95% CI 0.76 to 1.81; I² = 43%), at 12 months (Analysis 5.1.3 (1 study, 40 children): RR 0.89, 95% CI 0.68 to 1.16) and at 24 months (Analysis 5.1.4 (1 study, 40 children): RR 0.89, 95% CI 0.73 to 1.10). It is uncertain whether levamisole compared with cyclophosphamide makes any difference to the number of children with relapse at the end of therapy because the certainty of the evidence is very low (Analysis 5.1.1 (2 studies, 97 children): RR 2.14, 95% CI 0.22 to 20.95; I² = 79%).

• Adverse effects (infection, leucopenia) may not differ between treatment groups (Analysis 5.2).

The evidence was downgraded because of risk of bias issues, heterogeneity and imprecision (summary of findings Table 5).

Cyclosporin and prednisone versus prednisone alone

• The addition of cyclosporin to 12 weeks of prednisone therapy for treatment of the first presentation of nephrotic syndrome may reduce the risk of relapse at six months (Analysis 6.1 (1 study, 104 children): RR 0.33, 95% CI 0.33 to 0.83) and but not at 12 months (Analysis 6.2 (1 study, 104 children): RR 0.72, 95% CI 0.46 to 1.13) (low certainty evidence). It is uncertain if the number needing cytotoxic agents differed between groups (Analysis 6.3) because the certainty of the evidence was very low.

• It is uncertain if cyclosporin added to prednisone makes any difference to the end‐of‐follow‐up serum creatinine (very low certainty evidence).

• Hypertrichosis and gum hypertrophy were seen in 60% and 9% of children given cyclosporin, respectively. Psychological disturbances occurred in 27% of cyclosporin‐treated patients compared with 14% of patients treated with prednisone alone.

The evidence was downgraded due to risk of bias and imprecision (summary of findings Table 6).

Alkylating agents versus cyclosporin

• Cyclosporin when compared to cyclophosphamide given for eight weeks or chlorambucil given for six weeks, may make little or no difference to the risk of relapse during cyclosporin treatment (6 to 9 months) (Analysis 7.1 (2 studies, 95 children): RR 0.91, 95% CI 0.55 to 1.48; I² = 0%) (low certainty evidence).

• Alkylating agents may be more effective than cyclosporin in maintaining remission at 12 to 24 months after either has been ceased (Analysis 7.2 (2 studies, 95 children): RR 0.51, 95% CI 0.35 to 0.74; I² = 12%) (low certainty evidence).

• Cyclosporin compared with alkylating agents may result in raised serum creatinine, hypertrichosis and gum hypertrophy (Analysis 7.3) (low certainty evidence). It is uncertain if cyclosporin makes any difference to hypertension, and it is uncertain whether alkylating agents result in leucopenia because the certainty of the evidence is very low.

The evidence was downgraded due to risk of bias issues and imprecision (summary of findings Table 7).

Alkylating agents versus steroids or placebo or both

• Cyclophosphamide compared with prednisone or placebo probably reduces the number of children relapsing at six to 12 months. (Analysis 8.1.1 (4 studies, 161 children): RR 0.47, 95% CI 0.33 to 0.66; I² = 0%) (moderate certainty evidence), and may reduce the number of children relapsing at 13 to 24 months (Analysis 8.2.1 (2 studies, 27 children): RR 0.21, 95% CI 0.07 to 0.65; I² = 0%) (low certainty evidence).

• Similarly, chlorambucil compared with placebo or prednisone alone may reduce the number of children experiencing relapse at six months (Analysis 8.1.2 (2 studies, 41 children): RR 0.19, 95% CI 0.03 to 1.09; I² = 44%) and 12 months (Analysis 8.2.2 (2 studies, 32 children): RR 0.15, 95% CI 0.02 to 0.95; I² = 39%) (low certainty evidence) .

• When the studies evaluating cyclophosphamide or chlorambucil were combined, alkylating agents when compared to prednisone or placebo probably reduce the number of patients experiencing relapse at six to 12 months (Analysis 8.1 (6 studies, 202 children): RR 0.44, 95% CI 0.32 to 0.60; I² = 0%) (moderate certainty evidence) and may reduce the number suffering relapse at 13 to 24 months (Analysis 8.2 (4 studies, 59 children): RR 0.20, 95% CI 0.09 to 0.46; I² = 0%) (low certainty evidence).

• Leucopenia was reported in three studies (Analysis 8.3). Leucopenia developed in children receiving cyclophosphamide or chlorambucil. However, because of small numbers of events and included children, It is uncertain whether alkylating agents compared with prednisone or placebo result in leucopenia because the certainty of the evidence is very low.

The evidence was downgraded because of increased risk of bias and imprecision (summary of findings Table 8).

Cyclophosphamide: different durations, doses, route

• It is uncertain whether cyclophosphamide given for eight weeks compared with two weeks results in fewer children relapsing within 12 months (Analysis 9.1.1 (1 study, 32 children): RR 0.15, 95% CI 0.04 to 0.57), because the certainty of the evidence is very low.

• Cyclophosphamide given for 12 weeks compared to eight weeks may make little or no difference in the number of children relapsing at 12 months (Analysis 9.1.2 (1 study, 73 children): RR 1.04, 95% CI 0.75 to 1.44) or 24 months (Analysis 9.2.1 (1 study, 73 children): RR 0.98, 95% CI 0.74 to 1.28) or to the number of children with leucopenia (Analysis 9.3) (low certainty evidence).

• It is uncertain if cyclophosphamide given as 5 mg/kg/day over 6 weeks versus 2.5 mg/kg/day over 12 weeks makes any difference to the number of children relapsing at 12 months (Analysis 10.1 (1 study, 14 children): RR 2.33, 95% CI 0.11 to 48.99), or to adverse effects (Analysis 10.2), due to very low certainty evidence.

• IV cyclophosphamide (monthly for six months) when compared to oral cyclophosphamide (given for 12 weeks) may reduce the number of children with relapse at 6 months (Analysis 11.1 (2 studies, 83 children): RR 0.54, 95% CI 0.34 to 0.8; I² = 0%) but may make little or no difference to relapse at 12 to 24 months (Analysis 11.3 (2 studies, 83 children): RR 0.99, 95% CI 0.76 to 1.29; I² = 0%) (low certainty evidence). Comparing IV cyclophosphamide with oral cyclophosphamide, infections may be more common in the oral group (Analysis 11.4.3 (2 studies, 83 children): RR 0.14, 95% CI 0.03 to 0.72; I² = 0%) (low certainty evidence) (summary of findings Table 9).

The evidence was downgraded because of risk of bias issues and imprecision.

Chlorambucil (different dose regimens)

• Comparing chlorambucil regimens, it is uncertain if an increasing dose regimen versus a stable dose regimen makes any difference to relapse at 12 months (Analysis 12.1 (1 study, 21 children): RR 0.18, 95% CI 0.01 to 3.41) or to the number with leucopenia or thrombocytopenia (Analysis 12.2) (very low certainty evidence). However, there was a 34% increase in the incidence of leucopenia and an 18% increase in thrombocytopenia with the increasing dose regimen.

The evidence was downgraded due to risk of bias and serious imprecision.

Cyclophosphamide versus chlorambucil

• Chlorambucil when compared with cyclophosphamide may make little or no difference to relapse at 12 months (Analysis 13.1 (1 study, 50 children): RR 1.15, 95% CI 0.69 to 1.94) and 24 months (Analysis 13.2 (1 study, 50 children): RR 1.31, 95% CI 0.80 to 2.13) (low certainty evidence).

• Chlorambucil when compared with cyclophosphamide may result in more children with lymphopenia and thrombocytopenia (Analysis 13.3) (low certainty evidence)

The evidence was downgraded due to risk of bias and imprecision (summary of findings Table 10).

Alkylating agents versus vincristine

• IV cyclophosphamide when compared to IV vincristine may make little or no difference to the risk of relapse at 12 months (Analysis 14.1 (1 study, 39 children): RR 0.54, 95% CI 0.26 to 1.12) and 24 months (Analysis 14.2 (1 study, 39 children): RR 0.73, 95% CI 0.45 to 1.18) (low certainty evidence). Although no major side effects were reported, two children treated with vincristine had abdominal cramps and constipation, and five children receiving cyclophosphamide experienced vomiting that required anti‐emetics.

The evidence was downgraded to low certainty due to increased risk of bias and imprecision.

Cyclosporin dose

• After an initial 6‐month course of cyclosporin, cyclosporin used to maintain remission at doses to maintain levels at 60 to 80 ng/mL (mean dose 5.4 mg/kg/day), compared to cyclosporin at a fixed dose of 2.5 mg/kg/day, may reduce the number of children experiencing relapse at 12 months (Analysis 15.1.2 (1 study, 44 children): RR 0.33, 95% CI 0.16 to 0.70) and 24 months (Analysis 15.1.3 (1 study, 44 children): RR 0.65, 95% CI 0.45 to 0.94) (low certainty evidence).

• The variable dose schedule compared with fixed dose may make little or no difference to adverse effects (Analysis 15.2). (low certainty evidence).

The evidence was considered of low certainty due to risk of bias and imprecision.

• High‐dose cyclosporin (target C2 levels 600 to 700 ng/mL for first 6 months followed by 450 to 550 ng/mL for the next 18 months) compared with low‐dose cyclosporin (target C2 levels 450 to 550 ng/mL for the first 6 months and 300 to 400 ng/mL for the next 18 months) may make little or no difference to relapse at 2 years (Analysis 15.3.1 (1 study, 85 children): RR 0.74, 95% CI 0.45 to 1.22) or to adverse effects including hypertension, hirsutism and nephrotoxicity (Analysis 15.4). (low certainty evidence).

The evidence was considered of low certainty due to risk of bias and imprecision.

Mizoribine versus placebo

• Data on the number of children with relapse at 6 and 12 months who had received mizoribine or placebo could not be extracted. The reported hazard ratio of cumulative remission rate of 0.79 (95% CI 0.57 to 1.08) suggested that mizoribine when compared with placebo may make little or no difference to the risk of relapse.

• Mizoribine may increase the risk of hyperuricaemia (Analysis 16.1.2 (1 study, 197 children): RR 3.96, 95% CI 1.37 to 11.42) but not leucopenia or hepatic dysfunction (Analysis 16.1) (low certainty evidence).

The evidence was downgraded for adverse effects due to risk of bias issues and imprecision .

Azithromycin and prednisone versus prednisone alone

• Azithromycin and prednisone when compared with prednisone alone may make little or no difference to the risk of relapse by six months (Analysis 17.1 (1 study, 190 children): RR 0.55, 95% CI 0.30 to 1.02) (low certainty evidence).

• Adverse effects were not reported.

The evidence was downgraded due to risk of bias and imprecision.

Azathioprine versus steroids

• Azathioprine compared with steroids or placebo may make little or no difference to the number of children who relapse at six months (Analysis 18.1 (2 studies, 60 children): RR 0.90, 95% CI 0.59 to 1.38; I² = 3%) (low certainty evidence).

The evidence was downgraded due to risk of bias issues and imprecision.

Fusidic acid versus steroids

Fusidic acid and prednisone was compared with prednisone alone in a cross‐over study involving 18 children (Cerkauskiene 2005). The results for all courses of fusidic acid and prednisone (14 courses) and prednisone alone (17 courses) were combined so no meta‐analyses could be performed. There were no differences in the mean time to remission (12.6 ± 6.6 days for fusidic acid/prednisone versus 13.9 ± 7.4 days for prednisone alone) or in time to relapse (18.3 ± 23.9 weeks versus 17.8 ± 20.4 weeks). One child developed an allergic rash with fusidic acid.

ACTH

• ACTH when compared with placebo may make little or no difference to the risk of relapse at 6 months (Analysis 19.1 (1 study, 31 children): RR 1.00, 95% CI 0.83 to 1.20) (low certainty evidence).

• Adverse effects were reported for all 28 children, who received ACTH (those randomised to ACTH and those in the control arm receiving ACTH after relapse). The most common adverse effects were behavioural changes (25%), sleep disturbances (18%), cushingoid features (18%), injection site irritation (25%) and increased appetite (32%).

The evidence was downgraded due to risk of bias and imprecision.

Subgroup analyses

There were insufficient studies of any treatment combination to enable conducting detailed subgroup analyses.