Efficacy and safety of immunosuppressive treatment in IgA nephropathy: a meta-analysis of randomized controlled trials

This investigation required studies to meet the following inclusion criteria: the study was an RCT; the study compared different immunosuppressive agents versus non-immunosuppressive agents/placebo/no treatment; and study subjects were adult or pediatric patients with biopsy-proven IgAN.

Studies were rejected according to the following exclusion criteria: immunosuppressant not given orally or intravenously (intestinal steroid budesonide was excluded though 20% bioavailable); study subjects with secondary IgAN; no data available for this study in the article, data included in other articles, or data repeated in other articles; and article not in English.

The MEDLINE, EMBASE, and Cochrane Library medical databases were searched to retrieve relevant studies. Searches were performed in English, and each search retrieved studies that were published between establishment of the database and May 2018.

A comprehensive search strategy was established to ensure the comprehensive and accurate retrieval of studies. The MEDLINE and Cochrane Library databases were searched using the method described in the Cochrane Policy Manual [6], whereas EMBASE was searched using a sensitivity–specificity filter optimized [7]. The following search terms were used: IgAN, steroids, glucocorticoids, immunosuppressive agents, angiotensin-converting enzyme inhibitors, angiotensin receptor antagonists, and placebo. Furthermore, we also searched relevant professional journals manually.

Two investigators (ZZ and YY) independently selected studies from the retrieved literature and extracted the data and analytical results. If the two investigators had different opinions about the quality of a study, a third investigator (SMJ) examined the controversial literature and discussed it with the two aforementioned investigators. Data were included only if the three authors achieve consensus regarding the data.

If necessary, daily proteinuria was recalculated as g/day. Values for eGFR were based on the data provided by the authors of the included studies.

We assessed treatment-related changes based on mean values and standard deviations (SDs) changes between the pre-treatment and post-treatment. As the standard error of the mean (SEM) was used in some studies, we calculated the SD using the formula: SEM × square root of sample size. In addition, 95% confidence intervals (CIs) were used in some studies; we calculated the SD using the formula: ((upper limit of 95% CI – lower limit of 95% CI)/(2 × 1.96)) × √(n). Assessment of the risks of publication bias followed the Cochrane handbook.

Two authors (ZZ and YY) independently evaluated risk of bias using the Cochrane risk-of-bias tool [8]. We reviewed each trial and gave a score of bias according to the following criteria: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other bias.

To compare the effects of immunosuppressive agents and control treatment on proteinuria excretion and serum levels of creatinine, data on eGFR and end-stage renal disease (ESRD) were extracted for meta-analyses. Subgroup analyses were performed for each outcome based on the type of immunosuppressive agent.

For continuous outcomes, the differences in means and the 95% CI in mean change between baseline and end of treatment value were calculated for individual trials, and the weighted mean difference (WMD) was used as a summary estimator. Dichotomous outcome data from individual trials were analyzed using the relative risk (RR) measure and 95% CI. Heterogeneity of treatment effects between studies was investigated visually by examination of plots and statistically using the heterogeneity χ² and I² statistics. In all analyses, P < 0.05 was taken to indicate statistical significance. The fixed-effects and random-effects models were used for the meta-analysis of each indicator. Analyses were performed using Review Manager 5.2 (RevMan; Cochrane Collaboration, Oxford, UK).

To assess whether the current meta-analysis had a adequate sample size to draw firm conclusions about the effects of interventions, we performed trial sequential analyses (TSAs) for outcomes. TSAs involves a cumulative meta-analysis to create a Z curve of the summarized observed effect and the monitoring boundaries for benefit and harm and estimate the optimal sample size [9]. A sufficient level of evidence for the anticipated intervention effect may have been reached when the cumulative z curve crosses the trial sequential monitoring boundary. And there is insufficient evidence to reach a conclusion, if the z curve crosses none of the boundaries and the required information size has not been reached. These analyses were performed using the software TSA version 0.9 Beta (Copenhagen Trial Unit, Copenhagen, Denmark).

After performing electronic and manual searches, 4016 potentially relevant papers were obtained. After removing duplicated papers, 2639 papers remained. After browsing the titles and abstracts, 53 papers were selected. After reading the entire text of these 53 papers, 24 papers were excluded, and 28 papers describing 25 trials with a total of 1957 patients were ultimately included. The literature selection process is illustrated in Fig. 1, and detailed information regarding the examined studies is provided in Table 1 [10‐38].

By current standards, reporting of key indicators of trial quality was suboptimal. Some studies in particular provided few details on the process of randomization and concealment of allocation. Only six studies were double-blinded trials. Seven studies used an open-label design. The bias and overall risk diagrams of the included studies are presented in Fig. 2.

The difference in the means of urinary protein excretion between end of treatment and baseline was significantly lower in the steroid group than in controls (five trials [17, 21, 23, 29, 36], 222 patients; WMD –0.51, 95% CI − 0.73 to − 0.28, with a fixed-effects model; WMD –0.70, 95% CI − 1.2 to − 0.20, with a random-effects model; I² = 58%; Fig. 3). After removing Lai [23], heterogeneity I² changed to 0.

Patients receiving NSI alone showed a more significant reduction of urinary protein excretion after treatment compared to controls (seven trials [12, 26, 30‐34], 660 patients, WMD –0.43, 95% CI − 0.55 to 0.31, with a fixed-effects model; WMD –0. 43, 95% CI − 0.55 to − 0.31, with a random-effects model; I² = 0; Fig. 3).

With the S&NSI treatment approach, patients had a more significant reduction of urinary protein excretion after treatment compared to controls (three trials [10, 18, 28], 278 patients, WMD –0.16, 95% CI − 1.8 to − 1.4, I² = 83%, with a fixed-effects model; WMD –1.42, 95% CI − 2.18 to − 0.66, I² = 89%, with a random-effects model; Fig. 3). After removing Yoshikawa [18], heterogeneity I² changed to 0.

TSAs of steroids, NSI, and S&NSI all indicated that the cumulative z curve crossed both the conventional boundary and the trial sequential monitoring boundary (Fig. 4).

There were no statistically significant differences in creatinine changes between baseline and end of treatment between immunosuppressive treatment and control groups (nine trials [11, 13, 17, 19, 20, 23, 26, 29, 31], 420 patients, WMD –0.03, 95% CI − 0.11 to 0.15, with a fixed-effects model; WMD –0.03, 95% CI − 0.11 to 0.05, with a random-effects model; I² = 0%; Fig. 5).

TSAs of nine comparisons illustrated that the cumulative z curve did not cross the conventional boundary or the line of required information size, indicating that the evidence was insufficient. Therefore, further trials are required.

The differences in the means of eGFR between end of treatment and baseline were significantly higher in the NSI group than in controls (five trials [16, 20, 25, 32, 33], 817 patients; WMD 5.17, 95% CI 3.18 to 7.16, with a fixed-effects model; WMD 5.17, 95% CI 3.18 to 7.16, with a random-effects model; I² = 0%; Fig. 6). TSAs of five comparisons indicated that the cumulative z curve crossed the conventional boundary, but did not cross the trial sequential monitoring boundary.

However, when the steroid and S&NSI groups were added, there were no significant differences in eGFR changes in immunosuppressive treatment compared to controls (seven trials [16, 20, 25, 28, 29, 32, 33], 998 patients, WMD 0.26, 95% CI − 0.03 to 0.56, with a fixed-effects model; WMD 2.52, 95% CI − 0.49 to 0.53, with a random-effects model; I² = 76%; Fig. 6). TSAs of seven comparisons indicated that the cumulative z curve did not cross the conventional boundary or the line of required information size.

There was a lower risk of reaching ESRD in the immunosuppressive treatment group than in controls (12 trials [13, 17, 19, 24‐28, 31, 35, 37, 38], 1031 patients; RR 0.51, 95% CI 0.33 to 0.08, with a fixed-effects model; RR 0.55, 95% CI 0.33–0.90, with a random-effects model; I² = 8; Fig. 6). These analyses were dominated by the steroid treatment group (Fig. 7).

TSAs of steroids indicated that the cumulative z curve crossed both the conventional boundary and the trial sequential monitoring boundary.

A total of 20 articles reported adverse events during the observation period. The types of adverse events varied widely, and included infection, cardiovascular disease, respiratory disease, hepatotoxicity, and many others; the 13 most commonly reported are listed in Table 2. As the number of infections reported in Rauen [28] was greater than the total number, RR could not be calculated for infections. TSAs of infection, gastrointestinal disease, hematological disease, dermatological disease, impaired glucose tolerance or diabetes mellitus, and hyperkalemia indicated that the cumulative z curve crossed the conventional boundary but did not cross the trial sequential monitoring boundary. In addition, TSAs of the other seven diseases indicated that the cumulative z curve did not cross the conventional boundary or the line of required information size.

Previous studies have suggested that treatment with steroids or alkylating agents can significantly reduce proteinuria levels in patients with IgAN [42‐44]. Our meta-analysis also showed that immunosuppressive agents can significantly reduce the level of proteinuria. The levels of proteinuria in groups treated with steroids, NSI, or S&NSI were significantly reduced compared to controls. The heterogeneity of the steroid group was mainly derived from Lai [23], in which the inclusion criterion included nephrotic syndrome. In addition, the heterogeneity of the S&NSI group was mainly derived from Yoshikawa [18], in which the inclusion criterion included age < 15 years. Sequential analyses showed that immunosuppressive agents were effective for relieving proteinuria, and no additional sample size was required.

Our results suggest that non-steroidal immunosuppressive therapy may have a positive effect on eGFR. However, sequential analyses suggested that this is still inconclusive and further studies are required for confirmation. In addition, the treatment group showed a greater reduction in the risk for ESRD than the control group, and this effect was mainly due to the steroid treatment group. Sequential analyses showed that steroids could reduce the risk for ESRD without the need for a larger sample size. A relevant study [43] also suggested that high-dose short-course steroid therapy has a significant protective effect on renal function, while a low-dose long-course of steroids does not. Further studies are required to determine whether NSI or S&NSI can reduce the risk for ESRD.

The use of immunosuppressive agents is often accompanied by side effects. The immunosuppressive therapy group showed significant increases in gastrointestinal, hematological, dermatological, and genitourinary side effects, as well as impaired glucose tolerance or diabetes in this meta-analysis. As the number of infection events reported in the STOP study was too high, even exceeding the total number of patients, it was not possible to calculate the RR value. However, across all studies, the proportion of infections reported was still higher in the immunosuppressive therapy group than in controls. In addition, the TESTING study had to be discontinued because of the excessive number of serious adverse events, mostly infections. By contrast, hyperkalemia was more common in the control group, which may have been related to the application of ACEI and ARB. However, it should be noted that sequential analyses indicated that the statistical results of the above adverse events should be verified by further experiments. Although there were more adverse events in the immunosuppressant group, immunosuppressive agents should be used when necessary because they significantly reduce the risk of ESRD (which means fewer dialysis and transplantation).

Our study had several limitations that should be taken into consideration. The results of bias analyses indicated that nearly half of the studies did not explicitly report the methods used for randomization. In addition, few studies used blinded methodologies. The quality of the reports in the literature is unsatisfactory. In addition, there were some differences in the inclusion criteria between each study, such as age, proteinuria level, and renal function, and these confounding factors led to a high degree of data heterogeneity. Our results show that glucocorticoids therapy has no significant effect on serum creatinine or eGFR in patients with IgA nephropathy. However, because chronic administration of glucocorticoids significant muscle loss and endogenous creatinine production can ocurr, possible errors in estimation of GFR using serum creatinine based formulas [45] may have led to an over optimistic conclusion.