Background

Many children with steroid-sensitive nephrotic syndrome (SSNS) relapse frequently and receive immunosuppressive agents. In this systematic review of randomized controlled trials (RCTs), the benefits and harms of these immunosuppressive agents are evaluated.

Methods

RCTs with outcome data at six months or more that evaluated noncorticosteroid agents in relapsing SSNS were included. A summary relative risk for relapse at 6 to 12 months was calculated using a random effects model.

Results

Seventeen trials involving 631 children were identified. Cyclophosphamide [3 trials; relative risk (RR) 0.44, 95% confidence interval (CI), 0.26 to 0.73] and chlorambucil (2 trials; RR 0.13, 95% CI, 0.03 to 0.57) significantly reduced the relapse risk at 6 to 12 months compared with prednisone alone. In the single chlorambucil versus cyclophosphamide trial, there was no observed difference in relapse risk at two years (RR 1.31, 95% CI, 0.80 to 2.13). Cyclosporine was as effective as cyclophosphamide (1 trial; RR 1.07, 95% CI, 0.48 to 2.35) and chlorambucil (1 trial; RR 0.82, 95% CI, 0.44 to 1.53), but the effect was not sustained when cyclosporine was ceased. During treatment, levamisole (3 trials; RR 0.60, 95% CI, 0.45 to 0.79) was more effective than steroids alone, but the effect was not sustained.

Conclusions

Cyclophosphamide, chorambucil, cyclosporine, and levamisole reduce the risk of relapse in children with relapsing SSNS compared with prednisone alone. Clinically important differences in efficacy among these agents are possible, and further comparative trials are still needed. Meanwhile, the choice between these agents depends on physician and patient preferences related to therapy duration and complication type and frequency.

METHODS

Literature search

Randomized and quasi-RCTs were identified from MEDLINE (1966 to March 1999), EMBASE (1988 to March 1999), and the Cochrane Controlled Trials Register (Cochrane Library Issue 1, 1999) without language restriction. The optimally sensitive strategies developed for the Cochrane Collaboration for the identification of RCTs were used for MEDLINE¹⁸ and EMBASE searches (abstract; Lefebvre et al, Proceedings of the 4th International Cochrane Colloquium A28, 1996). These strategies were combined with the following medical subject heading and text word terms to limit the search to nephrotic syndrome in children: nephrotic syndrome, nephrosis, nephrosis-lipoid, and child. Reference lists of nephrology textbooks, review articles, relevant trials, and abstracts of scientific meetings were also searched. Investigators with a known interest in this area were contacted for information on any missed or unpublished trials.

Inclusion criteria

All randomized and quasi-randomized trials that evaluated noncorticosteroid agents in the treatment of relapsing NS in children and had outcome data at a minimum of six months were potentially eligible. Trials were included if all participants were aged 3 months to 18 years or if the outcomes of children included in the trial could be identified separately from those of the adults. Trials were eligible if a nonsteroid agent was given in comparison with placebo, steroids alone, no treatment, or another nonsteroid agent. Enrolled participants were required to be steroid responsive at the initial presentation and to have had at least one relapse. Trials enrolling children with congenital NS, steroid resistant NS, or secondary NS were excluded.

The primary outcome measure was the proportion of children who had relapsed at 6 to 12 and 13 to 24 months. Secondary outcome measures sought were mean relapse rate per patient per year, mean time to next relapse, and adverse effects of therapy.

Data extraction and analysis

All titles and abstracts were screened independently by two reviewers (A.D. and E.H.), and irrelevant studies were discarded. The full text of the remaining studies was assessed to determine if the inclusion criteria were met. The included studies were all assessed for trial quality, by the same two reviewers, without blinding to author or source. Any discrepancies in quality assessment were resolved in discussion. The quality items assessed were allocation concealment, intention to treat analysis, completeness of follow-up, and blinding of outcome¹⁹. Data were extracted onto standard preprepared forms. Any required information not available in the article was requested from the investigator(s) involved.

Statistical assessment

For dichotomous outcomes, the relative risks (RR) with 95% confidence intervals (CIs) were calculated in RevMan for individual studies²⁰, and the summary statistics were calculated using a random effects model. The random effects model takes into account the between-study variability as well as the within-study variability. A fixed-effect model was also used to test the robustness of the model chosen and susceptibility to outliers. Statistical significance was tested for using the test Z statistic. To determine whether effect sizes were equivalent in all studies, heterogeneity between studies was tested for using Cochran's Q statistic. The Q statistic has a chi-square distribution and is determined from the sum of the weighted differences between the summary effect measure and the measure of effect from each study. We hypothesized that certain between-study differences in participants (frequently relapsing vs. steroid dependent), interventions (duration of treatment), and trial quality may explain any observed heterogeneity of treatment effects. Examination of these possible between-study differences was attempted by subgroup analysis.

RESULTS

Literature search

Of the 832 titles and abstracts screened, 19 studies were identified by full-text review to be RCTs and to have follow-up data of six months or more. There was no disagreement between the two reviewers regarding the inclusion of trials. Two trials were subsequently excluded. One, assessing Chinese herbs, was excluded because both children and adults were included and the pediatric data could not be separated¹⁰. The other, evaluating levamisole, was in abstract form only, and the primary outcome measure could not be determined from the data included (abstract; Kirubakaran et al, Kidney Int 26:240, 1984). The requested information to allow inclusion of these trials could not be obtained from the authors. Seventeen trials were therefore included in this review.

Study characteristics

The characteristics of the 17 trials are shown in Tables 1 and 2. A total of 631 children was assessed, and the highest number of trials available for any one comparison was three: cyclophosphamide compared with steroid alone (N = 106 children)^21,22,23 and levamisole compared with placebo, steroid alone, or no treatment (N = 138)^12,24,25. Two trials compared azathioprine (N = 60)^9,26, and two trials compared chlorambucil with placebo or steroid alone (N = 41)^27,28. Three trials compared different cyclophosphamide regimes (N = 119)^29,30,31 and a further trial compared different chlorambucil regimes (N = 21)³². There were single trials comparing cyclosporine with cyclophosphamide (N = 55)¹¹, cyclosporine with chlorambucil (N = 40)³³, and cyclophosphamide with chlorambucil (N = 50)⁴. Prednisolone was used in all the trials either in combination with the trial agent or to treat relapses Tables 1 and 2. No eligible RCTs comparing levamisole or azathioprine with other noncorticosteroid agents or comparing Chinese medicines, IgG immunoglobulin, or disodium cromoglycate with corticosteroids or placebo were found.

Table 1 - Characteristics of the studies using cyclophosphamide in the treatment of relapsing nephrotic syndrome in childhood.

Full table

Table 2 - Characteristics of the studies using chlorambucil, levamisole and azathioprine in relapsing nephrotic syndrome in childhood.

Full table

Study quality

The overall study quality was poor Table 3. Only five trials had adequate allocation concealment^4,9,11,12,22, and three trials were double blinded^9,12,27. None of the studies reported an intention to treat analysis. Fourteen trials reported no loss to follow-up at six months, and the other three trials had losses of less than 10%^23,24,30. Four trials did not define relapse^25,27,32,33, and the remaining studies used a variety of definitions.

Table 3 - Quality assessment of the trials included in the meta-analysis of treatments used in relapsing nephrotic syndrome in childhood.

Full table

Study outcomes

Cyclophosphamide resulted in a decreased incidence of relapse at 6 to 12 months Figure 1^21,22,23 compared with prednisolone alone (RR 0.44, 95% CI, 0.26 to 0.73). In 19 children followed up beyond 12 months, the RR for relapse at 13 to 24 months was 0.13 (95% CI, 0.03 to 0.60)^22,23. A single trial showed that an eight-week course of cyclophosphamide resulted in fewer children relapsing within 12 months than a two-week course (RR 0.25, 95% CI, 0.07 to 0.92)²⁹. There was no evidence that prolonging the course from 8 weeks to 12 weeks further reduced the number of children experiencing a relapse (RR 1.04, 95% CI, 0.75 to 1.44)³⁰. A third trial showed that the same total dose of cyclophosphamide given over 6 weeks rather than 12 weeks did not reduce the number of children who relapsed by 12 months (RR 0.43, 95% CI, 0.02 to 9.00), but did increase the numbers experiencing side effects³¹.

Figure 1.

Meta-analyses of the relative risks (RR) and 95% confidence intervals (95% CIs) for relapse of NS by 4 to 12 months in 10 trials of immunosuppressive agents compared with prednisone, placebo, or no treatment in children with relapsing NS. Results shown ordered by trial weights. (A) Cyclophosphamide compared with prednisone. (B) Levamisole compared with prednisone or no treatment. (C) Chlorambucil compared with prednisone. (D) Azathioprine compared with prednisone. No heterogeneity was demonstrated using Cochran's Q statistic (chi-square). The test statistic Z indicates that cyclophosphamide, chlorambucil, and levamisole are significantly more effective in reducing the number of children who relapse compared with prednisone or no treatment, while azathioprine is no more effective than prednisone alone. The numbers in the experimental treatment column and control treatment column indicate the number of children (n) who suffered a relapse of nephrotic syndrome (NS) compared with the total number of children (N) who were treated. The relative risks (RR) with 95% CIs are shown graphically and numerically for individual studies and as a summary estimate (total) for the combined studies. The data were analyzed using the random effects model, which takes into account the between-study variability as well as the within-study variability. Individual effect sizes (weight percentage) are weighted according to the inverse of their variance. Cochran's Q statistic, shown as chi-square analysis with degrees of freedom (d.f.), provides the formal test of heterogeneity between studies. The test statistic Z is the formal test of significant effect. Studies are listed by first author and year of publication. The abbreviations ISKDC (in meta-analyses of cyclophosphamide and azathioprine) and BAPN (in meta-analysis of levamisole) refer to the International Study of Kidney Disease in Children and British Association for Paediatric Nephrology, respectively.

Full figure and legend (47K)

Chlorambucil treatment also reduced the risk for relapse at 12 months (RR 0.13, 95% CI, 0.03 to 0.57) compared with placebo or prednisone alone Figure 1^27,28. The study comparing different chlorambucil regimes found no significant decrease in relapse rates using an increasing dose regime over the stable dose regime (RR 0.18, 95% CI, 0.01 to 3.41), but there was a 34% increase in incidence of leucopenia and an 18% increase in thrombocytopenia with the higher dose³². There was no significant difference between chlorambucil and cyclophosphamide treatment in the risk of relapse at 12 and 24 months (RR at 24 months 1.31, 95% CI, 0.80 to 2.13; additional data provided by Professor J. Brodehl from the trial by the Arbeitsgemeinschaft für Pädiatrische Nephrologie)⁴. Because both agents belong to the same class of drug and because the single direct comparison did not show a statistically significant difference in efficacy, the results of the five trials of alkylating agents versus prednisone alone were combined (RR 0.32, 95% CI, 0.16 to 0.63). In the one trial that evaluated the benefits of alkylating agents in children with frequently relapsing steroid-sensitive NS (SSNS) compared with children who were steroid dependent, chorambucil and cyclophosphamide were more effective in preventing relapse in frequently relapsing children (RR at 24 months 0.34, 95% CI, 0.14 to 0.81)⁴.

Azathioprine did not cause a statistically significant reduction in the number of children who relapsed at six months compared with placebo or steroid alone (RR 0.90, 95% CI, 0.59 to 1.38; Figure 1)^9,26. Levamisole was administered for 4¹², 6²⁵, or 12²⁴ months. During administration, levamisole^12,24,25 was significantly more effective than prednisone alone (RR 0.60, 95% CI, 0.45 to 0.79) with no significant heterogeneity (Q = 0.36; Figure 1). There was no statistically significant benefit (RR 0.72, 95% CI, 0.47 to 1.11) over steroid alone at 6 to 12 months when levamisole treatment had been ceased for 3 months in two trials. However, there was significant heterogeneity of effect (Q = 7.80, P = 0.02), which could be explained by the duration of treatment, suggesting that levamisole is effective during treatment, but the effect is not sustained when treatment is ceased.

In the trial comparing cyclosporine with chlorambucil, cyclosporine was given for 24 weeks³³. At six months, both agents were equally effective in maintaining remission (RR 0.82, 95% CI, 0.44 to 1.53), but at 12 months, cyclosporine was significantly less effective in maintaining remission than chlorambucil (RR 0.47, 95% CI, 0.29 to 0.78). In the trial comparing cyclosporine with cyclophosphamide, cyclosporine was given for 12 months, and the risk for relapse at 9 months was similar for the two drugs (RR 1.07, 95% CI, 0.48 to 2.35)¹¹. At 24 months, cyclophosphamide was superior in maintaining remission (RR 0.40, 95% CI, 0.22 to 0.73).

Side effects were reported in 14 trials; in the other three trials, only the lack of serious infections and leukopenia sufficient to cause cessation of the medication were reported^23,26,29. Both alkylating agents were associated with leukopenia, thrombocytopenia, and infections Table 4. Hair loss was reported uncommonly, and cystitis was not seen with chlorambucil. There were two severe infections reported with cyclophosphamide⁴ and three serious viral infections with chlorambucil, the latter reported with the higher dose regime³². Gum hypertrophy and hirsutism were seen commonly with cyclosporine Table 4; elevated creatinine levels and hypertension occurred in 9 and 4% children^11,33. With levamisole, there was one case of gastrointestinal upset¹², and two of the three trials reported that no side effects occurred^24,25. There was a single case of pulmonary embolus associated with azathioprine treatment⁹.

Table 4 - Adverse effects during treatment of children with steroid-sensitive nephrotic syndrome with cyclophosphamide, chlorambucil or cyclosporine.

Full table

There were insufficient data to assess the mean relapse rate per patient per year and the mean time to next relapse. There were insufficient trials of any treatment combination to allow any subgroup analyses.

DISCUSSION

Randomized controlled trials in children with relapsing SSNS show that cyclophosphamide (2 to 3 mg/kg/day for 8 weeks), chlorambucil (0.2 mg/kg/day for 8 weeks), cyclosporine (6 mg/kg/day), and levamisole (2.5 mg/kg on alternate days) substantially reduce the risk of relapse compared with corticosteroid alone. These interventions typically reduce the risk of relapse by approximately 50% for one to two years during and after a treatment course. This benefit is sustained beyond the on-treatment period for the alkylating agents but only occurs during treatment with levamisole and cyclosporine.

Although these trial data show that immunosuppressive treatment is better than no immunosuppressive treatment, between-agent trials have not demonstrated a clear benefit of one over any other in preventing NS recurrence. The relative efficacy of levamisole is not known. Comparative trials of cyclophosphamide, chorambucil, and cyclosporine have been done, but because of insufficient power, clinically important differences in treatment effects have not been excluded. For example, using the upper and lower bounds of the 95% CIs of the RR estimate obtained from the single comparative trial of chorambucil versus cyclophosphamide, chorambucil could reduce the risk of recurrence by 20% or could double the risk of recurrence compared with cyclophosphamide. Similarly, compared with cyclophosphamide, cyclosporine could reduce the risk of relapse by 50% or could more than double the risk. Adequately powered randomized comparative trials are required to determine which of the four agents is most effective. Because cyclosporine and cyclophosphamide are the two interventions in most widespread use¹⁵, a comparative trial of these two medications would have the most applicability. Assuming a 50% recurrence rate in the cyclosporine-treated group, 130 patients would need to be recruited to a RCT to detect a 50% statistically significant RR reduction for relapse between the two agents. Until then, choice between these agents must be based on other noneffectiveness considerations, such as local availability or licensing, costs, and physician and patient preferences concerning duration of treatment and frequency and nature of complications. By stratifying recruited patients into frequent relapsers and steroid dependent, this trial could also test the hypothesis that alkylating agents are more effective in the frequent relapsing group and cyclosporine is more effective in the steroid dependent group, suggested by post hoc analysis of published trials of alkylating agents⁴ and uncontrolled studies of cyclosporine^34,35.

Our conclusions differ somewhat from recently published guidelines that recommend eight weeks of chlorambucil for frequently relapsing SSNS and 12 weeks of cyclophosphamide for steroid dependent SSNS¹⁵. In part, this may reflect the differences in information sources used. For example, one study that has been influential in shaping guidelines about cyclophosphamide use found that 12 weeks of cyclophosphamide was more effective than eight weeks in preventing relapse in children with steroid dependent SSNS³⁶. However, this study used historical controls, which may be associated with an overestimation of the treatment effect³⁷. In contrast, in a RCT, increasing the duration of cyclophosphamide from 8 to 12 weeks did not improve efficacy³⁰.

What would the benefits and harms be of using an immunosuppressive agent in a child with relapsing NS? Cohort studies show that between 35 and 53% relapse frequently at some time during their disease^2,3. Intervention with immunosuppressive agents would only be undertaken in this group of children whose risk for further recurrences approaches 100% with corticosteroid treatment alone. The meta-analysis shows that the RR for relapse is 0.44 following cyclophosphamide, so the risk for relapse is reduced from 100% to approximately 40%. Hence, on the benefit side of the equation, assuming that all children will relapse, 60 fewer children would relapse for every 100 children treated with cyclophosphamide³⁸. On the harm side of the equation, for every 100 children treated with cyclophosphamide about 1 child will suffer a significant infection, 4 will develop cystitis and 14 lose their hair. Children who relapse only once during the first six months after the initial course of prednisone treatment have only a 10% risk of becoming a frequent relapser³. Thus, only 10 of 100 such children are considered at risk of relapsing frequently. Since cyclophosphamide, if administered to such children would reduce the risk of relapse by 60%, only 6 of 100 children would benefit, while the number suffering adverse effects would be unchanged. Thus, the benefits of treatment would outweigh the harms only in children who relapse frequently.

This study has several potential problems because of the limitations of the primary data. Overall the study quality was poor with only 5 of 17 trials demonstrating adequate allocation concealment. Trials with inadequate allocation concealment can exaggerate the efficacy of the experimental treatment by 30 to 40%³⁹ and meta-analyzes of low quality trials may overestimate the benefit of therapy⁴⁰. This observation makes the need for an adequately powered, well-designed, and reported trial even more necessary. Because trials are not generally designed to evaluate harms of interventions unless the primary outcome is a harm-benefit composite such as death, the small number of serious adverse effects reported here may be an underestimate and may not be directly applicable to larger groups of children treated under nontrial conditions. The effects of publication bias could not be formally assessed because of the small number of studies for each agent. Key investigators in this field who were contacted did not reveal any unpublished data.

In conclusion, this systematic review and meta-analysis of randomized controlled trials show that eight-week courses of cyclophosphamide or chlorambucil and prolonged courses of cyclosporine or levamisole substantially reduce the incidence of relapse in children with NS. Published recommendations generally favor using courses of cyclophosphamide or chlorambucil initially in children with relapsing SSNS^15,16,17. However, using efficacy criteria, there are no data to show that alkylating agents should be preferred over cyclosporine or levamisole. Side effect profiles indicate serious infections, hair loss, and cystitis with the alkylating agents and hypertension and reduced renal function with cyclosporine. Few side effects were reported in the levamisole trials. However, important side effects reported in other studies include neutropenia and disseminated vasculitis^41,42. Thus, the decision as to which medication should be used in a child with frequently relapsing NS will largely depend on patient and physician preference following discussion of the possible side effects and costs of eight-week courses of alkylating agents and those of prolonged courses of cyclosporine or levamisole. Further comparative trials are still needed to identify clinically important differences in efficacy among the immunosuppressive agents in widespread use.

References

References

Acknowledgments

This work was supported by the Australian Kidney Foundation. This work was presented in part at the 36th Annual Scientific Meeting of the Australian and New Zealand Society of Nephrology (Melbourne, 2000). The authors thank Professors Barratt, Brodehl, and Ponticelli, and Dr. Beattie for providing further information about their trials.