Clinical Interventions and All-Cause Mortality of Patients with Chronic Kidney Disease: An Umbrella Systematic Review of Meta-Analyses

Patients with chronic kidney disease (CKD) have altered physiologic processes, which result in different treatment outcomes compared with the general population. We aimed to systematically evaluate the efficacy of clinical interventions in reducing mortality of patients with CKD. We searched PubMed, MEDLINE, Embase, and Cochrane Database of Systematic Reviews for meta-analyses of randomized controlled trials (RCT) or observational studies (OS) studying the effect of treatment on all-cause mortality of patients with CKD. The credibility assessment was based on the random-effects summary estimate, heterogeneity, 95% prediction intervals, small study effects, excess significance, and credibility ceilings. Ninety-two articles yielded 130 unique meta-analyses. Convincing evidence from OSs supported mortality reduction with three treatments: angiotensin-converting-enzyme inhibitors or angiotensin II receptor blockers for patients not undergoing dialysis, warfarin for patients with atrial fibrillation not undergoing dialysis, and (at short-term) percutaneous coronary intervention compared to coronary artery bypass grafting for dialysis patients. Two treatment comparisons were supported by highly credible evidence from RCTs in terms of all-cause mortality. These were high-flux hemodialysis (HD) versus low-flux HD as a maintenance HD method and statin versus less statin or placebo for patients not undergoing dialysis. Most significant associations identified in OSs failed to be replicated in RCTs. Associations of high credibility from RCTs were in line with current guidelines. Given the heterogeneity of CKD, it seems hard to assume mortality reductions based on findings from OSs.


Details of data analytic methods
. PRISMA Checklist Table S2. Details of meta-analyses of observational studies associating clinical intervention and all-cause mortality of patients with chronic kidney disease graded as suggestive evidence, weak evidence, or not significant Table S3. Details of meta-analyses of randomized controlled trials associating clinical intervention and all-cause mortality of patients with chronic kidney disease, having p-value>0.05 Table S4. Details of credibility assessment in meta-analyses of observational studies associating clinical intervention and all-cause mortality of patients with chronic kidney disease Table S5. Details of credibility assessment in meta-analyses of randomized controlled trials associating clinical intervention and all-cause mortality of patients with chronic kidney disease Table S6. Details of eligible meta-analysis unique in design but ineligible for re-analysis Table S7. Comparisons of effect of treatment on all-cause mortality between evidences from different chronic kidney disease stages Table S8. Sensitivity subset analysis of prospective studies only of evidence from observational studies graded as convincing or highly suggestive evidence

Assessment of heterogeneity
We performed Cochran's Q test and calculated the I2 statistic for evaluation of heterogeneity 1,2. I2 ranges from 0% to 100% and describes the percentage of variability in a study estimate that is due to between-study heterogeneity. I2 > 50% was regarded as large heterogeneity. Significant heterogeneity indicates presence of genuine heterogeneity or bias.

Estimation of the prediction interval
We estimated the 95% prediction interval, which is the range where a true effect of the intervention is to be expected for 95% of similar studies in the future 3. While the summary effects of random-effects meta-analysis represent the average effect of included studies, prediction interval estimates the treatment effect of individual studies in future settings 4. For example, a 95% prediction interval of risk ratio = (2 to 4) implies that 95% of future studies are expected to show a risk ratio between 2 and 4. Prediction intervals centers around random effects summary estimate, similar to confidence intervals. 95% prediction intervals corresponds to 95% confidence intervals when there is no in-between study heterogeneity and gets wider as in-between study heterogeneity increases. Prediction intervals including the null value suggests there may be settings where the intervention effect is null or even in the opposite direction and requires further study for identification of the causes of heterogeneity. 95% prediction interval excluding the null suggests that the treatment effect is beneficial in at least 95% of the future studies and concludes that results of treatment effects are consistent, even when some between-study heterogeneity is present.

Assessment of small study effects
We assessed small study effects, i.e. large studies having more conservative results than smaller studies, with the regression asymmetry test proposed by Egger, et al 5. Small-study effects were claimed at Egger p value < 0.1 with the effect of the largest study (the study with the smallest standard error) showing more conservative result than the summary effect of the metaanalysis under random model. Presence of small study indicates publication bias, selective reporting, or genuine heterogeneity 6.

Assessment of excess significance bias
We performed a test for excess significance to evaluate whether the number of studies reporting nominally significant results (p value < 0.05) is greater compared to the expected number of statistically significant studies 7. We assumed that the effect size of the largest study in a metaanalysis was plausible effect size of the individual studies 8. The expected probability that an individual study is statistically significant was assumed to be the power of the largest study at type I error rate = 0.05. Statistic A was calculated by the following χ2 statistic: where O is the number of observed statistically significant studies, E is the expected number of statistically significant studies, and N is the total number of individual studies. Excess significance was claimed at p value < 0.1 with the number of observed significant studies larger than the number of expected significant studies. Presence of excess significance indicates publication bias, selective analysis, or outcome reporting bias. Specify study characteristics (e.g., PICOS, length of follow-up) and report characteristics (e.g., years considered, language, publication status) used as criteria for eligibility, giving rationale.

2-3
Information sources 7 Describe all information sources (e.g., databases with dates of coverage, contact with study authors to identify additional studies) in the search and date last searched.

2-3
Search 8 Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated.

2-3, supplementary material
Study selection 9 State the process for selecting studies (i.e., screening, eligibility, included in systematic review, and, if applicable, included in the meta-analysis).

2-3
Data collection process 10 Describe method of data extraction from reports (e.g., piloted forms, independently, in duplicate) and any processes for obtaining and confirming data from investigators.

Data items 11
List and define all variables for which data were sought (e.g., PICOS, funding sources) and any assumptions and simplifications made.
3 Risk of bias in individual studies 12 Describe methods used for assessing risk of bias of individual studies (including specification of whether this was done at the study or outcome level), and how this information is to be used in any data synthesis.

3-4
Synthesis of results 14 Describe the methods of handling data and combining results of studies, if done, including measures of consistency (e.g., I 2 ) for each meta-analysis.

Study selection
17 Give numbers of studies screened, assessed for eligibility, and included in the review, with reasons for exclusions at each stage, ideally with a flow diagram.

15-16
Limitations 25 Discuss limitations at study and outcome level (e.g., risk of bias), and at review-level (e.g., incomplete retrieval of identified research, reporting bias).

16-17
Conclusions 26 Provide a general interpretation of the results in the context of other evidence, and implications for future research.