Correction: Limited Clinical Utility of Remote Ischemic Conditioning in Renal Transplantation: A Meta-Analysis of Randomized Controlled Trials

[This corrects the article DOI: 10.1371/journal.pone.0170729.].

Partial data from the study by MacAllister et al. (2015) was incorrectly extracted, which augmented the number of events including delayed graft function (DGF), acute rejection (AR), graft loss, and 50% fall in SCr. The authors have carefully re-extracted all the four inappropriate outcomes data and re-analyzed all involved results in the meta-analysis. The details of the modifications are listed below.
The incidence of DGF is set as the primary endpoint in the study. After reanalysis with the re-extracted data, there are some modifications in the text of the Abstract, Discussion, and Results sections with corrections shown here: In the Abstract and Discussion sections, "random-effects model: RR = 0.89; fixed-effect model: RR = 0.84" is corrected to "random-effects model: RR = 0.86; fixed-effect model: The correct sentence regarding the incidence of DGF in the "Study outcomes" section of the Results is: "After pooling data from all six trials, a trend of decline in the risk of DGF could be observed in the RIC group than the control group both under random-effects (RR = 0.86; Fig 2) and fixed-effect models (RR = 0.81; S1 Fig). However, unfortunately, the decrease did not demonstrate statistical significance under both models (random-effects model: 95% CI, 0.61-1.22, P = 0.41; fixed effect model: 95% CI, 0.57-1.13, P = 0.22;)".
In the fifth paragraph of the Discussion section, the needed sample size was also calculated based on the DGF rate. Thus, the revised sentence is: "Based on the DGF rate calculated in this meta-analysis (RIC group: 13.5%, Control group: 17.2%; relative risk reduction: 21.5%), a total of 2980 recipients are needed to draw a relatively stable statistical significant result in terms of DGF incidence".
Moreover, the secondary endpoints including the incidence of AR, graft loss and 50% fall in SCr are also modified. The correct sentence in the Results section entitled "Incidence of AR, graft loss, 50% fall in SCr" is: "No significant difference was detected in the incidence of AR between two groups (RIC vs. Control groups: RR = 0.87; 95% CI, 0.47-1.61; P = 0.65; Fig 3A). The incidence of graft loss within three months was documented in four studies, and no significant difference was observed between groups (RIC vs. Control groups: RR = 1.36; 95% CI, 0.45-4.10; P = 0.58; Fig 3B). With regard to the rate of 50% fall in SCr within 24h post-transplantation,  Fig 3C) was yielded after pooling extracted from three RCTs".
The Results text in the section "Hospital Stay" is also modified. The correct sentence is: "Among six included studies, four reported mean length of hospital stay [15,16,19,20]. There was no statistically significant difference in the hospital stay between two groups (WMD = -0.70; 95% CI, -1.53-0.14; P = 0.10; Fig 4C)".
Please see the corrected Figs 2 and 3 (random-effects models) and Table 2 below. The corrected versions of S2 Table (raw data)  It is important to note that the above changes do not affect the overall conclusions of the original article.
To elucidate the conclusions of the study more accurately, the conclusions are reiterated as follows: In summary, the present meta-analysis suggests that perioperative pre-RIP, per-RIP or post-RIP procedures showed no statistically significant protective impact on kidney transplant functions. Consequently, the clinical utility of RIP remains limited. Further well-designed RCTs with large sample size would need to be performed to document the efficacy and safety of RIC in renal transplantation. With regard to the methodological information of the article, the following should be clarified: 1) To avoid a possible effect of heterogeneity, the authors had also conducted a subgroup analysis based on different conditioning types, and the results remain unchanged; 2) If continuous data are presented as median and interquartile range (IQR), median was utilized to estimate mean when the sample size is larger than 25 [3], and interquartile range (IQR)/1.35 was treated as an estimator for standard deviation [4]; 3) In this meta-analysis both random-effects models and fixed-effect models were adopted. The authors mainly presented the results yielded under random effects model in the main text, and in the conclusions, while the fixed effect model was used as sensitivity analysis solely. Though the comparison could not be measurable, obvious differences between the two models could not be observed; 4) In the original article, one doublezero study (i.e., no AR events in both RIC and control patients) could be detected when the incidence of AR was evaluated, and the information from this study was not included in the summary effect for AR. The authors re-analyzed the outcomes of AR using SAS (Version 9.2) PROC NLMIXED with a beta-binomial model (the codes were obtained from supplementary of Furthermore, the sample size of the above double-zero study is also small. Thus, the original synthetic outcomes of AR could be acceptable after excluding the double-zero study; 5) The authors did not have a registered study protocol in a publicly accessible database.
A difference in DGF incidence could be observed between studies. This could be attributed to the following reasons: 1) Different definition of DGF; 2) Different study characteristics; and 3) Different renal allograft source: DBD, DCD or living donor.