Identification of trial reports
The Specialized Register is a register of all known reports of randomized trials in nephrology organized by trial. It contains reports identified from monthly searches of the Cochrane Central Register of Controlled Trials CENTRAL, weekly searches of MEDLINE OVID SP and searches of the current year of Embase OVID SP, and has been used in multiple published systematic reviews of immunosuppression. Hand searching of renal-related journals and the proceedings of major renal conferences, weekly current awareness alerts for selected renal journals and searches of the International Clinical Trials Register (ICTRP) Search Portal and ClinicalTrials.gov are also incorporated.
We judged that from the year 2000 trials should be reported in accordance with international guidelines including the consolidated standards of reporting trials (CONSORT) statement (http://www.consort-statement.org/) and International Committee of Medical Journal Editors' uniform requirements for manuscripts (http://www.icmje.org/).
Using a standardized data abstraction spread sheet, two reviewers independently examined each trial report and recorded key descriptive data. We collected details of year and journal of publication, drug-class comparison, trial location, phase, size, duration of follow-up and sponsorship, and journal impact factor and CONSORT endorsement at date of publication. Journal CONSORT endorsement required adoption of the statement at least 6 months before trial publication so as to allow for lag between manuscript submission and publication . We contacted journal editorial offices to determine date of CONSORT adoption. Where no reply was obtained we reviewed previous journal issues to determine author guidelines.
We recorded absolute measures of function: patient death and time to death, graft loss and time to loss, and measures of relative function; GFR measured directly, or by approximation of GFR with urinary creatinine clearance, or estimated by formulae from serum creatinine (estimated GFR [eGFR]). Relative measures of function are increasingly being used as surrogate markers of absolute end points in kidney transplant trials despite lacking proper validation, and so a selection of both types of measures was made . Where graft loss or time to loss was reported, we determined whether it included recipients who lost their graft and returned to dialysis together with those who died with a functioning transplant, or included only those who lost their graft and returned to dialysis, but not those dying with a functioning transplant (“censoring” for death with a functioning graft). Where GFR was directly measured, we recorded which chemical method was used, and where GFR was estimated, we recorded which formula was used to calculate eGFR.
Where outcomes were reported, we examined completeness of the published data using the outcomes reporting bias in trials (ORBIT) classification system, adapted for our study . Specifically, where outcomes were calculated using time-to-event data, to be graded as providing a complete report, the number in each group at the start of analysis, the proportion surviving at a fixed time point and either the number remaining at risk or a test of difference between the groups had to be stated, or alternatively, a hazard ratio with confidence interval (CI) or p-value presented. For continuous outcomes such as creatinine and eGFR, we expected that the number of participants contributing measurements, the point estimate result (e.g. mean or median) and an estimate of precision (e.g. standard deviation or error) would be stated. For dichotomous outcomes, such as number of patient deaths, we expected the numerator and denominator in each group to be stated. We then assessed the association between the frequency of core outcome reporting and the statistical significance of the results reported in included trials. Specifically, where a core outcome was reported, we determined the proportion of trials where a statistically significant difference between treatment arms was reported for that outcome (either in words “a statistically significant result was found” or quantitatively with a p-value of p < 0.05), and compared this with the proportion of all trials reporting that outcome. We hypothesized that we would find an inverse relationship, with the proportion of trials reporting a statistically significant association increasing with decreasing likelihood of complete data reporting, suggesting outcome reporting bias .
We assessed potential predictors of complete reporting using logistic regression for the outcomes of patient death or time to death, graft loss or time to loss, serum creatinine and eGFR. Potential predictors considered included publication year, journal impact factor (categorized as <3, 3–5.9, ≥6, adjusted for year of publication), journal endorsement of CONSORT guidelines at date of publication, geographical location of trial, number of trial participants (categorized <100, 100–199, ≥200) and funding source of the trial. Analysis was undertaken using STATA version 11.2 (StataCorp, College Station, TX) using the backwards stepwise approach, a likelihood ratio test significance of <0.05 warranting inclusion in the final model.