Loss to Analysis—a Problem in CKD Trials

Randomized trials of treatment for chronic kidney disease (CKD) have important quality shortcomings—including a high rate of loss of patients from the analysis, according to a study in the American Journal of Kidney Diseases.

The researchers performed a systematic evaluation of loss to analysis for primary outcomes of randomized controlled trials of patients with CKD undergoing dialysis or kidney transplantation. The analysis included 196 trials published in 2007 and 2008. Studies in which not all randomized patients were included in the primary outcome analysis were considered to have loss to analysis.

Twenty-seven percent of the trials specified no clear primary outcome. Five percent did not report numbers of patients randomized and analyzed, while 12 percent used time-to-event analysis. Of the remaining 110 studies, 58 percent had some loss to analysis. The median loss to analysis was 10 percent, with a range of one to 41 percent.

Fifty-four percent of trial reports said that analysis was by intention to treat. Yet 44 percent of studies making this claim did not include all randomized patients in the analysis. Imputation of missing data was reported by five percent of studies. Studies without loss to analysis tended to have smaller sample sizes: 128 versus 229.

Randomized trials of treatment for CKD pose unique challenges. Based on the new review, many CKD studies do not meet current standards for clinical trial reporting. Many trials do not specify a primary outcome of interest; those which do have high rates of data loss to analysis. Efforts to improve the quality of CKD randomized trials should include increased attention to transparency and reporting loss to analysis. [Deo A, et al: Loss to analysis in randomized controlled trials in CKD. Am J Kidney Dis 2011; 58: 349–355].


October-November 2011 (Vol. 3, Number 10 & 11)