In the News: Sub group analyses in Nephrology trials

A subgroup analyses is an evaluation of the treatment effect in subgroups of patients defined by baseline characteristics. The main purpose of doing these analyses is to provide a better understanding of the treatment effect heterogeneity. Often, these analyses tend to deviate from this purpose and mislead the reader and may potentially have adverse clinical practice implications.

To demonstrate fallacy of the subgroup analyses, a recent paper in CJASN by Fishbane etal conducted a thorough literature search of the nephrology RCT’s published between July1, 2010 and June 30,2011(included review of nephrology and transplantation journals and four general medicine journals-NEJM, Lancet, Annals of Internal Medicine and Archives of Internal Medicine) The authors found that approximately one third (37.3%) of the published nephrology RCT’s reported subgroup analyses. They found major deficiencies in reporting in the nephrology trials.

1. The majority failed to pre specify that a subgroup analyses would be performed (77.4%). Failure to prespecify diminishes the credibility of the results. 

2. There was an almost uniform failure to list all subgroup variables analyzed. Failure to do so suggests the possibility that numerous analyses may have been conducted in an effort to find a positive result.

3. Statistical testing was inappropriate, with a test of interaction reported in only 35.3% of subgroup analyses. 

4. It is very common to claim treatment effects based on separate tests for each level of subgroup. In addition, when subgroup analyses were discussed, there was a frequent failure to note the limitations of the analyses and discuss that the results should be viewed with caution. Positive claims are common in discussion sections, with general over interpretation of the importance of the subgroup results.

Fishbane et al summarize their recommendations for reporting of subgroup analyses. Subgroup analyses may be helpful for suggesting heterogeneous effects of treatments, but the results may be misleading and over interpreted. 

Post by Dr. Ashish Kataria, Fellow in Nephrology.