Adjusting to Imbalance: When to Statistically Adjust for Differences Between Treatment Groups in Clinical Studies

Abstract

Background/Objective: Surgeons conducting randomized trials are sometimes left with the dilemma of unbalanced characteristics between the treatment and control groups. It remains debatable how baseline differences between treatment and control groups should be handled in surgical studies. Some investigators advocate ignoring the imbalanced variable whiles others believe an “adjusted analysis” should be performed, correcting for the variable imbalanced between groups. We reviewed the rationale and conduct of adjusted and unadjusted analyses in surgical clinical trials. Methods: We conducted computerized and hand searches to identify published surgical randomized controlled trials in the British Medical Journal (BMJ), Journal of the American Medical Association (JAMA), New England Journal of Medicine, The Lancet, Journal of Bone and Joint Surgery (JBJS - American Volume), and Journal of Bone and Joint Surgery (JBJS - British Volume) between January 2000 and April 2003. Three reviewers abstracted information about imbalances in baseline variables and variable adjustment. Discrepancies were resolved by consensus. Results: We identified 72 randomized trials. Studies presented an average of 10.3 ± 7.7 baseline variables. Fifteen trials (20.8 %) reported imbalances in baseline characteristics of their study populations. Twenty-three trials (31.9 %) reported both unadjusted and covariate-adjusted results but unadjusted analyses received more emphasis in 18 trials (78.3 %). The studies' conclusions were changed in 3 trials (13 %) when an adjusted analysis was conducted. Conclusions: Our review has identified important problems with the reporting and rationale for adjusting for imbalances in patient groups in surgical clinical trials. Investigators conducting clinical comparative studies should endeavor to report the rationale for conducting adjusted analyses of their data. In the absence of such information, readers should rely more on the simple unadjusted results of a study.

Key words

randomized trials - baseline variables - adjustment - methodology - review

References

• 1 Assman S, Pocock S, Enos L, Kasten L. Subgroup analysis and other (mis) uses of baseline data in clinical trials.  Lancet. 2000;  355 1064-1069
• 2 Parker M J, Khan J K, Crawford J, Pryor G A. Hemiarthroplasty versus internal fixation for displaced intracapsular hip fractures in the eldery.  J Bone Joint Surg [Br]. 2002;  84 1150-1155
• 3 Pocock S, Assman S, Enos L, Kasten L. Subgroup analysis, covariate adjustment and baseline comparisons in clinical trial reporting: current practice and problems.  Stats Med. 2002;  21 2917-2930