Potential advantages of combination second-line drug therapy in rheumatoid arthritis (RA) may not be detectable in standard randomized controlled clinical trials, despite excellent design and performance. This results from intrinsic limitations of usual clinical studies, such as a short time frame, patient selection, and insufficient numbers of patients. Examples of selected clinical protocols in rheumatic diseases that illustrate these problems are presented. In systemic lupus erythematosus, many well-designed and -performed trials comparing combinations of cytotoxic drugs and corticosteroids with corticosteroids alone were inconclusive. However, combination therapy was superior in three types of studies: a 15-year clinical trial, a pooled analysis of multiple trials, and longitudinal databases of unselected patients. In RA, a 48-week study indicated no differences in results of treatment with auranofin, methotrexate, and the combination of these two drugs. In contrast, a clinical database from 15 rheumatology practices indicates that methotrexate was continued for 5 years by more than 50% of patients, compared with fewer than 10% for auranofin. A subset from the clinical database of courses of second-line drugs over 1 year (rather than 5 years), for only the initial course of a second-line drug (rather than any course), was examined. Using these selected data, which mimic the conditions of the clinical trial, no significant differences were seen in continuation of methotrexate versus auranofin. This observation suggests that intrinsic limitations, including patient selection, insufficient patient numbers, and a short time frame, may render it difficult (or impossible) to document the efficacy of combination therapy in RA using standard randomized controlled clinical trials.