Truth Survival in Clinical Research: An Evidence-Based Requiem?

Abstract

Purpose: Factors associated with the survival of truth of clinical conclusions in the medical literature are unknown. The authors hypothesized that conclusions derived from studies using better methodology should have a longer half-life.

Data Sources: MEDLINE and hand searches of journals with studies on cirrhosis and hepatitis.

Study Selection: Original articles and meta-analyses published from 1945 to 1999 about cirrhosis or hepatitis in adults.

Data Synthesis: In 2000, 285 of 474 conclusions (60%) were still considered to be true, 91 (19%) were considered to be obsolete, and 98 (21%) were considered to be false. The half-life of truth was 45 years. The 20-year survival of conclusions derived from meta-analysis was lower (57% ± 10%) than that from nonrandomized studies (87% ± 2%) (P < 0.001) or randomized trials (85% ± 3%) (P < 0.001). The survival of conclusions was not different when studies of high methodologic quality were compared with those of low quality. In randomized trials, the 50-year survival rate was higher for 52 negative conclusions (68% ± 13%) than for 118 positive conclusions (14% ± 4%) (P < 0.001).

Conclusions: Contrary to the authors' hypothesis, conclusions based on recognized, good methodology had no clear survival advantage. To better convince clinicians of the long-term utility of evidence-based medicine, better prognostic factors should be developed.