Trends in the quality of abstracts of articles describing randomized controlled cancer trials (RCTs) from 1975–2004

6541 Background: Busy oncologists may rely on abstracts to obtain most of their information from published reports. Here we evaluate the quality of abstracts in articles reporting RCTs, and whether quality of reporting has changed over time. Methods: Reports of RCTs evaluating systemic therapy for breast, colorectal and non-small cell lung cancer published 1975–2004 in 6 major journals were identified. The quality of abstracts was evaluated using a scale which awards 1 point for full and 1/2 point for partial reporting of each of 6 items: study rationale, statement of primary endpoint, description of intervention, treatment effect size, P-values or confidence intervals (CI), and toxicity. The scale was a modification of one previously developed to assess quality of reporting of conference abstracts. Trends in the reporting of individual items and composite scores were evaluated over three 10-year periods (1975–84, 1985–94, 1995–2004) using descriptive statistics. Results: Of 326 eligible articles, 319 had an abstract. Unstructured abstracts were common before 1994 ( Table ). The minimum and maximum quality scores assigned were 0.5 and 6/6. Median quality scores improved significantly with time (from 2.5 to 4.5/6, P < .001), as did inclusion of each of the individual items (P < .001). In the most recent 10-year period, three variables were adequately reported in less than half of abstracts (primary endpoint 38.0%, rationale 38.6%, toxicity 40.4%). Since the publication of the revised CONSORT statement in 2001, reporting of these items has improved (58.2%, 47.8%, 44.8% for 2002–2004). Conclusions: Quality of abstracts for published RCTs has improved over time, but deficiencies still exist. Items requiring improved reporting include identification of the primary endpoint, description of rationale, and a summary of toxicity. No significant financial relationships to disclose. [Table: see text]