The importance of full text screening when judging eligibility criteria in a systematic review

Abstract

Aim: To evaluate if statistically significant results are more likely to be reported in title/abstracts compared to non-significant outcomes. Methods: In this methodological survey, we reanalyzed 59 observational studies from a previous systematic review. The PECO question was: Patient (P): children with primary teeth; Exposure (E): low birth weight and/or preterm; Comparison (C): normal birth weight and/or full-term; Outcome (O): dental caries. We analyzed the presence of key terms in the titles and abstracts, such as gestational age, preterm, full-term, birth weight, low/normal birth weight. Full texts were analyzed for “positive outcomes” (statistically significant association, p < 0.05 or 95% CI not crossing the null effect line) related to the association between the outcome and the exposure; and “negative outcomes” (when the outcome had statistically similar occurrence between the exposure and the comparison group). The odds ratio (OR) was calculated between the presence of key terms in titles/abstracts and the type of outcome (positive or negative). Results: Of 59 studies, 66% cited the key terms in titles/abstracts, and 75% reported negative outcomes. Studies with positive outcomes were more likely to report key terms in the titles/abstracts compared to studies with negative outcomes (OR: 4.5; 95% CI: 0.9-22.4; Chi-square test: p = 0.06). Studies with statistically significant outcomes, favoring the exposure or the comparison, were more likely to report these data in the titles/abstracts. Conclusion: When conducting a systematic review, the final decision related to the inclusion of a study must be based on a full-text level.