Does quality of observational studies affect the results of a meta-analysis?: the case of cigarette smoking and pancreatic cancer

Objectives: A meta-analysis of observational studies on association between cigarette smoking and pancreatic cancer was performed to focus, particularly, on the role of the studies' quality in affecting meta-analysis results. Methods: A bibliographic search was carried out on PubMed and EMBASE databases until February 15, 2008. Key words were "pancreatic neoplasms," "pancreatic cancer," "smoking," "smoke," "cigarette," "case-control studies," and "cohort studies." Studies about cigarette smoking and pancreatic cancer were selected and assessed on quality. STATA 9.0 was used. Results: Six cohort studies and 24 case-control studies were selected, with median quality scores of 8 (range, 3) and 10 (range, 8), respectively. Pooled case-control studies' odds ratio (OR) and cohort studies' risk ratio were, respectively, 1.45 (95% confidence interval [CI], 1.33-1.57) and 1.78 (95% CI, 1.64-1.92). After stratifying for quality scoring, high-quality-scored case-control studies yielded an OR of 1.38 (95% CI, 1.27-1.49), whereas the others gave an OR of 1.52 (95% CI, 1.34-1.73). The results of meta-analysis for cohort studies showed a risk ratio of 1.74 (95% CI, 1.61-1.90) and of 2.10 (95% CI, 1.64-2.67), respectively, for high- and low-quality score studies. Conclusions: There is evidence that cigarette smoking is an important risk factor for pancreatic cancer, but the estimate of the association greatly relies on the studies' quality.