Methodological quality in pharmacogenetic studies with binary assessment of treatment response: a review.

Objective: To evaluate the reporting of critical design issues and methods of statistical analysis in pharmacogenetic studies published in the medical literature.

Study Design And Settings: Systematic review of 65 original pharmacogenetic studies published in the literature over the last 15 years.

Results: The sample size determination and the planned sample size were lacking in 63 papers. The study design characterization was lacking in 43 papers. The number of patients analyzed ranged from 36 to 1400 (median=161 and interquartile range of 119-250). The Pearson's χ2 test and the Fisher's test were the most common forms of analysis. Multiple statistical testing was relevant to 59 papers, but only 11 addressed the issue of multiplicity (Bonferroni correction). Sources of multiplicity were multiple association assessment (45 papers), analysis of both genotype and allelic frequencies (44), and multiple analysis methods (unadjusted and adjusted). Hardy-Weinberg equilibrium was tested in 12 of 45 papers performing allelic analysis and was fully reported in four of them. The results of association analyses were commonly reported as P values but rarely as estimates of an association measure (odds ratio or relative risk) and its accuracy.

Conclusions: These results show that there is considerable room for improvement in the current standards of design, analysis, and reporting of pharmacogenetic research.