Jessica Dennis
MSc Student, Epidemiology and
Community Medicine
McLaughlin Centre for Population
Health Risk Assessment
University of Ottawa
dennis.jessica@gmail.com

Bio>

Go to lecture>

The case-only study design is an efficient design for assessing gene-environment and gene-gene interactions.  However there are concerns about its validity.  This presentation summarized the extent and sources of bias in the case-only design, evaluated by means of a systematic review and meta-regression analysis.  Studies that assessed bias in the case-only design applied to the study of genetic interaction were identified through database searches.  Qualitative comments on the extent and sources of bias were extracted.  A meta-regression analysis of the ratio of the case-control and case-only interaction odds ratios was conducted based on studies in which both methods were applied to the same dataset.  The search yielded 365 unique articles, of which 38 met the final inclusion criteria.  Potential sources of bias in the case-only design included non-independence of genotype and exposure in the source population, violations of the rare disease assumption, and misclassification of genotype and/or exposure.  Included in the meta-regression analysis were 25 evaluations from 21 studies.  The mean of the ratio was 1.06 (95% CI, 0.93-1.21), indicating that the case-control and case-only designs gave similar estimates of interaction.  The I2 statistic indicated that 20.7% of the between-study variation was due to true heterogeneity, which was not explained by any methodological characteristics of the included studies.  In conclusion, although theoretically possible, bias in the case-only design does not appear to be common in practice.  Previous criticisms of the case-only design may have been over-stated and the design may be of considerable value.