TY - JOUR
T1 - Selecting controls is not selecting "normals"
T2 - Design and analysis issues for studying the etiology of polycystic ovary syndrome
AU - Bloom, Michael S.
AU - Schisterman, Enrique F.
AU - Hediger, Mary L.
PY - 2006/7
Y1 - 2006/7
N2 - Introduction: Polycystic ovary syndrome (PCOS) is a highly prevalent disorder among postmenarcheal premenopausal women, accounting for much of anovulatory infertility, and it is associated with several comorbidities. The natural history of PCOS, as well as hypotheses regarding its etiology, facilitates employment of the case-control study design. However, a review of the literature suggested that inconsistency of research findings in relation to the etiology of PCOS might be, in part, due to the different choice of controls for case-control design and application. Design: Investigators have often employed "healthy" control groups and neglected to consider the tenability of the "rare disease" assumption in the analysis when instituting the case-control strategy. This might result in the introduction of a positive bias or overestimation of odds ratios, producing an effect estimate that is more extreme than that in the underlying population. Using several quantitative, though hypothetical, literature-driven examples, this bias is described and demonstrated. In addition, recommendations are provided with regard to case-sampling strategy when the rare disease assumption is untenable, as it may frequently be in studies of PCOS. Conclusion(s): It is hoped that more consistent case-control methodology, in concert with recent consensus on case definition, will more effectively facilitate the elucidation of the causes and consequences of PCOS.
AB - Introduction: Polycystic ovary syndrome (PCOS) is a highly prevalent disorder among postmenarcheal premenopausal women, accounting for much of anovulatory infertility, and it is associated with several comorbidities. The natural history of PCOS, as well as hypotheses regarding its etiology, facilitates employment of the case-control study design. However, a review of the literature suggested that inconsistency of research findings in relation to the etiology of PCOS might be, in part, due to the different choice of controls for case-control design and application. Design: Investigators have often employed "healthy" control groups and neglected to consider the tenability of the "rare disease" assumption in the analysis when instituting the case-control strategy. This might result in the introduction of a positive bias or overestimation of odds ratios, producing an effect estimate that is more extreme than that in the underlying population. Using several quantitative, though hypothetical, literature-driven examples, this bias is described and demonstrated. In addition, recommendations are provided with regard to case-sampling strategy when the rare disease assumption is untenable, as it may frequently be in studies of PCOS. Conclusion(s): It is hoped that more consistent case-control methodology, in concert with recent consensus on case definition, will more effectively facilitate the elucidation of the causes and consequences of PCOS.
KW - Bias
KW - case-control study
KW - collider stratification
KW - methods
KW - odds ratio
KW - PCOS
KW - polycystic ovary syndrome
KW - prevalence
UR - http://www.scopus.com/inward/record.url?scp=33745514680&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=33745514680&partnerID=8YFLogxK
U2 - 10.1016/j.fertnstert.2005.09.066
DO - 10.1016/j.fertnstert.2005.09.066
M3 - Article
C2 - 16750830
AN - SCOPUS:33745514680
SN - 0015-0282
VL - 86
SP - 1
EP - 12
JO - Fertility and Sterility
JF - Fertility and Sterility
IS - 1
ER -