Confounding by indication: implications for implant research

When studying potential adverse effects of medical and surgical treatments, including prosthetic implants, it is essential to be able to isolate the effect of the treatment from that of the underlying illness in order to draw conclusions about cause and effect. Treatment involving surgical implants is typically not assigned randomly, but stems from a clinical indication related to underlying medical problems. When undertaking a research study comparing implant recipients to other groups ofsubjects (who often do not have the same underlying problems), it is of critical importance to account for these inherent differences between the study groups. Confounding by indication is the bias that arises in the observational study of treatment effects, which stems from the differing baseline risks, comorbidities, and prognostic factors between patients who receive treatment and those who do not. In this article, we examine several observational epidemiologic studies that have noted associations between prosthetic implants and subsequent connective tissue disease. However, it can be demonstrated that none of these studies has been able to adequately account for confounding by indication, resulting in findings that are likely biased. Because of the increasingly widespread use of implants at younger ages and the growing importance of studying their potential long-term effects, a thorough understanding of this bias is necessary in order to properly conduct and interpret implant research studies.