Experimental Methods: No Adjustment Method Fully Resolves Confounding By Indication

The study appears online in the Journal of Clinical Epidemiology.

According to researchers, conventional methods to adjust for confounding, such as restriction and multivariable regression, leave residual confounding because of unmeasured factors. Propensity scores (PS) and instrumental variable (IV) methods have become increasingly popular with the intent to address residual confounding by simulating a randomized environment.

Using data from the Breast Cancer Effectiveness in Older Women study, a collaboration of investigators from the National Cancer Institute-funded Cancer Research Network, the researchers compared methods used to reduce confounding to estimate incidence rates of breast cancer recurrence in older women who received adjuvant chemotherapy compared with women who did not.

The researchers examined women 65 years of age or older diagnosed with early stage breast cancer. For women classified as high risk for recurrence, 20 percent experienced a breast cancer recurrence. In the unrestricted cohort, receipt of adjuvant chemotherapy was associated with recurrence. The PS distributions among women who received chemotherapy among those who did not showed no substantial overlap. Using the IV method, recurrence yielded a protective estimate. However, imbalances of measured factors across levels of the IV suggested residual confounding.

"With minimal trial-based information available to inform clinical guidelines, which currently offer no guidance for treating older women with cancer, non-randomized studies are vitally important," said lead author Jaclyn Bosco, MPH, project director in the Section of Geriatrics at BUSM. "Non-randomized studies are only reliable when confounding by indication is handled adequately. When treatment with adjuvant chemotherapy among older patients is based on clinical judgment, controlling for prognostic factors alone leaves residual confounding by indication."

Researchers further stated that PS and IV analysis methods can be useful under specific situations, but neither method adequately controlled confounding by indication in this study.

This study was funded by the National Cancer Institute, National Institutes of Health, and the Department of Health and Human Services. These organizations had no role in the design and conduct of the study, the collection, management, analysis and interpretation of the data, or the preparation, review and approval of the manuscript.