Methods of Controlling Confounding Effects on Observational Studies: Examining Chemotherapy Effectiveness in Breast Cancer Patients

Objective: Decreased efficacy of chemotherapy is reported in older breast cancer patients in clinical trials. The overall objective of this research was to assess the real world situation of chemotherapy effectiveness in general population by maximum accounting the bias-causing and confounding factors such as data heterogeneity, observed and unobserved confounding factors. We compared the results obtained from general Cox proportional hazard models and other three models based on Cox proportional hazard model, including PS analysis, IV analysis, and semiparametric shared gamma frailty model. ^ Results: Patients treated with chemotherapy were more likely to be younger, with no comorbid condition, with poor/undifferentiated cancer grade, with negative hormonal status, having received radiation therapy, with AJCC stage of II or III, and with one or multiple lymph nodes. This study indicated that chemotherapy has survival benefit for selective groups of older breast cancer patients. Results from all four models consistently showed decreased RR for overall mortality for 65-69 year old CTP with positive lymph nodes in stage II and III, CTP with negative hormonal receptor status as well as CTP with larger tumor size with poor cancer grades. Our results indicated the diminishing efficacy of chemotherapy as patients grow older. Moreover, all four models showed increasing RR for breast cancer-specific mortality as well as decreasing RR for nonbreast cancer-specific mortality as patients grow older. Chemotherapy showed modest benefit in 5-15 year unadjusted overall survival rates, where CUP in higher AJCC stage had greater failure probability than CTP with same AJCC stage. ^ Conclusion: Our study suggests that chemotherapy has survival benefit for selective groups of older breast cancer patients, 65-69 years of age. Relative risks for overall mortality as well as breast cancer-specific mortality increase as patients grow older.