Abstract P5-15-01: Comparative effectiveness of four treatment modalities for early stage breast cancer patients

Abstract

Abstract Background Although recent work using observational data has shown survival benefits for breast conserving surgery plus radiation (BCS+RT) over mastectomy in early stage breast cancer for elderly women, only one report has compared all four treatment modalities available to these women: BCS+RT, BCS without radiation (BCS-RT), mastectomy with radiation (M+RT), and mastectomy without radiation (M-RT). This study utilized risk adjustment (RA) methods to control for variation in observed characteristics. However, RA methods may result in biased estimates if unobserved characteristics affect both the probability of receiving treatment and survival. Other statistical methods have been developed to address this issue. When valid, Instrumental Variable (IV) analysis can mitigate this bias by effectively randomizing patients into treatment through instruments. Thus our objective is to use IV methods to assess the comparative effectiveness of these treatment modalities on overall survival for elderly women with early stage breast cancer. Methods Using SEER-Medicare data, we identified 35,452 patients diagnosed 1992-1998 with Stage I or II breast cancer. We classified primary therapy as M+RT, M-RT, BCS+RT and BCS-RT using registry data and claims and estimated the effect of primary therapy on 7-year overall survival, with 6 and 8 years as robustness checks. We utilized an instrumental variables (IV) approach to address unobserved confounding. First, we estimated treatment rates for each treatment in an area by including the 50 patients closest to each beneficiary's zip code and adjusting for patient and area characteristics. These areas were then ranked and divided into quintiles, which formed the instruments for our IV analysis. In addition, we control for a number of patient (age, race, comorbidity, chemotherapy use, marital status, urban/rural residence, SEER area, diagnosis year), tumor (size, stage, grade, histology, use ER/PR status, # of positive nodes), physician (sex, specialty, medical school graduation year, foreign graduate status), and hospital characteristics (bed size, ownership, urban/rural location, teaching status, medical school affiliation, cancer center, participation in clinical trials). Results Our IV analysis showed overall survival benefits for elderly women who received BCS+RT compared to M-RT (Table 1). We estimate a 17.5% increase in the probability of surviving 7 years for those who were affected by our instrument. No survival benefit was shown for any other treatment. Conclusion We use IV analysis to address unobserved confounding in treatment and survival for elderly, early stage breast cancer patients. Consistent with previous RA estimates, we find that BCS+RT provides 7-year survival benefits over M-RT. The magnitude of the survival benefits continue to grow, even after 8 years. We find no difference in overall survival for those who received M+RT or BCS-RT compared to M-RT. Table 1. Instrumental Variables results reporting percent change in overall survival by treatment typeOverall SurvivalM-RTM + RTBCS + RTBCS aloneyears% change (p-value)% change (p-value)% change (p-value)% change (p-value)6ref-4.3 (0.61)10.7 (0.008)-0.008 (0.90)7ref-4.5 (0.66)17.5 (0.0003)0.005 (0.95)8ref-2.1 (0.84)19.0 (0.0001)0.008 (0.92) Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P5-15-01.