Impact of adjuvant radiation on survival: A note of caution when using cancer registry data to evaluate adjuvant treatments

Background: With increasing frequency, studies using cancer registries have evaluated the treatment effect of adjuvant radiation; however, these analyses generally do not include chemotherapy treatment data. Our objective is to evaluate the potential impact the absence of adjuvant chemotherapy data has on the estimated survival benefit attributed to adjuvant radiation therapy. Methods: Using the National Cancer Data Base, patients were identified who underwent surgery for cancers that often require radiation therapy: breast, esophageal, gastric, pancreatic, and rectal cancer. Cox proportional hazards modeling with and without chemotherapy as a predictor variable was used to assess the impact of radiation therapy on 5-year survival. Results: From 1998 to 1999, 295,206 patients underwent surgical resection for one of five cancers. Chemotherapy administration ranged from 27.5% for gastric to 56.1% for rectal cancer. For cancers where chemotherapy affected survival, the impact of radiation therapy was overestimated in the multivariate model when chemotherapy was not included. For example, radiation treatment for rectal cancer was associated with a 31% decrease in the risk of death in the model that did not control for chemotherapy; however, the addition of chemotherapy to the model resulted in only a 14% decrease in the risk of death associated with receiving radiation therapy. Conclusions: For selected tumor sites, the administration of chemotherapy is not evenly distributed among patients receiving and not receiving radiation. Survival analyses that do not include chemotherapy administration overestimate the beneficial impact of radiation on survival. Evaluating the effect of radiation on survival retrospectively without adjusting for chemotherapy administration should be done cautiously.