Aspirin Use, Colorectal Cancer Survival, and Loss to Follow-up

Abstract

To the Editor: Dr Chan and colleagues reported the results of an observational study of nonsteroidal antiinflammatory drug (NSAID) use in the Nurses’ Health Study (NHS), in which a significant reduction in colorectal cancer– specific mortality was observed with regular postdiagnostic aspirin use (odds ratio, 0.71; 95% confidence interval, 0.53-0.95), but not with prediagnostic use. We have concerns about the interpretation of these results because of patient losses to the study population. Research investigating the association between NSAID use and colorectal cancer survival should include all incident colorectal cancer cases to validly generalize results to the larger patient population. All study participants in the study by Chan et al belonged to the NHS cohort; in the colorectal cancer survival study, an NHS cohort member was identified as a diagnosed case of colorectal cancer via the biennial mailed questionnaire. After noting diagnosis of colorectal cancer in the biennial questionnaire, study participants then had to consent to release of medical and pathology records for use in the present study. Due to the potential for significant lag time between diagnosis of colorectal cancer in NHS cohort members and entry into the present study as a participant, a number of colorectal cancer patients may have died before being detected. The result is that information, including both NSAID exposure and vital status, may have been lost from the study. The authors do not include information on how many colorectal cancer cases were expected in the NHS cohort, how many colorectal cancer cases were actually diagnosed, or how many cases were brought to the attention of the survival study for specimen collection and mortality follow up. Much of this information should be available through links with state and national registries. Patient loss before specimen collection and follow-up may have resulted in an analysis population that reflected the experience of patients who survived at least long enough to report colorectal cancer diagnosis to the study, not the general patient population. It is reasonable to expect that this group of longer-term survival patients may have experienced improved survival overall and had disease that was either less severe or affected differently by medications, including NSAIDs. The consequence of losing information on mortality events that occur before full data collection is that the population selected for study follow-up does not represent the general patient population; rather, any observed associations, or lack thereof, are relevant only to those patients who survived long enough to be detected. Anna E. Coghill, MPH acoghill@fhcrc.org Polly A. Newcomb, PhD, MPH John D. Potter, MD, PhD Cancer Prevention Program Fred Hutchinson Cancer Research Center Seattle, Washington