Deceiving Numbers

Abstract

Background. Increased 5-y survival for screened patients is often inferred to mean that fewer patients die of cancer. However, due to several biases, the 5-y survival rate is a misleading metric for evaluating a screening’s effectiveness. If physicians are not aware of these issues, informed screening counseling cannot take place. Methods. Two questionnaire versions (“group” and “time”) presented 4 conditions: 5-y survival (5Y), 5-y survival and annual disease-specific mortality (5YM), annual disease-specific mortality (M), and 5-y survival, annual disease-specific mortality, and incidence (5YMI). Questionnaire version “time” presented data as a comparison between 2 time points and version “group” as a comparison between a screened and an unscreened group. All data were based on statistics for the same cancer site (prostate). Outcome variables were the recommendation of screening, reasoning behind recommendation, judgment of the screening’s effectiveness, and, if judged effective, a numerical estimate of how many fewer people out of 1000 would die if screened regularly. After randomized allocation, 65 German physicians in internal medicine and its subspecialities completed either of the 2 questionnaire versions. Results. Across both versions, 66% of the physicians recommended screening when presented with 5Y, but only 8% of the same physicians made the recommendation when presented with M (5YM: 31%; 5YMI: 55%). Also, 5Y made considerably more physicians (78%) judge the screening to be effective than any other condition (5YM: 31%; M: 5%; 5YMI: 49%) and led to the highest overestimations of benefit. Conclusion. A large number of physicians erroneously based their screening recommendation and judgment of screening’s effectiveness on the 5-y survival rate. Results show that reporting disease-specificmortality rates can offer a simple solution to physicians’ confusion about the real effect of screening.