To the Editor: I am writing to inform the editors and readers of JAMA of errors in a study of conventional-dose vs high-dose conformal radiation therapy in clinically localized prostate cancer, published in the September 14, 2005, issue of JAMA, for which I was principal investigator.1
These errors were uncovered in the process of working on a series of secondary analyses for which access to the raw data was needed.
All of the original data had been collected, stored, and analyzed by the statistics department of the American College of Radiology (ACR). For one secondary analysis, the raw data were obtained from the ACR and the original Kaplan-Meier analysis of cancer-free survival was rerun. This outcome had been defined as freedom from 3 successive increases in the level of prostate-specific antigen (PSA). We were unable to reproduce the curves reported in the JAMA article and brought this to the attention of the ACR statisticians.
They traced the problem to an error in the definition of PSA failure coded in the computer program. Although the primary outcome of the study was biochemical failure defined as 3 successive increases (as indicated in the “Methods” section of the article), the computer code identified any 3 increases as biochemical failure. This erroneously identified additional patients as “failures” along with those correctly classified.
Zietman AL. Correction: Inaccurate Analysis and Results in a Study of Radiation Therapy in Adenocarcinoma of the Prostate. JAMA. 2008;299(8):898-899. doi:10.1001/jama.299.8.898-c