When the data won’t get you there: The ethics of scientific error, and worse

Abstract

A surgical scientist is informed by his research assistant and statistician, after they've reviewed data for a new publication, that the conclusions of a previously published paper may be incorrect. The prospectively collected data were assembled from experiences at several hospitals. The criteria for inclusion were not precisely observed, and a number of experimental-arm patients at one institution were mistakenly excluded from the analysis after enrollment. A computer glitch has corrupted much of the original data, and the backups are not to be found. Huge study samples and data sets, and the long interval since the data were collected, make chart retrieval effectively impossible. The study's conclusion did not alter conventional surgical practice; it supported current assumptions and techniques. The statistician now cannot confirm whether the data did or did not support the conclusions. The people who worked on the project all feel that even if the contaminated data were excluded, the recalculated data would support the published paper. But a recalculation is not possible. What should be done?