On the potential cost effectiveness of scientific audits

The rationale for the routine performance of scientific audits has been previously discussed, and it has been proposed that independent professionals audit scientific data just as certified public accountants in independent public accounting firms audit financial data (1-4). Scientific audits would typically require the examination of data in laboratory notebooks and other work sheets, upon which research publications are based. Examples of such audits have been publicized recently, although these represent audits which have been conducted relatively inefficiently, over periods of several years per audit, and which have only been conducted due to the persistence of whistleblowers suspecting scientific fraud (5, 6). A detailed report has also appeared on the results of an audit of the research activities of a particular individual, where the audit was limited solely to an examination of the research publications themselves for errors and discrepancies (7). It should be emphasized that the purpose of conducting scientific audits is not only to detect fabrication of experimental results but also to monitor presumably more prevalent, non-fraudulent, inappropriate practices, such as misrepresentation of data, inaccurate reporting, and departure from institutional guidelines for handling hazardous materials, working with human subjects, etc. Two concerns which have been raised concerning the performance of scientific audits relate to cost. What would they cost, and who would pay for them? These questions, however, may be turned around. What does it cost not to conduct such audits, and who pays for that? An assumption often made is that science is self-correcting, that sooner or later the truth will be revealed because of the need to replicate experiments of others for independent verification of novel findings (8). Testimony recently presented at a U.S. congressional hearing suggests that the self-correcting manner in which science advances represents a very slow and inefficient process for uncovering scientific fraud (5, 6, 9). Data from a survey of university scientists was also presented, indicating ". . . a reluctance to take prompt, corrective action not only when an investigator suspects another of misconduct but also should the investigator discover flaws in his or her own published reports-whether the flaws were the result of honest error or fraud"; (10). The uncritical acceptance by established scientists that the self-correcting process works compounds the problem. The Editor of Science has written that";. . . 99.9999 percent of reports are accurate and truthful. . ."; (8). If indeed only 0.0001% of published reports were inaccurate or untruthful, there would be little justification for scientific audits. However, congressional testimony from the National Institutes of Health (NIH) revealed that";. . . the NIH Director's office has handled an average of 15-20 allegations and reports of misconduct annually in its extramural programs, which supports the work of approximately 50,000 scientists"; (11). As I shall attempt to demonstrate, since NIH alone receives fraud-related complaints concerning the work of at least 0.03% of scientists it supports in other institutions, and since evidence indicates that the incidence of fraud is considerably greater than 0.03% (10, 12), the need to audit data is justifiable on the basis of being cost effective.