Sir

I agree with the point made in your Special Report “Should journals police scientific fraud?” (Nature 439, 520–521; 200610.1038/439520a) that editorial offices are not the proper place to monitor fraud. This is why, 19 years ago, my colleague Zoltan Aannau and I proposed data audit (Nature 327, 550; 198710.1038/327550b0).

Research subject to data audit could include studies presenting possible risks to public health, or those questioned by a whistleblower or by peer review. Others could be subject to random audits. Up to 1% of all studies could be audited every three to five years, at less than 1% of the cost of the original study (Accountability Res. 1, 77–83, 1989). Auditing could be done by an independent body that would certify the validity of published results. Sponsoring institutions could choose to publish a transparent analysis of selected papers on the web.

Although these processes might not eliminate all fraud or misconduct, they could substantially reduce such unethical practices.