A note regarding Lee's checks for minimum numbers of subjects where relative risks have been calculated.

In an interesting and useful paper in this journal [1], Peter Lee provided some checks that could be carried out to examine the accuracy of reported statistics where odds ratios or relative risks had been calculated.
Our experience in using these checks is that where odds ratios are reported that they work well and are often useful. In fact, for this reason, they have been incorporated in a suite of SAS® macros for performing meta-analysis [2]. However, as has recently been noted [3], they are difficult to apply where relative risks are reported (a fact partially reflected in Lee’s discussion). This note is written to warn readers to be cautious in applying the checks to relative risks and to suggest a possible implementation that may be useful. It should be noted at the outset that Lee was specifically considering the case of epidemiological studies. This means that for the formulae for relative risks, he was assuming an application in cohort studies rather than clinical trials. For cohort studies, risks are often lower than in clinical trials, and in consequence, the formulae will generally be found to work better in the former type of study than in the latter.
1. Lee PN. Simple methods for checking for possible errors in reported odds ratios, relative risks and confidence intervals. Statistics in Medicine 1999;18:1973–1981.
2. Senn S, Weir J, Hua TSA, Berlin C, Branson M, Glimm E. Creating a suite of macros for meta-analysis in SAS (R): a case study in collaboration. Statistics & Probability Letters 2011;81:842–851.