There appear to be many methodological as well as logical inconsistencies and interpretational difficulties with the report of Coleman, ^{Reference Coleman1} which might have diminished reviewers’ enthusiasm for its conclusions. Many of these have already been addressed by previous correspondence. We believe, however, that one methodological problem that has not yet been raised – the use of the population attributable risk (PAR) measure – is very important and merits comment. This might help readers avoid misunderstanding this study, and also other studies where the PAR is used.

In Coleman’s synthesis, the PAR measure has been applied inappropriately and, we believe, reported misleadingly. For example, the reported PAR for completed suicide is particularly high at 35%. For several reasons, readers should not interpret this figure as meaning that over a third of all suicides among women of reproductive age could be prevented if none of them underwent abortion. An inherent assumption in the PAR is that all other things would remain equal after the removal of a risk factor, which is clearly not true for abortion in this instance. Further, the aetiology of suicide is extremely complex, and in most cases cannot be attributed to a single adverse life event that is the one measured in a particular study. Women who die by suicide at some time following an abortion are likely to carry multiple distal and proximal risk factors as they proceed along their life course, as is true for most people of any age or gender who die by suicide, and it is fallacious to suggest that abortion can be isolated from other causal factors in these limited data-sets.

Second, in the first paragraph of the Discussion (p. 183), Coleman states with apparent certainty that ‘… nearly 10% of the incidence of mental health problems was shown to be directly attributable to abortion.’ This is about as unambiguous a statement of causality as could possibly be made, in the face of clear guidance on the potential pitfalls of drawing such conclusions when applying the PAR. ^{Reference Rockhill, Newman and Weinberg2} Having stated the causality of the association with such certainty, the author then appears to backtrack in her concluding remarks (pp. 185–186) by making the following ambiguous statement, clearly contradicting the view expressed at the start of her Discussion:

Following publication of just such a ‘prospective study with numerous controls’ in the New England Journal of Medicine in 2011, ^{Reference Munk-Olsen, Laursen, Pedersen, Lidegaard and Mortensen3} it might be appropriate for Priscilla Coleman (and colleagues supportive of her views) to reconsider their conclusions. This recent study ^{Reference Munk-Olsen, Laursen, Pedersen, Lidegaard and Mortensen3} provides the best data available from the largest unbiased sample on the association (or lack thereof) between excess risk of mental illness and abortion because that study is based on a large population sample, with measurement of mental illness both before and after the abortion event study. ^{Reference Munk-Olsen, Laursen, Pedersen, Lidegaard and Mortensen3} That study ‘does not support the hypothesis that there is an increased risk of mental disorders after a first-trimester induced abortion’ (quoted from abstract).