We would like to thank Rowlands & Guthrie for their positive comments about our paper. ^{Reference Fergusson, Horwood and Boden1} We do not agree that the Dingle et al ^{Reference Dingle, Alati, Clavarino, Najman and Williams2} paper should be dismissed on the grounds that it uses an inappropriate comparison. Although this dismissal is consistent with the opinions stated in the review your correspondents cite, ^{Reference Charles, Polis, Sridhara and Blum3} it reflects a common misunderstanding. There are, in fact, two closely related causal questions that one can ask about abortion and mental health. The first concerns whether or not abortion is an adverse life event that increases risks of mental health problems. Answering this question is important for understanding the extent to which women having abortions are an at-risk population for subsequent mental health problems. The second question concerns whether any mental health risks of abortion are greater or less than the mental health risks of unwanted pregnancies that come to term. Answering this question is important for understanding the extent to which abortion may mitigate or exacerbate any mental health problems associated with unwanted pregnancy. Dingle and colleagues ^{Reference Dingle, Alati, Clavarino, Najman and Williams2} address the first question by showing that women having abortions are an at-risk population for mental health problems, and that these responses seem similar to those of women who experience pregnancy loss.

Rowlands & Guthrie suggest that our paper has a number of limitations relating to the assessment of abortion, the number of abortions and the social context of the research. However, these problems have different implications for interpreting our research. Any under-ascertainment of abortion is likely to have the effects of biasing estimates of relative risk downwards (providing that under-ascertainment is statistically independent of mental health outcomes); failure to show the effects of multiple abortions does not threaten the validity of our conclusions, but may call the precision of our conclusions into question; and the sociolegal context within which the research was conducted implies that it is important to take pre-abortion mental health into account (which we did).

Rowlands & Guthrie also suggest that an adequate review of this issue has been conducted by Charles et al. ^{Reference Charles, Polis, Sridhara and Blum3} We do not agree. This review has been criticised on the grounds of investigator bias ^{Reference Ertelt4} and these concerns are supported by the somewhat anomalous conclusions the authors draw. For example, the study by Schmierge & Russo ^{Reference Schmiege and Russo5} using the National Longitudinal Study of Youth (NLSY) data is ranked as one of the four ‘good’ studies despite the facts that only 40% of women having an abortion were included, and mental health was measured using a single scale of depression, with this outcome being assessed up to 13 years after the abortion. All of these features will bias results towards the null hypothesis. In addition, Charles et al ^{Reference Charles, Polis, Sridhara and Blum3} failed to distinguish between the different causal questions that may be asked about abortion, and misleadingly dismissed a number of studies showing links between abortion and mental health on the grounds that they did not include an appropriate comparison group.

Finally, Rowlands & Guthrie claim that conclusions cannot be drawn about the causal effects of abortion on mental health in the absence of randomised controlled trials. We believe that this argument overstates the case and there is now growing evidence for two major conclusions about abortion and mental health. First, exposure to abortion is an adverse life event which is associated with a modest increase in risks of mental health problems. Second, the mental health risks associated with abortion may be larger, and certainly are not smaller, than the mental health risks associated with unwanted pregnancies that come to term.