Vermetten et al repeat several of the limitations addressed in our original discussion. ^{Reference Dückers, Alisic and Brewin1} We agree that the comparison of prevalence data is complicated and requires caution. We are also aware that current scientific evidence does not allow us to address all potential issues, despite the checks we reported in the initial paper. However, we do not agree that the points raised by Vermetten et al in any way contradict or detract from our analysis.

Their first suggestion is that our analysis is contradicted by known facts about the under-recognition and under-treatment of PTSD in low- and middle-income countries. Under-recognition and under-treatment of PTSD, as well as problems with the PTSD construct, may well exist in certain countries; however, these issues are logically distinct and cannot explain the specific pattern of results we obtained. For example, under-treatment cannot explain why prevalence rates based on standardised population surveys are relatively high or low.

Their second concern is about the measurement of exposure to trauma. It is true that we could not distinguish between different exposure types, which we continue to see as an important limitation. But in our analyses – and also when the exposure rates by Benjet et al ^{Reference Benjet, Bromet, Karam, Kessler, McLaughlin and Ruscio2} are used – higher rates of trauma exposure were associated with higher prevalence in the expected way. It is not the exposure data but the country vulnerability data that generate the paradox. Vermetten et al do not raise concerns about the measurement of vulnerability.

We disagree with their suggestion that ‘drawing strong inferences from this single data-point is not a scientifically sound approach’. Table 2 and Figure 2 in our paper clearly show patterns in the data as a whole that are not reliant on one country. Vermetten et al suggest that ‘other analytic approaches do allow rigorous conclusions regarding the cross-national epidemiology of PTSD’. However, the example they give does not involve country-level variables, which are the focus of our analysis. It is not clear to us how their example is relevant to our quite different research question.

So far, we found indications that, regardless of exposure, PTSD and other mental health problems are more often observed in less vulnerable, more affluent countries. ^{Reference Dückers, Alisic and Brewin1,Reference Dückers and Brewin3} The analyses we have used are appropriate to the question asked. Rather than ignoring challenging findings, we believe it is scientifically responsible to explore them further. If reliable, they have potentially far-reaching implications from an international mental health perspective.