Response to “Aromatase Inhibitors and Dementia Risk: Putting Safety Into Perspective”

We have read and appreciated Rizk and Mokbel's comments on our study. They highlight some of the main limitations of pharmacovigilance database studies. We are aware of all of these limitations and mention them in our article. In this letter, we will try to respond to their comments in more detail. First of all, they mention the underreporting, which could be more important in disorders such as dementia, given its insidious nature and difficulty in associating with the medication. We agree that underreporting is always a problem with this type of study, and so we highlight it; but, in general, this bias could only mean that the real number of cases would be higher. They also point out that selective overreporting could be a problem, given the publication of previous studies associating cognitive problems with these treatments. They note that without solid data, disproportionality analysis alone can't establish causality or even true incidence rates. In terms of over-reporting due to heightened awareness of the potential adverse drug reaction (ADR), the so-called “notoriety bias”, it could effectively lead to and over-representation of the ADR drug association. We don't know how this bias is affecting the reporting of this adverse reaction, but we agree that a disproportionality analysis cannot establish causality and, of course, it cannot establish incidence rates, and we point this out in our article. Regarding confounding variables, it is difficult to control for them in pharmacovigilance databases due to the lack of information in many cases, but we have done so for age and sex, including only women over 45 years of age to exclude all non-menopausal patients. This selection minimizes indication bias, as the “control” group is women with the same risk of the potential reaction studied. In conclusion, we agree with Rizk and Mokbel about the limitations of studies in pharmacovigilance databases and therefore conclude our work saying that “more research is needed to understand this complex issue”. It is not our aim to warn patients, but we do believe that the safety of medicines should be continuously assessed to keep safer medicines on the market. Pharmacovigilance databases give us some limited information. We insist that more specific studies should be carried out to evaluate this issue. All authors have contributed to and reviewed the final version of the manuscript. The authors declare no conflicts of interest. The authors have nothing to report.