Reply to the letter to the editor by M. Dören and E.M. Greiser regarding the article “Risk of different histological types of postmenopausal breast cancer by type and regimen of menopausal hormone therapy”

Abstract

Dören and Greiser raise several objectives about our study “Risk of different histological types of postmenopausal breast cancer by type and regimen of menopausal hormone therapy.”1 In fact, the response proportion of the population controls in our study was 43.1% for full participants. While this clearly is not ideal, one does need to consider that this is the true response proportion, not some kind of “participation” or “cooperation” proportion often reported in other studies.2 We undertook several measures to try to improve participation (i.e. variations of contact letters, incentives), but no amount of effort increased the response proportion. In addition to the fact that in Western countries the response in epidemiological studies has substantially declined over the last decades,3 we believe that this population of females in the age range of our study (50–74 years) is especially difficult to contact. To address the problem of potential selection bias due to a low response proportion, we offered to the nonresponders of the full interview a short telephone interview covering the key variables of the study. A total of 2,997 controls participated. Thus, the total response proportion for the controls for the key variables of the study is, in fact, 61.0%. Furthermore, we performed extensive simulations to study the potential impact of different exposure scenarios on the risk estimates. Taken together we conclude that despite the low response proportion of the full participants, selection bias is unlikely to explain our reported results. This is also supported by the fact that our results are generally in good agreement with those of other publications. The authors question the numbers in Figure 1. But this figure does not include 4,359 cases as the authors assume but 2,187 cases which were ever exposed to HT. In the 2 other panels of the figure these cases are subdivided into 1,506 cases with past and 666 cases with current exposure (this does not sum up to exactly 2,187 due to missing values). With respect to the assessment of HT exposure in our study, Dören and Greiser presumably missed the respective paragraphs in the methods and statistical analysis section. Here we reported that our questionnaire aimed at eliciting the lifetime HT exposure by means of a list of all preparations with their photographs, which were marketed throughout the past 35 years in Germany. For each woman up to 9 episodes of HT use characterized by different preparations or combinations of preparations were documented. Past and current use, or duration of use, respectively, of any of the analyzed HT combination or constituent was simultaneously included in the logistic regression models, and by doing so each was controlled for by the others. For example, if a woman was past user of a cyclical combined preparation and was current user of continuous HT both pieces of information were included in the models. The concern of the authors, we would have analyzed only recent HT exposure, is therefore unfounded. Finally, we agree that it would be important to assess the HT-associated risks for premenopausal women or women aged less than 55 years, whose menopausal status was unclear because of hysterectomy or HT use. Our study was, however, not designed to address this question and therefore not adequately powered for this purpose. Yours sincerely, Dieter Flesch-Janys.