Authors’ reply to Marjoribanks and colleagues, Rossouw and colleagues, Schroll and Lundh, and McPhersonBMJ 2012; 345 doi: http://dx.doi.org/10.1136/bmj.e8164 (Published 03 December 2012) Cite this as: BMJ 2012;345:e8164
- Louise Lind Schierbeck, registrar1,
- Lars Køber, professor, consultant2,
- Jens-Erik Beck Jensen, associate professor, consultant1
- 1Department of Endocrinology, Hvidovre Hospital, 2650 Hvidovre, Denmark
- 2Department of Cardiology, Rigshospitalet, Copenhagen, Denmark
Although randomisation is an important part of any trial, we acknowledge that performance and detection bias exist in open label trials,2 3 but these biases are less important when endpoints are driven by mortality. Furthermore, no trial of hormone replacement therapy (HRT)—including the Women’s Health Initiative—conducted in women with a uterus is completely blinded.4 5 Thus, our results are probably as valid as other randomised HRT trials. In the Danish Osteoporosis Prevention Study (DOPS), randomised treatment was stopped in 2002 when it was suggested that HRT had negative effects on cardiovascular health. This should have resulted in a higher detection rate of cardiovascular disease in the subsequent post-randomisation years of follow-up. However, this was not the case.
In the original protocol cardiovascular safety was defined as safety measures. In the present study the primary focus was on cardiovascular outcome; this was defined before data analysis and only hard endpoints that are associated with poor prognosis were included. These were all cause mortality, acute myocardial infarction, and hospital admissions for heart failure because these are also events validated in national Danish registers; thus, this is not a weakness but a strength.2 3 6
We agree that power is low to detect breast cancer, but power is not a problem in relation to the significant findings driven by all cause mortality.3 Also, we did analyse the data separately for estradiol plus sequential norethisterone acetate and estradiol alone (labelled as intact uterus and hysterectomy in figs 3-61).3
Thanks to McPherson for bringing smaller randomised controlled trials to our attention.7 We did not intend to discuss the vast HRT data, but these studies are relevant to our discussions.
DOPS provides long term longitudinal randomised trial data in a cohort of the very women who are normally treated, for which no data previously existed. Thus, it directly provides information on the long term effects of HRT in close proximity to menopause, especially with regard to cardiovascular disease, total mortality, and to some degree cancer. Finding significant effects in small groups underscores the clinical relevance for efficacy in reducing cardiovascular disease and total mortality.
Cite this as: BMJ 2012;345:e8164
Competing interests: LLS, LK, and J-EBJ are authors of the discussed research article.