by ecancer reporter Clare Sansom

In recent years, several large epidemiological studies have asserted that hormone replacement therapy (HRT), at least with oestrogen combined with progesterone, is so strongly linked with an increase in breast cancer risk that it can be considered a cause of the disease. 

These studies were widely reported in the mainstream media and led to a significant decrease in the numbers of women taking HRT and, quite possibly, to an increase in other health risks for the women concerned.

A group of epidemiologists led by Samuel Shapiro from the University of Cape Town Medical School, Cape Town, South Africa has now re-examined evidence in these studies for a causal link between HRT and breast cancer using established epidemiological principles of causality. 

This paper is the fourth of a series of reports by the group; it examines the evidence presented by the Million Women Study.

Between 1996 and 2001, UK women between 50 and 64 years of age were invited to take part in this questionnaire-based, long term study of reproductive choices and health when invited for breast cancer screening.

More than one in four UK women in the target age group agreed to take part, making it the largest study of its kind in the world. In 2003, its first published results appeared to show a fairly significant risk in breast cancer incidence in those women who chose to take HRT that was more significant for combined therapy with oestrogen and progesterone than with oestrogen alone.

This was the most significant of the three considered by Shapiro and co-workers in terms of both the increase in cancer risk found and its influence on regulatory authorities and the general public, so Shapiro considered that evaluating its findings would be of particular importance. 

Four reports published by Million Women Study investigators between 2003 and 2011 all concluded that there was a causal link between HRT use and breast cancer development. 

However, it was simply a large observational study, and these have often been associated with uncertainties and defects in study design. Shapiro and his co-workers set out to evaluate it based on generally accepted and often inter-related principles of causality.

The researchers found defects in many areas of the study design, including problems with the time order of events; information bias; detection bias; confounding; and the effect of duration and dose of HRT.

In terms of time order, they found that some cancers were identified so early that they would have to have been present at the start of the study and so should have been excluded, and some participants who stopped taking HRT during the study were identified as current users throughout.

Defects in study recruitment led to both information bias and detection bias. Using a screening programme to identify participants ensured that women who were aware of breast cancer risks were more likely to participate, and wording the questionnaire with an emphasis on a possible connection between HRT and breast cancer ensured that HRT users and women already worried about possible symptoms were more likely to do so.

This led to a remarkable difference in HRT use between women who chose to participate (32%) and those who did not (19%). A relatively small number of additional cancers detected in women taking HRT on account of these biases would have nullified the results of the study.

Furthermore, inadequate allowance was made for the fact that some confounding variables, such as menopausal status and BMI, change over time.

The data on the duration of HRT use during the study was insufficient to provide adequate duration-response analysis, and did not appear to show the increase in relative cancer risk with duration that would be expected if the link between HRT and breast cancer was causative.

Shapiro and his co-workers concluded by highlighting that these defects in the Million Women Study show it is impossible to use it to prove causality.

They emphasised, however, that they have not disproved any causal association between hormone replacement therapy and breast cancer risk; they have merely shown that this flawed study cannot indicate whether one exists.

Reference

Shapiro, S., Farmer, R.D.T., Stevenson, J.C., Burger, H.G. and Mueck, A.O. (2012). Does hormone replacement therapy cause breast cancer? An application of causal principles to three studies Part 4. The Million Women Study. J. Family Planning & Reproductive Health Care, published online ahead of print 16 January 2012. doi: 1136/jfprhc-2011-100229