Million Women Study is wrong about HRT, say epidemiologists
A study long used to establish causal links between hormone replacement therapy (HRT) and breast cancer is severely flawed, a group of epidemiologists have charged.
The observational Million Women Study (MWS), conducted in the UK, doesn’t adequately satisfy several criteria for causality – including information bias, detection bias, and biological plausibility – and thus can’t be used to conclude that HRT causes breast cancer, according to Samuel Shapiro, PhD, of the University of Cape Town in South Africa, and colleagues.
“HRT may or may not increase the risk of breast cancer, but the MWS did not establish that it does,” they wrote in the Journal of Family Planning and Reproductive Healthcare.
Several experts not involved in the study, however, have emphasised that they’re well aware of the limitations of observational studies such as the MWS, and that the totality of evidence thus far has shown a strong association between HRT and breast cancer.
“This report would not change how I counsel women, as multiple studies, including a randomised trial, have shown an increased risk of breast cancer from combination hormone therapy,” Kathy Helzlsouer, MD, director of the prevention and research center at Mercy Medical Center in Baltimore, said in an email to ABC News.
The analysis of the Million Women Study is the latest in a series of four papers by the Shapiro group exploring the credibility of three studies – the MWS, the Women’s Health Initiative (WHI), and the collaborative reanalysis (CR) – that causally link HRT, particularly oestrogen plus progestogen therapy, with breast cancer.
The earlier papers similarly found that neither the CR nor the WHI could satisfy criteria for establishing causality.
In the latest analysis, Shapiro and colleagues evaluated whether the evidence in the MWS was consistent with generally accepted principles of causality: time order, information bias, detection bias, confounding, statistical stability, duration-response, internal consistency, external consistency, and biological plausibility.
The MWS included women in the UK ages 50 to 64 who were eligible for mammography every three years from 1966 to 2001, and who were subsequently followed via questionnaire. Analyses were released in 2003, 2004, 2006, and 2011, and found significantly increased risks of breast cancer with use of oestrogen plus progestogen HRT.
Shapiro and colleagues found the study was lacking across nearly all of the categories establishing causality.
For instance, the study didn’t exclude breast cancers that were already present when the women were recruited and HRT users who were already aware of breast lumps or suspected that they had cancer and were likely to selectively choose to participate, thus increasing the number of cancers in this group, Shapiro’s group said.
Such detection bias could also have been present during follow-up, since those on HRT were advised to have mammography more frequently than those not on HRT, they added.
At the same time, HRT diminishes the sensitivity of mammography, meaning mammograms may be more intensively scrutinised to find disease, they said.
Nor did the MWS researchers adequately control for confounding, they said, since factors such as menopausal status, time since menopause, age at menopause, and BMI changed during follow-up, but such data was missing for 57% to 62% of women by the third report.
In that vein, the small relative risks observed in the study, few of which exceeded a risk ratio of two, could have been due to bias or confounding, they reported.
Also, the MWS findings may not be biologically plausible: the average time from recruitment to detection of breast cancer among HRT users was 1.2 to 1.7 years, and the likelihood of the cancer being fatal was 22% higher among HRT users. However, Shapiro and colleagues said this is biologically implausible, because such cancers normally take an average of 10 years to develop into a tumour with a diameter of 1cm.
“The name ‘Million Women Study’ implies an authority beyond criticism or refutation,” they wrote. “Yet … size alone does not guarantee that the findings are reliable.
“If the evidence was unreliable, the only effect of its massive size would have been to confer spurious statistical authority to doubtful findings,” they wrote.
However, few experts contacted by ABC News seemed concerned by the analysis, calling attention to a trend in declining breast cancer incidence in recent years that coincides with a drop in the use of oestrogen plus progestogen HRT.
“The key study which we all refer to is the WHI, which is a randomised controlled study,” Susan Love, MD, president of the Susan Love Research Foundation, said in an email. “The warnings about HRT and breast cancer come from that. The observational data such as the Million Women Study were supportive but by no means critical to the shift in thinking about HRT.”
The study’s authors acknowledged that they “presently consult, and in the past have consulted, with manufacturers of products discussed in this article.”
This article was first published by ABC News.
Shapiro S, et al. “Does hormone replacement therapy cause breast cancer? An application of causal principles to three studies. Part 4: The Million Women Study Family Planning 2012; DOI: 1136/jfprhc-2011-100229.