Rapid Responses to:

EDUCATION AND DEBATE: Nathan J Coombs, Richard Taylor, Nicholas Wilcken, and John Boyages Hormone replacement therapy and breast cancer: estimate of risk BMJ 2005; 331: 347-349 [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Too much information?

Kevin F Donnelly   (6 August 2005)

Framing risk as minimal is wrong

Prof. Mayer Brezis, MD MPH, Professor of Medicine   (7 August 2005)

Uncertain estimation

Jane I Galbraith   (15 August 2005)

Framing breast cancer risk as minimal is wrong

Ellen C G Grant   (16 August 2005)

An unfinished debate

Mark J Garton   (21 August 2005)

Too much information? 6 August 2005
Kevin F Donnelly, General Practitioner Cromwell Place St Ives PE27 5JD Send response to journal: Re: Too much information?	This otherwise helpful article is illustrated by a photograph of a bare breasted “patient” (with her face visible). To my mind, this is unnecessary in an article which is supposedly of an Educational nature, unless you feel the readers of the BMJ may not yet have the necessary anatomical knowledge of where breast cancer occurs. If that is indeed the case, should you not also illustrate articles about prostatic cancer with a pictures howing a “patient” with a gloved finger up the ...? Competing interests: None declared
Framing risk as minimal is wrong 7 August 2005
Prof. Mayer Brezis, MD MPH, Professor of Medicine, Director, Center for Clinical Quality & Safety Hadassah - Hebrew University Hospital Send response to journal: Re: Framing risk as minimal is wrong	The qualification of the increased risk of breast cancer by HRT as minimal, slight or very small is troubling. This increased risk is not far from the increase of 0.2% of incidence of proven myocardial infarction with Vioxx - See JAMA 2001: The annualized myocardial infarction rates for COX-2 inhibitors in both VIGOR and CLASS were significantly higher than that in the placebo group of a recent meta-analysis of 23 407 patients in primary prevention trials (0.52%): 0.74% with rofecoxib (P = .04 compared with the placebo group of the meta-analysis) and 0.80% with celecoxib (P = .02 compared with the placebo group of the meta-analysis. Which led Topol to write in the NEJM (2004): “Considering the tens of millions of patients who were taking rofecoxib, we are dealing with an enormous public health issue. Even a fraction of a percent excess in the rate of serious cardiovascular events would translate into thousands of affected people.” So this is a major public health problem. But even at the individual level, this level of risk (1:100 to 1:1000) has been defined as moderate by a previous Education & Debate in this journal (Calman KC. Cancer: science and society and the communication of risk. BMJ. 1996;313(7060):799-802) and in fact, comparable to the risk of dying over a year if smoking 10 cigarettes a day (See Table 2 of this report). The risk of contracting HIV is estimated at 1:1000 unprotected heterosexual contacts, even in a nation such as South Africa, in which more than 22% of adults are HIV-positive (Singhal A, Rogers EM. Combating AIDS: communication strategies in action. Thousand Oaks, Calif.: Sage Pub.; 2003) Given this probability, is it likely that an individual will adopt safer sex practice? (as discussed in Rogers EM. Diffusion of innovations, 5th ed, New York: Free Press; 2003, p.176). Or should we say: this risk is very small - forget about prevention, live the day?! Shouldn’t we, health professionals, be able to communicate what appears to laymen low risk, in a more realistic manner? The framing of risk in a responsible way should include comparisons with other known risks (such as Vioxx, cigarettes and unsafe sex), rather than being qualified as small. Competing interests: None declared
Uncertain estimation 15 August 2005
Jane I Galbraith, Senior Teaching Fellow in Statistics London School of Economics, WC2A 2AE Send response to journal: Re: Uncertain estimation	Good practice requires estimates to be accompanied by an indication of uncertainty, such as quoting standard errors or presenting confidence intervals. It also requires possible sources of bias to be clearly identified, especially when the bias may be large compared to the standard error. On their way towards estimating the additional risk of breast cancer due to hormone replacement therapy (HRT) Coombs and colleagues attempt to estimate the baseline risk for different age groups by substituting results from three separate studies into an algebraic identity. This identity, equating the indirectly with the directly calculated "attributable risk", only holds if all the figures have been measured in a compatible manner on the same population (or at least on populations identical in relevant respects). Coombs and colleagues use relative risk estimates from the Million Women Study, estimated HRT use from the 2001 Australian Health Survey, and estimates of underlying breast cancer incidence obtained by some intricate adjustment of the New South Wales Central Cancer Registry records for 2001. Are they correct to assume that data collected using very different methodologies from very different populations will nonetheless give coherent results? They further assume that the relative risks do not vary with age, that half the Australian women on HRT in the 2001 survey were on combination therapy, that 1972 to 1989 was a "stable period", that past users have the same risk as never users, and that possible effects of oral contraceptive use are irrelevant. There are two problems: firstly unless the assumptions hold and the data are compatible the estimated baseline risks will be biased; secondly the estimates of relative risk, HRT use and cancer incidence are separately subject to estimation error and hence the estimated baseline risks will be subject to relatively larger estimation errors. The sizes of the biases and standard errors of the estimated baseline risks might, if they were known, render the supposed estimates of additional risk so uncertain as to be meaningless. Jane Galbraith, Department of Statistics, London School of Economics. Competing interests: None declared
Framing breast cancer risk as minimal is wrong 16 August 2005
Ellen C G Grant, physician and medical gyanecologist Kingston-upon-Thames, KT2 7JU, UK Send response to journal: Re: Framing breast cancer risk as minimal is wrong	Nathan Coombs and colleagues’ uncertain calculations of cumulative risks of breast cancer due to HRT received widespread publicity in the national media.1 This was unfortunate because the authors minimised proven high risks by making several unjustifiable assumptions. The Australian 2001 survey used listed 62.3% of 55-59 year old women as HRT non users. In 2000 nearly half of the HRT non users in this age group in a South Australian survey had previously used HRT (43%).2 Past use of progestagens and oestrogens for contraception increased breast cancer risks for up to 10 years after stopping in a reanalysis of 54 studies.3 Progestogen use increases risk of breast cancer irrespective of the reason for use. Most women used as controls in HRT studies, including the Women’s Health Initiative studies, previously used hormones for contraception, infertility or HRT which leads to underestimation of current and post-HRT risks. Furthermore, Coombs substantially lowered recorded breast cancer incidence in women in New South Wales in 2001 by estimating “underlying” population incidence from an age, period, cohort model based on 1972-1996 data, using the average of the stable period effect 1972-1989. This manipulation reduced the numbers of the breast cancers in women over 50 by as much as a fifth. Breast cancer incidences were not stable in the UK between 1972-89 but varied with changes in hormone use.4,5 Sharp increases in HRT use from 1986 and matching increases in breast cancer registrations led to the introduction of national breast cancer screening of 50-65 year-olds in 1992. Registration data from the National Statistics National Cancer Intelligence Centre and the Welsh Cancer Intelligence & Surveillance Unit from 1971 to 1997 show the largest increases in breast cancer were 72% for ages 25-29, 112% for ages 50-54, 83% for ages 55-59 and 78% for ages 60-64. The all age increase in England and Wales was 70%. The average exposure to progesterones and oestrogens, given for contraceptives or HRT, was 3 years and 2 years in 105 breast cancer studies reanalysed. Since 1962 relatively short durations of hormone use have contributed to large increases in breast cancer. In the Million Women Study current use of combined HRT for less than 12 months increased the risk breast cancer and of abnormal mammograms the Women’s Health Initiative Study. Both studies were terminated prematurely before risk estimates could be upgraded by further diagnoses of breast cancers, strokes, thromboses or heart attacks in users. It is unhelpful to women to encourage use of HRT by manipulating cancer data. 1 Coombs NJ, Wilcken N, Boyages J. Hormone replacement therapy and breast cancer: estimate of risk. BMJ 2005: 331: 347-349. 2 MacLennan AH, Wilson DH, Taylor AW. Hormone replacement therapy use over a decade in an Australian population. Climacteric 2002; 5: 351-6. 3 Breast cancer and hormonal contraceptives: collaborative reanalysis of individual data on 53 297 women with breast cancer and 100 239 women without breast cancer from 54 epidemiological studies. Collaborative Group on Hormonal Factors in Breast Cancer. Lancet 1996; 347: 1713-27. 4 Grant ECG. Increases in breast cancer incidence http://bmj.com/cgi/eletters/328/7445/921#55298, 1 Apr 2004 5 Grant ECG. Re: Rapid Responses: Author’s reply. http://bmj.com/cgi/eletters/328/7445/921#55843, 6 Apr 2004 Competing interests: None declared
An unfinished debate 21 August 2005
Mark J Garton, Consultant physician Department of Medicine, Perth Royal Infirmary, Perth PH1 1NX, UK Send response to journal: Re: An unfinished debate	Editor Coombs et al clearly illustrate the excess risk of breast cancer attributable to hormone replacement therapy (HRT). 1 As HRT remains the most effective remedy for hot flushes and vaginal dryness, 2 their data should comfort many women considering such treatment. However, the results would have been even more reassuring had they used risk estimates derived from the randomised controlled Women’s Health Initiative (WHI) trial, 3 rather than observational data from the Million Women Study (MWS). 4 In the WHI study, combined HRT carried a lower relative risk of invasive breast cancer than in the MWS (RR 1·24 vs 2·00), particularly among women without previous HRT exposure (1·09, 95% CI 0·86-1·39). Hazard ratios were higher in subjects with prior HRT exposure, but the trend for duration of use was not statistically significant. The annualised incidence of breast cancer among controls previously exposed to HRT was actually lower than never users, while the incidence among ever- users of HRT assigned to active treatment was close to background rates. This unexpected finding largely accounts for the apparent dose response relationship between HRT exposure and subsequent invasive breast cancer. Although this could reflect successful user bias (as suggested by the WHI investigators), similar findings could arise if combined HRT promotes, but does not initiate, tumour growth. The oestrogen-only arm of the WHI was even more surprising, suggesting that unopposed oestrogen might actually reduce the risk of invasive breast cancer in hysterectomised women.5 Decisions to initiate or continue HRT now require a broad assessment of likely risks and benefits. Simple advice to limit therapy to less than 5 years may not help individuals with severe climacteric symptoms, premature menopause or osteopenia, and in the light of available evidence, is misleading. In providing an easily understood framework for assessing and expressing risk, Coombs et al re-open an unfinished debate. mark.garton@tuht.scot.nhs.uk 1. Coombs NJ, Taylor R, Wilcken N, Boyages J. Hormone replacement therapy and breast cancer: estimate of risk. BMJ 2005; 331: 347-9. 2. Hickey M, Davis SR, Sturdee DW. Treatment of menopausal symptoms: what shall we do now ? Lancet 2005; 366: 409-21. 3. Chlebowski RT, Hendrix SL, Langer RD, Stefanick ML, Gass M, Lane D, et al for the WHI Investigators. Influence of estrogen plus progestin on breast cancer and mammography in healthy postmenopausal women: the Women's Health Initiative randomised trial. JAMA 2003; 289: 3243-53 4. Million Women Study Collaborators. Breast cancer and hormone- replacement therapy in the Million Women Study. Lancet 2003; 362: 419-27. 5. The Women’s Health Initiative Steering Committee. The effects of conjugated equine estrogen in postmenopausal women with hysterectomy. The Womens’ Health Initiative Randomised controlled trial. JAMA 2004; 291: 1701-12. Competing interests: I have received speaker fees from Merck Sharp and Dohme, and Proctor and Gamble, and research support from Lilly