Menopause Live - IMS Updates

Date of release: 30 November, 2009

New-onset breast tenderness is not a useful predictor of breast cancer

A recent paper by Crandall and colleagues, based on data from the Women’s Health Initiative (WHI) clinical trial, reported that ‘new-onset breast tenderness during conjugated equine estrogens plus medroxyprogesterone therapy was associated with increased breast cancer risk’ [1]. The authors assert that new-onset breast tenderness (NOBT) is a useful predictor of breast cancer and caution that ‘an increase in breast tenderness, easily detected by physicians or patients, identifies a population at particular risk for breast cancer. These findings should be considered by women who experience NOBT while receiving combined hormone therapy and by their prescribing physicians to inform decisions regarding continued combined hormone therapy’ [1]. 


Breast tenderness was assessed by a self-administered menopause symptom questionnaire that asked about the degree of bother in the past 4 weeks, using a 4-point scale (none, mild, moderate, severe). Of the 16,608 women in the trial of conjugated equine estrogens (CEE) 0.625 mg plus medroxyprogesterone acetate (MPA) 2.5 mg daily, 14,538 women reported no breast tenderness at study entry. Of these, 6555 were assigned to placebo and 6868 were assigned to active treatment. After 12 months on their assigned treatment, 770 women given placebo and 2477 women given CEE plus MPA reported any breast tenderness (mild, moderate, or severe) on repeat administration of the same questionnaire. Thus, the rate of breast tenderness within the prior 4 weeks was 11.8% in women given placebo and 36.1% in women given CEE plus MPA after 12 months of treatment, a roughly three-fold difference. Over the 5.6-year average follow-up interval, the annualized rates of breast cancer in women with NOBT were 4.4 and 6.0 per 1000 in women given placebo and active treatment, respectively [1].


In parallel with the three-fold difference in NOBT at 12 months, a basic Cox regression model testing NOBT as a predictor of breast cancer found a hazard ratio of 1.37 with a 95% confidence interval of 1.05–1.77. However, after the addition of treatment assignment to the model, the hazard ratio was reduced to 1.29 and the 95% confidence interval was no longer significant (0.99–1.70). The sensitivity of NOBT for predicting breast cancer risk was 41% and the specificity was 64%. The positive predictive value was 2.7% [1].


Breast tenderness is an extremely common symptom in women who start estrogen–progestin hormone therapy (EPT). This study reported a rate of 37% at 1 year. Another trial reported a cumulative rate of 58% during the 1st year for women aged 45–70 years assigned to oral estradiol and norethindrone acetate [2]. Importantly, the latter study showed that breast tenderness is most prominent during the first 3 months, and abates in nearly two-thirds of patients by 1 year. Thus, the assertion that this symptom is a useful predictor of breast cancer is of serious concern because of the large percentage of patients who could be affected, the likelihood that initial symptoms will be blurred with prevalence at 1 year (tripling the number of women affected), and the potential angst and panic such an association might promote. 
The assertion that NOBT is a reliable predictor of breast cancer is flawed and needlessly inflammatory. Breast cancer is by far the greatest health fear of women in this age group, perceived as three- to five-fold more threatening than coronary disease despite the fact that cardiovascular disease is an eight times greater threat [3, 4]. A risk marker with the potential to stir fear in this context should be reliable and strong. Having the risk factor should distinguish women likely to get the disease from others who will not. The positive predictive value (PPV) of a risk marker is the best measure of this association since it measures the likelihood of disease in a specific population, given that the marker is present. 
Typical risk markers with public health and clinical utility have PPVs well over 20%. For example, elevated body mass index has a PPV of 21.7–41.0% for hyperlipidemia in a community population of children [5]; silent myocardial ischemia on exercise stress-testing in asymptomatic type 2 diabetics has a PPV of 20% for incident coronary disease in 2.7 years [6], and the presence of abnormal ankle-brachial index in patients suspected of having coronary disease has a PPV of 94.8% for prevalent coronary disease by angiography [7]. The PPVs of these risk markers range from 20 to 95%, and, where time is involved, it was 20% in 2.7 years. In contrast, the PPV for NOBT in the paper by Crandall and colleagues is an order of magnitude less, at under 3% in twice the time (5.6 years). Another indication that this is not a robust association comes from the change in the hazard ratio for NOBT after adjustment for treatment. With this necessary adjustment, NOBT was no longer significant [1].
Of course, this all comes down to what it means for an individual woman. The PPV of 2.7% in 5.6 years means that less than 3% of women with NOBT developed breast cancer. More importantly, greater than 97% of women with NOBT did not develop breast cancer. If nearly all of the patients having a marker for a disease, particularly a disease as feared as breast cancer, will not get the disease, promoting this fear is likely to cause more harm than good. 
The paper by Crandall and colleagues offers the hypothesis that a three-way interaction between breast density, hormone therapy and breast cancer could account for the NOBT association through a link between breast ‘proliferation that is manifest radiographically as increased mammographic density’, and NOBT [1]. This intuitive concept has been raised in several articles in the literature. However, the only published paper to investigate this question failed to find an association [8]. This paper is cited and then dismissed with the statement ‘However, in one study’, after which the authors proceed to speculate about this unsupported hypothesis [1]. 
Another hypothesis offered by Crandall and colleagues to explain this association is that EPT ‘induced increases in serum estrone or estrone sulfate levels that could result in breast tenderness and increased breast cancer risk’ [1]. This statement is not supported by the results of the WHI CEE-alone trial which found no increased risk of breast cancer [9].
Finally, Crandall and colleagues assert that the PPV of 2.7% is substantial because it is similar to the sensitivity and specificity found for the Gail Model 2 in the Nurses’ Health Study cohort [10] and to the PPV for mammography in a large clinical population [11]. However, in the first paper, Rockhill and colleagues devote considerable discussion to the point that the Gail 2 score is not well-suited for this purpose. In the second paper, Elmore and colleagues are concerned about the high false-positive rate and conclude that new ‘techniques are needed to decrease false-positive results while maintaining high sensitivity.’ Of note, the PPVs cited from Elmore of 6.6 and 7.8% for women aged 50–59 and 60–69 years, respectively, are two- to three-fold greater than the 2.7% in the present study [1, 11]. Rather than applying the caution that was the primary point of both of these citations, the abstract and the discussion end with statements citing legitimacy based on similarity to these other findings. It is clear that Rockhill and colleagues would not sanction the application of Gail 2 for individual risk prediction because of its poor performance. Specifically, those authors wrote that the model ‘had modest discriminatory accuracy at the individual level. This finding has implications for use of the model in clinical counseling of individual women’ [10]. Elmore and colleagues end their paper by stating ‘we need to develop ways to reduce the false-positive rates of breast cancer screening and their associated psychological and economic costs’ [11]. Plainly, neither of these papers should be used to sanction a risk marker that is even poorer than those they studied.
The initiation of EPT and the continuation of use at subsequent evaluations require careful consideration of each individual woman’s health status, good clinician–patient communication about her current indications for therapy, and careful weighing of benefits and risks by both the prescriber and the patient. A frank discussion of research reports that have emerged since her last visit may be critical to address concerns. Those that demonstrate changes in the risk/benefit equation for that patient should be discussed for their potential impact on a decision to continue, modify or stop therapy. Reports that create unnecessary fear should be debunked. NOBT is a very common occurrence in women starting EPT; in the WHI cohort in this paper, 37% of women reported this symptom. Yet, over 97% of women with NOBT 12 months after initiating CEE plus MPA did not develop breast cancer over 5.6 years. Therefore, the use of this common symptom to predict breast cancer risk is inappropriate. It feeds a fear that is already out of proportion to actual risk and will cause undue angst in the vast majority of women affected.


Robert D. Langer
Principal Scientist and Medical Director, Jackson Hole Center for Preventive Medicine; Adjunct Scholar in Epidemiology, University of Pennsylvania Center for Clinical Epidemiology & Biostatistics; Professor of Family and Preventive Medicine, University of California San Diego (retired)


  1. Crandall CJ, Aragaki AK, Chlebowski RT, et al. New onset breast tenderness after initiation of estrogen plus progestin therapy and breast cancer risk. Arch Intern Med 2009;169:1684-91. Published October 12.

  2. Harvey J, Scheurer C, Kawakami FT, et al. Hormone replacement therapy and breast density changes. Climacteric 2005;8:185-92.

  3. Deeks A, Zoungas S, Teede H. Risk perception in women: a focus on menopause. Menopause 2008;15:304-9.

  4. Mosca L, Jones WK, King KB, et al. Awareness, perception, and knowledge of heart disease risk and prevention among women in the United States. American Heart Association Womens Heart Disease and Stroke Campaign Task Force. Arch Fam Med 2000;9:506-15.

  5. Eissa MA, Wen E, Mihalopoulos NL, et al. Evaluation of AAP guidelines for cholesterol screening in youth Project HeartBeat! Am J Prev Med 2009;37(1 Suppl):S717.

  6. Rutter MK, Wahid ST, McComb JM, et al. Significance of silent ischemia and microalbuminuria in predicting coronary events in asymptomatic patients with type 2 diabetes. J Am Coll Cardiol 2002;40:5661.

  7. Chang ST, Chen CL, Chu CM, et al. Ankle-disease is a useful test for clinical practice in outpatients with suspected coronary artery disease. Circ J 2006;70:68690.

  8. Boyd NF, Martin LJ, Li Q, et al. Mammographic density as a surrogate marker for the effects of hormone therapy on risk of breast cancer. Cancer Epidemiol Biomarkers Prev 2006;15:961-6.

  9. Stefanick ML, Anderson GL, Margolis KL, et al.; WHI Investigators. Effects of conjugated equine estrogens on breast cancer and mammography screening in postmenopausal women with hysterectomy. JAMA 2006;295:1647-57.

  10. Rockhill B, Spiegelman D, Byrne C, et al. Validation of the Gail et al. model of breast cancer risk prediction and implications for chemoprevention. J Natl Cancer Inst 2001;93:358-66.

  11. Elmore JG, Barton MB, Moceri VM, et al. Ten-year risk of false positive screening mammograms and clinical breast examinations. N Engl J Med 1998;338:1089-96.