Overdiagnosis in Breast Cancer Chemoprevention Trials Overdiagnosis in Breast Cancer Chemoprevention Trials

the reduction in the use of hormone replacement therapy, and the immediacy of the effect suggests the disappearance of established subclinical cancers rather than the prevention of new cancers 23–25. The complementary observations concerning the immediate effects of tamoxifen chemoprevention and hormone replacement withdrawal suggest that a slowly growing er-positive cancer might be induced to regress based on the sudden removal of an estro-genic signal. We question whether that paradigm is relevant for evaluating antiestrogen-based chemoprevention studies—in particular if the cohorts under study were being followed for incident cancers with annual mammography. If the paradigm were to hold true for patients assigned to an antiestrogen arm in the screening trials, then removal of the estrogen signal might provoke the regression of some er-positive nonpalpable tumours, while similar cancers in the placebo arm might remain detectable—but would not threaten survival (and might eventually regress). If all incident breast cancers in the chemoprevention trials were to have been diagnosed by mammography, overdiagnosis would be a concern. In contrast, if the great majority of cancers were detected by clinical breast exam or self-exam (being therefore palpable), overdiagnosis would be less of an issue. However, when the relevant studies were initiated, overdiag-nosis was not a recognized phenomenon. We therefore reviewed the incidence and mortality data from the breast cancer prevention trials, seeking to determine whether breast cancers were detected only by mammography or were palpable. A preventive effect restricted to women with cancers detected only by mammography would raise the question of whether the observed incidence benefit could be expected to translate into a mortality benefit. If so, then an excess of nonpalpable mammography-detected er-positive breast cancers would be expected in the placebo arm compared with the treatment arm, and restriction of the analysis to palpable cancers would attenuate or eliminate the Several randomized controlled trials have demonstrated that the preventive use of an antiestrogen agent such as tamoxifen 1–4 , raloxifene 5–7 , anastro-zole 8 , or exemestane 9 will reduce the incidence of estrogen receptor (er)–positive breast cancers by 50% or more. The reduction in risk becomes apparent shortly after tamoxifen initiation 10. However, no mortality benefit has yet been demonstrated with tamoxifen or any other agent, an effect that might be statistical: that is, the statistical power to detect a difference in mortality could be lacking because deaths from breast cancer are far fewer in number than cases of breast cancer, …


Overdiagnosis in breast cancer chemoprevention trials
V. Sopik msc* and S.A. Narod md* the reduction in the use of hormone replacement therapy, and the immediacy of the effect suggests the disappearance of established subclinical cancers rather than the prevention of new cancers [23][24][25] .The complementary observations concerning the immediate effects of tamoxifen chemoprevention and hormone replacement withdrawal suggest that a slowly growing er-positive cancer might be induced to regress based on the sudden removal of an estrogenic signal.
We question whether that paradigm is relevant for evaluating antiestrogen-based chemoprevention studies-in particular if the cohorts under study were being followed for incident cancers with annual mammography.If the paradigm were to hold true for patients assigned to an antiestrogen arm in the screening trials, then removal of the estrogen signal might provoke the regression of some er-positive nonpalpable tumours, while similar cancers in the placebo arm might remain detectable-but would not threaten survival (and might eventually regress).If all incident breast cancers in the chemoprevention trials were to have been diagnosed by mammography, overdiagnosis would be a concern.In contrast, if the great majority of cancers were detected by clinical breast exam or self-exam (being therefore palpable), overdiagnosis would be less of an issue.However, when the relevant studies were initiated, overdiagnosis was not a recognized phenomenon.
We therefore reviewed the incidence and mortality data from the breast cancer prevention trials, seeking to determine whether breast cancers were detected only by mammography or were palpable.A preventive effect restricted to women with cancers detected only by mammography would raise the question of whether the observed incidence benefit could be expected to translate into a mortality benefit.If so, then an excess of nonpalpable mammography-detected er-positive breast cancers would be expected in the placebo arm compared with the treatment arm, and restriction of the analysis to palpable cancers would attenuate or eliminate the Several randomized controlled trials have demonstrated that the preventive use of an antiestrogen agent such as tamoxifen [1][2][3][4] , raloxifene [5][6][7] , anastrozole 8 , or exemestane 9 will reduce the incidence of estrogen receptor (er)-positive breast cancers by 50% or more.The reduction in risk becomes apparent shortly after tamoxifen initiation 10 .However, no mortality benefit has yet been demonstrated with tamoxifen or any other agent, an effect that might be statistical: that is, the statistical power to detect a difference in mortality could be lacking because deaths from breast cancer are far fewer in number than cases of breast cancer, and because the average time to cancer is much shorter than the time to death 11 .In other words, it could be too early to see an effect.However, the lack of an observed survival benefit might also be a result of chemoprevention agents preferentially preventing cancers that would rarely lead to death.That paradigm extends the (controversial) concepts of overdiagnosis and of the potential for spontaneous regression of some lowgrade breast cancers 12 .
Overdiagnosis is a problem associated with mammography-detected breast cancers [13][14][15] , and some authors suggest that overdiagnosis helps to explain why the reduction in mortality associated with screening mammography might be less than expected [16][17][18] .In the report of the 25-year follow-up of the Canadian National Breast Screening Study, the authors concluded that 50% of all nonpalpable mammography-detected cancers are overdiagnosed (22% of all mammography-detected cancers) 17 .Breast cancers that are most likely to be nonprogressive (or regressive) and to be detected by mammography are er-or progesterone receptor-positive, small, nodenegative, and nonpalpable [19][20][21] .
The phenomenon of breast cancer regression was also observed in the wake of the widespread discontinuation of hormone replacement therapy around 2000, after hormone replacement therapy was reported to increase the risk of breast cancer 22 .A rapid decline in breast cancer incidence quickly followed effect.In the absence of data on palpability (or detection method), er status and lymph node status might to some extent indicate whether the incident cancers could be examples of overdiagnosis-under the assumption that most overdiagnoses are to be found in the node-negative luminal subgroup.
Table i summarizes er status, nodal status, mortality, and detection method (mammographydetected vs. palpable) of incident breast cancers diagnosed in the trials.No trial demonstrated a significant (or borderline significant) reduction in mortality, and in all trials, the benefit was restricted to the er-positive subgroup.The detection method of incident cancers was reported only in the 1998 report of the National Surgical Adjuvant Breast and Bowel Project Breast Cancer Prevention Trial.In that report, tamoxifen was shown to reduce the occurrence of er-positive tumours by 69% 26 .In 2008, Shen et al. reported the detection method of the cancers, identifying mammography-only detection of 54 er-positive cancers in the placebo arm and 19 in the tamoxifen arm 27 .Removing those nonpalpable breast cancers from the analysis resulted in a significant risk estimate of 0.29 (p < 0.0002), suggesting that overdiagnosis had not unduly inflated the results.However, the data came from only 4 years of follow-up 26 , and it would be useful to have data for the palpability of incident cancers reported in the long-term follow-up from 2005.Contrary to the 1998 report, the long-term follow-up identified a higher reduction in node-negative breast cancers and more breast cancer deaths in subjects assigned to tamoxifen than to placebo (12 vs. 11) 2 .
The remaining trials have published no information about the palpability or detection method of incident cancers.All trials (except the raloxifene trials) reported on the nodal status of incident breast cancers and show a reduction of both node-positive and node-negative breast cancers with the chemoprevention agent, but in all trials except one 4 , the reduction was more profound for the node-negative cancers 1,2,8,9 .Based on the data reported so far, the evidence that overdiagnosis might explain the discordance between cancer incidence ratios and cancer mortality ratios in the breast cancer prevention studies is insufficient-but overdiagnosis might still be a contributing factor.Unfortunately, the trials do not have adequate power to detect a small reduction in mortality (20% for example).However, this issue appears to be of increasing relevance, and we would be grateful if the authors of future study reports would provide data concerning the mammography detection and palpability of incident cancers in addition to er status and nodal status.It is commendable to prevent the occurrence of breast cancer, but ultimately, the greatest value of tamoxifen chemoprevention will come from its ability to prevent deaths from breast cancer.
Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).

table i
Estrogen receptor (er) status, mortality, nodal status, and detection method of incident breast cancers detected in tamoxifen, anastrozole, and exemestane prevention trials na Current OnCOlOgy-

VOlume 22, number 1, February 2015
Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).

e9 Current OnCOlOgy-VOlume 22, number 1, February 2015
Copyright © 2015 Multimed Inc.Following publication in Current Oncology, the full text of each article is available immediately and archived in PubMed Central (PMC).