Overdiagnosis in mammography screening: a 45 year journey from shadowy idea to acknowledged realityBMJ 2015; 350 doi: https://doi.org/10.1136/bmj.h867 (Published 04 March 2015) Cite this as: BMJ 2015;350:h867
- Alexandra Barratt, professor of public health
- Correspondence to:
- Accepted 28 January 2015
Clinical context—Breast cancer is the most common cancer in women worldwide with a significant burden of morbidity and mortality in developed and developing countries.
Diagnostic change—Mammography screening has provided expanded opportunities to detect breast cancer over the last three decades, and more recently through incremental improvements in detection provided by digital mammography and tomosynthesis
Rationale for change—Mammography screening can detect breast cancer earlier, leading to improved patient outcomes
Leap of faith—The benefits of early detection and treatment of breast cancer (lives saved) far outweigh the adverse physical and psychosocial effects of early detection and treatment
Impact on prevalence—Breast cancer incidence has increased significantly in recent decades, most strikingly among women aged 50-69 years in countries where there is good uptake of mammography screening
Evidence of overdiagnosis—Strong increases in the incidence of early breast cancer without proportionate reductions in advanced cancer incidence, and evidence from randomised trials and observational studies showing excess cancers detected among women during screening which are not compensated by decrements in incidence once women cease screening
Harms of overdiagnosis—All women diagnosed with early breast cancer are treated comprehensively for breast cancer, even though some have overdiagnosed (harmless) cancers; these women cannot benefit from treatment but are exposed to the physical and psychosocial harms of cancer treatments
Limitations—High quality evidence about overdiagnosis and overtreatment in the context of contemporary breast cancer screening and treatment is extremely limited, leading to divisive debates between experts
Conclusion—Agreement between experts about overdiagnosis in breast cancer screening is urgently needed so that women can be better informed; advances in mammographic technology should be restrained until incremental net improvements in health outcomes are demonstrated; screening programmes should not be expanded to other age groups (older and younger women) pending results from the UK trials.
Breast cancer is the most common cancer in women worldwide.1 Its incidence has been rising in developed countries since the 1960s and in less developed countries more recently. Changes in average body mass index, reproductive factors, use of hormone therapy, and consumption of alcohol have contributed to this rise.1 The considerable burden of morbidity and mortality of breast cancer and our inability to prevent it have driven efforts over more than five decades to detect it early to maximise the effectiveness of treatment.
Introduction of screening
Although the histological diagnosis of breast cancer has remained largely unchanged, the introduction of screening programmes has meant that we have, for more than two decades, systematically sought the diagnosis in asymptomatic women. Diagnostic expansion of breast cancer has meant expanding the opportunity to find it rather than widening the definition of the disease.
Large randomised trials of screening mammography were conducted between 1963 and 1990. The two earliest trials—the Health Insurance Plan in New York2 and the Swedish Two County Study3—both reported that screening reduced the risk of dying from breast cancer by approximately 30% in women over 50. These trials, and the UK Forrest Report,4 launched widespread uptake of screening mammography in the US in the mid-1980s and the establishment of publicly funded mammography screening programmes in Sweden (1986), the United Kingdom (1988), Australia (1991), Canada (1992), and Norway (1995).
It was thought that the benefits of early detection and treatment would outweigh any risks of treatment and would not introduce substantial harm through false positives, radiation, or overdiagnosis. However, evidence of the unanticipated harms of mammography screening has fuelled debate about whether such screening should be changed5 or even abandoned.6 7 More than any other debate about overdiagnosis the discussion of breast cancer has spilt from the pages of specialist medical press and become heated arguments squarely in the public eye.8 9 10
Impact on prevalence
The incidence of breast cancer among women aged 50-69 rose by 2-10% a year after implementation of mammography screening in the UK, Europe, the United States, and Australia.11 12 Studies in Australia,13 14 Canada, 15 Denmark,13 16 Norway,13 17 Sweden,13 18 and the US19 have shown marked increases in breast cancer incidence with the introduction of screening, even after the changes in body mass index, reproductive factors, use of hormone therapy, and consumption of alcohol were taken into account.
In an optimal screening scenario rates of advanced cancer should drop with increased detection and treatment of early stage disease.20 However, although large increases in detection of early breast cancer (ductal carcinoma in situ (DCIS) and early invasive cancer) have been observed,11 17 19 rates of advanced cancer have declined only modestly17 19 21 or remained stable.22 This strongly suggests that mammography screening is detecting low risk or non-progressing breast cancer that would never have become life threatening. Furthermore, the extension of the upper age limit to 75 years for mammography screening in the Netherlands has led to a smaller than expected decrease in advanced breast cancer, while the incidence of early breast cancer strongly increased (figure⇓).23
DCIS has long been considered a precursor lesion for invasive breast cancer, and mammography has proved highly effective at finding it. Kerlikowske showed a 500% increase in incidence of DCIS among women over 50 between 1983 and 2003.24 DCIS now accounts for 20-30% of screen detected breast cancer.12 24 Notably, incidence of invasive breast cancer has not declined, suggesting that DCIS is not a precursor lesion of most invasive cancers. DCIS detection has been further increased by wide adoption of digital mammography.25 26
Breast cancer is categorised into early and advanced cancer to help guide treatment decisions.27 Early breast cancer refers to in situ disease (mainly DCIS) and invasive breast cancer that has not spread outside the breast or axillary lymph nodes. Advanced breast cancer has spread beyond the breast and axilla—locally advanced cancer to adjacent tissues or regional lymph nodes and metastatic cancer to distant organs. DCIS is sometimes described as pre-invasive breast cancer but increasingly is regarded as a marker of risk of invasive breast cancer, rather than a precursor condition.
Evidence of overdiagnosis
The possibility that mammography screening might detect non-progressing breast cancer has been recognised since its inception.28 But the tendency of screening to disproportionately detect low risk cancer was largely ignored. Only recently has the detection of non-progressing early breast cancer that would not be found in the woman’s lifetime without screening—that is, overdiagnosis—gained serious attention and recognition.
Forrest, in his influential report on breast cancer screening for the UK health ministers in 1986, was also alert to the possibility of harm and indeed mentioned the potential for overdiagnosis in breast cancer screening.4 But he said that, although with improving sensitivity of mammography it could happen, “overdiagnosis was not a problem.”
In 2012, prompted by increasing debate, an independent UK panel reviewed the benefits and harms of screening mammography and estimated that about 19% of breast cancers (including both DCIS and early invasive breast cancer) diagnosed among women invited to screening were overdiagnosed.12 This estimate was based on data from three randomised trials that the panel considered to be the best available evidence. Estimates from a wide range of observational studies and modelling studies have been considerably higher (up to 50%) or lower (<5%).12 18
The existence of overdiagnosis is now generally accepted.29 But how often it happens remains contested, with experts expressing wildly divergent views. For example, the UK panel concluded that screening saves lives but that about three times as many women are overdiagnosed as have their life saved. Estimates by other expert groups range from two lives saved for every woman overdiagnosed30 to 10 women overdiagnosed for every breast cancer death avoided.31
A review of seven autopsy studies found a median prevalence of 1.3% (range 0-1.8%) for invasive breast cancer and 9% (range 0-15%) for DCIS among women not known to have breast cancer.32 Prevalence of DCIS was higher (10-39%) when restricted to women of screening age (40-70 years).
Furthermore, a few studies have shown that some breast cancers detected through screening may spontaneously regress rather than remain non-progressing, providing further evidence that screening is detecting “harmless” cancers.33
Evidence of harms
Screen detected breast cancer that is non-progressing cannot be distinguished from early cancer that will progress because of the lack of reliable prognostic markers. As a result treatment is recommended to all women to reduce the risks of recurrence and breast cancer mortality. Although it is not a life threatening condition, screen detected DCIS is treated similarly to invasive breast cancer. Thus, overdiagnosis inadvertently turns some well women into patients who are exposed to all the risks of cancer treatments.12
Data from the US show that surgical treatment for early breast cancer is expanding. More women are choosing mastectomy over breast conserving surgery, with the greatest increase among women diagnosed with DCIS.34 Furthermore, 20% and 7% of women diagnosed with early breast cancer are considering and receiving contralateral prophylactic mastectomy, respectively.35
In the NHS screening programmes 99% of women with screen detected breast cancer undergo surgery and approximately 70% have adjuvant radiotherapy and hormone therapy.12 If approximately 20% of screen detected breast cancers are overdiagnosed (based on the estimate of 19% made by the UK independent panel), then about 20% of these women are undergoing cancer treatments for cancers to “cure” a disease which they would never have had without screening.
Risk of death from breast cancer treatment is small in the short term—even with extensive surgery the risk of death is less than 0.3%.12 But there are longer term risks; for example, radiotherapy increases the risk of heart disease, especially for women with a left sided breast cancer. Rates of major coronary events seem to increase linearly with mean radiation dose to the heart, by approximately 7% per gray in a recently reported US study.36 In this study, the increased risk was apparent within 5 years of radiotherapy and persisted for 20 years; the mean estimated radiation dose to the heart was 2.9 Gy for women with right sided breast cancers and 6.6 Gy for women with left sided cancers. Absolute risks of future heart disease were higher for women with pre-existing heart disease or multiple risk factors.
Radiotherapy for early breast cancer also increases the risk of other cancers. A meta-analysis found that mortality rates from lung and oesophageal cancers were approximately doubled in women followed up over 15 years (rate ratio 1.78 (P=0.0004) for lung cancer and rate ratio 2.40 (P=0.04) for oesophageal cancer.37 Arguably more important are the effects of breast cancer treatments on quality of life. In a large adjuvant treatment trial, adjuvant hormone therapy resulted in hot flushes (30% of women), trouble sleeping (20%), weight gain (20%), loss of interest in sex (8-16%), and lack of energy (16-20%).38 Treatment can also potentially affect family members, relationships, financial stress, and insurance status among women diagnosed with breast cancer.39
How to do better
Estimating how often overdiagnosis occurs in screening mammography is difficult because the randomised trials are now decades old and observational studies are susceptible to biases that preclude accurate evaluation. But resolution should be an urgent priority because the wide variation in estimates blocks progress towards effective policy action.18 40 41 42 Achieving consensus seems unlikely in the short term until agreement can be reached among experts about which types of study can be relied on to quantify overdiagnosis.
As suggested recently,42 resolution may be assisted by the development of internationally agreed standards for conducting cohort and ecological studies to monitor overdiagnosis in screening programmes, as it seems unlikely that new randomised trials of mammography screening will be undertaken.
Less aggressive treatment options
Trials are needed to establish whether less aggressive treatments for screen detected early breast cancer are appropriate. Taking the lead from approaches to treating low risk prostate cancer, the LORIS trial will randomise women with low risk DCIS to surgery or to active monitoring with annual mammography.43 The primary outcome is ipsilateral invasive breast cancer free survival at 5 years.
New technology brings additional challenges. Tomosynthesis (three dimensional mammography) promises a 30-50% increase in detection of breast cancers.44 45 46 But before wide implementation we should establish whether it alters the balance of benefit and harm. In one single arm study, for example, breast cancer detection rates were 5.3 cancers per 1000 screens and 8.1 cancers per 1000 screens for two dimensional mammography alone and integrated two and three dimensional mammography, respectively.44 But a randomised trial showing a reduction in cancers detected within 12 months of screening (known as interval cancers) is needed to be confident that the extra detection represents incremental benefit.45 47 The opportunity to do this trial is limited because of pressure to rapidly adopt tomosynthesis into practice.45 Nonetheless, we should learn from the introduction of digital mammography, which has increased the detection of cancer without significantly reducing detection of interval cancers and therefore may have augmented overdiagnosis without conferring a mortality benefit.25 26 48 We should consider a moratorium on the introduction of new breast imaging technologies until incremental benefit to women has been demonstrated.
Quality information for patients
Many women continue to be “prescribed” or encouraged to undergo screening rather than being supported to make an informed choice.49 50 Women should be given information that has been carefully developed and tested, because information is an intervention that may have both positive and detrimental effects.51 52
Screening targets for screening services should be questioned, and consideration should be given to ensuring the provision of balanced information—for example, using the “consider an offer” approach outlined by Entwistle and colleagues in 2008.53 Practitioners should not be incentivised to achieve participation, nor should high participation in screening be regarded as a marker of health service quality.
Think twice before extending screening programmes
Extending screening to women in their 70s has been shown to significantly increase the incidence of early stage breast cancer,23 and this could have detrimental effects for older women. The risk of overdiagnosis and overtreatment of breast cancer is influenced by competing mortality risks, which increase with age. Thus, even if screening is beneficial for a woman in her 50s, this may change as she ages such that the balance of benefit to harm becomes increasingly less favourable.
Informing wider debate
Breast cancer research has led the way in developing awareness of the potential harms of overdiagnosis and overtreatment among asymptomatic people who participate in cancer screening programmes. Increasing awareness and understanding of overdiagnosis in relation to the early detection of lung and thyroid cancers, as well as breast and prostate cancers, is needed and should be prioritised in public communication initiatives.
Lessons learnt about overdiagnosis from three decades of mammography screening programmes
Realisation—It has taken three decades to begin to publicly acknowledge overdiagnosis in mammography screening. Future early detection programmes (especially for breast, prostate, lung, and thyroid cancer) should be alert to the potential for overdiagnosis from the outset and establish mechanisms to monitor its occurrence
Changes in technology may adversely alter outcomes—New technologies should be evaluated for their overdiagnosis potential, and not simply substituted for old technologies on the assumption that the benefit to harm ratio will be improved by increased detection
Participation in cancer screening is a choice—Cancer screening holds potential benefits and harms for healthy people. An ethical approach to screening requires respecting individual autonomy to decide to accept or decline an offer of screening. Participation targets should not be set and should not be regarded as a marker of health service quality
Information is an intervention—Information about cancer screening services should be developed and tested using evidence based approaches to effective and balanced risk communication. Offers to participate in screening should come to people from neutral, independent agencies, not from screening services
Extensions to cancer screening programmes—Extending the eligible age for screening programmes (to higher or lower age groups) has the potential to alter the balance of cancer deaths averted to cancers overdiagnosed. Extension should be based on demonstrated ability to achieve equivalent benefit to harm ratios and not on the assumption that extension will achieve a net benefit
Cite this as: BMJ 2015;350:h867
This article is part of a series on overdiagnosis looking at the risks and harms to patients of expanding definitions of disease and increasing use of new diagnostic technologies.
Competing interests: I have read and understood BMJ policy on declaration of interests and have no interests to declare.
Provenance and peer review: Commissioned; externally peer reviewed.