Overdiagnosis of cancer in Australia: the role of screening

The balance between benefits and risks could be improved if effective risk‐based screening protocols were developed

Overdiagnosis can be defined as the proportion of diagnosed cancers that would not otherwise have come to a person's attention during their lifetime.1,2 Overdiagnosis provides no benefit to the patient but can have financial, psychosocial, and health consequences.2 While advances in imaging and other screening and diagnostic technologies can lead to therapeutic benefit, they can also increase overdiagnosis. Because overdiagnosed cancers are generally indistinguishable from potentially lethal cancers, the imperative to treat is equivalent.2

To reduce the potential for overdiagnosis, personalised risk‐based screening has been advocated, whereby screening intensity would be adjusted according to individual cancer risk.3 Theoretically, the full benefits of screening would be retained but the potential for overdiagnosis would be reduced. The concept is attractive, but considerable research would be needed to implement it.3 Investigations are underway into finding how to customise screening in order to achieve the best trade‐off between benefit and overdiagnosis.3 The technical challenges are significant if protecting patients is to be maintained, public acceptance of the process assured, and medico‐legal liability appropriately managed.3 The social acceptability of different policy options would need to be assessed.

To weigh the benefits and consequences of screening, both must be accurately quantified. Doing so is not a simple task, nor is determining the overdiagnosis burden.2 In the study reported in this issue of the MJA, Glasziou and colleagues4 estimated the burden by comparing lifetime risks in 1982 (before large scale organised cancer screening and the introduction of newer imaging technologies) and in 2012. The authors focused on five cancers (prostate, breast, renal, thyroid, and melanocytic skin cancers) for which evidence of overdiagnosis has been reported.5 The investigators adjusted their analyses for potential confounding by demographic changes and alterations in the prevalence of cancer risk factors; for breast cancer, for example, these included childbearing, parity, and excess body weight.

Glasziou and his colleagues estimated that the lifetime proportion of overdiagnosed cancers was 24% for men and 18% for women. For men, it was estimated that 22% of invasive melanomas (and 58% of all melanomas), and 73% of thyroid, 42% of renal, and 42% of prostate cancers were overdiagnosed; for women, it was estimated that 15% of invasive melanomas (54% of all melanomas) and 73% of thyroid, 58% of renal, and 13% of invasive breast cancers (22% of all breast cancers) were overdiagnosed. The investigators performed their study carefully, supporting their conclusions when possible with sensitivity analyses and reference to outside literature. Uncertainty about their estimates of overdiagnosis nonetheless remains.

The Independent United Kingdom Panel on Breast Cancer Screening reported in 2013 that the extent of overdiagnosis was uncertain because reliable data on lifetime risk were not available.2 This problem still applies. The panel found that evidence from randomised controlled trials indicated that the frequency of overdiagnosis was about 11% from the population perspective and about 19% for the women invited for screening. It has been argued, however, that these estimates were inflated, because in many studies the follow‐up time after the cessation of screening was insufficient.6 Further, the relevance to contemporary screening programs of trial results for screening undertaken 40 or so years ago can be questioned.2 The UK panel also considered the findings of observational studies, and found that varying the assumptions and statistical methods applied to the same data resulted in widely varying estimates, from 0% to 36%, with no good reasons to prefer particular sets of assumptions.2 Many of the uncertainties discussed by the UK panel also apply to the observational study of Glasziou and colleagues.

Many factors other than overdiagnosis could have changed cancer incidence rates between 1982 and 2012, including alterations in cancer risk related to changes in physical activity, body weight, use of hormone therapies, reproductive behaviour, and diet and alcohol intake. Pathology interpretation could have been influenced by increased medico‐legal pressures.7 Completeness of cancer registration data may also have varied, but this effect is likely to have been minimal.

Glasziou and colleagues have practised the art of the possible in conducting their study, including adjusting for potential confounding, but the available data for this purpose were too limited to remove uncertainty.

Overdiagnosis, whatever its prevalence, is a negative outcome, and the balance between benefits and risks could be improved by effective risk‐based screening protocols.3 The question is how they might be developed. Studies that could produce data for informing a risk‐based approach have commenced in a number of countries, using questionnaire data, mammographic density, and the assessment of new biomarkers to estimate breast cancer risk.3 Such studies should be a priority, so that shifting to a risk‐based screening approach can be evidence‐based rather than moving ahead of the evidence.