These data show that personalised screening with eligibility for screening based on an absolute risk that is dependent on age and polygenic risk and equivalent to the risk threshold for eligibility based on age alone could reduce the number of people eligible for screening while detecting the majority of the cancers identified through a programme based on age alone. Alternatively, screening the same number of individuals in a personalised screening programme could potentially detect a greater number of cases than a screening programme based on age alone.
However, we have estimated the proportion of the population to be offered screening and the proportion of cancer cases that might be screen detectable in this subgroup of the population, from the distribution of genetic risk in the population. The estimate of potentially detectable cases is based on cancer incidence derived from cancer registration and is independent of the detection rate by screening. Given the normal distribution of polygenic risk among the cases, and the number of cases in single age group, we estimated the proportion and the expected number of cases that will occur above a certain absolute risk threshold. We have not estimated the expected number of cases to be detected following a screening programme, as this would depend on screening programme sensitivity. Screening programme sensitivity is the probability of detecting cancer by screening in a population subjected to screening. The programme sensitivity increases with decrease in the inter-screening interval, with increase in test sensitivity and with increase in the duration of the pre-clinical screen-detectable phase (Launoy et al, 1998
). In subjects of a given age at high genetic risk, the test sensitivity is likely to be the same or better than in those of the same age at low genetic risk. However, both the PSA test (Hoffman et al, 2002
) and mammogram are less sensitive in younger subjects (at lower risk). It is not known how test sensitivity will compare between younger and older subjects at the same absolute risk. The duration of the pre-clinical, screen-detectable phase may also vary by underlying genetic risk. Thus, the comparative sensitivity of the screening programme under the two approaches is not known, and empirical data will be needed in order to estimate this. Assuming equivalent or improved screening programme sensitivity, personalised screening has the potential for cost saving as the cost of the genetic test for risk profiling may be offset by savings on repeat screening and diagnostic work-up of false positives.
Reducing the number of screening tests may also reduce some of the harms associated with screening. Fewer screen tests will, at the population level, reduce the anxiety and inconvenience associated with having the test. Assuming that the probability of a false positive is independent of polygenic risk-profile, reducing the number of screen tests will also reduce the number of false-positive screens, with a reduction in the harms associated with a false positive and the benefit of saving further resources on diagnostic tests. Personalised screening also has the potential to reduce the harms associated with overdiagnosis and overtreatment, but this depends on the nature of the relationship between polygenic risk and disease aggressiveness. To date, there is equivocal evidence on the association of combination of prostate cancer susceptibility variants with disease aggressiveness (Xu et al, 2008
; Bao et al, 2010
Personalised screening may potentially confer additional benefits. It can detect cancer in younger subjects at high risk. Prostate and breast cancer detected in younger subjects may tend to behave more aggressively (Fredholm et al, 2009
; Lin et al, 2009
). If polygenic high risk is associated with disease aggressiveness, then potentially additional life years would be gained by early detection of cancer in younger subjects. The majority of breast cancer susceptibility variants identified to date confer risk for oestrogen receptor-positive breast cancers (Turnbull et al, 2010
), which are responsive to hormonal treatment and have a favourable prognosis (Dunnwald et al, 2007
). However, the nature of the complex interaction between disease risk, tumour subtype, natural history of disease, and benefit from screening are not understood and the true benefits of screening according to genetic risk cannot be estimated.
In addition to polygenic risk, there is scope for individualised screening based on phenotypic risk markers. Already, there is considerable screening activity below the age range of the UK national programme for women with a significant family history of breast cancer (Maurice et al, 2006
). There is also interest in tailoring screening to risk based on mammographic breast density. This might be used to prescribe screening frequency or indeed modality, since in addition to risk, density affects the sensitivity and the potential lead time of mammographic screening (Chiu et al, 2010
). Further studies are needed using empirical data to test the implications of adding information on PSA test level and family history to polygenic risk profiling for personalised screening in prostate cancer (Zheng et al, 2008
The threshold risk for personalised screening will be population specific. We have used data from England to estimate the proportion of men 45–79 years that would be eligible for screening and the proportion of cases potentially detectable within the eligible population at different risk thresholds. The optimum threshold risk for population of England will be different from that of another population with different incidence of cancer, such as of Asian population with low incidence of prostate cancer.
Other issues need to be considered. Screening based on a personalised risk-profile would add complexity to a screening programme. Perhaps of greater importance is the fact that eligibility for screening based on age is generally acceptable to both professionals and the public, but whether eligibility based on age and other risk factors would also be acceptable is not known. Furthermore, there are ethical and legal issues associated with genetic testing and risk prediction that would need to be addressed before personalised programmes could be implemented.
Personalised screening strategy based on age and genetic risk would potentially improve the efficiency of screening programmes and reduce their adverse consequences. Questions remain whether higher genetic risk affects cancer detection and cancer behaviour and so affecting test sensitivity, overdiagnosis and outcome. Further evidence from empirical data is needed. Nevertheless, this approach has the potential to inform public health policy decision making in the context of population screening.