Benefits of cancer screening take years to appreciate

There has been much debate in recent years about the relative benefits and risks of screening. Debates about breast cancer screening have been particularly heated, even though such screening is recommended in most developed countries. Concerns raised in a recent major review of the benefits and risks of the UK breast screening programmes are also at the heart of discussions about prostate specific antigen testing, where there is evidence of benefit but also of considerable risks.1 2 In a linked research paper (doi:10.1136/bmj.e8441), Lee and colleagues consider how long it might take for the benefits of screening to show in a population invited for breast or bowel cancer screening.3

Where benefits are considered to outweigh risks then screening can be recommended. In developed countries and jurisdictions where a population model of health is delivered, screening for breast, cervical, and bowel cancer is often offered on an organised basis.4 5 Evaluation of the effectiveness of such programmes is demanded ethically and by the programmes’ funders, as well as by the populations screened and their doctors. Although it has never been evaluated by a randomised controlled trial (RCT), cervical screening has been shown to have a large effect, with five yearly screening probably preventing 63-73% of cervical cancers in women over 50 years.6 Evidence from RCTs suggests that an invitation for bowel or breast cancer screening prevents around 20% of deaths from these cancers.1 7 The effect of cervical screening can be shown more easily using observational data than can the effect of breast or bowel cancer screening, because the effects of breast and bowel cancer screening are much smaller.

Although population screening is offered as a public health measure, the benefits and risks affect individuals. At invitation for screening, many of the risks are immediately evident to the person concerned. These include the anxiety associated with participation, awaiting results, and referral for diagnostic investigation—which often proves that cancer is not present. The testing and diagnostic procedures, such as mammography or needle biopsy of the breast, or endoscopy, also have associated complications. These immediate risks can be readily quantified and affected people identified. By contrast, almost all the benefits of screening take years to appear, and we can never know for certain who exactly has benefited.

The main benefit sought from breast and bowel cancer screening is the reduction in mortality from that cancer. This takes years to accrue, partly because even after symptomatic diagnosis of these cancers people normally survive for several years, and partly because of the additional lead time applied by earlier diagnosis through screening. In randomised trials where the population is free of cancer at the start, randomisation should account for temporal changes in other factors. Benefit, if it occurs, can be seen more clearly in such trials, whereas in public health screening programmes, where there is no control group and the population includes people diagnosed with cancer before screening started, analyses based on time trends can be difficult to interpret.

Lee and colleagues used data from RCTs of breast and bowel cancer screening to estimate how long it takes for one death to be prevented for each 1000 people screened.3 They conclude that, for patients over age 50 years, the time lag for either of these screening programmes is around 10 years. It is important to acknowledge such a time lag, so that benefit is not sought too soon.

The authors’ choice of one in 1000 to determine what is worth while underlies their conclusions, but it is an arbitrary choice, and they assume that the benefit of screening is constant at all ages. The risk of developing and dying from breast or bowel cancer, however, increases with age, and the absolute benefit of breast cancer screening rises as women get older.8 In addition, the confidence intervals around Lee and colleagues’ estimates of “10 years” are wide, with a range of 5 to 16 years. This shows that the estimates are too uncertain to justify any recommendation.

Doctors advising individual patients and public health practitioners considering population screening programmes should take account of the different time scales for benefits and for risks to accrue. Likely life expectancy is only one of many factors that may play a part in determining how someone responds to an invitation or recommendation for screening. However, suggesting that a person’s life expectancy should be an arbitrary 10 years for them to benefit from screening is as misguided as saying that everyone will benefit.