Commentary: Screening: a seductive paradigm that has generally failed us

Screening healthy people has face value and great public and political appeal. It looks so simple, and yet screening is fraught with difficulties. These start already with the terminology, and common slogans like, ‘Catch the disease early, before it has produced any symptoms!’ are misleading on two counts. First, disease means lack of ease, which is not what we understand by being healthy; but people who work with screening tend to forget that they deal with healthy people. For example, women being invited to mammography screening are often called patients in scientific articles. The second error is the assumption that the disease is caught early. That is rarely the case, and breast cancer is again a good example. If we assume that the growth rate for a particular cancer is constant, then the women have harboured the cancer for 21 years on average before it is large enough to be detected by mammography screening.1 Finding precursors to cancer is of course an entirely different matter. Screening with flexible sigmoidoscopy identifies polyps and vaginal smear finds carcinoma in situ. A third problem with screening is that it always causes harm. Sometimes it also leads to benefits, and sometimes the benefits are sufficiently large to outweigh the harms. The main focus in screening trials should therefore be to quantify the harms, but this has rarely been the case, if ever. Screening trials focus on disease-specific mortality, which may seem natural, but it is the wrong outcome. Screening leads to overdiagnosis, and interventions that are beneficial for real patients can be lethal for healthy overdiagnosed people. Radiotherapy of overdiagnosed women may kill at least as many as those who are spared dying from breast cancer by attending breast screening.2 Total mortality should therefore be the primary outcome in screening trials of mortality, and Saquib et al. report a systematic review in this issue of the journal that aimed at clarifying whether screening lowers total mortality for diseases that carry a high disease-specific mortality.3 They focused on cancer, cardiovascular diseases, type 2 diabetes and chronic obstructive pulmonary disease. They did not find any screening trials for hypertension or chronic obstructive pulmonary disease. Disease-specific mortality was reduced with ultrasound for abdominal aortic aneurysm in men, mammography for breast cancer and faecal occult blood test and flexible sigmoidoscopy for colorectal cancer, but the risk ratio point estimates for all-cause mortality were all very close to 1.00 (range 0.98–1.03). Screening proponents often say that disease-specific mortality is the right outcome, arguing that in order to show an effect on total mortality, trials would become unrealistically large. I believe this argument is invalid, for both scientific and ethical reasons. We do randomized trials in order to avoid bias, and our primary outcome should therefore not be a biased one. Drug interventions are usually more common in a screened group, and they tend to increase mortality for a variety of non-disease related reasons.4 From an ethical perspective, it is problematic to screen the whole population in a certain age group without knowing whether this makes people live longer, while knowing almost certainly that it makes people less happy. It took 50 years after the first randomized trial of mammography started before we knew what the psychological consequences are of the many false-positive findings.5 A specially designed questionnaire was developed using focus groups and women who had attended screening were followed up for 3 years. Even after so long a time, those who had experienced a false-positive diagnosis had an anxiety level (and other psychological problems) that fell between that for women with breast cancer and women who had always been told they did not have cancer. This study showed for the first time that the psychological harms of breast screening are substantial and long-lasting, and they affect a huge number of healthy women, as the cumulative risk of a false-positive result after 10 mammograms ranges from about 20% to 60%.6 Added to this comes the psychological harm inflicted on all the overdiagnosed women who do not know that they are overdiagnosed but think that they suffer from a fatal disease. It is therefore pretty clear that any utility analysis that takes the psychological harms of breast screening into account will come out negative, as was recently reported by the Swiss Medical Board.7 Saquib et al. found no screening trials for hypertension and only one for diabetes, ADDITION-Cambridge, for which the risk ratio for all-cause mortality was 1.06. In our systematic review of general health checks,8 7 of the 16 trials screened for diabetes, and likely all of them screened for hypertension (in one, the screening tests were not specified). Although we had 11 940 deaths, we did not find an effect on total mortality (risk ratio 0.99, 95% confidence interval 0.95 to 1.03). We could not include the most recent trial, as it was published in 2014.9 It investigated the effect of systematic screening for risk factors for ischaemic heart disease and lifestyle counselling. This trial was large but it also failed to find an effect on total mortality: 3163deaths occurred, and the hazard ratio was 1.00 (0.91 to 1.09). It is worth noting that when screening does not work, it might be because beneficial effects are outweighed by harmful ones. Diabetes drugs, for example, are approved on the basis of their glucose-lowering effect without knowing what they do to patients. And the only large trial of tolbutamide ever performed was stopped prematurely because the drug increased cardiovascular mortality.4 Rosiglitazone was once the most-sold diabetes drug in the world, but it was taken off the market in Europe in 2010 as it causes myocardial infarction and cardiovascular death; and pioglitazone has been linked to heart failure and bladder cancer.4 Screening is popular, but we need to be much more careful in the future when we contemplate approaching healthy people with our screening tests, and should demand much stronger evidence than when we treat patients. Conflict of interest: None declared.