Premature introduction of screening: Politicians can play a part in initiating inappropriate screening programmes. UK prime minister, Margaret Thatcher sanctioned nationwide breast screening in 1988, 2 weeks before a general election; some believe this was to garner the women’s vote. The decision was premature and based on insufficiently validated evidence from the Swedish two counties study and the UK Forrest report. In the 1970s, screening for cancer of the uterine cervix was widely introduced, also before its efficacy had been fully evaluated. Fortunately, it has proved successful despite the difficulty of engaging women at high risk. Sadly, the natural history of untreated dysplastic cervical cellular abnormalities was not properly established before the impact of widespread screening made this ethically impossible. This severely hampered scientific study of the disease, its early diagnosis and best treatment.

Criteria for assessing a screening programme: Many years ago the World Health Organization (WHO) realised that even beneficial screening could be expensive, unpleasant, inaccurate and unproductive, and could adversely affect psychological or physical well-being. In 1968, they published a list of criteria for effective screening programmes (Box 6.2) including attributes of the disease, the test and the treatment. These principles are still relevant today and have been added to by the UK National Screening Committee and other groups. Box 6.3 shows a summary of these modified criteria.

Evolution of screening programmes: Once begun, any screening programme must remain under constant evaluation and modified or discontinued when criteria are no longer being met. For example, in the 1950s and 1960s, screening for pulmonary tuberculosis (TB) by mass miniature chest X-ray was highly successful but it was disbanded in the 1970s when new cases fell below a level at which the unit cost per new case could be justified; interestingly, by then, the yield of new cases of lung cancer from screening began to exceed that of TB, but there was virtually no effective treatment for it at the time.

The disease: The screened condition should be an important health problem either because it is common (such as lung or prostate cancer) or has serious but preventable consequences such as carotid artery disease or abdominal aortic aneurysm (AAA). The prevalence (proportion of cases already in a population) and the incidence (the proportion of new cases) of the disease in the population at risk are discovered from pilot studies. There should be a truly early stage where treatment outcomes are better than at a late stage. Colorectal adenomas and early cancers are good examples.

The biological behaviour or natural history of the disease should be well understood, including how latent disease progresses to clinical disease, and the disease course should be reasonably predictable. For example, AAAs are known to expand smoothly for the most part and rarely rupture until they are large. The risks of untreated disease also need to be understood and there should be a long period between the first detectable stages and overt disease.

The diagnostic test: The test must be valid, i.e. reliable in detecting the disease. This is defined by sensitivity and specificity. Sensitivity is the capability of the test to identify affected individuals in the screened population, i.e. the proportion of people who have the disease and are detected. A test with many false negative results is insensitive and unreliable. A UK appeal court ruled that sensitivity is paramount in (cervical) screening and awarded damages to women with missed diagnoses at screening. Specificity is the degree to which a positive test can be relied upon to prove the disease is present; in other words, the higher the false positive rate, the lower the specificity.

The test must be simple and cheap and it must identify the disease by a reliable, validated and reproducible method. The distribution of test values should be understood well enough to define normality or relative risk associated with particular stages, e.g. for an AAA, the diameter with a high risk of rupture. Intervals for repeating the test should be worked out for normal subjects and for those with positive results near the threshold.

The complete screening programme must be clinically safe and acceptable socially and ethically to health professionals and the public. This includes the test and any diagnostic procedures, treatments or interventions screening initiates. For example, if a test is perceived as unpleasant, e.g. colonoscopy, uptake is low and the benefits are proportionately smaller.

The overall benefits should be greater than the risks; this includes any physical and psychological harm caused by the test, diagnostic procedures and treatment.

Diagnosis and treatment: Cases identified by the test must be amenable to effective, acceptable and safe diagnostic procedures and the potential benefits of medical or surgical intervention prompted by earlier diagnosis need to be understood.

There should be clear evidence from high-quality randomised controlled trials (RCTs) that early treatment produces better outcomes than treatment at later stages, i.e. ‘cure’ should be more likely, survival longer, or earlier treatment easier. Treatment should have minimal side-effects. There also needs to be agreement in advance about who should be offered treatment and its nature; this evidence may emerge from RCTs.