Skip navigation

Evidence-based screening

Bandolier has identified a screening blacklist ( Bandolier 16 ) highlighting those screening tests on which there is little evidence of cost-effectiveness. What evidence is needed (both clinical and cost effectiveness) before introducing new screening tests? (see also Bandolier 19 )

Fundamental criteria

The fundamental criteria on which effectiveness of screening should be judged was set out as long ago as 1968 by Wilson and Jungner (WHO Public Health Papers 34). Briefly, a screening test should fulfil ten criteria, including these important points:-
  • The condition has a recognisable early phase and early treatment can be shown to improve prognosis.
  • Effective treatment is possible and available.
  • The test for the condition should be relatively simple, not harmful and acceptable to the patient.
  • The test should achieve a balance between false positives and false negatives which is related to the severity of consequences of wrong diagnosis both for the health care system and the patient.
  • Screening must be sustainable once introduced and not just part of a limited specific initiative.

Randomised trials

However, even these criteria do not go far enough as we must also ask "Does the test affect the clinical outcome(s) of the disease within a whole population who are at risk?" For example, can screening for aortic aneurysm affect mortality in the community, or can screening for Helicobacter pylori in primary care improve the detection rate of early gastric cancer?

Invariably the answer to these questions can only be found in lengthy, expensive, randomised trials. As the Bandolier blacklist highlights, in the current climate it is unlikely that any commissioner would or should accept the introduction of a screening procedure not validated by such trials.

Devil in the details

This, though, is not the end of the story. Randomised trials address clinical outcomes, not whether our health care system can administer a test successfully. Will patients accept the test, what about defaulters and recalls and are all patients at-risk being offered the test? The devil, as always, is in the details. Experience with both breast and cervical cancer screening should warn of the consequences of neglecting these simple questions. To use another example, screening for diabetic retinopathy is unlikely to be effective if a diabetic register has not been compiled and validated regularly.

It is little use waiting for the results of expensive RCTs before introducing screening if more mundane issues about delivery of care are not investigated with equal rigour. Screening can only be effective if all those at risk are offered a test and followed up appropriately.

Effective screening is as much a matter of good administration as it is of good medicine. For those contemplating introducing screenings two crucial questions should be answered:-
  • Can the proposed test be shown to be cost-effective - is it worth doing?
  • Can the test be delivered and the results responded to consistently for all those at risk - can it be done?

Dr Gill Grimshaw
Department of General Practice
University of Leicester

previous or next story in this issue