Screening attempts to detect a disease in an early, curable state. There are guidelines published by the WHO regarding what makes the ideal screening test based on the condition, the test, the treatment and the program:
- The condition should be an important health problem
- The condition should have a detectable latent or early phase
- The progression of the condition from latent to advanced stages should be understood
- There should be a suitable test
- The test should be acceptable to the population
- There should be an accepted treatment for patients with the latent/early stages of disease
- There should be a policy on which patients are screened and treated
- Facilities for diagnosis and treatment must be available
- The cost (including detection, treatment and follow up) should be balanced against the health budget as a whole
- Screening should be an ongoing process and not a single action to detect all disease at a particular time
In Australia, there are two additional guidelines:
- There should be evidence supporting the diagnostic test and the treatment for the disease
- The screening program should provide a net benefit to harm ratio for the population
I have divided screening into several categories below.
Studies that examine the benefits of screening are prone to several biases:
- Lead time bias: In some cases, screening will pick up a cancer before it becomes symptomatic. If there is effective treatment, then mortality will be significantly improved. If there is ineffective treatment, then mortality will be the same as if it been picked up when symptomatic; the problem arises that patients will have a longer apparent survival due to the effect of picking up the cancer early (i.e. the patient has the diagnosis for longer) rather than the treatment having any impact.
- Length time bias: Within a particular tumour classification, some tumours will be inherently more aggressive and progress more quickly whereas others progress more slowly. This is a potential problem in breast cancer screening, where screening will detect the majority of women with slow growing, readily curable malignancies but miss the more aggressive cases that arise over a matter of weeks and metastasise rapidly.
- Compliance bias: Patients who volunteer for screening trials are more likely to be health conscious and follow recommended treatments; they will therefore tend to have better outcomes than those not included in trials. This may bias the results in favour of screening
These biases are minimised by using a randomised controlled trial (avoids compliance bias) that fully evaluates all deaths in both groups (avoids length time bias) with long term mortality follow up (avoids lead time bias).