Skip to content
Scintria
ToolsGrant writerFunding
HomeLiteratureClinical trialsEvidence LabProtocol packData analysisLearnProjectsLibraryReferencesApplications
Start free
ScintriaStart free

Learn / Epidemiology / Screening and diagnostic tests

Principles of screening and its hidden biases

14 minute read · 6-question quiz

Download this lesson as a colour Word or PDF study guide with the figures and answers. Included in paid plans; reading and quizzes stay free.Sign in

What you will learn

  • State the Wilson and Jungner criteria and apply them to a proposed programme

  • Explain lead-time, length and overdiagnosis biases

  • Calculate the yield and false-positive burden of a screening programme

Screening is the application of a test to people without symptoms to identify those at increased risk of a disease, so that early treatment improves their outcome. It differs from diagnosis, which investigates people with symptoms. Screening harms healthy people if done badly: false alarms, anxiety, unnecessary treatment.

The Wilson and Jungner criteria (WHO, 1968, summarised)

The condition

An important health problem, with a recognisable latent or early symptomatic stage and a known natural history.

The treatment

An accepted, effective treatment; earlier treatment must do better than treatment at the usual time. Facilities for diagnosis and treatment are available.

The test

Suitable, simple, safe and acceptable; adequate sensitivity and specificity; an agreed policy on whom to treat.

The programme

Cost balanced against benefit; case-finding is a continuing process, not 'once and for all'.

Worked example: a screening programme in 10,000 adults

Disease prevalence is 1%; the test has sensitivity 85% and specificity 95%.

DiseaseNo diseaseTotal
Test positive85495580
Test negative159,4059,420
Total1009,90010,000
  • The programme finds 85 of 100 cases, and misses 15.
  • It produces 495 false alarms: more than five false positives for every true case.
  • PPV = 85 / 580 = 14.7%: only about 1 in 7 positive results is a true case.
  • NPV = 9,405 / 9,420 = 99.8%: a negative result is very reassuring.

Those 580 people need a confirmatory test; each follow-up carries cost, anxiety and sometimes harm (a biopsy). The test is 'good', yet most positives are false. This is why screening is justified only when the disease is serious, early treatment helps and the confirmatory pathway is safe and available.

10,000Screened580Test positive85Confirmed diseaseunder 85Benefit from early treatment
From screened population to people who benefit: each step narrows.

Why screening studies are hard to interpret: three biases

Lead-time bias. Screening finds the disease earlier but the person may die at the same time. Survival measured from diagnosis then looks longer without any real benefit. Suppose a cancer is found by screening at age 60 and the patient dies at 66 (survival 6 years). Without screening, the same cancer is diagnosed by symptoms at 63 and the patient still dies at 66 (survival 3 years). Treatment gained nothing; the clock simply started earlier. The antidote: compare mortality in the screened and unscreened populations, not survival times.

Length bias. Screening, performed at intervals, is more likely to catch slow-growing, less aggressive disease (it spends longer in the detectable phase) than fast-growing disease that appears between screens. Screen-detected cases therefore look to have a better prognosis.

Overdiagnosis. Screening can detect 'disease' that would never have caused symptoms or death in the person's lifetime (some prostate and thyroid cancers). Treating it adds harm and no benefit. Overdiagnosis is invisible in an individual and only revealed by population comparisons.

Age 60Screening detectscancerAge 63Symptoms: clinicaldiagnosis withoutscreeningAge 66Death in bothscenarios
Lead-time bias: survival from diagnosis is 6 years with screening and 3 without, yet life is not prolonged.

The right outcome for a screening trial

Only disease-specific mortality (and all-cause mortality) in a randomised comparison of screening vs no screening can show benefit free of lead-time, length and overdiagnosis bias. Observational comparisons of 'screened vs unscreened' also suffer healthy-screenee bias: those who attend are healthier.

Harms are part of the assessment

Report false positives, overdiagnosis, complications from follow-up tests and psychological impact alongside lives saved. Participants need balanced information to make an informed choice.

Examples to know

Well-established: cervical screening (HPV testing or cytology) and treatment of precancer, bowel cancer screening, newborn screening for some metabolic disorders, antenatal HIV and syphilis testing. In Uganda, screen-and-treat for cervical precancer (VIA/HPV testing and thermal ablation) is a priority for HIV-positive women.

Summary

  • Screening tests asymptomatic people to find disease earlier; harms to healthy people must be weighed.

  • Wilson and Jungner criteria cover the condition, treatment, test and programme.

  • Even a good test has a low PPV when prevalence is low.

  • Lead-time, length and overdiagnosis biases make survival after diagnosis unreliable.

  • Mortality in a randomised comparison is the valid outcome of a screening programme.

Check your understanding

Question 1
1

Screening differs from diagnosis because it is applied to:

Question 2
2

With 1% prevalence, sens 85% and spec 95% in 10,000 people, false positives are:

Question 3
3

Lead-time bias makes screening appear to:

Question 4
4

Screen-detected cases having a better prognosis because slow-growing disease is more likely to be caught is:

Question 5
5

Overdiagnosis refers to:

Question 6
6

The best outcome for evaluating a screening programme is:

Draft study guide. Spotted an error or something unclear? Tell us at hello@scintria.com. This is educational material, not clinical or regulatory advice.

From association to causationValidity and reliability of tests