Interpreting a screening pilot for abdominal aortic aneurysm
A commissioning group pilots ultrasound screening for abdominal aortic aneurysm (AAA) in men aged 65 in a region of 50,000 eligible men. Over one year, 20,000 men attend. The aorta is ≥3.0 cm in 400 men (2.0% of attenders). Of these, 40 have an aneurysm ≥5.5 cm and are referred for surgery; 6 die within 30 days of operation. Over five years of follow-up, 30 men in the screened cohort die of ruptured AAA. In the unscreened comparison population (matched, 50,000 men), the five-year rupture death rate is 0.20%. The pilot costs £1.1 million in year one. A local clinician argues the programme 'does more harm than good' because of the operative deaths.
1. Calculate the key performance measures of this screening test and interpret them, stating any assumptions.
Read the worked answer
Data. Attenders 20,000; screen-positive (AAA ≥3.0 cm) 400 → prevalence 2.0%. Referred for surgery (≥5.5 cm) 40; 30-day operative deaths 6.
Positive predictive value. PPV = true positives / all screen positives = 400/400 = 100% if we define 'positive' as AAA ≥3.0 cm and treat ultrasound as the reference standard. This is circular - ultrasound is the diagnostic test, so PPV is not the right metric here. The meaningful figure is the detection rate: 2.0% of attenders had an AAA, consistent with expected prevalence in 65-year-old men (roughly 1.5–2.5%).
Operative mortality. 6/40 = 15% 30-day mortality among those operated on. This is high compared with contemporary elective AAA repair (typically 1–3% in high-volume centres), suggesting a case-mix or quality issue worth reviewing.
Assumptions. Attendance 40% (20,000/50,000) - low uptake biases everything; attendees are healthier than non-attendees (healthy screenee effect).
Interpretation. The test 'works' in the sense that it finds aneurysms at expected prevalence, but the programme's value depends on outcomes, not detection.
2. Using the comparison data, estimate the absolute effect of screening on five-year AAA rupture mortality and comment on whether the pilot 'does more harm than good'.
Read the worked answer
Screened cohort. 30 rupture deaths in 20,000 attenders over five years = 0.15%.
Unscreened comparison. 0.20% over five years.
Absolute risk reduction = 0.20% − 0.15% = 0.05 percentage points, i.e. about 5 per 10,000 over five years. Number needed to screen to prevent one rupture death ≈ 1/0.0005 = 2,000 men over five years.
Harm side. 6 operative deaths in 40 operations = 6 per 20,000 screened = 0.03% (3 per 10,000). So the mortality harm from surgery is of similar order to the mortality benefit - the net benefit is small and fragile.
Caveats.
- The comparison is not randomised; confounding and the healthy screenee effect can inflate apparent benefit.
- Rupture deaths are only part of the picture - screening also prevents emergency surgery and morbidity, and detects aneurysms that never rupture (overdiagnosis, surveillance burden, anxiety).
- The high operative mortality (15%) may be a local quality signal, not an inherent property of screening.
Verdict. The clinician is wrong to conclude 'more harm than good' from operative deaths alone - the arithmetic shows a net mortality benefit, though a small one. But the high operative mortality and low uptake mean the programme should be reviewed, not simply continued: centralise surgery, audit outcomes, improve uptake, and reconsider whether benefits exceed harms locally. This is a judgement call, not a calculation.