A very large study finding that NSAIDs prevent macular degeneration, and why to distrust it

Design

retrospective propensity score-matched cohort study using the TriNetX multicentre electronic medical records database, 2015-2024

Population

634,794 patients prescribed NSAIDs matched 1:1 with 634,794 non-users on age, sex, race, smoking and comorbidities; mean age about 60

Primary outcome

cumulative incidence and hazard ratio of incident age-related macular degeneration

Effect

overall HR 0.58 (95% CI 0.56-0.59); 6 months 0.31 (0.27-0.36), 1 year 0.36, 3 years 0.42, 5 years 0.48; aspirin 0.72, non-selective agents 0.41

This retrospective cohort used the TriNetX electronic records database to match 634,794 people prescribed NSAIDs one-to-one against the same number who were not, on age, sex, race, smoking and selected comorbidities, and followed them for incident age-related macular degeneration between 2015 and 2024. NSAID users had substantially lower risk overall (HR 0.58, 95% CI 0.56 to 0.59), for both non-exudative (0.56) and exudative (0.62) disease, with aspirin at 0.72 and non-selective agents at 0.41.

The timing pattern is what gives the game away. The apparent protection was strongest at six months (HR 0.31, 95% CI 0.27 to 0.36) and weakened steadily to 0.48 at five years. A drug that genuinely prevented a slowly developing degenerative disease would not deliver a two-thirds risk reduction within six months and then lose ground; that gradient is the signature of a detection or ascertainment artefact rather than biology. People prescribed NSAIDs are people engaging with healthcare for musculoskeletal problems, and the comparison group — patients with no NSAID prescription at all — is not a healthy control but an unusual one.

So the honest summary for a patient who asks is that observational data show an association, that the pattern of that association is more consistent with how the data were collected than with a drug effect, and that nothing here justifies taking an NSAID to protect vision. The known gastrointestinal, renal and cardiovascular risks of long-term NSAID use are established; this benefit is not.

• Do not recommend NSAIDs for macular degeneration prevention on this evidence
• Note the effect was largest at six months and shrank with time — the wrong shape for a preventive effect
• The comparison group is people with no NSAID prescription in a decade, which is not a normal population
• Propensity matching on recorded variables cannot balance healthcare-seeking behaviour, which drives both exposure and diagnosis
• Established NSAID harms are unchanged by this and remain the relevant consideration for long-term use

The statistics, in plain English

With over a million patients, confidence intervals become extremely tight (0.56 to 0.59) — which measures precision, not correctness, and a large biased study gives a precisely wrong answer. The time gradient is the diagnostic clue: a real preventive effect on a disease that takes years to develop should strengthen with follow-up, not weaken from 0.31 at six months to 0.48 at five years. Propensity matching equalises the variables entered into it and nothing else, so systematic differences in how often each group visited a clinician — and therefore how often anyone looked at their retina — survive the matching intact.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.