Not every glaucoma suspect needs to be seen every year

Design

retrospective cohort study using a de-identified US claims database, with Cox proportional hazards modelling and subgroup risk estimation

Population

83305 patients newly diagnosed as open angle glaucoma suspects between 2007 and 2021, with 3-year lookback and 5-year follow-up

Primary outcome

diagnostic conversion to primary open angle glaucoma

Effect

20.6% converted over 5 years, 6.1% per year (9.4% year 1, 5.3% years 2-5); 2.0% per year in untreated under-50s vs 16.7% in treated over-70s

Glaucoma suspect follow-up consumes a large share of clinic capacity, and the interval is usually set by habit. This retrospective cohort used a US claims database to measure what actually happens: 83305 patients with newly diagnosed open angle glaucoma suspicion between 2007 and 2021, all with continuous enrolment for three years before and five years after diagnosis, and all with at least one OCT or visual field test.

Over five years, 17134 (20.6%) converted to primary open angle glaucoma - an annual rate of 6.1%, front-loaded at 9.4% in the first year and 5.3% thereafter. On multivariable regression, conversion was associated with older age (hazard ratio 1.38 or above), male sex (1.12), Black race (1.14), location outside the Northeast (1.24 or above), a record of gonioscopy (1.90) and being treated as a suspect (1.31 or above).

The useful output is the stratification. The lowest-risk group - under 50 and untreated - converted at 2.0% a year; the highest - over 70 and treated - at 16.7%. Standardised to a common 5% per-visit conversion threshold, that implies monitoring every 6.1 years for the lowest-risk patients and every 0.6 years for the highest, against the roughly annual review both currently receive. The signal to act on is not the exact interval, which comes from claims data with all the misclassification that implies, but the spread: the same follow-up rhythm is serving two groups whose risk differs eight-fold. Lengthening the interval for the youngest untreated suspects is where clinic capacity is hiding.

• Set the follow-up interval by risk, not by convention - conversion risk differs about eight-fold across suspect subgroups.
• Front-load surveillance: conversion was 9.4% in year one and 5.3% thereafter.
• The lowest-risk group, under 50 and untreated, converted at 2.0% a year and is where intervals can safely lengthen.
• Read 'record of gonioscopy' and 'being treated' as markers of a clinician already suspicious, not as causes of conversion.
• These are claims-based diagnoses, so conversion means a coded diagnosis rather than an adjudicated one - use the stratification, not the precise intervals.

The statistics, in plain English

Two of the strongest predictors are markers of clinical suspicion rather than biology: a record of gonioscopy (hazard ratio 1.90) and being treated as a suspect (1.31 or above) both indicate that a clinician already thought this patient was likely to convert. That is confounding by indication, and it inflates those hazard ratios relative to any causal reading. The outcome is a coded diagnosis in a claims database, so 'conversion' includes reclassification by a new clinician as well as genuine progression. The stratified conversion rates are the durable part, because they describe what happened in each group rather than attributing it; the derived monitoring intervals of 6.1 and 0.6 years are a modelled illustration of a 5% per-visit threshold, not a recommendation validated against outcomes.

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.