Design: retrospective clinical cohort study with three visual field progression methods and a requirement for agreement between two
Population: 186 eyes with failed trabeculectomy — 94 same-site repeat trabeculectomy, 92 Ahmed glaucoma valve — with ≥4 postoperative fields and ≥2 years follow-up
Primary outcome: visual field progression rate after surgery, with risk factors for deterioration
Effect: progression in 35% of valve vs 31% of trabeculectomy eyes (P = 0.6); median mean deviation −0.43 vs −0.29 dB/year (P = 0.9); trabeculectomy reduced pre- to postoperative rate more (P = 0.03); postoperative peak pressure raised deterioration risk 11% per mmHg

When a trabeculectomy fails, the choice is usually framed as repeat trabeculectomy against a drainage device, and the argument is made on pressure. A retrospective cohort followed 186 eyes with a failed trabeculectomy — 94 having same-site repeat trabeculectomy, 92 receiving an Ahmed glaucoma valve — all with at least four postoperative visual fields and two years of follow-up, assessing progression by mean deviation slope, Glaucoma Rate Index and pointwise linear regression, and calling an eye progressing only where two or more methods agreed.

Both procedures lowered pressure significantly, with trabeculectomy achieving lower pressure on fewer medications. Field progression was similar: 35% of valve eyes and 31% of trabeculectomy eyes progressed (P = 0.6), with median mean deviation decline of −0.43 dB/year and −0.29 dB/year respectively (P = 0.9). But in the subset with adequate preoperative fields, comparing each eye's rate before and after surgery, trabeculectomy reduced the rate of progression more than the valve did (P = 0.03). And one risk factor stood out: higher postoperative peak intraocular pressure carried an 11% increase in the risk of field deterioration per mmHg — peak, not mean.

Two things follow for clinic. First, a repeat trabeculectomy at the same site is a reasonable option after a first failure and should not be abandoned in favour of a tube by default. Second, and more widely applicable, the number to act on is the highest pressure the eye has seen since surgery, not the average of the clinic readings. An eye with a good mean and an occasional spike is not a well-controlled eye, and a record that only reports averages will hide exactly the pressure that is doing the damage.

Record and act on the peak postoperative pressure, not the mean of the visit readings.
Consider same-site repeat trabeculectomy after a first failure rather than defaulting to a tube.
Treat a single high reading between good ones as a finding, not as an artefact.
Schedule fields densely enough to detect a rate change — four tests over two years is the minimum this analysis required.
Where a spike is found, look for its cause: bleb failure, steroid response, or a missed dose at a specific time of day.

Why it matters

Postoperative pressure control is judged on averages, and the measurement that predicted field loss was the one nobody records.

The statistics, in plain English

The primary comparison is negative: 35% against 31% with P = 0.6, and median decline rates of −0.43 against −0.29 dB/year with P = 0.9, which means these two operations could not be distinguished on field progression in 186 eyes. The before-and-after comparison favouring trabeculectomy (P = 0.03) comes from a subset with sufficient preoperative fields and is a secondary analysis, so it should be held more loosely than the headline result. The peak-pressure finding is the most useful because it applies regardless of which operation was done — though it is an association in a retrospective cohort, and eyes that spike may differ in ways the analysis could not capture.