The Field That Doesn't Match Its Own Reliability Indices

Fixation losses 4/12 (33%) - flagged by the >20% heuristic, but in practice fixation losses are the weakest of the three: the blind-spot check can be triggered by a slightly off-center starting position, a small head turn, or an inexperienced patient, and a genuine arcuate scotoma does not result from wandering gaze.

False positives 18% - this is the index that should worry you most here, and it is also the one most often ignored. False-positive responses are presses to nothing; the perimeter counts them as seeing. That artificially lifts the hill of vision and can flatten or partially fill in a real defect. In other words, the defect displayed is most likely no better - and possibly worse - than what is shown. FP error cannot manufacture an anatomically coherent inferior arcuate respecting the nasal raphe.

False negatives 33% - failures to respond to very bright stimuli. These exaggerate defect depth and can make MD look worse than the truth. Some FN are expected in a patient with an MD of −6.4 dB, because part of the test is being presented inside a deep scotoma.

Net reasoning: FP and FN push in opposite directions, and neither produces a meridian-respecting arcuate. Combined with the disc/rim correlation and the 3.3 dB decline from a prior reliable test, the defect is credible enough that it should not be dismissed as 'unreliable test, ignore.'

1. Keep the current printout - an unreliable test still contains data and is part of the record; never discard or re-run as if the first never happened.
2. Give the patient a break and re-instruct - explain that the test measures the dimmest light he can see and that guesses count against him. Directly address the false positives ('only press when you actually see the light').
3. Correct the refraction properly - trial-lens the distance prescription, verify the near add for the bowl distance, and confirm the lens is not fogging or decentered. Uncorrected cylinder is a common cause of a depressed, noisy field.
4. Optimize mechanics - patch the non-tested eye fully, seat height, chin rest, steady gaze on the fixation target, technician in the room watching the eye rather than the screen.
5. Repeat with SITA Standard (or 24-2C if macular involvement is a question) - slower and more sampling, but usually more reliable; reserve SITA-Fast for patients with proven stable reliability.
6. Correlate structurally - pull the disc photos/OCT RNFL and compare the defect's location and depth with the rim loss.

Reasoning: the goal is a defensible data point today, because the physician is deciding whether this is progression. Fixing the testing conditions is faster and cheaper than scheduling a return visit, and a second, cleaner field in the same session lets him judge agreement.

• A single flagged test is not a diagnosis in either direction. Declaring it 'false' because of the FP rate throws away a defect that fits the disc; declaring progression on one test commits the patient to a possible escalation of therapy (or surgery) on data with 33% FN.
• Progression is a trend, not an event. Glaucoma progression is conventionally judged on repeatable change across a series - typically at least three fields over time - or by event analysis against baseline, plus structural confirmation. Test-retest variability is largest exactly where this patient's defect sits (the steep slope at the edge of a deep scotoma), so a single 3 dB drop is within noise.
• What settles it: a reliable repeat field in the near term (weeks, not 6 months), agreement between the field location and the RNFL/rim location on OCT or photos, and IOP/target-pressure history. If the repeat is clean and concordant, treat it as real; if the repeat normalizes and structure is stable, the first was artifact.

Key judgment: act on concordance, not on a lone number - and document the reasoning so the next technician understands why the field was repeated.