Humphrey Field Analyzer 24-2/10-2 SITA testing — tracking glaucomatous scotoma development and structure-function correlation with OCT RNFL
The Humphrey Field Analyzer (HFA, Carl Zeiss Meditec) is the worldwide clinical standard for quantifying the visual field — the three-dimensional "hill of vision" whose peak sensitivity sits at fixation and slopes toward the periphery. Standard automated perimetry (SAP) measures differential light sensitivity at fixed retinal locations, converting a subjective perceptual task into a reproducible, normative-database-referenced number in decibels at each of dozens of points.
Full-threshold testing (the original Humphrey method) uses a 4-2 dB staircase bracketing procedure at every point independently, taking 12–15 minutes per eye and fatiguing patients enough to degrade reliability late in the test. SITA (Swedish Interactive Threshold Algorithm, Bengtsson & Heijl, 1997) replaced this with a Bayesian approach:
• A prior probability distribution of likely threshold at each location is built from a normative population, then continuously updated using a maximum-likelihood estimator as the patient responds • Information from neighboring points and from the growing pattern of responses across the whole test informs the algorithm in real time, not just point-by-point • Testing stops at each location once the estimated threshold's confidence interval is sufficiently narrow, rather than after a fixed number of reversals • SITA-Standard: ~5–6 minutes/eye (roughly 50% faster than full-threshold) with equivalent test-retest variability • SITA-Faster (2019): ~3 minutes/eye using a single-reversal strategy and reduced initial presentations, now standard of care for stable, cooperative patients
Stimulus size III (Goldmann, 4 mm² at 33 cm) on a 31.5 apostilb white background is standard; stimulus intensity ranges 0–51 dB (0 dB = brightest, 10,000 apostilbs; higher dB = dimmer, harder to see).
The 24-2 pattern places 54 test locations on a grid offset 3° from the vertical and horizontal meridians, spanning the central 24° (extending to 30° nasally), deliberately avoiding the fovea and physiologic blind spot while sampling the arcuate nerve fiber bundle trajectories most relevant to glaucoma.
Each completed test yields two global indices, referenced against age-matched normal eyes:
• Mean Deviation (MD): the average difference between the patient's measured sensitivity and the normal age-corrected value across all 54 points, weighted for the variance at each location. A single number in dB summarizing overall field health — negative values indicate loss. MD is sensitive to diffuse loss (cataract, generalized reduction) as well as focal defects.
• Pattern Standard Deviation (PSD): quantifies the irregularity of the hill of vision after removing overall (generalized) depression — it rises when SOME points are much worse than others, even while MD remains near-normal. PSD is therefore the more sensitive index for early, focal glaucomatous loss, often abnormal (p<5%) years before MD crosses into abnormal range.
Reliability indices — fixation losses (patient responds to stimuli aimed at the physiologic blind spot, meaning they are not fixating), false-positive rate (responding when no stimulus was presented), and false-negative rate (missing a stimulus brighter than a previously-seen threshold at the same point) — must all fall within acceptable limits or the test is flagged unreliable and should be repeated.
The Humphrey Field Analyzer, introduced in 1984, replaced manual kinetic Goldmann perimetry (a technician manually moving a light target and plotting isopters by hand) with fully automated, computer-controlled static threshold testing — a shift that, combined with normative databases, made visual field results comparable across clinics, countries, and decades in a way manual perimetry never could.
The nasal step is often the earliest detectable glaucomatous field defect: a horizontal, step-like difference in sensitivity between the superior and inferior nasal field, sharply demarcated at the horizontal midline. Its shape is not arbitrary — it is a direct printout of retinal nerve fiber layer anatomy, because RNFL bundles arcing from the temporal retina toward the optic disc do not cross the horizontal raphe.
Retinal ganglion cell axons converge on the optic nerve head following stereotyped arcuate paths: fibers from the superior retina sweep in an arc above fixation to enter the superior pole of the disc; fibers from the inferior retina mirror this below. Crucially, these two fiber populations essentially never cross the horizontal raphe — the anatomical seam running temporal-to-nasal through the fovea.
Because glaucomatous damage occurs at the level of the optic nerve head (axonal compromise at the lamina cribrosa), and because a damaged bundle carries information from only one side of the raphe, the resulting field defect is bounded by that same horizontal line. A nasal step therefore looks like a sharp horizontal "cliff" in sensitivity confined to the nasal field, with normal sensitivity on the other side of the midline at the same eccentricity.
This respect for the horizontal midline is a key clinical discriminator: neurologic field defects (from chiasmal or retrochiasmal lesions) instead respect the VERTICAL midline, because information from each hemifield is processed by the contralateral hemisphere. A field defect obeying the horizontal raphe points to the eye and optic nerve; one obeying the vertical meridian points to the brain.
A defect only becomes detectable on standard automated perimetry once a substantial fraction of the retinal ganglion cell population subserving that location has already been lost. Landmark histologic work by Kerrigan-Baumrind and colleagues (2000), correlating post-mortem axon counts with antemortem visual fields in monkey glaucoma models, estimated that approximately 25–35% of retinal ganglion cell axons must be lost before a reproducible defect appears on standard perimetry — and some estimates from earlier work (Quigley, Sommer) suggested this threshold could be as high as 40–50% in certain regions.
This "pre-perimetric" or "structure precedes function" window is why optical coherence tomography (OCT) of the retinal nerve fiber layer has become central to modern glaucoma diagnosis: RNFL thinning can be detected years before a corresponding functional defect crosses the threshold of standard perimetry's sensitivity, offering a critical head start for treatment. This same principle — structural change outrunning measurable function — recurs throughout the disease course and is revisited quantitatively in the final stage of this simulation, where field and OCT data are integrated directly.
As bundle damage progresses beyond a single hemifield step, it traces the full curved trajectory of the arcuate retinal nerve fiber bundle — producing a comma- or sickle-shaped scotoma that arcs from the blind spot, sweeps around fixation within the Bjerrum area (10–20° eccentricity), and terminates at the nasal horizontal raphe. This shape is so characteristic that its presence is almost pathognomonic for glaucomatous — as opposed to other causes of — optic neuropathy.
The arcuate nerve fibers originate from ganglion cells in the temporal retina and sweep in a gentle curve above or below the fovea (never crossing the temporal horizontal raphe extending nasally from fixation to the disc) before converging on the superior or inferior pole of the optic nerve head — the two regions of the disc rim that are, not coincidentally, most vulnerable to glaucomatous cupping because of the biomechanics of the lamina cribrosa's connective tissue architecture.
When an entire bundle is compromised, the resulting scotoma follows this same arc: beginning near the physiologic blind spot (15° temporal to fixation), curving around the macula within the Bjerrum area, and fading into a nasal step as it approaches the horizontal raphe. Isolated arcuate scotomas are usually asymptomatic — patients rarely notice them subjectively because the fellow eye and cortical filling-in compensate — which is precisely why routine perimetric screening, not symptom-triggered testing, is essential for early glaucoma detection.
Two complementary output formats characterize each visual field printout at this stage:
• The grayscale map: a purely visual, non-statistical representation shading each tested location from white (normal) to black (non-seeing), useful for rapid pattern recognition but influenced by media opacities like cataract
• The total deviation and pattern deviation probability plots: point-by-point statistical comparison against the age-matched normative database, displayed as symbols (·, ·, °, ×, ▪) coding significance from p<5% down to p<0.5%. The PATTERN deviation plot statistically removes any generalized (diffuse) depression first — for example from early cataract — isolating focal loss and making it the most sensitive map for detecting exactly this kind of arcuate defect
A cluster of ≥3 contiguous points depressed at p<5% on the pattern deviation plot, with at least one point at p<1%, meeting the Anderson-Patella criteria, is the standard research and clinical definition of a "glaucomatous visual field defect" — the threshold this simulation's arcuate stage is built to represent.
The Ocular Hypertension Treatment Study (OHTS, 2002) followed over 1,600 patients with elevated eye pressure but normal fields for years, finding that topical medication reducing intraocular pressure by ~20% cut the 5-year risk of developing a glaucomatous field defect roughly in half (from 9.5% to 4.4%) — direct proof that intervention before a visible arcuate scotoma forms materially changes the disease trajectory.
For decades, glaucoma was framed as a disease of peripheral vision, sparing central acuity until very late. That view has been substantially revised: a large fraction of patients, especially those with normal-tension glaucoma or myopia, develop paracentral defects — scotomas within 10° of fixation — early in the disease, sometimes before more peripheral field loss is obvious. Because standard 24-2 testing places only 12 of its 54 points within the central 10°, these defects can be under-sampled, motivating dedicated 10-2 testing.
Retinal ganglion cell density is extraordinarily non-uniform: roughly half of all ~1.2 million ganglion cells in a human retina serve the central 8–10° of visual field, packed into the parafoveal ring around the avascular fovea itself. A defect that removes only a small absolute NUMBER of cells here can therefore knock out a disproportionately large fraction of central function compared with an equivalent cell loss in the sparse peripheral retina.
Paracentral scotomas cluster inferonasally more often than any other single location — a pattern increasingly linked to disc hemorrhages, beta-zone parapapillary atrophy, and focal lamina cribrosa defects concentrated at the inferotemporal disc margin. Unlike more peripheral arcuate defects, paracentral loss can measurably impair reading speed, contrast sensitivity for faces, and fine visuomotor tasks — explaining why patient-reported quality of life correlates more strongly with central 10° field status than with the peripheral field alone, even at matched overall MD.
The 10-2 test pattern places 68 locations at 2° intervals across the central 10° radius — a spatial resolution three times finer than 24-2's 6° spacing in the same region. This matters because a small, deep scotoma can fall entirely between two 24-2 test points and be completely missed, while still being readily detected by the denser 10-2 grid.
Current clinical guidance (endorsed by multiple glaucoma subspecialty societies) recommends adding 10-2 testing whenever: (1) 24-2 MD is worse than roughly -6 dB, (2) any paracentral point within the central 24-2 locations is abnormal, or (3) structural OCT shows macular ganglion cell complex thinning — precisely the profile represented by this simulation stage. Missing an active paracentral scotoma has direct consequences: patients can fail to notice a defect straddling their reading fixation point because the fellow eye compensates, right up until bilateral disease removes that safety margin.
A single visual field is a snapshot; glaucoma management depends on the trajectory across many snapshots over years. Zeiss's Guided Progression Analysis (GPA) software formalizes two complementary statistical approaches — event-based analysis, which flags individual points that worsened beyond measurement noise, and trend-based analysis, which fits a regression line to Mean Deviation over time to estimate a rate in dB per year.
Event-based GPA compares each follow-up field to the average of two baseline exams, point by point. Because test-retest variability at any single location can itself exceed several dB — especially at already-damaged points, where variability is paradoxically HIGHEST — a naive point-to-point comparison would generate enormous false-positive "progression" noise.
GPA instead uses empirically-derived test-retest confidence limits (from large normative repeatability studies) specific to each starting sensitivity level, flagging a point only if it worsens beyond that point-specific noise threshold:
• "Possible progression": ≥3 points worsened beyond limits on a single follow-up field • "Likely progression": the same ≥3 points worsened on two consecutive follow-up fields — repeatability is what converts a possibly-spurious blip into a confirmed event
This two-tier system deliberately trades some sensitivity for specificity, because triggering an unnecessary escalation of glaucoma treatment (more drops, laser, or surgery, each with real side-effect burden) on a false alarm carries real cost.
Trend-based analysis performs linear regression of Mean Deviation (or VFI) against time across all available fields, yielding a slope in dB/year with a confidence interval and p-value for whether that slope differs from zero.
Population-level data (Heijl et al., Investigative Ophthalmology & Visual Science, 2009, following >2,000 field series) found untreated or under-treated glaucoma progresses at a median rate around -0.5 to -1 dB/year, but with wide variance: roughly 12% of eyes were "fast progressors" losing more than 1 dB/year, a rate that — left unchecked — can produce disabling field loss within a single decade. In contrast, healthy aging alone accounts for only about -0.05 to -0.1 dB/year of MD decline from lens and neural senescence.
Because VFI weights central points more heavily and is less confounded by cataract-driven diffuse loss than MD, many clinicians now track BOTH a Mean Deviation slope for overall sensitivity trend and a VFI slope as a functionally-weighted, cataract-resistant progression metric — and a linear extrapolation of either slope is used to estimate a patient's "time to blindness" at the current rate, directly informing how aggressively intraocular pressure needs to be lowered.
A meaningful fraction of "rapid progressors" identified by Heijl and colleagues were losing field fast enough that, without treatment intensification, simple linear extrapolation predicted disabling bilateral field loss (VFI approaching the tunnel-vision range) within the patient's expected remaining lifetime — the single strongest argument for calculating, not just eyeballing, a numeric progression rate at every glaucoma visit.
In advanced glaucoma, contiguous arcuate, nasal step, and paracentral defects merge, leaving only a small central island and a crescent of temporal field — classic "tunnel vision." At this stage, optical coherence tomography of the peripapillary retinal nerve fiber layer, plotted as a circular TSNIT (temporal-superior-nasal-inferior-temporal) profile around the optic disc, shows severe circumferential thinning whose spatial pattern mirrors the functional scotoma almost point for point.
Naively, one might expect RNFL thickness (a structural measure) and MD or VFI (functional measures) to fall in lockstep. They do not. In the earliest disease stages, large amounts of RNFL and ganglion cell loss occur while perimetric indices remain within normal limits — the pre-perimetric window discussed earlier. But once a "floor" of remaining axons is reached, the relationship inverts: OCT thickness measurements plateau near a noise floor (residual glial tissue and blood vessels contribute residual reflectance even with zero functioning axons) while perimetric sensitivity continues to fall in large, measurable decibel steps.
This produces the well-documented "structure-function curve": a shallow region at mild disease (function normal despite structural loss), a steep-slope middle region where the two measures track closely and are most mutually informative, and a flattened structural floor at advanced disease where OCT can no longer meaningfully quantify further loss and perimetry becomes the only sensitive monitoring tool remaining.
Spectral-domain OCT scans a circle (typically 3.4mm diameter) around the optic disc and unwraps the resulting RNFL thickness profile into a TSNIT curve — Temporal, Superior, Nasal, Inferior, back to Temporal — because the normal RNFL naturally follows a double-hump pattern, thickest at the superior and inferior poles (where the arcuate bundles converge) and thinnest temporally and nasally.
Glaucoma preferentially damages the inferotemporal and superotemporal RNFL first — the same bundles whose functional territory produces the arcuate scotomas mapped earlier in this simulation — so the TSNIT curve's normal double-hump shape progressively flattens as disease advances, with the inferior pole typically thinning slightly earlier and more severely than the superior pole in most population studies. By end-stage disease, the entire circumpapillary profile can fall below the 5th percentile of the normative database, and only the temporal RNFL sector (serving the nasal field, often relatively preserved until very late) remains within a measurable range.
A visual field diameter of 20° or less in the better-seeing eye meets the legal definition of blindness in the United States (and similar thresholds are used internationally), regardless of visual acuity, which can remain excellent (20/20) even with severe tunnel vision because the preserved central island still resolves fine detail. This dissociation — perfect reading vision with profound peripheral field loss — is precisely why glaucoma is called "the silent thief of sight": central acuity, the symptom patients and even routine eye chart screening most readily notice, is the last function lost, long after most of the optic nerve's axons are already gone.
Patients with tunnel fields report severe functional impact disproportionate to their measured acuity: loss of mobility confidence, elevated fall risk, and disqualification from driving, even while passing a basic acuity-only vision screening. This gap between acuity-based and field-based definitions of functional vision is a major reason perimetry, not just visual acuity testing, is mandated for glaucoma monitoring and for driver-licensing vision standards in most jurisdictions.