Device agreement breaks down in primary congenital glaucoma eyes
Corneal measurement in primary congenital glaucoma is not a routine optics exercise. These eyes often carry a history of treatment, altered corneal shape, and biomechanical changes that can distort the numbers clinicians rely on when judging visual quality and planning follow-up. In that setting, even modest disagreement between imaging platforms can complicate interpretation of higher-order aberrations, especially when the goal is to separate disease-related irregularity from instrument noise.
This cross-sectional study compared three devices — Pentacam HR, ACE, and iTrace — in 37 eyes from 23 treated PCG patients. Mean age was 9.4 ± 2.8 years, placing the cohort squarely in childhood, where cooperation, fixation stability, and pupil size can all influence optical measurements. The investigators examined anterior corneal HOAs at 3-, 4-, and 5-mm optical zones and used pairwise intraclass correlation coefficients together with Bland-Altman analysis, while also applying patient-level cluster-bootstrap confidence intervals and distribution-free limits of agreement.
The central finding was straightforward: the devices did not behave interchangeably in this disease population.
Coma was the most consistent aberration, but agreement still varied widely
Among the measured aberrations, coma performed best, yet even coma ranged from poor to excellent depending on the device pair and optical zone. The reported ICC values spanned 0.34 to 0.91, and agreement improved at larger zones for the Pentacam HR–ACE comparison. In practical optical terms, this pattern is not surprising. Larger analysis zones can average out some localized irregularity and may reduce the influence of central measurement noise, but they also introduce their own dependence on pupil size and peripheral corneal sampling.
Trefoil was a different story. Agreement was poor for most device pairs, with ICC values at or below 0.47. Trefoil often reflects subtle asymmetry in corneal shape, and in eyes already altered by congenital glaucoma and prior treatment, that subtlety may be exactly what the instruments fail to capture consistently. When a parameter is this unstable across platforms, it becomes a weak candidate for longitudinal comparison unless the same device is used every time.
Keratometric agreement was stronger than HOA agreement
Keratometry fared better than the HOA metrics. Between Pentacam HR and ACE, agreement was excellent, with ICC values of at least 0.93. After excluding one implausible iTrace acquisition, the iTrace comparisons also improved to good-to-excellent agreement, with ICC values from 0.88 to 0.92 across 21 to 22 eyes. Those numbers suggest that the basic curvature estimate is far more robust across platforms than the more intricate wavefront-derived aberration profile.
A single implausible acquisition on iTrace also serves as a reminder that instrumentation quality and patient cooperation matter more in pediatric glaucoma than in many adult corneal studies. Small fixation errors, tear film instability, or an off-axis scan can create an outlier large enough to distort a small cohort analysis. In an exploratory dataset of this size, one aberrant measurement can change the interpretation of an entire comparison.
Undilated pupils likely contributed to the mismatch
Restricting analysis to pupil-adequate eyes improved Pentacam HR–ACE horizontal coma agreement at 3 mm, with ICC improving from 0.34 to 0.83, although the confidence intervals overlapped. Even without a definitive statistical separation, the direction of change points to a biologically plausible mechanism: undilated pupil size partially shaped the disagreement. If the effective aperture is too small or inconsistent, the instrument may sample a different portion of the corneal optical surface on each device, and the result is less a true physiologic mismatch than a sampling problem.
That mechanism matters in pediatric glaucoma because pupil diameter can fluctuate with ambient light, attention, and medication exposure. The same child can present with different optical conditions from one visit to the next, and a device that depends on a stable aperture may perform well one moment and poorly the next. Measurement disagreement in this context is not merely technical; it can alter whether a clinician concludes the cornea is stable, irregular, or worsening.
Clinical interpretation should stay device-specific
For clinicians, the most actionable message is caution. The data do not support treating these three devices as interchangeable for corneal HOA assessment in treated PCG eyes. If follow-up decisions depend on subtle shifts in coma or trefoil, switching platforms could create the illusion of progression or improvement where none exists. Longitudinal monitoring should therefore stay anchored to a single device whenever possible, and any cross-device comparison should be interpreted with restraint.
Coma appeared to be the most dependable HOA metric in this cohort, yet the study does not establish a universal threshold for clinical acceptability. A statistically meaningful agreement coefficient is not the same as a bedside usable one, especially when visual quality, amblyopia risk, and surgical history all complicate interpretation. In pediatric glaucoma, a measurement can be reproducible and still be clinically fragile if it fails to track the optical reality that matters to the child’s vision.
A small exploratory cohort leaves the bigger question open
The analysis is necessarily limited by size and design. Only 37 eyes were studied, and the inferential conclusions were restricted to Pentacam HR versus ACE at 4–5 mm, while ACE at 3 mm and all iTrace comparisons remained descriptive. Treating that boundary seriously is important. Small pediatric ophthalmic cohorts can reveal useful signals, but they cannot settle whether one device is superior across broader glaucoma phenotypes, whether surgical status modifies agreement, or whether the same patterns persist in untreated disease.
Even so, the study fills an important gap. Corneal imaging in primary congenital glaucoma often focuses on anatomy and pressure control, while optical quality receives less attention. Here, the optical quality data showed that HOAs are not a stable common currency across devices, and trefoil in particular is too inconsistent to use as a shared metric without caution. Larger-cohort validation will need to determine whether the observed disagreement reflects a true limitation of the instruments in abnormal pediatric corneas or a narrower issue tied to pupil size, scan quality, and the optical geometry of treated PCG eyes.