Abstract
Purpose:
Optical coherence tomography (OCT) has been used to monitor papilledema. This study aims to determine which OCT-derived measures of the optic nerve head (ONH) detect resolving papilledema in children faster than standard OCT measures.
Methods:
Children (≤18 years of age) with papilledema who completed optic nerve SD-OCT pretreatment and had evidence of treatment response on one or more follow-up OCTs within 4 months were included. Standard (mean circumpapillary retinal nerve fiber layer [cpRNFL] thickness), device-derived (per-quadrant cpRNFL) and custom (ONH height, maximum Bruch's membrane displacement [BMD], ONH volume [ONHV], and BMD volume) OCT measures were calculated. Per-eye generalized estimating equations (GEEs) modelled changes in device-derived and custom measures as a function of mean cpRNFL to identify those measures that resolved faster during early (0–2 months) follow-up. Mean cpRNFL coefficients of greater than 1 indicated faster resolving papilledema.
Results:
We included 52 eyes of 29 children (mean age, 12.8 years; 72.4% female). In analysis of early follow-up visits (38 eyes from 22 children), nasal cpRNFL and maximum BMD in each quadrant resolved faster than mean cpRNFL (GEE coefficients range, 1.14–3.37). Inferior cpRNFL, superior, nasal, and inferior ONH heights and ONHV resolved slower than mean cpRNFL (GEE coefficients range, 0.67–0.87).
Conclusions:
Nasal cpRNFL is a promising device-derived OCT measure for the early detection of resolving papilledema in children compared with mean cpRNFL. Maximum BMD, a custom measure, also shows promise, but its calculation has not yet been incorporated into commercial OCT devices.
Translational Relevance:
This study guides the optimal use of OCT in capturing resolving papilledema in children.
This retrospective, longitudinal study evaluated children 18 years old or younger who presented to the Children's Hospital of Philadelphia between January 2016 and December 2020. The study adhered to the tenets of the Declaration of Helsinki and was approved with waiver of informed consent by the Institutional Review Board at the Children's Hospital of Philadelphia.
Subjects were included if they (1) were aged 18 years or younger, (2) underwent treatment for diagnosed or suspected papilledema in the setting of idiopathic intracranial hypertension (IIH) (i.e., primary pseudotumor cerebri syndrome), secondary pseudotumor cerebri syndrome, or papilledema secondary to a tumor, and (3) completed traditional cpRNFL and enhanced depth imaging raster scan protocol of the ONH using Heidelberg Spectralis SD-OCT (Heidelberg Engineering Inc., Heidelberg, Germany) on at least two separate visits in the first 4 months of treatment, including a pretreatment visit and one showing resolving papilledema, defined as a 10% or greater decrease in mean cpRNFL relative to pretreatment in an eye. The exclusion criteria were (1) worsening papilledema (i.e., >10% increase in mean cpRNFL relative to the pretreatment visit) before the first resolving visit and (2) poor quality OCT images precluding appropriate segmentation of the internal limiting membrane (ILM) and/or BM.
The thicknesses of standard (mean cpRNFL) and device-derived (per-quadrant cpRNFL) OCT measures were extracted from the consecutive circular B-scans using automated tools included with the device software (Heidelberg Eye Explorer: HEYEX, Heidelberg Engineering Inc., Heidelberg, Germany).
For custom OCT measures, segmentation of the retinal layers in each of the 24 radial B-scans for each eye was performed automatically using device software. We used ILM and BM, with BM margins on either side of the optic nerve opening joined together through interpolation. Raw format optic nerve OCT images (*.vol files) including the segmented ILM and BM curves were exported. Using customized MATLAB software (v2018a, The Mathworks, Inc., Natick, MA), custom shape (per-quadrant ONH height [ONHH] and maximum BM displacement [BMD]) and volumetric (ONH volume [ONHV] and BMD volume [BMDV]) measures of the ONH were calculated. Details of this process were adapted from a previously published technique.
20,30 Briefly, the
y (axial) and
x coordinates of the segmentations were scaled to micrometers using scan-specific pixel-to-µm scaling factors. Radial B-scans were truncated to 2.51 mm on either side of the scan midpoint based on the width of the shortest scan in the dataset. On central B-scans from each peripapillary quadrant (i.e., the 0°, 90°, 180°, and 270° radial scans), the ONHH was defined as the maximal difference between the ILM and the BM (
Fig. 1). The maximum BMD was defined as the maximal difference between the BM and a secant line formed by connecting the outermost BM points (
Fig. 1). More positive displacement values indicate that the BM is more anterior to the secant line. Negative (or less positive) maximum BMD values indicated a posteriorly displaced BM layer.
A trapezoidal method of approximation was used to calculate custom volumetric measures of the ONH as previously published.
20 Each of the 24 radial B-scans yielded an optic nerve height curve, that is, ILM–BM at each pixel, as a function of the distance from the optic nerve center. Trapezoidal prisms were then interpolated between each of these adjacent optic nerve height curves. The ONHV was defined as the sum of these trapezoidal prisms. The BMDV was calculated similarly except using the secant lines instead of the ILM. If the ILM and/or BM segmentations from some radial B-scans were unavailable owing to poor image quality, then 12, 8, or 6 radial B-scans were used to ensure 360° coverage of the ONH.
The combined ganglion cell layer and inner plexiform layer (GCIPL) thickness was extracted from the macular thickness map generated using the automated segmentation and then manually corrected by the same experienced operator. GCIPL thickness values from the inner and outer superior, nasal, inferior, and nasal quadrants were averaged to compute the mean GCIPL.
Each eligible eye in this study was regarded as a unit of analysis. First, changes in OCT variables were modelled as a function of time (in weeks after the pretreatment visit) to determine which OCT measures reflect decreasing swelling during the first four months of treatment. All available visits in this window were considered. Generalized estimating equation (GEE) models, accounting for intereye correlation and longitudinal correction for each subject, were constructed to calculate the mean change per week in each OCT measure. In each model, an OCT measure was the dependent variable and time was the independent variable, with a statistically significant slope being indicative of change. A P value of less than 0.05 was considered statistically significant.
Next, early follow-up visits (0–2 months after the pretreatment visit) were considered. Per-eye GEEs modelled the percent change in each OCT measure between pretreatment and these early follow-up visits, as a function of the corresponding percent change in mean cpRNFL over the same interval. OCT measures with 95% CI for a slope of greater than 1 were identified to have a greater decrease than mean cpRNFL in early stages of treatment and hence, capture resolving papilledema earlier than standard mean cpRNFL. A Bonferroni multiple comparison adjustment was made for these comparisons (P < 0.00357, translating to a confidence interval of 99.642%, was accepted as statistically significant).
For the macular analysis, we compared mean GCIPL thickness across the pretreatment, early (0–2 month) follow-up, and late (2–4 month) follow-up visits using GEE models accounting for within-subject inter-eye correlations. A Bonferroni multiple comparison adjustment was made for these pairwise comparisons (P < 0.0167 was accepted as statistically significant). All statistical analyses were performed using SPSS statistical software (v28, IBM, Armonk, NY).
Comparing Mean cpRNFL and OCT-based Device-Derived and Custom Measures in Early Follow-Up (0–2 Months)
In this retrospective longitudinal study, we evaluated the potential of device-derived and custom developed OCT measures of ONH morphology for the early detection of resolving papilledema in children. Our study provides novel longitudinal data for these measures of ONH swelling during the first several months of resolving disease. Our results indicate that the commercially available measure of cpRNFL thickness in the nasal quadrant can detect resolving papilledema earlier than mean cpRNFL. Of the custom measures, the maximum BMD in each peripapillary quadrant also resolved faster than mean cpRNFL during early treatment, but computation of this metric warrants intensive image processing methods that are not integrated currently with commercial systems in the clinical setting. Direct shape (i.e., per-quadrant ONHH) and volumetric (i.e., ONHV) measurements of the ONH demonstrated slower changes.
Similar longitudinal data in the pediatric population are sparse. In a study of 14 pediatric patients with papilledema secondary to IIH or other known etiologies, Landau et al.
31 reported that improvements in the mean cpRNFL thickness were evident as early as 1 week after optic nerve sheath fenestration. In a case series of two pediatric IIH patients, Loo et al.
32 noted resolving papilledema, as measured by mean cpRNFL thickness, as early as 3 months without corresponding changes on fundoscopy and with complete resolution evident by 6 months in both patients. Thus, although the longitudinal trends shown in this study are not particularly surprising, they corroborate and extend the results of previous longitudinal OCT studies in both the pediatric and adult
9,15 papilledema populations and contribute further evidence that the studied OCT-based standard, device-derived, and custom measures can capture improvements in papilledema over time.
Because the mean cpRNFL thickness summarizes information about structural changes around the optic disc, clinicians may consider this measure to diagnose papilledema; however, studies comparing longitudinal cpRNFL data across peripapillary quadrants with the traditionally used mean cpRNFL during resolving disease are limited. A prospective cohort study by Rebolleda and Muñoz-Negret
33 reported mean and per-quadrant cpRNFL data at the baseline visit and 3-, 6-, and 12-month follow-up visits in a cohort of 22 adult patients with mild papilledema during treatment. At the 3-month visit, the mean cpRNFL thickness decreased by 31.9% compared with baseline. The only peripapillary quadrant to report a greater decrease was the nasal quadrant (mean decrease of 34.2%), although a statistical comparison was not reported. Our findings in the present study parallel these observations, showing that OCT-measured changes in cpRNFL thickness after treatment for papilledema are not uniform along spatial meridians. The reason for such a differential response in cpRNFL decrease after therapy is not known, but it might reflect underlying differences in the anatomical structure of the ONH, including the acute angle of ONH insertion nasally, or from the fact that the axoplasmic pressure gradients that produce papilledema are themselves differentially distributed.
BM shape changes and deflection have been proposed as markers for early ICP treatment. In this study, displacement of the BM in each peripapillary quadrant showed a faster decrease in ONH swelling compared with mean cpRNFL in the early follow-up period. The mean slopes of the %∆ in maximum BMD/%∆ mean cpRNFL ranged from 2.68 to 3.37, which were greater than that of the %∆ nasal cpRNFL/mean cpRNFL (
m = 1.14), which suggests even better potential for early detection of resolving papilledema with these custom measures. BMDV showed an even higher mean slope of approximately 11.8, but this measure did not reach statistical significance owing to variability in the data. The use of shape and volumetric measures of the BM as potential biomarkers to measure the early success of lowering ICP has been reported in adults with papilledema.
21,24,34,35 These studies have reported that the geometric shape of peripapillary BM/retinal pigment epithelium in adult patients with intracranial hypertension is angulated anteriorly towards the vitreous (similar to the more positive absolute maximal displacement of the BM in our study) and later showed that, in patients who underwent interventions to lower ICP, there was a measurable posterior displacement of the BM as early as 1 hour (after lumbar puncture and shunting procedures) with an eventual return toward normal within the first 3 months of medical therapy. The principal component analyses used by these authors, however, warrant significant analysis efforts and do not necessarily reflect an intuitive method to monitor resolving edema in the real-world clinical setting. Simpler approaches to measure BM configuration, including directly calculated BM/retinal pigment epithelium angle, have reflected decreased ONH swelling in the hyperacute setting.
36 In the current study, we directly compared the early dynamic changes in image-based structural parameters of the BM with mean cpRNFL. Our results build on prior literature by contributing evidence that absolute displacement of the BM can detect resolving papilledema in children faster than the mean cpRNFL, especially in the early weeks to months of therapy.
Longitudinal data on maximum BMD in all peripapillary quadrants and BMDV were characterized by a degree of variability leading to concerns regarding the reliability and predictive capability of these measures in real-world clinical decision-making. A source of this variability includes our use of automatic segmentation of the BM, which is prone to artifact in the peripapillary region, especially when severe ONH swelling shadows the BM margins. In addition, accuracy in segmentation algorithms at the BM opening margins, a region prone to error, may be a contributing factor. In this study, we relied on a traditional method of BM representation, where a spline interpolation connected the BM opening margins on either side of the neural canal opening. Other segmentation methods that exclude regions of uncertain BM segmentation could possibly improve the precision of this measure in longitudinally monitoring papilledema.
20 Irrespective of possible improvements in segmentation or the statistically significant findings noted in the current study, the lack of a commercially available algorithm that can quickly compute BM changes precludes the immediate, widespread use of these measures to guide clinical care. However, as OCT technology continues to advance, it is possible that the calculation of custom shape and volumetric measures of BM morphology will be integrated into commercially available algorithms and subsequently may supplement the use of device-derived nasal cpRNFL in gauging early response to papilledema treatment in children.
Limited prior studies have reported longitudinal data on the protrusion height and volume of the ONH during resolving disease and showed ONHV to significantly decrease within 3.0 to 3.5 months of therapy.
9,15 However, the dynamic changes in ONHH and ONHV have not previously been directly compared with mean cpRNFL in the early stages of papilledema treatment in children. Our results revealed that superior, nasal, and inferior ONHH as well as ONHV resolved more slowly compared with mean cpRNFL in the first 2 months of follow-up. These findings suggest that, although direct OCT measures of the ONH may be used to monitor papilledema changes over extended follow-up periods, they are slow to respond to variations in ICP and, thus, are not candidates for detecting early resolving disease when compared with device-derived measures like nasal cpRNFL. Optic disc swelling affects the prelaminar region before affecting the peripapillary area, where cpRNFL thickness is measured.
37 Because ONHH and ONHV capture quantitative information regarding structures at or near this predominant site of axoplasmic stasis, then perhaps it is not too surprising that after treatment, they resolve more slowly than cpRNFL thickness, which measures more of the marginal zone of the pathologic process.
We acknowledge that decreases in the cpRNFL thickness (and in other studied OCT measures) seen in the initial weeks after treatment in the current study could reflect, at least in part, axonal loss during the disease course. However, the longitudinal analysis of mean GCIPL did not reach statistical significance and the mean GCIPL loss over the study duration in these patients was small (approximately 2.02 µm, corresponding with 2.6% of the mean GCIPL thickness at presentation). Compared with normative values of GCIPL thickness in healthy children (aged 5–15 years)
38 using the same OCT device (Heidelberg Spectralis), the GCIPL thickness in our study was abnormally thin (<5th percentile) in 5% of eyes before treatment. In a recent study of children with newly diagnosed brain tumors,
39 the authors noted relatively high negative predictive values for the mean GCIPL thickness, thus signifying that normal GCIPL structure confers a relatively high certainty of intact visual function. This finding helps to support that the children in our cohort, with a largely normal GCIPL thickness, likely retained visual function and that decreases in cpRNFL and changes in BMD seen in this study reflect decreasing edema with contributions from underlying axonal damage and degeneration being limited.
Limitations inherent to a retrospective study should also be considered when interpreting these results, including the potential for selection bias owing to the frequency of imaging. Subject follow-up times were also heterogenous owing to the clinical nature of the data analyzed.
In conclusion, there is a need to detect early treatment response in children with papilledema. Our results indicate that nasal cpRNFL, a measure automatically calculated by most commercial OCT devices, may detect papilledema treatment response faster than mean cpRNFL. BMD, a custom OCT measure, may also detect resolving papilledema earlier than mean cpRNFL. However, computation of this metric involves intensive imaging processing methods, often with manual components, that are not included on current versions of commercial OCT device software. These results contribute to the evolving literature on the use of OCT to monitor papilledema in children during treatment. Further research that can validate the clinical usefulness of these findings is warranted.
Funding provided by NIH P30 026877, an unrestricted grant from Research to Prevent Blindness, and The Richard Shafritz Endowed Chair in Pediatric Ophthalmology Research.
Disclosure: T.V. Majmudar, None; H.E. Moss, None; R.A. Avery, None