Retinal nerve fiber layer (RNFL) thickness depicted by heatmap (A) and along a peripapillary circumference with a 1.72-mm radius (B). An axial cross-section through the center of the optic disc (C) enables analysis of RNFL thickness (blue arrowhead), retinal thickness (red arrowhead) (D), and anterior retinal projection (E, yellow arrowhead). The asterisk in D indicates vascular elements causing posterior shadowing. The dotted white line is a vector connecting the posterior-most ILM adjacent to either side of the optic disc. GCL indicates ganglion cell layer; ILM, inner limiting membrane; IN, inferior nasal; INL, inner nuclear layer; IT, inferior temporal; NL, nasal lower; NU, nasal upper; ONL, outer nuclear layer; RPE, retinal pigment epithelium; SN, superior nasal; ST, superior temporal; TL, temporal lower; TU, temporal upper.
A, Near infrared reflectance (NIR-REF) and 3-dimensional spectral-domain (SD)–OCT. The white arrowheads indicate blurred optic disc margins. B, Fundal images of patients with normal (patient 1; ICP, 6 mm Hg) and elevated (patient 2; ICP, 18 mm Hg and patient 3; ICP, 20 mm Hg) ICP. C, OCT retinal parameters (maximal retinal thickness, anterior projection, and maximal retinal nerve fiber layer [RNFL] thickness) plotted as a function of ICP measured intraoperatively. D, Receiver operating characteristic (ROC) curves for each of the 3 OCT parameters, combined RNFL and maximal retinal thickness parameters, and a model combining all parameters.
eTable 1. Characteristics of Study Subjects.
eTable 2. Comparison of OCT Parameters Among Study Cohorts.
eTable 3. Distribution of OCT Parameters Among Negative Controls and Its Association With Age.
eTable 4. Sensitivity and Specificity for Detecting Elevated ICP Using OCT Parameters and Clinical Signs.
eTable 5. Area Under ROC Curve for the Detection of ICP>15 mmHg Among Craniosynostosis Patients Using OCT Parameters.
eTable 6. Intereye Agreement Among OCT Parameters.
eTable 7. Correlation Among OCT Parameters.
eTable 8. Intragrader and Intergrader Reliability for OCT Parameters (Based on Sample of 20 Eyes).
eFigure. Study Enrollment.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Swanson JW, Aleman TS, Xu W, et al. Evaluation of Optical Coherence Tomography to Detect Elevated Intracranial Pressure in Children. JAMA Ophthalmol. 2017;135(4):320–328. doi:10.1001/jamaophthalmol.2017.0025
What are the sensitivity and specificity of spectral-domain optical coherence tomography for diagnosing intracranial hypertension in children?
In this cross-sectional study that included 79 patients, spectral-domain optical coherence tomography parameters demonstrated a sensitivity of 89% and specificity of 62% for detecting intracranial hypertension in children with hydrocephalus or craniosynostosis.
Spectral-domain optical coherence tomography shows promise as a surrogate, noninvasive measure of intracranial pressure, and has the potential to advance current diagnostic and management paradigms for elevated intracranial pressure in children.
Detecting elevated intracranial pressure in children with subacute conditions, such as craniosynostosis or tumor, may enable timely intervention and prevent neurocognitive impairment, but conventional techniques are invasive and often equivocal. Elevated intracranial pressure leads to structural changes in the peripapillary retina. Spectral-domain (SD) optical coherence tomography (OCT) can noninvasively quantify retinal layers to a micron-level resolution.
To evaluate whether retinal measurements from OCT can serve as an effective surrogate for invasive intracranial pressure measurement.
Design, Setting, and Participants
This cross-sectional study included patients undergoing procedures at the Children’s Hospital of Philadelphia from September 2014 to June 2015. Three groups of patients (n = 79) were prospectively enrolled from the Craniofacial Surgery clinic including patients with craniosynostosis (n = 40). The positive control cohort consisted of patients with hydrocephalus and suspected intracranial hypertension (n = 5), and the negative control cohort consisted of otherwise healthy patients undergoing a minor procedure (n = 34).
Main Outcomes and Measures
Spectral-domain OCT was performed preoperatively in all cohorts. Children with cranial pathology, but not negative control patients, underwent direct intraoperative intracranial pressure measurement. The primary outcome was the association between peripapillary retinal OCT parameters and directly measured elevated intracranial pressure.
The mean (SD) age was 34.6 (45.2) months in the craniosynostosis cohort (33% female), 48.9 (83.8) months in the hydrocephalus and suspected intracranial hypertension cohort (60% female), and 59.7 (64.4) months in the healthy cohort (47% female). Intracranial pressure correlated with maximal retinal nerve fiber layer thickness (r = 0.60, P ≤ .001), maximal retinal thickness (r = 0.53, P ≤ .001), and maximal anterior retinal projection (r = 0.53, P = .003). Using cut points derived from the negative control patients, OCT parameters yielded 89% sensitivity (95% CI, 69%-97%) and 62% specificity (95% CI, 41%-79%) for detecting elevated intracranial pressure. The SD-OCT measures had high intereye agreement (intraclass correlation, 0.83-0.93) and high intragrader and intergrader agreement (intraclass correlation ≥0.94). Conventional clinical signs had low sensitivity (11%-42%) for detecting intracranial hypertension.
Conclusions and Relevance
Noninvasive quantitative measures of the peripapillary retinal structure by SD-OCT were correlated with invasively measured intracranial pressure. Optical coherence tomographic parameters showed promise as surrogate, noninvasive measures of intracranial pressure, outperforming other conventional clinical measures. Spectral-domain OCT of the peripapillary region has the potential to advance current treatment paradigms for elevated intracranial pressure in children.
Detecting elevated intracranial pressure (ICP) is of major importance for patients with either an expansive intracranial process, such as tumor or hydrocephalus, or a constricting skull pathophysiology, such as craniosynostosis. An ideal measure of ICP would be sensitive, specific, noninvasive, and easy to implement. Current conventional methods of measuring ICP often yield equivocal results, reflect limited time interval of assessment, and may be invasive. Fundus examination of the optic nerve head can detect papilledema or optic atrophy, either of which may be associated with elevated ICP. However, this subjective evaluation has low interclinician agreement (only 14%-40% sensitivity to detect increased ICP) and often detects only late manifestations of severe ICP elevation.1-3 On the other hand, direct ICP monitoring via a transcranial catheter is accurate but invasive, with significant risks and expense and typically requires sedation for 24 to 48 hours.4 Ultrasonography of the optic nerve sheath diameter shows promise in diagnosing cases of severely elevated ICP, especially in acute situations,5-8 but has a sensitivity of only 11%.9,10 An indirect, noninvasive measure of ICP that can be performed serially over extended periods would be ideal.
Because the optic nerve is primarily intracranial, variations in ICP may affect the anatomy of the intraocular optic nerve. Optical coherence tomography (OCT) can provide reliable and highly reproducible quantitative measures of the changes of the retinal structure surrounding the optic nerve (peripapillary region) associated with optic nerve edema.11
In this study, patients with craniosynostosis and hydrocephalus, disorders associated with increased ICP, were studied.12-14 We tested whether quantitative structural parameters of the peripapillary retina correlate with ICP and can be used to noninvasively monitor ICP.
Three cohorts of patients at The Children’s Hospital of Philadelphia were prospectively enrolled from the Craniofacial Surgery Clinic (Table 1). Retinal OCT was the main study procedure, and association between 8 retinal OCT parameters and directly measured ICP was the primary outcome. Written informed consent was obtained from parents/guardians for all patients; procedures were approved by The Children’s Hospital of Philadelphia institutional review board (13-010131).
After induction of general anesthesia and prior to the surgical procedure, imaging using portable spectral-domain (SD) OCT was performed (iVue software version 3.2; Optovue) in each eye using 2 imaging protocols: optic nerve head (ONH) and 3-dimensional disc. From the ONH scan, the iVue software determined the retinal nerve fiber layer (RNFL) thickness along a peripapillary circumference with a radius of 1.72 mm centered on the optic disc (Figure 1A and B).15 The mean peripapillary RNFL thickness and optic disc area and volume were recorded. On the 3-dimensional scan, segmentation boundaries were automatically fitted to the inner limiting membrane (ILM), RNFL, and retinal pigment epithelium (Figure 1C).16 A masked analyst (W.X.) selected the single SD-OCT cross-section that corresponded most closely to the center of the optic disc for thickness measurements (Figure 1C).
The ILM is automatically detected by SD-OCT segmentation algorithms and served as the vitread reference for defining thickness parameters (Figure 1D). Greater axon density of the RNFL near the ONH margin confers an elevated appearance to the ONH nerve edge; vascular elements cause posterior shadowing (Figure 1D, asterisk). The RNFL thickness was defined as the distance between the ILM and the posterior boundary of the RNFL on the horizontal cross-section centered on the ONH. The maximal retinal thickness, or total optic nerve elevation, was the length of a vertical segment of a line that extended from the maximal ILM elevation to a perpendicular intersection with the segmentation boundary of the retinal pigment epithelium/Bruch membrane as it connects the nasal and temporal sides of the neural canal (Figure 1D, red arrowhead). Similarly, the maximal anterior projection was measured along a line that extended from the maximal ILM elevation and the location where it crosses, perpendicularly, a line/plane parallel to the peripapillary ILM (Figure 1E, yellow arrowhead).
To determine intragrader and intergrader reliability, the initial grader (W.X.) and a separate grader (J.W.S.) independently measured a random sample of 20 scans after 1 month had elapsed.
Patients in the first 2 cohorts underwent direct ICP measurement during the initial craniotomy by neurosurgeons who were masked from the OCT findings. Following surgical skin incision and scalp dissection, the patient’s end-tidal carbon dioxide was standardized to 36 mm Hg and analgesia was optimized by the anesthesiologist17 before an initial burr hole was placed lateral to the midline. A flexible Camino ICP monitor catheter (model 110-4B; Integra Lifesciences) was introduced through a 2-mm durotomy and advanced 1 to 2 cm into the superficial subdural space, parallel to the plane of the dura. Waveform was evaluated on the monitor, ensuring characteristic pulsation pattern and respiratory variation. The waveform was followed until the ICP value had stabilized at a constant value for 1 minute, and that ICP was recorded. An ICP greater than 15 mm Hg was considered elevated, consistent with previous studies.12-14,18-23
Spectral-domain OCT parameters for each eye were used for analysis, and intereye correlation was accounted for by using generalized estimating equations with an exchangeable working correlation structure.24 Mean OCT parameters were compared among 3 cohort groups using a generalized linear regression model that accounts for the intereye correlation and the age difference across groups. Among the patients with craniosynostosis and positive control patients, we assessed the association of OCT parameters with ICP by using the generalized linear regression models (without and with adjustment by age), where ICP was modeled as an outcome and each OCT measure was modeled as a predictor. Models were executed by using PROC GENMOD (SAS version 9.4; SAS Institute). Among negative control participants, we assessed age relation to OCT parameters using Pearson correlation analysis and generalized linear regression models by using each OCT parameter as an outcome and age as a predictor. We determined the normal range of OCT parameters using the 95% CIs, the upper limits of which were applied to those with craniosynostosis to calculate the sensitivity and specificity for detecting elevated ICP (>15 mm Hg) using each OCT parameter. We performed receiver operating characteristic (ROC) curve analyses and calculated the sensitivity and specificity corresponding to the optimal cut point determined based on the largest value of the sum of sensitivity and specificity of the ROC curve. Intragrader, intergrader, and intereye agreement of OCT parameters were assessed using intraclass correlation and 95% CIs. All statistical analyses were performed with the use of SAS version 9.4 (SAS Institute Inc), and P < .05 (without correction for multiple comparisons) was considered significant.
Seventy-nine patients were analyzed: 40 patients (78 eyes; mean [SD] age, 34.6 [45.2] months; 33% female) in the craniosynostosis cohort, 5 (9 eyes; mean [SD] age, 48.9 [83.8] months; 60% female) as positive control patients, and 34 (68 eyes; mean [SD] age, 59.7 [64.4] months; 47% female) as negative control individuals (eFigure and eTable 1 in the Supplement). The most common surgical indications among the negative control cohort were superficial facial lesion excision (n = 16) and palatoplasty (n = 11). Fifty patients (63%) were younger than 2 years of age. The median age was 15 months (interquartile range, 9-37 months) for patients with craniosynostosis, 16 months (interquartile range, 11-17 months) for positive control patients, and 21 months (interquartile range, 10-117 months) for negative control individuals. Among patients with craniosynostosis, 15 (38%) had syndromic or multisuture synostosis, and 31 (78%) were undergoing initial intracranial surgical procedure. Nineteen patients with craniosynostosis (48%), and all 5 patients with hydrocephalus, had elevated ICP, defined as greater than 15 mm Hg.
Three patients with craniosynostosis exemplify the spectrum of abnormalities encountered: patients 1 and 2 had normal preoperative fundus examination findings, whereas patient 3 had obvious optic nerve edema (Figure 2A and B). Spectral-domain OCT demonstrated normal appearance in patient 1, but raised optic nerve contour in patients 2 and 3. Intracranial pressure was within normal limits in patient 1 (6 mm Hg), but elevated in patient 2 (18 mm Hg) and patient 3 (20 mm Hg).
Quantitatively, all 3 optic nerve SD-OCT parameters used in this study demonstrated moderate or high correlation with invasively measured ICP: maximal RNFL thickness (r = 0.60, P ≤ .001), maximal retinal thickness (r = 0.53, P ≤ .001), and maximal anterior retinal projection (r = 0.53, P = .003) (Table 2 and Figure 2C).25 The SD-OCT disc parameters differed significantly between the 3 patient cohorts (eTable 2 in the Supplement). Among patients with craniosynostosis, abnormal SD-OCT parameters were associated with elevated ICP at least 90% of the time (Figure 2C). When patients with craniosynostosis were stratified, those with elevated ICP greater than 15 mm Hg showed thickening of all 3 mean SD-OCT parameters compared with those with ICP of 15 mm Hg or less, even after adjustment for age (mean [SE], maximal RNFL thickness: 208.5 [6.8] µm vs 158.9 [6.6] µm, P < .001; maximal retinal thickness: 588.7 [23.0] µm vs 435.8 [22.3] µm, P < .001; and maximal anterior retinal projection: 240.0 [23.2] µm vs 108.4 [22.5] µm, P = .001; Table 3). Maximal RNFL and retinal thickness did not differ between patients with craniosynostosis without elevated ICP and negative control individuals (Figure 2C, bottom left quadrant). The OCT measurements in positive control patients were similar to patients with craniosynostosis with ICP greater than 15 mm Hg (mean [SE], maximal RNFL thickness: 228.1 [60.4] µm vs 208.6 [34.4] µm, P = .12; maximal retinal thickness: 690.1 [261.7] µm vs 584.7 [125.2] µm, P = .17; and maximal retinal anterior projection: 336.0 [218.0] µm vs 237.3 [114.4] µm, P = .12), and were significantly higher than negative control individuals (mean [SE], maximal RNFL thickness: 228.1 [60.4] µm vs 154.3 [27.4] µm, P < .001; maximal retinal thickness: 690.1 [261.7] µm vs 414.3 [82.9] µm, P < .001; and maximal retinal anterior projection: 336.0 [218.0] µm vs 46.3 [57.5] µm, P < .001). Further, refractive error was not associated with any of the 3 OCT parameters, both with and without age adjustment (r ≤ 0.11; P ≥ .62). Among ONH parameters, only optic cup volume differed significantly between patients with craniosynostosis with and without elevated ICP.
Among normal control individuals, the cut points for abnormal OCT parameters based on the upper limit of the 95% CI of normal approximation (irrespective of their age) were established as maximal RNFL thickness of 208 μm; maximal retinal thickness, 577 μm; and maximal anterior retinal projection, 159 μm (eTable 3 in the Supplement). Using the above-established SD-OCT cut-off points, elevated anterior retinal projection was 84% (95% CI, 62%-94%) sensitive and 67% (95% CI, 45%-83%) specific for elevated ICP among patients with craniosynostosis (eTable 4 in the Supplement; Figure 2D). The 2 other OCT parameters (maximal RNFL thickness and maximal retinal thickness) each showed higher specificity (81% and 86%, respectively) but lower sensitivity (79% and 63%, respectively; Figure 2D). A combination of maximal RNFL thickness and maximal anterior retinal projection (in which the test result is considered positive if either parameter is above its cut point) yielded 89% sensitivity (95% CI, 69%-97%) and 62% specificity (95% CI, 41%-79%). Participant age was weakly correlated with OCT parameters (Pearson correlation coefficient ≤0.30). Sensitivity analysis using age-specific cut points yielded similar results for sensitivity and specificity compared with a single, age-independent cut-off point for each parameter (data not shown).
Among clinical predictors, papilledema on indirect ophthalmoscopy had 11% sensitivity and 100% specificity for elevated ICP (eTable 4 in the Supplement). Radiographic findings and suspicious headaches similarly had low sensitivity (<42%) but high specificity (>78%).
Each SD-OCT parameter had a high area under the ROC curve for detecting elevated ICP (ranges from 0.81-0.86), and their optimal cut points determined from ROC curves provide high sensitivity (0.84-0.89) and specificity (0.71-0.76) (Figure 2D; eTable 5 in the Supplement).
There was high intereye agreement of SD-OCT parameters across all participants (intraclass correlation coefficients range, 0.83-0.92; eTable 6 in the Supplement). Similarly, there was high correlation among all OCT parameters (range, 0.87-0.93; eTable 7 in the Supplement). Intragrader (0.94-1.00) and intergrader (0.98-1.00) agreement of OCT parameters were also very high (eTable 8 in the Supplement).
This study establishes an association between quantitative parameters of the microscopic structure of the peripapillary retina, determined noninvasively with SD-OCT, and ICP, measured directly from the intracranial space. The measurements were made within minutes of each other. Retinal changes generally correlated in a linear fashion with increased ICP. Spectral-domain OCT demonstrated potential as a diagnostic tool for the noninvasive detection of elevated ICP, with 89% sensitivity and 62% specificity, and it may be used to advance the current diagnostic and treatment algorithms in diseases with elevated ICP. Finally, OCT parameters can be reliably measured as indicated by high intergrader and intragrader agreement.
The ability to reliably detect elevated ICP in at-risk patient cohorts using noninvasive techniques promises to be transformational. Parents and physicians of patients with factors suspicious for elevated ICP, but no objective findings using currently available methods, are understandably reluctant to proceed with invasive testing or surgical intervention.
We observed that papilledema in younger patients was not reliably detected by SD-OCT ONH parameters that have been studied in older children with craniosynostosis.26,27 We investigated a broader set of parameters; for example, anterior retinal projection was found to correlate with ICP in a small series of patients studied with handheld ultrasonography.5 Furthermore, because previous normative studies of OCT have not assessed children younger than 3 years of age,28-30 we evaluated age-matched control participants to clarify patterns and variability in a normal population. Our finding that only 1 retinal parameter enlarges with normal aging (and to only a small degree) partially affirms28,29 and refutes26 findings of previous studies limited to older patients. We found sufficient magnitude of retinal variation in patients with elevated ICP to propose retinal thickness thresholds independent of age and still achieve high sensitivity.
There is a wealth of information on the influence of refractive errors on OCT parameters, particularly in highly myopic eyes with long axial lengths. High hyperopia and shorter axial length are much less represented in the literature and are often associated with small nerves with elevated contours, which may confound the relationship between ICP measures and optic nerve OCT parameters. In this study, only 1 patient had a high refractive error (+6.00 D), the remainder of the patients averaged +1.00 D with no significant correlation between OCT thickness parameters and magnitude of refractive errors. Although possible RNFL overestimates from hyperopic errors are considered relatively small in relation to changes in optic nerve parameters found in this study and in other OCT studies in mild papilledema,31 care must be taken when using OCT as a surrogate measure of ICP in patients with high (>6 D) refractive errors.32-36 An attractive attribute of OCT as a surrogate measure of ICP is its repeatability over time. True papilledema secondary to increased ICP is expected to fluctuate with ICP and trend with repeated examinations compared with clinical cases of stably elevated optic nerve contours, such as constitutionally small optic nerves or optic nerve drusen.37 The longitudinal behavior and variability of the OCT parameters and specific metrics, such as the deformation of the lamina cribosa, may be less prone to anatomic variations, hold promise as potential discriminators between true papilledema and pseudopapilledema, and warrant future studies.38
We defined an abnormally elevated ICP as greater than 15 mm Hg, consistent with previous studies of craniosynostosis and hydrocephalus.4,12-14,18-23 However, our data affirmed that retinal thickness and ICP are continuous parameters: ICPs between 10 and 15 m Hg are likely not normal, particularly in young infants. Similarly, OCT demonstrates retinal thickening and projection in these patients that clearly deviates from age-matched normal patients, despite being below an ICP threshold of 15 mm Hg. We interpret this to suggest that patients with borderline ICP elevation likely embody pathophysiology, affirming the high incidence of intracranial hypertension associated with these conditions.
Secondary outcomes affirmed low sensitivity of direct fundus examination (11%), computed tomographic findings of thumb printing (42%) or ventriculomegaly (11%), or headache or neurocognitive changes (33%) to detect elevated ICP. The high specificity of these signs (78%-100%) may suggest that specialists at our institution have higher thresholds for diagnosing abnormality, but ultimately confirms the subjectivity of these measures.13 We noted a high (48%) rate of patients with craniosynostosis with elevated ICP, which may reflect the large syndromic patient proportion in our cohort as well as several patients initially operated on at later age. Nonetheless, it is consistent with a growing body of evidence documenting the high rates of elevated ICP in craniosynostosis.13,39,40
This study had several limitations that we attempted to mitigate. Our perioperative ICP measurement protocol was consistent with other subdural catheterization methods,6,41 but lacked the 24- to 48-hour duration or surveillance for multiple B-type waves reported in other protocols,12,13,19 which were not feasible in a perioperative setting. We sought to mitigate potentially confounding perioperative factors by controlling end-tidal carbon dioxide, ensuring adequate analgesia, standardizing timing of measurement relative to the case, and confirming authenticity of the waveform and ICP value over time. Sedation and analgesia are regularly used in overnight studies of direct ICP measurement; if anything, we would expect perioperative sedation and analgesia to depress ICP and reduce our detection of patients with elevated ICP. The precise relationship between the clinical ONH appearance and the OCT and ICP parameters was not approached quantitatively with fundus photography as such a procedure would have inadmissibly prolonged anesthetic time.42
Papilledema is thought to develop at a relatively slow rate after days or weeks of elevated ICP.9,41 The magnitude and time course of the ICP elevations will likely interact with anatomic factors, such as the distensibility of the intraocular optic nerve support and axonal elements, to ultimately result in clinically and/or SD-OCT–evident papilledema.38,43-45 In this study, care was taken to include patients with symptom stability for a minimum of 30 days to allow for sufficient time for papilledema to develop and stabilize. As a result, however, we may have excluded patients with earlier changes of the ONH anatomy in response to ICP elevation. More precise dissection of the ONH structure with SD-OCT and more frequent and earlier observations in larger groups of patients, complementing traditional ICP measures with intraoperative and ambulatory OCTs (without the need of general anesthesia), are needed to gain further insight into the complex relationships between ICP levels and the anatomy of the optic nerve.
The study design acknowledged that OCT measurements are not a standard component of this invasive procedure and needed to minimize obtrusiveness to timing and flow of surgery. This constrained sample composition and extent of measurements to focus on the central question: do findings support the potential for OCT as a surrogate, noninvasive measure of increased ICP? The relationships found are in agreement with previous work where cerebrospinal fluid opening pressures have been related to peripapillary OCT measures.46-48 We recognize that they are modest and potentially influenced by other variables that warrant further study in larger more heterogeneous samples and with evolvingly sophisticated resolution.36 However, we note a degree of sensitivity far superior to conventional noninvasive methods and believe that this is an important step toward maturation of a clinically essential tool and technology.
In this study, we found that noninvasive quantitative measures of the peripapillary retinal structure by SD-OCT were correlated with invasively measured ICP. Optical coherence tomographic parameters showed promise as surrogate, noninvasive measures of ICP outperforming other conventional clinical measures. Spectral-domain OCT of the peripapillary region has the potential to advance current treatment paradigms for elevated ICP in children.
Accepted for Publication: January 8, 2017.
Corresponding Author: Jesse A. Taylor, MD, Division of Plastic Surgery, The Children’s Hospital of Philadelphia, Colket Translational Research Bldg, 3501 Civic Center Blvd, Philadelphia, PA 19104 (firstname.lastname@example.org).
Published Online: February 23, 2017. doi:10.1001/jamaophthalmol.2017.0025
Author Contributions: Dr Taylor had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Swanson, Aleman, Liu, Storm, Bartlett, Katowitz, Taylor.
Acquisition, analysis, or interpretation of data: Swanson, Aleman, Xu, Ying, Pan, Lang, Heuer, Bartlett, Katowitz.
Drafting of the manuscript: Swanson, Aleman, Xu, Pan, Katowitz, Taylor.
Critical revision of the manuscript for important intellectual content: Aleman, Xu, Ying, Liu, Lang, Heuer, Storm, Bartlett, Katowitz.
Statistical analysis: Swanson, Xu, Ying, Pan.
Obtained funding: Katowitz.
Administrative, technical, or material support: Swanson, Xu, Liu, Lang, Storm, Bartlett, Katowitz.
Study supervision: Ying, Heuer, Storm, Bartlett, Katowitz, Taylor.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This study was supported by grant UL1TR000003 from the National Institutes of Health National Center for Advancing Translational Sciences for equipment acquisition, grants K12EY015398-10 and 2-P30-EY01583-26 from the National Institutes of Health National Eye Institute, and funding from Research to Prevent Blindness, Foundation Fighting Blindness, Hope for Vision, Macula Vision Research Foundation, Pennsylvania Lions Sight Conservation and Research Foundation, Mackall Trust Funds, the Foundation Fighting Blindness, and the Center for Human Appearance at the University of Pennsylvania for research assistance and biostatistical support. Funding from each of these sources was used for equipment acquisition and/or researcher salary support.
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: The study team is grateful to Graham Quinn, MD, and Lloyd Bender, MD, for their contributions to the design of this study, and to Brianne Mitchell, MD, for her assistance with data collection. They are affiliated with the Children’s Hospital of Philadelphia and none received compensation.