Utility of Ultrasound and Optical Coherence Tomography in Differentiating Between Papilledema and Pseudopapilledema in Children

The evaluation of children presenting with mild to moderate optic nerve head elevation can be challenging. Once confirmed, benign etiologies of optic nerve head elevation such as optic nerve head drusen (ONHD) and peripapillary hyper-reflective ovoid mass-like structures (PHOMS)^{1, 2}, frequently referred to as pseudopapilledema, do not require the patient to undergo further invasive or costly tests. On the other hand, children with optic nerve head elevation believed to be secondary to elevated intracranial pressure (ICP, termed papilledema) must be evaluated with brain MRI and, in most cases, a lumbar puncture. In the absence of distinctive features to help differentiate between these two etiologies (i.e., hemorrhage, cotton wool spots, visible druse), additional ophthalmic imaging is frequently performed.

B-Scan ultrasonography (BSUS) has been utilized for decades to help determine whether optic nerve head elevation is secondary to druse versus other etiologies. ^3–8 The absolute optic nerve sheath diameter (ONSD, mm), presence/absence of hyperreflective nodules at the ON head, ON sheath accentuation, and the “crescent sign” have been reported as helpful in differentiating between ONH elevation secondary to pseudopapilledema and true papilledema.^3–8 Device accessibility and ease of acquisition regardless of patient cooperation makes BSUS an appealing imaging modality for children—although certain features of ONHD may impede diagnostic accuracy.^{6, 9–11}

Spectral domain optical coherence tomography (OCT) has been recently utilized to differentiate between ONHD and papilledema. The circumpapillary retinal nerve fiber layer (cpRNFL) thickness and other OCT measures have demonstrated mixed results in their ability to differentiate between papilledema and ONHD.^12–20 Beyond the automated OCT segmentation software, the diameter of Bruch’s membrane opening (BMO), maximum papillary height and Bruch’s membrane deformation have also demonstrated mixed results in differentiating between causes of optic nerve head elevation.^{12, 14–21}

The purpose of this study was to investigate which individual and or combined features from BSUS and OCT are best able to differentiate between papilledema and pseudopapilledema in children.

METHODS

This study was approved by the Children’s Hospital of Philadelphia (CHOP) Institutional Board of Review. All data was collected and stored according to HIPAA guidelines. Children evaluated in the outpatient Neuro-Ophthalmology clinic between January 2016 and May 2018 at CHOP were retrospectively identified if they presented with optic nerve head elevation of unknown etiology. When the attending pediatric Neuro-Ophthalmologist (GTL or RAA) could not determine whether the optic nerve head elevation was secondary to pseudopapilledema versus elevated ICP (i.e., papilledema) and there were no other diagnostic features on dilated exam (i.e., vessel obscuration, hemorrhage, cotton wool spots, visible druse), the child underwent both BSUS and OCT performed by the same investigator (CT-H). A diagnosis of papilledema was confirmed by a combination of MRI and lumbar puncture results as well as resolution of optic nerve head elevation after treatment. Pseudopapilledema was confirmed by the absence of change in optic nerve head elevation on follow up examinations.

B-Scan Ultrasonography

A 12-MHz ultrasound (Accutome B-Scan Plus, Accutome Inc., Malvern, PA, USA; axial resolution of 0.015mm) was used for all children. Axial scans were also performed, along with vertical transverse (para-axial) scans, to obtain ONSD measurements. Subjects were classified as having ONHD regardless of superficial versus buried location. Fluid signs were categorized as the crescent sign/donut sign or as sheath accentuation/optic nerve doubling.²² The donut sign, or crescent sign, has been described as a hypoechoic, crescentic area surrounding the round hypoechoic focus which represents the optic nerve.³

ONSD was measured in each eye at 3mm behind the posterior retina perpendicular to the optic nerve head. The average of 3 unique measures were used to calculate the final ONSD.²³ An abnormally widened ONSD was defined as greater than 4.5 mm based on parameters for normal/abnormal ON sheath diameter set forth in previous studies.^{23, 24} BSUS acquisition and analysis was performed by the same investigator (CT-H).

OCT

Spectral domain OCT (Heidelberg Spectralis, Heidelberg Engineering GmbH, Heidelberg, Germany) was acquired in all subjects using the same protocol by the same investigator (CT-H). The average cpRNFL thickness, diameter of BMO, maximum papillary height, and the presence/absence of hyper/hyporeflective lesions at the optic nerve head were calculated from traditional and enhanced-depth imaging (EDI) OCT protocols, respectively. The EDI protocol utilized 24 radial line scans with each b-scan averaged nine times (automatic real-time = 9) centered over the optic nerve head. All scans were reviewed, and when necessary, segmentation errors were corrected.

Measurements of BMO diameter and maximum papillary height, performed by a masked investigator (MKF), were obtained at six clock-hour locations on an optic nerve radial scan. BMO diameter was defined as the transverse diameter of the neural canal at the level of the RPE/BM complex (Figure 1). The average of the six BMO measurements yielded the mean BMO diameter for each eye. Maximum papillary height was defined as the perpendicular line from the highest point of the optic nerve head to the BMO diameter line (Figure 1).

Figure 1.

Example of measurements of BMO diameter and maximum papillary height.

Statistical Analysis

We compared the BSUS and OCT imaging features between eyes with papilledema vs. eyes with pseudopapilledema using generalized linear models that accounted for the inter-eye correlation through generalized estimating equations.²⁵ We calculated the sensitivity, specificity, area under receiver operating characteristic curves (ROC) and their 95% confidence intervals (95% CI) from logistic regression models to determine the ability of BSUS and OCT imaging features, individually and in combination, for discriminating eyes having papilledema versus pseudopapilledema. For calculating the 95% CIs of sensitivity, specificity and area under ROC curve, the inter-eye correlation was accounted for generalized estimating equation.²⁵ All statistical analyses were performed in SAS v9.4 (SAS Institute Inc., Cary, NC).

RESULTS

One-hundred eighty-one eyes from 94 children (mean age: 11.0 years; range: 3.2–17.9) were included; 36 eyes from 20 subjects with papilledema and 145 eyes from 74 subjects with pseudopapilledema. There was no significant difference in patient sex (papilledema 64% female; pseudopapilledema 65% female) or patient age between subjects with papilledema and pseudopapilledema (mean 9.9 versus 11.1 years, respectively; p = 0.18). All patients diagnosed as having papilledema underwent MRI imaging and elevated ICP was secondary to an intracranial brain tumor in 4 subjects, the remainder met diagnostic criteria of pseudotumor cerebri syndrome after undergoing a lumbar puncture.²⁶ All patients diagnosed with papilledema were followed until the papilledema resolved confirming the diagnosis. Seventeen patients diagnosed as pseudopapilledema underwent MRI, 6 of whom also had a lumbar puncture all of which had opening pressure less than 25 centimeters of water. Subjects diagnosed with pseudopapilledema had follow up visits (median = 2 months, range 1–8) that typically occurred two months after their diagnosis (median = 2 months, range 1–68).

BSUS

All four BSUS features were significantly different between groups (Table 1). Among BSUS features, optic nerve sheath widening (>4.5mm) demonstrated the best sensitivity (86%, 95% CI: 64–96%) and specificity (88%, 95% CI: 79–94%) for papilledema. Sensitivity and specificity did not improve when considering the presence of optic nerve sheath widening (>4.5mm) along with any other the other 3 BSUS features. Given the importance of disease prevalence, optic nerve sheath widening (>4.5mm) demonstrated modest positive predictive value (65%, 95% CI: 45–80%) and excellent negative predictive value (96%, 95% CI: 89–99%) for papilledema. Drusen was detected in 50% of papilledema patients and 93% of pseudopapilledema patients.

Table 1:

Univariate Analysis for the Association of B-scan Ultrasonography with Papilledema versus Pseudopapilledema.

	Papilledema N=36 eyes (%)	Pseudopapilledema N=145 eyes (%)
Drusen present
No	18 (50.0%)	10 (6.9%)
Yes	18 (50.0%)	135 (93.1%)*
Sheath accentuation/Optic nerve doubling
No	18 (50.0%)	125 (86.2%)
Yes	18 (50.0%)	20 (13.8%)*
Optic nerve sheath diameter > 4.5 mm
No	5 (13.9%)	128 (88.3%)
Yes	31 (86.1%)	17 (11.7%)*
Crescent/donut sign
No	24 (66.7%)	138 (95.2%)
Yes	12 (33.3%)	7 (4.8%)*

^*P < 0.001, between groups.

OCT

Both mean values and distinct cut-points for cpRNFL thickness and maximum papillary height, along with the presence of hyper/hyporeflective nodules were significantly different between subject groups (Table 2). The mean values and percentages above distinct cut-points for the mean Bruch’s membrane opening did not significantly differ between groups (Table 2). When assessed as a continuous variable, cpRNFL thickness had the best diagnostic accuracy for discriminating between papilledema and pseudopapilledema (AUC=0.87, 95% CI: 0.82–0.93) whereas mean Bruch’s membrane opening and maximum papillary height performed poorly based on the area under the receiver operating characteristic curve (AUC=0.55, 95% CI: 0.43–0.67 and AUC=0.66, 95% CI: 0.55–0.76, respectively).

Table 2:

Univariate Analysis for the Association of Optical Coherence Tomography Features with Papilledema versus Pseudopapilledema.

	Papilledema N=36 eyes	Pseudopapilledema N=145 eyes
Circumpapillary Retinal Nerve Fiber Layer Thickness (microns)
Mean (SD)	164 (29)	123 (25)
Median (min, max)	162 (121, 261)	122 (69, 192)*
<120	0 (0.0%)	65 (44.8%)
≥120	36 (100%)	80 (55.2%)*
<140	6 (16.7%)	110 (75.9%)
≥140	30 (83.3%)	35 (24.1%)*
<160	18 (50.0%)	133 (91.7%)
≥160	18 (50.0%)	12 (8.3%)*
Hyper/hypo-reflective Round Nodules a
No	10 (27.8%)	4 (2.8%)
Yes	22 (61.1%)	128 (88.3%)*
Mean Bruch’s Membrane Opening b
Mean (SD)	1384 (172)	1404 (152)
Median (min, max)	1342 (1090, 1795)	1396 (1083, 1699)
<1500	24 (66.7%)	98 (67.6%)
≥1500	8 (22.2%)	39 (26.9%)
<1600	28 (77.8%)	121 (83.5%)
≥1600	4 (11.4%)	16 (11.0%)
Maximum Papillary Height c
Mean (SD)	974 (125)	902 (143) #
Median (min, max)	979 (672, 1223)	885 (565, 1387)
<900	6 (16.7%)	71 (49.0%)
≥900	24 (66.7%)	66 (45.5%) *
<1000	17 (47.2%)	106 (73.1%)
≥1000	13 (36.1%)	31 (21.4%)*
<1100	25 (69.4%)	129 (89.0%)
≥1100	5 (13.9%)	8 (5.5%)#

^*P < 0.001, between groups

^#P < 0.05, between groups

^aData unavailable (n = 17) due to patient cooperation and or inadequate image quality

^bData unavailable (n = 12) due to patient cooperation and or inadequate image quality

^cData unavailable (n = 14) due to patient cooperation and or inadequate image quality

When considering different cut-off points, cpRNFL thickness ≥140 microns demonstrated the best sensitivity (83%, 95% CI: 66–93%) and specificity (76%, 95% CI: 66–84%) for OCT measures to identify papilledema. Given the importance of disease prevalence, cpRNFL thickness ≥140 microns demonstrated poor positive predictive value (46%, 95% CI: 31–62%) and excellent negative predictive value (95%, 95% CI: 88–98%) for papilledema.

Combined BSUS and OCT Features

The combination of the best BSUS metric and best OCT metric (i.e. optic nerve sheath widening >4.5mm AND cpRNFL thickness ≥140) reduced the overall sensitivity (72%, 95% CI: 52–86%), but increased the specificity (95%: 95% CI: 88–98%, Table 3) subsequently improving the positive predictive value (79%, 95% CI: 55–92%) with a slight reduction in the negative predictive value (93%, 95% CI: 86–97%) relative to each individual metric. Reducing the cpRNFL thickness threshold to ≥120 microns in combination with the best BSUS metric demonstrated a small increase in sensitivity, along small reductions in specificity, positive and negative predictive values (Table 3).

Table 3:

Sensitivity, Specificity, Positive and Negative Predictive Value for the Best B-Scan Ultrasonography Measure and Optical Coherence Tomography Measure Alone or in Combination for Discriminating Between Papilledema and Pseudo-Papilledema.

	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)
Optic nerve sheath diameter > 4.5 mm	86.1 (64.0 – 95.6)	88.3 (79.3 – 93.7)	64.6 (45.0 – 80.2)	96.2 (88.5 – 98.8)
Circumpapillary Retinal Nerve Fiber Layer Thickness ≥140 (microns)	83.3 (65.9 – 92.8)	75.9 (66.0 – 83.6)	46.2 (31.2 – 61.9)	94.8 (87.5 – 98.0)
Optic nerve sheath diameter > 4.5 mm AND Circumpapillary Retinal Nerve Fiber Layer Thickness ≥140 (microns)	72.2 (52.3 – 86.1)	95.2 (87.5 – 98.2)	78.8 (54.6 – 92.0)	93.2 (85.9 – 96.9)
Optic nerve sheath diameter > 4.5 mm AND Circumpapillary Retinal Nerve Fiber Layer Thickness ≥120 (microns)	86.1 (64.0 – 95.6)	91.0 (82.2 – 95.7)	70.5 (49.4 – 85.3)	96.4 (88.8 – 98.9)

PPV = positive predictive value

NPV = negative predictive value

Analysis is at the eye-level, with inter-eye correlation accounted for using generalized estimating equations.

CONCLUSIONS

By acquiring both BSUS and OCT metrics in a large cohort of patients with optic nerve head elevation, our study found that specific measures from unique modalities demonstrate varying degrees of diagnostic accuracy when differentiating between children with papilledema and pseudopapilledema. While all BSUS tests showed a difference between groups, optic nerve sheath diameter >4.5mm demonstrated the best sensitivity and specificity in both detecting papilledema and in ruling it out. A cpRNFL thickness ≥140 microns demonstrated the best combination of sensitivity and specificity compared to other cpRNFL thickness thresholds (i.e., ≥120 or ≥ 160 microns). The prevalence of hyper- or hyporeflective round nodules indicative of PHOMS and/or drusen was different between groups, but it did not have good diagnostic accuracy in differentiating between them. Using different cut points, the maximum papillary height also differed between groups, but had poor diagnostic accuracy (i.e., AUC = 0.66).

Our study highlights the relative contribution, or lack thereof, of different BSUS findings believed to be present or absent in children with elevated ICP. Specifically, optic sheath accentuation and the crescent/donut sign have been reported to be strongly indicative of elevated ICP. While these signs were present in one-third to one-half of our papilledema subjects, they were also present in 5–14% of those with pseudopapilledema. While these false-positive signs may seem small, the much higher prevalence of pseudopapilledema results in a near equal number of patients with this sign from each group, producing a positive predictive value near chance.

While the positive predictive value of an optic nerve sheath diameter >4.5mm is slightly better than chance (65%), its excellent negative predictive value is the most clinically helpful, since in our study population the incidence of pseudopapilledema is over 3.5 times that of papilledema. Specifically, when the optic nerve sheath diameter is below 4.5 mm and there are no concerning signs or symptoms, the clinician can be reassured that there is a 96% chance that the patient does not have elevated ICP and can subsequently avoid additional invasive or expensive tests (e.g., lumbar puncture or brain MRI). In three patients (5 total eyes) with papilledema that did not demonstrate an optic nerve sheath diameter >4.5mm, other clinical features (i.e., visual acuity loss, headache) prompted the clinician to order a brain MRI. Individually, BSUS characteristics have been reported to be helpful in differentiating pseudopapilledema from papilledema. In agreement with our findings, multiple studies have found that increased optic nerve sheath diameter has a high negative predictive value in evaluation of optic nerve head elevation.^{5, 7, 8} The “crescent” or “doughnut” sign was not found to be a reliable indicator in our study due to the high number of false positive results.

In an attempt to create simple guidelines for clinicians, we evaluated three different thresholds of cpRNFL thickness. We found that a cpRNFL thickness of ≥140 microns demonstrated the best overall performance providing an excellent negative predictive value. Despite this high negative predictive value, numerous factors could certainly influence the utility of such a cut-off including the duration of elevated ICP, recent lumbar puncture, history of optic atrophy and prior treatment of elevated ICP. As with all diagnostic tests used to evaluate elevated ICP, clinicians should not rely on one single value or measure to confirm or refute the diagnosis, but instead consider all clinical features.²⁷ We found that our results were in agreement with several prior studies, which found that cpRNFL swelling is indicative of papilledema.^{12, 14, 28–31}

In order to improve the clinical utility of these measurements, we considered the contribution of both BSUS and OCT and found that the combination of increased cpRNFL thickness on OCT and increased optic nerve sheath diameter on BSUS were most helpful in differentiating pseudopapilledema from a true elevation in ICP. Our findings suggest that patients with thickened cpRNFL on OCT and widening of the optic nerve sheath on BSUS are more likely to have true elevations in ICP. Only one other study combined components of both BSUS and OCT, and based on 3 subjects with papilledema, the authors concluded that these tests were not sufficient to detect intracranial pathology.³²

BSUS has long been the diagnostic modality of choice in identifying ONHD. However, as OCT technology has developed to include enhanced-depth imaging, visualization of deeper drusen has improved.¹⁸ The prevalence of ONHD in children is estimated to be between 1–14.6%, depending on the morphologic definition used.^{10, 33} The morphology of ONHD on OCT has been debated, with some studies describing hyporeflective masses with a hyperreflective margin,^{18, 34–36} and some identifying hyperreflective structures.^{28, 37, 38} Recent consensus recommendations established that previously reported hyperreflective structures are likely PHOMS, while true ONHD are hyporeflective, with full or partial hyperreflective margins.¹

Although some have suggested PHOMS may be precursors to ONHD,³⁹ many features of PHOMS are inconsistent with ONHD and they appear to be a non-specific finding in optic nerve head elevation, perhaps representing distended axons.^{1, 2} In the present study, we classified the hypo- and hyperreflective round nodules visualized on OCT to be either present or absent (independent of cpRNFL thickness) since it has been our experience that the hyperreflective calcified borders of drusen are less commonly visualized in children compared to adults and that the lesions classified at PHOMS were also found to be hyperreflective on BSUS (Figure 2). This finding is in contrast to other investigators who believe PHOMS are not visible on BSUS.¹ While the debate about whether these lesions should be classified as PHOMS or drusen is important, it is somewhat irrelevant to the clinician attempting to determine whether this child’s optic nerve head elevation is due to papilledema and requires additional testing—especially since both PHOMS and drusen are frequently found in children with papilledema.⁴⁰ While PHOMS/ONHD were more prevalent in the pseudopapilledema group (88% and 93% respectively), diagnostic utility was poor. Our data reinforce the concept that the presence of PHOMS and or ONHD does not rule out true elevations in ICP.

Figure 2.

Mixed hypo- and hyperreflective nodules on OCT B-scan (Top middle panel) that are also hyper-reflective on B-scan ultrasound (Top right panel). Purely hyperreflective nodule on OCT B-scan (Bottom left panel) that is also hyperreflective on B-scan ultrasound (Bottom right panel).

Some studies have found that, in cases of mild papilledema, cpRNFL thickness may not be significantly different between those with buried ONHD, as buried drusen may cause cpRNFL thickening without obvious hypo-/hyperreflective foci.^{16, 17, 20} Specifically, one study found no statistically significant difference in RNFL thickness between eyes with buried ONHD and mild papilledema.¹⁷ There are several possible explanations for these discrepancies including small sample size and ambiguity on classification of ONHD. Our results demonstrated good diagnostic utility of cpRNFL thickening to identify papilledema in our study population.

In conclusion, we present the diagnostic utility of OCT and BSUS to differentiate between children with papilledema and pseudopapilledema. In patients with a reassuring clinical presentation and with an optic nerve sheath diameter < 4.5 mm on BSUS and cpRNFL thickness < 140 microns on OCT, clinicians may feel comfortable foregoing further invasive testing, as the likelihood of papilledema in this population is low.