Diagnostic Ability of Retinal Nerve Fiber Layer Thickness Measurements and Neurologic Hemifield Test to Detect Chiasmal Compression

Abstract

Purpose.: The purpose of the study was to investigate the diagnostic ability of the neurologic hemifield test (NHT) and retinal nerve fiber layer (RNFL) thickness measurements to detect chiasmal compression.

Methods.: Thirty-seven patients with chiasmal compression, 35 patients with glaucoma, and 30 patients with glaucoma suspect were enrolled. The NHT score was established from a 30-2 visual field pattern deviation probability plot. Each test point value was calculated for a number that was inversely proportional to its pattern deviation probability. The NHT score was the absolute value of the difference in the sum of the point scores for two symmetrical regions of 16 points on either side of the vertical meridian. RNFL thickness was scored from 0 to 10 according to the pattern of RNFL loss and probability of abnormality. Solely present nasal or temporal RNFL defects were scored high, and exclusively appearing superior or nasal RNFL defects were scored low. The differences in the NHT and RNFL scores among the three groups were compared. Diagnostic ability was assessed by receiver operating characteristic (ROC) analysis.

Results.: The NHT and RNFL scores were significantly higher in patients with chiasmal compression than in patients with glaucoma and glaucoma suspect. The area under the ROC curve (AROC) was 0.734 of the NHT score and 0.613 of the RNFL score. When AROC was calculated using the NHT and RNFL scores concurrently, AROC was increased to 0.807.

Conclusions.: The NHT score and RNFL score have diagnostic ability to detect chiasmal compression, and simultaneous assessment of NHT and RNFL scores improves the diagnostic power.

Introduction

Compressive lesion of the optic chiasm can lead to compromised visual function involving defects in the visual field. Since the crossing nerve fibers in the chiasm that originate from retinal ganglion cells (RGCs) in the nasal hemiretina are damaged, chiasmal compression produces visual field defects that respect the vertical midline, typically causing bitemporal hemianopsia. ^1–5 Although identifying this pattern of visual field defect is crucial for the diagnosis of chiasmal compression, clinicians may verify this by manual inspection. Recently, Boland et al. have developed a novel algorithm, the neurologic hemifield test (NHT), which is an automated analysis of the quantitative differences across the vertical midline, to assist in the recognition of neurologic field defects in data from automated perimetry. ⁶ They demonstrated that the NHT distinguished neurologic field defects from those of glaucoma and rivaled the performance of subspecialist clinicians. 

Chiasmal compression can also lead to retinal nerve fiber layer (RNFL) loss in the nasal and temporal sectors of the optic nerve with relative preservation of the superior and inferior sectors. This pattern of RNFL loss is called band atrophy (BA) of the optic nerve and is opposite to the preferential thinning of superior and inferior regions seen in glaucomatous eyes. RNFL defect in the nasal and temporal areas of the optic disc is an important clue in the diagnosis of chiasmal compression. ^7,8 Optical coherence tomography (OCT) is a noninvasive method for obtaining cross-sectional images and provides reliable and reproducible RNFL thickness measurements. ^9,10 Moreover, previous studies have demonstrated that OCT is able to identify the characteristic pattern of RNFL loss in eyes with BA. ^8,11,12 In the present study, we constructed a simple algorithm using this concept, an RNFL score, from RNFL thickness measured by OCT to improve the detection of chiasmal compression. The aim of the study was to ascertain the diagnostic ability of the NHT and RNFL scores to detect chiasmal compression. 

Methods

Subjects

In this study, 37 patients with chiasmal compression who showed a visual field impairment as determined by standard automated perimetry (SAP), 35 age-matched patients with glaucoma, and 30 patients with glaucoma suspect were enrolled from the Ophthalmology and Neurosurgery Departments in our institution between September 2008 and November 2011. Glaucoma was diagnosed when the patient showed a characteristic structural damage to the optic nerve head and associated visual dysfunction. ¹³ Diagnostic criteria of glaucoma suspects were (1) abnormally large optic disc area, (2) large cup/disc ratio, (3) normal neuroretinal rim area and configuration, (4) normal form of zone alpha, (5) no zone beta, and (6) normal parapapillary RNFL. ¹⁴ A chiasmal compressive lesion and deviation of chiasmal structures were confirmed by magnetic resonance imaging (MRI). In patients with glaucoma, subjects were selected to have a mean deviation within 15% from the neurologic field pair for matching the level of severity with the neurologic visual field defects. This research adhered to the tenets of the Declaration of Helsinki. The institutional review board approved the research, and informed consent was obtained. 

All patients underwent a complete ophthalmic examination, including visual acuity (Va), intraocular pressure, refraction, slit lamp biomicroscopy, dilated fundus examination, SAP, and OCT within 2 weeks before surgery. SAP and OCT were performed separately by an expert technician who was blinded to the disease entity of the patient. Eyes were excluded if they had any anterior segment, retinal, posterior segment, or optic nerve disease other than compressive optic neuropathy and glaucoma; an unreliable visual field testing >20% false positive, false negative, or fixation loss; or a spherical refractive error outside the range of ±5 diopter. 

NHT Score

The NHT score was established as described in the study of Boland et al. ⁶ Automated perimetry was conducted using the central 30-2 Swedish Interactive Threshold Algorithm (SITA) on a visual field analyzer (Humphrey Visual Field Analyzer II; Carl Zeiss Meditec, Dublin, CA) with a Goldmann size III stimulation on a 31.5-apostilb background. In the pattern deviation probability plot, each test location is graded as normal or abnormal at a defined level compared to that in a normative population. An abnormal test point probability is reported in four levels (<0.5%, <1%, <2%, <5%). Along with the reported probability level, each test point was calculated for a defined number (Table 1) that was inversely proportional to its pattern deviation probability. Then the points were grouped into two symmetrical regions of 16 points on either side of the vertical meridian (Fig. 1). The range of pattern deviation probability and NHT point score were slightly modified from those of Boland et al. The final NHT score was the absolute value of the difference in the sum of the point scores for right and left regions. The higher NHT score from the two eyes of each patient was analyzed. 

Figure 1. 

 

The test point of NHT from a visual field test. Automated perimetry was conducted using the central 30-2. The points were grouped into two symmetrical regions of 16 points on either side of the vertical meridian (solid circles). The regions affected by the physiological blind spot and nasal periphery (open circles), where initial glaucomatous field defect is prevalent, were excluded. Each test point was calculated for a number that was inversely proportional to its pattern deviation probability. The final NHT score was the absolute value of the difference in the sum of the point scores for right and left regions.

Table 1. 

 

Scores for the Neurological Hemifield Test (NHT) from Pattern Deviation Probability Values

Pattern Deviation Probability	NHT Score
≥5%	0
<5%, ≥2%	2
<2%, ≥1%	5
<1%, ≥0.5%	10
<0.5%	20

RNFL Score

OCT imaging was conducted after pupil dilation using Spectral Domain OCT (Cirrus, software version 4.5.1.11; Carl Zeiss Meditec). RNFL thickness measurements were obtained using the Optic Disc Cube 200 × 200 protocol. OCT provides information on the probability of abnormality in patient examination results after comparison with an internal normative database. Since clinicians usually rely on the printout for identification of abnormalities, we constructed an RNFL score according to the normal distribution percentile and pattern of RNFL defect (Table 2). As the distribution of RNFL loss in eyes with BA is predominantly in the nasal and temporal quadrant, solely present nasal or temporal RNFL defects were scored high. In contrast, RNFL thinning preferentially occurs in superior and inferior regions; thus exclusively appearing superior or nasal RNFL defects were scored low. The higher RNFL score from the two eyes of each patient was analyzed. 

Table 2. 

 

Scores for Retinal Nerve Fiber Layer Thinning from Optical Coherence Tomography

Normal Distribution Percentiles	RNFL Score
Nasal or temporal defect
Nasal or temporal <1%	10	More probably chiasmal lesion
Nasal or temporal <5%	8
Combined defect
Any nasal or temporal defect + superior or inferior <5%	6
Any basal or temporal defect + superior or inferior <1%	4
Superior or inferior defect
Superior or inferior <5%	2	More probably glaucoma
Superior or inferior <1%	1
Within normal distribution	0

Statistical Analysis

An analysis of variance (ANOVA) and Scheffé post hoc analysis were conducted to compare the visual field test and RNFL measurements among chiasmal compression, glaucoma, and glaucoma suspect groups. Receiver operating characteristic (ROC) curves of the NHT score and RNFL score were drawn both separately and concurrently to assess their ability to distinguish field loss corresponding to chiasmal compression from that of glaucoma or glaucoma suspect patients. Areas under the ROC curve (AROC) were also evaluated. 

The ROC calculations were written in JMP_IN 4.0.2 (SAS Institute Inc., Cary, NC), and the other statistical analyses were performed with SPSS software version 15.0 for Windows (SPSS Inc., Chicago, IL). All tests were two-tailed, and P < 0.05 was considered statistically significant. 

Results

General Characteristics

In the present study, 37 patients with chiasmal compression (16 men and 21 women), 35 patients with glaucoma (20 men and 15 women), and 30 patients with glaucoma suspect (15 men and 15 women) were recruited. Among the 37 patients with chiasmal compression, 26 patients had pituitary tumor, 7 patients had craniopharyngioma, and 4 patients had suprasellar meningioma. The mean age was 45.97 ± 15.71 years in patients with chiasmal compression, 49.38 ± 13.58 years in patients with glaucoma, and 49.37 ± 11.77 years in patients with glaucoma suspect. The differences in age (P = 0.204) and sex ratio (P = 0.102) among the groups were not significant. The mean LogMAR Va was 0.34 ± 0.50 in patients with chiasmal compression, 0.32 ± 0.48 in patients with glaucoma, and 0.10 ± 0.12 in patients with glaucoma suspect. Va was significantly worse in chiasmal compression (P = 0.000) and glaucoma (P = 0.000) than in glaucoma suspect. 

NHT Score

Mean deviation (MD) was −10.30 ± 10.18 in patients with chiasmal compression, −9.39 ± 7.70 in patients with glaucoma, and 0.67 ± 0.94 in patients with glaucoma suspect. MD was significantly worse in chiasmal compression (P = 0.000) and glaucoma (P = 0.000) than in glaucoma suspect. The difference in MD between chiasmal compression and glaucoma was not significant (P = 0.739) (Table 3). 

Table 3. 

 

Demographic Characteristics, NHT Score, and RNFL Score of Enrolled Patients

	Chiasmal Lesion	Glaucoma	Glaucoma Suspect	ANOVA
Number	37	35	30
Diagnosis	Pituitary tumor (26), craniopharyngioma (7), suprasellar meningioma (4)
Age	45.97 ± 15.71	49.38 ± 13.58	49.37 ± 11.77	0.204
Sex ratio (M/F)	0.73	1.30	1.00	0.102
Visual field test
MD	−10.30 ± 10.18	−9.39 ± 7.70	0.67 ± 0.94	0.000
PSD*	6.83 ± 4.85	8.02 ± 4.32	1.65 ± 0.34	0.000
NHT score	128.49 ± 80.65	61.67 ± 45.14	19.60 ± 10.53	0.000
Retinal nerve fiber layer thickness
Superior	108.27 ± 29.22	94.93 ± 18.41	115.37 ± 15.63	0.000
Inferior	115.41 ± 22.50	78.73 ± 18.61	118.10 ± 15.70	0.000
Nasal	61.22 ± 11.61	62.44 ± 7.48	67.10 ± 6.66	0.018
Temporal	57.44 ± 17.66	60.67 ± 9.38	65.40 ± 12.17	0.040
Average	86.03 ± 16.19	74.01 ± 7.96	91.50 ± 9.18	0.000
RNFL score	3.52 ± 2.39	2.30 ± 1.89	0.63 ± 0.32	0.000

*  PSD, pattern standard deviation.

The NHT score was 128.49 ± 80.65 in chiasmal compression, 61.67 ± 45.14 in glaucoma, and 19.60 ± 10.53 in glaucoma suspect. The NHT score was significantly higher in chiasmal compression than in glaucoma (P = 0.000) and glaucoma suspect (P = 0.000). The difference in the NHT score between glaucoma and glaucoma suspect was not significant (P = 0.121) (Fig. 2). 

Figure 2. 

 

Comparison of NHT (A) and RNFL (B) score by diagnostic group. Data points indicate mean ± one standard deviation. The NHT score and RNFL score were significantly higher in chiasmal compression than in glaucoma and glaucoma suspect.

RNFL Score

Superior and inferior RNFL thicknesses were significantly worse in patients with glaucoma than in patients with chiasmal compression (P = 0.000) and glaucoma suspect (P = 0.000). Nasal RNFL thickness was significantly worse in patients with chiasmal compression than in patients with glaucoma suspect (P = 0.020); however, nasal RNFL thickness was not significant compared to values in glaucoma patients (P = 0.781). Temporal RNFL thickness was also significantly worse in patients with chiasmal compression than in those with glaucoma suspect (P = 0.042), but it was not significant compared to values in glaucoma patients (P = 0.477). 

RNFL score was 3.52 ± 2.39 in patients with chiasmal compression, 2.30 ± 1.89 in patients with glaucoma, and 0.63 ± 0.32 in patients with glaucoma suspect. RNFL score was significantly higher in chiasmal compression than in patients with glaucoma (P = 0.027) and glaucoma suspect (P = 0.000). The difference in RNFL score between glaucoma and glaucoma suspect was also significant (P = 0.014) (Fig. 2). 

ROC Analysis

The NHT score showed 94.44% sensitivity and 42.19% specificity for distinguishing visual field loss corresponding to chiasmal compression at a cutoff value of 135. The AROC of the NHT score was 0.734. Among 37 patients with chiasmal compression, 63% of 24 patients had higher NHT scores than the cutoff value of 135. 

The RNFL score demonstrated 95.83% sensitivity and 27.59% specificity for diagnosing chiasmal compression at a cutoff value of 6. The AROC of the RNFL score was 0.613. Among 37 patients with chiasmal compression, 53% of 20 patients had higher RNFL scores than the cutoff value of 6. 

When AROC was calculated from the NHT score and RNFL score concurrently, AROC was increased to 0.807. In addition, among 37 patients with chiasmal compression, chiasmal compression in 90% of 34 patients was detected when NHT and RFNL scores were examined simultaneously (Fig. 3). 

Figure 3. 

 

Receiver operating characteristic curve of NHT (A), RNFL (B), and combined (C) scores for diagnosing chiasmal lesion from glaucoma. The AROC for the NHT score was 0.734, and that for the RNFL score was 0.613. AROC increased to 0.807 when AROC was calculated from the NHT and RNFL scores concurrently. Cutoff value for the NHT score was 135; that for the RNFL score was 6.

Discussion

In the present study, we investigated the diagnostic ability of NHT and RNFL scores to detect chiasmal compression, and the results demonstrated that the NHT and RNFL scores can effectively distinguish patients with chiasmal compression from patients with glaucoma and glaucoma suspect. 

The NHT score evidenced an ability to discriminate chiasmal compression from glaucoma. Several algorithms have been developed to assist the interpretation of automated perimetry, including the Glaucoma Hemifield Test (GHT). However, these algorithms tend to concentrate on glaucoma. ^15,16 The NHT is an automated analysis of the quantitative differences across the vertical midline. It was proposed to aid identification of a certain type of visual defect that is caused by chiasmal or postchiasmal neurologic disease. ⁶ Boland et al. enrolled various patients with either a chiasmal lesion or a postchiasmal lesion causing homonymous field loss to test the ability of the NHT score to distinguish neurologic field defects from those of glaucoma. The NHT score was significantly higher in patients with chiasmal compression than in patients with glaucoma and glaucoma suspect. ROC analysis also showed significant diagnostic ability to differentiate the visual field defect caused by chiasmal compression from that caused by glaucoma. These results suggest that the NHT score can be a useful tool for detecting chiasmal compression. However, the sensitivity, specificity, and diagnostic power of the NHT score in this study were lower than in the study by Boland et al. 

The RNFL score also showed an ability to distinguish chiasmal compression from glaucoma. Crossing nerve fibers in the chiasm originating from RGC axons in the nasal hemiretina, including the papillomacular bundle, generally enter the disc through the nasal and temporal poles. In contrast, uncrossed fibers in the chiasm originating from RGC axons in the temporal hemiretina generally enter the disc through the superior and inferior poles. ^7,12 Therefore, chiasmal compression can induce RNFL loss in the nasal and temporal sectors of the optic nerve with relative preservation of the superior and inferior sectors, and this pattern of RNFL loss is called BA of the optic nerve. A diagram of an optic nerve cross section with BA is shown in Figure 4. ^7,17 We established an algorithm to improve the detection of chiasmal compression, an RNFL score using the pattern of RNFL loss, that automatically provided information on the probability of abnormality from the OCT (Table 2). The RNFL score was significantly greater in patients with chiasmal compression than in patients with glaucoma and glaucoma suspect. ROC analysis showed a significant diagnostic ability to discriminate the RNFL loss caused by chiasmal compression from that in glaucoma and glaucoma suspect. These results suggest that the RNFL score can also be a valuable test to detect chiasmal compression. 

Figure 4. 

 

Diagram of optic nerve cross section with band atrophy. Crossing nerve fibers in the chiasm originating from retinal ganglion cell (RGC) axons in the nasal hemiretina (gray) generally enter the disc through the nasal and temporal poles. Uncrossed fibers in the chiasm originating from RGC axons in the temporal hemiretina (white) generally enter the disc through the superior and inferior poles. Chiasmal compression induces RNFL loss in the nasal and temporal sectors of the optic nerve with relative preservation of the superior and inferior sectors.

The RNFL score is based on the specificity of the pattern of RNFL loss in chiasmal compression, not on the sensitivity or severity of RNFL loss. Exclusive loss of RNFL in the temporal and/or nasal portion is not more common than loss in the nasal and/or temporal areas in association with some loss in the superior and/or inferior sector; however, the pattern of RNFL loss that is more limited to the nasal and/or temporal area is more specific for chiasmal compression than broad losses of RNFL, including losses in the superior or inferior sections. Thus the exclusive loss of RNFL in the temporal/nasal portions was scored higher. It is obvious that even in BA, some superior or inferior RNFL losses are accompanied by temporal or nasal RNFL losses. ¹⁷ However, the more the chiasmal compression is restricted to the center of the chiasm, the more limited the RNFL defects are to the temporal or nasal sectors. This can be fully expected based on the histologic study of Unsöld et al. ⁷ Moreover, Kanamori et al. ¹² divided the eyes of chiasmal compression into three groups according to the degree of temporal hemianopia: namely, group 1, normal; group 2, ≤1 quadrant defect; and group 3, >1 quadrant defect. In group 1, patients showed a relatively larger RNFL reduction rate of 26.6% in the temporal portion and 29.1% in the nasal portion, which was more than twice the small RNFL reduction rates of 10.0% in the inferior and 13.9% in the superior portion. However, in groups 2 and 3, both the nasal/temporal and superior/inferior portions showed large RNFL reduction rates of over 30%, and the discrepancy between nasal/temporal and superior/inferior areas was reduced. These results suggested that RNFL defects were more limited and prominently present in the nasal or temporal portion when the chiasmal compression was small and restricted to the chiasm, with subtle visual field defects. As the chiasmal compression became larger, the superior/inferior portions were increasingly involved, and the RNFL scores decreased. However, in this situation, the visual field defects extend toward typical complete temporal hemianopia, and the NHT scores increases. Therefore, the RNFL score and NHT score are complementary for diagnosing chiasmal compression. 

Concurrent evaluation of the NHT score and RNFL score can improve the diagnostic ability to detect chiasmal compression. In the ROC curve, the NHT and RNFL scores corresponding to the highest average sensitivity and specificity were 135 and 6, respectively. Using these cutoff values, among 37 patients with chiasmal compression, 63% of 24 patients were diagnosed with the NHT score, and 53% of 20 patients were diagnosed with the RNFL score. However, when the NHT and RNFL scores were applied simultaneously, 90% of 34 patients were diagnosed with chiasmal compression. AROC analysis also revealed that the diagnostic power was higher when the NHT score and RNFL score were used concurrently than when they were used individually. These results suggest that the diagnostic power of the RNFL score is on a par with and complementary to that of the NHT score. 

Many studies have been performed using the visual field tests and OCT in differential diagnosis and evaluation of chiasmal compression, and this study is in the same context as the previous studies. ^{4,5,12,18–20} NHT and RNFL scores are based on well-established concepts of bitemporal hemianopia and BA that is caused by chiasmal compression. The novelty of this study is the systematic quantification of pattern of visual field and RNFL defect using the scoring algorithms. The results of this study evidenced that NHT and RNFL scores can be useful for diagnosing chiasmal compression and that concurrent evaluation of the NHT and RNFL scores improves the diagnostic performance. These algorithms for identifying chiasmal compression may be useful especially to clinicians who are not neuro-ophthalmologic specialists. 

There are several limitations to the present study. Chiasmal compression induces binocular visual field defects; however, NHT and RNFL scores are based on monocular analysis. A binocular analysis algorithm might improve the diagnostic ability. In the ROC analyses, NHT and RNFL scores showed low specificity. The specificity was 42.19% for the NHT score and 27.59% for the RNFL score. Further investigations, including use of a greater number of individuals and adjustment in cutoff values, are expected to enhance the diagnostic ability and value of the test. 

In conclusion, the NHT score and the RNFL score have diagnostic ability to detect chiasmal compression, and concurrent assessment of the NHT and RNFL scores improves the diagnostic power. NHT and RNFL scores can be useful supplementary tools for detecting a chiasmal compression. 

Footnotes

 Disclosure: C.H. Moon, None; S.H. Lee, None; B.-T. Kim, None; S.C. Hwang, None; Y.-H. Ohn, None; T.K. Park, None