Introduction

Retinal nerve fiber layer cleavages (RNFLCs) are dark, spindle-shaped spaces along the major retinal vessels [1]. In 1992, Chihara et al. firstly reported RNFLCs in three highly myopic patients [2]. Subsequent clinical studies demonstrated the characteristic features of RNFLCs, and used different names to describe these lesions, such as inner retinal cleavage [1], paravascular retinal rarefaction [3], pseudo-defects of the retinal nerve fiber layer [4], paravascular abnormalities [5], paravascular lamellar hole [5] or retinal cyst [6], paravascular inner retinal defect [7], and paravascular defect [8]. Comparing with glaucomatous wedge-shaped nerve fiber layer defects, RNFLCs usually are disconnected from the optic nerve head and have empty spaces with irregular margin over the inner retina [1].

The association of RNFLCs with visual field (VF) defects was controversial [1, 4, 7]. Chihara et al. reported that the presence of RNFLCs was not associated with a significant visual dysfunction [2]. However, later studies showed different results on the structural-functional relationships in RNFLCs. Currently, the proposed causes of RNFLCs included epiretinal membrane (ERM) [5, 9], myopia [2, 10, 11], vitreous traction [3], and glaucoma [8]. Previous studies conducted by Hood et al. demonstrated an association between paravascular defect and arcuate glaucomatous defect. Local axon loss caused mechanical traction to the nerve fiber near the vessels was speculated to be the cause [8]. The functional impact and clinical relevance of RNFLCs in glaucoma patients require further elucidation.

A longitudinal follow-up study conducted by Hwang et al. examined ten eyes with various baseline characteristics and showed no functional and structural changes during a median follow-up of 9 months [1]. However, there is no information on the long-term stability of RNFLCs. In the present study, we collected long-term follow-up data and aimed to identify the structural and functional features of RNFLC in glaucoma suspects and patients during longitudinal follow-up.

Methods

This retrospective and observational case series was approved by the Institutional Review Board at National Taiwan University Hospital (NTUH) and followed the tenets of the Declaration of Helsinki. We enrolled cases with the diagnosis of primary open angle glaucoma (POAG), glaucoma suspect, or ocular hypertension (OHT) from the NTUH glaucoma clinics from January 2006 to December 2016 consecutively. Glaucoma suspect was defined by enlarged cupping without wedge or diffuse RNFL loss, either on fundus photography or OCT deviation map. Ocular hypertension was defined by elevated intraocular pressure (IOP) > 21 mmHg, cup-disc ratio ≤ 0.5, and the absence of wedge or diffuse RNFL loss, either on fundus photography or OCT deviation map. POAG was defined by the presence of wedge shape or diffuse RNFL defect, elevated IOP > 21 mmHg recorded at least once, and open angle on gonioscopic examination. Patients with preperimetric glaucoma were also enrolled. Basic ophthalmic examinations, including refractive status and biometry measurements (Lenstar LS 900, Haag-Streit, Koeniz, Switzerland), were recorded. Color fundus photography was taken by a fundus camera (Canon CR-DGi non-mydriatic fundus camera, Canon Inc., Tokyo, Japan). RNFLC was evaluated by an experienced glaucoma specialist (T.H.W.) on color fundus photography centered on disc and macula with a 40° range. Mydriatic fundus examination was also performed. Structures surrounding RNFLCs were assessed by the Cirrus OCT (Cirrus HD-OCT 4000, Carl Zeiss Meditec Inc., Dublin, CA, USA) with high-definition images HD 5-line raster scan horizontally. Patients with signal strength lower than seven, segmentation error, or staphyloma affecting OCT analysis were excluded. VF was examined by the SITA-Fast 24-2 program (Humphrey Field Analyzer II; Carl Zeiss Meditec Inc., Dublin, CA, USA). All examinations were performed at routine follow-ups scheduled at 6-month intervals for all patients.

The morphology of RNFLCs was classified as (1) cystoid space (CS), (2) vascular microfold (VM), and (3) lamellar hole (LH) (Fig. 1) [11]. Lesions with intraretinal hyporeflective spaces only were classified as CS. CSs that had the communications with vitreous cavity would be characterized as LH. VM indicated the tenting of the retinal vessels toward vitreous cavity. LH was classified as a higher grade lesion than CS in previous studies [5, 9]. The locations of RNFLCs were divided into superotemporal arcade (ST), inferotemporal arcade (IT), superonasal arcade (SN), and inferonasal arcade (IN). The corresponding VF defects were assessed by the Garway-Heath Map [12] and were defined by the pattern deviation map of two consecutive, reliable VF field tests that showed a cluster of three points with probabilities of < 5%, or a cluster of two points with probabilities of < 2%, or one point with probabilities of < 1% on the corresponding area of RNFLC. We defined that a RNFLC was associated with a defect on OCT deviation map by the presence of yellow or red signals within an area spanning 45° clockwise and 45° counterclockwise of the RNFLC on the OCT deviation map (Fig. 2). We defined that a RNFLC was corresponding to a defect on OCT deviation map by the presence of yellow or red signals at the same location (Fig. 3). The structural and functional progression in the cleaved area was assessed using color fundus photography, VF test, and OCT results by two experienced glaucoma specialists (J.Y.H. and T.H.W.) and one retina specialist (C.M.Y.) without knowing each other’s assessment first. If the result of assessment was not consistent, final consensus would be reached after discussion among three interpreters. The follow-up period was calculated from the date of first photo taken to that of the last fundus photography. Structural change was defined by the widening or narrowing of the pre-existing cleavage, development of new area or change in bridging tissues presented on color fundus photography, or the red pixels of the corresponding area on guided progression analysis (GPA) of OCT. Development of new corresponding scotoma and widening or deepening of pre-existing scotoma were coined as functional declines. The area of RNFLC at the beginning and at the end of follow-up was analyzed by Fiji (an extension of ImageJ software) [13]. The area of RNFLC and optic disc was delineated on the color fundus photography and the numbers of pixel of the circled area were calculated with Fiji. The number of pixels of RNFLC was converted into the area based on the ratio of the number of pixels of optic disc on fundus photography and the area measured by OCT.

Fig. 1
figure 1

The morphological classification of RNFLCs. a Cystoid space (CS). b Vascular microfold (VM). c Lamellar hole (LM)

Full size image
Fig. 2
figure 2

A representative image of a RNFLC which was associated with a defect on OCT deviation map. a Association is defined if the abnormality (empty arrowhead) was presented on the OCT deviation map (inset) in the area 45° clockwise and 45° counterclockwise (dotted line) encompassing the RNFLC (arrow). b The high definition images of 5-line raster scans using Cirrus HD-OCT in the cleaved area. The glaucomatous wedge defect (empty arrowhead) and a cystoid space (arrow) were located in the area between two vessels (arrowhead)

Full size image
Fig. 3
figure 3

A representative image of a RNFLC with a corresponding defect on OCT deviation map. a There is a RNFLC (arrow) noted near the superotemporal arcade on the color fundus photography. There is another RNFL (arrowhead) near the inferotemporal arcade on the color fundus photography but showed no corresponding defect on OCT deviation map. b The defect on OCT deviation map (arrow) corresponded with the location of RNFLC on fundus photography (arrow). c The high-definition images of 5-line raster scans using Cirrus HD-OCT showed a cystoid space (arrow) at the cleaved area

Full size image

We assigned patients to the high myopia group if their axial length measurements were longer than 26.0 mm. We used linear mixed model to compare continuous variables, and Pearson’s chi-squared test and Fisher’s exact test to evaluate categorical variables. A p value of < 0.05 was considered statistically significant. All statistical analyses were performed using SPSS version 19 (IBM Corp. Released 2010. IBM SPSS Statistics for Windows, Armonk, NY, USA).

Results

We enrolled 30 subjects and observed 62 locations of RNFLCs in 43 eyes. There were 14 men (46.7%) and 16 women (53.3%) with a mean age of 42.0 ± 12.6 years (median 44, range 15–70). The follow-up period averaged 66.8 ± 37.8 months (median 70, range 7–126.8 months). The diagnosis of patients included POAG, OHT, and glaucoma suspect. The mean deviation (MD) in VF results was − 2.2 ± 2.4 dB, with one moderately glaucomatous eye (MD = − 6.77 dB) and one severely glaucomatous eye (MD = − 12.8 dB). No patients received any ocular surgical interventions other than cataract operation. Detailed information was listed in Table 1.

Table 1 Ocular characteristics of study population
Full size table

Locations and types of RNFLC

For the 62 defect locations, RNFLCs occurred most commonly along ST (53.2%), followed by IT (43.5%), SN (1.6%), and IN (1.6%) (Table 1). The possible combinations of RNFLC locations in the same eye included ST and IT (37.2%), IT and SN (2.3%), and ST and IN (2.3%). In eyes having RNFLCs in both ST and IT, the ST lesions were more prominent in 12 eyes (75%). RNFLC had the association with vein in 30 locations (48.4%), with artery in seven locations (11.3%), and involved both vein and artery in 25 locations (40.3%). Focal vitreoretinal traction was noted by OCT in 26 locations (41.9%) and 12 eyes (27.9%) showed evidence of localized separation of vitreoretinal adhesion along vessels. Among focal vitreoretinal tractions, there were 18 locations (69.2%) at ST, seven locations (26.9%) at IT, and one location (3.8%) at IN. Morphological classification delineated CS in all 62 locations (100%), VM in 44 locations (71%), and LH in 24 locations (38.7%). Some locations demonstrated more than one type of RNFLC. Twenty-three locations (37.1%) had CS and VM, three locations (4.8%) had CS and LM, and 15 locations (24.2%) had all three types of RNFLCs. Among the lesions of LH type, 14 locations (58.3%) had focal vitreoretinal traction nearby, nine locations (37.5%) had localized separation of vitreoretinal adhesion along the vessels, and one location (4.2%) had neither.

Functional and structural characteristics of RNFLC

Nine locations (14.5%) of RNFLC had corresponding defects on OCT deviation map. Thirty-one locations (50%) of RNFLC were associated with defects on OCT deviation map. Three locations (4.8%) of RNFLC had corresponding VF defects meeting the criteria. There was no functional or structural progression in the study subject during follow-up visits as determined by color fundus photography, VF test, and GPA of OCT interpreted by three ophthalmologists. There was no statistically significant change of RNFLC area during follow-up judged by paired t test (p = 0.268).

High myopia vs. non-high myopia

Among the 43 eyes included in this study, 36 eyes (83.7%) had axial lengths longer than 26.0 mm and were classified as highly myopic eyes. The clinical characteristics of the highly myopic and non-highly myopic eyes are listed in Table 2. There were no significant differences in the incidence of association with OCT defect, corresponding VF defect, presence of ERM, localized separation of vitreoretinal adhesion along the vessels, and focal vitreoretinal traction between highly myopic and non-highly myopic eyes. The non-high myopic eyes had more defect locations with LH than the highly myopic eyes (p = 0.038). More of the highly myopic eyes possessed more than one lesion in the same eye (47.2%). However, due to small sample size in the non-high myopia group, this finding did not achieve statistical significance (p = 0.209).

Table 2 Clinical characteristics of the eyes enrolled categorized by axial length
Full size table

Glaucoma vs. non-glaucoma

Comparing eyes with and without definite diagnosis of glaucoma (Table 3), there was no difference in axial length, spherical equivalent, presence of localized separation of vitreoretinal adhesion along the vessels, vitreous traction, and corresponding VF defect. Most glaucoma patients had mild VF defect with the mean deviation of − 2.9 ± 3.0 dB. There was only one moderately glaucomatous eye (MD = − 6.77 dB) and one severely glaucomatous eye (MD = − 12.8 dB). Therefore, the MD in VF test was not significantly worse than the control group (p = 0.060). The RNFLCs in glaucomatous eyes were more frequently associated with OCT defects on deviation map, and this was the only significant difference between these two groups (p = 0.021). There was no difference in the presence of LH (p = 1.000) or the number of eye possessing more than one lesion (p = 0.818).

Table 3 Clinical characteristics of the eyes enrolled categorized by diagnosis of glaucoma
Full size table

Discussion

RNFLC is frequently misinterpreted as a glaucomatous retinal fiber layer defect [1]. Our study demonstrated that RNFLC had predilection for superotemporal location and paravenous association. All cleaved area showed CS on OCT, while VM, LH, and mixed type lesions can also be detected. Corresponding VF defects and OCT deviation map defects were detected in a small portion of patients. However, long-term follow-ups revealed no functional or structural progression.

The tangential traction force produced by ERM or the locations of glaucomatous defect may affect the distribution of RNFLC [5, 8, 9]. However, there is no well-established explanation for the predilection for ST. In our study, more than half of RNFLCs were located along the superior arcade. In the eyes having RNFLCs in both ST and IT, 75% of them had more prominent lesions at ST. Meanwhile, nearly 70% of the lesions with focal vitreoretinal traction occurred at ST. Ito et al. demonstrated by OCT that the posterior vitreous detachment (PVD) process initially enlarges upward to the superior arcade and then extends downward to the inferior arcade in eyes with idiopathic macular hole [14]. The authors suggest that PVD may occur at superior vascular arcade earlier than inferior arcade because of gravity pulling down the upper part of the vitreous. We therefore inferred that RNFLC is more likely to be detected at the upper arcade in eyes that are early in the progression of PVD because localized separation of vitreoretinal adhesion along the vessels may occur first.

In the literatures, the results of VF defect and RNFLC show conflicting relationships [1, 4, 7, 9, 10]. Muraoka et al. reported that 85.4% of myopic patients with paravascular inner retinal defect had corresponding VF defects using Goldmann perimetry [7]. Miyoshi et al. showed that higher grade of paravascular inner retinal defect was associated with more VF defects in patients having ERM [9]. In contrast, Hwang et al. [1] and Tuulonen et al. [4] found no corresponding VF defects on standard automated perimetry. Our study showed corresponding VF defect in 4.9% of the lesions by standard automated perimetry. Different factors associated with RNFLC may account for the discrepancy. In patients with ERM, functional defects associated with RNFLC may be more prominent because of stronger tangential traction. However, in RNFLC associated with high myopia or glaucoma, functional defects seem to be minimal. The discrepancy may be also related to different examination methods. For example, corresponding VF defect was detected more frequently by Goldmann perimetry [7, 9] and less by standard automated perimetry [1, 4]. Komeima et al. also reported a case of paravascular inner retinal cleavage with corresponding relative scotoma on microperimetry [10]. Future study can consider to utilize microperimetry, as its’ correlation with retinal morphology may help to precisely locate the corresponding defect.

Based on previous studies, the causes of RNFLC can be categorized into tangential traction and anteroposterior traction. ERM exerted tangential traction to the retinal vessels, which were relatively resistant to the stretching force. In response to the traction force, paravascular tissues split and formed paravascular cysts [5]. Liu et al. showed that RNFLC disappeared on OCT and ophthalmoscope examination after ERM removal [5]. Therefore, ERM could be an independent factor of RNFLC.

In highly myopic eyes, axial elongation, posterior staphyloma formation, and the inflexibility of retinal vessels may all contribute to apply an anteroposterior traction and induced the formation of paravascular abnormalities, including paravascular cyst, VM, LH, and paravascular retinoschisis [2, 11, 15]. Previous studies have shown those paravascular abnormalities were more frequently detected in highly myopic eyes [11, 16, 17]. Ohno-Matsui et al. reported a case of RNFLC formation in 15 years with vitreoretinal adhesion and an increase of 1.5 mm in axial length [6]. We also found an 11-year-old subject who developed RNFL during follow-up. On the initial examination, axial length in the right eye was 25.01 mm and spherical equivalent was − 2.25 D, with no detection of RNFLC. RNFLC was first detected by the fundus photography when the subject was 16 years of age. There was no increase in his axial length after the age of 18, with the axial length 27.41 mm and the refraction − 8.25 D. The RNFLC had no functional or structural change during follow-up. The subject was diagnosed with OHT, and had neither ERM nor vitreoretinal adhesion on OCT examination. This case demonstrated that myopia may be an independent factor of RNFLC formation. Furthermore, the paravascular abnormalities in highly myopic eyes shared similar contributing factors with myopic foveoschisis; however, the latter may be visually disabling [11, 16, 17]. Myopic foveoschisis was reported to be associated with the presence of paravascular LH, different types of paravascular abnormalities, or paravascular retinoschisis [11, 17]. The splitting of paravascular retinal layers may transmit the stretching force to the macular area and contribute to the myopic foveoschisis [17].

Vitreoretinal traction also played an important role in RNFLC formation. Spencer et al. showed by autopsies that paravascular vitreoretinal attachment resulted in retinal rarefaction, and PVD further transformed this lesion to partial thickness retinal tear [3]. Using OCT, Shimada et al. demonstrated that the paravascular LH originated from the paravascular cyst after its inner wall was pulled away by vitreous traction [11]. Miyoshi et al. identified vitreous adhesion on 50% of paravascular cyst but none on LH [9]. Liu et al. also showed vitreoretinal adhesion alone could result in the formation of paravascular abnormalities [5]. In current study, the presence of LH was more common in non-highly myopic eyes, possibly having less liquefied vitreous and thus had more intense vitreoretinal traction comparing with highly myopic eyes. The presence of localized separation of vitreoretinal adhesion along the vessel or adjacent focal vitreoretinal traction was noted in nearly all locations with paravascular LH except one. This supported that vitreoretinal traction may lead to the formation of RNFLC.

The role of glaucoma in RNFLC is not well established. In glaucoma patients and suspects, some small hyporeflective spaces in RNFL detecting by circumpapillary OCT scans (1.7-mm-radius) were found to be associated with major blood vessels on wide field cube scan (9 × 12 mm) of swept source OCT [8, 18] and had similar morphological characteristics as RNFLC. Hood et al. proposed that local loss of axons might create a force pulling axons away from the nearest vessels and resulting in the formation of those paravascular defects [8]. Recently, Lee et al. proposed the gliosis scar-contraction-traction theory. Reactive gliosis occurred at peripapillary region with glaucomatous damage. Contraction of the scar tissue may produce tractional force transmitting to the RNFL defect margin [19]. This tangential traction may play a role in RNFLC formation. This hypothesis is supported by a higher level of association between RNFLCs and defects on OCT deviation map in glaucoma patients than in patients without definite diagnosis of glaucoma in our study. However, glaucomatous RNFL defects may be a less potent associated factor of RNFLC. Regarding the presence of LH or not and the number of RNFLC, we did not identify any difference between glaucoma patients and suspects in our study. Hood et al. also stated that the length of paravascular defect was shorter in eyes with arcuate defect or high myopia, while it was longer in eyes with ERM [8]. Because of the high proportion of highly myopic eyes and small number of the study population, the mechanism and the independent role of glaucoma in RNFLC cannot be elucidated in this study and more research is needed.

This study was limited by the heterogeneity of the study group, enrolling the glaucoma and glaucoma suspect patients, who may be highly myopic or not. However, the number of non-myopic patients with RNFLC was very few in our area due to the high prevalence of myopia in Asian population and the close association between myopia and RNFLC. Therefore, the independent role of glaucoma in RNFLC formation could not be elucidated in this study. Moreover, since the presence of RNFLCs was detected by color fundus photography, some smaller lesions may have not been noticed and the prevalence of RNFLC may be underestimated. A recent study of paravascular inner retinal abnormalities in normal healthy eyes using OCT showed the lesions were presented in 43% of eyes but none of them was detected on fundus photography [20]. This study also had some limitations due to the nature of retrospective study. For the analysis of RNFLC by OCT, the horizontal OCT scans were only performed at the cleaved areas; therefore, the authors were not able to perceive the vitreoretinal relationship thoroughly in areas without evident RNFLC. In 2015, Muraoka et al. suggested that longitudinal scans may reveal the morphology of inner retinal defect better than the horizontal scans [7]. As most of patients in this study were enrolled before 2015, only horizontal scans were performed in the cleaved area. In addition, most glaucoma patients in this study were in the early stage, and it is possible that advanced glaucoma exert more potent tractional force to the RNFL. However, it is worth to be mentioned that RNFLCs in glaucoma patients are usually detected in their early stages. As the condition of glaucoma progresses, RNFLCs may escape from being detected due to RNFL thinning. Since only two patients showed glaucoma progression, we could not further elucidate whether glaucoma progression would lead to functional and structural progression in RNFLC.

Despite of the weakness, we described the clinical characteristic of RNFLCs in early glaucoma patients and suspect, which were different from the glaucomatous RNFL defect. The predilection for superior arcade location may be explained by the sequence of PVD. Compared to ERM, glaucomatous defect may be a possible but less potent contributor to RNFLC development. But the independent role of glaucoma in RNFLC needs further research. Although some of RNFLCs may result in VF or structural defect, we found no evidence of structural or functional progression with the longest longitudinal follow-up in the literature, suggesting that RNFLC is stable in early glaucoma patients and suspect.