Neuro-Ophthalmology

Risk factors associated with progression from papilloedema to optic atrophy: results from a cohort of 113 patients

Abstract

Background The aim of this study was to assess the risk factors for atrophic progression of patients with papilloedema secondary to intracranial hypertension, using optical coherence tomography parameters.

Methods A retrospective study was conducted at Marseille University Hospitals’ Ophthalmology departments between December 2015 and December 2021. All patients with papilloedema resulting from elevated intracranial hypertension at the initial presentation were included. Ophthalmological evaluations included analysing retinal nerve fibre layer (RNFL), ganglion cell layer (GCL) and total peripapillary retinal thickness (RT).

Results The study included 222 eyes from 113 patients. The main aetiologies of intracranial hypertension were idiopathic intracranial hypertension (49/113), intracranial tumours (33/113) and cerebral venous thrombosis (15/113). The initial RNFL and RT showed significant correlations with optic atrophy. The mean RNFL was 199.63 µm in the ‘no atrophy’ group and 365.28 µm in the ‘atrophy’ group (p<0.001). Similarly, the mean RT was 483.72 µm in the ‘non-atrophy’ group and 796.69 µm in the ‘atrophy’ group (p<0.001). The presence of peripapillary haemorrhages showed a strong correlated with optic atrophy with an OR=19.12 (p<0.001). Impaired initial visual acuity was also associated with final optic atrophy with an OR=7.76 (p=0.020). Furthermore, impaired initial GCL was a major predictor of optic atrophy (OR=18.25 (p=0.021)).

Conclusion Our study highlights the risk factors for optic atrophy in papilloedema, aiming to facilitate the early detection of patients at a high risk of vision loss and enable more aggressive medical or surgical management.

What is already known on this topic

  • Recognised risk factors for atrophic progression of papilloedema is the initial grade of oedema, other risk factors are poorly identified. Despite this, previous studies tend to include Frisen classification and visual acuity tests, without the use of optical coherence tomography.

What this study adds

  • This study identified risk factors for optic atrophy in papilloedema. The initial oedema grade quantifiable with optical coherence tomography, severe visual acuity decline, ganglion cell damage and peripapillary haemorrhages predict optic atrophy progression.

How this study might affect research, practice or policy

  • It would be interesting to follow up after aggressive medical treatment those patients who show these characteristics.

Introduction

Papilloedema is the swelling of the optic nerves, caused by increased intracranial pressure (ICP). It is a major health problem, particularly because of its disabling symptoms, its aetiologies and its complications. Indeed, papilloedema can lead to optic nerve atrophy with visual field alterations, or even permanent vision loss due to optic nerve damage if patients are not promptly and appropriately treated.1

Some of the main causes include idiopathic intracranial hypertension (IIH)—an increasingly common disease in our society as obesity rates increase,2 brain tumours and cerebral sinus venous thrombosis. Surgical treatments for IIH—such as venous stenting or shunt surgeries—are currently only performed in refractory papilloedema and chronic stages of papilloedema. Similarly, brain tumours often do not initially require surgery. The resulting chronic elevation of ICP is highly linked to visual prognosis and presents a high threat to the patient’s vision, while invasive surgery also presents a risk for vision.3 4 Being able to quickly and accurately identify papilloedema with or without risk of optic nerve atrophy is thus essential for guiding neurologists, neurosurgeons and neuroradiologists in pathologies requiring a more aggressive management.

Currently, the stages of papilloedema are usually classified using a scoring system proposed by the Frisén scale.5 However, this scale is susceptible to high interevaluator and intraevaluator variability.6 Further, while one of the recognised risk factors for atrophic progression of papilloedema is the initial grade of oedema,7 other risk factors are poorly identified. Optical coherence tomography (OCT) is a quick, non-invasive, easily quantifiable technique that could allow objective measures of papilloedema ‘progression’, as well as identification of risk factors for atrophic progression. It has been shown, for example, that OCT volumetric measurements of papilloedema correlate with Frisén scale grade.8 Further, OCT can also help distinguish between normal optic nerve or pseudo-optic disc oedema and true optic disc oedema.9 Despite this, previous studies tend to include Frisén classification and visual acuity tests, without the use of OCT.7

The aim of this study was to assess the risk factors for atrophic progression of papilloedema patients with papilloedema secondary to IH, using OCT parameters.

Methods

Participants

We conducted a retrospective analysis, incorporating longitudinal follow-up data from patients with optic disc oedema who were recruited from the ophthalmology departments of Marseille University Hospitals, including both The Hospital of La Timone and North Hospital between December 2015 and December 2021. Opt-out consent was obtained from all patients. This is a method used in the recruitment of participants into research where information is provided to the potential participant regarding the research and the involvement, and where their participation is presumed unless they express to decline to participate. All patients, irrespective of age, with optic disc oedema secondary to elevated ICP at initial presentation, were included. Additional inclusion criteria were a recent diagnosis of papilloedema (less than 1 week after first finding of papilloedema if diagnosis was made with an in-office ophthalmologist), feasible OCT examination, and OCT follow-up for at least 3 months.

Exclusion criteria for correlation analyses were the absence of at least two retinal nerve fibre layer thickness (RNFL) OCT examinations from different days (patients had to have at least one follow-up 3 months after the initial visit), the presence of an ophthalmological disease or ophthalmological feature that could influence OCT imaging (eg, glaucoma, history of optic neuropathy, papillary dysversion, epiretinal membrane) or a brain tumour adjacent to the visual pathways as defined by experts from the Department of Neuroradiology at the Timone Hospital (tertiary centre).

Clinical measures

The following clinical data were collected: age, sex, symptoms (headache, nausea, vomiting, visual blur, diplopia, tinnitus, visual eclipses, epilepsy) and their speed of onset, body mass index (BMI), aetiology of papilloedema, presence of peripapillary haemorrhage at initial presentation, association with sixth cranial pair palsy (abducens nerve), visual field, visual acuity, opening ICP at lumbar puncture in strict lateral decubitus and therapeutic management. Initial diagnosis of IIH was confirmed according to the revised Dandy criteria.10

OCT protocol

OCT data were collected using a Heidelberg Spectralis Spectral Domain OCT (Heidelberg Engineering, Germany), Spectralis software V.5.3.3.0, Eye Explorer software V.1.6.4.0. The following OCT data were collected as part of a routine clinical OCT protocol: RNFL, total peripapillary retinal thickness (RT) (TR: total retina), ganglion cell layer (GCL) and scleral canal size. The RNFL was measured using a 3.4 mm circular scan around the optic nerve with the segmentation algorithm and the standardised eye-tracking enabled device protocol (TrueTrack). The TR was measured using the same image as for the RNFL, and corresponded to the thickness between the internal limiting membrane (ILM) and Bruch’s membrane (BM). RNFL and TR were automatically segmented by the device. Spectralis software of Heidelberg OCT was used to perform manual corrections of RNFL and TR segmentation when the automatic segmentation by the OCT showed incorrect layer segmentation of the ILM, RNFL or BM (figure 1). The verifications and modifications of the RNFL segmentations were performed by two ophthalmologist. The GCL was analysed when a macular OCT was available and automatically segmented by the device. It was thoroughly checked, and in cases of inaccurate automatic segmentation, the values were not considered. OCT data were collected several times during the follow-up: at diagnosis, at 3 months, at 6 months, at 12 months and at the patient’s last visit.

Figure 1
Figure 1

Comparison of automatic and manual segmentation of optic nerve head scans. (A) The infrared reflectance image illustrating the localisation of the 3.4 mm circular scan centred on the optic nerve head with the automatic segmentation. The automatic segmentation of internal limiting membrane (red superior), Bruch’s membrane (red inferior) and RNFL (blue) is inappropriate and shows a false global thickness (G). (B, C) Equivalent segmentation images and global thickness (G) after manual correction. RNFL, retinal nerve fibre layer.

Our primary outcome measure was the presence of optic nerve atrophy defined by the presence of an RNFL beyond 95% of the OCT database CI, and/or severe sectoral atrophy defined by a value beyond 99% of the OCT database CI

Statistical analysis

Data were analysed using RStudio software V.4.2.0 (2022-04-22 ucrt) (R Core Team 2022. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria) and the ggplot and dplyr packages. Descriptive results are presented using the mean±SD, minimum and maximum values, or the number and associated percentage (%) dependent on the data. Correlation analyses of quantitative variables were measured using the Pearson correlation coefficient. Analysis of the primary outcome was performed using logistic regression and Wilcoxon tests for univariate analyses, and multiple logistic regression. Relevant variables were included in the multiple model approach and were selected through a stepwise procedure using the Akaike information criterion. A p<0.05 indicated a statistically significant result.

Results

Cohort details

The study included data for 222 eyes across 113 patients with IH, 93 of whom were at La Timone Hospital and 20 at the North Hospital in Marseille, France. Data from 54 eyes were subsequently excluded from the correlation analyses: 34 were lost to follow-up due to having a follow-up of less than 3 months, 16 eyes were removed due to lack of OCT data, 3 eyes had an optic disc dysversion and 1 eye had intravitreal haemorrhage. Out of the 168 eyes finally included, 35 showed optic atrophy according to the primary outcome.

Demographic and clinical data

The average age of the patients was 24.3 years (±16,72), with a mean BMI of 25.9 kg/m2. The majority of patients had bilateral involvement, with only four patients presenting unilateral involvement, and mean follow-up was 13.9 months. The cohort included more women (64.6%).

The various aetiologies of IH in the cohort are detailed in figure 2. The main causes of IH, accounting for more than 80% of cases, were IIH (43.36%), intracranial tumours (29.20%) and cerebral venous thrombosis (13.27%). Among the IIH cases, only 3 were classified as fulminant, 15 as subacute and 31 as chronic. Of the intracranial tumours, these were found to be gliomas (10/33), meningiomas (5/33), cysts (5/33), medulloblastomas (3/33), craniopharyngiomas (3/33), metastases (2/33), pineal tumours (2/33), haemangioblastoma (1/33), schwannoma (1/33) and neuroblastoma (1/33). Among the gliomas, we found pilocytic astrocytomas (4/10), tectal plate gliomas (2/10), gangliogliomas (2/10), one brainstem glioma and one posterior fossa glioma.

Figure 2
Figure 2

Bar plot of the different aetiologies of intracranial hypertension in the cohort.

Out of the 113 patients, 52 were children under the age of 18. The aetiologies of IH in children were primarily intracranial tumours (19/33), IIH (18/46), cerebral venous thrombosis (5/15), cerebral haemorrhages (4/7), meningitis (3/6) and congenital stenosis of the aqueduct of Sylvius (3/3).

Headache (73%) was the most common symptom. Patients had the following ophthalmologic symptoms during their illness: visual blur (35%), diplopia (19%) and visual eclipses (10%). They also described nausea and/or vomiting (37%), tinnitus (5%) and seizures (5%). Sixty patients were treated pharmacologically with acetazolamide (n=60), sometimes in combination with topiramate (n=7) or furosemide (n=1). Repeated therapeutic lumbar punctures were performed on nine patients. The most frequently performed surgeries were excisional surgery (n=18), ventriculocisternostomy (n=14) and ventriculoperitoneal shunts (n=13). In rarer cases, stenting (n=1), external lumbar shunt (n=1), subdural-peritoneal shunt (n=1) and haematoma evacuation (n=1) were performed.

Initial visual acuity and follow-up

Initial visual acuity was less than or equal to 20/25 on the Snellen scale in 14.6% of eyes. Sixty-nine per cent (87/126) of the eyes with an initial visual field test had only a enlargment of the blind spot. The visual field was normal in 9.5% of patients (12/126). Finally, 21.5% (27/126) had severe impairment such as central or paracentral scotomas or even diffuse deficits.

At the time of the last follow-up, 5.4% of the eyes (12/222) had a visual acuity of 20/32 or less. Among them, three eyes had a visual acuity lower than 20/200.

OCT follow-up

The evolution of OCT parameters (RNFL, GCL, TR) is detailed in table 1. Statistical analyses of RNFL and RT follow-up showed that papilloedema decreased significantly during the first 3 months following diagnosis. The size of the scleral canal at diagnosis was 1880.15±272.06 µm.

Table 1
|
Evolution of the cohort*

Correlation between RNFL, OCT and demographic measures

Correlation analyses were performed between RNFL at last follow-up and demographic and OCT measures. Age at diagnosis, BMI and time between the appearance of first symptoms and the diagnosis were not correlated with final RNFL (table 2). A subgroup analysis in the IIH category was performed and also showed no correlation between BMI and final RNFL with a correlation coefficient of −0.19 (p=0.09).

Table 2
|
Univariate correlation analysis of quantitative variables against the final RNFL

The initial RNFL and the initial TR were significantly negatively correlated with the final RNFL with respective correlation indices of −0.65 (p<0.001) and −0.61 (p<0.001). The initial RNFL and the initial TR were also significantly negatively correlated with the final GCL, but with smaller correlation indices, respectively, −0.5 (p<0.001) and −0.45 (p<0.001).

Furthermore, we conducted correlation analyses between an initial altered GCL (defined by an initial GCL at D0 lower than 1mm3) and the final RNFL. The RNFL had an average of 102 µm in the group with intact GCL, whereas it was 78 µm in the group with altered GCL, with a p<0.01.

Other significant associations were found. Scleral canal size was moderately negatively correlated with final RNFL (cor=−0.39; p<0.01) as was opening ICP at lumbar puncture in strict lateral decubitus (cor=−0.28; p<0.01).

Demographic and clinical risk factors for final optic nerve atrophy

Correlation analyses were performed between qualitative variables and the primary outcome (table 3). Results showed that sex was not correlated with the presence of final optic nerve atrophy in the total cohort, nor in the IIH subgroup analysis with an OR=0.84 (95% CI 0.12 to 3.73, p=0.839). Furthermore, the presence of peripapillary haemorrhages at the initial fundus was strongly correlated with the presence of optic nerve atrophy at the final follow-up with an OR=19.12 (95% CI 3.92 to 111.37, p<0.001).

Table 3
|
Univariate and multivariate analysis of the risk factors for optic atrophy by primary outcome

The aetiology of IH did not correlate with the primary outcome. Idiopathic or secondary IH was not showing any risk of progression towards optic nerve atrophy. Initial visual acuity inferior or equal to 20/25 on the Snellen scale was also strongly correlated with final optic nerve atrophy with an OR=7.76 (95% CI 1.35 to 46.39, p=0.02). Visual field damages were significantly correlated with optic nerve atrophy in univariate analysis with an OR=37.51 (95% CI 10.97 to 176.88, p<0.001). However, they were not retained in the multivariate model. Some patients had an impaired GCL since the initial stage with a volume of less than 1 mm3. This impairment was a significant risk factor for optic nerve atrophy with an OR=18.25 (95% CI 1.94 to 416.59, p=0.021).

Analyses of first symptoms were also performed to determine whether certain symptoms were potentially suggestive of severity and thus of progression to optic atrophy. No significant association was found.

Relationship between OCT and primary endpoint in atrophic versus non-atrophic patients

Quantitative variables were analysed in relation to the primary outcome. The initial RNFL was significantly correlated with optic atrophy with a mean RNFL at 199.63 µm in the ‘non-atrophy’ group and 365.28 µm in the ‘atrophy’ group. Initial TR was also correlated with optic atrophy with a mean TR at 483.72 µm in the ‘non-atrophy’ group and 796.69 µm in the ‘atrophy’ group. However, in the multivariate model, only one of these two variables was retained by the analysis because they are highly dependent on each other. ICP and scleral canal size were significantly correlated with optic atrophy in univariate analysis but were not retained in the multivariate analysis due to a large amount of missing data.

Conclusions

This study on 113 patients from a real-life setting identifies risk factors long-term optical nerve atrophy after an increased intracerebral pressure, independent from the underlying aetiology and using OCT.

Total peripapillary retinal thickness has been used to quantify initial papilloedema because it is a more reliable value for high-grade papilloedema and because it compares favourably with the clinical staging of Frisén.6 Indeed, RNFL treatment algorithms often fails to provide a proper segmentation in high-grade papilloedema while total retinal thickness works more better.

Albrecht et al demonstrated that initial RNFL thickness is a predictor of RNFL loss over time in patients with IIH.11 The similar study by Kabatas et al prospectively followed eyes with papilloedema due to IIH. Patients with optic atrophy had the following characteristics: higher cerebrospinal fluid opening pressure, higher grade of papilloedema, thicker mean peripapillary RNFL thickness, thinner GCL layer and severe visual field loss.12 Our study confirms these findings in stasis papilloedema regardless of the underlying aetiology.

Older studies have already identified clinical risk factors for optic atrophy based on visual acuity impairment but without the use of OCT. They include the grade of papilloedema according to the Frisén classification, the presence of peripapillary haemorrhage and advanced age.7 Initial visual acuity is impaired in 5%–17% of patients in the literature.13 This clinical presentation is a major risk factor for visual pathway damage. A more recent study has also shown that the Frisen grade of papilloedema is also a risk factor for optic atrophy in the paediatric population.14

We chose to use OCT parameters as the primary outcome measure because it is an objective and reproducible criterion that reliably reflects anatomical optic loss. Regarding the functional consequences of optic atrophy, it has been demonstrated that visual acuity is generally preserved and is not the first visual function affected.15 16 It is not a reproducible and discriminative criterion reflecting the occurrence of optic atrophy. Therefore, we did not select it as the primary outcome measure. Nevertheless, it remains an important evaluation criterion since there is still a risk of decreased visual acuity, as it was the case in 5% of our patients.

Visual field testing is a much better assessment of visual function in patients with optic atrophy. Optic atrophy can lead to visual field defects characterised by diffuse and unsystematic scotomas with a decrease in mean deviation. Because this is a retrospective study, most of the visual fields conducted were Goldman visual fields rather than automated visual fields due to the routine practices of the services. Furthermore, there were missing values of mean deviation, which prevented us from choosing it as the primary outcome measure. Because of this methodological difference, it is not possible to directly compare the data from Goldman visual fields with that of automated visual fields. However, it has been shown that RNFL thickness is strongly correlated with mean deviation in optic atrophy, regardless of the underlying causes,15 confirming the relevance of analysing the final RNFL as a marker of optic atrophy occurrence. Future analyses are required to confirm the correlation between final RNFL and visual field in cases of optic atrophy due to papilloedema.

Severe optic disc oedema with elevated ICP makes the optic disc highly vulnerable to ischaemia because of compression of the vessels within. There is an increased vascular resistance due to the compression of capillaries by swollen axons and accumulation of extracellular fluid in the oedema and venous stasis.17 According to Micieli et al, the severity of papilloedema is not associated with age, race, gender or BMI in patients with IH but is significantly higher with elevated CSF opening pressure.18 In this study, there was no observed correlation between BMI and optic atrophy, which aligns with the findings in Orcutt et al.'s study, where no significant correlation was found in the idiopathic intracranial hypertension (IIH) subgroup.7 On the other hand, a retrospective study of 414 patients found a tendency to loss of vision associated with a severe BMI (greater than 40 kg/m2) with an OR=1.4 (95% CI 1.03 to 1.91, p=0.03).19

Unilateral involvement represented 3.5% of our patients. This is a rare phenomenon but described in the literature as being around 4% in papilloedema.20 One study suggests that asymmetric papilloedema stems from asymmetries in the bony optic canal that interfere with the flow of cerebrospinal fluid between the perioptic subarachnoid spaces.20

This study also found a strong association between the alteration of the initial GCL and the final optic atrophy. We used GCL volume instead of the thickness because we did not have access to thickness values at the inception of the study. Furthermore, both in our clinical practice and in the literature, we observe that GCL volume is a highly relevant metric for assessing the entire GCL layer, as it reflects a potential localised loss that could be missed if we were to rely solely on the average thickness value. Hence, we employed this data to evaluate the MGCL. This strong association between the alteration of the GCL and the final optic atrophy, which is consistent with the study by Marzoli et al, indicates that, in patients with IH, signs of optic neuropathy can be identified even in the presence of papilloedema.21 Thus, OCT, which is already useful for quantifying papilloedema, may show retinal ganglion cell damage in an early phase of the disease.22

The sex ratio in our study in favour of females is explained by a preponderance of females in IIH and cerebral venous thrombosis with the important role of sex hormones.23,24 However, male or female sex is neither a risk factor nor a protective factor for optic atrophy, which partly contradicts the Bruce et al study in which men with IIH were twice as likely as women to develop severe visual loss, but this study used visual acuity as the principal outcome.25

The size of the scleral canal usually measures about 1.5 mm in healthy patients. This cohort presents a larger size because of a higher cerebrospinal fluid pressure in the optic nerve sheath for patients with IH. It also appears to be moderately correlated with progression to optic atrophy.

OCT is a marker of treatment efficacy and disease progression in IH.26 It is a quick and easy alternative to repeated lumbar punctures for monitoring patients with IIH, and facilitates monitoring of disease progression and response to treatment. In addition, several studies have shown that the rise in ICP correlates with the hike in RNFL during the follow-up of these patients.27 28

However, this study shows some limitations. It is a retrospective study and some clinical data from the interview or clinical examination were not always specified, such as the presence of certain symptoms or peripapillary haemorrhages. Also, children were mostly excluded from the study because OCT was often impossible to perform due to lack of compliance with the examination. The prevalence of IIH was low in our cohort because the more serious aetiologies (tumours, thromboses) are overestimated in connection with the region’s reference neurosurgery department attached to our ophthalmology department. However, the aetiology of IH had no correlation with the primary outcome. Finally, non-significant relationship could be due to an underpowered assessment. For example, ICP and scleral canal size are significantly correlated with atrophy in univariate analysis but were not included in the multivariate analysis due to a large amount of missing data.

In conclusion, this study identified risk factors for optic atrophy in papilloedema. The initial grade of oedema is easily quantifiable with OCT and is an important reflection of the risk of optic suffering. The presence of an initial severe visual acuity drop, early GCL damage or peripapillary haemorrhages are also associated with progression to optic atrophy. It would then be interesting to follow up after the medical treatment those patients who show these characteristics.