Retinal layer segmentation in multiple sclerosis

a systematic review and meta-analysis

ERN-EYE IMSVISUAL

Research output: Contribution to journalArticle

65 Citations (Scopus)

Abstract

Background Structural retinal imaging biomarkers are important for early recognition and monitoring of inflammation and neurodegeneration in multiple sclerosis. With the introduction of spectral domain optical coherence tomography (SD-OCT), supervised automated segmentation of individual retinal layers is possible. We aimed to investigate which retinal layers show atrophy associated with neurodegeneration in multiple sclerosis when measured with SD-OCT. Methods In this systematic review and meta-analysis, we searched for studies in which SD-OCT was used to look at the retina in people with multiple sclerosis with or without optic neuritis in PubMed, Web of Science, and Google Scholar between Nov 22, 1991, and April 19, 2016. Data were taken from cross-sectional cohorts and from one timepoint from longitudinal studies (at least 3 months after onset in studies of optic neuritis). We classified data on eyes into healthy controls, multiple-sclerosis-associated optic neuritis (MSON), and multiple sclerosis without optic neuritis (MSNON). We assessed thickness of the retinal layers and we rated individual layer segmentation performance by random effects meta-analysis for MSON eyes versus control eyes, MSNON eyes versus control eyes, and MSNON eyes versus MSON eyes. We excluded relevant sources of bias by funnel plots. Findings Of 25 497 records identified, 110 articles were eligible and 40 reported data (in total 5776 eyes from patients with multiple sclerosis [1667 MSON eyes and 4109 MSNON eyes] and 1697 eyes from healthy controls) that met published OCT quality control criteria and were suitable for meta-analysis. Compared with control eyes, the peripapillary retinal nerve fibre layer (RNFL) showed thinning in MSON eyes (mean difference −20·10 μm, 95% CI −22·76 to −17·44; p<0·0001) and in MSNON eyes (–7·41 μm, −8·98 to −5·83; p<0·0001). The macula showed RNFL thinning of −6·18 μm (–8·07 to −4·28; p<0·0001) in MSON eyes and −2·15 μm (–3·15 to −1·15; p<0·0001) in MSNON eyes compared with control eyes. Atrophy of the macular ganglion cell layer and inner plexiform layer (GCIPL) was −16·42 μm (–19·23 to −13·60; p<0·0001) for MSON eyes and −6·31 μm (–7·75 to −4·87; p<0·0001) for MSNON eyes compared with control eyes. A small degree of inner nuclear layer (INL) thickening occurred in MSON eyes compared with control eyes (0·77 μm, 0·25 to 1·28; p=0·003). We found no statistical difference in the thickness of the combined outer nuclear layer and outer plexiform layer when we compared MSNON or MSON eyes with control eyes, but we found a small degree of thickening of the combined layer when we compared MSON eyes with MSNON eyes (1·21 μm, 0·24 to 2·19; p=0·01). Interpretation The largest and most robust differences between the eyes of people with multiple sclerosis and control eyes were found in the peripapillary RNFL and macular GCIPL. Inflammatory disease activity might be captured by the INL. Because of the consistency, robustness, and large effect size, we recommend inclusion of the peripapillary RNFL and macular GCIPL for diagnosis, monitoring, and research. Funding None.

Original languageEnglish (US)
Pages (from-to)797-812
Number of pages16
JournalThe Lancet Neurology
Volume16
Issue number10
DOIs
StatePublished - Oct 1 2017

Fingerprint

Multiple Sclerosis
Meta-Analysis
Optic Neuritis
Nerve Fibers
Optical Coherence Tomography
Ganglia
Atrophy

ASJC Scopus subject areas

  • Clinical Neurology

Cite this

Retinal layer segmentation in multiple sclerosis : a systematic review and meta-analysis. / ERN-EYE IMSVISUAL.

In: The Lancet Neurology, Vol. 16, No. 10, 01.10.2017, p. 797-812.

Research output: Contribution to journalArticle

@article{36dc520d8a0a4bfeb1b28cd3f866845c,
title = "Retinal layer segmentation in multiple sclerosis: a systematic review and meta-analysis",
abstract = "Background Structural retinal imaging biomarkers are important for early recognition and monitoring of inflammation and neurodegeneration in multiple sclerosis. With the introduction of spectral domain optical coherence tomography (SD-OCT), supervised automated segmentation of individual retinal layers is possible. We aimed to investigate which retinal layers show atrophy associated with neurodegeneration in multiple sclerosis when measured with SD-OCT. Methods In this systematic review and meta-analysis, we searched for studies in which SD-OCT was used to look at the retina in people with multiple sclerosis with or without optic neuritis in PubMed, Web of Science, and Google Scholar between Nov 22, 1991, and April 19, 2016. Data were taken from cross-sectional cohorts and from one timepoint from longitudinal studies (at least 3 months after onset in studies of optic neuritis). We classified data on eyes into healthy controls, multiple-sclerosis-associated optic neuritis (MSON), and multiple sclerosis without optic neuritis (MSNON). We assessed thickness of the retinal layers and we rated individual layer segmentation performance by random effects meta-analysis for MSON eyes versus control eyes, MSNON eyes versus control eyes, and MSNON eyes versus MSON eyes. We excluded relevant sources of bias by funnel plots. Findings Of 25 497 records identified, 110 articles were eligible and 40 reported data (in total 5776 eyes from patients with multiple sclerosis [1667 MSON eyes and 4109 MSNON eyes] and 1697 eyes from healthy controls) that met published OCT quality control criteria and were suitable for meta-analysis. Compared with control eyes, the peripapillary retinal nerve fibre layer (RNFL) showed thinning in MSON eyes (mean difference −20·10 μm, 95{\%} CI −22·76 to −17·44; p<0·0001) and in MSNON eyes (–7·41 μm, −8·98 to −5·83; p<0·0001). The macula showed RNFL thinning of −6·18 μm (–8·07 to −4·28; p<0·0001) in MSON eyes and −2·15 μm (–3·15 to −1·15; p<0·0001) in MSNON eyes compared with control eyes. Atrophy of the macular ganglion cell layer and inner plexiform layer (GCIPL) was −16·42 μm (–19·23 to −13·60; p<0·0001) for MSON eyes and −6·31 μm (–7·75 to −4·87; p<0·0001) for MSNON eyes compared with control eyes. A small degree of inner nuclear layer (INL) thickening occurred in MSON eyes compared with control eyes (0·77 μm, 0·25 to 1·28; p=0·003). We found no statistical difference in the thickness of the combined outer nuclear layer and outer plexiform layer when we compared MSNON or MSON eyes with control eyes, but we found a small degree of thickening of the combined layer when we compared MSON eyes with MSNON eyes (1·21 μm, 0·24 to 2·19; p=0·01). Interpretation The largest and most robust differences between the eyes of people with multiple sclerosis and control eyes were found in the peripapillary RNFL and macular GCIPL. Inflammatory disease activity might be captured by the INL. Because of the consistency, robustness, and large effect size, we recommend inclusion of the peripapillary RNFL and macular GCIPL for diagnosis, monitoring, and research. Funding None.",
author = "{ERN-EYE IMSVISUAL} and Axel Petzold and Balcer, {Laura J.} and Calabresi, {Peter A.} and Fiona Costello and Frohman, {Teresa C.} and Frohman, {Elliot M.} and Martinez-Lapiscina, {Elena H.} and Green, {Ari J.} and Randy Kardon and Olivier Outteryck and Friedemann Paul and Sven Schippling and Patrik Vermersch and Pablo Villoslada and Balk, {Lisanne J.} and Orhan Aktas and Philipp Albrecht and Jane Ashworth and Nasrin Asgari and Graeme Black and Daniel Boehringer and Raed Behbehani and Leslie Benson and Robert Bermel and Jacqueline Bernard and Alexander Brandt and Jodie Burton and Peter Calabresi and Jonathan Calkwood and Christian Cordano and Ardith Courtney and Andr{\'e}s Cruz-Herranz and Ricarda Diem and Avril Daly and Helene Dollfus and Christina Fasser and Carsten Finke and Jette Frederiksen and Elenaw Garcia-Martin and Su{\'a}rez, {In{\'e}s Gonz{\'a}lez} and Gorm Pihl-Jensen and Jennifer Graves and Joachim Havla and Bernhard Hemmer and Huang, {Su Chun} and Jaime Imitola and Hong Jiang and David Keegan and Eric Kildebeck and Alexander Klistorner",
year = "2017",
month = "10",
day = "1",
doi = "10.1016/S1474-4422(17)30278-8",
language = "English (US)",
volume = "16",
pages = "797--812",
journal = "The Lancet Neurology",
issn = "1474-4422",
publisher = "Lancet Publishing Group",
number = "10",

}

TY - JOUR

T1 - Retinal layer segmentation in multiple sclerosis

T2 - a systematic review and meta-analysis

AU - ERN-EYE IMSVISUAL

AU - Petzold, Axel

AU - Balcer, Laura J.

AU - Calabresi, Peter A.

AU - Costello, Fiona

AU - Frohman, Teresa C.

AU - Frohman, Elliot M.

AU - Martinez-Lapiscina, Elena H.

AU - Green, Ari J.

AU - Kardon, Randy

AU - Outteryck, Olivier

AU - Paul, Friedemann

AU - Schippling, Sven

AU - Vermersch, Patrik

AU - Villoslada, Pablo

AU - Balk, Lisanne J.

AU - Aktas, Orhan

AU - Albrecht, Philipp

AU - Ashworth, Jane

AU - Asgari, Nasrin

AU - Black, Graeme

AU - Boehringer, Daniel

AU - Behbehani, Raed

AU - Benson, Leslie

AU - Bermel, Robert

AU - Bernard, Jacqueline

AU - Brandt, Alexander

AU - Burton, Jodie

AU - Calabresi, Peter

AU - Calkwood, Jonathan

AU - Cordano, Christian

AU - Courtney, Ardith

AU - Cruz-Herranz, Andrés

AU - Diem, Ricarda

AU - Daly, Avril

AU - Dollfus, Helene

AU - Fasser, Christina

AU - Finke, Carsten

AU - Frederiksen, Jette

AU - Garcia-Martin, Elenaw

AU - Suárez, Inés González

AU - Pihl-Jensen, Gorm

AU - Graves, Jennifer

AU - Havla, Joachim

AU - Hemmer, Bernhard

AU - Huang, Su Chun

AU - Imitola, Jaime

AU - Jiang, Hong

AU - Keegan, David

AU - Kildebeck, Eric

AU - Klistorner, Alexander

PY - 2017/10/1

Y1 - 2017/10/1

N2 - Background Structural retinal imaging biomarkers are important for early recognition and monitoring of inflammation and neurodegeneration in multiple sclerosis. With the introduction of spectral domain optical coherence tomography (SD-OCT), supervised automated segmentation of individual retinal layers is possible. We aimed to investigate which retinal layers show atrophy associated with neurodegeneration in multiple sclerosis when measured with SD-OCT. Methods In this systematic review and meta-analysis, we searched for studies in which SD-OCT was used to look at the retina in people with multiple sclerosis with or without optic neuritis in PubMed, Web of Science, and Google Scholar between Nov 22, 1991, and April 19, 2016. Data were taken from cross-sectional cohorts and from one timepoint from longitudinal studies (at least 3 months after onset in studies of optic neuritis). We classified data on eyes into healthy controls, multiple-sclerosis-associated optic neuritis (MSON), and multiple sclerosis without optic neuritis (MSNON). We assessed thickness of the retinal layers and we rated individual layer segmentation performance by random effects meta-analysis for MSON eyes versus control eyes, MSNON eyes versus control eyes, and MSNON eyes versus MSON eyes. We excluded relevant sources of bias by funnel plots. Findings Of 25 497 records identified, 110 articles were eligible and 40 reported data (in total 5776 eyes from patients with multiple sclerosis [1667 MSON eyes and 4109 MSNON eyes] and 1697 eyes from healthy controls) that met published OCT quality control criteria and were suitable for meta-analysis. Compared with control eyes, the peripapillary retinal nerve fibre layer (RNFL) showed thinning in MSON eyes (mean difference −20·10 μm, 95% CI −22·76 to −17·44; p<0·0001) and in MSNON eyes (–7·41 μm, −8·98 to −5·83; p<0·0001). The macula showed RNFL thinning of −6·18 μm (–8·07 to −4·28; p<0·0001) in MSON eyes and −2·15 μm (–3·15 to −1·15; p<0·0001) in MSNON eyes compared with control eyes. Atrophy of the macular ganglion cell layer and inner plexiform layer (GCIPL) was −16·42 μm (–19·23 to −13·60; p<0·0001) for MSON eyes and −6·31 μm (–7·75 to −4·87; p<0·0001) for MSNON eyes compared with control eyes. A small degree of inner nuclear layer (INL) thickening occurred in MSON eyes compared with control eyes (0·77 μm, 0·25 to 1·28; p=0·003). We found no statistical difference in the thickness of the combined outer nuclear layer and outer plexiform layer when we compared MSNON or MSON eyes with control eyes, but we found a small degree of thickening of the combined layer when we compared MSON eyes with MSNON eyes (1·21 μm, 0·24 to 2·19; p=0·01). Interpretation The largest and most robust differences between the eyes of people with multiple sclerosis and control eyes were found in the peripapillary RNFL and macular GCIPL. Inflammatory disease activity might be captured by the INL. Because of the consistency, robustness, and large effect size, we recommend inclusion of the peripapillary RNFL and macular GCIPL for diagnosis, monitoring, and research. Funding None.

AB - Background Structural retinal imaging biomarkers are important for early recognition and monitoring of inflammation and neurodegeneration in multiple sclerosis. With the introduction of spectral domain optical coherence tomography (SD-OCT), supervised automated segmentation of individual retinal layers is possible. We aimed to investigate which retinal layers show atrophy associated with neurodegeneration in multiple sclerosis when measured with SD-OCT. Methods In this systematic review and meta-analysis, we searched for studies in which SD-OCT was used to look at the retina in people with multiple sclerosis with or without optic neuritis in PubMed, Web of Science, and Google Scholar between Nov 22, 1991, and April 19, 2016. Data were taken from cross-sectional cohorts and from one timepoint from longitudinal studies (at least 3 months after onset in studies of optic neuritis). We classified data on eyes into healthy controls, multiple-sclerosis-associated optic neuritis (MSON), and multiple sclerosis without optic neuritis (MSNON). We assessed thickness of the retinal layers and we rated individual layer segmentation performance by random effects meta-analysis for MSON eyes versus control eyes, MSNON eyes versus control eyes, and MSNON eyes versus MSON eyes. We excluded relevant sources of bias by funnel plots. Findings Of 25 497 records identified, 110 articles were eligible and 40 reported data (in total 5776 eyes from patients with multiple sclerosis [1667 MSON eyes and 4109 MSNON eyes] and 1697 eyes from healthy controls) that met published OCT quality control criteria and were suitable for meta-analysis. Compared with control eyes, the peripapillary retinal nerve fibre layer (RNFL) showed thinning in MSON eyes (mean difference −20·10 μm, 95% CI −22·76 to −17·44; p<0·0001) and in MSNON eyes (–7·41 μm, −8·98 to −5·83; p<0·0001). The macula showed RNFL thinning of −6·18 μm (–8·07 to −4·28; p<0·0001) in MSON eyes and −2·15 μm (–3·15 to −1·15; p<0·0001) in MSNON eyes compared with control eyes. Atrophy of the macular ganglion cell layer and inner plexiform layer (GCIPL) was −16·42 μm (–19·23 to −13·60; p<0·0001) for MSON eyes and −6·31 μm (–7·75 to −4·87; p<0·0001) for MSNON eyes compared with control eyes. A small degree of inner nuclear layer (INL) thickening occurred in MSON eyes compared with control eyes (0·77 μm, 0·25 to 1·28; p=0·003). We found no statistical difference in the thickness of the combined outer nuclear layer and outer plexiform layer when we compared MSNON or MSON eyes with control eyes, but we found a small degree of thickening of the combined layer when we compared MSON eyes with MSNON eyes (1·21 μm, 0·24 to 2·19; p=0·01). Interpretation The largest and most robust differences between the eyes of people with multiple sclerosis and control eyes were found in the peripapillary RNFL and macular GCIPL. Inflammatory disease activity might be captured by the INL. Because of the consistency, robustness, and large effect size, we recommend inclusion of the peripapillary RNFL and macular GCIPL for diagnosis, monitoring, and research. Funding None.

UR - http://www.scopus.com/inward/record.url?scp=85029184210&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85029184210&partnerID=8YFLogxK

U2 - 10.1016/S1474-4422(17)30278-8

DO - 10.1016/S1474-4422(17)30278-8

M3 - Article

VL - 16

SP - 797

EP - 812

JO - The Lancet Neurology

JF - The Lancet Neurology

SN - 1474-4422

IS - 10

ER -