Article Text

Download PDFPDF

Original article
Impairment of contrast visual acuity as a functional correlate of retinal nerve fibre layer thinning and total macular volume reduction in multiple sclerosis
  1. Markus Bock1,2,
  2. Alexander U Brandt1,3,
  3. Jörn Kuchenbecker4,
  4. Jan Dörr1,2,
  5. Caspar F Pfueller1,2,
  6. Nicholetta Weinges-Evers1,
  7. Gunnar Gaede1,
  8. Hanna Zimmermann1,
  9. Judith Bellmann-Strobl2,
  10. Stephanie Ohlraun1,
  11. Frauke Zipp2,5,
  12. Friedemann Paul1,2
  1. 1NeuroCure Clinical Research Center, Charité—Universitätsmedizin Berlin, Berlin, Germany
  2. 2Cecilie Vogt Clinic, Charité—Universitätsmedizin Berlin, Berlin, Germany
  3. 3gfnmediber GmbH, Berlin, Germany
  4. 4Department of Ophthalmology, HELIOS Klinikum Berlin-Buch, Berlin, Germany
  5. 5Department of Neurology, University Medicine Mainz, Mainz, Germany
  1. Correspondence to Dr Friedemann Paul, NeuroCure Clinical Research Center, Charité—Universitätsmedizin Berlin, Charitplatz 1, 10117 Berlin, Germany; friedemann.paul{at}


Objectives To analyse the association between retinal nerve fibre layer thickness (RNFLT) and total macular volume (TMV) as measured by optical coherence tomography, and contrast sensitivity (CS) measured by Functional Acuity Contrast Testing (FACT) in relapsing-remitting multiple sclerosis; and to investigate whether FACT testing by a contrast box device is feasible in multiple sclerosis (MS).

Methods fact was performed using the Optec 6500 P vision testing system with best correction under photopic and mesopic conditions without glare. The Area Under the Log Contrast Sensitivity Function (AUC) was calculated. RNFLT and TMV were assessed by Stratus optical coherence tomography. All participants underwent visual acuity testing (Snellen), spherical refractive error testing and cylindrical refractive error testing.

Results 85 relapsing-remitting multiple sclerosis patients (170 eyes) and 35 healthy controls (HC, 70 eyes) were measured. AUC Day and Night were lower in MS than in HC (p<0.001) when correcting for age, as were mean RNFLT and TMV (p<0.001 and p=0.018, respectively). Both RNFLT and TMV predicted contrast sensitivity in MS (AUC Day: standardised coefficient β=0.277, p<0.001, and β=0.262, p<0.001, respectively; AUC Night: β=0.202, p=0.009 and β=0.222, p=0.004, respectively, linear regressions). In HC, there was no correlation between RNFLT or TMV and contrast sensitivity.

Conclusion (1) Contrast sensitivity is reduced in MS versus HC; (2) RNFL and TMV as morphological measures of retinal axonal loss are predictors of contrast sensitivity as a functional visual parameter in MS but not in HC; and (3) FACT with the contrast box is a novel, feasible and rapid method to assess contrast sensitivity in MS.

  • Multiple sclerosis
  • vision
  • low contrast sensitivity
  • optical coherence tomography
  • optical quality
  • mesopic vision
  • day vision
  • night vision
  • retinal nerve fibre layer
  • retina
  • vision

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Visual dysfunction is the presenting symptom in 50% of multiple sclerosis patients (MS) and occurs in up to 80% of patients over the course of disease.1–3 Despite this high prevalence and the negative impact on quality of life,4 assessment of visual dysfunction in the majority of clinical trials was limited to Snellen visual acuity (VA) testing, which is insufficient to detect more subtle but relevant features of visual abnormalities such as contrast sensitivity (CS).5 6 Recent investigations suggested that contrast letter acuity and CS may be more sensitive measures of visual dysfunction in MS and optic neuritis (ON), showing abnormalities even in patients with normal Snellen acuity and visually evoked potentials.3 7 8 Thus, additional tools for improved detection of visual impairment in clinical trials are warranted.

Optical coherence tomography (OCT) is increasingly accepted as a tool for in vivo assessment of retinal axonal loss in isolated ON and various MS subtypes with or without ON.9 10 However, OCT's sensitivity in discriminating MS eyes from healthy eyes is still limited,11 and its appropriateness as an outcome measure in clinical trials on neuroprotective therapies remains to be proven. To date, only a few studies have investigated the association of morphological changes in retinal nerve fibre layer thickness (RNFLT) with CS as a functional parameter of visual dysfunction in MS.12 13 Against this background, we aimed to analyse the association between morphological parameters of axonal retinal loss (RNFLT and total macular volume (TMV)) from OCT, and functional parameters of impaired contrast acuity, using Functional Acuity Contrast Testing (FACT), in both MS patients and healthy controls (HC). FACT investigates a patient's ability to perceive linear sine-wave gratings in different spatial frequencies (SF), and is performed in a variety of light conditions intended to imitate those which the patient would encounter in everyday life.14 15 Since one difficulty of FACT in clinical trials is the incorporation of CS at all SF, we evaluated FACT by means of the Area Under the Log Contrast Sensitivity Function (AUC).16 17 The AUC elegantly combines CS at all SF into one single parameter by calculating the AUC between all measurements (figure 1). AUC has been widely applied and validated in several trials.18–20 It has been proven not only to be reliable in comparison with single CS measurements, but also to better discriminate between healthy subjects and patients.14 18 21 22 Finally, since AUC incorporates CS at all SF, it is more resistant to floor or ceiling effects than single CS. Thus, a further aim of our study was to investigate whether FACT is feasible in MS patients, by means of a contrast box not previously applied to this condition.

Figure 1

Calculation of the area under the curve (AUC): sample calculation of AUC as described in Methods. The two axes show the logarithmic expressions of the measured contrast sensitivity (y-axis) at each of the five spatial frequencies (x-axis). The five black dots represent the original five measurement points by functional acuity contrast testing. The AUC (grey area) is then calculated as the area under the curve of a third-degree polynomial function of the five measurement points that was calculated using a least-square curve fit algorithm.16 17 21 22 The AUC's low boundary equals the logarithmic expression of the lowest spatial frequency x0=log10(1.5), and the high boundary equals the logarithmic expression of highest spatial frequency x4=log10(18). The AUC describes the overall performance of a patient under either photopic or mesopic conditions with only a single expression. The higher the AUC number, the better is the patient's performance.


Study participants

Relapsing-remitting MS (RRMS) patients23 on a stable immunomodulatory therapy were consecutively recruited for this cross-sectional study from screening visits for ongoing clinical trials, which were approved by the local ethics committee and to which all participants gave informed written consent according to the Declaration of Helsinki. Patients met the following criteria: age ≥18 years, spherical refractive error between −6 and +4 dioptre, astigmatism <−3.5, and intraocular pressure ≥8 mm Hg and <22 mm Hg. Exclusion criteria comprised all ocular diseases including an acute episode of ON less than 3 months prior to screening. HC without history of neurological or ophthalmological diseases were consecutively recruited within the hospital.

Clinical and visual assessment

All participants underwent VA testing, spherical refractive error testing (REFs), cylindrical refractive error testing (REFc), OCT and FACT examination. VA testing outcomes in decimals were converted into Snellen equivalents using the standard protocol VA Conversion Chart. Neurological disability in patients was assessed by the Expanded Disability Status Scale.24 The examiners were aware of the participants' belonging to the MS or HC group but were blinded to further relevant clinical data in the MS group (eg, Expanded Disability Status Scale, history of relapses or ON).

Contrast sensitivity

FACT was measured by trained personnel using the Optec 6500 P system (Stereo Optical, Chicago, Illinois) with best correction under photopic (‘daylight’ with target luminance value of 85 cd/m2) and mesopic (‘night light’ with target luminance value of 3 cd/m2) conditions without glare. The linear sine-wave grating charts tested for five SF of 1.5, 3, 6, 12 and 18 cycles per degree (cpd). Each SF was tested with nine levels of contrast, with a decrease of 0.15 log steps per contrast level. The gratings were located in square patches arranged in two rows with five patches above and four patches below within the contrast box. CS was recorded as the lowest contrast level achieved by the patient for each SF. An SF was final, and the results recorded after a subject's first wrong reply. All assessments started using the lowest SF. Each eye was tested in monocular mode with nine contrast levels and five SF under photopic (day) and mesopic (night) conditions without glare, following the standards published by the American National Standards Institute.15 The total examination time was typically below 10 min for both eyes in sequential monocular testing.

Calculation of AUC of FACT

AUC was calculated as previously described by Applegate et al as the Area Under the Log Contrast Sensitivity Function (AULCSF) as a combined expression for all data points of each FACT session16 17 21 22 but modified for the Optec format. The following calculations were performed using the statistical software R First, the spatial frequencies (1.5, 3, 6, 12 and 18 cpd) and CS at each spatial frequency were converted into logarithmic (base 10) expressions. For each data set of measurements, a least-square curve fit was performed, using a third-order polynomial function with R's ‘nls()’ routine. nls() calculates the best-fitting curve to all measurement points of one given data set via least-square regression. AUC was then calculated as the area under this curve between the lowest logarithmic spatial frequency (x0=log10(1.5)) and the highest logarithmic spatial frequency (x4=log10(18)) (figure 1).

Optical coherence tomography

RNFLT and TMV were measured with Stratus 3000 OCT (Carl Zeiss Meditec, Dublin, California) using ‘Fast RNFL 3.4’ and ‘Fast Macula Thickness Map’ protocols (software V4.0) by trained personnel. For the first protocol, three 3.4 mm diameter circular scans were acquired circumferentially to the optic disc, and for the second protocol six radial lines were taken, centred within the fovea. A good-quality image was defined as having a generalised signal distribution, a reflectance signal from either RNFL or retinal pigment epithelium strong enough to identify either layer, no missing parts caused by eye movements and a signal strength of ≥8 of 10. Segmentation lines for upper and lower borders of RNFL were required to be on the internal limiting membrane and lower border of the RNFL. Images which did not meet these criteria were excluded. A-scan data were digitally exported in a blinded fashion; the mean RNFLT was calculated in μm, and TMV in mm3.

Statistical analysis

Differences between MS patients and HC were analysed using the Pearson χ2 test for gender, and analysis of variance (ANOVA) for age. Comparison of REFs, REFc and Snellen VA was performed using Generalised Estimating Equation (GEE) models to account for intereye dependencies.

Differences in AUC, RNFLT and TMV between MS and HC were calculated using GEE to account for intereye dependencies and using age as covariate. To assess the correlation between RNFLT or TMV and AUC Day or Night, (a) a Spearman ρ analysis was performed between AUC Day or Night and RNFLT or TMV, and (b) linear regressions (LR) were performed with RNFLT or TMV with VA and age as independent values and AUC Day or Night as dependent values. To calculate R2 changes attributable to RNFLT or TMV and VA and age, LR were performed with three consecutive models with (1) only RNFLT or TMV as an independent value, (2) RNFLT/TMV and VA as independent values and (3) RNFLT/TMV, VA and age as independent values. To further confirm results from LR and account for intereye/intrapatient dependencies, GEE were performed with RNFLT or TMV and VA and age as independent values and AUC Day or AUC Night as dependent values.

For the analyses of AUC for the prediction of RNFLT or TMV loss, GEEs with RNFLT or TMV as dependent and AUC Day or Night as independent and correcting for VA and age were performed. Predicted changes were calculated based on a change of 0.1 of AUC.

Age related changes in CS were analysed for non-linear effects using AUC/age scatter plots and LR with a residual analysis. AUC Day showed minor age effects not suggestive for non-linear effects, while AUC Night showed minor age effects with a possible non-linear compound (data not shown). The reader should be aware that minor non-linear effects in AUC Night might confound results from GEE.

For all calculations, statistical significance was established at p<0.05. All tests should be understood as constituting exploratory data analysis, as no previous power calculation and adjustments for multiple testing were made. Statistical analysis was performed with PASW 18 (SPSS, IBM Corporation, Somers, New York, USA).


We measured 170 MS and 70 HC eyes (table 1). Five right and one left MS eyes were excluded from further analysis because of acute unilateral ON. One HC left eye and one MS left eye were excluded due to amblyopia. Patients and HC differed in age (p=0.002, ANOVA) but not in gender (p=0.623, Pearson χ2); statistical tests therefore included age as a covariate whenever applicable. There were no differences between HC and MS patients in REFs (p=0.617, GEE), REFc (p=0.375, GEE) and high-contrast Snellen VA (p=0.205, GEE). However, since even small differences in high-contrast VA might influence low CS, high-contrast VA was used as a covariate.

Table 1

Demographics and main visual characteristics for healthy controls and multiple sclerosis patients

Contrast sensitivity and OCT

CS was measured at five different spatial frequencies for night and day settings and transformed into single comparable values of AUC Day and Night (see Methods). AUC Day and Night were lower in MS patients than in HC (table 2, figure 2A) when corrected for age differences.

Table 2

Overview of contrast box and optical coherence tomography measurements

Figure 2

Contrast sensitivity, retinal nerve fibre layer thickness (RNFLT) and total macular volume (TMV) in multiple sclerosis patients (MS) and healthy controls (HC). (A) Comparison of the area under the curve (AUC) day and night (left), RNFLT (middle) and TMV (right) between HC and MS patients. Whiskers represent the 5th and 95th percentiles, the lower box limit the 25th percentile, and the upper box limit the 75th percentile. Outliers are not plotted. The middle line represents the mean value. ***p<0.001; *p<0.01 (GEE correcting for intereye dependencies and age). (B) Scatter plots showing the correlation between RNFLT and AUC Day (left) and AUC Night (right) between the HC represented by black triangles and MS patients shown by grey squares. The line shows the coefficient from linear regression of MS patients' data. (C) Same as (B) but for the correlation between TMV and AUC Day and Night. AULCSF, Area Under the Log Contrast Sensitivity Function.

All OCT measurements fulfilled the minimum required signal strength criteria and were included in a further analysis. No errors in layer segmentation were observed. The mean RNFLT and TMV were reduced in MS eyes compared with HC (table 2, figure 2A).

Correlation between contrast sensitivity, RNFLT and TMV

RNFLT correlated with AUC Day (Spearman ρ=0.251, p=0.001) and AUC Night (ρ=0.185, p=0.017). TMV correlated with AUC Day (ρ=0.306, p<0.001) and AUC Night (ρ=0.226, p=0.003) in a similar fashion. In separate, multistep linear regressions, both RNFLT and TMV were predictors of CS in MS patients when correcting for VA and age (table 3, figure 2C). RNFLT, VA and age reached a combined R2=0.397 for AUC Day and combined R2=0.389 for AUC Night. For TMV, VA and age, the combined R2 for AUC Day was 0.357, and the combined R2 for AUC Night was 0.372. In all models, RNFLT or TMV and VA were significant predictors of AUC Day or Night, whereas age was not. R2 changes attributable to RNFLT or TMV and VA are given in table 3. To confirm these results in a model accounting for intereye/intrapatient dependencies, we performed GEEs with AUC Day or AUC Night as dependent variables and RNFLT or TMV with VA and age as independent variables (table 3). There was no correlation between RNFLT or TMV with CS in HC (not shown).

Table 3

Results of linear regression analyses (LR) and Generalised Estimating Equations (GEEs) (from multiple sclerosis patients only) with retinal nerve fibre layer thickness (RNFLT) (first column) or total macular volume (TMV) (second column) and visual acuity (VA) and age as independent variables, and area under the curve (AUC) day (first row) or AUC Night (second row) as dependent variables

Predictive value of AUC for RNFLT or TMV

To assess how changes in AUC may predict RNFLT loss, we performed GEEs with RNFLT or TMV as dependent values and AUC Day or Night as independent values while correcting for VA and age. A reduction of 0.1 in AUC Day was associated with an RNFLT loss of 1.852 μm (95% CI 0.830 to 2.874, p<0.001) and TMV loss of 0.044 mm3 (95% CI 0.014 to 0.075, p=0.004). A reduction of 0.1 in AUC Night was associated with a reduction of RNFLT of 0.846 μm (95% CI 0.189 to 1.502, p=0.012) and a reduction of TMV of 0.020 mm3 (95% CI 0.004 to 0.036, p=0.013).


We investigated the relationship between morphological (RNFLT/TMV) and functional parameters (CS) of contrast vision dysfunction in MS and compared these with HC. Key findings are (1) that photopic and mesopic CS is significantly reduced in MS versus HC; (2) that RNFLT and TMV as morphological measures of retinal axonal loss are predictors of CS as a functional visual parameter only in MS but not in HC; and (3) that contrast box FACT-based AUC testing is a novel, feasible and rapid method to assess CS in MS. Thus, our study is the first to investigate the impact of both RNFL and TMV loss on CS. This measure was recently shown to be clinically meaningful for the patient's daily life.6 CS was assessed using the Optec 6500 P vision testing system, which has shown a moderate to high retest reliability and has yielded reliable FACT under photopic and mesopic conditions in healthy individuals and ophthalmological conditions for screening purposes16 17 but has yet to be applied in neurological diseases. Using this technique, we were able to show significant reductions of contrast acuity in MS versus HC despite comparable visual parameters between the two groups (Snellen VA, refractive error), demonstrating that the method of CS testing performed here is capable of detecting MS-related visual dysfunction. Our results underscore the notion that visual impairment with relevance for the patient's daily life may exist in spite of otherwise normal VA.5 6 It is noteworthy that the findings of Regan and colleagues, who described alterations in sine-wave grating CS despite normal Snellen acuity in MS by a technique similar to that applied in the Optec 6500 P vision testing system almost 30 years ago,5 did not lead to a broad clinical application of CS testing, which may indicate that impairments of this aspect of vision are still under-recognised. Awareness of this clinically relevant measure seems to have increased recently. In a previous study, Fisher et al investigated the association of reduced RNFLT, but not of TMV, with changes in CS measured by Sloan and Pelli–Robson charts12 and reported mean RNFLT differences of 4 μm for every one-line change in low-contrast letter acuity and CS scores, and a significant but modest correlation between these parameters. These results are in line with our findings, and both may point to a pathophysiological link between morphological damage to retinal axons and subsequent impact on visual function. This is further substantiated by the fact that in our MS cohort, both RNFLT and TMV predict CS in a similar order of magnitude, as well as by our observation that a significant correlation between RNFLT and TMV measures on the one hand and CS on the other was detected only in MS patients but not in HC. However, despite statistical significance, these correlations are no more than modest, which suggests that the anterior visual pathway damage is not the only contributor to impaired contrast acuity visual function, but damage to the optic radiations and the occipital lobes may also play a role. Moreover, OCT may capture only some qualitative and quantitative alterations in the retina (eg, axonal and ganglion cell loss) but not all disease-related changes.26 Differences in baseline OCT scores between individuals and acquisition errors in the time-domain OCT or FACT scores could be further contributors to the only modest correlation between OCT measures and CS scores.

A methodological weakness of our study is that we did not compare our FACT measurements with those from Sloan and Pelli–Robson charts previously used for testing low-contrast letter acuity and CS.3 7 27 Although aimed at assessing similar visual functions, the results from FACT and Pelli–Robson and Sloan charts are probably not interchangeable, a problem which we are currently addressing in a separate trial. A potential advantage of FACT over Arden plates or Sloan and Pelli–Robson charts could be a better standardisation which might improve the comparability of results between centres,28–30 and the experimental setup of FACT that simulates both photopic and mesopic conditions. Adding mesopic test conditions, which have not been specifically investigated in MS, might extend the utility of CS testing because mesopic conditions may be closer to the patients real-world experience.31 A further advantage of FACT could be that this method tests the most relevant range of spatial information, while the sizes tested by Pelli–Robson charts are too large to be meaningful to most everyday viewing.32 Wall chart tests such as Pelli–Robson or Sloan charts are likely to be impacted by the rate by which the examiner uncovers the plates or by ambient lighting conditions, and results are influenced by a high misclassification rate, or insufficient spectrum of spatial frequencies.33–35 It therefore seems worthwhile to investigate the potential of FACT for CS analysis in MS in comparison with Sloan and Pelli–Robson charts,3 7 especially in the context of clinical trials and in relation to patient-reported aspects of vision.6 Future investigations will include a larger variety of patients with different levels of disease activity and with various disease courses, as the generalisability of our results is limited by the recruitment of RRMS patients from clinical trials which may have caused a meaningful selection bias. A further limitation of the current work is the fact that HC failed to match by age with patients, which required correction for age in our statistical analysis. However, as the mean age difference between groups was only 5 years, and both our analysis and previous work suggest that this small difference is unlikely to have a relevant impact on CS,18 we believe that this age difference does not substantially weaken our findings. One problem with FACT, however, is that the three-alternative forced-choice mode to present the grating charts raises the probability of guessing correctly up to 33.3%.

In sum, our study shows that functional contrast vision in MS is influenced by morphological changes in the anterior visual pathway, and that contrast vision testing with the Optec 6500 contrast box is capable of detecting differences from HC. Increased awareness among physicians with regard to contrast vision impairment and conceptual consideration of this aspect in clinical MS trials is therefore necessary.


We thank our study nurses A Els, M Glass, F Lipske and C Rudolph, for their support, and all our patients who contributed to our study.


View Abstract


  • MB and AUB contributed equally.

  • Funding This work was supported by the German Research Foundation (DFG Exc 257 to JD, SO, CFP and FP) and grant KF2286101FO9 from the German Ministry of Economics to NeuroCure Clinical Research Center and gfnmediber.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Ethics approval was provided by the Ethik Kommission der Charité.

  • Provenance and peer review Not commissioned; externally peer reviewed.