Article Text

This article has a correction. Please see:


One year changes in disability in multiple sclerosis: neurological examination compared with patient self report
  1. E L J Hoogervorst1,
  2. M J Eikelenboom1,
  3. B M J Uitdehaag1,2,
  4. C H Polman1
  1. 1Department of Neurology, VU Medical Centre, Amsterdam, Netherlands
  2. 2Department of Clinical Epidemiology and Biostatistics, VU Medical Centre
  1. Correspondence to:
 Dr E L J Hoogervorst, Department of Neurology, VU Medical centre, Academisch Ziekenhuis Vrije Universiteit, Postbox 7057, Amsterdam 1007 MB, Netherlands;

Statistics from

The most common neurological outcome measure for multiple sclerosis (MS) patients is the Expanded Disability Status Scale (EDSS). Originally, Kurtzke developed this disability status scale,1 later revised to a more refined classification system, known as the EDSS.2 The EDSS is based on a neurological examination of eight functional systems usually performed by a medical doctor. While problems of standardisation, resulting in suboptimal interrater reliability, marginal sensitivity to change and bias to locomotor function have been described3–6 the EDSS remains a useful tool for classifying disability in MS patients.

In a longstanding invalidating neurological disease like MS where cure is not yet possible, disability as perceived by the patient is an important measure and for this purpose the Guy’s Neurological Disability Scale (GNDS)7 was recently developed. The GNDS measures disability based on patient self report, embraces the whole range of disabilities that can be encountered in the course of MS, is patient orientated, user friendly, and not biased to any particular disability.7,8 The GNDS is a questionnaire, divided in 12 subcategories, directed to assess a patient’s disability in the previous month by patient interview and can be applied by any healthcare worker. It is therefore extremely practical and capable of incorporating patients’ views of their disability in a structured manner.8,9

In a previous study we showed good cross sectional correlations between EDSS and GNDS: these correlations were mainly attributable to the importance of spinal cord related neurological functions for both scales.9 That study also confirmed that the GNDS measures other dimensions of disability than the EDSS and incorporates patients’ perspectives, both of which can cause marked discrepancies between scores on the two scales. Such discrepancies between objective measurements and subjective complaints have been demonstrated in many studies. To give just one example, Sharrack and Hughes found that the assessment of a patient’s walking ability is very different between doctors and patients, and also very different from measured distances.10

The aim of this study was to prospectively characterise the relation between one year changes in neurologist rating of neurological examination of abnormalities as measured by the EDSS and changes in patient self report as measured by the GNDS. More specifically, we studied whether discrepancies between changes in EDSS and GNDS were attributable to differences in the change perceived by the patient and that measured by the physician or the result of symptoms covered by the GNDS but not included in the EDSS. In addition, we studied how significant changes in the EDSS were associated with changes in patient perceived disability.



Two hundred and fifty patients with clinically definite MS11 underwent longitudinal examinations of EDSS and GNDS at our outpatient clinic. Patients were diagnosed as having relapsing-remitting MS (n=126), secondary progressive MS (n=60), or primary progressive MS (n=64).12

Test procedures

Disability was assessed at baseline and after one year using EDSS and GNDS. Data from EDSS and GNDS were collected in the same visit, under standardised conditions by well trained physicians. The two measurements were performed in the same order, first the GNDS followed by the EDSS.

The EDSS is divided in 20 half steps ranging from 0 (normal) to 10 (death due to MS).2 Each subcategory of the GNDS was scored separately ranging from 0 (normal) to 5 (maximum help required). The GNDS score is the sum score of the 12 subcategories (range 0 to 60).7


Results were analysed in several ways. We studied cross sectional correlations between EDSS and GNDS sum score at baseline and follow up. Longitudinal correlations were studied between Δ EDSS and Δ GNDS sum score, Δ functional systems, and Δ GNDS subcategories, as well as between Δ functional systems and the corresponding Δ GNDS subcategories (Δ sphincter compared with Δ bladder and bowel function; Δ brain stem compared with Δ speech and swallowing; Δ pyramidal, sensory and cerebellar compared with Δ upper and lower limb function; Δ cerebral compared with Δ cognition and mood and Δ vision compared with Δ visual).9

As this analysis may be confounded by the fact that the EDSS is an ordinal scale, we also studied the total number of patients showing a clinically significant change on the EDSS and GNDS, as well as the mean change in GNDS for patients with a change on the EDSS. A clinically significant change of the EDSS was defined as a change of 1.0 point or more at EDSS levels <5.5 or 0.5 point or more at EDSS levels ⩾5.5.13 For the GNDS a change of three or more points in the sum score was considered as clinically significant.7,14 As the threshold of significance is not well defined we performed a sensitivity analysis in which we defined a significant change on the GNDS as a change of at least 2, 4, and 5 points, as well.


Correlations were calculated using the Spearman rank correlation coefficient (r). In this study we considered p values of less than 0.01 as significant and p values of less than 0.05 as a trend only. The strength of correlation was labelled as follows: correlations <0.40 as weak to marginal; 0.40 to 0.60 as moderate; 0.60 to 0.80 as good, and >0.80 as excellent. Agreement between clinically significant EDSS and GNDS changes was calculated using linear weighted Cohen’s κ.


Table 1 summarises patient characteristics and scores on EDSS and GNDS at baseline and follow up. Mean age at baseline was 44.3 years (SD 11.0); 38% of the patients were male and 62% were female. Average time interval from baseline to follow up measurement was 13.6 months (SD 1.9). Baseline EDSS scores ranged from 0 to 9.0 and median EDSS scores at baseline and follow up were the same (4.0). The median GNDS sum score at follow up (14.0) showed a deterioration of 1.0 point compared with baseline (13.0).

Table 1

GNDS sum score expressed as median (interquartile range)

Changes in GNDS occurred independent from baseline GNDS sum scores; both frequency and magnitude of the Δ GNDS were equally distributed along the full range of baseline GNDS sum scores, whereas changes in EDSS were more frequent and larger in the lower part of the scale (data not shown). This emphasises an important measurement scale characteristic of the GNDS.

Cross sectional correlations between EDSS and GNDS sum score were 0.69 (p<0.01) at baseline and 0.77 (p<0.01) at follow up. With respect to changes over time, only a weak to marginal correlation was found between Δ EDSS and Δ GNDS sum score (r=0.19; p<0.01).

Correlations between change in the EDSS functional systems and change in the corresponding GNDS subcategories were marginal to weak (Δ brain stem compared with Δ speech and swallowing, r=0.12, not significant; Δ pyramidal, sensory and cerebellar compared with Δ upper and lower limb function, r=0.16, p<0.05; Δ cerebral compared with cognition and mood, r=0.25, p<0.01 and Δ vision compared with visual r=0.23, p<0.01). A moderate correlation was found between Δ sphincter of the EDSS and Δ bladder and bowel function of the GNDS (r=0.58, p<0.01).

Table 2 shows that the only statistically significant correlation between change in the different GNDS subcategories and Δ EDSS was found for lower limb function, whereas (as expected) many more GNDS subcategories were correlated with Δ GNDS. Table 3 shows that of all functional systems only changes in brain stem, sphincter, and cerebral functional systems correlated significantly with changes in GNDS sum score, while with respect to the EDSS the situation is opposite, with the exception of the visual system, which is not correlated to either overall score.

Table 2

Correlations between Δ GNDS subcategories and Δ EDSS and Δ GNDS for the total population

Table 3

Correlations between Δ functional systems and Δ EDSS and Δ GNDS for the total population

Table 4 shows that a total of 59 patients showed a clinically significant worsening on the EDSS, whereas 37 patients showed a clinically significant improvement. A clinically significant worsening or improvement of the GNDS sum score (based on the definition of a change of at least three points being significant) was observed in 86 and 52 patients, respectively. Strikingly, opposite changes in EDSS compared with GNDS sum score were observed in 20 patients (8%): in 11 patients of 37 (29.7%) in whom there was a clinically significant improvement on the EDSS there was a 3 point or more worsening in the GNDS, whereas, vice versa, in 9 patients of 59 (15.2%) in whom there was a clinically significant worsening on the EDSS there was a 3 point or more improvement on the GNDS sum score. Poor agreement (linear weighted Cohen’s κ=0.12) was found between change in disability as measured by neurologist rating of neurological examination abnormalities and patient self report.

Table 4

Number of patients showing clinically significant changes in EDSS and GNDS sum score

Varying the definition of significant change in GNDS to at least 2, 4, or 5 points resulted in corresponding percentages of patients with a significant worsening on the EDSS accompanied by a significant improvement on the GNDS of 22.0% (13 of 59), 10.2% (6 of 59), and 10.2% (6 of 59) respectively. Agreement between the two disability measures remained poor with respective linear weighted Cohen’s κ values of 0.11, 0.13, and 0.16.

Figure 1 shows the mean GNDS change per category of EDSS change: the profile of changes clearly indicates that for patients with a significant improvement on the EDSS there is the largest improvement on the GNDS and vice versa that for patients with a significant worsening on the EDSS there is the largest worsening on the GNDS.

Figure 1

Mean change in GNDS sum score per (significant) EDSS change. 1, Significant EDSS improvement; 2, not significant EDSS improvement; 3, no EDSS change; 4, not significant EDSS worsening; 5, significant EDSS worsening.


Assessment of the patient’s perspective of disability is important, as reflected by the increasing number of studies incorporating such measures.15 To our knowledge, until now, no other large study has reported on the relation between change in neurologist rating of neurological examination abnormalities as measured by the EDSS and change in patient perceived disability.

Two important pieces of information can be obtained from the data collected in this study. In the first place we show that the correlation between one year longitudinal changes in the EDSS compared with GNDS is substantially lower than the cross sectional correlation between these measures; in the second place we show that, dependent on the criterion applied for significant change in the GNDS, 10% to 22% of the patients who have a significant worsening on the EDSS show a significant improvement in perceived disability.

It is obvious from our data that despite the profile of mean GNDS sum score changes being nicely rank ordered in relation to EDSS changes (significant or not according to commonly applied criteria, fig 1), the longitudinal correlation between EDSS and GNDS is much lower than the cross sectional correlations. Although this is partially attributable to the magnitude of one year changes being considerably smaller than the range of scores at a given point in time, detailed analyses of our data also gives rise to some other explanations. Correlations between changes in functional systems and changes in GNDS subcategories are only moderate to good for those functional systems that are (mainly) based on patient self report; those functional systems that are based on neurological examination are poorly or not at all correlated with the subcategory scores based on the patient’s perception. Table 3 shows that whereas changes in the EDSS are only correlated to changes in lower limb function, changes in the GNDS are significantly correlated to a wide range of disability subcategories, including categories that theoretically could have lead to EDSS changes (bladder function and bowel function) and categories that are not at all incorporated in the EDSS (that is, fatigue).

In conclusion there seem to be three reasons for the discrepancies between changes in EDSS and GNDS: (1) differences in the change perceived by the patient and that measured by the physician, (2) changes in many disability subcategories not leading to changes in EDSS, and (3) the GNDS incorporating aspects of disability (for example, fatigue) that are not at all considered by the EDSS. Of course it is well known from other studies that the EDSS is heavily biased to locomotor function,3,4 but this study adds a unique dimension based on how disability is perceived by MS patients.

Next, to compensate for the fact that our results might have been influenced by the changes in disability that occurred over a follow up period of about one year being comparatively small, we performed additional analyses on those patients who showed significant EDSS changes. For these analyses we used widely applied definitions for significant EDSS change.13 In many recent clinical trials in MS this definition of a significant change has been used to define significant worsening that served as primary outcome measure to determine whether disease modifying interventions were effective or not.16–18 In an attempt to obtain information on individual patients we also analysed how many patients with significant EDSS changes have corresponding or opposite changes in the GNDS of at least 2, 3, 4, or 5 points. Based on the data by Sharrack and Hughes7,9 we defined a change of at least three points to be primary for this purpose; clinically significant worsening in the physician derived EDSS was accompanied by a clinically significant increase in disability as perceived by the patient in only less than 50% of patients. Remarkably, 15% of these patients in whom a clinically significant worsening of EDSS was documented reported a significant improvement in their own subjective perception of disability compared with 30% who reported a significant worsening whereas a significant improvement in EDSS was found.

Performing these analyses for GNDS changes of at least 2, 4, and 5 points the percentages of patients showing a significant worsening on their EDSS with an associated improvement on the GNDS would have been 22%, 10%, and 10% respectively, indicating that this phenomenon is not dependent on the specific cut off of GNDS change.

In our opinion it is highly unlikely that our observations are attributable to measurement errors with respect to the EDSS: EDSS scores were obtained by a restricted number of physicians who had undergone multiple training sessions. In addition, the percentage of patients in this study showing significant worsening in EDSS score over one year (24%) is very similar to that in many other studies.7,19–21 Of course the patient’s perspective is subjective by nature and may change over time without corresponding objective changes. This may lead to response shift and thus complicate the comparison between changes in GNDS and EDSS.

In conclusion, we present longitudinal data supporting the idea that patient’s impression of change in disability may differ not just quantitatively but also qualitatively from that of an examining physician. This seems to be because there are true differences in the change perceived by the patient and that measured by the physician and to the fact that many more dimensions of disability have an impact on the GNDS than on the EDSS. Longer follow up of our cohort and additional studies in other patient groups are required to better understand this phenomenon and to determine what consequences it can have for future clinical trial design.


View Abstract

  • .


    Please note that the abtsract for this article was omitted.
    The abstract is shown here as 'On screen text' and PDF.

    The error is much regretted

    Files in this Data Supplement:


  • Competing interests: none declared.

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.

Linked Articles

  • Correction
    BMJ Publishing Group Ltd