Expanded Disability Status Scale

Expanded Disability Status Scale in Multiple Sclerosis Trials and Models

The Expanded Disability Status Scale, usually shortened to EDSS, is the neurologist's rating that multiple sclerosis (MS) trials have long used to record whether disability has worsened. For a health economist the EDSS matters twice over. Trial results for disease-modifying therapies are reported as delays in confirmed disability progression on the EDSS, and MS cost-effectiveness models in NICE technology appraisals such as TA767 have built their health states out of EDSS levels, with a utility and an annual cost attached to each level. In that appraisal of ponesimod the company's Markov model had 20 health states, 10 of them EDSS states for relapsing-remitting MS, 9 for secondary progressive MS and one for death, so the scale's step definitions became the skeleton of the QALY calculation.

Steps, functional systems and scoring rules

Kurtzke built the EDSS on his earlier Disability Status Scale (DSS), which graded people with MS 0 for normal and 10 for death due to MS in whole steps. The 1983 paper in Neurology kept those anchors and split each of the steps 1 to 9 into two half steps, so the expanded scale reads 0, 1.0, 1.5 and so on up to 9.5, then 10.

The neurologist first grades eight functional systems, which are pyramidal, cerebellar, brain stem, sensory, bowel and bladder, visual, cerebral and other functions. Kurtzke revised the sensory and the bowel and bladder systems in the same paper. The lower steps of the EDSS are set by the pattern of functional system grades, while walking ability takes over higher up the scale. In the standardised Neurostatus definitions developed by Kappos and colleagues and used in many trials, steps 0 to 1.5 apply only when walking is unrestricted, and steps of 5.5 and above are defined exclusively by the ability to walk, the assistance required or the use of a wheelchair.

The Neurostatus scoring material was revised to clarify definitions and add a separate ambulation score while remaining compatible with earlier versions. A trial report or model that states which version raters used makes comparisons across studies easier, because small differences in how a step is assigned move people across the boundaries that later become model states.

Rater agreement and the size of a meaningful change

The EDSS is a clinician-reported outcome, so its precision depends on the examining neurologist. Studies in Canada and London measured how far raters agree and how well the scale detects change.

Noseworthy and colleagues had two physicians, blinded to each other and to earlier scores, examine 168 patients in a Canadian clinical trial, giving 545 paired examinations. Exact agreement ranged between 48 per cent for the cerebellar functional system and 69 per cent for the EDSS and the pyramidal system, and once chance agreement was removed the kappa statistic values lay between 0.31 and 0.62. Agreement within one step reached at least 92 per cent except for the cerebellar and sensory systems. The authors recommended that a change of at least two steps, 1.0 point on the EDSS, is needed to be confident that disability has genuinely changed.

Sharrack and colleagues at Guy's Hospital reported in 1999 that inter-rater and intra-rater reliability of the EDSS was high when a difference of up to 1.0 point was counted as agreement. They found the score distribution bimodal and judged the EDSS unresponsive to clinical change, although valid as a measure of impairment and disability. Hobart, Freeman and Thompson concluded in 2000 that the lack of psychometric input into the development of the EDSS and functional systems had limited their usefulness as evaluative outcome measures.

Confirmed disability progression as a trial endpoint

Because a single examination can be off by a step, trials require a rise in EDSS to persist before it counts. The resulting endpoint, confirmed disability progression (CDP), is analysed as time-to-event data and summarised as a hazard ratio.

The protocol for ORATORIO, the trial of ocrelizumab in primary progressive MS, treated progression as confirmed when the higher EDSS was seen again at a regularly scheduled visit at least 12 weeks after it was first recorded. The size of the rise depends on where the patient starts. The same protocol defined progression as an increase of at least 1.0 point from a baseline EDSS of 5.5 or less, or at least 0.5 point from a baseline above 5.5.

Montalban and colleagues reported 12-week CDP in 32.9 per cent of patients on ocrelizumab and 39.3 per cent on placebo in ORATORIO, a hazard ratio of 0.76 (95 per cent confidence interval 0.59 to 0.98). For progression confirmed over 24 weeks the figures were 29.6 per cent and 35.7 per cent, a hazard ratio of 0.75. In the paired OPERA trials in relapsing MS, which enrolled people with an EDSS of 0 to 5.5 at screening, Hauser and colleagues reported both 12-week and 24-week confirmed endpoints. Economic models then choose which confirmation window feeds the treatment effect. In TA767 the treatment effects from the company's network meta-analysis adjusted progression through each EDSS state using 6-month confirmed disability accumulation.

EDSS states in multiple sclerosis cost-effectiveness models

The ponesimod model in TA767 was a cohort Markov model in which each EDSS level was a health state. In such a model, each cycle the cohort moves between states according to transition probabilities estimated from untreated natural history data, the treatment effect slows those movements, and relapses are layered on top.

Natural history cohorts behind the transition matrices

Untreated progression has come from long-running Canadian cohorts. Weinshenker and colleagues followed 1,099 consecutive patients in London, Ontario between 1972 and 1984, grading disability on the original Kurtzke DSS, which has whole steps only. The British Columbia Multiple Sclerosis database supplied the natural history in TA767, and the UK Multiple Sclerosis Risk Sharing Scheme compared its treated patients with eligible British Columbia patients whose data were collected in 1980 to 1996.

In TA1126, on natalizumab for highly active relapsing-remitting MS, NICE noted that previous technology appraisals in relapsing-remitting MS had used the British Columbia or the London Ontario natural history models. The external assessment group in that appraisal instead used real-world data from the UK MS Register, calculating event rates for an EDSS increase confirmed over six months and for relapse, and applying relative effects from the network meta-analysis.

Utilities and costs attached to each EDSS level

Each EDSS state carries a health state utility value, and here the steepness of the gradient drives much of the QALY gain. Orme and colleagues sent a postal questionnaire to 12,968 people on a UK MS Trust database and used the EQ-5D to estimate the disutility of progression and of an acute relapse. Hawton and Green reported in 2016, from a UK longitudinal cohort, mean EQ-5D values of 0.846 at EDSS 0 and 0.025 at EDSS 8, against SF-6D means of 0.702 and 0.529 at the same two levels. A relapse in the previous six months lowered mean utility by 0.076 on the EQ-5D and 0.052 on the SF-6D.

The choice of instrument therefore changes the QALY value of slowing progression, since the EQ-5D gradient across EDSS levels is far steeper than the SF-6D gradient. Hawton and Green also found EQ-5D values slightly higher at EDSS 4 than at EDSS 3, the same inconsistency Orme and colleagues had reported. Used directly, such values would give the more disabled state the higher utility, so the analysis has to decide whether to keep, smooth or pool them.

Committee concerns about EDSS-based models

NICE appraisal committees, in TA767 and TA656, have raised points that bear directly on EDSS-based structures. In TA767 the committee concluded that further sensitivity analysis was needed to explore unlikely numbers of people in high EDSS health states, and noted that fatigue was not modelled explicitly but captured only through health-related quality of life measured by EDSS score. In TA656 on siponimod, clinical experts explained that somebody usually diagnosed with secondary progressive MS at EDSS 6 might be diagnosed at EDSS 4 once the drug was available, a shift that matters for any model in which the progressive phase begins at a fixed EDSS level.

Illustrative one-cycle calculation with two EDSS bands

All figures here are illustrative and are not taken from any appraisal. A cohort of 1,000 people starts in an EDSS 3 band with an annual probability of 0.10 of moving to an EDSS 4 to 5 band. A treatment cuts that probability by a relative risk of 0.70. The EDSS 3 band carries an EQ-5D-based utility of 0.70 and an annual cost of GBP 5,000, and the higher band an EQ-5D-based utility of 0.55 and a cost of GBP 9,000, before drug costs.

1. Treated transition probability. The treated probability is 0.10 × 0.70 = 0.07, so 1,000 × 0.07 = 70 people progress with treatment, against 1,000 × 0.10 = 100 without.

2. QALYs in the cycle, counting state membership at the end of the cycle.

$$Q = n_3 u_3 + n_4 u_4$$

where $Q$ is the total QALYs for the cohort in the cycle, $n_3$ and $n_4$ are the numbers in the lower and higher bands, and $u_3$ and $u_4$ are the band utilities. Untreated, 900 × 0.70 = 630 and 100 × 0.55 = 55, giving 685 QALYs. Treated, 930 × 0.70 = 651 and 70 × 0.55 = 38.5, giving 689.5 QALYs. The gain is 689.5 − 685 = 4.5 QALYs per 1,000 people.

3. Disease management costs. Untreated, 900 × 5,000 = 4,500,000 and 100 × 9,000 = 900,000, a total of GBP 5,400,000. Treated, 930 × 5,000 = 4,650,000 and 70 × 9,000 = 630,000, a total of GBP 5,280,000. The saving is 5,400,000 − 5,280,000 = 120,000, or GBP 120 per person.

One avoided progression is worth 0.70 − 0.55 = 0.15 QALYs in a year under these utilities, and the 30 avoided progressions give 30 × 0.15 = 4.5 QALYs. If the higher band's EQ-5D utility were replaced by a flatter SF-6D gradient, the gain would shrink even though the clinical effect is unchanged, which is why the utility source sits alongside the hazard ratio in a cost-utility analysis of an MS therapy. Trials report the treatment effect as a hazard ratio, which applies to the rate of progression, so a full model would convert the annual probability to a rate, apply the hazard ratio and convert back, where this example simply multiplies the probability.

How the EDSS differs from neighbouring measures

The EDSS rates impairment and walking ability by neurological examination, whereas a patient-reported measure records what the person says about their own health. Like any measure with ordinal scaling, its steps rank people without equal spacing, and the meaning of a step changes along the scale, with functional systems governing the low range and walking distance and aids governing the upper range. The modified Rankin Scale is the comparable clinician-rated grading of disability in stroke, where the disability follows an acute event rather than accumulating over years of MS. The DSS used in the London, Ontario cohort is the whole-step predecessor of the EDSS, so transition matrices built on it need care before half-step trial data are applied.

Limitations and common misreadings

Several errors recur when EDSS data move into economic evaluation. Most come from treating a coarse clinical rating as if it were a precise interval measure.

• Treating EDSS points as equal intervals. A one-point change near the top of the scale reflects walking aids or wheelchair use, while the same change at the bottom reflects functional system grades, so averaging EDSS points across patients has no clear meaning.
• Ignoring rater error. Noseworthy and colleagues found exact agreement in only 69 per cent of paired EDSS ratings, so an unconfirmed one-step change in a single patient may be noise.
• Mixing confirmation windows. A hazard ratio for 12-week CDP and one for 24-week or 6-month CDP are not interchangeable inputs to the same transition matrix.
• Taking old cohorts as current natural history. The London, Ontario patients were followed in 1972 to 1984 and the British Columbia comparison patients in the Risk Sharing Scheme were recorded in 1980 to 1996, and the committee in TA1126 accepted UK MS Register data instead.
• Expecting responsiveness. Sharrack and colleagues judged the EDSS unresponsive to clinical change, so short trials can miss real improvement or deterioration.

Sources

• Hauser SL, Bar-Or A, Comi G, et al. Ocrelizumab versus interferon beta-1a in relapsing multiple sclerosis. New England Journal of Medicine. 2017;376(3):221-234. doi:10.1056/NEJMoa1601277
• Hawton A, Green C. Health utilities for multiple sclerosis. Value in Health. 2016;19(4):460-468. https://pubmed.ncbi.nlm.nih.gov/27325338/
• Hobart J, Freeman J, Thompson A. Kurtzke scales revisited: the application of psychometric methods to clinical intuition. Brain. 2000;123(5):1027-1040. doi:10.1093/brain/123.5.1027
• Kurtzke JF. Rating neurologic impairment in multiple sclerosis: an expanded disability status scale (EDSS). Neurology. 1983;33(11):1444-1452. doi:10.1212/WNL.33.11.1444
• Montalban X, Hauser SL, Kappos L, et al. Ocrelizumab versus placebo in primary progressive multiple sclerosis. New England Journal of Medicine. 2017;376(3):209-220. doi:10.1056/NEJMoa1606468
• National Institute for Health and Care Excellence. Ponesimod for treating relapsing-remitting multiple sclerosis (TA767). 2022. https://www.nice.org.uk/guidance/ta767/chapter/3-Committee-discussion
• National Institute for Health and Care Excellence. Siponimod for treating secondary progressive multiple sclerosis (TA656). Published 18 November 2020. https://www.nice.org.uk/guidance/ta656/chapter/3-Committee-discussion
• National Institute for Health and Care Excellence. Natalizumab (originator and biosimilar) for treating highly active relapsing-remitting multiple sclerosis after disease-modifying therapy (TA1126). Published 28 January 2026. https://www.nice.org.uk/guidance/ta1126
• Neurostatus. Definitions for a standardised, quantified neurological examination and assessment of Kurtzke's Functional Systems and Expanded Disability Status Scale in multiple sclerosis (Kappos L). Version 04/10.2. https://www.neurostatus.net/media/specimen/Definitions_0410-2_s.pdf
• Noseworthy JH, Vandervoort MK, Wong CJ, Ebers GC. Interrater variability with the Expanded Disability Status Scale (EDSS) and Functional Systems (FS) in a multiple sclerosis clinical trial. Neurology. 1990;40(6):971-975. doi:10.1212/WNL.40.6.971
• Orme M, Kerrigan J, Tyas D, Russell N, Nixon R. The effect of disease, functional status, and relapses on the utility of people with multiple sclerosis in the UK. Value in Health. 2007;10(1):54-60. doi:10.1111/j.1524-4733.2006.00144.x
• Palace J, Duddy M, Bregenzer T, et al. Effectiveness and cost-effectiveness of interferon beta and glatiramer acetate in the UK Multiple Sclerosis Risk Sharing Scheme at 6 years: a clinical cohort study with natural history comparator. Lancet Neurology. 2015;14(5):497-505. doi:10.1016/S1474-4422(15)00018-6
• F. Hoffmann-La Roche. A study of ocrelizumab in participants with primary progressive multiple sclerosis (ORATORIO), protocol, NCT01194570. https://cdn.clinicaltrials.gov/large-docs/70/NCT01194570/Prot_000.pdf
• Sharrack B, Hughes RA, Soudain S, Dunn G. The psychometric properties of clinical rating scales used in multiple sclerosis. Brain. 1999;122(1):141-159. doi:10.1093/brain/122.1.141
• Weinshenker BG, Bass B, Rice GP, et al. The natural history of multiple sclerosis: a geographically based study. I. Clinical course and disability. Brain. 1989;112(1):133-146. doi:10.1093/brain/112.1.133