Article Text

Download PDFPDF

Original research
Clinically stable disease is associated with a lower risk of both income loss and disability pension for patients with multiple sclerosis
  1. Thor Ameri Chalmer1,2,
  2. Mathias Buron1,2,
  3. Zsolt Illes3,4,
  4. Viktoria Papp3,
  5. Asta Theodorsdottir3,
  6. Jakob Schäfer5,
  7. Victoria Hansen5,
  8. Nasrin Asgari6,7,
  9. Pernille Bro Skejø8,
  10. Henrik Boye Jensen9,10,
  11. Per Soelberg Sørensen1,2,
  12. Melinda Magyari1,2
  1. 1Danish Multiple Sclerosis Center, Department of Neurology, University of Copenhagen, Rigshospitalet Glostrup, Copenhagen, Denmark
  2. 2Danish Multiple Sclerosis Registry, Department of Neurology, Rigshospitalet Glostrup, Copenhagen, Denmark
  3. 3Department of Neurology, Odense University Hospital, Odense, Denmark
  4. 4Institute of Clinical Research, University of Southern Denmark, Odense, Denmark
  5. 5Department of Neurology, Aalborg University Hospital, Aalborg, Denmark
  6. 6Department of Neurology, Slagelse Hospital, Slagelse, Denmark
  7. 7Department of Neurobiology, Institute of Molecular Medicine, University of Southern Denmark, Odense, Denmark
  8. 8Department of Radiology, Slagelse Hospital, Slagelse, Denmark
  9. 9Department of Brain and Nerve Diseases, Lillebealt Hospital, Kolding, Denmark
  10. 10Institute of Regional Health Research, University of Southern Denmark, Odense, Denmark
  1. Correspondence to Dr Thor Ameri Chalmer, Danish Multiple Sclerosis Center, Department of Neurology, University of Copenhagen, Rigshospitalet Glostrup, Copenhagen 2100, Denmark; thor.ameri.chalmer.01{at}


Objective To assess the risk of losing income from salaries and risk disability pension for multiple sclerosis patients with a clinically stable disease course 3 years after the start of disease-modifying therapy (DMT).

Methods Data from the Danish Multiple Sclerosis Registry were linked to other Danish nationwide population-based databases. We included patients who started treatment with a DMT between 2001 and 2014. Patients were categorised into a clinically stable group (No Evidence of Disease Activity (NEDA-2)) and a clinically active group (relapse activity or 6-month confirmed Expanded Disability Status Scale worsening). Outcomes were: (1) loss of regular income from salaries and (2) a transfer payment labelled as disability pension. We used a Cox proportional hazards model to estimate confounder-adjusted HRs, and absolute risks were plotted using cumulative incidence curves accounting for competing risks.

Results We included 2406 patients for the income analyses and 3123 patients for the disability pension analysis. Median follow-up from index date was ~5 years in both analyses. The NEDA-2 group had a 26% reduced rate of losing income (HR 0.74; 95% CI 0.60 to 0.92). HRs were calculated for 5-year intervals in the disability pension analysis: year 0–5: a 57% reduced rate of disability pension for the NEDA-2 group (HR 0.43; 95% CI 0.33 to 0.55) and year 5–10: a 36% reduced rate (HR 0.64; 95% CI 0.40 to 1.01).

Conclusion Clinically stable disease course (NEDA-2) is associated with a reduced risk of losing income from salaries and a reduced risk of disability pension.

View Full Text

Statistics from


Multiple sclerosis (MS) is a neurological disease affecting the central nervous system that may lead to physical disability, cognitive impairment and lower socioeconomic status.1–3 Patients with MS have a higher risk of disability pension and a lower average income compared with the background population.2 In addition, MS does influence the life of patients and their families and has pronounced economic consequences for society.4 Several disease-modifying therapies (DMTs) are available for the treatment of relapsing-remitting MS (RRMS)5 and recently also for a subset of patients with primary progressive MS6 and secondary progressive MS.7 8 The short-term effect of DMTs on disability worsening, typically measured as changes in the Expanded Disability Status Scale (EDSS), was studied in placebo-controlled randomised controlled trials (RCTs).6 7 9–15 The long-term effect of DMT exposure is more difficult to assess. Most observational studies and RCT extension studies agree that both treatment in general and in particular treatment initiated in the early stage of the disease are associated with a better prognosis.16–19 Moreover, the current treatment guidelines encourage therapy escalation to a highly effective DMT in patients with breakthrough disease.20 The new classification of MS disease courses categorise RRMS patients into patients with active disease (breakthrough disease) and stable disease based on the occurrence of clinical relapses and MRI activity (gadolinium-enhancing lesions or new or enlarging T2 lesions).21 Although disability worsening in RRMS is primarily a result of incomplete remission of relapses, a part of the EDSS worsening does not seem to be associated with clinical relapse activity.22 Whether this non-relapse-related worsening is a result of ‘sub-clinical’ inflammatory activity that may be seen in MRI or the worsening represents a neurodegenerative process independent of inflammation is still not fully known. An increasingly proactive treatment strategy has led to the concept No Evidence of Disease Activity (NEDA).23 NEDA is characterised by the number of parameters included. NEDA-3 refers to a disease course with no relapses, no EDSS worsening and no MRI activity, while NEDA-2 refers to clinically stable disease (or clinical NEDA).24 NEDA should always be assessed during a defined period—usually 1 year. The prognostic value of NEDA on EDSS worsening has previously been investigated.24–26 However, the association between NEDA-2 among patients exposed to DMT and the subsequent risk of losing the ability to maintain a job has not previously been studied. The aim of this study was to estimate the risk of losing income from salaries and risk of disability pension for patients with NEDA-2 compared with patients with a clinically active disease 3 years after starting their first DMT.


Data sources and study population

Data were retrieved from the Danish Multiple Sclerosis Registry (DMSR), a nationwide population-based registry in Denmark, a country with a population of 5.7 million. DMSR has been collecting data at diagnosis on all patients with MS since 1948. From 1996, MS physicians from all Danish MS clinics started collecting data prospectively on all patients treated with DMT.27 Data entry for the authorities are mandatory ensuring high completeness. We retrieved information about dates for starting and stopping DMT, relapse dates, EDSS records and MRI data. Data from DMSR were linked by the personal identification number (allocated to all Danish citizens or persons living in Denmark for more than 3 months)28 to data from registries hosted by Statistics Denmark29: the income statistics register,30 the Danish Rational Economic Agents Model (DREAM) database,31 the Danish National Patients Register32 and the population statistics register.28 Data were pseudoanonymised by Statistics Denmark and were accessed on their servers.

We included patients diagnosed with RRMS who started treatment with DMT after 1 January 2001. At treatment start, we required a disease duration of less or equal to 10 years, age between 18 and 55 years, EDSS less or equal to 4 and treatment with DMTs for RRMS (excluding ‘Fampridine’ ‘Hematopoietic Stem Cell Transplantation’, ‘Azathioprine’ ‘Methotrexate’, ‘Mitoxantrone’, ‘Treosulfan’, ‘Ofatumumab’, ‘Rituximab’, ‘Intravenous Immunoglobulins’ or ‘Study Drugs’). The 3 years (±1 year) following treatment start were labelled ‘definition period’ and defined the study groups as patients were divided into patients with clinically stable disease (NEDA-2) and patients with a clinically active disease course. An active disease course was defined as the occurrence of a relapse and/or a 6-month confirmed EDSS worsening. During the definition period, we required that patients had received DMTs for at least 75% of the time and at least three EDSS records. The index date was defined as the date of the EDSS record closest to 3 years after treatment start (±1 year). Accordingly, the last date of the definition period was also the first date of the follow-up period ensuring no overlap between the two periods (figure 1). Relapse dates registered within 30 days of a previous relapse date were considered as one relapse. EDSS records registered 30 days or less after a registered relapse date were not taken into consideration. EDSS worsening was defined as a 1.5-step increase if EDSS at treatment start was 0, and as a 1.0-step increase if EDSS at treatment start was between 1.0 and 4.5. The EDSS increase had to be confirmed by a subsequent EDSS record separated by at least 6 months. Patients who received disability pension before the index date were excluded for the disability pension analysis, and patients without income from salaries (from now on referred to as income) at the index date were excluded for the income analysis.

Study endpoints

The first outcome was loss of regular income from salaries defined as the loss of income for a full calendar year. Data on income were obtained from the income statistics register. The register contains detailed information on persons who have submitted a tax return to the Tax Administration in Denmark. The variable used for our outcome is updated annually as the accumulated income for each year. Income includes taxable income including perks, severance pay, fees for being a member of a board of directors, tax-free salaries and value of stock options. All these income categories relate to job maintenance or immediately after leaving a job (severance pay). The variable is the best proxy available for income from salaries.

The second outcome was the risk of disability pension defined as the first transfer payment labelled as disability pension. Data on disability pension were obtained from the DREAM database, which is an administrative database that contains weekly updates on transfer payments for all Danish citizens who have received a social transfer payment. Data have been collected since 1991, and all transfer payments are labelled (eg, sick leave, educational support and disability pension). The frequent updates allow precise monitoring of changes in transfer payments during follow-up.

Statistical analysis

Clinical and demographic characteristics are given as median (IQR), mean (SD) or percentage as appropriate. The risk of losing income and the risk of receiving disability pension were plotted as cumulative incidence curves accounting for competing risks (emigration, state pension and death).33 We followed patients until they reached either an outcome, a competing risk or study end (censoring). Study end was 1 January 2017, for the income analysis and 25 June 2018, for the disability pension analysis. The difference in study termination was due to data availability. We calculated the annual mean income as 2015-indexed euros and plotted two graphs reflecting changes in income during follow-up: (1) the annual mean income (patients with no income at index date were excluded) and (2) the annual mean income with censoring in the occurrence of a year without income (patients with no income at index date were excluded).

We estimated HRs using Cox proportional hazards models. We adjusted the estimates for the following confounders: age, sex, EDSS, calendar year, disease duration and socioeconomic status (all at treatment start) and relapse rate 2 years before treatment start. Patients were followed from index date and until the occurrence of an outcome or a censoring event (state pension, emigration, death or study end). Proportional hazards were tested using cumulative martingales residuals.34 We categorised numeric variables if the linearity assumption was not met.

We tested for any sex differences in rates. Finally, we conducted a subanalysis of patients with available data on MRI activity. In this subanalysis, we compared NEDA-3 with patients with active disease.

Sensitivity analyses

We defined three sensitivity analyses and applied them to both outcomes. First, we split the composite definition of clinically active disease into two variables: a categorised ‘relapse count’ and the presence of a 6-month EDSS worsening. Second, we redefined EDSS worsening as a numeric variable based on the following calculation:

Embedded Image

Lastly, we merged the two outcomes into one composite outcome. Accordingly, we analysed the time from index date to either disability pension, loss of income or censoring, whichever occurred first.

Ethics, approvals and data access

Data were analysed on the servers of Statistics Denmark. Accordingly, we cannot share data publicly as they contain sensitive, identifying participant information. These data are, however, available on request by researchers meeting the criteria for access to confidential information.

All analyses were conducted using SAS software, V.9.4.


The study population

A total of 3123 patients were included in the analysis of disability pension and 2406 patients in the income analysis (figure 2). Clinical and demographic data for patients in the disability analysis are presented in table 1 and data for patients in the income analysis in online supplementary table 1. The proportion of females was slightly higher and the mean age 1.5 years lower in the active group. The median duration of the definition period was 3 years (range 2–4 years) for both groups. More patients in the active group escalated treatment during both the definition period and during follow-up compared with the NEDA-2 group. The distribution of EDSS scores at treatment start was similar, while active patients had a higher level of EDSS 3 years after treatment start (at index date). We found no differences in socioeconomic status or comorbidity at treatment start. The The international Statistical Classification of Diseases 10 (ICD10) diagnoses defining comorbidities are presented in online supplementary table 2.

Figure 2

Patient disposition chart. DMT, disease-modifying therapy; EDSS, Expanded Disability Status Scale; RRMS, relapsing-remitting multiple sclerosis.

Table 1

Demographic and clinical information

We excluded 603 patients as they received disability pension prior index date. Interestingly, most of these patients had an active disease course during the definition period (active: 71.3%; NEDA-2: 28.7%).

Income from salaries

The median follow-up in the income analysis was 4.9 years (NEDA-2 group: 4.7 years; active group: 5.2 years) and 20.6% reached the outcome (NEDA-2 group: 17.3%; active group: 23.8%). The cumulative incidence curve indicated a slightly higher risk of losing income for the active group (figure 3A). After 10 years, the model estimated that 27.7% (95% CI 23.8% to 31.7%) of NEDA-2 patients and 34.0% (95% CI 30.2% to 37.7%) of active patients had once lost their income for a year. We calculated the mean income for each year after index date and plotted the means stratified by study groups. Figure 4A shows total income regardless of events after the index date (eg, disability pension), and it indicates that the mean income dropped during follow-up in both groups. In figure 4B, we censored patients if they encountered a year without income. Consequently, the plot reflects annual mean income for patients who still have income. In this plot, the mean income increased during follow-up.

Figure 3

Cumulative incidence curves of losing income from salaries (A) and disability pension (B). NEDA, No Evidence of Disease Activity.

Figure 4

Annual mean income from salaries during follow-up (A). Annual mean income from salaries with censoring in the occurrence of a year without income (B).

The Cox proportional hazards model showed proportional hazards for the study group variable. Sex did not have proportional hazards and was instead included in the unspecified baseline function. Age at treatment start (numeric variable) did not show linearity and was categorised into six categories.

We found that the NEDA-2 group had a 25% reduced rate of losing income compared with the group of active patients (HR 0.75; 95% CI 0.62 to 0.90; p=0.002) (table 2). The output from the regression analysis is presented in table 3. We used ‘Employee, low skilled’ as the reference category for socioeconomic status at treatment start and found that the ‘Employee, moderate-high skilled’ group had a markedly lower hazard rate of losing income. A high EDSS score at treatment start was also strongly associated with subsequent loss of income. All variables included in the regression analysis were statistically significant associated with the outcome except relapse count before treatment start. We tested the interaction between sex and the study groups variable using a likelihood ratio test and found no evidence of sex-specific effects on the risk of losing income (p=0.23).

Table 2

Primary results

Table 3

Output from the regression analysis (income)

Lastly, we conducted a subanalysis including patients with available MRI data. Three hundred and forty-seven patients were available for the analysis. Due to the limited study size and few events, we adjusted for sex and age only. The analysis had insufficient power to detect a difference in hazard rates; the HR for NEDA-3 was 0.87 (HR 0.87; 95% CI 0.36 to 2.11; p=0.75).

Disability pension

The median follow-up in the disability pension analysis was 4.9 years (NEDA-2: 4.7 years; active group: 5.1 years). The proportion of patients with a registered outcome was 12.8% (NEDA-2 8.3%; active group: 17.0%). The cumulative incidence curve indicates a clear difference in the risk of disability pension (figure 3B). After 10 years of follow-up, the estimated proportion of patients reaching the outcome was 13.6% (95% CI 11.1% to 16.4%) for the NEDA-2 group and 22.6% (95% CI 20.1 to 25.3) for the active group.

In the Cox proportional hazards model, the study group variable did not have proportional hazards. Consequently, we split the follow-up into 5-year intervals. In year 0–5 after index date, the NEDA-2 group had an 57% reduced rate of disability pension (HR 0.43; 95% CI 0.33 to 0.55; p<0.001); in year 5–10 after index date, the rate of disability pension was 33% reduced for the NEDA-2 group (HR 0.67; 95% CI 0.40 to 1.01; p=0.06); and for >10 years of follow-up, the association became uncertain (HR 0.92; 95% CI 0.26 to 3.22; p=0.89). Few patients (~15%) contributed with 10 years of follow-up or more. The output from the regression analyses is presented in table 4. As seen, a high EDSS score at treatment start was strongly associated with later disability pension. Moreover, the effect of socioeconomic status at treatment start influenced the rate of disability pension as well. However, as the distribution of socioeconomic status at treatment start was similar, the confounding effect was small. All variables were associated with disability pension except relapse count before treatment start. When testing the interaction between sex and the study group variable, we found no evidence of sex-specific effects on the risk of disability pension (p=0.68).

Table 4

Output from regression analysis (disability pension)

Lastly, we conducted the analyses only including patients with available MRI data. A total of 612 had available data on MR activity in the definition period. Due to the limited study size and few events, we adjusted for sex and age only. We found a marked difference in hazard rates in favour if NEDA-3, but the estimate had wide CIs (0.41; 95% CI 0.14 to 1.17; p=0.09).

Sensitivity analyses

We fitted a model with relapse count (four categories) and EDSS worsening as independent variables instead of the combined variable ‘active disease’. In the income analysis, we found that EDSS worsening (adjusted for relapse count) during the definition period was clearly associated with losing income (HR 1.63; 95% CI 1.27 to 2.09). However, the association between the categorised relapse count (adjusted for EDSS worsening) and losing income was statistically insignificant (p=0.69) (table 5). In the disability pension analyses, both relapse rate and EDSS worsening were associated with the outcome. Patients with EDSS worsening had a 2.6-fold higher rate of disability pension (HR 2.59; 95% CI 2.04 to 3.29). (table 5). The overall association between relapse count and disability pension had a p value of 0.002.

Table 5

Results from sensitivity analyses

Second, we substituted the explanatory variable ‘EDSS worsening’ with ‘delta-EDSS’ and found that for each 1-point increase in EDSS per ‘definition period’ (~3 years), the hazard rate of losing income increased by 28% (HR 1.28; 95% CI 1.18 to 1.38) and the hazard rate of disability pension increased by 57% (HR 1.57; 95% CI 1.46 to 1.69) (table 5).

Lastly, when merging disability pension and loss of income into one composite outcome, we found a 25% lower hazard rate of reaching either disability pension or loss of income for the NEDA-2 group (HR 0.74; 95% CI 0.62 to 0.89).


We conducted a nationwide and population-based cohort study estimating the association between NEDA-2 3 years after starting DMT and two socioeconomic outcomes; a full year without income from salaries and disability pension. We found that NEDA-2 was strongly associated with a reduced risk of disability pension, and particularly so during the first 5 years after the index date. The declining rate of disability pension during follow-up may have occurred as the most disabled patients receive disability pension early during follow-up. Accordingly, the comparison of groups after 10 years of follow-up was between less disabled patients. The rate of losing income was also reduced for patients with NEDA-2, but the effect size was less pronounced. When dividing ‘active disease’ into EDSS worsening and relapse activity, we found that the risk of both outcomes was mainly associated with EDSS worsening. We found no evidence of different risks between male and female patients.

The unpredictable nature of MS makes it difficult to accurately predict the prognosis early in the disease course. This uncertainty is unsatisfying for both patients and healthcare professionals. Consequently, the optimal treatment strategy is continuously being debated. There is an increased tendency of a more proactive therapeutic approach; the tolerance of disease activity has markedly decreased and NEDA has been proposed as a treatment goal.23 However, the use of NEDA as a therapeutic goal has limitations. Most importantly that the predictive accuracy of NEDA (and especially for a patient without NEDA) on future disability accumulation is unclear.24 25 35 36 In relation, studies assessing the effect of relapse activity on later disability worsening are inconsistent.22 37 Most studies that evaluate the prognostic value of NEDA measure disability worsening as changes in the EDSS score—the most widely used disability score in MS research.34 However, the EDSS score’s ability to assess changes in symptoms associated with unemployment such as cognitive impairment, upper limb function and fatigue is limited.38 Consequently, we searched for alternative ways of measuring disease worsening. We selected two outcomes that evaluate the ability for patients to maintain a job, but they also reflect two different socioeconomic situations. In Denmark, patients receiving disability pension can apply at their local municipality and become approved for working a few hours per week and thereby having income. In contrast, patients who lose their income are not necessarily receiving disability pension. The occurrence of a year without income can be related to various situation and not all are related to disability: maternity leave, studying, a high unemployment rate for specific professions or a deliberate life choice. However, long-term unemployment is often a negative life event, and the probability of it being related to MS is likely high. We only had access to annualised data on income, and the precise date of losing income is unknown. Disability pension transfer payments are more likely to be related to disability worsening. The allocation of disability pension is based on various information about the applicant including medical assessments of working capacity. However, a potential limitation of disability pension as the outcome is that the legislation that gives patients access to their pension can change over time. In our study, a change in legislation occurred once during follow-up (2004–2018) in 2013. To mitigate this potentially confounding factor, we adjusted all our analyses for the calendar year at treatment start. Moreover, we analysed data including only patients with index date after 31 December 2012 and found similar results as in the main analysis (data not shown). Another complicating feature of disability pension as the outcome is that patients can receive it for reasons not related to MS. However, the distribution of comorbidities at treatment start did not indicate marked differences between groups, and the proportion of patients with comorbidity was low. Accordingly, few patients are expected to receive disability pension due to other diseases than MS.

Interestingly, we found that socioeconomic status is an important prognostic factor of both disability pension and loss of income. The definition of socioeconomic status, the data source and the data quality influence the strength of potential associations. We had access to nationwide individualised data on all patients and merged 18 socioeconomic subcategories into the eight categories presented (to reduce the number of variables in the statistical models). Our data suggest that even in a country with high equality like Denmark, socioeconomic status at treatment start is a confounder of the risk of disability pension. A Canadian/UK based observational study also found an association between socioeconomic status and disability worsening.39 In addition, a French study indicated that access to second-line therapies was lower for patients of low socioeconomic status.40 However, socioeconomic status was not our primary interest, and further studies are needed to clarify the strength of this finding.

We required that the included patients had not received disability pension at the index date in the disability pension analysis and that they had income at index date in the income analysis. Furthermore, we assessed the effect of on-treatment NEDA-2 after 3 years. The distribution of prescribed DMTs was not similar between groups during the definition period. Consequently, our results describe the prognostic value of 3-year NEDA-2 for patients treated with DMT who do not receive disability pension. A similarly restricted generalisation applies for the income analyses.

In conclusion, the results from this nationwide study indicate that NEDA-2 after 3 years of treatment is associated with a higher chance of maintaining a job for patient with RRMS.


View Abstract


  • Contributors TAC: study concept and design, analysis and interpretation of data and drafting/revising the manuscript. MM and PSS: study concept and design, acquisition of data, analysis or interpretation of data and revising the manuscript. MB: study design and revising the manuscript. ZI, VP, AT, JS, VH, NA, PBS and HBJ: acquisition of data and revising the manuscript

  • Funding This study was funded by Ejnar Jonasson called Johnsen and wife’s memorial fund, Danish Multiple Sclerosis Society, Fonden for neurologisk forskning.

  • Competing interests TAC has received support for congress participation from Merck, Novartis, Biogen and Roche. MB has received support for congress participation from Roche. ZI has served on scientific advisory boards, served as a consultant, received support for congress participation, received speaker honoraria and received research support from Biogen, Merck-Serono, Sanofi-Genzyme, Lundbeck and Novartis. VP has received support for congress participation from Merck. AT has received support for congress participation from Merck, Novartis, Biogen and Roche. JS has received support for congress participation from Genzyme, Biogen, Roche and Merck. VH has received support for congress participation from Roche, Biogen, Merck, Sanofi Genzyme and Almirall. NA has nothing to disclose. PBS has nothing to disclose. HBJ has nothing to disclose. PSS has received personal compensation for serving on advisory boards for Biogen, Merck, Novartis, Teva, MedDay Pharmaceuticals and GSK; has served on steering committees or independent data monitoring boards in trials sponsored by Merck, Teva, GSK and Novartis; and has received speaker honoraria from Biogen, Merck Serono, Teva, Sanofi-Aventis, Genzyme and Novartis. MM has served on scientific advisory boards for Biogen, Sanofi, Teva, Roche, Novartis and Merck; has received honoraria for lecturing from Biogen, Merck, Novartis, Sanofi and Genzyme; and has received support for congress participation from Biogen, Genzyme, Teva and Roche.

  • Patient consent for publication Not required.

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

  • Data availability statement Data may be obtained from a third party and are not publicly available

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