Article Text

Download PDFPDF

Research paper
Excess mortality among patients with multiple sclerosis in Denmark has dropped significantly over the past six decades: a population based study
  1. Nils Koch-Henriksen1,2,
  2. Bjarne Laursen3,
  3. Egon Stenager2,4,
  4. Melinda Magyari2,5
  1. 1 Department of Clinical Epidemiology, Clinical Institute, University of Aarhus, Aarhus, Denmark
  2. 2 The Danish Multiple Sclerosis Registry, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark
  3. 3 National Institute of Public Health, University of Southern Denmark, Copenhagen, Denmark
  4. 4 Institute of Regional Health Research, University of Southern Denmark, Odense, Denmark
  5. 5 Department of Neurology, Danish Multiple Sclerosis Center, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark
  1. Correspondence to Dr Nils Koch-Henriksen, The Danish Multiple Sclerosis Registry, Copenhagen University Hospital Rigshospitalet, Section 7801 Blegdamsvej 9, DK 2100 Copenhagen, Denmark; koch-henriksen{at}


Background Lifetime expectancy in multiple sclerosis (MS) is reduced. Few studies have had sufficient follow-up or sufficient number of patients to assess if survival has improved with time. However, a recent meta-analysis found no time-dependent change in MS excess mortality across studies over recent decades.

Objective To investigate whether short-term all-cause excess mortality in patients with MS in the total Danish population has changed over the last six decades.

Patients and methods We included all patients with MS recorded in the nationwide Danish MS Registry with definite or probable MS and onset from 1950 through 1999. The Danish Civil Registration System provided date of death for all deceased patients with follow-up in 2015, and Statistics Denmark supplied specific population mortality. We calculated excess number of death per 1000 person-years (EDR) and standardised mortality ratio (SMR).

Results We included 18 847 patients among whom 6102 had died as opposed to 2492 expected deaths. EDR was 10.63 (95% CI 10.19 to 11.09) and a SMR was 2.45 (95% CI 2.39 to 2.51). The 15-year EDR dropped gradually from 11.29 (95% CI 9.95 to 12.73) in the 1950–1959 onset cohort to 2.56 (95% CI 1.98 to 3.18) in the 1990–1999 onset cohort, and SMR dropped from 4.48 (95% CI 4.06 to 4.92) to 1.80 (95% CI 1.62 to 1.99).

Conclusion The decline in short-term excess mortality in MS started decades before disease-modifying treatment of MS became available, before use of MRI became widespread, and before the McDonald diagnostic criteria were introduced. A change in the MS cohorts with fewer malignant cases may be a significant contributor.

  • Multiple sclerosis
  • outcome
  • mortality
  • natural history

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.


Multiple sclerosis (MS) is not by itself regarded as a lethal disease, but a large number of follow-up studies have indicated shorter lifetime expectancy in patients with MS relative to the background population.1 2 During recent decades, MS epidemiology and treatment has changed in ways that could affect lifetime expectancy: increasing incidence,3 lower general population mortality and a range of new disease-modifying treatments.

From a methodological point of view, death is an important endpoint. It is precise and usually complete, and comprises in many cases a proxy parameter for prior disease activity and disability.

A recent review and meta-analyses of mortality in patients with MS4 based on published studies over several decades have indicated that standardised mortality ratios (SMRs) of MS relative to the general populations have remained virtually constant over time. Two studies provided the majority of cases for this meta-analysis: a Canadian study5 that showed constant MS mortality and an earlier Danish study6 that indicated improvement of survival in MS over time relative to the background population.

The primary aim of the present study is to analyse whether excess mortality in Danish patients with MS has changed over the last six decades.

Patients and methods

Study design

Cohort follow-up study with death as endpoint.

Study population

The nationwide population based Danish MS Registry has at an almost complete level collected information on all Danish citizens who were alive at a prevalence date in 19497 and all cases with onset of MS since 19488. To ensure sufficient follow-up, we selected patients with onset 1950–1999.

Vital status

Vital status and date of death for all MS cases were up to 1968 provided by national registers and since by the Danish Civil Registration System.9

Population mortality

We calculated expected numbers of deaths (expD) from sex-specific, age-specific and calendar year-specific population life table entries back to 1950, published by Statistics Denmark:

Embedded Image

where n is number of patients in a selection; start(i) and end(i) are calendar year of onset and death or censoring for the i’th patient; popD is the 1 year death density in the population; y is calendar year, and a(y) is age of the i’th patients at year y.

Because population life tables for the first-onset cohorts were available only up to the age of 90 years, we right truncated all analyses that included background population mortality at that age, thereby missing 39 deaths. Only patients who complied with the diagnostic criteria of probable or definite MS according to the criteria of Allison and Millar,10 the Poser criteria (from 1994)11 and the McDonald criteria (from 2005)12 13 were included. No patients with onset before 1950 were included in the cohort analyses to avoid immortal time bias because patients with onset before 1950 would have been selected by survival up to this year. We followed patients from onset until death, emigration or end of follow-up in 2015, whichever came first.

Measures of excess mortality

Excess mortality was calculated as SMR which is the observed number of deaths divided by the expected number as it would have been if the patients were subjected to the sex-specific, age-specific and calendar year-specific background population mortality, and as excess death rate (EDR) which is observed minus expected deaths per 1000 person-years of observation. EDR gives an impression of the numbers of lost lives and is in our view better for comparing subsets with different background mortality.

Trends in mortality

For analysing trends in mortality, we only included the first 15 years after onset because we had observed all included patients for at least 15 years, and because the background population mortality in the affected age ranges is low. For analyses including median age at death and median lifetime from onset, we used the full observation length to follow-up. Patients were divided into five 10-year onset cohorts from 1950–1959 to 1990–1999 defined by the year of clinical onset. We right truncated the analyses at 15 years after onset to ensure the same observation time for all the onset cohorts.

Statistical analyses

We tested trends in EDR by linear regression weighted by the inverse variance, calculated as (number of person-years)2/number of deceased patients in each onset cohort. For the cumulative surviving probability, we used life table analyses with 1 year steps. We calculated 95% CIs for cumulative survival by the exponential variant of the method of Greenwood.14 For SMR and EDR, we calculated 95% CIs by assuming Poisson distribution of counts of deaths. For analysis of mortality by type of onset symptom, we also used Cox regression with inclusion of sex, age at onset and calendar year as covariates. We calculated median time to death and median age at death and their 95% CIs as the interpolated time or age where cumulative surviving probability and its 95% CI intersected the 50% probability line. We used SPSS V.22 and Excel Visual Basic for data management and analyses.


Overall survival

We included 12847 persons who had had clinical onset of definite/probable MS from January 1st 1950 through December 31st 1999, and in this part of the analyses we followed them to death or follow-up by 8th March 2015. Of the patients, 62.8% were female. We followed the patients for 339483 years of observation until death, age of 90 years or follow-up, whichever came first. The number of observed deaths was 6102. On the basis of matched population mortality, the expected number of deaths was 2492. SMR was 2.45 (95% CI 2.39 to 2.51) and EDR was 10.63 (95% CI 10.19 to 11.09). Because we did not have population mortality figures beyond the age of 90 years for all periods, we censored all surviving patients at this age, losing 39 deaths. Figure 1 shows for both sexes life-table curves calculated from onset and from birth with 95% CI as well as the matched population surviving probability. The median remaining life expectancy from onset was 35.0 years in patients with MS (95% CI 34.5 to 35.5) compared with 49.1 years if the patients had been subjected to the matched population mortality from time of onset onwards. The observed median lifetime from birth was 69.1 years among the patients (95% CI 68.6 to 69.5) and the matched population median lifetime was 80.2 years. For both observed and expected surviving times, we had assigned a cumulative surviving probability of 1.00 from birth and up to the age at onset of the patient with MS. As the immortal time error was the same for both the observed and expected mortality, there was no immortal time bias. Figure 1 is for illustration only, as the data are very heterogeneous: the left part of the curves reflects both old and new onset cohorts, whereas the right part reflects lagged mortality based only on the oldest onset cohorts.

Figure 1

Life-table curves from onset and from birth. Survival of patients and corresponding background population are set to unity up to the year of onset of MS. Thus, immortal time errors are present for both patients with MS and population.

Gender and excess mortality

For the whole cohort and through the whole observation period to death or follow-up, male patients with MS had a median surviving time from onset of 29.7 years (matched population: 43.8 years), and the median age at death was 66.0 years (matched population: 73.3 years). The corresponding figures for female patients were 36.7 years (matched population: 50.8 years) and 71.2 years (matched population: 81.3 years) (figure 1). For male patients, SMR was 2.36 and EDR was 12.81. For female patients, the SMR was 2.53 and EDR was 9.41. The two methods of calculating excess mortality gave opposite trends between the sexes, which reflects that SMRs are quotients and EDRs are differences. EDR more directly reflects the difference in number of lives lost. Differences between the sexes in excess mortality during the first 15 years after onset for the whole 1950–1999 onset cohort are shown in the rightmost column of table 1. Findings on excess mortality during the first 15 years were similar with the EDR consistently higher for male than female patients.

Table 1

Fifteen-year excess mortality by onset cohort

Age at onset and mortality

Also in this analysis, we used the full observation time from onset to death or follow-up. With low age at onset, the patients died at a lower age, but they survived longer from onset compared with patients with onset at higher age. For example, patients with onset at <25 years of age died at a median age of 64 years, and the median surviving time from onset was 43 years. In comparison, patients with onset at the age of 55 years or more died at a median age of 75 years, but their median surviving time after onset was less than 17 years. Patients with high age at onset had a higher excess mortality in terms of EDR. Table 2 displays results for five age-at-onset cohorts.

Table 2

Median age at death and median surviving time by age at onset

Presenting symptom and survival

Information on disease course from onset (relapsing–remitting MS or primary progressive MS) was not available in the older part of the Danish MS Registry. Instead, we compared survival in patients with presenting symptoms typical for relapsing–remitting MS: optic, brainstem or sensory, with survival if the presenting symptoms were typical for primary progressive MS: pyramidal, cerebellar or sphincter symptom onset. In 11 386 of the patients, the onset symptom could be classified as either pyramidal/cerebellar/sphincter (n=3612; 58.3% female, mean age at onset 37.0 years) or as optic/brainstem/sensory (n=7774; 64.9% female; mean age at onset 32.3 years). Calculated from onset to death or follow-up, SMR for pyramidal/cerebellar/sphincter onset was 2.64 (95% CI 2.53 to 2.76) and EDR was 14.75 (95% CI 13.76 to 15.78). With optic/sensory/brainstem onset SMR was 2.33 (95% CI 2.25 to 2.41) and EDR was 8.55 (95% CI 8.02 to 9.08). We also compared the mortality of the two groups with Cox regression analysis. Without adjustment, the HR for death with pyramidal/cerebellar symptoms versus brainstem/sensory symptoms at onset was 1.65. However, sex and age at onset had their share in the HR: adjustment for age at onset, sex and calendar year reduced the HR to 1.29 (95% CI 1.22 to 1.36; p<0.0001).

Trends in survival

In this analysis, we only looked at excess mortality in the first 15 years after onset in the five onset cohorts beginning with the 1950–1959 cohort and ending with 1990–1999 cohort. Table 2 shows the SMR and EDR of all-cause mortality for the five onset cohorts and for both sexes along with details. The censoring at the age of 90 years had no influence, because none of the patients had reached that age within 15 years of onset.

The 15-year EDR dropped gradually from 11.29 (95% CI 9.95 to 12.73) in the 1950–1959 onset cohort to 2.56 (95% CI 1.98 to 3.18) in the 1990–1999 onset cohort, and SMR dropped from 4.48 (95% CI 4.06 to 4.92) to 1.80 (95% CI 1.62 to 1.99) through the same onset cohorts where the mean age at onset had increased from 32.9 to 35.4 years (table 2). Among patients in the 1980–1989 onset cohort, 10.6% had been treated with disease-modifying drugs within the 15-year observation time, and in the 1990–1999 onset cohort, 50.6% had been treated.

Figure 2 shows life-table curves for each onset cohort in five panels, depicting the conspicuous improvement of 15-year cumulative surviving probability for each of the five onset cohorts compared with the matched background population. Note that the scale on the Y-axes starts at 0.75. Figure 3 shows the sex-specific EDRs with 95% CI for the five onset cohorts.

Figure 2

Life tables for patients with MS in the first 15 years after onset compared with a matched background population by decade of onset. Note that the scale on the y-axes starts with 0.75.

Figure 3

EDR (excess number of deaths per 1000 person-years of observation) by onset cohort and sex.

Test for trend

With linear regression, weighted by inverse variance for each data point, EDR proved to change significantly with −2.290 per decade after adjustment for sex (95% CI −3.902 to −0,678; p=0.013; R2=0.979). Sex influenced EDR significantly (p=0.010), but it did not interact with cohort (p=0.513).

Mean age at death and retrospective mean disease duration by decade 1950–2009

In this analysis, we looked at dates of death for all patients who died within the six decades between 1950 and 2009 and calculated mean age and mean disease duration at death (table 3). As this calculation was retrospective, we also included cases who were prevalent in 1949 to prevent bias. Patients who died in the 1950’s died at a considerably lower age and after a shorter disease duration than patients who died in the 2000’s.

Table 3

Age at death and retrospective disease duration in all patients with MS who died in 1950–2009


Our study, using data from the Danish MS Registry from 1950 up to the present, has in agreement with numerous other studies5 6 15–25 shown significantly reduced survival in patients with MS compared with the background population. Our key finding is, however, that excess mortality in terms of EDR decreased from 11.29 to 2.56 and SMR from 4.48 to 1.80 between the 1950–1959 and 1990–1999 onset cohorts. We analysed all-cause mortality in the first 15 years after onset in five 10-year onset cohorts. Our reasons for restricting analyses of trend to the first 15 years after onset were to ensure comparability between the five 10-year onset cohorts and that, within the age ranges covered by this interval, typically from 35 to 50 years of age, the background population mortality is rather low. Hence, we may attribute a larger proportion of deaths in this interval to MS.

We chose to compute survival from time of clinical onset rather than from time of diagnosis. Our arguments were that irrelevant circumstances influence the time of diagnosis, and patients who die from MS-related causes will probably always have received an MS diagnosis. Hence, we expect deaths between onset and diagnosis to occur at a rate similar to the background population mortality.

The strong temporal trend shown in our study is in some disagreement with a recent meta-analysis of mortality in MS over time,4 where 12 European and Canadian studies were compared and listed by their midyear of follow-up. SMR varied from study to study according to some heterogeneity, but there was no temporal trend from study to study. Three of the included studies covered themselves long ranges of enrolment or onset periods enabling analyses of trends in survival within the study: a study from Norway19 which covered periods from 1953 to 1967 up to 1983–2003 showed without statistical significance that SMR in the most recent period was a factor of 0.79 lower than in the first period. However, the statistical power was low because of small numbers of deceased cases. In the previous Danish study,6 excess mortality was shown to decrease through time. In the Canadian study from British Columbia,5 three 5-year onset cohorts were compared with enrolment from 1980 to 1985 through 1992–1997. The 10-year surviving probability did not seem to change significantly, but immortal time bias may have inflated survival estimates in the first-onset period. The difference between our study and the meta-analysis4 could also be attributed to the fact that we only looked at the first 15 years after onset.

The increase in lifetime expectancy, found in our study, was not paralleled by slower clinical progression to Expanded Disability Status Scale26 score of 6 (using walking aids) in a MS cohort study from the British Columbia MS Database.27

Our study has also confirmed that, relative to high age at onset, young age at MS onset carries a better prognosis in terms of longer surviving time but a worse prognosis in terms of lower age at death,19 a pattern also seen with other cohort studies28 29 with the endpoints Disability Status Scale score of 6 or 7.

One weakness of our study is that information on disease course from onset (relapsing–remitting MS or primary progressive MS) was not available in the older part of the Danish MS Registry. Instead, we compared survival in patients with presenting symptoms typical of relapsing–remitting MS: optic, brainstem or sensory, with patients with presenting symptoms typical of primary progressive MS: pyramidal, cerebellar or sphincter symptom onset. Patients with pyramidal/cerebellar/sphincter onset had a higher mortality, which is in keeping with the higher mortality among patients with primary progressive MS.5 19 21

The strengths of our study are that we have been able to follow nationwide complete successive onset cohorts of patients with MS from the same relative genetically homogenous population over 65 years in a society with free access to health services for all citizens through more than 60 years and with complete follow-up of vital status. Moreover, the registration of treatment with disease-modifying drugs is complete and nationwide.30 We believe that Danish patients with MS are representative of patients with MS from northern Europe.

There is no obvious explanation for the improved short-time survival in MS in Denmark. Better care of disabled patients, rehabilitation and symptom treatment may play a role; although, the proportion of severely disabled patients within 15 years after onset is low. The median age at assignment of Disability Status Scale score of 7 (wheelchair) was, according to the Lyon European Database for MS, 63 years,28 but in our study only 8.6% of the patients had reached that age within the first 15 years of onset. Better treatment of comorbidities may play a role, but the background population should also have benefited from better treatment of the same diseases. Disease-modifying drugs have shown to improve survival in MS,31 32 but they have only a modest share in the improvement in our study, as the excess mortality started to abate three decades before disease-modifying drugs became available. Only in the most recent onset cohort from 1990 to 1999, a substantial proportion of the patients had been treated with a disease-modifying drug. Better living conditions and welfare during the last 60 years may have benefited chronically ill people more than the background population, but this can only explain a part of the improvement and probably only in the first two or three decades of this survey. Decades ago, benign cases could be more difficult to identify, given the less sophisticated techniques available at that time, and better ascertainment and diagnosis of benign cases may play a role, but still, mortality started to decay in cohorts who had onset long before the use of MRI became widespread in Denmark around 1990, or new diagnostic criteria were applied in the 2000s. The global increase in MS incidence and female to male ratio of incidence3 may have been accompanied by a change in the MS cohorts with fewer malignant cases. Danish patients with MS are probably representative of Northern European patients with MS.


We thank all Danish neurological departments for collaboration through the last 60 years, and we are grateful to Mrs Tina Rømer Olsen who had played a great role in maintenance of the databases and archives of the Danish MS Registry.



  • Contributors NK-H designed the study, contributed to data acquisition, performed the analyses and drafted the manuscript. BL critically revised the manuscript and the statistical methods. ES contributed to data acquisition, critically revised the manuscript and interpreted the results. MM contributed to the concept and design, critically revised the manuscript and interpreted the results.

  • Funding We are grateful to the Danish Multiple Sclerosis Society, which has supported and financed the Danish MS Registry since 1966 without restrictions.

  • Competing interests None declared.

  • Ethics approval The Danish MS Registry is approved for medical research by the Danish Data Protection agency, journal 2008540482. Data extracted for the present study contained no personal identification. According to the Danish regulations, ethical committees need not be involved in observational register studies. For the processing of register data, patient consent is not required.

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

Linked Articles