Estimation of annual probabilities of changing disability levels in Australians with relapsing-remitting multiple sclerosis

Abstract

Background:

Transition probabilities are the engine within many health economics decision models. However, the probabilities of progression of disability due to multiple sclerosis (MS) have not previously been estimated in Australia.

Objectives:

To estimate annual probabilities of changing disability levels in Australians with relapsing-remitting MS (RRMS).

Methods:

Combining data from Ausimmune/Ausimmune Longitudinal (2003–2011) and Tasmanian MS Longitudinal (2002–2005) studies (n = 330), annual transition probabilities were obtained between no/mild (Expanded Disability Status Scale (EDSS) levels 0–3.5), moderate (EDSS 4–6.0) and severe (EDSS 6.5–9.5) disability.

Results:

From no/mild disability, 6.4% (95% confidence interval (CI): 4.7–8.4) and 0.1% (0.0–0.2) progressed to moderate and severe disability annually, respectively. From moderate disability, 6.9% (1.0–11.4) improved (to no/mild state) and 2.6% (1.1–4.5) worsened. From severe disability, 0.0% improved to moderate and no/mild disability. Male sex, age at onset, longer disease duration, not using immunotherapies greater than 3 months and a history of relapse were related to higher probabilities of worsening.

Conclusion:

We have estimated probabilities of changing disability levels in Australians with RRMS. Probabilities differed between various subgroups, but due to small sample sizes, results should be interpreted with caution. Our findings will be helpful in predicting long-term disease outcomes and in health economic evaluations of MS.

Keywords

Multiple sclerosis progression transition probabilities EDSS Australia

Background

Multiple sclerosis (MS) is an inflammatory, neurodegenerative, progressive disease of the central nervous system leading to increasing disability over time. The majority (85%–90%) of people with MS are diagnosed with a clinical subtype characterised by episodes of acute neurological deterioration, followed by partial or complete recovery, referred to as relapsing-remitting MS (RRMS).¹ MS has its onset in the most productive years of life, with typical age of onset between 25 and 40 years.² The impact of MS on people’s lives is multifaceted, profoundly disrupting careers, relationships, families and health-related quality of life (HRQoL).³ Transition probabilities are widely used in the assessment of disease progression between different levels of disease severities.^4–7 Transition probabilities are numbered between 0 and 1, where 0 = impossible and 1 = certain.⁸ Country-specific health economic evaluations of available treatment options guide evidence-based reimbursement decisions, with transition probabilities an important input to modelled evaluations.^4,8

In the medical decision analysis field, multi-state Markov models are commonly used to describe processes by which individuals move through a finite number of health states in a unit of time.⁹ Markov transition probabilities are presented in the form of a matrix, which is often complicated to estimate.¹⁰ The probabilities of changes in disability levels (both forwards and backwards) due to MS have been documented in the United States and other parts of the world.^5,7,11–14

While other researchers have attempted to estimate probabilities of changing disability levels due to ageing for the overall (not specifically MS) Australian population,^15–17 probabilities of disability progression in Australians with MS have not been estimated. Estimation of transition probabilities from an Australian population with the most common form of MS (RRMS) at varying levels of disability severity will be useful not only to predict disease outcomes but also to assess the cost-effectiveness of various MS interventions in Australian and similar populations. We aimed to bridge this important gap by utilising data from two longitudinal epidemiological studies of Australians with MS and to investigate the impact of baseline characteristics on transition probabilities to identify people with RRMS who are more likely to progress or improve disability states.

Materials and methods

Study design

Data from the Ausimmune Longitudinal Study (AusLong) and Tasmanian MS Longitudinal study (TasMSL) were combined to estimate the probabilities of RRMS progression and to adjust these probabilities for important risk factors. The Ausimmune/AusLong Study included 279 people with a first clinical diagnosis of central nervous system demyelination, about 69% and 75% of whom had converted to MS by 2/3-year review and 5-year review, respectively.¹⁸ Participants were recruited as part of the Ausimmune case–control study between 2003 and 2006 and have been followed up in the AusLong Study.¹⁹ Face-to-face assessments occurred in the Ausimmune Study at baseline and at 2/3-year reviews and in the AusLong Study at year 5; year 10 reviews are in progress. The analysis presented here includes data from entry into the Ausimmune Study (‘study entry’) until the 5-year (AusLong) review. Of all Ausimmune/AusLong cases, 258 had an eventual diagnosis of relapsing-onset disease type. Of those, 169 participants had a ‘classic first demyelinating episode (FDE)’, with their initial onset event occurring just prior to study participation. The remaining 89 participants had their initial onset event in the more distant past, but this was not recognised at that point as an FDE. The Ausimmune Study was approved by nine regional Human Research Ethics Committees, with informed consent from all study participants obtained prior to participation.^18,19 The AusLong Study was approved by five regional Human Research Ethics Committees.

The TasMSL Study is a prospective cohort study of 198 prevalent cases with definite MS living in Tasmania, Australia.²⁰ Of all TasMSL cases, 149 had RRMS at onset. The TasMSL Study followed people with MS for an average of 2.3 years (from 2002 to 2005), with face-to-face reviews every 6 months (for patient interview, collection of biological samples and clinical assessments). Clinical disability and cognitive function were assessed at annual review, while relapses were reported in real time by telephone. Ethics approval was obtained, and informed consent was obtained from all study participants prior to participation.²¹

Data setup and cleaning

We combined both datasets (n = 407 with RRMS) to calculate probabilities and adjust for important covariates. Expanded Disability Status Scale (EDSS) is the most widely used scale to quantify disability in MS.²² EDSS scores were available for at least three time points for the majority of our aggregate sample. The details on how EDSS was measured for each study are provided elsewhere.^21,23 In line with previous studies in the health economics of MS,^3,24–27 we classified participants into three broad disability groups based on their EDSS scores: no/mild disability (EDSS 0–3.5), moderate disability (EDSS 4–6.0), and severe disability (EDSS 6.5–9.5). Our analysis excluded cases with EDSS data missing at two or more time points (n = 25, 16 from AusLong). We also excluded the 47 AusLong Study participants who had not converted to MS by 5-year review. We further excluded five cases from the remaining sample who had severe disability at baseline to avoid any doubts whether these patients could be classified as RRMS at inclusion, leaving 330 cases in this analysis.

Measurement of covariates

Age, sex, disease duration from symptom onset and immunotherapy usage were available at study entry. In order to investigate the impact of individual characteristics of our sample on transition probabilities, we created four binary covariates: age (age ⩾41.0 years at study entry = 1, and 0 otherwise), sex (1 = female and 0 = male), disease duration from symptom onset (⩾5.2 years at study entry = 1, and 0 otherwise), and immunotherapy usage (on-treatment greater than 3 months during the study = 1, and 0 otherwise). The binary cut-points for age and disease duration were based on variable means (Table 1). Notably, the mean and median age of our sample were the same (41 years). However, the median disease duration of our sample (1.1 years) was much smaller. Because the median-based cut-point did not make a lot of theoretical sense, we opted to go for a mean-based cut-point (5.2 years) for disease duration (and also for ‘age’ to stay consistent). A further binary variable ‘relapse’ was created (1 = history of relapse, and 0 otherwise) to discriminate between people with and without a history of relapse during the study period.

Table 1.

Demographic and other features of the study sample by MS severity.

Variable names	Descriptive statistics
	No/Mild (n = 279)	Moderate (n = 51)	Severe (n = 0)	Total (n = 330)
Female (%)	220 (79)	40 (78)	N/A	260 (79)
Mean age at first EDSS measurement, years (SD)	39.2 (9.8)	50.0 (9.9)	N/A	40.9 (10.6)
Median age at first EDSS measurement, years (IQR)	39.0 (33.0–46.0)	49.0 (43.0–57.0)	N/A	41.0 (34.0–48.0)
Mean EDSS score at first measurement (SD)	1.5 (1.0)	4.4 (0.9)	N/A	2.0 (1.5)
Median EDSS score at first measurement (IQR)	1.5 (1.0–2.0)	4.0 (3.5–5.0)	N/A	2.0 (1.0–2.5)
Mean duration of MS from symptom onset to study entry, years (SD)	3.6 (5.7)	13.3 (12.0)	N/A	5.2 (7.9)
Median duration of MS from symptom onset to study entry, years (IQR)	0.9 (0.3–4.0)	12.0 (3.0–20.0)	N/A	1.1 (0.4–7.0)
Had relapse during study period^a (%)	214 (77)	31 (61)	N/A	245 (74)
Any use of immunotherapy greater than 3 months during the study (%)	116 (42)	38 (75)	N/A	154 (47)

IQR: interquartile range; MS: multiple scleroses; SD: standard deviation.

No/Mild severity includes Expanded Disability Status Scale (EDSS) levels 0–3.5, moderate includes levels 4–6.0 and severe includes levels 6.5–9.5.

Study period varied from individual to individual (2 to 5 years).

Analysis

We fitted a three-state continuous-time Markov model to describe the progression of RRMS under the assumption of time homogeneity.^{5,6,12,28–30} The analysis was performed using the ‘msm’ package for R,⁹ fitting a multi-state model based on panel data under a continuous-time Markov model. First, we combined data on a selected set of variables from the TasMSL and AusLong databases. Second, we defined a 3 × 3 matrix of transition intensities based on the disability categories considered and specified the allowable transition. Any health state was allowed to transit to itself or to the next or previous levels. Third, we used the ‘msm’ numerical procedure to find the maximum likelihood estimates of transition intensities.⁹ Annual transition probabilities were calculated by taking the matrix exponential of the resultant scaled transition intensity matrix.⁹ Notably, our probabilities estimates are not just the numbers of transitions with specific outcomes divided by the total number of transitions at risk as this simple procedure fails to account for the irregular follow-up times (between 1 and 3 years in our data). Our choice of a state-transition Markov model suitable for longitudinal data analysis overcomes this problem by enabling the scaled maximum likelihood estimates of transition intensities on the face of intermittently observed short-term data.⁹

In the next step, we adjusted the transition probabilities for age, sex and relapse history. Disease duration was excluded from this stage of analysis due to being positively correlated with age. Furthermore, the immunotherapy history was excluded due to being positively correlated with relapse history. The effects of covariates were modelled under the proportional intensities assumption.³¹ Covariates were assumed to be constant between the observation times of the Markov process. The covariate-adjusted transition intensity matrix was obtained by treating the baseline transition intensities as a log-linear function of covariates.⁹ Bootstrap confidence intervals (CIs) were obtained for the overall transition probability matrices.

Separate transition probability matrices were next derived for each category of the five binary covariates based on univariate models. Numerical overflow (as a result of the small sample sizes) prevented estimation of CIs for transition probability matrices for these subgroups. Importantly, because death due to MS in RRMS phase is not very likely,³² probabilities of death were not calculated.

We undertook two sensitivity analyses to exclude the potential bias that could be introduced if the measured EDSS reflected a recent or current relapse. The disability level of ‘classic FDE’ cases was theoretically zero the day before their FDE. We therefore set their disability level as no/mild at study entry. For those who had an unrecognised previous event sometime in the past, we could not make that assumption and used the measured EDSS at study entry. Examination of the data showed that there were only 3 of the 85 cases with prior events who had a disability state at study entry that was more severe than their disability state at next review. We undertook a sensitivity analysis excluding these cases to determine the extent of any potential bias. A second issue that we identified was that 28 people with MS reported having a relapse at the time of disability assessment; again, there was a possibility of relapse-induced increase in their disability. However, examination of the data for most of these individuals (25 out of 28) showed that the EDSS category was not higher during the relapse compared to pre-relapse. A further sensitivity analysis was undertaken by excluding the three cases where it may have been an issue.

Results

An aggregate sample of 330 participants provided evaluable data (1297 person-years of follow-up and 660 year-to-year transitions, on average 2 per person). Just over three-quarters of the participants were women (79%) (Table 1). Men were slightly older than women (42 vs 40 years) at study entry. Additionally, the moderately disabled group on average was more than 11 years older than the no/mildly disabled. The average duration of MS disease was 6.2 years for men and 4.9 years for women. The mean study entry EDSS for the moderately disabled group (4.4) was higher than for the no/mildly disabled (1.5) groups. Mean duration of MS differed as expected between the no/mildly disabled and moderately disabled groups, with the moderately disabled group having the disease duration of 13.3 years (on average 10 years longer than no/mild group). Eighty-five percent and 15% of our sample had no/mild and moderate disability, respectively, at study entry.

Table 2 shows the total number of transitions from any one disability state to the other. Most participants remained in their study entry disability states throughout the follow-up period. On 61 occasions, a transition was observed from no/mild to moderate disability state, while only on 1 occasion was there an observation of severe disability following a no/mild disability observation. On a small number of occasions, an observation of moderate disability was followed by one of no/mild disability or severe disability. There was no transition observed from severe state (to moderate and no/mild states).

Table 2.

Total number of transitions between three health states during the study period.

		To
		No/Mild	Moderate	Severe
From	No/Mild	460	61	1
	Moderate	9	120	6
	Severe	0	0	3

No/Mild severity includes Expanded Disability Status Scale (EDSS) levels 0–3.5, moderate includes levels 4–6.0 and severe includes levels 6.5–9.5.

In all, 261, 63 and 7 people with MS contributed 0, 1 and 2 year-to-year transitions, respectively.

The baseline unadjusted transition probabilities as well as covariate-adjusted transition probabilities are presented in Table 3. From the covariate-adjusted transition probability matrix, the no/mild disability group had a 6.4% probability of transitioning to the moderate state, less than 1% probability of transitioning to a severe state, and 93% probability of remaining in the no/mild disability state 1 year later. The moderate disability group was more than twice as likely to improve to the no/mild state (probability: 6.9%) than to progress to the severe state (probability: 2.6%). The severe disability group had 0% probability of improvement to the moderate and no/mild states.

Table 3.

Annual transition probabilities of MS progression.

Unadjusted
To
		No/Mild	Moderate	Severe
From	No/Mild	0.940 (0.923–0.954)	0.059 (0.045–0.075)	0.001 (0.000–0.002)
	Moderate	0.067 (0.026–0.117)	0.900 (0.845–0.943)	0.030 (0.010–0.055)
	Severe	0	0	1.000 (1.000–1.000)
Adjusted^a
To
		No/Mild	Moderate	Severe
From	No/Mild	0.934 (0.915–0.953)	0.064 (0.047–0.084)	0.001 (0.000–0.002)
	Moderate	0.069 (0.010–0.114)	0.904 (0.864–0.978)	0.026 (0.007–0.048)
	Severe	0	0	1.00 (1.000–1.000)

No/Mild severity includes Expanded Disability Status Scale (EDSS) levels 0–3.5, moderate includes levels 4–6.0 and severe includes levels 6.5–9.5.

Probabilities adjusted for age, sex and relapse rate.

Our sensitivity analyses (Supplemental Table 1) where we excluded a subset of those (1) with an unrecognised previous event well before study entry and a higher disability at study entry than on subsequent review and (2) having a relapse at the time of disability assessment did not materially alter the key results.

The next step was to derive separate transition probability matrices for each category of the five binary covariates (sex (male vs female), age at study entry (⩾41.0 years vs <41.0 years), duration from symptom onset (⩾5.2 years vs <5.2 years), relapse (ever had relapse vs no relapse) and immunotherapy (greater than 3 months of use of immunotherapy vs no immunotherapy or <3 months of use)). Table 4 reports 10 sets of transition probabilities in sections a, b, c, d and e. Baseline transition intensities and the estimated hazard ratios for each covariate and their CIs are presented in Supplemental Table 2. Being male or older, having a longer disease duration or a history of relapse, and not being on immunotherapy were associated with a higher likelihood of forward progression and lower likelihood of backward progression. For example, people aged ⩾41.0 years had probabilities of progression of 9% (from no/mild to moderate), 0.2% (no/mild to severe), and 4% (from moderate to severe). By contrast, for people aged <41.0 years the comparable transition probabilities were 4%, 0.1% and 2.5%. Furthermore, older people were less likely to improve (e.g. 5% for older vs 7% for younger improved from moderate to no/mild).

Table 4.

Crude annual transition probability matrices by subgroups.

	No/Mild	Moderate	Severe
Female/Male
No/Mild	0.940/0.928	0.058/0.069	0.001/0.002
Moderate	0.064/0.035	0.909/0.923	0.027/0.042
Severe	0/0	0/0	1.000/1.000
Young (age <41 years)/Older (age ⩾41 years)
No/Mild	0.960/0.908	0.039/0.090	0.001/0.002
Moderate	0.072/0.051	0.903/0.907	0.025/0.042
Severe	0/0	0/0	1.000/1.000
Disease duration <5.2 years/Disease duration ⩾5.2 years
No/Mild	0.958/0.765	0.040/0.228	0.000/0.007
Moderate	0.59/0.050	0.929/0.897	0.012/0.052
Severe	0/0	0/0	1.000/1.000
No relapse during study period/Relapse during study period
No/Mild	0.943/0.909	0.055/0.090	0.001/0.001
Moderate	0.068/0.038	0.896/0.947	0.036/0.015
Severe	0/0	0/0	1.000/1.000
Immunotherapy/No immunotherapy
No/Mild	0.971/0.864	0.028/0.133	0.000/0.003
Moderate	0.120/0.028	0.867/0.934	0.013/0.038
Severe	0/0	0/0	1.000/1.000

Discussion

Our study estimated probabilities of disability progression for a sample of Australians with RRMS at varying levels of disability severity. Importantly, our study was based on longitudinal data from two well-conducted Australian MS studies to calculate transition probabilities. The longitudinal design should provide more accurate assessment of disability progression than previous cross-sectional studies, while the focus on a single disease provides data specific for MS rather than for the general Australian population.^15,17 We found that a very high proportion of people with MS remained in the same disability states across the course of follow-up (93% no/mild, 91% moderate and 100% severe), indicating that MS progresses very slowly on average. The probabilities of moving between distant disability states (from no/mild to severe (0.1%) or from severe to no/mild (<0.0%)) were quite low, which aligns with other MS progression studies.⁵ Moreover, similar to other studies, we found a higher probability of disability progression for males^33,34 and older people with MS.³⁵ Our findings regarding the higher likelihood of progression of people with longer disease duration (⩾5.2 years) support the existing evidence that considers longer disease duration as an important predictor of greater EDSS increase.³⁶ Also, the higher disability progression of those with a relapse history in our sample is in line with the observations from other authors who found a positive relationship between relapses and disability progression.^36–38 Finally, the lower probabilities of disability progression of those on immunotherapy in our sample align well with the existing evidence suggesting that immunotherapy is effective in reducing disability progression in people with RRMS.^36,39

Transition probabilities of disability progression in MS have been estimated in other countries. The comparability of our probability estimates with those from other studies is, however, limited because those studies^11,12,40 defined disease classes and progression of disability differently. For instance, the transition probabilities of experiencing and sustaining an EDSS score ⩾3 within 5 years of study entry in 267 people with MS were found to be 0.18 for people with baseline EDSS ⩽1 and 0.29 for a baseline EDSS = 2.¹¹ Another study classified 140 people with MS into three disability categories, namely, no disability (EDSS ⩽1.5), mild disability (EDSS 2–2.5) and moderate to severe disability (EDSS ⩾3).¹² The study reported that the probability of sustained MS progression among those having no disability at baseline was 0.46 (for those treated with intramuscular injection (IMI) Betaferon-1a at baseline) versus 0.55 (for placebo). The comparable figures for those having mild disability at baseline were 0.30 (IMI Betaferon-1a) versus 0.60 (placebo).

Certain limitations need to be considered when interpreting our results. First, our sample was over-represented by Tasmanian people with MS and did not include any participants from Western Australia, South Australia, the Australian Capital Territory, or the Northern Territory. Second, the AusLong Study comprises people in the early phase of MS. Thus, many participants in our analysis had low EDSS scores (i.e. EDSS <3) at study entry, with people with no/mild disability levels tending to have lower forward progression probabilities and conversely a higher probability of remaining in that state.⁵ Third, transitions could be history-dependent, but our model did not consider history of previous events (i.e. past health states or how long people have been in any particular health state). Fourth, by grouping the participants into three broad disability groups, some of the transitions may have been due to measurement error around EDSS cut-points. Fifth, because of the small sample size and short follow-up period, particularly those from the subgroup analyses in which in some cases the likelihood of improvement is sometimes greater than that of progression, the results should be interpreted with caution. Sixth, transition probabilities may depend on treatment approaches that may differ between patients, but our analysis was not able to provide estimates of probabilities broken down by treatment type. Finally, we did not independently consider a category of ‘no disability’ due to small numbers⁴¹ and to match up with our previous work.^3,27 However, future work on costs, utilities, probabilities and modelling will attempt to split ‘no disability’ from ‘mild disability’ as we include larger sample sizes.

Our choice of a state-transition Markov model suitable for longitudinal data analysis aligns with other studies in the field that recommend transition models over standard (survival) models, especially given the irregular follow-up times and intermittently missing short-term observed data.^5,9,29,42 In addition, survival analysis generally considers the EDSS changes that are sustained and thus may not be well-suited to studies in MS that involve both relapse and remission as cardinal features.⁵

While uncertainties exist in our results, the estimates reported in our study are consistent with the findings of the existing literature in the field. We recommend the use of these specific Australian transition probability estimates (with some degree of caution) in future local health economic evaluations due to three key reasons: (1) no alternative data are available from other similar nations that match up with our previous work on costs and health utilities in MS;^3,27 (2) transition probabilities may differ between nations due to different treatment patterns and hence the different intervention effects; and (3) the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) recommends the use of locally driven input parameters when they are available, especially when local treatment patterns may differ from international data.⁴³

The methods that we used are capable of producing increasingly reliable results subject to the availability of more data. Future studies are recommended to investigate the impact of time-varying characteristics of people with MS (when better long-term observed data become available), especially given that covariate values and their effects might change substantially in the long run. In addition, future probability estimates should consider history of previous events, and estimation of transition probabilities for progressive forms of MS.

Conclusion

Our study has quantified transition probabilities for an Australian sample of people with MS and has demonstrated that progression of MS is slow. Furthermore, the probabilities differ between various subgroups based on participants’ characteristics. Male sex, age at onset ⩾41.0 years, longer disease duration (⩾5.2 years), not being on immunotherapy for greater than 3 months and having a history of relapse increased the risk of worsening disability. Because of the small sample size and short follow-up period, particularly those from the subgroup analyses, the results should be interpreted with caution. The estimates that we have generated will be useful in developing a Markov state-transition model of the progression of MS to calculate the impact of MS on life expectancy, quality-adjusted life years and total lifetime costs of MS in Australian and similar populations.

Supplemental Material

MSJ806103_supplemental_table_1 – Supplemental material for Estimation of annual probabilities of changing disability levels in Australians with relapsing-remitting multiple sclerosis

Supplemental material, MSJ806103_supplemental_table_1 for Estimation of annual probabilities of changing disability levels in Australians with relapsing-remitting multiple sclerosis by Hasnat Ahmad, Ingrid van der Mei, Bruce V Taylor, Robyn M Lucas, Anne-Louise Ponsonby, Jeannette Lechner-Scott, Keith Dear, Patricia Valery, Philip M Clarke, Steve Simpson and Andrew J Palmer in Multiple Sclerosis Journal

Supplemental Material

MSJ806103_supplemental_table_2 – Supplemental material for Estimation of annual probabilities of changing disability levels in Australians with relapsing-remitting multiple sclerosis

Supplemental material, MSJ806103_supplemental_table_2 for Estimation of annual probabilities of changing disability levels in Australians with relapsing-remitting multiple sclerosis by Hasnat Ahmad, Ingrid van der Mei, Bruce V Taylor, Robyn M Lucas, Anne-Louise Ponsonby, Jeannette Lechner-Scott, Keith Dear, Patricia Valery, Philip M Clarke, Steve Simpson and Andrew J Palmer in Multiple Sclerosis Journal

Footnotes

Acknowledgements

We express our heartfelt thanks to the participants in the Ausimmune and AusLong studies for their time and energy, without which we could not have realised this work. The authors also thank the paid research personnel, including the local research officers: Susan Agland, BN, Hunter New England Health, Newcastle, New South Wales; Barbara Alexander, BN, Queensland Institute for Medical Research, Queensland; Marcia Davis, MD, Queensland Institute for Medical Research, Queensland; Zoe Dunlop, BN, Barwon Health, Geelong Hospital, Victoria; Rosalie Scott, BN, Royal Brisbane and Women’s Hospital, Queensland; Marie Steele, RN, Royal Brisbane and Women’s Hospital, Queensland; Catherine Turner, MPH&TM, Menzies Research Institute, Tasmania; Brenda Wood, RN, Menzies Research Institute, Tasmania; and the Ausimmune Study project officers during the course of the study: Jane Gresham, MA (Int Law), National Centre for Epidemiology and Population Health, The Australian National University, Canberra; Australian Capital Territory; Camilla Jozwick, BSc(Hons), National Centre for Epidemiology and Population Health, The Australian National University, Canberra; Australian Capital Territory; Helen Rodgers, RN, National Centre for Epidemiology and Population Health, The Australian National University, Canberra; Australian Capital Territory. The members of the Ausimmune/AusLong Investigators Group are as follows: Robyn M Lucas (National Centre for Epidemiology and Population Health, Canberra); Keith Dear (Duke Kunshan University, Kunshan, China); Anne-Louise Ponsonby and Terry Dwyer (Murdoch Children’s Research Institute, Melbourne, Australia); Ingrid van der Mei, Leigh Blizzard, Steve Simpson, Jr and Bruce V Taylor (Menzies Institute for Medical Research, University of Tasmania, Hobart, Australia); Simon Broadley (School of Medicine, Griffith University, Gold Coast Campus, Australia); Trevor Kilpatrick (Centre for Neurosciences, Department of Anatomy and Neuroscience, University of Melbourne, Melbourne, Australia); David Williams and Jeanette Lechner-Scott (University of Newcastle, Newcastle, Australia); Cameron Shaw and Caron Chapman (Barwon Health, Geelong, Australia); Alan Coulthard (University of Queensland, Brisbane, Australia); Michael P Pender (The University of Queensland, Brisbane, Australia); and Patricia Valery (QIMR Berghofer Medical Research Institute, Brisbane, Australia). We also thank Mr Christopher Jackson (the author of ‘msm package for R’ and a Senior Investigator Statistician at the MRC Biostatistics Unit, University of Cambridge) and Mr Petr Otahal (Statistical Officer, Menzies Institute for Medical Research, University of Tasmania) for their expert feedback in successfully achieving the statistics reported in this paper.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by Multiple Sclerosis Research Australia (Grant Number: 14-039)

ORCID iD

Steve Simpson

References

Miller

Leary

. Primary-progressive multiple sclerosis. Lancet Neurol 2007; 6: 903–912.

Tao

Simpson

van der Mei

, et al. Higher latitude is significantly associated with an earlier age of disease onset in multiple sclerosis. J Neurol Neurosurg Psychiatry 2016 87: 1343–1349.

Ahmad

Taylor

van der Mei

, et al. The impact of multiple sclerosis severity on health state utility values: Evidence from Australia. Mult Scler 2017; 23: 1157–1166.

Fleurence

Hollenbeak

. Rates and probabilities in economic modelling. Pharmacoeconomics 2007; 25: 3–6.

Mandel

Mercier

Eckert

, et al. Estimating time to disease progression comparing transition models and survival methods – An analysis of multiple sclerosis data. Biometrics 2013; 69: 225–234.

Mandel

Betensky

. Estimating time-to-event from longitudinal ordinal data using random-effects Markov models: Application to multiple sclerosis progression. Biostatistics 2008; 9: 750–764.

Gauthier

Mandel

Guttmann

, et al. Predicting short-term disability in multiple sclerosis. Neurology 2007; 68: 2059–2065.

Gray

Clarke

Wolstenholme

, et al. Applied methods of cost-effectiveness analysis in healthcare. Oxford: Oxford University Press, 2010.

Jackson

. Multi-state modelling with R: The MSM package. 1.6.6 ed. Cambridge: School of Clinical Medicine, MRC Biostatistics Unit, 2018.

10.

Craig

Sendi

. Estimation of the transition matrix of a discrete-time Markov chain. Health Econ 2002; 11: 33–42.

11.

Mandel

Gauthier

Guttmann

CRG

, et al. Estimating time to event from longitudinal categorical data: An analysis of multiple sclerosis progression. J Am Stat Assoc 2007; 102: 1254–1266.

12.

Mandel

Betensky

. Estimating time-to-event from longitudinal categorical data using random effects Markov models: Application to multiple sclerosis progression 2008; 9: 750–764.

13.

Holmoy

Celius

. Cost-effectiveness of natalizumab in multiple sclerosis. Expert Rev Pharmacoecon Outcomes Res 2008; 8: 11–21.

14.

Gani

Giovannoni

Bates

, et al. Cost-effectiveness analyses of natalizumab (Tysabri®) compared with other disease-modifying therapies for people with highly active relapsing-remitting multiple sclerosis in the UK. Pharmacoeconomics 2008; 26: 617–627.

15.

Leung

. Projecting the needs and costs of long term care in Australia. Aust Actuar J 2004; 10: 343.

16.

Hariyanto

Dickson

Pitt

. Estimation of disability transition probabilities in Australia I: Preliminary. Ann Actuar Sci 2014; 8: 131–155.

17.

Hariyanto

Dickson

Pitt

. Estimation of disability transition probabilities in Australia II: Implementation. Ann Actuar Sci 2014; 8: 156–175.

18.

Saul

Ponsonby

A-L

Lucas

, et al. Stressful life events and the risk of initial central nervous system demyelination. Mult Scler 2017; 23: 1000–1007.

19.

Pan

Simpson

van der Mei

, et al. Role of genetic susceptibility variants in predicting clinical course in multiple sclerosis: A cohort study. J Neurol Neurosurg Psychiatry 2016; 87: 1204–1211.

20.

Pittas

Ponsonby

A-L

van der Mei

, et al. Smoking is associated with progressive disease course and increased progression in clinical disability in a prospective cohort of people with multiple sclerosis. J Neurol 2009; 256: 577–585.

21.

Lin

Taylor

Simpson

, et al. Association between multiple sclerosis risk-associated SNPs and relapse and disability–A prospective cohort study. Mult Scler 2014; 20: 313–321.

22.

Prosser

Kuntz

Bar-Or

, et al. Cost-effectiveness of interferon beta-1a, interferon beta-1b, and glatiramer acetate in newly diagnosed non-primary progressive multiple sclerosis. Value Health 2004; 7: 554–568.

23.

Simpson

Tan

Otahal

, et al. Anxiety, depression and fatigue at 5-year review following CNS demyelination. Acta Neurol Scand 2016; 134: 403–413.

24.

Karampampa

Gustavsson

Miltenburger

, et al. Treatment experience, burden and unmet needs (TRIBUNE) in MS study: Results from five European countries. Mult Scler 2012; 18: 7–15.

25.

Kobelt

Berg

Lindgren

, et al. Costs and quality of life of multiple sclerosis in Switzerland. Eur J Health Econ 2006; 7: 86–95.

26.

Kobelt

Berg

Lindgren

, et al. Costs and quality of life in multiple sclerosis in Europe: Method of assessment and analysis. Eur J Health Econ 2006; 7: 5–13.

27.

Palmer

Colman

O’Leary

, et al. The economic impact of multiple sclerosis in Australia in 2010. Mult Scler 2013; 19: 1640–1646.

28.

Andersen

Borgan

Gill

, et al. Statistical models based on counting processes. New York: Springer-Verlag, 1995.

29.

Welton

Ades

. Estimation of Markov chain transition probabilities and rates from fully and partially observed data: Uncertainty propagation, evidence synthesis, and model calibration. Med Decis Making 2005; 25: 633–645.

30.

Cox

Miller

. The theory of stochastic processes. New York: Chapman and Hall, 1965.

31.

Marshall

Jones

. Multi-state models and diabetic retinopathy. Stat Med 1995; 14: 1975–1983.

32.

Scalfari

Knappertz

Cutter

, et al. Mortality in patients with multiple sclerosis. Neurology 2013; 81: 184–192.

33.

Bergamaschi

. Prognostic factors in multiple sclerosis. Int Rev Neurobiol 2007; 79: 423–447.

34.

Ribbons

McElduff

Boz

, et al. Male sex is independently associated with faster disability accumulation in relapse-onset MS but not in primary progressive MS. PLoS One 2015; 10: e0122686.

35.

Trojano

Liguori

Bosco Zimatore

, et al. Age-related disability in multiple sclerosis. Ann Neurol 2002; 51: 475–480.

36.

Jokubaitis

Spelman

Kalincik

, et al. Predictors of long-term disability accrual in relapse-onset multiple sclerosis. Ann Neurol 2016; 80: 89–100.

37.

Hirst

Ingram

Pearson

, et al. Contribution of relapses to disability in multiple sclerosis. J Neurol 2008; 255: 280–287.

38.

Fahrbach

Huelin

Martin

, et al. Relating relapse and T2 lesion changes to disability progression in multiple sclerosis: A systematic literature review and regression analysis. BMC Neurol 2013; 13: 180.

39.

Tsivgoulis

Katsanos

Grigoriadis

, et al. The effect of disease modifying therapies on disease progression in patients with relapsing-remitting multiple sclerosis: A systematic review and meta-analysis. PLoS One 2015; 10: e0144538.

40.

Gani

Nixon

Hughes

, et al. Estimating the rates of disability progression in people with active relapsing-remitting multiple sclerosis. J Med Econ 2007; 10: 79–89.

41.

Hiligsmann

Ethgen

Bruyère

, et al. Development and validation of a Markov microsimulation model for the economic evaluation of treatments in osteoporosis. Value Health 2009; 12: 687–696.

42.

Kalbfleisch

Lawless

. The analysis of panel data under a Markov assumption. J Am Stat Assoc 1985; 80: 863–871.

43.

Drummond

Barbieri

Cook

, et al. Transferability of economic evaluations across jurisdictions: ISPOR good research practices task force report. Value Health 2009; 12: 409–418.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.10 MB

0.12 MB