Epidemiology and therapeutic management of highly active relapsing-remitting multiple sclerosis adults in the French national health insurance database

Arnaud Kwiatkowski Neurologie, Groupement des Hôpitaux de l’Institut Catholique de Lille Hôpital Saint Vincent de Paul Marianne Payet Direction des Affaires Médicales, Merck s.a.s., 37 rue Saint Romain, 69008 Lyon, France. Emmanuelle Préaud Direction Affaires Publiques et Accès au Marché, Merck s.a.s., 37 rue Saint Romain, 69008 Lyon, France. Ludovic Lamarsalle HEVA, 186 Avenue Thiers, 69006 Lyon Fanny Raguideau HEVA, 186 Avenue Thiers, 69006 Lyon Olivier Chevreuil (  olivier.chevreuil@merckgroup.com ) Direction des Affaires Médicales, Merck s.a.s., 37 rue Saint Romain, 69008 Lyon, France. Benoit van Hille Direction des Affaires Médicales, Merck s.a.s., 37 rue Saint Romain, 69008 Lyon, France. Olivier Vandhuick Clinique de l’Europe, 28 Rue Méridienne, 76100 Rouen, France


Introduction
Multiple sclerosis (MS) is a chronic, in ammatory, demyelinating, neurodegenerative disorder of the central nervous system. It is the leading cause of non-traumatic neurological disability in young adults in the USA and Europe. 1 Over 2010-2015, about 110,000 persons had MS in France. 2 MS is classi ed into relapsing or progressive forms, with Relapsing-Remitting MS (RRMS) accounting for 85% of all initial diagnoses. RRMS is characterized by intermittent disease relapses followed by periods of remission. Over time, 80% of patients with RRMS progress to secondary progressive MS, characterized by an accumulation of permanent disability without signi cant relapses or remissions. In some cases, the disease may be a progressive accumulation of disability from onset (primary progressive MS). 3,4 There is yet a group of patients which remains di cult to de ne and for which there is no consensus about their treatment. 5,6 These patients have frequent relapses and/or Magnetic Resonance Imaging (MRI) activity (either when untreated or while on a Disease-Modifying Drug [DMD] therapy) and have what might be termed as a Highly Active RRMS (HA-RRMS). 6,7 Several DMDs have been approved for the treatment of MS over the last 20 years. 8 Although no cure for MS exists, current treatments aim to reduce the frequency and severity of relapses, while preventing disability progression. 9 These therapies, such as interferons β and glatiramer acetate, teri unomide and dimethyl-fumarate are available for MS. At the time of the study, for patients with active disease and an insu cient response to at least one drug, as well as for patients with rapidly evolving severe RRMS, two immunosuppressants called high e cacy DMD (HE-DMD) had approval for use in France. These HE-DMDs, natalizumab and ngolimod, became available in France in 2007 and December 2011, respectively.
Several studies have been conducted on the French National Health Insurance (NHI) database to characterize the whole MS population through several identi cation algorithms. 2,10 However, none of these studies described HA-RRMS patients, whose disease management is more complex and still being debated. 11 Furthermore, most of these studies had a short follow-up period (1 year). Therefore, a nationallevel retrospective cohort study was set up, using NHI data, to describe incident HA-RRMS adult patients living in France, over 2010-2015, and their disease management, and followed them up until 2017.

Study design
We performed a longitudinal retrospective cohort study based on the National Health Data System (SNDS) from the NHI, and speci cally, on data extracted from the SNIIRAM (Système National d'Information Inter-Régimes de l'Assurance Maladie) database.
Patients were included in the study on an index date. The index date was de ned as either the date of the rst reimbursement of HE-DMDs or the date of relapse, during the inclusion period (January 1st, 2010 to December 31st, 2015). Patients were followed until criteria were met (death, one year without any medical care expenses, end of the study (December 31st, 2017)), whichever occurred rst. Hence, patient followup lasted up to seven years. Data was extracted between January 1st, 2008 (to allow for the identi cation of HA-RRMS patients) and December 31st, 2017.

Data sources
The SNIIRAM is brie y described in Supplement 1 and in detail elsewhere. 12 In short, it contains comprehensive, individual-level, anonymized data on all healthcare expenses reimbursed by the French NHI. The NHI covers primary and secondary care expenses, medications, medical devices, and hospital costs 13 (additional information in Supplement 2). The General Scheme, covering 76% of inhabitants, is the largest scheme of the NHI.

Study population
We included adults (≥ 18 years) continuously a liated to the General health insurance Scheme between two years prior to the index date and 2017. According to an algorithm developed with two neurologists, HA-RRMS patients were included when meeting at least one of the following criteria: 1) Treatment criterion: At least one reimbursement of the only two available HE-DMD in France for HA-RRMS (natalizumab, ngolimod); 2) Relapse criterion: At least two relapses in one year and no previous treatment for MS during the previous two years and an MRI of the central nervous system performed during the three months following the 2nd relapse and a DMD during the six months following the MRI; 3) Relapse criterion: One relapse in one year despite a DMD, with an MRI of the central nervous system performed during the three months following the relapse and a switch of DMD during the six months following the MRI.
Relapses were de ned using a validated US claims-based algorithm 14  Relapse events that occurred within the same 30-day period were treated as a single relapse. Exclusion criteria are listed in Supplement 3.
New HA-RRMS patients were de ned as patients without reimbursement for HE-DMDs and/or without MS relapse(s) twelve months prior to the index date. HE-DMD discontinuation was de ned as three months without natalizumab infusion or three months after end of last dispensed ngolimod tablets. Discontinuation for pregnancy was de ned by the occurrence of a delivery within fteen months after treatment discontinuation. Switch of DMD or HE-DMD was de ned as dispensing of the alternative DMD or HE-DMD six months following the date of discontinuation of the initial treatment.

Outcomes of interest
Baseline characteristics of the study population (demographic data, comorbidities, long-term disease (LTD) status, disability pension) and therapeutic management (drugs [Supplement Table 1], hospitalizations [Supplement 4], primary and secondary care, laboratory tests, imaging procedures, transportation, medical devices, and date of care) were collected. Of note, all secondary care visits accounted for were those held in the community and not at the hospital; all hospital practitioners' visits were combined together.

Statistical analyses
Continuous, quantitative variables were summarized with the number of patients (n), mean, and standard deviation (SD). Categorical, qualitative variables were summarized with the percentage of patients per category. Con dence intervals were computed using the normal approximation. The de nition of HA-RRMS prevalence, incidence, mortality and the Charlson Comorbidity Index are in Supplement 5.
Treatment duration was assessed with the Kaplan-Meier method. The SNIIRAM covers almost the entire French population, 12 thus no sample size calculation was necessary.

Ethics
Patient consent was not necessary because this study uses secondary data and protection of patients' rights and freedom were guaranteed by the French data protection authority (CNIL: decision DR-2018-099). Study protocol was approved by the committee for research, studies and evaluations in the eld of health (CEREES) and all methods were performed in accordance with the corresponding guidelines. STROBE cohort reporting guidelines 15 were used. Almost ¾ of incident patients were women (7,067 women, 73.7%) and the mean age at index date was 39.9 years old (SD 10.5, median 39.0) ( Table 1). Most patients had obtained LTD status (9,539 patients, 99.4%), and 45.6% (4,352 patients) of those with an LTD status got it within ve years prior to their index date. Over the study period, 6,646 (69.3%) incident patients had a Charlson Comorbidity Index of 0 and 632 (6.6%) had an index of 3 or above. Of the 9,238 incident patients under the age of 60, 1,835 (19.1%) had a disability pension at index date. Most of them (1,153 patients, 62.8%) had a level 2 disability pension. When considering the rst change in pension level over the study period, 1,863 (20.2%) had an increase and 126 (1.4%) a decrease in pension level. Among the patients with an increase to level 2 or 3 disability pension, the change occurred, on average, after 6.4 years.

Epidemiology
Among adults 20 years old and above, the age-standardized prevalence rate increased from 11.0 HA-RRMS cases per 100,000 in 2010 to 26.5 HA-RRMS cases per 100,000 in 2015 ( Table 2). The agestandardized incidence rate was 3 new cases per 100,000 in 2010. It increased up to 4.5 new cases per 100,000 in 2012, then remained above 3.6 case per 100,000 until 2015. Between 2010 and the end of the study follow-up in 2017, 122 HA-RRMS patients over 20 died (4 patients aged 18 to 20) and the crude mortality rate varied between 113 deaths per 100,000 HA-RRMS patients in 2010 and 389 deaths per 100,000 HA-RRMS patients in 2015 (Table 3). The mortality rate of HA-RRMS patients was signi cantly about twice as high as the mortality rate of the French population in 2011, 2014 and 2015.
Disease management Drugs Within the two years prior to the index date, 7,202 (75.0%) patients had taken a DMD (Table 4). During the study follow-up, 3,809 patients (39.7%) took a DMD; interferon beta was again the most common DMD (22.6% of patients treated) (  (Fig. 1). Patients took HE-DMDs (initiated at index date or later) for, on average, 3.5 years prior to discontinue it. This includes 225 women who stopped HE-DMDs and gave birth; 190 (84.4%) of them restarted an HE-DMD after childbirth. During the study, 32.5% of patients treated with natalizumab at index date switched to ngolimod, after a mean treatment duration of 4.2 years. Conversely, 7.0% of those treated with ngolimod at index date switched to natalizumab, after a mean treatment duration of 3.5 years. About a third of patients (32.8%, 1,926 patients out of 5,865) ever treated with ngolimod during the study discontinued their treatment and 63.5% (2,328 out of 3,664 patients) ever treated with natalizumab discontinued their treatment, after a mean treatment duration of 3.5 and 2.7 years, respectively. Of the patients ever treated with ngolimod, 7.4% switched to a DMD and 4.4% to natalizumab (Table 6). Meanwhile, of the patients ever treated with natalizumab, 10.0% switched to a DMD and 27.9% to ngolimod. During the study, 1,308 of 2,892 patients (45.2%) not treated with an HE-DMD at index date initiated an HE-DMD (on average, 4.4 months after the index date).
Of those with ngolimod as their rst HE-DMD, 102 switched to rituximab and 352 switched to a DMD; of those with natalizumab as their rst HE-DMD, 96 switched to rituximab and 305 switched to a DMD. Taken as a whole, 83.5% (8,012 patients) took an HE-DMD during the study.
The following results are based on the 8,045 patients with at least two years of follow-up. We observed a marked increase in the annual age-standardized prevalence rate of HA-RRMS over the study period, together with an increase in the age-standardized incidence rate starting in 2012. Rather than an increase in disease prevalence, this surge is most likely explained by the launch of ngolimod (used as inclusion criterion) on the French market in 2011. This likely also explains why more patients switched from natalizumab to ngolimod than the opposite and more patients not treated with an HE-DMD at the index date initiated ngolimod than natalizumab.
A previous study, based on the SNIIRAM database and conducted by Foulon et al., 16 estimated that about 98,500 persons older than 20 had MS in France in 2012, corresponding to a national crude MS prevalence of 199 per 100,000. We estimated that about 7,300 persons older than 20 had an HA-RRMS in 2012, corresponding to a crude prevalence of 17 per 100,000. Based on these data, HA-RRMS patients would represent around 8% of MS patients. This proportion is slightly lower than that reported by the French Health Authorities around the same period among MS patients followed-up by physicians (11%). 17 The median age at MS diagnosis was 40 years old in the Roux et al. study. 2 The median age of new HA-RRMS cases was 39 years old in our study (adults only) and the estimated median time to MS diagnosis was about 5 years prior to HA-RRMS. These results could imply that, among patients who experience HA-RRMS, HA-RRMS happens early in the MS disease history and that HA-RRMS patients are diagnosed with MS at a younger age than the overall MS population.
The crude HA-RRMS mortality rate increased over the study period with the highest crude rate of 389 deaths per 100,000 HA-RRMS cases (age 20 and above) in 2015 and an SMR of 2.1 compared to general population. The 2013 crude MS mortality rate was 1,370 deaths per 100,000 MS cases (all ages) with an SMR of 2.6. 16 That year, the HA-RRMS crude mortality rate was 189 deaths per 100,000 HA-RRMS cases (age 20 and above) and the SMR 1.2 (95% CI 0.6-1.7). These results are in favor of a lower mortality among HA-RRMS patients compared to the whole MS patient population. This could be explained by the fact that HA-RRMS are younger than MS patients, or improved disease management when HE-DMD are taken early.
In this study with a mean follow-up of 4.0 years, HE-DMD therapy lasted, on average, for 3.5 years and most but not all patients were treated with an HE-DMD at one point during the study. This highlights the need for additional HE-DMD options, as they should be initiated as early as possible. 7,11 Patients who have not been observed to have taking HE-DMD could be relying on other treatments (DMD, symptomatic treatments, other immunosuppressants, rituximab and mitoxantrone) or perhaps be refusing HE-DMD due to several reasons (not available in the claims database): suboptimal e cacy, tolerability issues, or concern about potential tolerability issue. 9 Another possibility is that, despite using multiple criteria and having been validated by experts, our algorithm also detects a small proportion of MS patients who experience at least one relapse, have an MRI, a DMD and a switch of DMD, and yet do not have an HA-RRMS. Nevertheless, the algorithm successfully identi ed genuine HA-RRMS patients, the therapeutic management of these patients and its evolution with the introduction of ngolimod. Such algorithm is all the more needed than the identi cation of these patients is still being debated, 5,6 earlier management improves patient outcomes, 7 and new drugs are now available.
Finally, our study showed a large spectrum of healthcare professionals involved in the management of HA-RRMS. The high number of hospital care visits and the higher number of visits the rst year of followup among the patients with at least one visit is partly related to drug dispensing. The high number of relapses in our population is at least partly due to our selection criteria (30% of patients were included for having one or more relapses).
This study has several strengths. The rst one is the use of comprehensive data from the French administrative healthcare claims database, speci cally from the General Scheme, covering 76% of the French population, offering improved coverage compared to studies based on the MS French observatory 18 and the MS regional registry in Lorraine. 19 The second one is the absence of patient participation bias because the data are routinely collected for reimbursement purposes and not speci cally for this study. The third one is the longer follow-up compared to other studies which allows accurate capture of disease management, including treatment changes.
This study also has limitations that must be considered when interpreting the results. In the absence of clinical data (such as MS relapses) in the database which would allow to ascertain HA-RRMS patients, we had to develop an algorithm. This algorithm should be used in additional studies to con rm its capacity to capture all HA-RRMS patients and updated with new HE-DMDs. Our results apply to that period only, as new treatments have since been released. Using an anonymous database hinders access to additional information on patient and disease characteristics through patient questionnaires (e.g. the Expended Disability Status Scale used in MS studies). In that respect, linking claims data to MS registry data would further our understanding of the HA-RRMS patient population and their therapeutic management.
This real-world study around the time when HE-DMDs became available offers some valuable insight into the HA-RRMS population. Our estimated proportion of HA-RRMS patient in the whole MS population is similar to that previously reported, thereby bringing consistency to our algorithm. Based on three criteria, our method can contribute to the much needed and still debated de nition of HA-RRMS. Despite the need for a long-term care, some patients never took HE-DMDs, highlighting the need for additional treatment options.

Declarations
Ethics approval and consent to participate

Consent for publication
The SNIIRAM contains comprehensive individualized and anonymous data on all healthcare expenditures reimbursed by the French NHI. Thus, no consent for publication was needed.

Availability of data and material
Data from the National Health data system in France (Système National des Données de Santé, SNDS) are publicly available (https://www.snds.gouv.fr/SNDS/Processus-d-acces-aux-donnees) after acceptance from the French data protection authority (Comité National d'informatique et Liberté, CNIL).
The datasets analysed during the current study are available from co-author Ludovic Lamarsalle (llamarsalle@hevaweb.com) on reasonable request. All data generated or analysed during this study are included in this published article.          Table 8 Multiple sclerosis-related outpatient, hospital practitioner, and other medical professional visits in incident highly active relapsing-remitting multiple sclerosis adults with at least 2 years of follow-up (n = 8,045), 2010-2017 Neurosurgeon 0.0 (0.1) 18 (0.2%) 0.0 (0.1) 83 (1.0%) All secondary care visits accounted for were those held in the community and not at the hospital. All hospital practitioners' visits were combined due to inconsistent reliability of the hospital practitioner specialty in the database. a Excludes visits during hospital stays because they are included in the price of the hospital stay.