Persistence with and adherence to fingolimod compared with other disease-modifying therapies for the treatment of multiple sclerosis: a retrospective US claims database analysis

Abstract

Objective:

Achieving therapeutic goals in multiple sclerosis (MS) requires strict adherence to treatment schedules. This retrospective study analyzed persistence with, and adherence to, fingolimod compared with injectable/infusible disease-modifying therapies (DMTs) in patients with MS.

Methods:

Patients in the PharMetrics Plus™ US administrative claims database with at least one prescription for, or administration of, fingolimod, glatiramer acetate (GA), interferon (IFN), or natalizumab (index DMT) between October 1, 2010 and September 30, 2011 were included. Patients were naïve to index DMT (no claim in the previous 360 days) and had an MS diagnosis code within 360 days of the first index DMT prescription. Outcomes were persistence, risk of discontinuing index DMT (evaluated by a Cox proportional hazards model), adherence (measured using the medication possession ratio [MPR] and proportion of days covered [PDC] in patients with at least two index DMT prescriptions), and the risk of being non-adherent (MPR <80% and PDC <80%, assessed using a logistic regression model).

Results:

The study included 3750 patients (fingolimod, n = 889; GA, n = 1233; any IFN, n = 1341; natalizumab, n = 287). Discontinuation rates (fingolimod, 27.9%; GA, 39.5%; IFN, 43.7%; natalizumab, 39.5%; all p < 0.001) and risk of discontinuation were significantly higher (hazard ratios vs fingolimod [95% confidence interval]: GA, 1.75 [1.49–2.07]; IFN, 2.01 [1.71–2.37]; natalizumab, 1.53 [1.22–1.91]) for patients receiving other DMTs compared with fingolimod. The risk of being non-adherent was also lower for patients in the fingolimod cohort than the other treatment cohorts, irrespective of whether non-adherence was defined as MPR <80% (p < 0.05 for all) or PDC <80% (p < 0.05 for GA and IFN).

Limitations:

As with all studies assessing real-world treatment patterns it is unclear if medications were used as prescribed.

Conclusions:

In a real-world setting, persistence with, and adherence to, oral fingolimod was higher than for injectable and infusible DMTs.

Keywords::

• Multiple sclerosis
• Persistence
• Adherence
• Fingolimod
• Interferon
• Glatiramer acetate
• Natalizumab

Introduction

Multiple sclerosis (MS) is an incurable, chronic, inflammatory, degenerative disease of the central nervous system (CNS)1,2. It is the leading cause of disability in young and middle-aged people in the developed world3, affecting ∼2.1 million people worldwide and 400 000 people in the US4. Most patients (80–85%) present with the relapsing–remitting form of MS, which is characterized by self-limiting episodes of neurological dysfunction (relapses) followed by periods of remission and recovery5. The course of MS is highly unpredictable but, as endogenous repair mechanisms fail to restore damaged CNS tissue6, the disease progressively worsens to result in irreversible disability and premature death7,8.

Current treatments for MS aim to prevent relapses, slow disease progression, and postpone long-term disability. Glatiramer acetate (GA) and interferons (IFNs) are recommended as first-line disease-modifying therapies (DMTs) for MS9–11, but these injectable therapies are increasingly perceived to have only partial effectiveness and a sub-optimal tolerability profile, mainly driven by influenza-like symptoms or injection site reactions12. Several new therapies with different mechanisms of action and modes of administration have been introduced in recent years12,13. These include natalizumab, a recombinant anti-alpha 4 monoclonal antibody that is administered by monthly infusions14,15, and fingolimod, a sphingosine 1-phosphate receptor modulator, which is the first oral therapy approved for the treatment of relapsing forms of MS16.

Achieving therapeutic goals requires prompt treatment initiation and stringent adherence to both the medication dose and administration schedule17–20. Adherence is a measure of the extent to which patients take medication in accordance with the prescribed regimen21. Real-world insight into medication-taking behaviors can be gained from analyses of medical claims data. Claims databases also provide valuable information on persistence, which provides a complementary assessment of medication use, measuring the duration of continuous use rather than the proportion of days that a patient took the prescribed medication21.

Several studies have shown that adherence to, and persistence with, DMTs are associated with a reduced risk of relapse22,23, reduced healthcare resource utilization22, reduced incidence of MS-related hospitalization19, and improved health-related quality-of-life24,25 in patients with MS. First-line injectable DMTs are associated with sub-optimal rates of adherence and persistence20,22,26–29, and treatment-related issues, such as unresponsiveness and intolerance to GA or IFNs, prompt patients to discontinue or switch therapy30,31. A review of DMT discontinuation rates across several countries found that 16–27% of patients discontinued GA or IFN therapy prematurely within 24 months of follow-up, increasing to 43% for GA and up to 34% for IFNs when patients were followed for longer than 24 months32. The most common reasons given for IFN discontinuation were side-effects and lack of effectiveness32. A US claims database analysis reported discontinuation rates of 43.8% for GA and 44.8–53.9% for IFNs for treatment-naïve patients within 12 months of treatment initiation33. A recent retrospective claims database analysis of nearly 7000 patients with MS in the US reported that only 55% and 52–62% of patients were adherent (measured as a medication possession ratio [MPR] ≥80%) to GA and IFNs, respectively, over 12–36 months27. Another study that included only IFNs reported that 27–41% of patients were adherent (MPR ≥85%) in each year of the study, with only 4% of patients being continuously adherent throughout the 3-year study22. Natalizumab is administered as an intravenous infusion once monthly and has improved adherence compared with the injectable DMTs, with one study reporting a mean adherence rate (measured using MPR) of 85% compared with 73–78% for GA and IFNs26. The authors of that study suggested that this improved adherence was probably due to the requirement for active physician involvement and monitoring during natalizumab administration26.

Effective management of MS is an ongoing process in which factors that support patient persistence and adherence are important. Phase 3, randomized, controlled trials have shown that fingolimod is more effective than IFN or placebo at reducing relapses, slowing disability progression and improving magnetic resonance imaging (MRI) and brain atrophy outcomes34,35. In addition, the favorable tolerability profile36,37 and once-daily oral administration of fingolimod may result in improved medication-taking behaviors compared with injectable/infusible DMTs. A recent claims database analysis demonstrated that patients with MS initiating fingolimod therapy had higher rates of persistence (discussed above) and adherence than patients using GA and IFNs (MPR ≥80% in treatment-naïve patients; fingolimod, 87.4%; GA, 81.5%; IFNs, 72.8–80.2%; all p < 0.05)33. However, that study did not include patients treated with natalizumab.

The objective of this retrospective cohort analysis of a US health insurance claims database was to assess persistence and adherence rates among patients with MS treated with once-daily, oral fingolimod in comparison with injectable (GA and IFNs) and infusible (natalizumab) DMTs in a real-world setting.

Patients and methods

Data source

Adjudicated medical and pharmacy claims data from the PharMetrics Plus™ claims database were used in this retrospective cohort analysis. More than 87 million health-plan members across the US have been enrolled in the database from 2006 onwards, providing longitudinal data on ∼22 million patients with at least 4 years of continuous enrollment in their health plan. The database has broad geographical coverage, with patients in each three-digit zip code in the US, and is representative of the US commercially insured population. It includes data from 80% of all US doctors and 90% of US hospitals, and representation from 85% of the Fortune 100 companies. By including integrated claims data (i.e. data from multiple places of service [inpatient, outpatient and pharmacy]) for over 100 000 patients with MS in 2011, the database is also considered to be generally representative of the population of patients with MS in the US.

Inpatient and outpatient diagnoses were recorded using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes; other data recorded include outpatient and inpatient procedures, detailed information on pharmacy and medical benefits (co-payment/co-insurance amount, deductible, in-network vs out-of-network), inpatient stay (admitting vs other diagnoses, admission type and source, discharge status) and provider details (specialty, zip code, attending, referring, rendering, prescribing, primary care provider), and retail and mail-order prescription records. ICD-9-CM diagnosis and procedural codes were available for each hospital stay, although names and descriptions of specific medications during inpatient stays were not available. Demographic variables (age, gender, and geographical region), insurance product type (e.g. health maintenance and preferred provider organizations), payer type (e.g. commercial, self-pay) and start and stop dates of health-plan enrollment were additional data elements considered. Amounts charged by providers, and amounts allowed and paid by health plans, were available for all services rendered, and included the dates of all services. The names and descriptions of specific medications used during hospital stays were not available, however, making it impossible to identify DMTs used during an inpatient stay. DMTs administered in an outpatient clinic or infusion center could be identified when an overnight hospital stay was not required. Dates of service for all claims, and fees charged by providers (and amounts allowed and paid by health plans) were available for all services. Other recorded fields included demographic data (patient age, gender, and geographical region), insurance product type (e.g. health maintenance and preferred provider organizations), payer type (e.g. commercial, self-pay), and start and stop dates of health-plan membership.

The database is fully compliant with the Health Insurance Portability and Accountability Act (HIPAA) privacy regulations. Patient informed consent and Institutional Review Board approval were not required for this study because all patient-level data are certified to be anonymous and no direct contact with the patient was required.

The study was designed, implemented, and reported in accordance with the ethical principles laid down in the Declaration of Helsinki, the Guidelines for Good Pharmacoepidemiology Practices (GPP) of the International Society for Pharmacoepidemiology, and the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines38.

Patient selection

Patients who had at least one prescription or administrative claim for fingolimod (Gilenya® ), GA (Copaxone®^† ), IFN therapy (IFN beta-1a [intramuscular Avonex®^‡ or subcutaneous Rebif®^§ ] or IFN beta-1b [Extavia®^¶ and Betaseron®** , both administered subcutaneously]) or natalizumab (Tysabri®^†† ) between October 1, 2010 and September 30, 2011 (index period) were identified in the database. National Drug Codes used are listed in Supplementary Table 1 and procedural codes for administering DMTs in the clinical setting are listed in Supplementary Table 2. The first prescribed DMT in this period was defined as the index DMT, and the date on which the index DMT was first prescribed was defined as the index date. Patients with evidence of both fingolimod and another DMT were preferentially assigned to the fingolimod cohort to preserve the sample size, while the other treatment cohorts were assigned based on the first observed DMT. Medical and pharmacy records for eligible patients were then studied for 360 days following the index date.

For inclusion, patients needed to have continuous health-plan enrollment for a minimum of 360 days before and after the index date (the pre- and post-index periods, respectively); to have at least one claim with an MS diagnosis (ICD-9-CM code 340) within 360 days of the index date (pre- or post-index); and to be aged 18 years or older on the index date. Patients were excluded from the analysis if they had received more than one DMT on the index date, had a claim for the index DMT in the previous 360 days (i.e. patients must have been naïve to index DMT), had data quality issues (e.g. invalid enrollment date, missing or invalid age or missing gender; or coverage information) or had a gap in the claims data indicative of missing days of supply information for the index therapy (Figure 1). Days’ supply is not reported on claims for physician-administered injectable/infusible treatments so, in order to measure persistence, days’ supply was derived from the prescribing information for each treatment. Administration of once-daily GA equalled 1 days’ supply; administration of 4-weekly natalizumab equalled 28 days’ supply; administration of once-weekly Avonex® equalled 7 days’ supply; administration of Betaseron® and Extavia® (prescribed for use every other day) equalled 2 days’ supply; and Rebif® administered 3-times weekly equalled 2.3 days’ supply.

Figure 1. Attrition of the study sample, by reason.

Study measures

The primary measures of interest in this study were persistence and adherence rates in the fingolimod, GA, IFN, and natalizumab cohorts. Additional analyses were performed for the patient cohorts receiving IFN beta-1a and IFN beta-1b. Persistence was measured, based on treatment practice patterns, as the time in consecutive days from index DMT monotherapy initiation until discontinuation, receipt of another DMT of interest (fingolimod, GA, IFN, or natalizumab), or the end of available data (360 days post-index), whichever occurred first. Discontinuation was defined as a gap in exposure to the index DMT of at least 60 days following the date when the index DMT should next have been dispensed or administered.

The proportion of patients persistent with therapy, total number of therapy claims, and total number of persistence days were assessed for all patients and for those patients with at least two claims for the index DMT. Adherence to index DMTs was assessed over the post-index persistence period in the sub-set of patients with at least two claims for their index DMT. It was assessed using two measures.

• MPR was defined as the total number of days supply of the index DMT during the persistence period divided by the number of days between the first prescription fill and the last refill plus the days supply of the last refill.

• Proportion of days covered (PDC) was defined as the number of days on the index DMT during the persistence period divided by the number of days that a patient was persistent with the index DMT. Compared with the fixed-interval MPR, the PDC provides a more conservative estimate, as it counts only the number of days with index therapy. If a patient refills a prescription early, any days with double counting or overlapping from the early refill would be counted only once.

For each index DMT, MPR and PDC were presented as a point estimate and were provided for those patients with at least two claims for their index DMT. Patients with an MPR or PDC of at least 80% were considered adherent and those with an estimated MPR or PDC of less than 80% were considered non-adherent.

Other data recorded included patient demographics and baseline characteristics, which were evaluated during the pre-index period. These included age at index date, gender, previous use of non-index DMT, use of dalfampridine, Charlson comorbidity index score39,40, comorbidities of interest (e.g. dyslipidemia, depression), evidence of any MS relapse, and number of MS relapses.

Statistical analyses

For categorical measures, data are presented as counts and proportions. Continuous variables were summarized by providing the mean, 95% confidence interval (CI), standard deviation (SD) and median. Differences in the distribution of these variables were tested for statistical significance using χ² tests for categorical variables and the Wilcoxon rank-sum test for continuous variables. The risk of discontinuing treatment with the index DMT was estimated using a Cox proportional hazards model, adjusting for differences in pre-index demographic and clinical characteristics. Variables included in the final model were age group, gender, payer type, physician/prescriber specialty, pre-index total healthcare cost (as a logarithmic continuous measure), pre-index number of relapses (two or more vs one), pre-index use of dalfampridine, and comorbidities and symptoms present in at least 5% of patients and those thought to be associated with the dependent variable. Other demographic and clinical characteristics were originally included, but later removed from the model due to multi-collinearity. The risk of being non-adherent to the index DMT during the persistence period was estimated in patients with at least two claims during the persistence period with a logistic regression model using a similar approach. The dependent variable was non-adherence to DMT, with MPR/PDC <80% = 1 and MPR/PDC ≥80% = 0.

Results

Study attrition

In total, 39 552 patients were initially identified, of whom 35 802 (90.5%) were excluded for the reasons listed in Figure 1. Of the 3750 eligible patients, 889 (23.7%) received fingolimod as their index DMT, 1233 (32.9%) received GA, 1341 (35.8%) received an IFN (Avonex®, n = 527; Rebif®, n = 551; Betaseron®, n = 247; Extavia®, n = 16) and 287 (7.7%) received natalizumab.

Baseline demographics and clinical characteristics

The baseline demographics and clinical characteristics of the fingolimod, GA, IFN and natalizumab cohorts are shown in Table 1. Over three-quarters (77.0%) of the patients included in the study were women. Patients in the fingolimod cohort were significantly older (all p < 0.01) and were more likely to have previously used DMTs (all p < 0.001) than patients in the other DMT cohorts. In addition, patients treated with fingolimod had lower Charlson comorbidity index scores than patients treated with the other DMTs.

Table 1. Baseline demographics and clinical characteristics (in the 360-day pre-index period) of patients with MS included in the index cohorts.

Characteristic	Fingolimod (n = 889)	GA (n = 1233)	IFN (n = 1341)	Natalizumab (n = 287)
Age, years
Mean (SD)	46.0 (9.8)	44.0 (10.8)	44.1 (10.7)	44.0 (10.0)
Median	47.0	44.0***	45.0***	44.0**
Female, n (%)	692 (77.8)	965 (78.3)	1029 (76.7)	201 (70.0)**
Previous use of non-index DMT, n (%)^†	628 (70.6)	59 (4.8)***	106 (7.9)***	78 (27.2)***
GA	186 (20.9)	0 (0.0)	64 (4.8)	26 (9.1)
IFN	366 (41.2)	52 (4.2)	40 (3.0)	57 (19.9)
Natalizumab	117 (13.2)	7 (0.6)	5 (0.4)	0 (0.0)
Previous use of dalfampridine, n (%)	130 (14.6)	24 (1.9)***	26 (1.9)***	34 (11.8)
Charlson comorbidity index score, mean (SD)	0.50 (0.89)	0.68 (1.22)	0.58 (1.04)	0.55 (1.24)
Symptoms affecting ≥10% of patients, n (%)
Fatigue	270 (30.4)	404 (32.8)	418 (31.2)	113 (39.4)**
Headache	206 (23.2)	342 (27.7)*	314 (23.4)	58 (20.2)
Walking, balance and co-ordination problems	189 (21.3)	291 (23.6)	272 (20.3)	63 (22.0)
Numbness	165 (18.6)	502 (40.7)***	479 (35.7)***	60 (20.9)
Visual symptoms	157 (17.7)	232 (18.8)	217 (16.2)	44 (15.3)
Bladder dysfunction	128 (14.4)	101 (8.2)***	97 (7.2)***	38 (13.2)
Muscle weakness/spasm/spasticity	126 (14.2)	201 (16.3)	202 (15.1)	48 (16.7)
Respiration/breathing problems	107 (12.0)	108 (8.8)*	121 (9.0)*	36 (12.5)
Dizziness and vertigo	94 (10.6)	205 (16.6)***	222 (16.6)***	36 (12.5)
Bowel dysfunction	89 (10.0)	122 (9.9)	127 (9.5)	31 (10.8)
Co-morbidities affecting ≥10% of patients, n (%)
Dyslipidemia	246 (27.7)	278 (22.5)**	306 (22.8)**	69 (24.0)
Depression	184 (20.7)	229 (18.6)	194 (14.5)***	67 (23.3)
History of CVD	49 (5.5)	154 (12.5)***	168 (12.5)***	23 (8.0)
Patients experiencing a relapse in the pre-index period, n (%)	368 (41.4)	395 (32.0)***	427 (31.8)***	130 (45.3)
Outpatient relapse^‡	348 (94.6)	314 (79.5)***	357 (83.6)***	112 (86.2)**
Inpatient relapse^‡	37 (10.1)	89 (22.5)***	81 (19.0)***	26 (20.0)**
Number of pre-index relapses, mean (SD)	0.67 (1.01)	0.39 (0.71)	0.39 (0.67)	0.74 (1.03)
Patients experiencing relapses in the pre-index period, n (%)
0 relapses	521 (58.6)	838 (68.0)***	914 (68.2)***	157 (54.7)
1 relapse	225 (25.3)	343 (27.8)	365 (27.2)	81 (28.2)
2 relapses	89 (10.0)	35 (2.8)	42 (3.1)	27 (9.4)
≥3 relapses	54 (6.1)	17 (1.4)	20 (1.5)	22 (7.7)
Total healthcare cost, US$
Mean (SD)	$33 288 ($24 837)	$16 871 ($28 327)	$14 128 ($17 812)	$26 417 ($27 691)
Median	$31 450	$10 902***	$9413***	$20 167***

*p < 0.05, **p < 0.01, ***p < 0.001 vs fingolimod.

^†Percentages are based on the total number of patients in each cohort.

^‡Among those patients who had a relapse; percentages do not add up to 100%, as some patients experienced both inpatient and outpatient visits.

CVD, cardiovascular disease; DMT, disease-modifying therapy; GA, glatiramer acetate; IFN, interferon; MS, multiple sclerosis; SD, standard deviation.

A history of dyslipidaemia was significantly more common in the fingolimod cohort than in the GA and IFN cohorts, but a significantly lower proportion of patients in the fingolimod cohort had a history of CVD compared with patients in the GA and IFN cohorts. A history of depression was more common in the fingolimod and natalizumab cohorts than in the GA and IFN cohorts. During the pre-index period, significantly more patients in the fingolimod and natalizumab cohorts experienced at least one relapse compared with patients in the GA and IFN cohorts (fingolimod, 41.4%; GA, 32.0%; IFN, 31.8%; natalizumab, 45.3%). The mean number of relapses in the pre-index period was highest in the natalizumab cohort (0.74), followed by 0.67 in the fingolimod cohort and 0.39 in both the GA and IFN cohorts.

Discontinuation of index DMT

Of the 3750 patients included in the study, 38.2% discontinued their index therapy during the post-index period. The proportion of patients who discontinued their index therapy was significantly lower in the fingolimod cohort (27.9%) than in the other DMT cohorts (GA, 39.5%; IFN, 43.7%; natalizumab, 39.0%; all p < 0.001; Figure 2). Similar observations were made for the individual IFN sub-cohorts, with 44.2% of patients in the IFN beta-1a sub-cohort and 41.8% of patients in the IFN beta-1b sub-cohort discontinuing therapy (all p < 0.0001 vs fingolimod).

Figure 2. Proportion of patients with MS who discontinue index DMT within 360 days of initiating therapy. The risk of discontinuing treatment with the index DMT was estimated using a Cox proportional hazards model, adjusting for pre-index demographic and clinical characteristics of interest. DMT, disease-modifying therapy; GA, glatiramer acetate; IFN, interferon; MS, multiple sclerosis.

After controlling for baseline characteristics, the risk of discontinuation from index DMT was 1.5–2-fold higher in patients taking GA, IFN, or natalizumab than in those taking fingolimod (hazard ratios vs fingolimod [95% CI]: GA, 1.75 [1.49–2.07]; IFN, 2.01 [1.71–2.37]; natalizumab, 1.53 [1.22–1.91]; Figure 2). The adjusted estimates for the risk of discontinuing are very similar to the unadjusted estimates (Table 2). Additional multivariate analyses assessing the impact of independent variables on the risk of discontinuation showed that the hazard ratio is influenced by gender, number of pre-index relapses, pre-index total healthcare costs, and several symptoms (numbness, visual symptoms, bowel dysfunction and pain; Table 2).

Table 2. Cox proportional hazards model of variables affecting risk of discontinuing index disease-modifying therapy in patients with MS.

	Hazard ratio	95% CI	p Value
Unadjusted estimates:
GA vs fingolimod	1.515	1.300–1.766	<0.0001
IFN vs fingolimod	1.748	1.506–2.028	<0.0001
Natalizumab vs fingolimod	1.508	1.206–1.885	0.0003
Adjusted estimates:
GA vs fingolimod	1.753	1.487–2.067	<0.0001
IFN vs fingolimod	2.013	1.713–2.366	<0.0001
Natalizumab vs fingolimod	1.526	1.218–1.913	0.0002
Independent variables:
Age group (years)
35–44 vs 18–34	0.918	0.790–1.065	0.2587
45–54 vs 18–34	0.890	0.765–1.036	0.1334
55–64 vs 18–34	0.894	0.746–1.072	0.2270
≥65 vs 18–34	0.944	0.610–1.463	0.7978
Gender: female vs male	1.168	1.025–1.330	0.0197
Payer type: non-commercial vs commercial	1.049	0.943–1.168	0.3757
Prescriber physician specialty:
General/family practice vs neurology	0.967	0.748–1.252	0.8012
Internal medicine vs neurology	0.878	0.664–1.160	0.3589
Other vs neurology	1.011	0.894–1.143	0.8637
Unknown vs neurology	0.961	0.826–1.118	0.6054
Pre-index total healthcare costs (log)	1.037	1.003–1.071	0.0307
Number of pre-index relapses:
1 vs 0	0.992	0.876–1.123	0.8994
≥2 vs 0	1.233	1.010–1.504	0.0391
Pre-index use of dalfampridine: yes vs no	1.129	0.885–1.441	0.3279
Symptoms
Fatigue: yes vs no	1.084	0.964–1.219	0.1792
Headache: yes vs no	1.034	0.911–1.173	0.6084
Walking, balance, co-ordination problems: yes vs no	0.902	0.788–1.032	0.1326
Numbness: yes vs no	0.746	0.661–0.841	<0.0001
Visual symptoms: yes vs no	0.777	0.671–0.900	0.0007
Bladder dysfunction: yes vs no	0.900	0.745–1.087	0.2737
Respiration/breathing problems: yes vs no	1.111	0.935–1.319	0.2323
Dizziness and vertigo: yes vs no	0.995	0.856–1.156	0.9450
Bowel dysfunction: yes vs no	1.197	1.012–1.416	0.0357
Pain: yes vs no	1.322	1.083–1.614	0.0061
Co-morbidities
Dyslipidemia: yes vs no	0.982	0.861–1.120	0.7858
Tobacco use (including disorder): yes vs no	0.931	0.771–1.125	0.4607

CI, confidence interval; GA, glatiramer acetate; IFN, interferon; MS, multiple sclerosis.

Non-adherence to index DMT

Adherence to index DMT was assessed over the post-index persistence period in the sub-set of patients with at least two fills/administrations of their index DMT; the mean persistence period for these patients was longer than that for the full study population (fingolimod, 318 [SD = 88 days]; GA, 301 [SD = 99 days]; IFN, 290 [SD = 105 days]; natalizumab, 288 [SD = 110 days]). Using an MPR of less than 80% as a measure of non-adherence, a lower proportion of patients receiving fingolimod were non-adherent over the post-index persistence period than patients receiving other DMTs (Figure 3a). When the individual IFNs were considered as separate sub-cohorts, 11.2% in the IFN beta-1a sub-cohort and 14.5% of patients in the IFN beta-1b sub-cohort had an MPR of less than 80%. In multivariate adjusted analyses, the risk of being non-adherent (MPR <80%) to GA, IFN, or natalizumab was 1.9–2.3-fold higher than fingolimod (odds ratio vs fingolimod [95% CI]: GA, 2.30 [1.60–3.30], p < 0.0001; IFN, 2.12 [1.48–3.04], p < 0.0001; natalizumab, 1.88 [1.16–3.05], p = 0.0103). Similar results were observed when non-adherence was defined as a PDC of less than 80%. The proportion of non-adherent patients was lower in the fingolimod cohort than in the other treatment cohorts (Figure 3b). When the individual IFNs were considered as separate sub-cohorts, 14.2% in the IFN beta-1a sub-cohort and 17.6% of patients in the IFN beta-1b sub-cohort had a PDC of less than 80%. Furthermore, after adjusting for baseline differences patients receiving GA or IFN had a 1.4–1.8-fold increased risk of being non-adherent (PDC <80%) over the 360-day follow-up than patients receiving fingolimod (odds ratio vs fingolimod [95% CI]: GA, 1.76 [1.30–2.38], p = 0.0002; IFN, 1.52 [1.13–2.06], p = 0.0060; natalizumab, 1.36 [0.89–2.08], p = 0.1534). The adjusted estimates for the risk of non-adherence are very similar to the unadjusted estimates (Table 3).

Figure 3. Proportion of non-adherent patients with MS, measured as MPR <80% (a) or PDC <80% (b) during the post-index persistence period and the risk of being adherent to therapy for each cohort over the 360-day follow-up. The risk of being non-adherent to index DMT was estimated in patients with at least two claims during the persistence period using a logistic regression model, adjusting for pre-index demographic and clinical characteristics of interest. GA, glatiramer acetate; IFN, interferon; MPR, medication possession ratio; MS, multiple sclerosis; PDC, proportion of days covered.

Table 3. Regression analysis of variables affecting odds of being non-adherent (MPR <80%, PDC <80%) to index DMT in patients with MS with two or more claims for their index DMT.

	MPR <80% vs MPR ≥80%			PDC <80% vs PDC ≥80%
	Odds ratio	95% CI	p Value	Odds ratio	95% CI	p Value
Unadjusted estimates:
GA vs fingolimod	2.034	1.451–2.853	<0.0001	1.654	1.252–2.186	0.0004
IFN vs fingolimod	2.022	1.447–2.826	<0.0001	1.525	1.155–2.013	0.0029
Natalizumab vs fingolimod	1.917	1.194–3.080	0.0071	1.416	0.936–2.143	0.0998
Adjusted estimates:
GA vs fingolimod	2.301	1.604–3.300	<0.0001	1.759	1.303–2.375	0.0002
IFN vs fingolimod	2.120	1.481–3.036	<0.0001	1.524	1.128–2.059	0.0060
Natalizumab vs fingolimod	1.882	1.161–3.051	0.0103	1.361	0.892–2.077	0.1534
Independent variables:
Age group (years)
35–44 vs 18–34	0.835	0.611–1.140	0.2561	0.948	0.717–1.253	0.7051
45–54 vs 18–34	0.666	0.480–0.924	0.0151	0.737	0.550–0.988	0.0413
55–64 vs 18–34	0.646	0.434–0.962	0.0316	0.707	0.496–1.008	0.0551
≥65 vs 18–34	0.938	0.369–2.382	0.8927	1.186	0.532–2.645	0.6759
Gender: female vs male	1.409	1.053–1.884	0.0209	1.289	1.004–1.655	0.0462
Payer type: non-commercial vs commercial	0.989	0.783–1.251	0.9296	1.043	0.849–1.281	0.6872
Pre-index total healthcare costs (log)	0.965	0.913–1.019	0.1986	0.948	0.902–0.995	0.0320
Prescriber physician specialty:
General/family practice vs neurology	0.815	0.454–1.463	0.4933	0.752	0.446–1.265	0.2827
Internal medicine vs neurology	0.682	0.355–1.310	0.2507	0.621	0.345–1.115	0.1108
Other vs neurology	0.904	0.690–1.184	0.4633	0.890	0.703–1.127	0.3318
Unknown vs neurology	0.831	0.598–1.154	0.2686	0.805	0.602–1.078	0.1454
Number of pre-index relapses:
1 vs 0	0.952	0.720–1.258	0.7287	1.103	0.796–1.290	0.9141
≥2 vs 0	1.303	0.826–2.055	0.2551	1.356	0.918–2.002	0.1259
Pre-index use of dalfampridine: yes vs no	0.597	0.304–1.176	0.1359	0.505	0.278–0.918	0.0250
Symptoms
Fatigue: yes vs no	0.964	0.737–1.260	0.7871	0.973	0.771–1.229	0.8201
Headache: yes vs no	0.717	0.531–0.968	0.0301	0.774	0.598–1.001	0.0512
Walking, balance, co-ordination problems:  yes vs no	0.819	0.597–1.124	0.2159	1.111	0.857–1.441	0.4261
Numbness: yes vs no	0.478	0.360–0.635	<0.0001	0.516	0.404–0.659	<0.0001
Visual symptoms: yes vs no	0.835	0.623–1.167	0.3208	0.866	0.658–1.139	0.3021
Bladder dysfunction: yes vs no	1.163	0.770–1.756	0.4734	1.131	0.790–1.619	0.5002
Respiration/breathing problems: yes vs no	0.979	0.644–1.488	0.9215	1.061	0.745–1.512	0.7437
Dizziness and vertigo: yes vs no	0.811	0.565–1.164	0.2553	0.919	0.680–1.242	0.5811
Bowel dysfunction: yes vs no	1.103	0.742–1.640	0.6272	0.860	0.595–1.243	0.4212
Pain: yes vs no	1.381	0.868–2.198	0.1727	1.428	0.951–2.143	0.0860
Co-morbidities
Dyslipidemia: yes vs no	0.910	0.673–1.230	0.5397	0.936	0.772–1.214	0.6193
Tobacco use (including disorder): yes vs no	0.952	0.619–1.465	0.8238	0.956	0.658–1.388	0.8113

CI, confidence interval; DMT, disease-modifying therapy; GA, glatiramer acetate; IFN, interferon; MPR; medication possession ratio; MS, multiple sclerosis; PDC, proportion of days covered.

Multivariate adjusted analyses assessing the impact of independent variables on non-adherence to therapy showed that predictors of non-adherence were different to predictors for the risk of discontinuation. Odds ratios for the risk of being non-adherent (common to both MPR and PDC) were age, gender, and numbness (Table 3). Significant predictors also differed between the two measures of adherence, with age, gender, headache, and numbness influencing adherence measured by MPR, and age, gender, pre-index healthcare costs, pre-index use of dalfampridine, and numbness influencing adherence measured by PDC.

Discussion

This retrospective cohort database analysis showed that rates of persistence and adherence in a real-world setting were significantly higher for patients with MS treated with fingolimod than for those treated with injectable/infusible DMTs. In this US managed care population, the risk of discontinuing treatment was lower with fingolimod than with GA, IFNs, and natalizumab. Adherence outcomes, as assessed by both MPR and PDC, were also significantly improved for the fingolimod cohort compared with the cohorts for the other DMTs.

By reporting on both persistence and adherence, this study provides a comprehensive assessment of DMT-taking behavior in a real-world setting. The higher rates of persistence and adherence observed in patients receiving fingolimod than in those receiving injectable/infusible DMTs in our study may reflect the favorable tolerability profile of fingolimod36,37 and the convenience of its once-daily oral administration. In addition, the superior efficacy of fingolimod over IFN beta-1a i.m. with regard to reducing relapses, and improving MRI and brain atrophy outcomes, as demonstrated in the phase 3, randomized, controlled TRANSFORMS study34, may contribute to improved rates of medication use.

The main reasons for discontinuing IFN and GA treatment are a perceived lack of efficacy30 and the occurrence of adverse events30,32. Discontinuations due to these reasons can lead to a switch in DMT. Switches can result in increased non-pharmacy costs41 and may be associated with a break in disease control, as some DMTs take time to become fully effective; for example, GA may take to up 9 months to reach its maximum effectiveness42. However, switches can also be beneficial to patients, particularly if they have switched due to experiencing breakthrough disease on injectable therapies to a more efficacious therapy. A switch to a more efficacious therapy may promote adherence, which has been shown to be associated with improved clinical outcomes19, reduced healthcare resource utilization22, reduced incidence of MS-related hospitalization19, and improved health-related quality-of-life24,25. Tan et al.19 found that patients who were adherent (measured as an MPR of ≥0.80) had significantly lower odds of relapse than non-adherent patients (odds ratio = 0.71; 95% CI = 0.59–0.85). These findings suggest that treatment with highly effective and tolerable DMTs that promote patient adherence, such as fingolimod, may result in better clinical outcomes.

Only one other US claims database study, by Agashivala et al.33, has compared discontinuation and adherence rates for fingolimod with those for the first-line injectable DMTs. However, that study did not include patients treated with natalizumab. Agashivala et al. reported that, in patients who had previously used DMTs, those treated with fingolimod had lower rates of discontinuation than those treated with injectable DMTs (fingolimod, 25.7%; GA, 37.4%; IFNs, 45.9–57.1%; p < 0.05). These rates are comparable to those in our study, which also showed that significantly fewer patients in the fingolimod cohort discontinued therapy than in the other DMT cohorts (fingolimod, 27.9%; GA, 39.5%; IFN, 43.7%; natalizumab, 39.0%; all p < 0.001). Agashivala et al. also assessed rates of adherence using both MPR and PDC, and reported that, among patients who had previously used DMTs, those treated with fingolimod had higher rates of adherence than patients treated with injectable DMTs. Rates of adherence using MPR in the Agashivala et al. study are generally in agreement with those observed in our study, although rates using a PDC of at least 80% to identify adherent patients are lower than those in our study (fingolimod, 73.7% vs 89.7%, respectively). This could be explained by the differences in calculating PDC between the two studies; in our study, PDC was calculated using the number of days that a patient was persistent to the index DMT, while Agashivala et al. calculated PDC over a specified time interval, regardless of whether the patient was persistent.

The rates of adherence reported for GA and IFN in our study are higher than those reported in previous studies that used an MPR of at least 80% as the definition of adherence (88% for both GA and IFN). Halpern et al.26 reported lower rates of adherence for GA and IFN (52.2–62.3% of patients) and Tan et al.19 reported a combined rate of 59.6%; however, both of these studies required patients to have only one claim for the index DMT (not at least two claims as in our study), and both had substantially longer follow-up periods than in our study (1–3 years and 4 years, respectively). Another 4-year study reported adherence rates of 52.2–62.3% for GA and IFNs27. Similar rates were also reported in a review that included 24 studies of adherence to DMTs, with weighted mean adherence rates for GA and IFNs ranging between 56.8% and 69.4%43. This review also noted that the highest rates of adherence were seen with the DMTs that required least frequent administration43, indicating that patients may be averse to frequent injections and that DMTs with a convenient mode of administration and dosing schedule may improve patient adherence.

To our knowledge, no other retrospective claims database analyses have compared rates of discontinuation and non-adherence in patients treated with fingolimod and natalizumab. Fingolimod was associated with lower rates of discontinuation and non-adherence than natalizumab during the post-index period in our study. While the reasons for changes in medication use are not recorded in the database, the difference in these rates for fingolimod and natalizumab may reflect the convenience of oral administration or potentially the limitations of long-term natalizumab treatment due to increasing risk of progressive multi-focal leukoencephalopathy (PML) over time44. In addition, there may be a stronger incentive for patients to stay on fingolimod compared with natalizumab owing to the requirement for cardiovascular monitoring for at least 6 h after the first dose of fingolimod and the need for this to be repeated with any lapse in therapy of over 2 weeks45. Our study has demonstrated using multivariate adjusted analyses that the risk of discontinuation or non-adherence is influenced by age, gender, some clinical characteristics, and several symptoms. It is possible that other factors not accounted for in this study, such as formulary management and out-of-pocket expenses, could also influence persistence with and adherence to DMTs.

Our results for adherence with natalizumab are slightly higher than those reported by Halpern et al.26 in their retrospective claims analysis comparing persistence with, and adherence to, second-line DMTs. Our study found that rates of persistence and adherence with natalizumab were similar to those observed with IFN and GA, whereas Halpern et al. showed that, when used as second-line treatments, rates are significantly improved for natalizumab compared with the injectable DMTs.26 These differences may be explained by subtle differences in the calculation of persistence between the two studies.

This retrospective database analysis has both strengths and limitations relating to the study design. A major strength was that data were derived from a large US administrative health-plan database, which comprises adjudicated claims for more than 60 million active patients across the US, including over 100 000 patients with MS, and is considered to be representative of the US national, commercially insured population. These data provide an excellent resource for assessing treatment patterns and outcomes in a real-world setting.

Most of the limitations of this database analysis are inherent in all retrospective claims database analyses. In particular, results are based on pharmacy and physician outpatient claims and do not provide information on whether medications were used as prescribed. There is the potential for the miscoding of diagnosis codes by clinicians and administrators, and chart review and verification of data were not possible. However, the likelihood of including non-MS patients in this study is minimized owing to the requirement for both a diagnosis of MS and a prescription for/administration of a DMT.

To preserve the sample size of the fingolimod cohort, patients with claims for both fingolimod and another DMT during the index period were preferentially assigned to the fingolimod cohort. However, as most of the preferentially assigned patients would have been excluded from the GA and IFN cohorts owing to not being newly treated with their index DMT or not having pre-index continuous enrollment, this assignment does not significantly affect the size of cohorts for the other DMTs. In addition, there was an absence of clinical measures, such as MS severity and the risk of disease progression, which were not readily available in the claims database. Such end-points would have provided additional detail and insight into treatment patterns reported in this study.

Finally, the analytical focus was on patients who met continuous enrollment criteria (12 months pre- and post-index therapy), which may have excluded patients with different treatment patterns incompatible with enrollment criteria. However, this type of continuous enrollment restriction was necessary to ensure that full details of patients’ treatment were captured in the pre- and post-index periods, and to allow comparison between cohorts.

Conclusions

This real-world analysis demonstrated that patients prescribed fingolimod benefited from significantly improved rates of persistence and adherence over all injectable/infusible DMTs. These improved outcomes are likely to facilitate the core MS treatment goals of preventing relapses, slowing disease progression and postponing disability.

Transparency

Declaration of funding

This study was funded by Novartis Pharma AG, Basel, Switzerland.

Declaration of financial/other relationships

NB, RL, and GC are paid employees of Novartis Pharma AG, Basel, Switzerland. NA and AP are paid employees of Novartis Pharmaceuticals Corporation, East Hanover, NJ. JRK, AAP, SUK, CBM, and CM are paid employees of IMS Health, Plymouth Meeting, PA. JME Peer Reviewers on this manuscript have no relevant financial relationships to disclose.

Acknowledgments

The authors take full responsibility for the content of the paper. The authors thank Dr Gemma Carter and Sebastian Reynolds (Oxford PharmaGenesis, Ltd) for medical writing support, editorial assistance, and collation and incorporation of comments from all authors.

Notes

*Gilenya is a registered trademark of Novartis Pharma AG, Basel, Switzerland.

^†Copaxone is a registered trademark of Teva Pharmaceuticals, Petach Tikva, Israel.

^‡Avonex is a registered trademark of Biogen Idec, Cambridge, MA, USA.

^§Rebif is a registered trademark of Merck Serono SA, Genf, Switzerland.

^¶Extavia is a registered trademark of Novartis Pharma AG, Basel, Switzerland.

**Betaseron is a registered trademark of Bayer Healthcare, Leverkusen, Germany.

^††Tysabri is a registered trademark of Biogen Idec, Cambridge, MA, USA.