Using observational analysis of multiple sclerosis relapse to design outcomes-based contracts for disease-modifying drugs: a feasibility assessment

Abstract

Objective:

To assess predictors and costs of multiple sclerosis (MS) relapse, a potential outcome measure in payer-manufacturer risk-sharing agreements for disease-modifying drugs (DMDs).

Methods:

A retrospective cohort analysis of medical/pharmacy claims was used. Study patients had ≥1 DMD (interferon beta, glatiramer, natalizumab) claim, without DMD claims in a 6-month pre-period before DMD initiation; were aged 18–64 years and continuously enrolled from the pre-period through a 24-month post-period; and had ≥2 MS medical claims during the 30-month study period. Post-period relapse cohorts included: (1) severe (hospitalization with MS diagnosis); (2) moderate (outpatient services including intravenous methylprednisolone); and (3) none. Poisson regression modeled severe relapse frequency, logistic regression modeled ≥1 severe relapse, and generalized linear modeling predicted healthcare costs. Tested predictors included demographics, insurance type, index DMD, pre-period health status, and DMD medication possession ratio (MPR).

Results:

Severe relapse was experienced by 14.5% and moderate relapse by 13.8% of 2291 patients. In logistic regression, severe relapse was predicted by plan type; age (odds ratio [OR] = 1.018, 95% confidence interval [CI] = 1.005–1.031); pre-period Charlson Comorbidity Index (OR = 1.307, 95% CI = 1.166–1.464); pre-period proxy measure indicating impaired activities of daily living (OR = 1.470, 95% CI = 1.134–1.905); pre-period MS hospitalization (OR = 2.174, 95% CI = 1.537–3.074); and DMD non-adherence (MPR OR = 0.101, 95% CI = 0.068–0.151). Poisson regression results were similar. Predicted mean [standard deviation] all-cause healthcare expenditures were tripled for patients with severe compared with moderate relapse ($48,173 [$8665] and $13,334 [$1929], respectively).

Limitations:

Commercially insured patients from a single payer; use may have been inconsistent with approved indications; proxy relapse measure may have misclassified patients.

Conclusions:

Severe MS relapses requiring hospitalization, although affecting less than 15% of patients initiating DMD treatment, are associated with high medical costs. The only actionable predictor of severe relapse identified in observational analysis was MPR, raising questions about the feasibility of using observational data to guide outcomes-based contracting.

• Outcomes-based contracts
• Risk-sharing contracts
• Multiple sclerosis
• MS relapse

Introduction

Because of heterogeneity in the clinical manifestations and severity of multiple sclerosis (MS), it is not uncommon to observe some patients experiencing high levels of disease activity and rapid disease progression while other patients remain relatively symptom-free1,2. The rate of relapse, defined as symptoms or signs of an acute central nervous system demyelinating event lasting at least 24 h3, is among the most commonly used measures of disease activity1,4. Clinical trial data suggest that relapse can influence disability, particularly short-term disability with afferent pathway involvement5–7, although evidence about the relationship between relapse and long-term disability is equivocal8–12.

Regardless of its effects on long-term disability, preventing MS relapse is important because of its negative effects on patients’ short-term quality-of-life (QoL) and functioning2,13,14. For example, in a sample of patients treated with disease-modifying drugs (DMDs), Kobelt et al.2 measured health utility (0 = death, 1 = full health) with the EQ-5D (EuroQol Group, The Netherlands), a validated and commonly used QoL scale, and found mean scores of 0.648 for patients with a relapse in the previous 3 months compared with 0.742 for patients without relapse. In a sample not limited to DMD-treated patients, Karampampa et al.13 found a smaller QoL effect measured in a sub-group of patients with Expanded Disability Status Scale scores of 5 or less (0.730 and 0.744, respectively, for patients with and without relapses in the previous year); however, relapse was strongly associated with annual mean hours of informal care provided by family and/or friends (259 vs 49 for patients with and without relapse, respectively) and sick leave (211 vs 48, respectively). Patient survey data also suggest that relapse prevention is important to patients with MS. For example, using conjoint analysis, Johnson et al.15 found that in return for a reduction from 4 to 1 relapses in 5 years, coupled with a reduction in time to progression of disability from 3 to 5 years, patients with MS (n = 651) were willing to accept annual risks of 0.38% for progressive multifocal leukoencephalopathy (PML)-related death or disability, 0.39% for death from liver failure and 0.48% for death from leukemia.

When relapse does occur, treatment choice depends on severity and may include outpatient services, inpatient care, and/or oral or intravenous (IV) corticosteroids3,16,17. For patients not requiring acute observation, home health services may be used to provide IV methylprednisolone (IVMP) treatments. Additional treatment options include plasma exchange for patients whose relapse remains unresponsive to steroid treatment and the use of IV immunoglobulin in combination with IVMP18,19. The costs to manage relapse may be high; O’Brien et al.16 estimated costs (2002 US$) of $12,870 for ‘high-intensity management’ requiring hospitalization; $1847 for a ‘moderate episode’ requiring emergency department, observational unit, or ‘acute treatments’, such as IVMP; and $243 for a ‘mild episode’ requiring physician office visit care and medications to treat symptoms.

Given the importance of relapse both to patients and to payers, it is not surprising that relapse is frequently considered an outcome of potential interest in risk-sharing contracts between health plans and pharmaceutical manufacturers of MS medications20. Both in the US and in Europe, these contractual arrangements have been the subject of great interest among healthcare payers and drug manufacturers as a way to pay for performance, thereby potentially improving or maintaining patient outcomes in the face of limited drug budgets21. However, for health plans and pharmaceutical manufacturers to develop mutually beneficial risk-sharing arrangements, credible evidence necessary to predict and affect the outcomes of interest must be available so that both parties can gauge the potential risks and benefits of these arrangements with a reasonable level of accuracy. With the exception of the UK’s highly controversial, ongoing study of a risk-sharing arrangement for the use of DMDs in patients with MS22, evidence about the experiences of payers and manufacturers in outcomes-based contracting is generally limited to anecdote, case study, and modeling21,23. The question of whether it is possible to identify actionable predictors of MS relapse using administrative data, the most readily available source of information for most organizations, has not been examined in research to date.

The present study sought to provide evidence about the implications of using severe relapse as an outcome of interest in risk-sharing contracts related to the care of patients with MS. Severe relapse requiring hospitalization was chosen as the primary study outcome because the study by O’Brien et al.16 suggested that it is the principal cost driver among relapsing patients; however, because that study was published 10 years ago and may not reflect contemporary utilization outcomes, we also calculated costs for patients with moderate relapse requiring outpatient IVMP. To assess the predictability and cost of relapse, the study addressed the following research questions in a sample of patients enrolled in commercial health plans: (1) Do baseline demographic or clinical characteristics differ for patients who experience moderate or severe relapse compared with those who do not? (2) What factors are associated with the occurrence and number of severe relapses? (3) What are the economic effects of moderate or severe relapse on health plans?

Methods

Data source

The study used integrated medical and pharmacy claims from the i3 InVision^TM Data Mart databases. The i3 Invision^TM Data Mart contains data collected as part of normal business operations for a national insurance organization with ∼14 million enrollees each year, located predominantly in the southern and midwestern US. Data are provided to investigators with unique de-identified coding to ensure compliance with the patient privacy standards of the Health Insurance Portability and Accountability Act (HIPAA) of 1996. The database is commonly used in health services research and is generally considered to represent the experiences of commercially-insured patients in the US24–26.

Subject selection

The study was based on claims with dates of service from January 1, 2006, through March 31, 2011, and a sample identification period of July 1, 2006, through March 31, 2009. All study patients met the following a priori inclusion-exclusion criteria: (1) at least one claim in the sample identification period for a DMD, identified by national drug code (NDC) numbers and J codes, with no DMD claims in the previous 6 months (pre-period) and with the first DMD claim date designated as the index date; (2) aged 18–64 years on the index date; (3) continuously enrolled for a 30-month study period including the 6-month pre-period and the 24 months post-index (post-period); and (4) at least two claims at any time during the 30-month study period with a primary or secondary diagnosis of MS, designated by International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code of 340.xx.

Study outcomes

Relapse events were defined in a manner similar to that of O’Brien et al.16, who were among the first to attempt to categorize relapse severity using administrative data sources. Specifically, a relapse was identified as a medical claim meeting the following criteria: (1) date of service in the 24-month post-period, (2) either a primary or secondary diagnosis of MS and (3) either inpatient hospital care or IVMP. Patients were categorized according to whether a relapse had occurred and, if so, the level of intensity required to treat the event: (1) severe relapse requiring hospitalization, (2) moderate relapse requiring outpatient IVMP use or (3) no relapse meeting these criteria.

Healthcare resource utilization was measured using claims with dates of service either within the 90 days following the relapse or until the next relapse, whichever occurred first, and was defined as plan-allowed amount. All cost data were normalized to 2012 US dollars using the medical care component of the Consumer Price Index (CPI), the inflation adjustment measure used in the study by O’Brien et al.16 Annualized costs were calculated based on follow-up time (e.g., 90-day costs were multiplied by 4.1).

Statistical analysis

Descriptive statistics, including Pearson χ² tests for categorical variables and Analysis of Variance (ANOVA) for continuous variables, were used to compare baseline patient characteristics across the three cohorts. To assess predictors of relapse, two different types of models were used, each reflecting an outcome measure that might be used in a risk-sharing contract. Poisson regression models, which are appropriate for count data, were used to model frequency of severe relapses because patients with MS commonly experience more than one event within any given 24-month period. Additionally, logistic regression analysis was used to model the occurrence of at least one severe relapse during the 24-month study period. For each predictor variable, the intensity (incidence) rate ratio (IRR) was estimated as the exponentiated coefficient from the Poisson regression, and odds ratios (ORs) and 95% confidence intervals (CIs) were estimated in the logistic regression.

In both Poisson and logistic regression models, predictors included patient characteristics (age, sex, region of residence, and insurance product); intent-to-treat (index) DMD (interferon beta 1-a intramuscular, interferon beta 1-a subcutaneous, interferon beta 1-b, glatiramer, or natalizumab); three measures of pre-period health status; and medication possession ratio (MPR), a commonly used measure of treatment adherence. The three health status measures included (1) the Charlson Comorbidity Index (CCI), (2) occurrence of at least one severe relapse in the pre-period, and (3) a proxy measure of activities of daily living (ADL) impairment developed by Chan et al.27. The proxy measure assesses the number of disability-associated conditions (e.g., depression, decubitus ulcers) based on ICD-9-CM codes and was validated by Chan et al. using ADL and mobility impairment findings from the Medicare Current Beneficiary Survey. For the present study analysis, the proxy measure was recoded as a binary variable indicating presence of at least one disability-associated condition. MPR was defined from the index date to the relapse date for patients who experienced a relapse or from baseline to the end of the post-period for patients who did not have a relapse. The MPR numerator was the total days supply for the index DMD.

Because cost data are frequently not normally distributed, generalized linear models were used to calculate adjusted (predicted) mean paid amounts for each of the two relapse groups, controlling for patient demographic and insurance characteristics, year of index date, CCI, ADL limitation proxy, pre-period relapse, and index DMD. Cost data were sub-divided into the following components: (a) inpatient, (b) emergency department, (c) home health, (d) physician office visits, (e) physical and speech therapy, (f) oral steroids, (g) IVMP, (h) other services, and (i) total expenditures. The ‘other’ category encompassed all services not included in (a)–(g), such as prescription drugs (other than steroids); laboratory tests; imaging, including magnetic resonance imaging (MRI); and outpatient surgeries and procedures, including those necessary to address bladder or bowel dysfunction. For cost components (a)–(g), 2-part models were used to estimate costs, where the first part was a probit regression to estimate the probability of a service being utilized and the second part estimated costs for utilizers based on a generalized linear model with gamma distribution and log link. The 2-part modeling approach is common in statistical analyses of data with numerous values of zero (0), such as healthcare services that are used by a minority of patients28. For components (h) and (i), generalized linear models with gamma distribution and log link were employed. All cost regressions were estimated only for the sub-group of patients who had at least one relapse in the post-period.

All analyses assumed an a priori level of significance of 0.05 and were conducted using SAS version 9.2.

Results

Based upon the a priori inclusion and exclusion criteria, 2291 patients aged 18–64 years initiated a DMD (at least one DMD medical or pharmacy claim with no DMD use in the previous 6 months), were continuously enrolled with the insurance company, and had at least two primary or secondary diagnoses of MS during the study period (Table 1). The majority of patients, 71.7% (n = 1642), experienced no relapses within the 24-month post-index period. The percentages of patients with severe or moderate relapse were similar, with 14.5% (n = 333) of patients receiving inpatient care and 13.8% (n = 316) receiving outpatient IVMP.

Table 1. Inclusion–exclusion criteria and sample size.

Inclusion–Exclusion Criteria
1. Pharmacy claim for a DMD between July 1, 2006, and March 31, 2009,^a with the first claim date identified as the index date and no DMD use in the previous 6 months (n = 26,612)
2. At least two medical claims with a primary or secondary diagnosis of MS (ICD-9-CM of 340.xx) from 6 months pre-index (i.e., the pre-period) through 24 months post-index (i.e., the post-period) (n = 24,509)
3. Aged 18–64 years on the index date (n = 23,702)
4. Continuous insurance coverage from the start of the pre-period through the end of the post-period (n = 2291)

^aThe study was based on claims with dates of service from January 1, 2006, through March 31, 2011. DMDs included glatiramer, interferon beta 1-a subcutaneous, interferon beta 1-a intramuscular, interferon beta 1-b, and natalizumab.

DMD, disease-modifying drug; ICD-9-CM, International Classification of Diseases, Ninth Revision, Clinical Modification; MS, multiple sclerosis.

The mean age for patients who experienced either no relapse or a moderate relapse was ∼43 years, compared with 45 years for patients in the severe relapse group (p < 0.001; Table 2). The proportions of patients enrolled in either health maintenance organizations (HMOs) or point-of-service (POS) plans were smaller in the severe relapse group compared with the moderate and no relapse groups (p = 0.003). There was no statistically significant relationship between severe relapse and intent-to-treat (index) DMD (p = 0.286).

Table 2. Baseline patient characteristics of study cohort (n = 2291).

	Severe relapse requiring hospitalization (n = 333)	Moderate relapse requiring outpatient IVMP (n = 316)	No relapse (n = 1642)	p-value
Age (years; mean ± SD)	45.13 ± 11.51	42.65 ± 9.63	42.88 ± 10.29	<0.001
Gender (% Female)	77.8	80.7	78.2	0.579
Insurance product type (%)				0.003
EPO	17.1	18.4	16.3
HMO	8.7	9.5	9.3
Indemnity	2.7	1.0	0.6
POS	63.4	65.5	69.2
PPO	7.8	5.7	4.6
Other	0.3	0	0
Index DMD				0.286
Glatiramer	41.74	36.39	37.58
IFN beta 1-a IM	18.92	24.05	23.14
IFN beta 1-b	11.11	14.56	12.12
IFN beta 1-a SC	23.42	21.84	21.44
Natalizumab	4.80	3.16	5.72
CCI (mean ± SD)^a	0.84 ± 1.29	0.41 ± 0.76	0.41 ± 0.91	<0.001
Possible ADL impairment (%)^a,^b	53.2	42.7	36.6	<0.001
MS hospitalization (%)^a	21.0	9.5	10.3	<0.001
MS hospitalization or IVMP (%)^a	31.5	36.4	18.9	<0.001

^aPre-period measure of health status.

^bProxy measure that indicates whether patient had at least one condition (medical, psychosocial, musculoskeletal or dysphagia-related) suggesting possible ADL impairment. Source: Chan et al.27

ADL, activities of daily living; CCI, Charlson Comorbidity Index; DMD, disease-modifying drug; EPO, exclusive provider organization; HMO, health maintenance organization; IFN, interferon; IM, intramuscular; IVMP, intravenous methylprednisolone; POS, point-of-service; PPO, preferred provider organization; SC, subcutaneous; SD, standard deviation.

Pre-period health status was significantly associated with post-period relapse events. Mean pre-period CCI scores and percentages of patients with a diagnosis indicating possible ADL impairment27 were higher for patients with severe relapse than for those without relapse or with moderate relapse (p < 0.001). Additionally, 31.5% of patients with severe post-period relapse and 36.4% with moderate post-period relapse had experienced an MS hospitalization and/or IVMP in the pre-period, compared with 18.9% of patients without post-period relapse (p < 0.001).

In the Poisson regression model, counts of severe relapses were predicted by older age (IRR = 1.013, p = 0.019); indemnity insurance or other insurance type (reference category) instead of POS (IRR = 0.468, p = 0.027); and all three measures of baseline health status: CCI (IRR = 1.190, p < 0.001), proxy measure of ADL impairment (IRR = 1.353, p = 0.009) and pre-period MS hospitalization (IRR = 1.701, p < 0.001; Table 3). DMD non-adherence was strongly associated with number of severe relapses (IRR for MPR = 0.155, p < 0.001).

Table 3. Poisson regression analysis of number of severe relapses (n = 2291).

	Poisson model-estimated IRR	p-value
Age	1.013	0.019
Female relative to male	0.947	0.682
Geographic region relative to South
Northeast	0.992	0.966
Midwest	1.070	0.629
West	1.108	0.520
Insurance product type relative to indemnity/other^a
EPO	0.527	0.077
HMO	0.480	0.057
POS	0.468	0.027
PPO	0.724	0.401
Treatment relative to Glatiramer
IFN beta 1-a IM	0.823	0.202
IFN beta 1-b	0.756	0.133
IFN beta 1-a SC	0.979	0.881
Natalizumab	1.051	0.852
Pre-period health status
CCI	1.190	<0.001
Possible ADL impairment^b	1.353	0.009
MS hospitalization	1.701	<0.001
DMD MPR	0.155	<0.001

^aOne patient with ‘Other’ insurance was included in the indemnity group.

^bProxy measure that indicates whether patient had at least one condition (medical, psychosocial, musculoskeletal or dysphagia-related) suggesting possible ADL impairment. Source: Chan et al.27

ADL, activities of daily living; CCI, Charlson Comorbidity Index; DMD, disease-modifying drug; EPO, exclusive provider organization; HMO, health maintenance organization; IFN, interferon; IM, intramuscular; IRR, intensity (incidence) rate ratio; IVMP, intravenous methylprednisolone; MPR, medication possession ratio; MS, multiple sclerosis; POS, point-of-service; PPO, preferred provider organization; SC, subcutaneous.

The logistic regression analysis produced similar results for most of the predictor variables (Table 4). Each 1-year increase in age was associated with a statistically significant increase in the probability of severe relapse (OR = 1.018, 95% CI = 1.005–1.031). One-point increments in the pre-period CCI and having at least one ADL impairment proxy condition were associated with increases in the odds of post-period relapse of ∼31% and 47%, respectively. MS hospitalization prior to DMD initiation more than doubled the odds of post-period relapse (OR = 2.174, 95% CI = 1.537–3.074). As in the Poisson regression analysis, higher MPR was associated with a significantly lower risk of severe relapse (OR = 0.101, 95% CI = 0.068–0.151). All insurance types except preferred provider organization (PPO) were associated with reduced risk relative to indemnity insurance.

Table 4. Logistic regression analysis of ≥1 severe relapse (n = 2291).

	Logistic model estimated odds ratio (95% CI)
Age	1.018 (1.005–1.031)
Female relative to male	0.919 (0.678–1.247)
Geographical region relative to South
Northeast	0.999 (0.644–1.548)
Midwest	1.133 (0.824–1.560)
West	1.137 (0.795–1.626)
Insurance product type relative to Indemnity/other^a
EPO	0.348 (0.131–0.919)
HMO	0.303 (0.110–0.837)
POS	0.297 (0.116–0.756)
PPO	0.529 (0.190–1.478)
Treatment relative to Glatiramer
IFN beta 1-a IM	0.773 (0.550–1.087)
IFN beta 1-b	0.696 (0.458–1.057)
IFN beta 1-a SC	0.952 (0.688–1.319)
Natalizumab	1.037 (0.569–1.890)
Pre-period variables
CCI	1.307 (1.166–1.464)
Possible ADL impairment^b	1.470 (1.134–1.905)
MS hospitalization	2.174 (1.537–3.074)
DMD MPR	0.101 (0.068–0.151)

^aOne patient with “Other” insurance was included in the indemnity group.

^bProxy measure that indicates whether patient had at least one condition (medical, psychosocial, musculoskeletal or dysphagia-related) suggesting possible ADL impairment. Source: Chan et al.27

CCI, Charlson Comorbidity Index; DMD, disease-modifying drug; EPO, exclusive provider organization; HMO, health maintenance organization; IFN, interferon; IM, intramuscular; IVMP, intravenous methylprednisolone; MPR, medication possession ratio; MS, multiple sclerosis; POS, point-of-service; PPO, preferred provider organization; SC, subcutaneous.

Adjusted mean total all-cause healthcare charges for patients experiencing a severe relapse were ∼3.6-times those of patients who were treated for relapse in the outpatient setting ($48,173 vs $13,334, respectively, p < 0.001; Table 5). The majority of the incremental cost difference was due to the cost of hospitalization (p < 0.001). Patients with a severe relapse also incurred higher physical/speech therapy costs (p < 0.001) but lower physician office visit costs (p = 0.006). Expenditures for home health visits were similar for patients with severe relapse treated in the inpatient setting ($991) or moderate relapse treated in the outpatient setting ($953; p = 0.265).

Table 5. Annualized adjusted all-cause healthcare charges (2012 US$) for patients with severe relapse requiring hospitalization or moderate relapse requiring outpatient IVMP.

Estimated Annual Mean (±SD) Charges (2012 US$)^a	Severe relapse requiring hospitalization (n = 333)	Moderate relapse requiring outpatient IVMP (n = 316)	p-value
Inpatient	$31,641 ± 10,207	$480 ± 538	<0.001
Emergency department	$54 ± 82	$40 ± 57	0.008
Home health	$991 ± 473	$953 ± 395	0.265
Physician office visit	$2776 ± 1434	$3079 ± 1332	0.006
Physical/Speech therapy	$111 ± 97	$51 ± 44	<0.001
Oral steroids	$3 ± 2	$2 ± 1	<0.001
IVMP	$19 ± 25	$188 ± 84	<0.001
Other^b	$12,666 ± 3821	$8742 ± 2078	<0.001
Total annualized charges	$48,173 ± 8665	$13,334 ± 1929	<0.001

^aCharges were estimated using a multivariate, generalized linear model with gamma distribution and log link. Regressions controlled for patient characteristics (age, sex, region of residence, and type of insurance coverage), year of index date, general health (Charlson Comorbidity Index), pre-period relapse, proxy for possible ADL limitation27, and index DMD.

^bAll services not included in (a)–(g), such as prescription drugs (other than steroids); laboratory tests; imaging, including magnetic resonance imaging (MRI); and outpatient surgeries and procedures, including those necessary to address bladder or bowel dysfunction.

DMD, disease-modifying drug; IVMP, intravenous methylprednisolone; SD, standard deviation.

Discussion

In a sample of commercially insured patients with MS, we found that the rates of severe and moderate relapse in the 24 months after DMD initiation were 14.5% and 13.8%, respectively. Severe relapse was both associated with high medical cost—adjusted annualized all-cause mean health plan payments of $48,173—and difficult to predict, with only a few significant independent variables identified in multivariate analyses.

The most actionable finding of this study suggests that patients with MS are at greater risk of severe relapse when they do not adhere to therapy, and health plans and pharmaceutical companies have an opportunity to provide patient support so that optimal adherence levels can be achieved. In this respect, the present study’s findings are similar to those of a previous study by Ivanova et al.29, who found 24-month relapse rates of 12.5% and 19.5% for DMD-adherent (MPR ≥ 80%) and non-adherent patients, respectively, using a slightly broader definition of relapse (emergency department or hospital care for MS). When the definition by Ivanova et al. was limited to MS-related inpatient care, the resulting relapse rate of 9.1% (7.6% and 12.5% for adherent and non-adherent patients, respectively) was somewhat lower than the 14.5% severe relapse rate identified in the present study, perhaps because the sample in the former study consisted of active employees (rather than employees and dependents). Similarly, in a sample comparable to that assessed in the present study but using a shorter follow-up period, Tan et al.30 found MS hospitalization rates of 7.8% and 12.0%, respectively, for DMD-adherent and non-adherent patients during the 1 year after DMD initiation.

Study findings have potentially important implications for the use of observational research as a tool to generate ‘real-world’ evidence about resource-intensive medical events in the context of payer–pharmaceutical manufacturer contracting. Specifically, a limited number of actionable predictors identified in observational analyses of typical administrative claims databases may make outcomes-based contracts for MS difficult to negotiate and manage. In addition to MPR, significant predictors of severe relapse included age, insurance type, and baseline health status measures. Payers and pharmaceutical companies have little ability to influence the level of comorbidity or disability of patients insured within a given health plan. Both parties also have no influence over the patient’s age and little influence on other pre-existing conditions such as prior severe relapse. Clinically, our findings are unsurprising, because it is widely acknowledged that predicting the disease course of any individual patient with MS is difficult1. Laboratory-based measures, such as tissue biomarkers and MRI results, are potentially useful in clinical settings31,32, but unlikely to be available to payers and manufacturers engaging in contract negotiation and follow-up.

The present study’s findings are consistent with the limited evidence about risk-sharing arrangements reported to date, which has generally suggested that these schemes, although seemingly promising, are difficult to implement21. First, they are often administratively complex and require high-quality information systems. For example, a risk-sharing contract between a large US health plan and the manufacturer of risedronate, a bisphosphonate, provides for rebate payments to the health plan for patients who incur a non-vertebral fracture, confirmed by x-ray, while compliant on the drug. The scheme requires tracking of incident fractures, retrospective determination of medication compliance and links from the medical system to the financial system to trigger accurate payments for confirmed cases21,33. A second common problem is difficulty in identifying an unconfounded measure, that is, an outcome that is unlikely to be affected by factors other than the drug. Risk-sharing arrangements using validated biomarkers of disease progression are more likely to be successful than schemes relying on less disease-specific measures (e.g., mortality), albeit still administratively complex21. An example is a risk-sharing arrangement undertaken by the UK for bortezomib, a proteasome inhibitor indicated for multiple myeloma treatment. The agreement provides for refunds made by the manufacturer to the government for non-responding patients, identified using serum M protein values21,34.

In addition to these common problems, payers and manufacturers considering a risk-sharing arrangement for MS face the challenges associated with measuring outcomes for a disease with long-term functional consequences. For example, a preliminary evaluation of the UK’s MS risk-sharing arrangement, a cohort study with historical comparator, produced inconclusive findings largely because of complex methodological issues35. In addition to statistical problems associated with missing data, these included the effect of short-term disability fluctuations on calculation of study outcomes and problems in identifying an appropriate comparator group.

Limitations

The present study has several limitations. First, the analysis was limited to commercially insured patients aged 18–64 years, and its results may not be generalizable to older patients and beneficiaries of either Medicare or Medicaid. Similarly, the data were drawn from a single health plan; it is possible that relevant medical or pharmacy benefit design features (e.g., formulary, pre-authorization for hospital stays) could have affected study outcomes. However, the data are commonly used in healthcare research intended to represent the commercially insured population of the US25,26. Second, we used a proxy measure of relapse that, although consistent with previous claims-based research, may have resulted in misclassification. Third, like any claims-based analysis, this study may have been affected by coding errors or omissions. However, we think it unlikely that coding problems would have affected the study groups differentially, and we used both primary and secondary diagnoses to identify relapses. Fourth, ICD-9-CM coding did not enable us to distinguish US Food and Drug Administration (FDA)-approved from non-approved uses (e.g., primary progressive MS or clinically isolated syndrome). Similarly, because of the relapsing nature of MS, we were not able to determine whether natalizumab was used only as second-line therapy, as recommended in FDA labeling.

Conclusion

An analysis of administrative claims data for 2291 patients with MS found that severe relapses requiring hospitalization, although affecting less than 15% of patients initiating DMD treatment, were associated with high all-cause medical costs. A lack of actionable predictors of MS relapse may complicate the development of mutually beneficial outcomes-based contracts between payers and pharmaceutical manufacturers.

Transparency

Declaration of funding

This study was funded by Teva Pharmaceuticals. One author is an employee of Teva, and the remaining two authors are consultants to Teva. No assistance in the preparation of this article is to be declared.

Declaration of financial/other relationships

The authors report no relevant relationships other than Teva, described above.