Cost-Effectiveness of Rasagiline Compared With First-Line Early Parkinson Disease Therapies

Parkinson disease (PD) is a neurologic disorder that affects approximately 1 million individuals in the United States.¹ The incidence of PD is approximately 60,000 new cases per year.¹ The disease is associated with limitations in physical functioning and autonomy, and it leads to severe disability. As PD progresses, patients and their families experience substantial health and economic burdens. Multiple analyses have shown that the economic burden of PD is directly proportional to increases in levels of disease severity or disability and is greater in patients with motor fluctuations or involuntary bodily movements called dyskinesias.^2-4

Parkinson disease requires a pharmacologic intervention; but rather than choosing the most clinically effective drug, physicians may choose less clinically effective but better-tolerated and better-adapted drugs to fi t a patient’s profile. This issue is particularly relevant to patients withearly PD. In the United States, indicated treatments for early PD include levodopa (LD), dopamine agonists (DAs), and selective irreversible monoamine oxidase type-B (MAO-B) inhibitors. Levodopa is the gold-standard pharmacologic treatment for controlling motor symptoms of PD; however, it has disadvantages as a fi rst-line therapy due to the likelihood of inducing long-term irreversible motor complications (ie, dyskinesias), which interfere with the normal functioning of a patient.^5,6 DAs are available as first-line therapies and allow relatively good control of motor symptoms; however, DA treatment eventually becomes insufficient and must be replaced by or co-administered with LD, typically within a few years of initially being prescribed. Dopamine agonist treatment is more costly than LD treatment⁷; exposes patients to serious adverse events such as psychiatric disorders, cardiovascular fibrosis, and sleep attacks8; and has a more rigorous titration schedule that can be more difficult to follow for both the patient and the caregiver. Dyskinesias also may develop in patients taking a DA, although the risk of developing dyskinesias is less than that with LD.^9,10

Rasagiline mesylate (rasagiline), a once-daily, well-tolerated, second-generation selective irreversible MAO-B inhibitor, is indicated as monotherapy treatment in early PD. Like other PD treatments, treatment with rasagiline inevitably will be supplemented or replaced by a DA or LD to control PD symptoms as the disease progresses. Initiating treatment with rasagiline prior to treatment with a DA, however, may delay motor complications associated with LD. Although more costly than generic DAs or LD, rasagiline has a better side effect profi le. Introducing rasagiline into the treatment pathway as a fi rst-line therapy could be an effective strategy for postponing the need for a DA and/or LD, thereby delaying the appearance of disabling motor complications that are linked to poor quality of life¹¹ and higher healthcare costs.^3,4,12

The purpose of this study was to examine whether rasagiline is a cost-effective first-line treatment strategy when compared with initiating PD therapy with a DA (in the extended-release dose or in either the generic or the branded version of the standard-release dose) or first-line LD. To our knowledge, this is the first analysis to compare all approved, early PD monotherapy treatment strategies from a US managed care perspective.

METHODSModel Structure

A Markov health-state transition model was developed in Microsoft Excel (Microsoft Corporation, Redmond, Washington) to calculate the costs and outcomes of patients starting therapy for early PD. This design is similar to that of a previously published economic analysis.¹³ The Markov design was chosen for its appropriateness in estimating costs and outcomes for progressive diseases such as PD. The model time horizon was 5 years, which was consistent with the long-term follow-up from the TVP-1012 in Early Monotherapy for Parkinson Disease Outpatients (TEMPO) trial¹⁴ and the pivotal ropinirole trial.⁹

Figure 1

displays the Markov model structure. The model was designed as a therapy-switching model, in which health states are represented by the current therapy. Patients switch therapy when the current therapy is no longer suffi cient to manage their disease symptoms, as determined by their clinician.

Patients entered the model in fi rst-line therapy with rasagiline, a DA, or LD. Every 6 months, patients could either remain on current therapy or switch according to the treatment patterns displayed in Figure 1. Patients starting on rasagiline could switch to a DA or LD because rasagiline has a mode of action distinct from that of DAs, and a decrease in rasagiline efficacy does not prevent switching to a DA. Patients who were prescribed LD for any reason were assumed to switch to LD, even if the drug was co-administered with a DA or LD. Patients taking a DA could develop dyskinesias or switch to LD. Patients taking LD could develop dyskinesias. All patients could transition to the death state, according to a PD-adjusted, therapy-independent mortality risk.

The model predicted total costs from a US managed care perspective and total quality-adjusted life-years (QALYs) after 5 years from the start of PD therapy. Costs and outcomes were presented on a per-person basis and were discounted at 3% per annum.¹⁵ Other key outcomes were the percentage of patients requiring LD, the percentage of patients with dyskinesias, and the incremental cost per QALY. Four treatment strategies (ropinirole XL, pramipexole, generic ropinirole, and fi rst-line LD) were individually compared with rasagiline as fi rst-line therapy. We assumed that all DAs had equivalent effi cacy. Therefore, second-line DA therapy was determined by fi rst-line DA therapy in the comparator arm. If treatment was initiated with LD, then ropinirole XL was assumed to be the DA following rasagiline, to bias the model against rasagiline.

Inputs

Patients entering the model were assumed to be 61 years old, to have Hoehn and Yahr (H&Y) stage 1.5 early PD, and to require a pharmacologic intervention for their condition. We based this assumption on the patient characteristics in the TEMPO trial.¹⁴

Health-state transition probabilities are summarized in

Table 1A

. Transition probabilities from rasagiline to a DA and from rasagiline to LD were calculated based on the TEMPO (H. Lundbeck, unpublished data, 2009) trial,¹⁴ using the percentage of patients initiating any clinicianselected DA (or those requiring LD) during each 6-month cycle following the start of rasagiline therapy. For this analysis, we assumed all DAs to be equally efficacious.

Patients in both arms of the TEMPO trial received adjunctive LD as needed. We assumed that any requirement for LD resulted in the discontinuation of the previous therapy (either rasagiline or DA) and that patients continued with LD monotherapy. The transition probabilities from DA to LD were calculated from the ropinirole trial.⁹ In that trial, 65.88% of patients were reported to have received supplemental LD by the end of the 5-year period due to such factors as adverse effects of DA therapy or poor response to DA monotherapy. ⁹ The probability of switching from a DA to LD in 6 months was estimated by assuming an exponential distribution.

Patients could experience dyskinesias while on treatment with a DA or LD. The probabilities of developing dyskinesias while on LD or DA therapy were obtained from previous studies.^9,10 Time-dependent probabilities of developing dyskinesias were derived from a 5-year survival distribution function showing the percentage of patients remaining dyskinesia free on each therapy where those on a DA did not receive supplemental LD.

Death could occur from any health state. Risk of death was based on US population—level, sex- and age-specific, all-cause mortality¹⁶ and adjusted using a PD-specifi c relative mortality adjustment of 2.3.¹⁷ Mortality risk was assumed to be independent of treatment strategy, as no data were available to suggest any treatment-specific difference in mortality.

Pharmaceutical costs are presented in

Table 1B

. Rasagiline was assumed to be dosed as indicated, 1 mg once daily.¹⁸ Ropinirole XL, pramipexole, and generic ropinirole costs were estimated from an escalating dosing schedule.^9,19,20 We assumed equivalent effi cacy among DAs and used ropinirole as the primary DA comparator because it was the least expensive of the DA options. Patients taking LD were assumed to take a co-formulation of carbidopa and LD, using a 1:4 ratio of carbidopa to LD,²¹ as well as an LD dosage of 400 mg per day.²¹ As carbidopa and LD are generic, we assumed the least expensive generic option. All drug prices were based on wholesale acquisition costs from the Red Book.²²

In addition to drug costs, we considered other medical costs, including inpatient admissions, emergency department and outpatient visits, long-term care, outpatient therapy, and medical equipment. We assumed all patients in nondyskinetic health states incurred the same nonpharmaceutical direct medical costs, regardless of treatment. These costs were estimated from the nonpharmaceutical components of a managed care database analysis of early PD patients²³ and are presented in Table 1B. We assumed that patients in dyskinetic health states had higher direct medical costs.^3,4,12 We estimated the relative cost incurred by patients with any evidence of dyskinesias (Unified Parkinson Disease Rating Scale domain IVa score >0) to be 1.679 times those costs incurred by patients with no evidence of dyskinesias (Unifi ed Parkinson Disease Rating Scale domain IVa score = 0).¹² Costs were infl ated from 2002 US dollars to 2010 US dollars.²⁴

Health-state utility weights were used to measure the benefit of therapy. Utility weights allow for an objective measurement of the desirability of a health state in a cost-utility analysis. A utility value of 1.0 represents perfect health, whereas a value of 0.0 represents death. When combined with life-years, utilities produce QALYs. Utility weights were obtained from the visual analogue scale utility estimates presented by Palmer et al.¹¹ The rasagiline, DA, and first-line LD health states assumed that patients were at H&Y 1.5 with no off time, because this was consistent with the starting H&Y scores in both the clinical trials.^9,14 The second-line LD health state assumed that patients were at H&Y 2.5 with no off time because patients who required LD following either rasagiline or a DA were assumed to have more advanced disease. We assumed that patients on a DA with dyskinesias and patients on LD with dyskinesias had H&Y scores of 1.5 and 2.5, respectively, with off time. A weighted average was calculated for those patients with various percentages of off time by assuming the dyskinesias were correlated with classification of off time motor fluctuations by Palmer et al.¹¹ Table 1B presents the utility weights used in each modeled health state.

Sensitivity Analyses

To test the robustness of our assumptions and the effects of uncertainty-specifi c parameter estimates, we examined the effects of changes to parameter values in 1-way, scenario, and probabilistic sensitivity analyses. All sensitivity analyses were tested on the base results of rasagiline versus generic ropinirole because this comparison resulted in the least cost-effective results for rasagiline in the base-case analysis. The sensitivity of results to individual parameters was examined by varying each parameter value to its upper and lower bounds in a 1-way sensitivity analysis. Plausible ranges for the sensitivity analysis can be seen in Table 1B. We also performed a probabilistic sensitivity analysis (second-order Monte Carlo simulation) in which all parameters were varied simultaneously.²⁵ Costs and the dyskinesia multiplier were varied according to a gamma distribution, and transition probabilities and utilities were varied according to a beta distribution. All parameters were estimated assuming a standard error of 10% of the base case value. The analysis was run 10,000 times in order to capture stability in the results.

RESULTS

Results for rasagiline compared with the other 4 early PD pharmacologic interventions are presented in

Table 2

. Initiating treatment with rasagiline resulted in better outcomes, in terms of QALYs, compared with ropinirole XL, pramipexole, generic ropinirole, and LD (0.10, 0.10, treatment with rasagiline was cost saving compared with ropinirole XL, pramipexole, and LD ($3141, $833, and

$571, respectively). Compared with generic ropinirole, initiating treatment with rasagiline resulted in slightly increased total costs ($2692), with a cost per QALY of $25,939.

Figure 2

displays the distribution of living patients across therapies and the clinical outcome of dyskinesias after 5 years. Initiating treatment with rasagiline delayed the need for LD, thereby reducing the incidence of costly and uncomfortable dyskinesias by a relative 38% and 73% when compared with initiating therapy with a DA or LD, respectively.

Figures 3A

and

3B

display the results of the 1-way and scenario analyses when rasagiline was compared with initiating therapy with generic ropinirole. The results were most sensitive to variations in utility weights and the dyskinesia cost multiplier. Varying the utility weight for rasagiline to its lower bound or generic ropinirole to its upper bound resulted in rasagiline being less effective and more costly. Varying the utility weight for rasagiline to its upper bound or generic ropinirole to its lower bound resulted in rasagiline being more costeffective, with incremental cost-effectiveness ratios of $5571 and $9395, respectively. Due to readability issues, utility weights of rasagiline and generic ropinirole are not shown in Figure 3. Using the estimates of standard gamble utility weights, setting transition probability from DA to LD to its lower bound, and setting the dyskinesia cost multiplier to its lower bound (ie, assuming no impact of dyskinesias on costs) each resulted in a cost per QALY of slightly above $50,000 ($52,441, $53,452, and $52,547, respectively). Assuming a standard cost-effectiveness acceptability threshold of $50,000 per QALY,26 the cost-effectiveness of rasagiline compared with generic ropinirole was robust to all other individual parameter and scenario variations. A dyskinesia cost multiplier of 2.35 would result in cost savings when initiating treatment with rasagiline compared with initiating treatment with generic ropinirole. In the probabilistic sensitivity analysis, rasagiline was cost-effective (assuming a threshold of $50,000 per QALY) in 60.5% of simulations when compared with generic ropinirole. Rasagiline was more effective regardless of cost in 69.1% of simulations.

DISCUSSION

Our model considered, from a third-party US managed care perspective, 61-year-old patients with early PD who were given rasagiline as a fi rst-line treatment. Initiating treatment with rasagiline for early PD is predicted to be a cost-effective strategy compared with initiating treatment with ropinirole XL, pramipexole, generic ropinirole, or LD. Patients initiating therapy with rasagiline had improved outcomes (higher QALYs, fewer patients requiring LD, and fewer dyskinesias) compared with patients initiating therapy with a DA or LD. Rasagiline resulted in higher costs over the 5-year period compared with generic ropinirole but was considered cost-effective. Rasagiline resulted in reduced costs compared with ropinirole XL, pramipexole, and LD. These results suggest that initiating treatment with rasagiline will improve patients’ quality of life and potentially reduce total healthcare costs for the payer.

The analysis considered only direct medical costs associated with PD treatment; societal costs were not considered in the model. Including societal costs would increase the cost-effectiveness of rasagiline, as societal costs would presumably be greater for the treatment pathway with a higher incidence of dyskinesias.For example, if additional societal costs were 2.5 times greater than direct medical costs (as estimated by Huse and colleagues²⁷), that would result in the rasagiline strategy being dominant compared with generic ropinirole.

A limitation in the model is the lack of data for the incremental cost of dyskinesias. As observed in the sensitivity analyses, the results of the model are most sensitive to the assumptions regarding the dyskinesia cost multiplier (ie, the factor used to estimate dyskinetic health-state costs). We identifi ed 3 publications that compared nondyskinetic costs with dyskinetic costs, all of which took a European perspective.^3,4,12 Because none of these studies estimated US costs, we took a conservative approach in our estimation of the cost multiplier of 1.679. An examination of the 3 studies indicated a relative dyskinetic cost of 2 times that of nondyskinetic costs. The higher the value of the cost multiplier, the more cost-effective rasagiline was compared with other treatments, due to the higher prevalence of dyskinesias in the comparator compound treatment strategies after 5 years. A cost multiplier of 2.35 resulted in rasagiline being a cost-saving strategy against all comparators. Further research on the relative direct medical costs of PD patients with and without dyskinesias in a US setting would enhance the robustness of the model results.

An additional limitation is that the model captures the benefits of initiating treatment of PD over a 5-year period, the same time period observed in the TEMPO and the ropinirole trials. As such, we do not capture the lifetime costs and outcomes associated with care for PD. Given the starting age of our PD population, our model predicted that 87.6% of the population would still be alive after 5 years. Therefore, if a lifetime analysis were performed, the majority of the patient population would continue to reap the benefi ts of further delaying LD in the treatment pathway. As seen in Figure 2, more than half of the patients in the rasagiline arm had not progressed to LD therapy at 5 years, and more than 80% of patients in the rasagiline arm had not experienced dyskinesias. Extending the time horizon would further magnify the clinical benefi ts of initiating treatment with rasagiline. Thus, the 5-year time horizon may be considered a conservative approach.

Finally, we were limited to published clinical trial data that may or may not refl ect real-world efficacy. We incorporated clinical data from 2 pivotal phase III trials in our analysis.^9,14 Both trials allowed for the supplemental use of LD based on clinical need at the clinician’s discretion. In the TEMPO trial, clinicians could supplement or switch patients on rasagiline to either a DA or LD. We assumed that any use of a DA or LD signifi ed a switch to DA or LD monotherapy rather than combination therapy. Although combination therapy would potentially be more expensive, assuming a full switch to the next treatment biases the model against rasagiline, as this approach moves patients through the treatment pathway faster than may actually occur in clinical practice.

Our analysis suggests that initiating treatment of early PD with rasagiline is a cost-effective treatment strategy for first-line monotherapy and offers patients an additional therapy option in the PD treatment pathway. Delaying DA treatment (thus, subsequently delaying LD treatment) delays the dyskinesias that will inevitably occur. Our model suggests that starting PD therapy with rasagiline may provide patients with an additional treatment in the limited arsenal of therapy options, thereby providing additional time free from disabling dyskinesias that are linked to quality of life and increased costs. Results from this cost-effectiveness model and prior clinical trial data provide ongoing support for rasagiline’s use as monotherapy in the United States for patients with early PD.