Cost-Effectiveness and Cost-Utility of Early Levodopa in Parkinson’s Disease

General design

To separate possible disease-modifying effects from the direct symptomatic effect of levodopa, a multi-center, randomized delayed-start, double-blind placebo-controlled trial design was used. Patients with early PD whose functional health did not yet warrant initiation of PD-medication were randomized to either 80 weeks of treatment with levodopa/carbidopa 100/25 mg three times per day (early-start group) or to 40 weeks placebo followed by levodopa/carbidopa 100/25 mg three times per day for 40 weeks (delayed-start group). A total of eight study-visits were performed. To be able to detect a minimal clinically important difference of four points on the Unified Parkinson’s Disease Rating Scale (UPDRS) [2] at 80 weeks, 446 patients needed to be included [3]. This prospective study was approved by the Amsterdam University Medical Centers Institutional Ethics Committee under number 2011_057.

Early levodopa treatment was economically evaluated versus delayed-start levodopa as the current standard of care in patients of the LEAP-study, where CEA and CUA were performed from the societal perspective. The societal perspective included the health care costs among providers, the non-reimbursable out-of-pocket expenses of patients, and the employer costs of sick leave from work (‘absenteeism’) and lowered productivity while at work (‘presenteeism’). As the UPDRS is frequently used in PD trials to assess treatment efficacy, the extra cost per unit decrease on the UPDRS 80 weeks after baseline was chosen as the CEA outcome measure. The extra costs per additional quality adjusted life year (QALY) during follow-up was chosen as the CUA outcome measure. The time horizon was restricted to the planned period of clinical follow-up of 80 weeks as patients from the early and delayed-start groups clinically did not differ significantly at this point, and no marked differences in resource use could be observed during the last half year of follow-up. Therefore, projections from a lifetime perspective based on modelling of the future disease course were considered obsolete. With a horizon of more than one year, the costs and effects during the weeks that fell in the second year of follow-up were discounted against 4% and 1.5% respectively –these yearly discount rates are common in Dutch health care research.

Cost components and resources

We included the health care costs of (a) levodopa treatment, including pharmacy delivery costs, (b) consultations by the neurologist, psychiatrist, rehabilitation specialist, general practitioner, company physician, emergency care, psychologist, Parkinson nurse, physiotherapist, occupational therapist, speech therapist, and social worker, (c) hospital day care treatment or inpatient stay, (d) institutionalized care other than hospitals, like a nursing home or rehabilitation center, (e) formal home care, and (f) devices supporting a patient’s autonomy, like walking aids or an adjusted telephone. Out-of-pocket expenses concerned over-the-counter medication for extra, private help and adaptations at home. Questionnaires were used to collect data regarding the impact of early levodopa treatment on sickness leave from work. Both work absenteeism and presenteeism were questioned to be able to study productivity losses.

Measurements took place at weeks 22, 40, 56, 68, and 80. The recall period in the questionnaires was four weeks. The results were generalized to the period in-between successive measurements by multiplying the reported data with the length of the preceding time interval in weeks divided by four. Whether this led to an over- or underestimation was qualitatively assessed by a separate question on the representativeness at the individual patient level of the use of resources during the recall period being substantially below, somewhat below, somewhat above or substantially above average.

Unit costs and costing

Unit costing was done in accordance with the most recent Dutch guideline on costing in health care research [4]. Units of volume measurement and unit costs are shown in Table 2. All unit costs were expressed for the reference year 2017 after price-indexing with general consumer price-index figures for the Netherlands, if sources with different base years were used.

The costs of levodopa/carbidopa treatment were based on prototyping, assuming an overall mean 100/25 mg three times per day dose per patient, including pharmacy dispensing fees depending on the mean total length of drug use during the follow-period per study group, starting with a higher fee for a first prescription and lower fees for each subsequent one at quarterly intervals. Unit costs for a consultation of the Parkinson nurse-specialist were aligned with unit costs of a social worker. Out-of-pocket expenses were gathered as reported by patients as monthly averages at the time of completion of the questionnaire.

In order to quantify productivity losses resulting from absenteeism and from presenteeism at work, we first determined the number of daily working hours for each patient with a paid job. Absenteeism was costed by multiplying the number of days absent from work with the patient’s daily working hours. In case of presenteeism the score between 0 (could not do the work at all) and 10 (worked us usual) was used to derive the patient’s daily working hours lost. The elasticity between hours absent and productivity is implicitly included in the valuation in Table 2. In case of sick leave, the average unit costs per lost working hour were taken, irrespective of age and sex. Costs were calculated as the sum of products of the volumes of resources used and their respective unit costs. Volumes of resource use and the costs are reported separately.

Patient outcomes

EQ-5D-3L health status data were gathered at baseline and at weeks 22, 40, 56, 68, and 80 [5]. The health utility of each EQ-5D health status scoring profile was determined by applying the available Dutch valuation algorithm [6]. The algorithm was based on the time trade-off elicitation technique of health state preferences during interviews with adults from the general population. The number of QALYs a patient generated during the 80 weeks of follow-up was determined by interpolation between successive measurements and taking the area under the curve as the best estimate, weighted for the lengths of the periods between measurements and accounting for periods with utilities below zero. If only one of two neighboring measurements was available, no interpolation was performed and the available measurement was considered to represent the interpolated value. Because of the weighting by period length, this approach effectively restores interpolation for each triple of successively planned measurements of which the middle one is missing. Only in case of two non-available successive measurements interpolated data were missing. The patients’ final scores were calculated by taking the weighted mean of available period scores. The maximum number of QALYs a patient could generate during 80 weeks of follow-up was 80/52.175 or 1.533 (undiscounted), or 79.58/52.175 or 1.525 (discounted).

Analyses

The health economic analyses were based on the intention-to-treat analysis data set. Results on costs, patient outcomes and differences between treatment groups are reported along with bias-corrected and accelerated 95% confidence intervals (BCa 95% CI). These confidence intervals were derived by drawing 2,500 samples of the same size as the original samples and with replacement, stratified by treatment group, academic versus non-academic hospital, age below or at least 65 years, and symptom duration shorter than or at least six months. P-values for differences by types of resources used are not reported, as p-values for the corresponding costs are identical.

No incremental cost-effectiveness and cost-utility ratios were calculated for the extra costs per extra unit decrease in UPDRS score or the extra costs per additional QALY. Both health outcome measures showed a near zero absolute difference between the study groups –see Results. Consequently, incremental analyses easily render into positive or negative infinity, leaving the health care policy maker and the interested medical professional in doubt whether early levodopa treatment should be regarded as respectively an inefficient or efficient alternative to delayed-start treatment. This could be overcome if a substantial cost difference was to be observed with a clear preference for one strategy over the other. In absence of a disease modifying effect of levodopa however, the hopes of observing such contrast were low.

Data sharing

De-identified data is available for academic research upon request by the corresponding author.

Patients

In the LEAP-study, 445 patients were assessed at baseline after initial enrolment, with 222 and 223 patients randomly assigned to the early-start, respectively delayed-start group. The CONSORT flow-diagram and descriptions of the demographic and baseline clinical characteristics of patients are reported elsewhere with the study groups showing similar profiles at baseline [1].

Resources use and costs

Use of health care resources was reported by 212 patients in the early-start group and 219 patients in the delayed-start group. The other 14 patients did not complete any questionnaire on health resource use during follow-up and were considered drop-outs for the cost analyses. Including drop-out, data were missing for 51 of 1,110 originally planned periods (4.59%, or 4.05% of person-years of follow-up) in the early-start and for 35 of 1,115 originally planned periods (3.14%, or 2.67% of person-years of follow-up) in the delayed-start group. Between 96.3% and 98.6% of patients at the successive measurements over time indicated that the use of resources during the recall period of the latest four weeks was similar to the preceding period.

Table 2 shows the mean use of resources by study group. Table 3 shows the corresponding costs. Tables 2 and 3 do not contain data on inpatient stays at a hospital ward, rehabilitation center or nursing home, because admissions in this cohort were rare, which is expected in early PD. Informal care from neighboring care givers was also still limited in scale. Walking aids and adjusted telephones too yet had to become popular in use.

Patients in the early-start group used levodopa for the full length of the follow-up period of 80 weeks. In the delayed-start group, 61% of patients started treatment in week 41; 17.9% already started with levodopa treatment between weeks 22 and 40 and were assumed to be on treatment for an average of 49 weeks; 21.1% already started with levodopa between weeks 4 and 22 and were assumed to be on treatment for an average of 64 weeks.

Patients in both treatment groups received most attention from the neurologist, Parkinson nurse, general practitioner, occupational therapist, psychologist, and, most dominantly, the physiotherapist. Patients in the delayed-start group more often visited the physiotherapist (28%), occupational therapist, speech therapist, neurologist, Parkinson nurse, and social worker. This was the main reason that the mean health care costs (Table 3) were lower in the early-start group, saving on average € 517 (undiscounted, P = 0.017; discounted, € 512 saved) per patient, despite the € 105 higher levodopa drug costs in this group. The out-of-pocket expenses by patients for over-the-counter medication and extra help at home were limited in this relatively healthy Parkinson population and were similar for both study groups.

At baseline, 24% indicated to have a paid regular job (early-start group: 23.9%; delayed-start group: 24.2%). More patients in the early-start group (N = 24; 10.8%) worked as an entrepreneur, compared with the delayed-start group (N = 9; 4%). Patients in the early-start group generated 14.4 hours more productivity losses due to absenteeism than patients in the delayed-start group during the follow-up period 80 weeks. These extra two days of absence from work were averaged over all patients, meaning that the difference in the subpopulation with paid employment may be about 6 extra working days of productivity lost during 80 weeks of follow-up. On average, the costs of productivity loss because of absenteeism were € 512 (undiscounted, p = 0.046) higher in the early-start group. The costs of productivity losses due to presenteeism were similar for the two groups. The societal costs of early-start levodopa were similar to the costs of delayed-start levodopa: –€ 59 (P = 0.89) per patient.

Health outcomes

Total UPDRS-scores at 80 weeks were available for 207 patients in the early-start group and 210 patients in the delayed-start group. The mean total UPDRS-score was 26.97 (BCa 95% CI: 25.10, 28.88) in the early-start group and 27.00 (BCa 95% CI: 25.11, 28.87) in the delayed-start group, a non-significant difference of –0.04 (BCa 95% CI: –2.76, 2.74; p = 0.99).

QALY scores for the follow-up period of 80 weeks could be calculated for all patients. Patients in the early-start group generated a weighted mean of 1.30 (BCa 95% CI: 1.26, 1.33) QALYs during follow-up versus 1.30 (BCa 95% CI: 1.27, 1.33) QALYs for patients in the delayed-start group, the difference of 0.00 (BCa 95% CI: –0.05, 0.04; p = 0.89) QALYs being negligible. The discounted QALY results were quite identical (not shown).