Generating evidence to inform health technology assessment of treatments for SLE: a systematic review of decision-analytic model-based economic evaluations

This study aimed to understand and appraise the approaches taken to handle the complexities of a multisystem disease in published decision-analytic model-based economic evaluations of treatments for SLE. A systematic review was conducted to identify all published model-based economic evaluations of treatments for SLE. Treatments that were considered for inclusion comprised antimalarial agents, immunosuppressive therapies, and biologics including rituximab and belimumab. Medline and Embase were searched electronically from inception until September 2018. Titles and abstracts were screened against the inclusion criteria by two reviewers; agreement between reviewers was calculated according to Cohen’s κ. Predefined data extraction tables were used to extract the key features, structural assumptions and data sources of input parameters from each economic evaluation. The completeness of reporting for the methods of each economic evaluation was appraised according to the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) statement. Six decision-analytic model-based economic evaluations were identified. The studies included azathioprine (n=4), mycophenolate mofetil (n=3), cyclophosphamide (n=2) and belimumab (n=1) as relevant comparator treatments; no economic evaluation estimated the relative cost-effectiveness of rituximab. Six items of the CHEERS statement were reported incompletely across the sample: target population, choice of comparators, measurement and valuation of preference-based outcomes, estimation of resource use and costs, choice of model, and the characterisation of heterogeneity. Complexity in the diagnosis, management and progression of disease can make decision-analytic model-based economic evaluations of treatments for SLE a challenge to undertake. The findings from this study can be used to improve the relevance of model-based economic evaluations in SLE and as an agenda for research to inform future health technology assessment and decision-making.


Supplementary Appendix 2: Full Data Extraction
. Full Data Extraction

Alternatives:
(i) Usual full dose azathioprine; (ii) PCR genotype test of TPMT activity to inform dose of azathioprine.

Model type:
Decision tree.

Time horizon:
One year.

Perspective:
Third party payer.

Benefit measure:
Adverse drug reactions avoided.
Direct costs included: PCR testing; Treatments; Routine laboratory tests; Physician office visits.

Indirect costs included:
Not applicable.

Effectiveness:
Adverse events from azathioprine [2]; Effectiveness of testing based on an assumption; Test accuracy [3].
Health-related quality of life: Not applicable.

Resource use:
Probability and duration of hospitalisations estimated by experts; Number of visits based on an assumption.

Unit costs:
Cost of testing by proxy with other PCR tests; Hospital cost model [4]; Provincial Guide to Medical Fees; IMS Health Canada Database.

Discount rate:
Not applicable.

Deterministic sensitivity:
Probability of preventable adverse drug reactions; Cost of PCR testing; Probability of hospitalisation; Test accuracy.

Value of information:
No.

Base-case:
Genotype test to inform azathioprine dose was dominant.

Probabilistic analysis:
Not applicable.

Study Design
Study Characteristics Data Sources Analysis Results Target population: Patients, aged 40, newly diagnosed with active, severe lupus nephritis and receiving immunosuppressive therapy.

Alternatives:
(i) IV cyclophosphamide (1,000mg/m 2 per month) for 6 months induction and IV cyclophosphamide every 3 months for 3 years maintenance; (ii) IV cyclophosphamide (1,000mg/m 2 per month) for 6 months induction and azathioprine (50mg per day) for 3 years maintenance; (iii) IV cyclophosphamide (1,000mg/m 2 per month) for 6 months induction and MMF (1000mg per day) for 3 years maintenance; (iv) MMF (2,000mg per day) for 6 months induction, then MMF (1,000mg per day) for 6 months, then azathioprine (50mg per day) as maintenance for up to 3 years.
Model type: Markov model.

Direct costs included:
Treatments; Laboratory tests; Administrative costs.

Health-related quality of life:
Direct observation (18 patients) in 4 Thai hospitals.

Resource use:
Medical records in 4 Thai hospitals.

Unit costs:
Thailand Ministry of Public Health; Standard Cost List for Health Technology Assessment; Thailand Hospital Database; ESRD [10].

Discount rate:
3% for costs and health.

Currency (Price year): Thai baht (2012)
Deterministic sensitivity: One-way sensitivity analysis on all input parameters.

Value of information:
No.

Base-case:
Probabilistic analysis: Strategy (ii) had the highest probability of cost-effectiveness across a range of costeffectiveness thresholds.

Value of information:
Not applicable.

Key drivers of relative cost-effectiveness:
Relative risk of complete remission, partial remission, and renal failure.   [11].

Study Design
Study Characteristics Data Sources Analysis Results Target population: Patients with lupus nephritis, between 20 and 40 years, that had responded to their induction therapy.

Country:
United States of America.

Model type:
Markov microsimulation model.

Time horizon:
3 years and lifetime.

Direct costs included:
Treatments, patient visits, laboratory studies, imaging studies, emergency visits, outpatient surgery, hospitalisations.

Indirect costs included:
Patient labour and non-labour time, caregiver time delivering care.

Deterministic sensitivity:
Single trial for treatment effectiveness; indirect costs in remission; utility in remission; cost of treatments; probability of infection; probability of end stage renal disease; direct costs during remission; disutility from infection; probability of relapse.

Value of information:
Population expected value of perfect information for the three-year model.

Base-case:
In the lifetime model, MMF had an ICER of $6,454 per QALY gained, relative to azathioprine.

Probabilistic analysis:
The probability that MMF was cost-effective was close to 100% for thresholds of $50,000 and $100,000 per QALY gained.

Value of information:
Population EVPI was $2,058,206 at a threshold of $100,000 per QALY gained for the three-year model.    [22].

Study Design
Study Characteristics Data Sources Analysis Results Target population: Adults (50kg body weight) with moderate to severe rheumatoid arthritis or SLE who are unsuitable for methotrexate or cyclophosphamide.

Alternatives:
(i) Conventional weight-based dose of azathioprine; (ii) PCR genotype test of TPMT activity to inform dose of azathioprine.

Model type:
Decision tree.

Time horizon:
One year.

Benefit measure:
Azathioprine discontinuation due to severe adverse events.

Direct costs included:
Treatments, routine laboratory tests, PCR genotype test, hospital admissions.

Indirect costs included:
Not reported.

Health-related quality of life:
Not applicable.

Resource use:
Treatment guidelines from the Physician's Desk Reference, Hospitalisations from four cases observed at Hanyang University Hospital.

Unit costs:
Korean insurance system, Four cases observed at Hanyang University Hospital.

Discount rate:
Not reported.

Deterministic sensitivity:
Prevalence of TPMT activity, Cost of hospital admission, Cost of PCR genotype test, Incidence of severe adverse events.

Value of information:
No.

Base-case:
Genotype test to inform azathioprine dose was dominant.

Probabilistic analysis:
Not applicable.

Value of information:
Not applicable.

Key drivers of relative cost-effectiveness:
Result was not sensitive to changes in input parameters.   [25].

Study Design
Study Characteristics Data Sources Analysis Results Target population: 50,000 patients with SLE that had active disease and a positive autoantibody test (anti-dsDNA positive and low complement).

Perspective:
Italian National Health Service and Societal.

Direct costs included:
Diagnostic tests; specialist visits; organ damage.

Indirect costs included:
Not reported explicitly.

Effectiveness:
Accompanying systematic review of the literature; BLISS trials [26] and the Johns Hopkins observational cohort [27].

Health-related quality of life:
Published literature, based on the BLISS trial sample of UK patients.

Resource use:
Not reported.

Discount rate:
3% for costs and health.

Deterministic sensitivity:
Change in SELENA-SLEDAI at 1-year; change in SELENA-SLEDAI according to the natural history model; discontinuation rate; probability of response; mortality and organ damage probabilities; mortality rates; utility; organ damage disutility; costs of each SELENA-SLEDAI score; organ damage cost; indirect costs.

Value of information:
No.

Base-case:
For NHS perspective, belimumab had an ICER of €32,859 per QALY gained, and €22,990 per life year gained, compared with the standard of care.

Probabilistic analysis:
At a threshold of €30,000 per QALY gained, the probability that belimumab was cost-effective was 29.1%. At €40,000 per QALY gained, the probability was 84.3%.

Value of information:
Not applicable.

Study Design
Study Characteristics Data Sources Analysis Results Target population: 10,000 patients with lupus nephritis that were due to receive induction therapy for a flare.

Country:
United Kingdom.

Model type:
Patient-level simulation.

Direct costs included:
Treatments and administration; major and minor infections.

Indirect costs included:
Not applicable.

Deterministic sensitivity:
Response rate to MMF and IV cyclophosphamide; rate of major and minor infections from both alternatives; duration of major and minor infections; rate of switching between alternatives; dose of treatments; disutility associated with each level of response; disutility associated with major and minor infections.

Value of information:
No.

Base-case:
The base-case analysis indicated that MMF dominated IV cyclophosphamide.

Probabilistic analysis:
At a cost-effectiveness threshold of between £25,000 to £35,000 per QALY gained, MMF had an 81% probability of being cost-effective.

Value of information:
Not applicable.

Key drivers of relative cost-effectiveness:
The response rate to IV cyclophosphamide.