Pharmacokinetic-pharmacodynamic-pharmacoeconomic analysis of rituximab for follicular lymphoma
J. Pink(1), S. Lane(2), D. Hughes(1)
(1) Centre for Economics and Policy in Health, Bangor University, UK; (2) Centre for Medical Statistics and Health Evaluation, University of Liverpool, UK
Objectives:
The early determination of economic value has become an important component of the drug development process. Clinical trial simulations based on PKPD models are already used during this process, for such purposes as exploring drug efficacy and safety, optimising trial design for later phases, and considering the effect of different dosing regimens and patient demographics[1]. We look at extending such simulations from merely considering the question of clinical efficacy to considerations of cost-effectiveness.
This is done using a mechanism-based economic modelling approach which incorporates data obtained during phase II clinical studies on the relationships between dose, exposure and response[2]. Specifically, the outputs from population PKPD models are used as the inputs for economic decision analyses, to generate estimates of cost-effectiveness at a much earlier stage than would be possible using convention health economic techniques.
As a proof-of-concept, we describe three case studies of rituximab for the treatment of follicular non-Hodgkin's lymphoma. The first two relate to previously published economic evaluations, to compare simulation- and trial-based estimates. The third forecasts the clinical and economic outcomes of the PACIFICO study, a phase III randomised controlled trial comparing R-FC (rituximab, fludarabine and cyclophosphamide) and R-CVP (rituximab, cyclophosphamide, vincristine and prednisolone) chemotherapies for the treatment of follicular lymphoma[3].
Methods:
We utilised population pharmacokinetic[4] and pharmacodynamic[5] models linking serum rituximab concentration to progression-free survival (PFS), to simulate the effectiveness of rituximab in various clinical contexts. The PK model is a two-compartment linear model, with body surface area and gender as significant covariates[4]. It was based on a phase II study of rituximab in 102 patients with rheumatoid arthritis.
The PD component is an exponential hazard model that links mean rituximab concentration since the last infusion to PFS[5], with the results dependent on the adjuvant chemotherapy. This model was built by fitting data from two studies of rituximab and validated by predicting the results of two additional separate studies. The combined PKPD model was used to generate simulated PFS data for each clinical scenario.
These data served as inputs to economic models of follicular lymphoma, based on National Institute for Health and Clinical Excellence (NICE) appraisals, to assess the cost-effectiveness of rituximab. The first two, of rituximab as maintenance[6] and first-line therapy[7] respectively, directly replicated the economic models from these appraisals (costs, health state utilities etc).
In a further analysis, the results of the ongoing PACIFICO trial were simulated to generate predictions of cost-effectiveness. This case study more closely replicates how this method might be implemented in practice, with the need to estimate costs and clinical effectiveness before phase III trial results are available. We also conduct a value of information analysis, to estimate the value of future research, based on reducing the uncertainty in results.
Results:Our analyses suggest an acceptable degree of concordance between simulation- and trial-based estimates of cost-effectiveness. For first-line and maintenance therapy, deviations of £2,099 and £1,355 per quality-adjusted life-year (QALY), respectively, from trial-based incremental cost-effectiveness ratio (ICER) estimates of £8,290 and £7,721 per QALY gained would not affect reimbursement decisions.
The probability of these rituximab-containing regimens being cost-effective at £20,000 and £30,000 per QALY thresholds was 1 for both first-line and maintenance therapy in both simulated and trial-based analyses. The range of cost-effectiveness thresholds over which more than 5% of simulations give different results between trial and simulation based analyses is £3,247-£16,256/QALY for maintenance therapy and £6,168-£13,872/QALY for first-line therapy.
Sensitivity analyses, performed to quantify the relative impact of different sources of parameter uncertainty on cost-effectiveness, demonstrated that changes in individual parameters resulted in deviations that were very similar in both trial and simulated methods.
For the PACIFICO case study, the ICER for R-FC versus R-CVP is £19,950/QALY, with an 80% probability of being cost-effective at £30,000/QALY threshold for cost-effectiveness.
Conclusions:Our analyses demonstrate the feasibility of mechanism-based economic analyses. Trial-based and PKPD-based estimates of cost-effectiveness were concordant, and decision uncertainty (the probability of cost-effectiveness at the payer's threshold) was equivalent. The deviations between simulation- and trial-based estimates of cost-effectiveness are no greater than between analyses based on different clinical trials.
Such an analysis may have utility during drug development in: (i) assessing the effect on cost-effectiveness of considering different sub-groups or dosing schedules; (ii) exploring the impact of protocol deviations; (iii) determining pricing structures, particularly in the context of value-based pricing. A pharmacokinetic-pharmacodynamic-pharmacoeconomic approach has distinct advantages over conventional economic evaluations which are empirical, and generally reliant on the results of phase III trials.
There are limitations to our approach: the results of such an analysis will be inherently uncertain, because of the extensive parameterisation of both the pharmacological and economic models. However, this uncertainty can be studied using value of information analysis, which quantifies the cost of reducing parameter uncertainty. Early indications of cost-effectiveness can thus be used to direct future research based on these costs, both by informing the design of phase III trials, and indicating important parameters for accurate quantification.
The population PKPD-based approach described here is consistent with Sheiner's "learning and confirming" paradigm for the clinical phases of drug development and consequently might help facilitate a coordinated modelling approach across pharmaceutical industry Research & Development, Pricing & Reimbursement, Health Economic & Outcomes Research, and Strategic Planning sections.
References:
[1] Holford N, Ma SC, Ploeger BA. Clinical trial simulation: a review. Clin Pharmacol Ther 2010; 88 (2): 166-182.
[2] Hughes DA, Walley T. Economic evaluations during early (phase II) drug development: a role for clinical trial simulations? Pharmacoeconomics 2001; 19: 1069-1077.
[3] Pettitt A. Purine-alkylator combination in follicular lymphoma immuno-chemotherapy for older patients. International Standard Randomised Controlled Trial Number Register - ICRCTN99217456 <http://www.controlled-trials.com/ISRCTN99217456/pacifico> 2009.
[4] Ng CM, Bruno R, Combs D, et al. Population pharmacokinetics of rituximab (anti-CD20 monoclonal antibody) in rheumatoid arthritis patients during a phase II clinical trial. J Clin Pharmacol 2005; 45: 792-801.
[5] Ternant D, Hénin E, Cartron G, et al. Development of a drug-disease simulation model for rituximab in follicular non-Hodgkin lymphoma. Br J Clin Pharmacol 2009; 68: 561-573.
[6] Roche. Rituximab for the treatment of relapsed follicular lymphoma. Single Technology Appraisal submission to the National Institute for Health and Clinical Excellence. <http://www.nice.org.uk/nicemedia/live/11730/38897/38897.pdf> 2007.
[7] Liverpool Reviews and Implementation Group. Rituximab for the first line treatment of stage III-IV follicular non-Hodgkin's lymphoma. Evidence Review Group Report for the National Institute for Health and Clinical Excellence <http://www.hta.ac.uk/erg/reports/1636.pdf> 2006.
