Modeling in economic evaluation: an unavoidable fact of life

Cost-effectiveness analyses using real-world data: an overview of the literature

Objectives: Real-world evidence (RWE) may provide good estimates of absolute event probabilities and costs in patients in actual clinical practice, but their use in decision-analytic models poses many challenges. A literature review based on a systematic search was conducted to summarize the limitations of using RWE in decision-analytic modeling reported in the literature, but also to identify existing recommendations about real-world modeling. Methods: A literature search was performed on Medline and Embase databases, as well as relevant websites. No restrictions in language or geographical scope were imposed. Results: A total of 14 references were included. RWE is recognized as a valuable source of data for market access and reimbursement, and as a complement to clinical trial evidence for treatment pathways, resource use, long-term natural history, and effectiveness. The main limitations identified in the literature were: confounding bias, missing data, lack of accurate data related to drug exposure and outcomes, errors during the record-keeping process, protection of private data, and insufficient numbers of patients. Although most submission guidelines recognized the potential biases associated with RWE, guidance on the appropriate methods to deal with these biases, and approaches to review different relevant evidence to inform model development, were scarce. Several initiatives have attempted to provide guidance on the use of RWE in decision-modeling. Conclusions: RWE is likely to be particularly valuable for informing healthcare policy-makers when formulating appropriate treatment pathways, encouraging the optimal allocation of scarce resources, and improving aggregate patient outcomes. However, little guidance is available on the relative merits of using efficacy and/or effectiveness evidence in Health Technology Appraisal submissions. Further research is needed to better understand these methods and their potential applications in a broader range of scenarios and simulation studies, and their impact on economic modeling.

A Critique and Impact Analysis of Decision Modeling Assumptions

Background . Numerous guidelines have been published defining good practice for the conduct of economic evaluations in general and model-based evaluations in particular. The extent to which guidelines are accepted is unknown, and the impact of deviations from good practice is not generally recorded. The authors identified 4 specific issues in applied studies that may affect the accuracy and comparability of different evaluations. Methods . A descriptive analysis of 4 modeling issues (inclusion of incident cases over a model time horizon, appropriate time horizon, parsimonious model structure, and the handling of age-specific subgroups) is presented. A case study model is analyzed to illustrate the quantitative impact of 3 of the issues. Results . In the case study model, alternative specifications of the modeling framework are shown to alter the estimated cost-effectiveness by large percentages. The combined effect of including incident cases and reduced follow-up yielded the highest divergence from the reference case results, by between 20% and 40%, depending on the age group. Reference case results of an age-weighted population were almost 14% different from the middle single age cohort. Discussion . The identified issues are all generalizable to a wide range of treatment areas and are, or should be, addressed by evaluative guidelines. The authors call for the continued development, dissemination, and application of guidelines for the conduct of economic evaluation in general and model-based economic evaluations in particular.

Cost-Effectiveness of a Lidocaine 5% Medicated Plaster Relative to Gabapentin for Postherpetic Neuralgia in the United Kingdom

BACKGROUND

Approximately 50% of elderly patients develop postherpetic neuralgia (PHN) after herpes zoster infection (shingles). A lidocaine 5% medicated plaster marketed in the United Kingdom in January 2007 has been shown to be an effective topical treatment for PHN with minimal risk of systemic adverse effects.

OBJECTIVE

This paper assessed the cost-effectiveness of using a lidocaine plaster in place of gabapentin in English primary care practice to treat those PHN patients who had insufficient pain relief with standard analgesics and could not tolerate or had contraindications to tricyclic antidepressants (TCAs). The analysis took the perspective of the National Health Service (NHS).

METHODS

The costs and benefits of gabapentin and the lidocaine plaster were calculated over a 6-month time horizon using a Markov model. The model structure allowed for differences in costs, utilities, and transition probabilities between the initial 30-day run-in period and maintenance therapy and also accounted for add-in medications and drugs received by patients who discontinued therapy. Most transition probabilities were based on non-head-to-head clinical trials identified through a systematic review. Data on resource utilization, discontinuation rates, and add-in or switch medications were obtained from a Delphi panel; cost data were from official price tariffs. Published utilities were adjusted for age and were supplemented and validated by the Delphi panel.

RESULTS

Six months of therapy with the lidocaine plaster cost pound 549 per patient, compared with pound 718 for gabapentin, and generated 0.05 more quality-adjusted life-years (QALYs). The lidocaine plaster therefore dominated gabapentin (95% CI, dominant- pound 2163/QALY gained). Probabilistic sensitivity analysis showed that there was a 90.15% chance that the lidocaine plaster was both less costly and more effective than gabapentin and a 99.99% chance that it cost < pound 20,000/QALY relative to gabapentin. Extensive deterministic sensitivity analyses confirmed the robustness of the conclusions.

CONCLUSION

This study found that the lidocaine 5% medicated plaster was a cost-effective alternative to gabapentin for PHN patients who were intolerant to TCAs and in whom analgesics were ineffective, from the perspective of the NHS.

Populating decision-analytic models: The feasibility and efficiency of database searching for individual parameters

Objectives: The aim of the study was to investigate the feasibility and effectiveness of searching selected databases to identify information required to populate a decision-analytic model.

Methods: Methods of searching for information to populate a decision-analytic model were piloted using a case study of prophylactic antibiotics to prevent recurrent urinary tract infections in children. This study explored how the information requirements for a decision-analytic model could be developed into searchable questions and how search strategies could be derived to answer these questions. The study also assessed the usefulness of three published search filters and explored which resources might produce relevant information for the various model parameters.

Results: Based on the data requirements for this case study, 42 questions were developed for searching. These questions related to baseline event rates, health-related quality of life and outcomes, relative treatment effects, resource use and unit costs, and antibiotic resistance. A total of 1,237 records were assessed by the modeler, and of these, 48 were found to be relevant to the model. Search precision ranged from 0 percent to 38 percent, and no single database proved the most useful for all the questions.

Conclusions: The process of conducting specific searches to address each of the model questions provided information that was useful in populating the case study model. The most appropriate resources to search were dependent on the question, and multiple database searching using focused search strategies may prove more effective in finding relevant data than thorough searches of a single database.

Pharmacoeconomic investigations in intensive care

Financial constraints in health care in general and critical care services in particular have placed increasing demands upon health care professionals and decision-makers to ensure the highest quality of care with the best possible outcomes are attained for the least possible expenditure of resources. As such, pharmacoeconomic methods have become increasingly applied to pressing clinical problems to facilitate cost-effective delivery of health care services to the critically ill. This manuscript briefly details the basic tenets of pharmaco-economic analysis and provides an overview of recent applications to critical care issues.

Changing the first line drug for malaria treatment--cost-effectiveness analysis with highly uncertain inter-temporal trade-offs

Access to effective treatment would substantially reduce the burden of malaria in sub-Saharan Africa, but resistance to chloroquine, the most commonly used first line drug, is now widespread. There has been considerable debate over the level of chloroquine resistance at which a new first line drug should be adopted. Two issues make this an extremely complex decision: it involves trade-offs in costs and health outcomes over time; and many of the parameters are uncertain. A modelling approach was identified as appropriate for addressing these issues. The costs and effects of changing from chloroquine to sulphadoxine-pyrimethamine (SP) as the first line drug were modelled over 10 years, allowing for growth in drug resistance. Probabilistic sensitivity analysis was used to allow for the high levels of parameter uncertainty. The optimal year of switch was highly dependent on both empirical values, such as initial resistance and resistance growth rates, and on subjective values, such as the time preferences of policy-makers. It was not possible to provide policy-makers with a definitive threshold resistance level at which to switch, but the model can be used as an analytical tool to structure the problem, explore trade-offs, and identify areas for which data are lacking.

The economics of selective serotonin reuptake inhibitors in depression: a critical review

The prevalence of depression and the high costs associated with its treatment have increased interest in pharmacoeconomic evaluations of drug treatment, particularly in the 1990s as the use of selective serotonin (5-hydroxytryptamine; 5-HT) reuptake inhibitors (SSRIs) expanded substantially. This review presents results from specific studies representing the key study designs used to address the pharmacoeconomics of SSRI use: retrospective administrative database analyses, clinical decision analysis models, and randomised clinical trials. Methodological considerations in interpreting results are highlighted. In retrospective administrative database analyses, most comparisons have been made between SSRIs and tricyclic antidepressants (TCAs). A few studies have addressed differences between SSRIs. The studies focused on healthcare cost (to payer) and cost-related outcomes (e.g. treatment duration, drug switching). Although SSRIs are generally associated with higher drug acquisition costs than are TCAs, total healthcare costs are at least offset, if not decreased, by reductions in costs associated with use of SSRIs. Although studies from the early 1990s show some advantage for fluoxetine, the results are limited by use of data from shortly after the introduction of paroxetine and sertraline; studies from the mid- 1990s on that compare drugs within the SSRI class show general equivalence in terms of cost. Important methodological advances are occurring in retrospective studies, with selection bias and other design limitations being addressed statistically. Clinical decision analysis models permit flexibility in terms of ability to specify different alternative treatment scenarios and varying durations. Sensitivity analysis aids interpretability, although model inputs are limited by data availability. Results from short term (1 year duration or less) studies comparing SSRIs and TCAs suggest that SSRIs are more cost effective or that there is no difference. Longer term studies (lifetime Markov models) focus more on the impact of maintenance antidepressant therapy and show more mixed results, generally favouring SSRIs over TCAs. The results indicate that the effect of SSRIs is mainly through prevention of relapse. Important assumptions of these models include fewer serious adverse effects and lower treatment discontinuation rates with SSRIs. Naturalistic clinical trials provide greater generalisability than traditional randomised clinical trials. One naturalistic trial found that nearly half of TCA-treated patients switched to another antidepressant within 6 months; only 20% of SSRI-treated patients switched. Cost differences between groups were minimal. These studies indicate few differences in medical costs, depression outcomes and health-related quality of life between TCAs and fluoxetine, although fewer fluoxetine-treated patients switched treatment.

Economic evaluations relating to diabetes: a descriptive review and their compliance with guidance

From a search of the Health Economics Evaluations Database (HEED), 301 studies relating to diabetes-related interventions were identified. Of these, 223 studies contained some original data, i.e., were applied studies. Over the 1990s, the number of studies undertaken in this area appears to be increasing year on year, and more of the studies are applicable to the USA than to all other countries combined. Most studies are peer reviewed, the main type of analysis is cost-consequence, and the interventions most often assessed are pharmaceutical and care services. Both the public and the private sectors sponsor many of these studies. However, only a fraction of these studies appear to accord with guidance for good economic evaluation. For example, when the search criteria were tightened to capture only those studies that consider both outcomes and costs, 173 studies were identified. Narrowing the search criteria to capture only those studies that were based on randomized controlled trial (RCT) or modeling data-the methods recommended by the US Panel and the National Institute of Clinical Excellence (NICE)-reduced the sample to 56 studies. With specific focus on those studies that evaluated pharmaceuticals, only the modeling approach appeared to show any compliance at all with another key recommendation of the existing guidance on methods; that of using outcome measures that are applicable to a long-term time horizon. Though modeling does offer certain advantages over RCTs (for example, a larger range of comparators can be more readily incorporated, long-term outcome measures can be more easily and inexpensively estimated), modeling results have less internal validity. Also, modeling in itself relies on good RCT data. Therefore, the results of economic evaluations should be interpreted with care, irrespective of whether they are principally based on RCT or model-based methods.

Testing the validity of cost-effectiveness models

A growing body of recent work has identified several problems with economic evaluations undertaken alongside controlled trials that can have potentially serious impacts on the ability of decision makers to draw valid conclusions. At the same time, the use of cost-effectiveness models has been drawn into question, due to the alleged arbitrary nature of their construction. This has led researchers to try and identify ways of improving the quality of cost-effectiveness models through identifying 'best practice', producing guidelines for peer review and identifying tests of validity. This paper investigates the issue of testing the validity of cost-effectiveness models or, perhaps more appropriately, whether it is possible to objectively measure the quality of a cost-effectiveness model. A review of the literature shows that there is much confusion over the different aspects of modelling that should be assessed in respect to model quality, and how this should be done. We develop a framework for assessing model quality in terms of: (i) the structure of the model; (ii) the inputs of the model; (iii) the results of the model; and (iv) the value of the model to the decision maker. Quality assessment is investigated within this framework, and it is argued that it is doubtful that a set of objective tests of validity will ever be produced, or indeed that such an approach would be desirable. The lack of any clearly definable and objective tests of validity means that the other parts of the evaluation process need to be given greater emphasis. Quality assurance forms a small part of a broader process and is best implemented in the form of good practice guidelines. A set of key guidelines are presented.