Increasingly, economic data are being considered in formulary decisions. In oncology, pharmacoeconomic evaluations are essential to help decision makers weigh the associated costs and outcomes of competing
Increasingly, economic data are being considered in formulary decisions. In oncology, pharmacoeconomic evaluations are essential to help decision makers weigh the associated costs and outcomes of competing chemotherapeutic interventions. In this article, we present a four-step pharmacoeconomic research model that can be customized for specific provider or payer systems. The model encompasses problem identification, clinical management analysis, three pharmacoeconomic analyses (cost consequence, expected cost, and cost effectiveness), and a sensitivity analysis-the rank order stability analysis (ROSA)-to validate the findings.
Health care professionals, policymakers, and formulary committee members are increasingly asked to augment formulary consideration of drug efficacy and safety with economic information. In this era of dwindling health care budgets, pharmacoeconomic analysis facilitates choices, in terms of overall outcomes between therapies competing for the same health care resources. This method of economic evaluation informs the programmatic medical decision maker of the appropriateness and value of health care procedures, including drugs [1]. In no other field has the necessity and relative value of such analyses proved to be more applicable than in oncology.
Comparisons of competing new chemotherapeutic interventions warrant a pharmacoeconomic evaluation to weigh associated costs and outcomes and compare these "offsets" with those of traditional therapy. Our four-step research approach described in this article is a flexible economic model for evaluating costs and consequences that
will accommodate provider-specific parameters.
One very useful analytic component of such a model, discussed below, is cost-effectiveness analysis, which compares the health effects of a treatment strategy with the resources that must be invested to adopt the strategy [2]. Comparison of effectiveness in cancer is particularly difficult, since the criteria for diagnosis vary with pathologists, and the criteria for prognosis vary with the extent of disease [3]. The ability to customize the economic model for specific provider or payer systems is essential in generating valid and generalizable pharmacoeconomic data.
Authoritative pharmacoeconomic research requires a coherent data set to ensure an effective valuation of costs and outcomes. Often, assumptions and biases are immersed within a data set, thereby compromising the integrity of an analysis. To compensate for this intrinsic uncertainty, the four-step model incorporates a comprehensive sensitivity analysis, described below, to identify "cost drivers" and specific points of model instability.
It should be emphasized that pharmacoeconomics is a prescriptive science, employed to facilitate choices in allocating scarce resources. When properly executed and validated, economic research studies provide essential information as input into the decision-making process [1]. Traditional considerations, such as safety, efficacy, equity, and access, should continue to serve as inputs into medical decisions.
Worldwide inflation of health care budgets has prompted many cost investigations, predominantly emphasizing drug therapy. New chemotherapeutic interventions have provided monumental improvements in patient care, but not without associated increases in drug therapy costs. Clinically significant outcomes (which may be curative, palliative, or preventive) require a pharmacoeconomic analysis to quantify cost-effectiveness of therapy. Desired outcomes attributed to new therapies may be appreciated by the provider, patient, and payer alike. An effective comparison of competing therapies warrants a valuation of desired and undesired outcomes of therapy, to ensure that a marginal outcome merits incremental cost.
Chemotherapy has resulted in improved palliation of symptoms, higher response rates, and extended time to treatment failures. These successful clinical results are allied with financial benefits that may be realized by the provider (eg, hospital), third-party payer (eg, Medicare), or patient. From the perspective of a provider or third-party payer, the cost savings generated by chemotherapy are partially derived from patients who have failed on primary therapy. The decreased follow-up care costs associated with treating failed therapy patients may be directly quantified as cost avoidance. Successful treatment, or successful period of care, may also be evaluated from the patient perspective as represented by quality of life. Cost effectiveness, possibly measured in terms of cost per event-free days, may provide a retrospective proxy for patient quality of life.
A longer-term research objective is to prospectively examine the economic implications of competing therapeutic interventions. Better efficacy and longer effective treatment periods translate into optimal resource utilization and improved patient quality of life. Validated survey instruments are employed to evaluate physical, social, and emotional aspects of a patient's well-being that are relevant and important to the patient [1]. This prospective technique will allow for a determination and financial quantification of patient disease-free periods, further demonstrating cost-effective or cost-utility benefits.
The four steps in our research approach are (1) problem identification, (2) clinical management analysis, (3) pharmacoeconomic analysis, and (4) rank order stability analysis (ROSA). This Four-Step Pharmacoeconomic Model, developed 2 years ago, has proven to be the state-of-the-art prototype for pharmacoeconomic research [4]. It has been successfully utilized in numerous cost-effectiveness studies worldwide. The flexibility and versatility of this model has enabled researchers to apply it to several diverse clinical areas, including chemotherapy, dermatology, psychology, anesthesiology, neuromuscular blockade, imaging (contrast media utilization), and contraceptives.
Step 1: Problem Identification
The problem identification segment of our study model dictates the key research issues associated with the pharmacoeconomic comparison of competing therapeutic interventions. The following issues must be explicitly addressed before beginning pharmacoeconomic evaluations:
Research Question-As in any scientific research project, the research question must be stated initially. Inherent in the research question are study objectives to map the path for future methodology decisions.
Research Perspective-The perspectives of a pharmacoeconomic study will prescribe the definition and valuation of cost and outcomes for the remaining analyses. All studies should begin with the society perspective and then be clearly subdivided for all relevant decision-making parties [1].
Analytic Time Horizon-The analytic time horizon should also be explicitly delineated during this phase. This time frame ideally extends far enough into the future to capture the major clinical and economic outcomes.
Treatment Comparators-The drug treatment should be compared with existing practice, relevant historical comparators, and minimum practice [1].
Analytic Technique-Researchers should identify the relevant and applicable pharmacoeconomic tools to be employed when analyzing the competing therapies.
Outcome and Cost Measures-Outcome measures should be identifiable, quantifiable, and consistent with current medical practice. Cost measures should be identifiable, quantifiable, and consistent with current hospital accounting practice. Both of these utility measures must be consistent and applicable to the defined perspective; that is, certain cost and outcome measures will be included or excluded, depending on the perspective. One traditional paradigm is to include measures that will influence the target audience's decision in choosing between therapies. Intuitively, this is a subjective decision, thus warranting a sensitivity analysis of alternative measures.
Data Sources-Clinical and economic data sources are defined after appropriate analytic techniques, costs, and outcomes are delineated. Data sources may range from retrospective analysis of government databases and the literature to prospective economic clinical trials and "time and motion" studies.
Potential Limitations-Potential limitations of the research study should be discussed during the problem identification segment of the analysis. The analytic environment described above already represents one limitation of data generalizability. Pharmacoeconomic studies are encouraged to investigate local and global cost effectiveness (ie, all relevant comparators, analytic perspectives, and health care systems) of the drug therapy in question [1].
Step 2: Clinical Management Analysis
The clinical management analysis involves three steps: clinical appraisal, decision-analytic model construction, and the application of a clinical input (eg, metaanalysis) for clinical safety and efficacy outcomes. The objective of this step in our research approach is to develop a clinical and economic profile of the standards of practice for the disease management topic of interest in various health care environments (eg, a managed care setting).
Clinical Appraisal-A clinical appraisal is performed by interviewing clinicians with structured instruments to establish patient profiles and to define health care resources consumed in the implementation of therapy and in the management of adverse events.
Decision-Analytic Model-A decision-analytic model is constructed to reflect standards of care and project uncertainty for competing treatment modalities. The decision-analytic model accounts for all therapeutic pathways and outcomes, probability data for obtaining each outcome, and the cost of each outcome. This chronologic arrangement of therapy events and outcomes will actuate the subsequent pharmacoeconomic analyses.
This model should then be programmed in a spreadsheet format to accommodate user-specific inputs. An expert panel of clinicians, economists, and pharmacists will be instrumental in developing and reviewing this model. The model that will be developed during clinical management analysis may be adapted to a particular institutional setting. For example, to study an aspect of health care delivery-variations in practice patterns-would require data on the following parameters: physician procedures, drug therapy, ancillary services, adverse events management, inpatient services, and outpatient services. To review variations in provider environments of a health care system, the model would consider the providers' payer mix, reimbursement schedules, prescription fees, and treatment restrictions.
Metaanalysis-After appraising clinical algorithms and then customizing them to specific health care providers and systems, the clinical input(s) must be identified. To capture this clinical data initially, and simulate real-world practice patterns, we recommend the use of metaanalysis.
Metaanalysis is a systematic method for finding, evaluating, and combining results from different scientific studies. It mathematically aggregates therapeutic success rates and provides summary statistics based on weighted averages. It considers sample size in each study, as well as between-study differences, in providing an overall summary estimate. We suggest that results be combined using the method of DerSimonian and Laird, modified for single group analysis, as presented by Velanovich (1991) [6]. This method produces a sample-size weighted-average value for each rate for each comparator, along with a standard error so that a 95% confidence interval may be constructed (95% CI = the mean ± 1.96 × SE). Einarson and colleagues state that metaanalysis is useful for integrating independent research results, but they also emphasize the importance of proper application [5].
An expert panel of clinicians should be assembled to assess and critique all ensuing phases of the analysis. Both the protocol and results of the metaanalysis should be compared with the opinions of the clinical experts.
Step 3: Pharmacoeconomic Analyses
The pharmacoeconomic analysis section of our research approach determines the expected cost and benefits of competing therapeutic interventions. In doing so, costs and probabilities of desired and undesired outcomes are considered. We first identify the cost inputs (ie, parameters that influence an economic analysis associated with the drug therapy) and outcomes (ie, therapy consequences identified in the clinical management analysis). We routinely employ the following pharmacoeconomic analyses: cost consequence, expected cost, and cost effectiveness.
Cost-Consequence Analysis-This analysis is comprised of two distinct research assessments: cost identification and outcome identification. Cost identification entails delineating all treatment parameters and then measuring the respective cost of each input. The definition of "cost" is sometimes ambiguous and will depend on the perspective defined in the clinical management analysis. For example, the allocated portion of cost may be of interest to the provider but not the third-party payer. The cost identification or cost of regimen analysis includes a calculation of the cost of drug therapy, medical care, facility use, and management of adverse events.
Drug therapy cost should not be misinterpreted as acquisition cost only (ie, the purchase cost of a drug to an institution, pharmacy, or patient). This parameter may also include the cost of labor time, overhead, supplies, equipment, and wastage associated with the preparation and administration of a drug regimen.
The cost of medical care is determined by calculating all costs concerning routine procedures and tests administered to the patient. It includes physician services and any ancillary service expended in treating the patient. Capital, overhead, labor, and related operating costs may all be considered.
Inpatient and outpatient services are also considered in this full costing approach. Time spent in routine care units, intensive care units, intermediate care units, rehabilitation units, outpatient clinics, etc, must be tracked and costed accordingly. Ideally, these costs are disaggregated into variable, fixed, and mixed classifications, to more precisely account for volume or service mix changes.
Adverse event management costs are calculated on a component basis for each adverse effect. Total management costs, inclusive of physician and hospital costs, are multiplied by percent incidence and percent treated.
After identifying cost parameters, outcomes and consequences must then be defined and measured. These definitions should be consistent with present standards of care, quantifiable, and valuable to the medical decision maker who is allocating health care resources.
Cost-consequence analysis aggregates these cost and outcome data, but does not weigh the parameters for the end-user of the analysis. The decision maker must interpret the data and infer any interrelationships. All costs and outcomes identified should be calculated as increments and totals, to be used in the subsequent pharmacoeconomic analysis.
Expected-Cost Analysis-This pharmacoeconomic calculation represents a hybrid analysis of competing therapies that incorporates efficacy and cost data defined in the above-mentioned analyses. Decision analytic principles form the underlying methodologic structure. Decision analysis is an explicit quantitative approach for choosing among competing strategies under conditions of uncertainty. The process involves calculating and summing expected utilities for each therapy. To do so, all patient outcomes are identified, valued, and classified for relevance to each analytic perspective.
A decision tree that depicts the information is used in the analysis. Each branch represents an outcome to which a probability is assigned. Probabilities may be success rates, failure rates, or adverse event incidence rates. This prescriptive tool provides a structure for depicting relationships between actions (therapeutic interventions) and their possible consequences (outcomes).
Future outcomes and costs should both be discounted, preferably at equivalent rates, as a baseline to be varied in sensitivity analyses [1]. The result of this systematic approach to manage uncertainty is an expected cost of competing therapies.
Cost-Effectiveness Analysis-This type of pharmacoeconomic analysis compares therapeutic interventions that generate common health outcomes. Cost-effectiveness analysis expresses outcomes as physical units, such as life-years saved. Cost-identification and cost-consequence analyses financially quantify the care rendered to the patient and the resulting treatment consequences. The patient perspective has yet to be considered in either the cost-consequence or expected-cost analysis. In a retrospective analysis, it is difficult to assess patient desirability of a specified health outcome. As a proxy, we have applied event-free days as our effectiveness measure in our oncology studies.
Certain situations will merit an incremental cost-effectiveness analysis in which marginal effectiveness must be weighed against incremental or decremental cost associated with an extra unit of health outcome achieved. In this case, the willingness of a provider or third-party payer to pay must be approximated as some threshold value. A contingent valuation may be employed to evaluate the appreciated benefit of a therapy to the decision maker. This debated calculation represents a component of cost-benefit analysis.
Health-related quality of life (HRQOL) or quality adjusted life-years (QALYs) represent a better gauge of patient preferences and are most commonly determined prospectively. Cost-utility analysis measures benefits in these utility units and utility-adjusted life-years.
Step 4: Rank Order Stability Analysis (ROSA)
Pharmacoeconomic analyses routinely conclude with some type of variability measure. This calculation is necessary to ensure the validity of the data in question. We recommend a comprehensive sensitivity analysis to identify "cost drivers" and specific points of model instability.
Rank order stability analysis (ROSA) is an approach for examining the sensitivity of pharmacoeconomic analyses. It is a comprehensive and clear method for validating results based on estimates and, in effect, a break-even analysis that identifies the specific point of insensitivity for all parameters in the pharmacoeconomic model. ROSA includes the following steps:
1. Identification of outcome of interest
2. Identification of input parameters to use as variables
3. Calculation of upper and lower limits of robustness
4. Calculation of parameter elasticities
5. Calculation of 95% confidence intervals for clinical rates
Substantial alteration in the analytic results should induce further examination of relevant data to ascertain the actual value of the uncertain variable. If study conclusions are upheld or remain stable with ROSA, a higher degree of validity is assigned to the analysis. However, these results do not represent impregnable conclusions and should not be construed in such a fashion. Regardless of the amplitude of instability, the relevant range of all parameter values should be stated explicitly in the study report. Only then may the decision maker interpret and utilize the information to more efficiently allocate health care resources.
1. Guidelines for Economic Evaluation of Pharmaceuticals. Canada, Draft 1.4, June 27, 1994. Center for Health Outcomes and Economics, East Brunswick, New Jersey.
2. Peck JR: How to evaluate drugs: Cost-effectiveness analysis. JAMA 264:83-84, 1990.
3. Lamm DL, Blumenstein BA, Crawford ED, et al: A randomized trial of intravesical doxorubicin and immunotherapy with Bacillus Calmette-Guerin for transitional-cell carcinoma of the bladder. N Engl J Med 325:1205-1209, 1991.
4. Clayton YM, Hay RJ: Epidemiology of fungal skin and nail disease: Roundtable discussion held at Dermatology 2000. Br J Dermatol 130(suppl 43):9-11, 1994.
5. Einarson TR, McGhan WF, Bootman JL, et al: Meta-analysis: Quantitative integration of independent research results. Am J Hosp Pharm 42: 1357, Sept 1985.
6. Velanovich V: Meta-analysis for combining Bayesian probabilities. Med Hypotheses 35 (3) 192-195, July 1991.