Article type
Year
Abstract
Background: Increasingly, healthcare economic evaluation is being conducted alongside multinational trials to inform reimbursement decisions at country/jurisdiction level. These studies often combine efficacy and resource use data from all participating countries to produce a single measure of cost-effectiveness, which is considered to be generalisable. It is argued that such a practice is inappropriate. Even in the absence of significant interaction between country and treatment, decision-makers may be interested in country specific cost-effectiveness ratios.
Objectives: Using alternative statistical models, we considered several assumptions concerning differences among countries in resource utilisation and examined the impact on cost-effectiveness estimates.
Methods: We used data from a multinational trial of an antibiotic in community-acquired pneumonia. Regression methods are used to determine treatment impact on cure and resource use by country, controlling for baseline characteristics. This framework is extended
Results: There was no statistically significant interaction between country and treatment on resource utilisation. For 4 countries with unit prices available, there were mean savings estimated using combined resource data, and allowing for interactions between country and treatment effect on resource use. When the outcome was valued 2,000 per additional patient cured, cost-effectiveness ranged from 373 to 488 for pooled data across 4 countries; from 270 to 544 when allowing for an interaction between country and cure on resource utilisation; from 251 to 1,194 when allowing for interactions by cure and by treatment, independent of cure; and from 111 to 981 when incorporating random effects for centres. Cost-effectiveness results variability increased with increasing levels of country-specificity. Between country variability could be partly explained by differences in case mix and centres.
Conclusions: Allowing for different impacts of treatment on resource use may uncover important differences in cost-effectiveness across countries, even in the absence of statistically significant interaction between countries and treatment on resource use. When an economic evaluation is conducted alongside a multinational clinical trial, combining resource use data may be an option, but uncertainty around the results and variability between countries are likely to be underestimated if exploring interactions between treatment and resource utilisation is not considered.
Objectives: Using alternative statistical models, we considered several assumptions concerning differences among countries in resource utilisation and examined the impact on cost-effectiveness estimates.
Methods: We used data from a multinational trial of an antibiotic in community-acquired pneumonia. Regression methods are used to determine treatment impact on cure and resource use by country, controlling for baseline characteristics. This framework is extended
Results: There was no statistically significant interaction between country and treatment on resource utilisation. For 4 countries with unit prices available, there were mean savings estimated using combined resource data, and allowing for interactions between country and treatment effect on resource use. When the outcome was valued 2,000 per additional patient cured, cost-effectiveness ranged from 373 to 488 for pooled data across 4 countries; from 270 to 544 when allowing for an interaction between country and cure on resource utilisation; from 251 to 1,194 when allowing for interactions by cure and by treatment, independent of cure; and from 111 to 981 when incorporating random effects for centres. Cost-effectiveness results variability increased with increasing levels of country-specificity. Between country variability could be partly explained by differences in case mix and centres.
Conclusions: Allowing for different impacts of treatment on resource use may uncover important differences in cost-effectiveness across countries, even in the absence of statistically significant interaction between countries and treatment on resource use. When an economic evaluation is conducted alongside a multinational clinical trial, combining resource use data may be an option, but uncertainty around the results and variability between countries are likely to be underestimated if exploring interactions between treatment and resource utilisation is not considered.