OBJECTIVES

Korea, Japan and China have updated their pharmacoeconomic evaluation (PE) guidelines in 2011, 2019 and 2019 respectively. This study aims at comparing these recent updates.

METHODS

Each guideline was extracted and summarized using the comparative table of pharmacoeconomic guidelines developed by ISPOR; characteristics were compared.

RESULTS

The guidelines all considered the following elements: perspective, target population, comparator, time horizon, analytical technique, costs to be included, modelling, outcome measure, method to derive utility, discount and sensitivity analysis. While absent from the Japanese guidelines, guidance on generalizability and budget impact analysis was presented in Korean and Chinese guidelines.

Similar guidance was found for target population (consistent with indications), comparator (most frequently used treatment) and time horizon (long enough). However, differences were identified regarding perspectives (China: both societal and healthcare system; Japan: public healthcare payer; Korea: limited societal), and discount rates (China: 5% [0-8%]; Japan: 2% [0-4%]; Korea: 3% [3%-7.5%]). Japanese and Chinese guidelines mentioned a preference for cost-utility analysis (CUA) considering quality-adjusted life year (QALY) as the health outcome, while Korean guidelines did not mention any preference between CUA and cost-effectiveness analysis (CEA). Regarding the responder for quality of life scores, the guidelines in Japan and Korea suggested general public, while the guidelines in China suggested patients.

CONCLUSIONS

The PE guidelines in the China, Korea and Japan present similar guidance in most of the aspects; differences were identified regarding perspective, discount rates, type of analysis, and preferred responder for the estimation of quality of life score. Overall, these three countries have similar PE framework, therefore researchers working on PE across these countries should pay more attention to the other aspects, such as whether the model structures represent the local situation and whether the model inputs are representative.