OBJECTIVES

Although economic evaluations are increasingly being used in healthcare decision-making, their statistical quality is far from optimal. Oftentimes, baseline imbalances, skewed costs, the correlation between costs and effects, and missing data are not adequately accounted for, leading to possible biased results. The current study aims to evaluate the impact of using different statistical methods on results of trail-based economic evaluations.

METHODS

Data from REALISE study were used (n=169 low back pain patients). 14 economic evaluations were performed, in which more advanced statistical methods were applied, step by step. In the simplest approach, cost and effect differences were estimated using t-tests and only patients with complete cost and effect data were included. In the most advanced approach, baseline imbalances were accounted for using regression-based adjustment, skewed cost data were accounted for using bias-corrected and accelerated bootstrapping, the correlation between costs and effects was accounted for using “seemingly unrelated regression”, and missing data were multiply imputed. For all approaches, incremental costs and QALYs, confidence intervals around incremental costs and QALYs, incremental cost-effectiveness ratios (ICERs) and cost-effectiveness accessibility curves were estimated and compared.

RESULTS

The ICER in the simplest approach, was €636,744 and €-48,097 in the most advanced approach. The probability of the intervention being cost-effective at a willingness-to-pay of €0/QALY gained was 0.67 in the simplest approach and 0.70 in the most advanced approach. Adjusting for: baseline imbalances and missing data, changed the point estimates; skewness and missing data, had an impact on the level of uncertainty; correlation between cost and effects, had no large implications.

CONCLUSIONS

This study suggest that failure to adequately account for baseline imbalances, skewed costs, correlated costs and effects, and missing data in trial-based economic evaluations may under- or overestimated cost-effectiveness results. Further research is warranted to check if these results are generalizable in different settings.