ASSESSING THE COST-EFFECTIVENESS OF HBA1C TESTING IN THE DIAGNOSIS OF TYPE II DIABETES AND THE IMPACT OF AN IMPERFECT DIAGNOSTIC REFERENCE STANDARD

Purpose: To conduct an economic evaluation as part of a health technology assessment to determine whether public funding of glycated haemoglobin (HbA1c) testing is justified in Australia.

Method: A Markov model was developed to simulate the long term costs and benefits (measured in quality-adjusted life years) of HbA1c compared to fasting plasma glucose (FPG) and/or oral glucose tolerance (OGT) testing. Accuracy inputs were based on a systematic literature review and meta-analysis of studies which compared HbA1c to FPG and/or OGT testing (known imperfect diagnostic reference standards). The modelled benefit from testing included i) the diagnosis of diabetes prior to symptom development; and ii) the identification of pre-diabetes, to introduce more frequent monitoring.

   Two HbA1c testing scenarios were considered. In the base case scenario, HbA1c testing would diagnose diabetes only; while in the alternative scenario, HbA1c testing would diagnose both diabetes and pre-diabetes.  As the current tests for diabetes and pre-diabetes (FPG and/or OGT) have imperfect accuracy, it was expected that the relative accuracy of HbA1c testing would be uncertain. As a consequence, provision was made for the results for each scenario to be considered both with the inclusion of the accuracy inputs and without. The poor quality evidence identified in the review considered HbA1c, FPG and OGT testing to be equally predictive of retinopathy (the reference standard).

Result: When accuracy data was included, the base case use of HbA1c testing was less expensive and less effective than the comparator, as HbA1c testing cannot identify pre-diabetes in this scenario (which is possible with the comparator). The alternative scenario was dominated by the comparator, due to the poor accuracy of HbA1c, relative to FPG and/or OGT testing. When test accuracy was excluded from the analysis, the base case continued to be less costly and less effective; however, the alternative scenario became dominant.

Conclusion: The inclusion of test accuracy inputs did not alter the cost-effectiveness in the base case scenario (diagnosis of diabetes only): HbA1c testing was consistently less expensive and less effective than the comparator.  However the inclusion of these inputs introduces considerable uncertainty in the cost-effectiveness of the alternative scenario (diagnosis of pre-diabetes and diabetes); the true measure of cost-effectiveness is likely to lie between modelled estimates that include or do not include these data.