Looking back to look forward: can analytical performance be successfully linked to health outcomes in laboratory...

The 2015 Milan consensus proposed three models for analytical performance (APS) specifications for laboratory tests. Model 1 recommends defining analytical performance specifications based on impact on clinical outcomes. We undertook a targeted literature review to identify and evaluate studies using indirect methods to evaluate the impact of test analytical performance on downstream outcomes. We identified 69 studies, 38 of which covered a single measurand: blood glucose. The outcome used was: test results in 33 studies (diagnostic classification n=28, prognostic classification n=5), change in clinical decision in 20 studies, intermediate outcome post treatment in 11 studies, and an actual health outcome in 5 health economic modelling studies of monitoring blood glucose (treatment benefit n=4, treatment harm n=5). In these five studies, the impact of analytical performance on subsequent health outcomes was estimated by extrapolating analytical performance (AP) impacts on treatment effects on intermediate outcomes, to impacts on health outcomes. Supporting empirical evidence for these extrapolations was not well described. Extrapolation of improved AP impacts on treatment effects may be more credible for decreasing short-term harms (e.g. insulin dosing errors causing hypoglycemia requiring external assistance) than for increasing long-term benefits (e.g. improved glycaemic control resulting in fewer cardiovascular events). Future Milan Model 1 APS studies using actual health outcomes may be most useful for monitoring tests of measurands closely linked to short-term harms from treatment. For other measurands, APS may be set using earlier outcomes in the causal chain between test and health outcome, such as diagnostic classification.