Desirable standards for laboratory tests if they are to fulfill medical needs

Many strategies to define desirable standards for laboratory tests to fulfill medical needs have been proposed over the last three decades. Traditional approaches are based on reference (normal) values, opinions of clinicians, the state of the art, views of experts, data on biological variation, and assessment of the effect of error on clinical use. All these approaches have advantages and disadvantages, but the consensus of experts reached over a decade ago that imprecision desirably be less than one-half of the within-subject biological variation still seems to provide the best set of generally applicable performance standards. Desirable bias is less than one-quarter of the group (within-subject plus between-subject) biological variation. Recent proposals are either restatements of traditional recommendations, further empirical suggestions, or models based on assessment of clinical needs, and have not been widely accepted. Both old and new studies on clinical opinions, sought by using structured questionnaires containing clinical vignettes designed to seek views on the magnitude of significant change, are flawed in design, execution, and data analysis. Until clinicians are more aware of test-result variability and clinical chemists gain quantitative knowledge on the interpretation of test results, it will be difficult to set desirable standards that fulfill actual medical needs, except in a few well-defined screening situations.