Certification is important—modeling and measurement are more important

An editorial in this journal about uncertainty [1] reflects similar views to mine for in vitro diagnostic assays. I offer the following comments for the case of uncertainty intervals and for quality provided by ISO 9001. Recommendations have been made [2, 3] to use uncertainty intervals for diagnostic assay results based on the ISO Guide to the Expression of Uncertainty in Measurement [4]. These intervals are common for reference materials that are assayed by definitive or reference methods but to try to use them for commercial diagnostic assays is unwarranted and will likely lead to intervals that are too narrow. Commercial diagnostic assays have been developed with tradeoffs. Accuracy that is available in reference methods has been traded off for ease of use, lower cost, or other attributes with the understanding that the lower accuracy still meets medical acceptability criteria. Unfortunately, commercial assays have a small percentage of results that contain large errors, whose root cause is often unknown. There is no practical way to construct GUM uncertainty intervals to account for these results; hence if uncertainty intervals are provided, they will likely be incorrect and they may mislead users in attaching more faith in the results than are warranted. In spite of this, the modeling that is required by GUM is useful in helping to pinpoint assay problems.