Reference interval evolution: what moving to the 99th percentile looks like for a comprehensive metabolic panel in an outpatient setting.

BACKGROUND

In clinical laboratory practice, reference intervals (RIs) for chemistry analytes are commonly adopted directly from manufacturer package inserts. These intervals are typically verified locally according to regulatory requirements rather than independently derived, under the assumption that manufacturer-reported RIs are broadly applicable across patient populations. In low pre-test probability settings such as routine outpatient testing or wellness testing, this approach can generate a high frequency of flagged abnormalities that more likely reflects biological variation rather than disease. Expanding RIs to the central 99% has been proposed as a strategy to improve specificity.

METHODS

The 99% RIs were derived for CMP analytes using retrospective data from 23,422 unique adult outpatients tested at a single institution in September 2025. Results were compared with 95% RIs and manufacturer-provided RIs across analytes. Flagging performance was evaluated using an independent retrospective dataset of 28,006 adult outpatients tested between December 2025 and January 2026. Analyses focused on the proportion of results flagged outside each RI.

RESULTS

Electrolytes, including sodium, chloride, calcium, and CO₂, showed minimal differences in flagging between manufacturer RIs and 99%-derived intervals, with differences consistently under 7%. In contrast, kidney and liver analytes demonstrated substantial variability. Manufacturer RIs flagged a markedly higher proportion of results outside the interval for creatinine, AST, ALT, ALP, BUN, potassium, and total protein compared with the 99% approach.

CONCLUSIONS

Electrolyte RIs are consistent across manufacturer-provided RIs and population-derived 99% intervals. In contrast, kidney and liver analytes show pronounced differences between historical package insert RIs and contemporary percentile-based distributions, resulting in substantial variation in abnormal result flagging. These findings highlight the impact of using alternate approaches to derive our kidney and liver RIs.