B-133 Optimizing Reference Intervals in Clinical Laboratories: Insights from Indirect Method Implementation

Abstract Background Reference intervals support physicians in interpreting test results and are a key concern in laboratory medicine. Accrediting bodies, such as College of American Pathologists, require laboratories to study their own reference intervals for their specific populations. The CLSI EP28 document provides guidelines for verifying or establishing these intervals. Recently, statistical tools have been explored for indirect determination using big data from laboratory results. This study presents the experience of a large Brazilian laboratory that has applied the indirect method for reference interval determination over the past two years. Methods A committee was established for the continuous evaluation of reference intervals used in the laboratory. Analytes for study were prioritized based on their clinical relevance. Laboratory test result data were extracted from varying periods within the recent historical database, ensuring a minimum of 1,000 results per analyte and platform. Hospital-origin samples and business-to-business service data were excluded. Data mining was performed using exclusion criteria derived from the laboratory’s historical records, including associated analytes and physician order notes. For all analytes, reference intervals were verified or established using indirect techniques through the LabRI™ algorithm, applying parametric, non-parametric, and robust statistical methods for outlier exclusion. These analyses were performed using a series of algorithms programmed in R Language. Verified reference intervals were derived from manufacturer package inserts and available literature. All results were reviewed by a multidisciplinary team of experts before adoption by the laboratory. Results During this period, reference intervals were determined for eleven analytes: Uric Acid, Creatinine, CPK, Urine Specific Gravity, Homocysteine, Insulin, Magnesium, Transferrin, Vitamin B12, Vitamin C, and Vitamin A. Of these, ten were studied in response to clinical concerns regarding the potential inadequacy of manufacturer-provided reference intervals. The number of results per partition ranged from 686 to 31,914, with an average of 8,446 results. For most analytes (8/11), verification studies using manufacturer or literature data were not approved, requiring the laboratory to establish its own reference values. Conclusion Our experience indicates that manufacturer-proposed reference intervals, as well as those derived from literature in different populations, are often unsuitable for direct adoption in laboratory reports. Laboratories should prioritize processes that enable the continuous evaluation of reference intervals. Additionally, we observed that the use of indirect approaches based on laboratory database analysis allows for a more efficient and cost-effective determination of reference intervals. We also emphasize the importance of using appropriate statistical tools for this purpose to ensure the reliability and accuracy of the results obtained.