= Emerging. More than 5 years before clinical availability. (16.60%, 2024)
= Expected to be clinically available in 1 to 4 years. (37.02%, 2024)
= Clinically available now. (46.38%, 2024)
MSACL 2024 : Najar

MSACL 2024 Abstract

Self-Classified Topic Area(s): Small Molecule > Data Analytics > none

Podium Presentation in Steinbeck 1 on Thursday at 10:50 (Chair: Angela Fung / Nicholas Spies)

Of Peaks and Valleys: Deconvolution in the Service of Clinical Reference Intervals

Ahmed Najar, Lilly Lim, Kara Lynch, Alan Wu
University of California San Francisco

Ahmed Najar, PhD (Presenter)

Presenter Bio: Trained as an engineer, I gravitated towards natural products chemistry in grad school to find myself in clinical chemistry thanks to MSACL. When I'm not troubleshooting a MS, I'm enjoying nature through hiking and swimming


INTRODUCTION: Establishing meaningful reference intervals (RI) for relevant clinical tests can be a daunting task for both adult and pediatric populations. In early 2023, our adult hospital clinical laboratory validated and implemented new chemistry and immunoassay instrumentation which required the verification of established reference intervals or the implementation of new reference intervals depending on the test. Clinical Laboratory Standards Institute (EP28-A3c) outlines best practice guidelines for “direct” reference interval determination which includes obtaining measurements in >120 healthy individuals per each RI partition. This task was impossible for our individual laboratory, therefore a combination of verification of historical reference intervals in 40 healthy subjects and adoption of the manufacturer recommendations was used with the intent to revisit the new RIs, particularly for some endocrine tests, 6 months after implementation of the new testing platform. During the same time frame, our neonatologists at our pediatric hospital asked our pediatric clinical laboratory to consider changes to the glucose interval for neonates. This led to an investigation of the appropriateness of all glucose pediatric ranges.
Recent developments in deconvolution algorithms render the establishment of “indirect” reference intervals more feasible by using readily available data in clinical laboratory information systems (LIS). Robust statistical methods are required to parse the healthy reference population from the pathological results.

OBJECTIVES: This study is describes the use of two different indirect reference intervals algorithms for three use cases: 1) With the implementation of a new Thyroid Stimulating Hormone (TSH) assay, the manufacturer’s RI was adopted. Upon consultation with the endocrine service, they noted an increase in consults from primary care regarding subclinical hyperthyroidism. 2) The neonatologist suggested that alternative glucose reference intervals should be implemented. Data was gathered and new RIs were established based on indirect RI algorithms in conjunction with expert opinion and practice guidelines from the neonatologists and pediatric endocrinologists. 3) With the implementation of a new testosterone immunoassay, the manufacturer’s RI was adopted, however, there was concern from the laboratory and endocrinologists that the lower limit of normal was too low for our patient population.

METHODS: 1) A deidentified dataset for 6 months of TSH results was obtained from the LIS (over 6000 unique patient entries). 2) a large deidentified multiyear dataset of glucose was collected from the LIS (over 180000 unique patient entries). 3) A deidentified dataset for 6 months of testosterone results was obtained from the LIS (over 1100 unique patient entries. All results were evaluated using two algorithms for calculating indirect reference intervals. The statistical analysis was conducted in R using the RefineR package which uses a Box-Cox transformation of the data along with the Swiss BioRef algorithm which uses the Ichihara iterative truncation-correction algorithm.

RESULTS: In all 3 cases the indirect reference intervals deviated from the historical ranges and more closely aligned with expert clinical opinion. For TSH, the lower limit of normal, as adopted by the manufacturer's recommendations, proved to be too low leading to an increase in suspected subclinical hyperthyroidism. For glucose, there was a difference between pediatric and adult ranges as suspected, but use of the indirect RI algorithms provided statistical data to support the recommendations from the neonatologists and pediatric endocrinologists. Also, the lower limit of the RI for testosterone was determined to be too low for our patient population. Both indirect RI algorithms used provided clinically similar results. In all cases, consort with endocrinologists and a validation with published direct reference intervals were necessary to link the results to clinical decision points.

CONCLUSION: Indirect reference intervals showcase that the use of deconvolution algorithms goes beyond mass spectrometry in the clinical laboratory. This work also highlights the necessity of expert knowledge along with the data generated from the indirect reference interval models.

Financial Disclosure

Board Memberno
IP Royaltyno

Planning to mention or discuss specific products or technology of the company(ies) listed above: