That night, Aliyah wrote a new lab policy. They would adopt the manufacturer’s broader interval for patients over 65—not out of laziness, but out of a deeper respect for EP28’s core principle: A reference interval is only as good as its reference population.
Three weeks later, Mrs. Park was in the ER with atrial fibrillation—a known risk of overtreatment in the elderly.
The lower limit of her in-house reference interval was 0.6 mIU/L. The upper limit was 3.2. clsi ep28
She called Mrs. Park’s family. The levothyroxine was stopped. The arrhythmia resolved.
And Aliyah learned that “normal” is not a number printed in a manual or even a percentiles from a tidy dataset. It is a fragile, shifting border between biology and statistics—and the job of a clinical chemist is not just to measure, but to interpret who, exactly, is in the room when you draw the line. That night, Aliyah wrote a new lab policy
Then came the case that changed everything.
Mrs. Eleanor Park, 68, came in for fatigue. Her TSH was 3.9 mIU/L—within the manufacturer’s range but above Aliyah’s verified upper limit of 3.2. Using the lab’s new narrow interval, the computer flagged it as Abnormal-High . The junior resident started her on low-dose levothyroxine. Park was in the ER with atrial fibrillation—a
Aliyah nodded. “But EP28 says if we have 120 subjects, nonparametric ranking is the gold standard. The 2.5th and 97.5th percentiles are 0.6 and 3.2. That’s our truth.”
Mrs. Park wasn’t abnormal. Aliyah’s reference population was just too young.
“That’s too narrow,” her senior technologist, Marcus, said, frowning at the scatter plot. “Manufacturer says 0.4 to 4.0. If we use ours, we’ll flag half our outpatients as abnormal.”