The key finding
A 2025 laboratory study revealed that the prostate health index (phi)—a blood test used to assess prostate cancer risk—has been reporting unnecessarily high measurement uncertainty. When researchers applied a more sophisticated mathematical approach that accounts for how the test’s three components correlate with each other, they cut the uncertainty estimate in half: from 7.2% down to 3.6% for a representative phi value of 24.48. This matters because lower uncertainty means doctors can have more confidence in borderline test results that might otherwise lead to unnecessary biopsies or missed cancers.
What the study looked like
This wasn’t a patient study but rather a methodological analysis of laboratory measurement techniques. The researchers compared two different approaches for calculating measurement uncertainty in the phi test. The phi combines three separate blood measurements—total PSA, free PSA, and [-2]proPSA—through a mathematical formula. The first approach, currently used by some labs, calculates uncertainty directly from internal quality control (IQC) data. The second method follows the “Guide to the Expression of Uncertainty in Measurement” (GUM), an international standard that allows scientists to account for correlations—the ways different measurements might systematically vary together. The team applied both methods to the same phi data and compared the resulting uncertainty estimates.
Why researchers think this happened
The dramatic difference stems from statistical correlation. When a lab measures total PSA and free PSA from the same blood sample, these values don’t vary completely independently—if one runs slightly high due to instrument drift or temperature, the other tends to shift similarly. The simpler IQC method treats each measurement as completely independent, which artificially inflates the combined uncertainty. The GUM approach incorporates a “correlation term” that mathematically accounts for this shared variation. When the researchers included correlation, the uncertainty dropped from 5.99% to 3.60%—a 40% reduction. The 7.2% figure from the IQC method was even higher because that approach captures additional sources of variation not directly related to the analytical measurement precision itself.
How to read this carefully
This study analyzed mathematical methods rather than clinical outcomes, so it doesn’t directly show whether the more accurate uncertainty estimates lead to better patient care. The correlation values the researchers used came from limited quality control data, and these correlations might vary between laboratories, instruments, or even different batches of testing reagents. The study also focused on a single phi value; the improvement in uncertainty might differ at very low or very high phi levels. Importantly, reducing measurement uncertainty doesn’t change the actual phi result—it just gives a more honest picture of how much we should trust that number. Laboratories would need to invest time in calculating correlation coefficients for their specific equipment and conditions.
What this means for everyday life
If you’re a man considering or undergoing prostate cancer screening, this finding suggests that phi test results might be more reliable than previously thought—but only if your lab adopts the better calculation method. This matters most when results fall in the gray zone: a phi around 25 to 35 might prompt a biopsy at some centers but watchful waiting at others. Tighter uncertainty bounds could help doctors make more confident recommendations in these borderline cases. The broader lesson extends beyond prostate testing: many medical tests combine multiple measurements, from hormone panels to metabolic scores, and the way labs calculate uncertainty can significantly affect clinical interpretation. Given this research, patients might reasonably ask their healthcare providers whether calculated test results include correlation adjustments—though implementing such changes requires laboratory-wide standardization efforts rather than individual patient action.