Lab Medicine Significance 7/10

Outlier Removal Methods Dramatically Alter Cardiac Troponin 99th Percentile Reference Limits

Using data from the Canberra Heart Study, this investigation demonstrates that the choice of statistical outlier removal method can change the troponin 99th percentile by more than threefold in healthy males and nearly twofold in females. The findings extend beyond troponin to any analyte using upper reference limits, revealing an urgent need for standardized statistical procedures in reference interval determination.

The original study

Statistical considerations for determining high-sensitivity cardiac troponin reference intervals.

Authors
Hickman PE, Koerbin G, Potter JM, Abhayaratna WP
Journal
Clinical biochemistry
Type
Clinical Trial, Journal Article, Multicenter Study
PMID
28263716
Read the original study →

Original abstract

The troponin 99th percentile is used as the laboratory decision point in the diagnosis of acute myocardial infarction. A recent publication has shown that the statistical treatment for outlier removal may dramatically change the calculated troponin 99th percentile. We have used our large database from the previously reported Canberra Heart Study to independently assess the effect of various methods for removing outliers on the calculated 99th percentile. We have performed the same exercise using the troponin 97.5th percentile as an exercise to assess how outlier removal may affect calculated upper reference intervals for any analyte which uses this boundary. For healthy males aged <75years the hs-cTnI troponin 99th percentile varied by a factor>3× depending upon the outlier removal method chosen and for the 97.5th percentile the variation was >50%. For women the variation in the hs-cTnI 99th percentile varied by a factor of nearly 2×. Qualitatively similar results were obtained forhs-cTnT. This is not simply a problem for troponin reference intervals. All analyte reference intervals have the potential to be significantly affected by the method chosen for outlier removal. To ensure that studies can be meaningfully compared, guidance on procedures for removing outliers needs to be standardized as a matter of urgency.