Reference intervals
Reference intervals vary considerably from one laboratory to another, and are dependent on the methodology and instrumentation utilized. This is especially true for serum enzymes, where methods may vary in pH, temperature, specific co-factors and substrate used in the reaction. As a consequence, published "normal" values may not be valid for results generated by your lab. Reference intervals should be established by each laboratory, as they are instrumentation and reagent dependent. You should not compare one laboratory's results to another laboratory's reference intervals.
Small deviations outside the reference interval may not be significant. In other words, clinically healthy animals may have mildly reduced or increased analyte values compared to a reference interval, but the values are actually normal for that animal. This is somewhat dependent on the analyte, e.g. mild elevations in liver enzymes are probably not significant (depending on clinical signs) in many animals, however electrolytes are maintained within fairly narrow limits and elevations in these are more likely to be clinically significant.
Reference intervals are usually determined from a population of healthy adult animals. There are two general methods for determining reference intervals, based on the distribution of the data from these healthy animals.
- Gaussian distribution: This is when the data is normally distributed, i.e. distributed symmetrically around the mean as illustrated to the right. The reference interval is calculated as the mean ± 2 SD, which encompasses 95% of the observations in healthy animals. The top 2.5% and bottom 2.5% results from healthy animals will fall outside an established interval.
- Non-Gaussian distribution: For data that is non-gaussian (i.e. skewed), the data can be mathematically transformed, e.g. to logarithms. If this yields a normal or Gaussian distribution, the geometric mean ± 2 SD can be used for reference interval determination (geometric means are based on the log-transformed data). However, in most instances, percentiles are used to create reference intervals from non-Gaussian data. The top 97.5 and bottom 2.5 percentiles are used as the limits of the reference range. Sufficient numbers of animals are required before percentiles are used (minimum of 50, although 120 is ideal). As most biochemical data is not normally distributed, this is the optimal technique for reference intervals determination.