ABSTRACT
I. Introduction.......................................................................................... 453
II. Reference Intervals and Percentile Estimators ................................... 454
III. Traditional Normal-Theory Approach ................................................ 455
IV. Data Transformation to Achieve Normality ....................................... 457
V. Nonparametric Approach Using Order Statistics................................ 458
VI. Precision of Reference Interval Endpoints ......................................... 462
VII. Outliers................................................................................................. 464
VIII. Summary of Methods to Derive Reference Intervals ......................... 466
References........................................................................................................ 467
The reference interval, also known as the reference range or normal range,
includes the central 95% of an analyte (where an analyte is “any substance or
chemical constituent of blood, urine, or other body fluid that is analyzed”)
measured in a presumed healthy population. The reference interval is used when
there is no clear definition of disease. By this definition, unhealthy, or abnormal
patient analyte values, would lie in the lower and upper 2.5%. In cases where the
disease is defined, for example,
glucose concentration levels above 110 mg/dL
are associated with diabetes,
the reference interval derived for the central 95% of
the population is not used. It should be noted that the reference interval is by its
nature a guideline and when combined with other clinical data helps define the
clinical status of a patient. However, often in clinical practice the reference
interval is misused as an absolute definition of health status. For this reason,
the traditional term “normal range” has been mostly replaced by “reference
range,” or more correctly, “reference interval,” in current practice.
