Selection of Reference Animals

Specific reference intervals are needed for each species of animal being tested. Less often, a different reference interval is needed for an analyte from a specific breed of animal (e.g., hematocrit values in greyhound dogs are higher than those in most other dog breeds). Values may vary with the age of the animal, with major changes occurring prior to puberty (e.g., 3-week-old pups have lower hematocrits than adults). Consequently some analytes need different reference intervals for different age groups. Some analytes also vary with sex, pregnancy, emotional state, and activity level. The types of animals sampled and environmental conditions present during the establishment of a reference interval should be defined, along with the methods and equipment used, so that the user can make appropriate evaluations. Ideally, a reference interval should be established using a population of healthy animals with a composition (age, breed, sex, diet, etc.) like the population of ill animals being evaluated. Homogeneous populations generally have more narrow reference intervals than heterogenous populations. Establishing a reference interval for a blood analyte using a group of male foxhound dogs housed in a research colony, fed the same diet, and conditioned to phlebotomies would likely result in reference intervals too narrow for the population of dogs examined in a typical small-animal practice. Reference intervals are generally established for a species by utilizing samples from apparently healthy adult animals of both sexes and various breeds. Monogastric animals should have been fasted overnight prior to blood sample collection.

Determination of Reference Intervals

Specific reference intervals should be established for each instrument and each test evaluated. Ideally, each animal would have its own reference intervals established by multiple assays done over time when the animal was healthy. In some instances, limited numbers of baseline values are available for an animal that can be helpful, but rarely are analytes measured often enough to establish an accurate reference interval for an individual animal. Consequently population-based reference intervals are used.

When the frequency diagram of test results from a healthy population is examined, many analytes exhibit a Gaussian or bell-shaped distribution (Fig. 1-1). When a Gaussian distribution is present, a minimum of 40 individuals (100 or more is preferred) should be assayed for statistical validity.² In this case, the reference interval is calculated using the mean ±2 standard deviations (SD). This interval approximates the 95% confidence interval. In other words, about 95% of healthy animals have test values within this reference interval, with about 2.5% of healthy animals having values above and about 2.5% of healthy animals values below the reference interval. A common mistake made by novices is to calculate the reference interval from the mean ±1 SD. When this is done, about 32% of healthy animals will have values outside the calculated interval. If less than 40 healthy animals are available, the upper and lower values measured should be used to create an estimated reference interval.⁵

FIGURE 1-1 Frequency diagram of a hypothetical plasma analyte with Gaussian distribution. The central (tallest) vertical line denotes the mean. Each additional vertical line represents one standard deviation (SD) from the adjacent vertical line. The reference interval calculated using mean ±2 SD (21 ± 4 mmol/L) is 17 to 25 mmol/L.

Some analytes do not exhibit a Gaussian distribution. Most commonly there is a skew toward the higher values. The use of mean ±2 SD to calculate reference intervals results in inappropriate reference intervals for skewed populations, as shown in Figure 1-2. Data may be manipulated (e.g., log or square root transformation) so that the frequency distribution of the transformed data approximates a Gaussian distribution. The boundaries are determined as before and results are retransformed to determine the reference interval. Alternatively, one can use percentiles to determine upper and lower limits, especially if large numbers of healthy animals are evaluated. Values are listed in ascending order. The lower limit is determined by the formula (n + 1) × 0.025, and the upper limit is determined by the formula (n + 1) × 0.975, where n = the number of normal animals assayed.² If 119 animals were used, the value for the 3rd lowest animal would be used as the lower limit and the value from the 117th animal (3rd from the top) would be used as the upper limit.

FIGURE 1-2 Frequency diagram of hypothetical absolute blood cell counts with a skewed population. The central (tallest) vertical line denotes the mean. Each additional vertical line represents one standard deviation (SD) from the adjacent vertical line. The use of mean ±2 SD to calculate the reference interval is inappropriate, as demonstrated by the lower limit being an impossible negative value.

Interpretation of Test Results Relative to Reference Intervals

The common usage of the 95% confidence interval to establish reference intervals means that 5% of healthy animals will be reported as abnormal for a given test. When multiple tests are done in laboratory medicine profiles, the probability of at least one test being abnormal increases with the number of tests done. For example, there is a 64% chance that at least one abnormal test result will be obtained when 20 analytes are measured from a healthy animal.⁶ The degree to which a test result is above or below the reference interval is generally important in deciding whether a high or low value should be taken seriously.

Use of Published Reference Intervals

Routine hematology test results are usually similar between laboratories; consequently published reference intervals for values such as total leukocyte counts and hematocrits are often used to interpret results from a species (e.g., wallaby) when reference values have not been established in the laboratory conducting the test. Hematology indices such as the red cell distribution width (RDW) vary more between laboratories, making the use of published reference intervals less acceptable.

The units used in reporting values can vary by laboratory and a conversion factor may be needed to compare a measured value to a published reference interval. For example, blood iron might be reported as 100 µg/dL or 18 µmol/L. Most U.S. laboratories continue to use conventional units, such as mg/dL; Canadian and European laboratories use the International System of Units (SI units), such as mmol/L. Where possible, moles are used rather than weight (e.g., mg) for SI units. This cannot be done for analytes, such as serum protein concentration, where the molecular weight is variable and/or unknown. For enzymes, an SI enzyme unit is defined as 1 µmol/min of substrate utilized or product formed. SI units are reported per liter.

For many wild animal species, reference intervals may not be published for some or all tests. The simultaneous measurement of a healthy “control” animal from the same species, preferably a cohort, can be used as a rough guideline reference value and therefore can aid interpretation of the patient’s results.