Here is the information we publish along with our physiological reference intervals:
The ISIS Physiological Data Project
At ISIS we take reasonable precautions to ensure that the reference intervals provided on this disk are of the highest possible quality. We remove extreme outlying values and test the validity of the submitted information in several ways. However, with more than 6.41 million test results from 342,000 samples, it is not possible to manually review all the test values submitted by member institutions. The reference intervals are provided to you on an AS IS basis. Some unlikely test values may have inadvertently been included in the calculations, leading to reference intervals for some individual tests and species that experienced clinicians will recognize as implausible. Even so, we are confident that the vast majority of reference values provided on this disk will provide useful diagnostic assistance for interpretation of test results obtained at your institution. Clinicians should always be aware that these are reference intervals, not diagnostic criteria. Any test result falling outside the appropriate reference range should result in a clinical evaluation of that result and a determination of the clinical significance of that result, but should not be taken as an absolute indicator of disease or even of an abnormal health status.
The ISIS Physiological Values database:
Test results for blood samples, along with associated body temperature and weight information, have been submitted to the ISIS database by member institutions for almost 40 years. Originally, paper forms were used to submit results to the database, but starting in 1992, member institutions using the MedARKS software gained the ability to submit records electronically. Database growth over the past 20 years has averaged over 15,000 blood sample records annually and the number of participating institutions continues to increase. The 342,000 sample records in the database represent 2788 species and provides a unique collection of physiological information from captive wildlife. This publication provides reference intervals for the 913 species with at least 50 samples in the database.
How to view the physiological reference intervals on this CD-ROM:
This CD-ROM actually contains two publications of reference intervals that differ only in the units of measurement. One set is in Standard International units used by most of the world, while the second is in the units commonly used by laboratories in the USA (conventional American units). There is an index page for each version of the reference intervals. Select the appropriate index page and double-click on the file name to open the index page in your browser. The scientific name of the species usually provides the easiest means to get to the appropriate page of reference intervals, but all pages are also cross-referenced by common name. Click on any letter in the quick navigation section at the top of the index page to jump you to the start of that alphabetic section on the index page. Click on a species name to get to the appropriate reference interval page.
Organization of the reference pages:
A reference intervals page starts with a header section that identifies the species, any age or gender partitioning of the samples used for the calculations, and the type of units, followed by a table that contains the calculated reference intervals. Each row in the table contains the test name, the units of measurement, the mean and median of test result values, the calculated central 95% reference interval (when the sample size is sufficient), the minimum and maximum data values used to calculate the reference interval, the number of samples used in the calculations and, finally, the number of individual animals that were sampled.
Hematology tests are at the top of the table, followed by blood chemistry tests and body temperature at the bottom of the table. The number of rows in this table varies greatly between species and this is entirely a reflection of the information submitted to the ISIS database by member institutions. In general, there is an inverse relationship between average body size for a species and the number of tests on a reference page for that species (smaller animals yield smaller samples, which allows for fewer tests).How the reference intervals are calculated:
There are several potential problems associated with calculating physiological reference intervals from a database assembled from multiple sources. Member institutions can use different methods to collect and handle the blood samples and different laboratories to process the samples. The health status assessment recorded at the time of sample collection is based on medical history and the experience and expertise of the clinician, but never-the-less, can be incorrect. Data entry errors can also degrade the quality of the available information. However, few institutions will ever have the resources or the number of values needed to calculate "de novo" reference intervals for even a fraction of the species held at their institution. So, despite these potential limitations, the reference intervals calculated from the ISIS database remain a unique and valuable resource.
Steps have been taken to minimize the impact of potential problems and maximize the quality of the reference intervals. Prior to starting the calculations, the global database was processed with the following filters applied at the sample level:
All samples where the indicated health status was abnormal (at the time of collection) were excluded from further analysis
Samples where the animal was only identified to the genus level were excluded from further analysis
Serum samples where the sample was frozen and stored before analysis were excluded from further analysis
Samples marked as deteriorated were excluded from further analysis
Species with less than 50 samples in the database were excluded from further analysis
The remaining samples were then processed and individual result values were examined. Clinical experience was used to identify individual result values that were judged to be incompatible with normal health (e.g., a serum calcium result of zero) and these individual results were excluded from future calculations as probable data entry errors. This process was designed to remove only the most egregious data errors.
Finally, all retained samples were examined for the presence of three or more excluded test results within a single sample. All samples meeting this criterion were considered to be significantly compromised and were excluded from further analysis.
After all the preliminary filtering and processing, the remaining numeric entry results in the database were used to calculate the reference interval pages for each species; non-numeric test result entries such as "NR", "ND" and "QNS" were ignored.
The list of species with at least 50 samples in the database was compiled and used as the basis for the reference interval pages in this publication.
A set of result values was extracted from the database, filtering by the selected species and test, and this set of data points was tested for outlier values using Tukey's inter-quartile range method. Fences were set at 3 times the IQ range to favor retention of values. All identified outlier values were excluded from the calculation for that reference interval
Note: Tukey's outlier detection method is relatively insensitive to the shape of the distribution curve and is normally an excellent method to detect multiple outlier values. However, some very highly skewed distributions will result in an excessive number of values being classified as outlier values when this detection method is used. Specifically, when given a results set with a large number of zero values and a relatively small number of positive values, such as is common for a set of basophil count results, Tukey's method will often classify all or almost all the positive values as outliers. Excluding all these results from the reference interval calculation is not appropriate. These highly skewed distributions are characterized by a Median Absolute Deviation (MAD) on the lower quantile of zero (or very close to zero), with a positive MAD for the upper quantile. To improve retention of results for highly skewed distributions, when the software detected a lower quantile MAD of zero combined with a positive upper quantile MAD, then Tukey's method was applied only to the results in the upper quantile and only the results greater than the upper fence value were classified as outliers; all other result values were retained for the calculations. This modification to Tukey's method resulted in increased retention of result values with only the very highest values (true outliers) being excluded for these highly skewed distributions. When highly skewed distributions were detected, the next step was skipped and the remaining results in the set were not retested for outliers prior to calculating a reference interval.
If outliers were present, the remaining results in the set were tested for outliers again (same method) and any additional outliers values were also excluded from the calculation for that reference interval (as recommended in ASVCP Guidelines for the Determination of Reference Intervals in Veterinary Species)
Reference intervals were calculated using a non-parametric, 500-count, resampling bootstrap method when the final results set (after discarding outlier values) was over 119 data points. For data sets with between 40 and 119 remaining result values, the reference interval was calculated using the Robust method with tuning constants of 3.7 and 205.6 (Clinical and Laboratory Standards Institute recommendation). When the final results set for a test fell below 40 results, a reference interval was not calculated, but the mean, median, and lowest and highest remaining test result values are reported as long as more than 30 data points remained. If there were less than 30 remaining data points for a test, nothing is reported.
As always, clinicians need to exercise their own judgement and use their clinical experience when using these reference intervals to interpret test results from captive wildlife.
International Species Information System
7900 International Drive
Bloomington, MN 55425