Sessions

Background: contemporary veterinary laboratory results heavily rely on population based reference intervals (RIs) estimated from results obtained from healthy individuals. The use of routine laboratory data for reference intervale stimation was made possible with novel statistical methods (indirect RI). This new approach allows the use of larger, laboratory-representative datasets and avoids cost- and labour requirements associated with the estabilishment of direct RIs.

Aim: direct hematological RIs estimated and verified from stored results carefully selected based on predefined criteria were compared with indirect RIs derived form various patient groups under real world circumstances.

Materials and methods: red blood cell count, hemoglobin concentration, hematocrit, mean cell volume (MCV), red cell distribution width (RDW), white blood cell count and various white blood cell fractions as well as reticulocyte count were investigated. The direct RI was compared to RIs derived with the indirect approach (refineR) from population groups sampled in the same timeframe but in different populations: a group with expanded healthy individuals, another derived from the same clinic’s full patient base, a third which added other clinics’ data and lastly one to which hospital population was added too. Agreement was assessed based on the confidence intervals and permissible differences of the reference limits.

Results: Agreement between the indirect and direct RIs varied considerably among analytes and between reference limits. For the indirect RIs estimated from the complete patient population of the same clinic, both the lower and upper reference limits for hemoglobin, MCV, and RDW remained within the permissible difference. For other analytes, only one reference limit showed acceptable agreement, while some leukocyte parameters and the reticulocyte count showed larger differences, often with a monotonic increasing or decreasing trend.

Conclusions: The robustness of indirect hematological reference limits estimated using the refineR algorithm depended on the analyte and the composition of the source population. For some analytes, the indirect method yielded reference limits comparable or close to those of the direct RI. Our results indicate that when applying indirect hematological RIs, not only dataset size but also its clinical composition and disease-associated changes in the distribution of the respective analyte should be considered.