A multilaboratory study on the variability of bovine semen analysis

Abstract

To evaluate the variability of semen analysis, five replicates of 10 different bovine frozen semen batches were coded with different identification numbers and submitted to various laboratories for evaluation. Three studies were conducted: study I included eight laboratories in semen processing centers in the United States; study II included one laboratory in one semen processing center and five veterinary university laboratories in the United States; and study III included five veterinary university laboratories in Brazil. Evaluation methodology, sample classification criteria, and reporting format varied considerably among laboratories. There were laboratory effects (P < 0.05) on sperm concentration, motility, and morphology results in all studies. When Bland–Altman plots were evaluated, differences in sperm concentration were approximately between −5 and +5 × 106 sperm/mL in study I, when the same method of evaluation was used by all laboratories but ranged between −30 and +30 × 106 sperm/mL in studies II and III. Differences in the proportions of motile sperm were approximately −30% to +30%, and differences in the proportion of normal sperm were −15% to +15% in studies I and II; these differences were −15% to +15% and −10% to +10%, respectively, in study III. Mean absolute (one tail) proportional differences in estimates across all laboratories ranged from 9% to 31%, 16% to 37%, and 9% to 14% for sperm concentration, motile sperm, and normal sperm across studies; much larger (48%–86%) differences were observed for sperm abnormality categories. Intralaboratory and interlaboratory precision varied considerably across laboratories and seemed to be at least in part related to methods used for evaluation; precision was better when the NucleoCounter was used for evaluation of sperm concentration, whereas the use of computer-assisted sperm analysis for evaluation of sperm motility resulted in greater precision in some but not all laboratories. None of the laboratories that classified samples as satisfactory or unsatisfactory achieved complete consistency for all replicates within all batches. In addition, consistent classification among laboratories was observed for just three batches in studies II and III. These observations put the reliability of semen analysis in check and make it very difficult, if not impossible, to meaningfully interpret evaluation results.