Fonseca Mariana, Kurban Daryna, Roy Jean-Philippe, Santschi Débora E, Molgat Elouise, Yang Danchen Aaron, Dufour Simon
Department of Pathology and Microbiology, Faculty of Veterinary Medicine, Université de Montréal, Saint-Hyacinthe, QC J2S2M2, Canada; Regroupement FRQNT Op+lait, Saint-Hyacinthe, QC J2S2M2, Canada.
Regroupement FRQNT Op+lait, Saint-Hyacinthe, QC J2S2M2, Canada; Department of Clinical Sciences, Faculty of Veterinary Medicine, Université de Montréal, Saint-Hyacinthe, QC J2S2M2, Canada.
J Dairy Sci. 2025 Apr;108(4):3917-3928. doi: 10.3168/jds.2024-25403. Epub 2024 Dec 16.
Mastitis, an inflammation of the udder primarily caused by an IMI, is one of the most common diseases in dairy cattle. Somatic cell count has been widely used as an indicator of udder inflammation, assisting in the detection of subclinical mastitis. More recently, differential SCC (DSCC), which represents the combined proportion of lymphocytes and polymorphonuclear leukocytes, has become available for routine dairy milk screening, although it was not yet widely studied. Therefore, the objective of this study was to assess and compare the usefulness of quarter-level SCS or DSCC to predict the probability of subclinical mastitis. Additionally, our goals included estimating the sensitivity (Se) and specificity (Sp) of SCS and DSCC across all potential cut-off values. The current study was an observational study conducted on commercial dairy farms. Five dairy herds were selected using a convenience sampling. A Gaussian finite mixture model (GFMM) was applied to investigate the latent quarter subclinical mastitis status with either measurement, SCS or DSCC. Posterior values for SCS and DSCC obtained from the GFMM were used for predictive estimation of the parameters. The estimated SCS distribution for healthy quarters had a mean (SD) of 1.4 (1.3), and, for quarters with subclinical mastitis, it was 4.5 (2.4). For DSCC, the estimated mean was 55.6% (15.2) for healthy quarters, whereas it was 80.4% (6.4) for quarters with subclinical mastitis. The most discriminant cut-off for SCS, as indicated by the Youden index, was 3.0, corresponding to exactly 100,000 cells/mL. At this threshold, the Se and Sp of SCS were 0.73 (95% Bayesian credible interval [BCI]: 0.70-0.77) and 0.90 (95% BCI: 0.89-0.91), respectively. The most discriminant cut-off point for DSCC was 70.0%, with corresponding Se and Sp values of 0.95 (0.93, 0.96) and 0.83 (0.81, 0.85), respectively. For the SCS analysis, we obtained predictive probabilities of subclinical mastitis approaching 0 and 100%, with only a narrow range of SCS results yielding intermediate probabilities. On the other hand, predictive probabilities ranging from 0 to 90% were obtained for DSCC analysis, with a large range of DSCC results presenting intermediate probabilities. Thus, SCS seemed to surpass DSCC for predicting subclinical mastitis. These findings provided a foundation for future studies to further explore and validate the efficacy of GFMM for diagnostic tests yielding quantitative results.
乳腺炎是主要由乳房内感染引起的乳腺炎症，是奶牛最常见的疾病之一。体细胞计数已被广泛用作乳腺炎症的指标，有助于检测亚临床乳腺炎。最近，代表淋巴细胞和多形核白细胞综合比例的差异体细胞计数（DSCC）已可用于常规牛奶筛查，尽管尚未得到广泛研究。因此，本研究的目的是评估和比较乳区水平的体细胞评分（SCS）或DSCC预测亚临床乳腺炎概率的有用性。此外，我们的目标包括估计SCS和DSCC在所有潜在临界值下的敏感性（Se）和特异性（Sp）。本研究是在商业奶牛场进行的一项观察性研究。采用便利抽样法选取了5个奶牛群。应用高斯有限混合模型（GFMM）来研究潜在的乳区亚临床乳腺炎状态，同时测量SCS或DSCC。从GFMM获得的SCS和DSCC的后验值用于参数的预测估计。健康乳区的估计SCS分布均值（标准差）为1.4（1.3），亚临床乳腺炎乳区的估计SCS分布均值为4.5（2.4）。对于DSCC，健康乳区的估计均值为55.6%（15.2），亚临床乳腺炎乳区的估计均值为80.4%（6.4）。尤登指数表明，SCS的最具判别力的临界值为3.0，对应于每毫升恰好100,000个细胞。在此阈值下，SCS的Se和Sp分别为0.73（95%贝叶斯可信区间[BCI]：0.70 - 0.77）和0.90（95% BCI：0.89 - 0.91）。DSCC的最具判别力的临界点为70.0%，相应的Se和Sp值分别为0.95（0.93, 0.96）和0.83（0.81, 0.85）。对于SCS分析，我们获得的亚临床乳腺炎预测概率接近0和100%，只有很窄范围的SCS结果产生中间概率。另一方面，DSCC分析获得的预测概率范围为0至90%，有很大范围的DSCC结果呈现中间概率。因此，在预测亚临床乳腺炎方面，SCS似乎优于DSCC。这些发现为未来进一步探索和验证GFMM在产生定量结果的诊断测试中的有效性的研究奠定了基础。
