Evaluation of rapid culture, a predictive algorithm, esterase somatic cell count and lactate dehydrogenase to detect intramammary infection in quarters of dairy cows at dry-off

Abstract

Our objective was to compare four tests to standard milk culture followed by MALDI-ToF in quarters of cows at dry-off. Cows (n = 432) were randomly selected from seven U.S. dairy herds already participating in a multi-site clinical trial in summer 2018. Aseptic foremilk samples were collected from quarters (n = 1728) two days prior to dry-off, and subjected to index and reference tests. The four index tests included rapid culture, a predictive algorithm, an esterase strip test measuring somatic cell count (SCC) and a cow-side lactate dehydrogenase (LDH) test. Rapid culture was performed by inoculating quarter milk samples onto a commercial rapid culture plate. Plates were evaluated by technicians after 30−40 h of incubation at 37 ± 2 °C. Quarters were classified as infected if any bacterial growth was observed. For the algorithm test method, all quarters were classified as infected if the cow met any of the following criteria: 1) any Dairy Herd Improvement Association (DHIA) test with a SCC > 200,000 cells / ml during the current lactation or 2) two or more clinical mastitis cases during the current lactation. Esterase-SCC and cow-side LDH tests involved adding milk to the test strip and reading for color changes. For esterase-SCC and cow-side LDH tests, low (≥250 cells / ml and ≥100 U / L) and high (≥500 cells / ml and ≥200 U / L) thresholds were used to classify quarters as infected or not. Composite samples (4 × 2 mL quarter-milk samples commingled) were also tested for rapid culture, esterase-SCC and cow-side LDH tests, such that if a composite sample was positive, then all quarters contributing to that sample were classified as infected. The reference test was traditional aerobic culture conducted in an accredited laboratory using MALDI-ToF for identification of isolates. Traditional culture was only conducted on quarter-milk samples, and consequently, IMI was always considered at the quarter-level. Unconditional logistic regression was used to estimate sensitivity (SE), specificity (SP), apparent prevalence, positive predictive values (PPV) and negative predictive values (NPV) for each index test. Cohen’s Kappa (κ) was used to measure agreement between tests. Algorithm, esterase-SCC and cow-side LDH tests had poor agreement with the reference test (κ ranging from 0.01 to 0.12), while rapid culture had fair agreement (κ = 0.28). No test had concurrently high SE and SP. Negative predictive values were moderate to high for all tests.