Study of an outbreak of off-flavours in bulk -tank milk

Off-flavours in milk can be classified using the following general categories: oxidized flavour, feed flavour, rancidity, and other flavours (e.g., unclean, malty, salty, flat or chemical). Off-flavours in milk occur at a low incidence in all milk-producing areas of Canada, with "outbreaks" occurring periodically in certain localities for no clearly defined reason. However, in the late 1990s, the incidence of off-flavours in bulk-tank milk was relatively high in Prince Edward Island (PEI). From the dairy company records it appeared that about 50/330, or 15% of herds, were affected during the winter season of 1999–2000. This presented a major economic problem for producers and dairy industry personnel, and attracted a degree of unwelcome media attention in PEI.

To evaluate the reliability of the method used for identifying off-flavours in milk, a sensory study was carried out with a panel of four milk-tank operators, who constituted the milk flavour quality control personnel. Results showed that the panelists had satisfactory agreement in differentiating off-flavoured milk from milk of good quality. The inter-panelist agreement ranged from substantial (Kappa statistic > 0.61) to almost perfect (Kappa statistic > 0.81), whereas the intra-panelist agreement range was moderate (Kappa statistic > 0.41) to almost perfect, suggesting that in the absence of a more objective diagnostic tool, a panel of trained milk graders was appropriate for the monitoring of the flavour quality of bulk-tank milk.

Results from clustering analyses revealed that this outbreak had not only a seasonal pattern, but also a limited geographical distribution with cases concentrated mostly in intensive dairy farming regions (Queens and Prince counties), and most importantly, a spatial-temporal pattern that usually peaked during fall–early winter months (September to January). Three high-rate space-time clusters (two composed of herds that experienced feed off-flavour and one composed of herds with rancid off-flavour) and two low-rate (areas with low rate of off-flavour occurrence) were identified. It appeared that high-rate clusters tended to receive more precipitation than low-rate clusters during the clustered time frame; temperature data were not as conclusive. (Abstract shortened by UMI.).