Consequences of diagnostic errors in mastitis therapy trials

Summary

The effect of errors that occur in the diagnosis of intramammary infectious mastitis on the precision of experiments measuring the efficacy of mastitis therapy has been investigated. Diagnostic errors within the range found by experienced workers can create large biases in the apparent cure rate of therapy particularly at cure rates of less than 0·5. Using confirmed methods of diagnosis rather than single samples and reducing the probabilities of false positive and false negative diagnoses to 0·01 and 0·05 respectively, the biases in the apparent cure rates are reduced to acceptable levels. A method is given for calculating the rates of occurrence of false positive and false negative diagnoses from the results of trials using confirmed diagnoses. These errors cannot be calculated from therapy trial data when diagnosis is based on single milk samples.

Because the bias in the measurements of the cure rate is greatest at the lowest levels of elimination, estimates of spontaneous recovery in untreated quarters have the greatest error. For this reason experiments incorporating an untreated control group of infected quarters usually reduce the precision of the therapy trials. An experiment in which the efficacy of a test product is measured relative to a reference product has advantages. It minimizes the difficulties arising from scale of measurement, diagnostic errors and herd differences in response rate, and makes possible comparisons between trials. Further investigations are required on the importance of spontaneous recovery, particularly for studies of Escherichia coli therapy and dry period therapy. The results of this investigation have relevance to all types of mastitis investigation that measure the change in mastitis status of udder quarters, i.e. new infection rates.