Evaluating the performance of sensory quality control: The case of boar taint

- The agreement between sensory evaluation and chemical analysis is moderate.
- High inter-rater variability was found for sensitivity and specificity of sensory evaluation.
- Skatole above acceptance thresholds was better detected compared to androstenone.
- Performance of single assessors is lower than the panel average.

Detection of malodours referred to as 'boar taint' in entire male pigs is essential for quality control when refraining piglet castration. This study analysed the sensitivity and specificity of sensory evaluation by trained assessors (n = 18) compared to chemical analysis of two marker compounds (androstenone, skatole) in backfat (n = 794). Taking the measurement uncertainty into consideration, several cut-off thresholds for chemical analysis were exemplarily evaluated. Using the panel average score, sensitivity and specificity of sensory analysis ranged from 61 to 69% and 77 to 85%, respectively. Performance of individual assessors varied highly (sensitivity: 47 to 86%; specificity: 45 to 88%) and correlated to olfactory acuity to the compounds. According to receiver operating characteristic-curves, the average panel performed better than single assessors regardless of the sensory criterion shift. Agreement plots illustrate that high skatole levels are better detected than high androstenone levels (useful for assessor feedback). The agreement between sensory and chemical analyses was moderate. Assessors need to be selected carefully.