Development of Objective Training and Performance Indicators for Gas Chromatography Olfactometry Panels

D. Frank, G. Giannikopoulos, C. Forde, C. Delahunty
Food Science Australia, PO Box 52, North Ryde NSW 1670, Australia

The accuracy of gas chromatography olfactometry (GC-O) is very much dependent on the performance of olfactory assessors (sniffers). This is particularly the case when one wishes assessors to rate intensity of perceived odours, rather than simply indicate detection. Objective training methods and assessor performance benchmarks are therefore desirable to ensure high quality data. In the current study we sought to develop a robust training method and also objective GC-O performance scores. Ten assessors were selected for specific training in time intensity (TI) rating of gas chromatograph effluents using a computer mouse and a sliding 100 mm line scale. Assessors were subjected to the ‘Sniffin’ Sticks’ basic olfactory acuity test before the GC-O experiments. To demonstrate performance improvement as a result of training, initial GC-O analyses (triplicate, randomly assigned over separate days) were carried out where only minimal safety and theoretical instruction in the performance of GC-O were provided (pre-training).  A liquid extract of a natural product was used as the sample type for this pre-training analysis.

Following pre-training, assessors were given specific individual feedback on their performance, by comparing individual aromagram data to the average group aromagram - with specific feedback on intensity scaling, number of peaks detected, and number of and suitability of descriptors given. A mixture of 30 common food and beverage volatiles, included at concentrations to give a range of odour activity, was used in the training phase. Assessors were given feedback and coaching in odour recognition and description using reference chemicals. Assessors were also given real-time coaching in correct mouse and sniffing technique. Finally, the liquid extract assessed pre-training was sniffed in triplicate by each panellist over separate days and different sessions (post training). 

Simple statistics, comparable to those typically used in GC-O, were used to measure performance. These included the number of compounds identified, a comparison of the order of perceived intensity across assessors, and ability to accurately describe the odour perceived. In addition, the replicate TI aromagram from each assessor, and the panel average TI- ‘principal aromagram’ (PA), was modelled using non-centred Principal Component Analysis. The pre- and post-training two dimensional Euclidean distances of the triplicate assessor PC loadings compared to that of the PA were subjected to ANOVA. Significant improvement in performance was measured for all assessors after training. Both ‘Sniffin Sticks’ and ‘Training Mix’ scores were significantly correlated to the post-training performance, and hence provide to an extent objective criteria to predict GC-O panel performance. Additionally an ideal  reference PA based on optimal  ‘training mix’ performance can be used as an objective reference point for assessing, calibrating and monitoring changes in individual GC-O panellist and overall GC-O trained panel performance over time.