Sensory analysis of characterising flavours: evaluating tobacco product odours using an expert panel

Abstract

Objectives: Tobacco flavours are an important regulatory concept in several jurisdictions, for example in the USA, Canada and Europe. The European Tobacco Products Directive 2014/40/EU prohibits cigarettes and roll-your-own tobacco having a characterising flavour. This directive defines characterising flavour as 'a clearly noticeable smell or taste other than one of tobacco […]'. To distinguish between products with and without a characterising flavour, we trained an expert panel to identify characterising flavours by smelling.

Methods: An expert panel (n=18) evaluated the smell of 20 tobacco products using self-defined odour attributes, following Quantitative Descriptive Analysis. The panel was trained during 14 attribute training, consensus training and performance monitoring sessions. Products were assessed during six test sessions. Principal component analysis, hierarchical clustering (four and six clusters) and Hotelling's T-tests (95% and 99% CIs) were used to determine differences and similarities between tobacco products based on odour attributes.

Results: The final attribute list contained 13 odour descriptors. Panel performance was sufficient after 14 training sessions. Products marketed as unflavoured that formed a cluster were considered reference products. A four-cluster method distinguished cherry-flavoured, vanilla-flavoured and menthol-flavoured products from reference products. Six clusters subdivided reference products into tobacco leaves, roll-your-own and commercial products.

Conclusions: An expert panel was successfully trained to assess characterising odours in cigarettes and roll-your-own tobacco. This method could be applied to other product types such as e-cigarettes. Regulatory decisions on the choice of reference products and significance level are needed which directly influences the products being assessed as having a characterising odour.

Keywords: advertising and promotion; prevention; public policy.

Figure 1

Principal component analysis biplot of the PC1 and PC2 product space and, marked in red, the attribute factor map. Reference products Cig-RP1, Cig-RP2, Cig-RP3, Cig-RP4, Cig-RP5, Cig-RP6, RYO-RP1, RYO-RP2, CM6, Virginia, Burley and Oriental were expected to have a tobacco flavour. Test products Cig-TP1 and Cig-TP2 were expected to have a menthol flavour, Cig-TP4 and RYO-TP1 were marketed as vanilla, Cig-TP5 as cherry, and Cig-TP3, Cig-TP6 and Cig-TP7 were expected to potentially have a flavour other than tobacco. RP, reference product; RYO, roll-your-own; TP, test product.

Figure 2

Results of hierarchical cluster analysis of the 20 tobacco products tested with four (A) and six (B) clusters. Each colour represents a cluster that is significantly discriminated from other clusters (p<0.05) by one or more odour attributes (odour attributes not shown in this figure). Open squares represent the clusters’ average. Cig-RP1, Cig-RP2, Cig-RP3, Cig-RP4, Cig-RP5, Cig-RP6, RYO-RP1, RYO-RP2, CM6, Virginia, Burley and Oriental are products expected to have a tobacco flavour. Cig-TP1 and Cig-TP2 were expected to have a menthol flavour, Cig-TP4 and RYO-TP1 were marketed as vanilla, Cig-TP5 as cherry and Cig-TP3, Cig-TP6 and Cig-TP7 were expected to potentially have a flavour other than tobacco. RP, reference product; RYO, roll-your-own; TP, test product.

Figure 3

Confidence ellipses with a 95% CI (figure 3A, α=0.05) and a 99% CI (figure 3B, α=0.01) of the 20 tobacco products tested. Reference products Cig-RP1, Cig-RP2, Cig-RP3, Cig-RP4, Cig-RP5, Cig-RP6, RYO-RP1, RYO-RP2, CM6, Virginia, Burley and Oriental were expected to have a tobacco flavour. Test products Cig-TP1 and Cig-TP2 were expected to have a menthol flavour, Cig-TP4 and RYO-TP1 were marketed as vanilla, Cig-TP5 as cherry, and Cig-TP3, Cig-TP6 and Cig-TP7 were expected to potentially have a flavour other than tobacco. RP, reference product; RYO, roll-your-own; TP, test product.