Comprehensive overview of common e-liquid ingredients and how they can be used to predict an e-liquid's flavour category

Erna J.Z. Krüsemann*, Anne Havermans, Jeroen L.A. Pennings, Kees De Graaf, Sanne Boesveldt, Reinskje Talhout

*Corresponding author for this work

Research output: Contribution to journalArticleAcademicpeer-review

43 Citations (Scopus)


Objectives Flavours increase e-cigarette attractiveness and use and thereby exposure to potentially toxic ingredients. An overview of e-liquid ingredients is needed to select target ingredients for chemical analytical and toxicological research and for regulatory approaches aimed at reducing e-cigarette attractiveness. Using information from e-cigarette manufacturers, we aim to identify the flavouring ingredients most frequently added to e-liquids on the Dutch market. Additionally, we used flavouring compositions to automatically classify e-liquids into flavour categories, thereby generating an overview that can facilitate market surveillance. Methods We used a dataset containing 16 839 e-liquids that were manually classified into 16 flavour categories in our previous study. For the overall set and each flavour category, we identified flavourings present in more than 10% of the products and their median quantities. Next, quantitative and qualitative ingredient information was used to predict e-liquid flavour categories using a random forest algorithm. Results We identified 219 unique ingredients that were added to more than 100 e-liquids, of which 213 were flavourings. The mean number of flavourings per e-liquid was 10±15. The most frequently used flavourings were vanillin (present in 35% of all liquids), ethyl maltol (32%) and ethyl butyrate (28%). In addition, we identified 29 category-specific flavourings. Moreover, e-liquids' flavour categories were predicted with an overall accuracy of 70%. Conclusions Information from manufacturers can be used to identify frequently used and category-specific flavourings. Qualitative and quantitative ingredient information can be used to successfully predict an e-liquid's flavour category, serving as an example for regulators that have similar datasets available.

Original languageEnglish
Pages (from-to)185-191
JournalTobacco Control
Issue number2
Early online date22 Feb 2021
Publication statusPublished - 2021


  • electronic nicotine delivery devices
  • public policy
  • surveillance and monitoring
  • tobacco industry


Dive into the research topics of 'Comprehensive overview of common e-liquid ingredients and how they can be used to predict an e-liquid's flavour category'. Together they form a unique fingerprint.

Cite this