Article Text

Download PDFPDF
Identifying dietary patterns using a normal mixture model: application to the EPIC study
  1. Michael T Fahey1,
  2. Pietro Ferrari2,
  3. Nadia Slimani3,
  4. Jeroen K Vermunt4,
  5. Ian R White1,
  6. Kurt Hoffmann5,
  7. Elisabet Wirfält6,
  8. Christina Bamia7,
  9. Mathilde Touvier8,9,
  10. Jakob Linseisen10,
  11. Miguel Rodríguez-Barranco11,
  12. Rosario Tumino12,
  13. Eiliv Lund13,
  14. Kim Overvad14,
  15. Bas Bueno de Mesquita15,
  16. Sheila Bingham16,
  17. Elio Riboli17
  1. 1Biostatistics Unit, Medical Research Council, Cambridge, UK
  2. 2Data Collection and Exposure Unit, European Food Safety Authority, Parma, Italy
  3. 3Dietary Exposure and Assessment Group, IARC, Lyon, France
  4. 4Department of Methodology, University of Tilburg, The Netherlands
  5. 5German Institute of Human Nutrition, Postdam-Rehbrücke, Germany
  6. 6Department of Community Medicine, Lund University, Malmö, Sweden
  7. 7Department of Hygiene and Epidemiology, University of Athens Medical School, Greece
  8. 8INSERM ERI-20, Institut Gustave-Roussy, Villejuif, France
  9. 9AFSSA, Maisons Alfort, France
  10. 10Institute of Epidemiology, Helmholtz Centre Munich, Neuherberg, Germany
  11. 11Andalusian School of Public Health, Granada, Spain
  12. 12Cancer Registry, Azienda Ospedaliera Civile-M.P. Arezzo, Ragusa, Italy
  13. 13Institute of Community Medicine, University of Tromsø, Norway
  14. 14Department of Clinical Epidemiology, Aarhus University Hospital, Aalborg, Denmark
  15. 15National Institute for Public Health and the Environment (RIVM), Bilthoven, The Netherlands
  16. 16MRC Dunn Human Nutrition Unit and MRC Centre for Nutritional Epidemiology and Cancer Prevention, UK
  17. 17Division of Epidemiology, Public Health and Primary Care, Imperial College London, UK
  1. Correspondence to Michael T Fahey, Private Bag 33, Clayton South, VIC 3169 Australiamichael.fahey{at}


Background Finite mixture models posit the existence of a latent categorical variable and can be used for probabilistic classification. The authors illustrate the use of mixture models for dietary pattern analysis. An advantage of this approach is taking classification uncertainty into account.

Methods Participants were a random sample of women from the European Prospective Investigation into Cancer. Food consumption was measured using dietary questionnaires. Mixture models identified latent classes in food consumption data, which were interpreted as dietary patterns.

Results Among various assumptions examined, models allowing the variance of foods to vary within and between classes fit better than alternatives assuming constant variance (the K-means method of cluster analysis also makes the latter assumption). An eight-class model was best fitting and five patterns validated well in a second random sample. Patterns with lower classification uncertainty tended to be better validated. One pattern showed low consumption of foods despite being associated with moderate body mass index.

Conclusion Mixture modelling for dietary pattern analysis has advantages over both factor and cluster analysis. In contrast to these other methods, it is easy to estimate pattern prevalence, to describe patterns and to use patterns to predict disease taking classification uncertainty into account. Owing to substantial error in food consumptions, any analysis will usually find some patterns that cannot be well validated. While knowledge of classification uncertainty may aid pattern evaluation, any method will better identify patterns from food consumptions measured with less error. Mixture models may be useful to identify individuals who under-report food consumption.

  • Mixture model
  • dietary pattern
  • latent variable
  • food consumption
  • factor analysis
  • cluster analysis
  • EPIC
  • epidemiology
  • multivariate analysis
  • nutrition
  • statistics

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • The co-authors Kurt Hoffmann and Sheila Bingham are now deceased.

  • Funding The EPIC study was supported by the ‘Europe Against Cancer’ Programme of the European Commission (SANCO); Ligue contre le Cancer (France); Société 3M (France); Mutuelle Générale de l'Education Nationale; Institut National de la Santé et de la Recherche Médicale (INSERM); German Cancer Aid; German Cancer Research Center; German Federal Ministry of Education and Research; Danish Cancer Society; Health Research Fund (FIS) of the Spanish Ministry of Health; Spanish regional governments of Andalusia, Asturias, Basque Country, Murcia, Navarra and ISCIII; Red de Centros RCESP, C03/09; Cancer Research UK; Medical Research Council, UK; the Stroke Association, UK; British Heart Foundation; Department of Health, UK; Food Standards Agency, UK; the Wellcome Trust, UK; Greek Ministry of Health; Greek Ministry of Education; Italian Association for Research on Cancer; Italian National Research Council; Dutch Ministry of Public Health, Welfare and Sports; Dutch Ministry of Health; Dutch Prevention Funds; LK Research Funds; Dutch ZON (Zorg Onderzoek Nederland); World Cancer Research Fund (WCRF); Swedish Cancer Society; Swedish Scientific Council; Regional Government of Skane, Sweden; and the Norwegian Cancer Society. IRW was supported by Medical Research Council grant U.1052.00.006.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.