Article Text

Download PDFPDF
OP88 What Factors Drive Unhealthy Diet? Novel Analysis of Food Frequency Questionnaire Data using Canonical Correspondence Analysis
  1. L A Goffe1,
  2. S P Rushton1,2,
  3. M White1,3
  1. 1Institute of Health and Society, Newcastle University, Newcastle-upon-Tyne, UK
  2. 2Biology, Clinical and Environmental Systems Modelling, School of Biology, Newcastle University, Newcastle-upon-Tyne, UK
  3. 3Fuse, UK Clinical Research Collaboration (UKCRC) Centre for Translational Research in Public Health, Newcastle-upon-Tyne, UK

Abstract

Background Diet is a major contributor to obesity, but the processes that influence intake remain unclear. Understanding the factors driving food consumption at the population scale requires detailed knowledge of the relationships between dietary patterns and socio-demographic factors. Objectively measuring diet is problematic, but food frequency questionnaires (FFQ) are an established, if imperfect, method for population studies. Such data is inherently multifactorial and various methods have been used to reduce dimensionality. We aimed to evaluate the role of socio-demographic factors in determining different dietary patterns using novel methods.

Methods Data were from a survey of dietary habits of 5044 adults in Newcastle-upon-Tyne, 2001–2002. This included the European Prospective Investigation of Cancer (EPIC) 134 item FFQ and socio-demographic data. We used canonical correspondence analysis (CCA), an extension of correspondence analysis that constrains the ordination to relate FFQ responses to relevant socio-demographic variables. Ordination methods summarise continuous trends according to socio-demographic variables, creating ‘components’ to describe trends in multivariate data. Results can be displayed graphically using scatter plots, showing the position of each food item and arrows, varying in length and direction, to show the strength of influence of each demographic variable.

Results Two main axes emerged from the FFQ data, describing dietary patterns. The first axis related to the level of processing; highly processed food items generally scoring low and raw ingredients scoring high. The second axis related to a Mediterranean diet; relevant items (e.g. pasta, rice, fruits and vegetables, cheese, yoghourt and wine) scored high, whereas those related to a traditional British diet (i.e. ‘meat and two veg’) scored low. Food consumption is highly individual but by constraining the ordination with reported age, dietary knowledge, educational attainment, socio-economic status and receipt of welfare benefits, it was possible to show their relationship to eating habits. These correlates accounted for 5.8% of variation, age exerting the strongest influence. Older people showed a strong preference for a British, low-processed food diet. University education was associated with consumption of a low-processed Mediterranean diet, whereas those living in deprived areas were associated with a highly-processed traditional British diet.

Discussion CCA provides a robust, descriptive and conceptually superior method for dietary pattern analysis using FFQ data. Results confirm findings of qualitative research, suggesting that diet is strongly culturally patterned among specific, identifiable population subgroups. Such work is limited by cross-sectional data, but CCA may enable better tailoring and targeting of interventions for those at greatest risk from an unhealthy diet.

Statistics from Altmetric.com

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.