- Click here for explanation of classification scheme.

To identify consumption patterns or habits that classify people according to the types of food they choose to consume, rather than based on self-definition or on the basis of nutrient intake alone.

Women from the UK Women's Cohort Study.

No apparent exclusion criteria within the UK Women's Cohort Study.

Recruitment

Women for the UK Women's Cohort Study were identified from questionnaires sent between 1995 and 1998 to 61,000 potential subjects. Subjects were identified based on responses to a 217-item food frequency questionnaire (FFQ).

Design

The goal was to identify homogeneous clusters of individuals based on responses to the FFQ using cluster analysis techniques.

The authors then examined the relationships between each cluster and sociodemographic characteristics, indicators of health and diet, and nutrient intakes of the different clusters of women identified by the cluster analysis.

Blinding used

None.

Intervention

None.

Statistical Analysis

Cluster analysis was done using a k-means cluster algorithm. Robustness of the identified clusters was checked by trying different initial cluster centers and by using different rules of convergence. Other characteristics were then tested for association with the identified clusters by using non-parametric analysis of variance for continuous variables and chi-square tests for dichotomous variables.

Because of the large size of the cohort, the relationship between clusters and other characteristics were all statistically significant at P<0.001, even when the differences among clusters were of little practical importance.

Timing of Measurements

 Questionnaires were sent between 1995 and 1998.

Factors

• Diet cluster: Based on results from a 127-item FFQ designed to capture food intake patterns over the previous 12-month period. The FFQ was adapted from the one used in the European Prospective Investigation into Cancer (EPIC) study
• Food frequency: Measured on a 10-point scale from "never" to "six+ times per day"
• Sociodemographic characteristics
• Indicators of health and diet
• Nutrient intakes.

 

Initial N

61,000 questionnaires originally sent. 33,971 women were selected.

Attrition (final N)

NA.

Age

35 years to 69 years at recruitment; mean 52±9 years.

Ethnicity

Not reported.

Other relevant demographics

• Generally well-educated (43% educated to "A" level standard or beyond)
• Mean age leaving full-time education was 18±4 years
• 55% held managerial, administrative, professional, technical or association professional positions
• 30% described themselves as vegetarian
• 58% take vitamins, minerals, fish oils, fiber or other food supplements (substantially higher than in the national population)
• 11% were current smokers.

Anthropometrics

• BMI mean 24.4±4.3 [compared to 24.8 (age 35-49) and 26.2 (age 50-64) among women in the national population]
• 9% were obese (BMI ≥30)
• Mean waist-to-hip ratio is 0.75±0.06.

Location

UK.

The authors identified seven natural clusters based on relatively high or low levels of consumption of particular foods. This served to define clusters based on diet diversity as well as consumption of types of foods.

Food Consumption Patterns

The authors identified cluster membership based on food consumption patterns, not on the subjects' self-definition. The seven clusters and the characteristic food types consumed are presented in Table 1.

Cluster	High-quantity foods	Moderate-quantity foods	Low-quantity foods
1. Monotonous low-quantity omnivores	White bread, milk, sugar	Potatoes, meat	Most other foods, especially bran and whole-meal food, soy products, yogurt, nuts, pulses, fruit and vegetables, alcohol
2. Health-conscious omnivores	Bran, potatoes, whole-meal food, yogurt, low-fat dairy products, pulses, fish, vegetables, salad, fruit	Most other foods including meat and wine	Chips, sugar
3. Traditional meat, chips and pudding eaters	White bread, chips, meat, sugar, high-fat and creamy food, biscuits, cakes	Most other foods including alcohol	Whole-meal food, soy products, vegetables, salad, fruit
4. Higher diversity, traditional omnivores	Chips, white pasta and rice, high-fat and creamy food, eggs, meat, fish, chocolate, biscuits, crisps. More fish and salad than cluster 3. More diversity than cluster 3	Vegetables, fruit, alcohol	Less cakes and puddings than cluster 3
5. Conservative omnivores	None reported	Most food, including potatoes, meat, fish, eggs, fruit, vegetables	Cereals, chips, whole-meal food, nuts, pulses, spreads and dressings, chocolate, crisps, biscuits. Less red meat, less chips and less puddings than clusters 3 and 4.
6. Low-diversity vegetarians	Whole-meal bread, soy products, pulses, fruit (apart from exotic fruit) vegetables	Cereals
7. High-diversity vegetarians	Whole-meal bread, cereals, whole-meal pasta and rice, soy products, spreads, nuts, pulses, vegetables, fruit herbal tea (generally higher consumption of these products than low diversity vegetarians)	None reported

The authors point out that it is not just the food type chosen, but the diversity of foods eaten that help define the clusters. In the case of the two vegetarian groups, it is the range of different foods eaten, and not just the frequency of specific food items, that differentiates the two groups.

 

Defining a Vegetarian: Self-Description Vs. Empirical Identification

 

The authors find that some women who identify themselves as vegetarian fell within the non-vegetarian clusters based on their actual consumption patterns. Similarly, some women who defined themselves as non-vegetarian fell into the vegetarian clusters.

How did this happen?

“Vegetarians” in the Non-Vegetarian Group

Even though some women define themselves as vegetarian, their food consumption patters are more similar to women in the non-vegetarian clusters. The authors found that:

• 59% of self-described vegetarians who fell into a non-vegetarian group eat meat (most of which is fish)

• 12% of “vegetarians” in the non-vegetarian group eat red meat, although less frequently that self-described non-vegetarians in the non-vegetarian groups [0.02 portions per day compared to 0.87 portions per day (P<0.001)]

Women who call themselves vegetarians may have food-intake patterns more similar to non-vegetarians in terms of overall food intake and diversity, but reduce or cut out red meat. It may be the way that the individual thinks about their red meat intake (rather than their actual red meat intake) that leads them to define themselves as vegetarian.

“Non-Vegetarians” in the Vegetarian Group

Non-vegetarians in the two vegetarian groups have overall food consumption patterns more like vegetarians than members of any of the three non-vegetarian groups. The authors surmise that this group may follow a vegetarian diet as far as possible, but occasionally allow themselves fish.

It may be for these women who describe themselves as non-vegetarian that the presence of any meat in their diet immediately disqualifies them as vegetarian.

 

How Do Groups Compare of Various Health and Dietary Characteristics?

The authors describe the dietary and health characteristics of the different clusters. Even though they fall into an omnivore category, health-conscious omnivores were more likely to consume at least five portions of fruits and vegetables per day as well as spend more time vigorously exercising then members of the two vegetarian groups. Also, fully one fifth of the members of the monotonous low-quantity omnivore group described themselves as vegetarians.

Cluster	Describes herself as vegetarian	Uses food supplements	Consumes five portions of fruits and vegetables per day	Median time in vigorous physical exercise (minutes per week)
Monotonous low-quantity omnivores	20%	53%	5%	0
Health-conscious omnivores	35%	65%	30%	70
Traditional meat, chips and pudding eaters	3%	50%	4%	20
Higher diversity, traditional omnivores	5%	55%	12%	60
Conservative omnivores	16%	56%	14%	30
Low-diversity vegetarians	71%	66%	12%	60
High-diversity vegetarians	75%	69%	23%	60

In terms of practical import, the authors point out that

• Not all diets that are healthful in one respect are healthful in all respects (for example, a higher fat diet may also be high in fruits and vegetables)

• There are complex interactions between nutrients, other dietary constituents, and the processing and preparation of foods that influence the health effects of diets.

 

Classifying diets in more pragmatic terms than just nutrient intake should provide valuable insight to complex diet-disease relationships. Dietary advice, although based on nutrient content of meals, needs to take account of the combinations of different food types that people naturally choose to use together.

Not-for-profit

0 Foundation associated with industry:

Strengths

• Well-designed and executed study
• Authors were thorough in reporting the limitations of the study.

Weaknesses

The sample was demonstrably different than the larger population on several factors--most important of which may be the level of health consciousness, income and education. It is unclear how the clusters identified in the sample might change if a representative sample from the larger population was examined.