Comparison of an Australian food-frequency questionnaire with diet records: implications for nutrition surveillance

Our plain-English summary of Comparison of an Australian food-frequency questionnaire with diet records: implications for nutrition surveillance, by G.L. Ambrosini, D. Mackerras, N.H. de Klerk, et al (Public Health Nutrition, 2003).

The study

• Authors: G.L. Ambrosini, D. Mackerras, N.H. de Klerk, A.W. Musk
• Journal: Public Health Nutrition
• Published: 2003;6(4):415-22
• Cited by: 37 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

Objectives: To compare a widely used Australian food-frequency questionnaire (FFQ) with diet records and consider the results in relation to its use in nutrition surveillance.

Design: Inter-method reliability study.

Setting: A randomised trial in subjects with past asbestos exposure.

Subjects: Seventy-two adults living in Western Australia.

Methods: A semi-quantitative FFQ developed by the Commonwealth Scientific Industrial Research Organisation in South Australia was administered after the completion of four 7-day diet records (DRs).

Results: Mean agreement between methods was not significantly different from 100% for many nutrients, but the limits of agreement indicated that, at the individual level, the FFQ over- or underestimated the DR by at least 50%. Mean agreement between methods decreased significantly with increasing intakes for the majority of nutrients. Pearson's correlation coefficients were less informative indicators of agreement compared with the limits of agreement.

Conclusions: These results indicate poor agreement between the FFQ and DR when estimating absolute intakes. Therefore, comparing intakes collected using this FFQ with specific cut-off points such as Recommended Dietary Intakes for nutrition surveillance may lead to seriously flawed conclusions about population intakes.

In plain English

This study did not measure asbestos exposure or disease. It was carried out among 72 adults (48 men and 24 women) in Western Australia who were taking part in a randomised trial of beta-carotene and retinol supplements. The trial participants were former workers and former residents of a blue asbestos mining town. The authors compared a widely used Australian food-frequency questionnaire, developed by CSIRO, with four 7-day diet records kept between December 1990 and October 1991. One man was excluded from the analysis, leaving 47 men. For men, the questionnaire's average estimate ranged from 70% (beta-carotene) to 127% (sugars) of the diet-record estimate. For women, it ranged from 90% (saturated fat) to 163% (vitamin C). For individuals, the questionnaire could under- or overestimate many nutrients by at least 50%. For many nutrients, agreement became worse as intakes increased. By the authors' criteria, agreement in women was acceptable only for starch and percentage of energy from carbohydrate. The authors concluded that this questionnaire does not permit confident comparison of absolute intakes against cut-off points such as Recommended Dietary Intakes. They noted that their numbers were small.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Inter-method reliability study comparing a food-frequency questionnaire with diet records, conducted within a randomised trial
• Who was studied: Volunteers from a Western Australian randomised trial of beta-carotene and retinol supplements. Trial participants were ex-workers and ex-residents of a blue asbestos mining town. Of 118 volunteers, 83 (57 men, 26 women) completed four 7-day diet records, and 72 of these (48 men, 24 women) also completed the questionnaire. One man was excluded, leaving 47 men in the analysis. Diet records were kept from December 1990 to October 1991.
• Exposure: Not an asbestos exposure study. The comparison was between dietary assessment methods: the 172-item 'Questan' version of the CSIRO semi-quantitative food-frequency questionnaire, and four 7-day diet records. Participants had past asbestos exposure, but asbestos was not analysed.
• Main results:
  • Men (n=47): the questionnaire's mean estimate ranged from 70% (beta-carotene) to 127% (sugars) of the diet-record estimate.
  • Women (n=24): mean agreement ranged from 90% (saturated fat) to 163% (vitamin C).
  • In men, the limits of agreement showed that the questionnaire could under- or overestimate many nutrients by at least 50%. For example, a man with a diet-record energy intake of 5000 kJ could have a questionnaire estimate anywhere between 2850 kJ (57%) and 7000 kJ (140%).
  • Many nutrients showed a significant dependency between agreement and the size of intake. Most often agreement became worse as intakes increased.
  • By the authors' criteria, agreement in women was acceptable only for starch and percentage of energy from carbohydrate. In men, agreement was adequate for all nutrients except alcohol, beta-carotene, cholesterol, total fat and all fat types, percentage of energy from fat, potassium, retinol, retinol equivalents, sugars and vitamin C.
  • In both men and women, most correlation coefficients between the two methods were strong enough that they would often be read as 'good agreement'. Generally, the nutrients with the strongest correlations were not the ones identified as having adequate agreement.
  • Among women, the proportion with calcium intakes below the Recommended Dietary Intake was 58% by the questionnaire and 92% by the diet records.
• The authors' conclusion: The authors concluded that the large discrepancies between the questionnaire and the diet records mean that questionnaire intakes do not permit confident assessment of absolute intake against specific cut-off points such as Recommended Dietary Intakes or dietary goals.
• Limitations the authors state:
  • This was the third food-frequency questionnaire participants completed that year. Any training effect could mean agreement was better than would be expected in the wider population.
  • The number of women was small. The authors said this probably explains the wider limits of agreement among women.
  • The apparent differences in questionnaire performance between men and women need confirmation because numbers were small.
  • Both methods generally underestimated intakes, as shown by low ratios of energy intake to basal metabolic rate.

Topics

Medical terms in this study, explained in our Lexicon: Asbestos Exposure.

Related studies

• Agreement between a brief food frequency questionnaire and diet records using two statistical methods — G.L. Ambrosini, N.H. de Klerk, et al. (2001)
• Records of the Australian Mesothelioma Surveillance Program have been lost! — M. Hendrickx (2008)
• Migration and work in postwar Australia: mortality profile comparisons between Australian and Italian workers exposed to blue asbestos at Wittenoom — A. Reid, E. Merler, et al. (2018)
• Surveillance of Australian workplace Based Respiratory Events (SABRE): notifications for the first 3.5 years and validation of occupational asthma cases — D. Elder, M. Abramson, et al. (2004)
• Comparison of exposure assessment methods in a lung cancer case-control study: performance of a lifelong task-based questionnaire for asbestos and PAHs — E. Bourgkard, P. Wild, et al. (2013)
• Health-related quality of life in patients with lung cancer: validation of the Mexican-Spanish version and association with prognosis of the EORTC QLQ-LC13 questionnaire — Ó. Arrieta, C. Núñez-Valencia, et al. (2012)

Browse return to medical research.