Epidemiologist and Research Scientist, Alberta Health Services - CancerControl
Should We Still Be Collecting Self-Reporting Dietary Data? Of Course We Should!
Diet is a key determinant of health, as it plays a pivotal role at all stages of the lifecycle and is relevant to chronic disease prevention, treatment and management, as well as healthy aging. Informative and reliable assessment of dietary intake is thus fundamental to health research. Yet, this is an area plagued by considerable misconceptions and challenges. Through this symposium, we will identify fundamental gaps in knowledge and resources that researchers are facing when attempting to assess dietary intake in their research, with a focus on issues of particular salience to Canadian researchers. For example, there will be discussion of the value of investing significant efforts to enhance the comprehensive quality of the Canadian Nutrient File to support surveillance efforts such as the Canadian Community Health Survey but also other types of research in Canada. There will also be discussion of various dietary assessment methods, including self-report tools and biomarkers, and their relevance to various research applications and settings. The overarching goal of this symposium is to engage the nutrition community in discussions about issues identified by the Partnership for Advancing Nutrition and Dietary Assessment in Canada (PANDA-C) initiative, which aims to support the evolution of nutritional and dietary assessment in Canada, with a longer-term goal of ensuring that valid and reliable assessment tools and strategies are available for use by Canadian researchers.
Considering and enhancing the value of self-report dietary intake data
Dr. Sharon I. Kirkpatrick
Measuring diet with as much accuracy as possible is fundamental to nutrition research. Given that objective measures of habitual diet (e.g., biomarkers) are known for only a few dietary components and tend to be cost prohibitive, there is a continued need to rely on self-report measures of dietary intake. In addition to their feasibility, self-report data provide details that biomarkers cannot, including the foods and drinks consumed, patterns of eating, and contextual factors such as the location of eating. However, it has long been recognized that data collected using self-report instruments are affected by varying types and degrees of measurement error, depending on the specific measure, population, and research context. This error contributes to inconsistencies in the literature and can obscure our understanding of relationships between diet and disease as well as the usefulness of interventions for improving diet. The growing evidence base demonstrates that study design (including the selection of the most appropriate measure for a given purpose) and analytic strategies can mitigate measurement error, yet these strategies remain underutilized. Improving the translation and application of existing and emerging evidence related to dietary assessment has the potential to enhance the impact of research for informing interventions to improve nutrition and health. Increased uptake of the best possible appropriate diet assessment strategies for particular purposes can foster consistent methodology across studies, helping to build a more coherent evidence base as well as facilitating innovative pooled analyses to answer questions that are unable to be addressed in a single study.
Food Frequency Questionnaires: Not Irrelevant Yet – the Debate 10 Years Later
Dr. Ilona Csizmadi
Because they are easy to administer and are designed to capture usual food intake, food frequency questionnaires (FFQs) are commonly used to assess diet in etiologic studies of chronic disease. However, comparison of FFQs against objective recovery biomarkers has revealed substantial systematic error associated with certain nutrients. This observation has led to intensive debate over the past decade about their utility in dietary assessment. While some groups have called for the abandonment of FFQs, along with other self-reporting methods in dietary assessment, others have emphasized the importance of better understanding the sources of measurement error in self-reporting methods with the aim of developing novel strategies to minimize their impact on intake and disease risk estimates. These strategies include the use of multiple instruments with complementary strengths (e.g., 24-hour recalls and FFQs), biomarker sub-studies to calibrate self-report assessments and/or sensitivity analyses that examine differences in study findings across misreporting categories. While advances in analytical methods are critical, they cannot compensate for deficiencies in the questionnaire design itself. Hence, the fundamental principles of FFQ design remain a high priority. These include ensuring that FFQs are designed and validated for the populations in which they are used, include important and discriminating foods for nutrients of interest, and that underlying nutrient databases reflect the composition of foods consumed in the populations of interest. Issues specific to the Canadian context will be highlighted.
Validation of a popular calorie counting tablet application with a weighed food record
Mavra Ahmed (Abstract Presentation)
Adopting innovative technology could overcome some of the barriers associated with traditional dietary assessment methods (DAM). The objective was to assess the relative validity of a popular tablet application, My Fitness Pal®, in comparison to a weighed food record and biomarkers. Participants were Canadian Armed Forces personnel (mean age 33 y), who recorded their usual dietary intake for 3 days using both the weighed food record (wFR) and tablet application. Mean differences in macronutrient intakes between the methods were evaluated with Student's t test. Pearson's and Spearman's correlation coefficients were used to examine relationship between nutrient intake data and biomarkers. The mean intake of energy was not significantly different (p>0.05) between the tablet application (2649kcal/d) and the wFR (2600kcal/d). Macronutrient intake did not differ between the two reporting methods. Significant, positive correlations were observed for energy (r=0.85), carbohydrates (r=0.81), protein (r=0.51), and fat (r=0.94) (p<0.05). Micronutrient intakes were significantly correlated, ranging from 0.54 to 0.98. Significant correlation was observed between urinary urea and dietary protein intake recorded by the wFR method (r=0.52, p<0.05) but not by the tablet method. This research aimed to fill the gap on the potential of innovative technologies in enhancing dietary assessments through cost-‐ and time-‐ effective data collection while minimizing the burden on both the participant and coordinator. Although promising as an alternate DAM, future studies should explore the validity of tablet applications in larger sample sizes and with respect to accuracy of micronutrient/urinary biomarkers.
Validation of a web-based dietary recall performed in the context of a fully-controlled feeding study
Jacynthe Lafrenière (Abstract Presentation)
Robust evaluation of self-reported food intake is the cornerstone of many areas related to nutrition research. As technological instruments have demonstrated a good acceptance and facilitate data collection on large cohorts, we have developed a new self-administrated web-based 24-hour food recall (R24W). Inspired by the USDA Automated Multiple-Pass Method, this tool includes 2865 food items and 687 recipes for mixed dishes. One of the validation strategies used to test this tool was to ask 20 women and 14 men for whom true intakes were known to fill out the R24W. These participants were already engaged in fully-controlled feeding studies where all meals were prepared and provided by our research team according to a 7-day cyclic menu. We examined the proportion of matches between true and reported food items as well as the proportion of exclusions (items consumed but not reported) and inclusions (items not consumed but reported). We also used correlation analysis and weighted kappa value for the classification into quartiles to assess the association between true and reported portion sizes. The proportion of matches was 86.4% while the proportion of exclusions was 11.9%. It was also found that 1.7% of items reported were inclusions. Reported portion sizes were correlated with weighted offered portions (r=0.867, P<0.001) and the weighted kappa value was 0.64 indicating a good agreement. These observations indicate that the R24W provides a good representativeness of portion sizes and performs well by proposing enough options to match real intakes for the majority of the respondents.
Epidemiologist and Research Scientist, Alberta Health Services - CancerControl
Registered Dietitian and Assistant Professor, School of Public Health and Health Systems, University of Waterloo
Write your notes here: