Doctor of Philosophy (PhD)
Department of Biomedical Science - Faculty of Health & Behavioural Sciences
Martin, Gina S, The interviewer-administered, open-ended diet history method for assessing usual dietary intakes in clinical research: relative and criterion validation studies, Phd thesis, Department of Biomedical Science, UNiversity of Wollongong, 2004. http://ro.uow.edu.au/theses/204
In this thesis three studies are reported, which aim to investigate the performance of an interviewer-administered, open-ended diet history for the use in clinical studies where dietary variables are manipulated. The studies presented here are focused on validating self-reported macronutrient and energy intakes in Australian adults who have volunteered to participate in dietary intervention research. The evaluations are performed using food records, 24-hour recalls and biochemical markers of intake for comparison. In addition, a pilot study is presented, which examines the ability of hair and sebaceous lipid composition to reflect dietary fatty acid intake in healthy volunteers. Clinical trials are the highest form of evidence for examining both the treatment and prevention of disease by dietary nutrients. Given this, it is imperative that these nutrients are measured as accurately as possible to avoid erroneous conclusions regarding their capabilities. Despite the importance of data from clinical trials, the literature relating to the performance of dietary assessment methods in this context is sparse. Despite evidence to show that food records and 24-hour recalls are not always able to measure intake accurately, many dietary intervention trials are using these methods to measure dietary intake during the intervention period. Since the diet history method is used in clinical practice, albeit in a less rigorous manner, and the clinical trial is, in essence, a highly controlled form of clinical practice, examination of the diet history method in this research context is warranted. The first study in this thesis examined the validity of an interviewer-administered, open-ended, non-structured diet history in adults with Type II diabetes mellitus. The trial examined the viability of a high monounsaturated fat diet for the treatment of associated blood and insulin abnormalities. The dietary data were obtained retrospectively and the diet history was examined relative to a 3-day food record using a number of statistical techniques. The results showed that the diet history method was able to measure both energy and macronutrient intakes reasonably well at both the group level and the individual level. There was, however, a failure to measure the fatty acid composition of the diet, which was due to a few outliers who reported discrepant values. Nethertheless, the relationship between paired data for these variables was significant once outliers were removed. There was no relationship between bias and intake indicating that systematic error was not operating in this group. An important finding from this study was the large proportion of individuals underreporting energy intake. This was not method-specific, but rather subject-specific and tended to occur in the larger individuals, however the use of a nutrient density model diluted in the effects of underreporting for comparisons of the diet history with the food record. The non-structured diet history method was found to estimate energy and nutrient intakes relatively well in a clinical population with Type II diabetes. The second study also used retrospective data from both the former study and another clinical trial, which examined the effects of a high monounsaturated fat diet on the prevention of diabetes in Australian adults with mild insulin resistance. The aim of this study was the compare the performance of the interviewer-administered, non-structured diet history between healthy individuals and individual with Type II diabetes mellitus undergoing similar intervention protocols with differing intensity. The results showed that the values measured by the DH in the healthy individuals were closer to those of the food record than in the individuals with Type II diabetes. Moreover, the degree of underreporting in the healthy individuals was significantly lower than in individuals with diabetes. Biases in the measurements of target nutrients also showed movement in different directions between the two sample groups. The third study was designed to fit within the research protocol of a dietary supplementation trials, which examined the effects of fish oil and soy isoflavone on the risk profile for cardiovascular disease. The study presented in this thesis aimed to assess the validity and reproducibility of a diet history method using both food records and multiple 24 hour recalls for reference, in addition to urine nitrogen and erythrocyte membrane phospholipid fatty acid composition as biomarkers of dietary intake. The biochemical markers not only provided information as to the performance of diet history method, but also the reference methods to determine if discrepancies between paired data were indeed due to diet history failings. The diet history method used in this study was slightly different to the aforementioned non-structured approach with the record of foods taking place on a structured meal-based form with food prompts for the dietitian. The allowed the dietitian to prompt for the context of eating as well as the foods consumed. Since the concern in this study was the accuracy of the data for a valid trial outcome, prompting was not considered to introduce bias, but rather to improve memory of foods consumed. The results showed that the diet history method was measuring higher absolute intakes of energy and macronutrients than both the food records and the 24-hour recalls, however, both reference methods showed a higher degree of underreporting and an underestimation of nitrogen intake of energy and macronutirents than both the food record and the 24 hour recalls, however, both reference methods showed a higher degree of underreporting and an underestimation of nitrogen intake during the course of the intervention indicating that the reference methods were prone to underestimation. The diet history values also moved in line with those of the food record and the 24-hour recall. In addition, the diet history was shown to be better at estimating long chain fatty acid intake than the food record for variables, which were of importance in a fish oil supplementation trial. Nevertheless, this study showed that the structured diet history method was both valid and reproducible for estimating macronutrient and energy intake in a sample of Australian adults participating in a dietary supplementation trial. The final study was designed in response to the lack of biomarkers available for fat intake. This study was pilot study, which aimed to investigate if the fatty acid composition of both hair and sebaceous secretions was related to the fatty acid composition of the diet. If a relationship was shown then a potential non-invasive method for assessing what may be a single or a few fatty acids in the usual diet was possible. The results showed that there were no significant relationships between hair or sebaceous fatty acid composition and corresponding dietary fatty acids due, in part, to large numbers of undetectable fatty acids in the different tissues, which prevented correlation. Future research will indicate if hair or sebaceous fatty acid composition is valuable for assessing changes in dietary fat. As a result of the studies presented in this thesis, the interviewer-administered, structured diet history method has been used in numerous studies to date conducted through the author�s institution and requests for its use continue as clinical dietitians conduct research within their own institutions. If anything, this thesis has shown that relative validation of dietary methods is valuable and should be done in context. The issues that affect persons participating in trials with large subject burdens are not the same for those who undertake just a validation study. The methods used in intervention trials needs to be validated within the context of the trial and its participants. However, objective markers should always be used as an adjunct to any validation study especially those that inform on absolute nutrient intake such as urine nitrogen. This thesis reports on a number of interesting limitations associated with measuring dietary intake in free-living individuals especially in larger individuals in the sample group. A number of interesting macronutrient-specific biases are also reported on in this work. Further research in dietary methodology should focus on the need for contextual validation of dietary methods since the administration environment affects the accuracy of the test method. In addition, clinical trails, which are dependent on nutrient targets for outcome, should make validation studies a priority and should question the accuracy of the reference methods used for validation purposes. This thesis has shown that the diet history interview is a valuable method for measuring dietary intake in clinical trials in free-living individuals. Of importance, is the contextual aspect of administration where the diet history would need to be validated in specific research environments, however, the studies presented here provide insight into the performance of the diet history method in clinical research.