Nutritional Epidemiology Module assignment

Instructions

The Module assignment is as detailed below. Please submit a single Word file or PDF that addresses all components of the assignment. The entire task should be completed individually (it is not intended to be a group exercise). An approximate proportion of scoring/grading (of 100%) is supplied for each question or task, e.g. (5%).

Any numeric information or figure must not be submitted in an artificial intelligence (AI) tool.

Your answers are expected to consist of up to 1500 words in total and to be submitted in a single file.

Selected publications that are relevant to this assignment are available as PDFs. Each file name starts with the first author’s name. Some publications are essential to enabling responses to the questions posed in the assignment. Not all publications are directly relevant, but uploaded partly as resources to learn.

Assignment

Research setting:

In an ongoing prospective cohort study in the UK, participants were recruited from the general practice (family doctor) registries. At its baseline (1993-1997), 24,857 participants underwent dietary assessment with a food-frequency questionnaire and many other measurements. Then, they were followed up over years for further assessment and for identification of deaths and disease onset. Recently, a research group has conducted an epidemiological analysis to address the following hypotheses:

(i) Dementia incidence would vary by the degree of adherence to a plant-based dietary pattern. Greater adherence is associated with lower dementia incidence when the plant-based dietary pattern is characterised by high consumption of healthy plant-based foods or beverages.

(ii) Dementia incidence would vary by the degree of adherence to a plant-based dietary pattern. Greater adherence is associated with higher dementia incidence when the plant-based dietary pattern is characterised by high consumption of unhealthy plant-based foods or beverages.

The investigators calculated dietary pattern scores of the current UK study participants (n=24,857) so that the scores reflected the degrees of adherence to two types of the plant-based diet patterns according to the approach previously taken by other researchers, e.g. Satija et al. (Satija et al., Healthful and Unhealthful Plant-Based Diets and the Risk of Coronary Heart Disease in U.S. Adults, J Am Coll Cardiol, 2017;70(4):411-422 doi: 10.1016/j.jacc.2017.05.047). The two scores were named healthy plant-based diet index (“healthy PDI”) for the diet high in healthy plant-based foods and unhealthy plant-based diet index (“unhealthy PDI”) for the diet high in unhealthy plant-based foods.

In the current UK study, dementia cases from the baseline to the end of follow-up (31 March 2022) were ascertained by record linkage with death certificates and hospital records and, using diagnostic records of International Classification of Diseases (ICD) codes for dementia: A81.0, F01, F01.0-F01.3, F01.8, F01.9, F02, F02.0-F024, F02.8, F03, F04, F05.1, F10.7, G31.0, G31.8, and I67.3. Other data collection methods for covariates are available in numerous publications, such as Tong et al., (Prospective association of the Mediterranean diet with cardiovascular disease incidence and mortality and its population impact in a non-Mediterranean population: the EPIC-Norfolk study, BMC Med, 2016;14(1):135 (doi: 10.1186/s12916-016-0677-4)). Briefly, socioeconomic factors, behavioural factors, medication use, and disease histories of participants and their family members were assessed with self-administered questionnaires. Anthropometric variables were obtained by trained research staff.

The investigators conducted descriptive analyses to characterise the study participants and multivariable-adjusted Cox proportional hazard regression analysis to characterise the association of each of healthy and unhealthy PDIs at baseline (1993-1997) with incidence of dementia ascertained by 2022. Some of the results are presented below.

Assignment questions:

Q1-1. Provide the PICO for each of the two hypotheses (PICO: Population, Intervention hypothesised, Comparison undertaken, and Outcome assessed) (3%).

Q2. Table S1 presents food grouping in the UK study. The food groups and their scoring were conducted in a consistent way (as done by Satija et al.). Certain plant-based food groups received positive scores while others were scored negatively.

Q2-1. Identify one food group or food item that is classified as healthy or unhealthy plant-based item and could be controversial in its scoring in the context of different countries, populations or population sub-groups. Then, please explain why the classification could be controversial and how the specific item could be classified differently. (5%)

Q3. Assume you are the investigator analysing the data from the UK study and decide to add one more component or split one food group into two (in Table S1).

Q3-1. Please nominate one dietary component you may include by adding it or splitting one component in the two PDIs (i.e. healthy and unhealthy PDIs). (6%)

Q3-2. Suppose you calculate each PDI after the modification (adding one component), according to the approach taken by Satija et al. What is the possible range of each PDI score? (4%)

Q4. Table 1 shows descriptive statistics by the healthy PDI.

Q4-1. According to the results in Table 1, select two covariates and discuss how each could cause confounding in the longitudinal analysis relating the healthy PDI to dementia incidence. Clarify the direction of confounding. (6%)

Q4-2. Covariates that indicate a crude correlation with the healthy PDI score do not necessarily act as confounders in the association of the PDI with dementia incidence. Describe two reasons for it. (6%)

Q5. Table 1 includes the descriptive statistics for trans fat intake. Suppose that Trans fat was consumed from plant-based foods. While it could be a part of any of the PDI, the investigators adjusted for the variable in the longitudinal analysis of the UK study dataset (as described in the footnote of Table 2 shown below).

Q5-1. The statistical adjustment was also made by Satija et al. Discuss why Satija et al. conducted the adjustment to examine the association independent of trans fat intake or separate out the potential effect of trans fat intake. (5%)

Q5-2. Discuss whether the same adjustment should be made in the UK study analysis and provide the rationale. (5%)

Q6. The investigators conducted Cox proportional hazard regression analyses and presented the results in Table 2 (for the UK study).

Q6-1. In this study, the investigators were interested in the diet-dementia association independent of obesity, even if obesity could be a mediator. They reasoned that obesity could be a confounder because it can cause measurement errors in self-reported dietary intakes systematically and may increase dementia incidence independently. Suppose that obesity caused under-reporting of all of the unhealthy plant-based items and animal-based items. Discuss which direction of confounding bias this measurement error would cause for an association of the unhealthy PDI with dementia incidence. (6%)

Q6-2. Using the numeric information in Table 2, describe, in two ways, the main result that is most relevant to the study hypothesis about a healthy plant-based diet: one, under the assumption of causality as if the estimate was obtained from a perfect trial (or a meta-analysis of trials); and the other, without any assumption of causality. (8%)

Q6-3. The investigators statistically adjusted for total energy intake. State at least three aims of the statistical adjustment for total energy intake in this study: through the adjustment, what specific bias did the investigators attempt to minimise in this particular study or what study-specific aim did the investigators attempt to achieve? If necessary, refer to the course materials and the published information on energy adjustment in nutritional epidemiology, such as Willett et al., Adjustment for total energy intake in epidemiologic studies, Am J Clin Nutr, 1997;65(4):1220S-1228S. (8%)