Machine learning methods have been used in identifying omics markers for a variety of phenotypes. We aimed to examine whether a supervised machine learning algorithm can improve identification of alcohol-associated transcriptomic markers. In this study, we analysed array-based, whole-blood derived expression data for 17 873 gene transcripts in 5508 Framingham Heart Study participants. By using the Boruta algorithm, a supervised random forest (RF)-based feature selection method, we selected twenty-five alcohol-associated transcripts. In a testing set (30 % of entire study participants), AUC (area under the receiver operating characteristics curve) of these twenty-five transcripts were 0·73, 0·69 and 0·66 for non-drinkers v. moderate drinkers, non-drinkers v. heavy drinkers and moderate drinkers v. heavy drinkers, respectively. The AUC of the selected transcripts by the Boruta method were comparable to those identified using conventional linear regression models, for example, AUC of 1958 transcripts identified by conventional linear regression models (false discovery rate < 0·2) were 0·74, 0·66 and 0·65, respectively. With Bonferroni correction for the twenty-five Boruta method-selected transcripts and three CVD risk factors (i.e. at P < 6·7e-4), we observed thirteen transcripts were associated with obesity, three transcripts with type 2 diabetes and one transcript with hypertension. For example, we observed that alcohol consumption was inversely associated with the expression of DOCK4, IL4R, and SORT1, and DOCK4 and SORT1 were positively associated with obesity, and IL4R was inversely associated with hypertension. In conclusion, using a supervised machine learning method, the RF-based Boruta algorithm, we identified novel alcohol-associated gene transcripts.

Ultraprocessed foods provide the majority of energy content in the American diet, yet little is known regarding consumption trends over time. We determined trends in diet processing level and diet quality from 1991 to 2008 within the prospective Framingham Offspring Cohort. Dietary intakes were collected by FFQ quadrennially 1991–2008 (total of four examinations). The analytical sample included 2893 adults with valid dietary data for ≥3 examinations (baseline mean age = 54 years). Based on the NOVA framework, we classified foods as: unprocessed/minimally processed foods; processed culinary ingredients (salt/sugar/fats/oils); and processed foods and ultraprocessed foods. We evaluated diet quality using the Dietary Guidelines for Americans Adherence Index (DGAI) 2010. Trends in consumption of foods within each processing level (servings/d) and diet quality over the four examinations were evaluated using mixed effects models with subject-specific random intercepts. Analyses were stratified by sex, BMI (<25 kg/m2, 25–29·9 kg/m2, ≥30 kg/m2) and smoking status. Over 17 years of follow-up, ultraprocessed food consumption decreased from 7·5 to 6·0 servings/d and minimally processed food consumption decreased from 11·9 to 11·3 servings/d (P trend < 0·001). Changes in intakes of processed foods, culinary ingredients and culinary preparations were minimal. Trends were similar by sex, BMI and smoking status. DGAI-2010 score increased from 60·1 to 61·5, P < 0·001. The current study uniquely describes trends in diet processing level in an ageing US population, highlighting the longstanding presence of ultraprocessed foods in the American diet. Given the poor nutritional quality of ultraprocessed foods, public health efforts should be designed to limit their consumption.

Existing studies addressing alcohol consumption have not captured the multidimensionality of drinking patterns, including drinking frequency, binge drinking, beverage preference and changes in these measures across the adult life course. We examined longitudinal trends in drinking patterns and their association with diet over four decades in ageing US adults from the Framingham Offspring Study (n 4956; baseline mean age 36·2 years). Alcohol intake (drinks/week, drinking frequency, beverage-specific consumption, drinks/occasion) was assessed quadrennially from examinations 1 to 8. Participants were classified as binge drinkers, moderate drinkers or heavy drinkers (4+ and 5+ drinks/occasion; ≤1 and ≤2 drinks/d and >7 and >14 drinks/week for women and men, respectively). Dietary data were collected by a FFQ from examinations 5 to 8 (1991–2008). We evaluated trends in drinking patterns using linear mixed effect models and compared dietary intake across drinking patterns using heterogeneous variance models. Alcohol consumption decreased from 1971 to 2008 (3·7 v. 2·2 oz/week; P < 0·05). The proportion of moderate (66 v. 59·3 %), heavy (18·4 v. 10·5 %) and binge drinkers (40·0 v. 12·3 %) declined (P < 0·05). While average wine consumption increased (1·4 v. 2·2 drinks/week), beer (3·4 v. 1·5 drinks/week) and cocktail intake (2·8 v. 1·2 drinks/week) decreased. Non-binge drinkers consumed less sugary drinks and more whole grains than binge drinkers, and the latter consumed more total fat across all examinations (P < 0·05). There was a significant difference in consumption trends of total grains by drinking level (P < 0·05). In conclusion, alcohol drinking patterns are unstable throughout adulthood. Higher intakes were generally associated with poorer diets. These analyses support the nuanced characterisation of alcohol consumption in epidemiological studies.

Higher carbohydrate intake, glycaemic index (GI), and glycaemic load (GL) are hypothesised to increase cancer risk through metabolic dysregulation of the glucose-insulin axis and adiposity-related mechanisms, but epidemiological evidence is inconsistent. This prospective cohort study investigates carbohydrate quantity and quality in relation to risk of adiposity-related cancers, which represent the most commonly diagnosed preventable cancers in the USA. In exploratory analyses, associations with three site-specific cancers: breast, prostate and colorectal cancers were also examined. The study sample consisted of 3184 adults from the Framingham Offspring cohort. Dietary data were collected in 1991–1995 using a FFQ along with lifestyle and medical information. From 1991 to 2013, 565 incident adiposity-related cancers, including 124 breast, 157 prostate and sixty-eight colorectal cancers, were identified. Cox proportional hazards models were used to evaluate the role of carbohydrate nutrition in cancer risk. GI and GL were not associated with risk of adiposity-related cancers or any of the site-specific cancers. Total carbohydrate intake was not associated with risk of adiposity-related cancers combined or prostate and colorectal cancers. However, carbohydrate consumption in the highest v. lowest quintile was associated with 41 % lower breast cancer risk (hazard ratio (HR) 0·59; 95 % CI 0·36, 0·97). High-, medium- and low-GI foods were not associated with risk of adiposity-related cancers or prostate and colorectal cancers. In exploratory analyses, low-GI foods, were associated with 49 % lower breast cancer risk (HR 0·51; 95 % CI 0·32, 0·83). In this cohort of Caucasian American adults, associations between carbohydrate nutrition and cancer varied by cancer site. Healthier low-GI carbohydrate foods may prevent adiposity-related cancers among women, but these findings require confirmation in a larger sample.

We aimed to examine the longitudinal association of dairy consumption with the changes in blood pressure (BP) and the risk of incident hypertension (HTN) among adults. This study included 2636 Framingham Heart Study Offspring Cohort members who participated in the 5th through 8th examinations (1991–2008) and were free of HTN at their first examination during the follow-up. Data collected at each examination included dietary intake (by a validated FFQ), BP (following standardised procedures) and anti-hypertensive medication use (by physician-elicited self-report). HTN was defined as systolic BP (SBP)≥140 mmHg, or diastolic BP (DBP)≥90 mmHg or anti-hypertensive medication use. We used repeated-measure and discrete-time hazard regressions to examine the associations of dairy consumption with the annualised BP change (n 2075) and incident HTN (n 2340; cases=1026), respectively. Covariates included demographic, lifestyle, overall diet quality, metabolic factors and medication use. Greater intakes of total dairy foods, total low-fat/fat-free dairy foods, low-fat/skimmed milk and yoghurt were associated with smaller annualised increments in SBP and a lower risk of projected HTN incidence. However, with the exception of total dairy foods and yoghurt, these inverse associations with HTN risk were attenuated as the follow-up time increased. For yoghurt, each additional serving was associated with 6 (95 % CI 1, 10) % reduced risk of incident HTN. Total dairy and total low-fat/fat-free dairy intakes were found to be inversely related to changes in DBP. Dairy consumption, as part of a nutritious and energy-balanced diet pattern, may benefit BP control and prevent or delay the onset of HTN.

Endothelial dysfunction and arterial stiffness are early predictors of CVD. Intervention studies have suggested that diet is related to vascular health, but most prior studies have tested individual foods or nutrients and relied on small samples of younger adults. The purpose of the present study was to examine the relationships between adherence to the 2010 Dietary Guidelines for Americans and vascular health in a large cross-sectional analysis. In 5887 adults in the Framingham Heart Study Offspring and Third Generation cohorts, diet quality was quantified with the 2010 Dietary Guidelines Adherence Index (DGAI-2010). Endothelial function was assessed via brachial artery ultrasound and arterial stiffness via arterial tonometry. In age-, sex- and cohort-adjusted analyses, a higher DGAI-2010 score (greater adherence) was modestly associated with a lower resting flow velocity, hyperaemic response, mean arterial pressure, carotid–femoral pulse wave velocity (PWV), and augmentation index, but not associated with resting arterial diameter or flow-mediated dilation (FMD). In multivariable models adjusting for cardiovascular risk factors, only the association of a higher DGAI-2010 score with a lower baseline flow velocity and augmentation index persisted (β = − 0·002, P= 0·003 and β = − 0·05 ± 0·02, P< 0·001, respectively). Age-stratified multivariate-adjusted analyses suggested that the relationship of higher DGAI-2010 scores with lower mean arterial pressure, PWV and augmentation index was more pronounced among adults younger than 50 years. Better adherence to the 2010 Dietary Guidelines for Americans, particularly in younger adults, is associated with a lower peripheral blood flow velocity and arterial wave reflection, but not FMD. The present results suggest a link between adherence to the Dietary Guidelines and favourable vascular health.

The intake of carbohydrates has been evaluated cross-sectionally, but not longitudinally in an ageing American adult population. The aim of the present study was to examine trends in the intake of dietary carbohydrates and their major food sources among the Framingham Heart Study Offspring (FOS) cohort, which had been uniquely tracked for 17 years in the study. The FOS cohort was recruited in 1971–1975. Follow-up examinations were conducted, on average, every 4 years. Dietary data collection began in 1991 (examination 5) using a validated semi-quantitative FFQ. The study included 2894 adults aged ≥ 25 years with complete dietary data in at least three examinations from 1991 to 2008. Descriptive statistics were generated using SAS version 9.3, and a repeated-measures model was used to examine trends in the intake of carbohydrates and their food sources in the whole sample, and by sex and BMI category. Over 17 years of follow-up, the percentage of energy from total carbohydrates (51·0–46·8 %; P for trend < 0·001) and total sugars (18·2–16·6 %; P for trend < 0·001) decreased. There was a decrease in the percentage of energy from fructose (5·4–4·7 %; P for trend < 0·001) and sucrose (9·8–8·8 %; P for trend < 0·001). Dietary fibre intake increased (18·0–19·2 g/d; P for trend < 0·001). The number of weekly servings of yeast bread, soft drinks/soda, cakes/cookies/quick breads/doughnuts, potatoes, milk, pasta, rice and cooked grains, fruit juice/drinks, potato chips/maize chips/popcorn, and lunch foods (e.g. pizzas and burgers) decreased significantly (P for trend < 0·001), while the intake of ready-to-eat cereals, legumes, fruits, dairy products, candy and ice cream/sherbet/frozen yogurt increased significantly (P for trend<0·04). Similar trends were observed when the analyses were stratified by sex and BMI. The present results suggest favourable trends in dietary carbohydrate consumption, but dietary guidelines for fruits, vegetables and fibre were not met in this cohort.

Few longitudinal studies carried out in US adults have evaluated long-term dietary fat intakes and compared them with the national recommendations during the two-decade period when the prevalence of obesity and insulin resistance increased substantively. In the present study, we examined trends in the intakes of dietary fats and rich dietary sources of fats in the Framingham Heart Study Offspring Cohort over a 17-year period. The cohort was established in 1971–75 with follow-up examinations being conducted approximately every 4 years. Dietary data were collected using a semi-quantitative FFQ beginning in 1991 (exam 5). We included 2732 adults aged ≥ 25 years with complete dietary data in at least three examinations from 1991 to 2008. Descriptive statistics were generated using SAS version 9.3, and a repeated-measures model was used to examine trends in macronutrient and food intakes using R. Over the 17 years of follow-up, the percentage of energy derived from total fat and protein increased (27·3–29·8 % of energy and 16·8–18·0 % of energy, respectively) and that derived from carbohydrate decreased (51·0–46·8 % of energy; P-trend < 0·001). Increases in the percentage of energy derived from all fat subtypes were observed, except for that derived from trans-fats, which decreased over time (P-trend < 0·001). Trends were similar between the sexes, although women exhibited a greater increase in the percentage of energy derived from saturated fat and less reduction in the percentage of energy derived from trans-fats (P interaction < 0·05). Trends in fat intake were similar across the BMI categories. The number of weekly servings of cheese, eggs, ice cream desserts, nuts, butter and sausages/processed meats increased, whereas the intake of milk, margarine, poultry, confectioneries, chips and breads decreased (P-trend < 0·001). In this cohort of predominantly Caucasian older adults, the percentage of energy derived from dietary fats increased over time, but it remained within the national recommendations of less than 35 % of total energy, on average.

This is the third in a series of three papers. In the first paper we describe a comprehensive stochastic model of an individual's lifetime that includes diagnosis with ischaemic heart disease and stroke and also the development of the major risk factors for these conditions. The second paper discusses in some detail models for changes in body mass index (BMI) and also the effects of these changes, in particular the current trend towards increasing prevalence of obesity, on diabetes, cardiovascular diseases and expected future lifetime. This paper is devoted to the following applications of the model described in the first paper:

(a) quantifying the effects of smoking and of changes in smoking habits, and,

(b) quantifying the effects of treatment with statins (drugs designed to lower cholesterol).

This is the second in a series of three papers. In the first paper we describe a comprehensive stochastic model of an individual's lifetime that includes diagnosis with ischaemic heart disease and stroke and also the development of the major risk factors for these conditions. The third paper is devoted to some applications of this model.

In this paper we discuss the effect of BMI on diabetes, heart disease and mortality and we use data from the Framingham Heart Study and the Health Survey for England to develop stochastic models for changes in an individual's BMI. Some of these models incorporate time trends leading to increasing prevalence of obesity. We then use these models to investigate how future expected lifetime and future expected healthy lifetime depend on BMI.

Our conclusions are that if the prevalence of obesity increases, even to an extreme degree, then the prevalence of diabetes and hypertension in the population will increase, possibly to a significant extent, but the prevalence of heart disease and stroke will increase by a much smaller amount and the effect on expected future lifetime will be small.

We construct a stochastic model of an individual's lifetime that includes diagnosis with ischaemic heart disease and stroke and also the development of the major risk factors for these conditions: hypercholesterolaemia, hypertension, diabetes and obesity. Smoking, another major risk factor, is treated deterministically. Mathematically, the model is a continuous time, finite state space Markov process, with the individual's age playing the rôle of time. The model is parameterised using data from the Framingham Heart Study, with parameter values adjusted so that the model is appropriate for UK conditions in the early 21st century. The model has been designed so that it can be used to quantify the effects of:

(i) trends, in particular increasing prevalence of obesity.

(ii) changes in behaviour, in particular smoking patterns, and

(iii) treatments, in particular statins for hypercholesterolaemia.

These applications are covered in two accompanying papers.