Old age is a major risk factor for disease and poor health and the number of older people is increasing, leading to a growing recognition of the need to develop strategies to reduce the health burden associated with aging( 1 ). Nutrition is one of the most important and modifiable factors affecting health in older age( Reference Mathers 2 ). Using cross-sectional baseline data, the Melbourne Longitudinal Study on Healthy Aging identified nutrition at baseline as an independent predictor of independence in daily living, good self-rated health and psychological wellbeing (i.e. ‘ageing well’) in older community-dwelling individuals( Reference Kendig, Browning and Thomas 3 ). Ethnicity has also been identified as a predictor of successful ageing( Reference Kendig, McDonald and Piggott 4 , Reference Milte and McNaughton 5 ). For example, people who consume diets associated with particular cultures, such as the Mediterranean and Okinawan diets, appear to have improved health outcomes and longevity( Reference Le Couteur, Solon-Biet and Wahl 6 , Reference Sofi, Cesari and Abbate 7 ). On the negative side, older individuals tend to have suboptimal diets( Reference Wakimoto and Block 8 – Reference Meydani 11 ). This particularly applies to older men who are at an even higher risk of nutritional inadequacies than women secondary to limited involvement in the planning and preparation of meals( Reference Hankin 12 ) and nutritional knowledge( Reference Baker and Wardle 13 ). Therefore, it is important that studies examining the relationship between diet and health take age, sex and ethnicity into consideration.
There are many approaches to investigating the relationship between dietary intake and health. Traditionally, ‘single nutrient’ or ‘one-variable-at-a-time’( Reference Simpson, Le Couteur and Raubenheimer 14 ) approaches have been used to explore associations between individual nutrients and health outcomes and this has been very effective for identifying the effects of nutritional deficiencies( Reference Hu 15 , Reference Raubenheimer and Simpson 16 ). However, with increasing rates of diet-related obesity and cardio-metabolic diseases( Reference Raubenheimer and Simpson 16 ), and a growing recognition of the complexity of diets and interactions between nutrients, there has been a shift towards exploring the relationship between health and broader classifications of dietary composition. Dietary patterns analysis has emerged as one such method to investigate the association between dietary intake and risk of chronic diseases( Reference Hu 15 ). This type of method focuses on foods and their intake rather than specific nutrients.
The ‘dietary index’ is a type of dietary pattern analysis in which individuals’ diets are scored according to how well they comply with established dietary guidelines( 17 ). The revised Australian Dietary Guideline Index (DGI-2013)( Reference Thorpe, Milte and Crawford 18 ) is a food-based dietary index developed to investigate the compliance of adults to the Australian Dietary Guidelines (ADG)( 17 ) (Table 1). The ADG are based on evidence related to the prevention of diet-related conditions and chronic diseases and the Australian National Health and Medical Research Council nutrient reference values( 17 ).
%E, percentage contribution to total energy.
* Criteria for maximum score has been derived from Australian Dietary Guidelines (1) according to age (70+) and sex (male) unless otherwise noted and total possible score=130.
† Food variety scores were calculated based on the number of different food items consumed in a day; food item was only considered if it belonged to a core food group (grains, fruit, vegetable, protein foods and dairy products); if participant was to consume a different food item to meet their requirements of each food group, he would consume a minimum of nineteen different food items (rounded as one cannot consume half of a new food item.
‡ Fluid intake included water and water present in milk, fruit juice, tea and coffee.
§ Amount of SFA as a percentage of total energy.
|| Since added sugar intake is not recommended there are no cut-off values for the number of recommended servings, instead half of the maximum discretionary food cut-off were used consistent with the original DGI.
¶ Na intake derived from salt added before or after cooking, packaged food items and salt naturally present in food.
Using the dietary index approach we investigated the association between food intake and health measures and indices in older men using data from the Concord Health and Ageing in Men Project (CHAMP). This cohort study was established to investigate geriatric syndromes of older men and the relationship between nutrition and health( Reference Cumming, Handelsman and Seibel 19 ). One-quarter of the CHAMP participants have Italian or Greek (Mediterranean) backgrounds, which has provided an opportunity to investigate whether the relationship between the DGI-2013 and health is influenced by ethnicity. Therefore, the aims of this study were to evaluate diet quality of older men using the DGI-2013, to discover sociodemographic and lifestyle factors associated with it, to investigate the associations between diet quality and some health measures and indices common in older age while accounting for the effect of ethnicity on these relationships.
Methods
Participants
The original selection of CHAMP subjects has been described in detail elsewhere( Reference Cumming, Handelsman and Seibel 19 ). In brief, 1705 men aged 70 years and over living in the suburbs of Burwood, Canada Bay and Strathfield in Sydney, Australia who were on the electoral roll enrolled in CHAMP at baseline (2005–2007). Participants have been followed up since 2005( Reference Cumming, Handelsman and Seibel 19 ), and in 2012 during the 5-year follow-up (third follow-up wave), the nutritional data were collected from 794 participants who completed a diet history questionnaire( Reference Waern, Cumming and Blyth 20 ).
All participants gave written informed consent. The study was approved by the Sydney South West Area Health Service Human Research Ethics Committee, Concord Repatriation General Hospital, Sydney, Australia.
Dietary intake
The dietary assessment method used in CHAMP has been described elsewhere( Reference Waern, Cumming and Blyth 20 , Reference Waern, Cumming and Travison 21 ). In brief, typical dietary intake was assessed using a diet histories questionnaire (DHQ) and covered usual intake over the past 3 months( Reference Burke 22 ) A research dietitian conducted and recorded all diet histories in the participant’s residence with the process averaging 45 min for completion. The validity of DHQ was established by comparison to a 4-d weighed food record collected in a subgroup of fifty-six CHAMP men and results published previously( Reference Waern, Cumming and Travison 21 ).
Data handling and database conversion
Participants’ daily dietary intakes were initially analysed using FoodWorks 7 Professional for Windows (2012; Xyris Software (Australia) Pty Ltd), which uses the Australian food, supplement and nutrient database 2007 (AUSNUT 2007). The national database has recently been updated (AUSNUT2001-13) therefore we have used the matching file( Reference Neale, Probst and Tapsell 23 ) to update all foods and recipes information from reported intake in the CHAMP study; this database also converts participants’ number of serves of each food group. Branded products were matched to AUSNUT 2007 food name using AUSNUT 2007 Brand Match File( 24 ), then AUSNUT 2007 food name was matched to AUSNUT 2011-13 using matching file( Reference Neale, Probst and Tapsell 23 ). Foods that were not completely matched via AUSNUT 2007 to 2011-13 matching file were manually matched by an experienced research dietitian (R. V. R.).
Revised Dietary Guideline Index
The original Australian DGI( Reference McNaughton, Ball and Crawford 25 ) is a food-based dietary index developed to investigate the compliance of adults to the Dietary Guidelines for Australian Adults( 26 ). Thorpe et al. ( Reference Thorpe, Milte and Crawford 18 ) revised the original DGI after the release of the new ADG containing changes in terminology and recommendations according to age and sex as well as a new component related to unsaturated fats.
Details of DGI-2013 has been provided elsewhere( Reference Thorpe, Milte and Crawford 18 ), in brief, the DGI-2013 is comprised of thirteen components each scored out of 10 (overall possible maximum score=130), with 0 indicating low compliance to ADG and 10 better compliance, and therefore, higher diet quality. DGI-2013 is divided into categories of adequate intake (intake encouraged) and moderate intake (restrict intake recommended). Cut-offs for maximum and minimum scores are presented in Table 1 which was adapted from Thorpe et al. ( Reference Thorpe, Milte and Crawford 18 ) publication.
The DGI (original and revised) were developed using data obtained through a 111-item FFQ. However, in the present study we have used DHQ to assess food intake that means there are no limitations on the number and quantities of food and beverages reported as is the case with FFQ. Therefore because of the nature of the data obtained – more quantitative than qualitative information – the following adaptations of DGI criteria were made:
-
(1) Added salt intake was not assessed in this study; therefore overall dietary Na intake was used to measure compliance with ADG. Individuals who consumed between 460 and 2300 mg (lower cut point of adequate intake to upper level of intake)( 27 ) of Na a day received the maximum score.
-
(2) For the guideline about limiting intake of foods high in SFA, the percentage of total energy derived from SFA was used as information on fat trimming and type of milk was not systematically collected in the present study.
-
(3) Variety of food intake was measured based on the number of different foods participants consumed a day. For this calculation, only core food groups were taken into consideration because they are the basis of a balanced diet containing essential macro and micronutrients( 28 ). A participant received the maximum score for food variety if they consumed at least nineteen different foods from core food groups.
-
(4) Total fluid intake was calculated by summing participant’s intake of water (including water present in coffee and tea), milk and fruit juice( 17 ).
-
(5) Solid fat equivalents was calculated by summing fats naturally occurring in meat, poultry, eggs, dairy products, fully or partially hydrogenated oils, shortening, palm oil and coconut oil.
-
(6) Number of serves of discretionary foods was determined by summing the number of serves of added sugar (1 teaspoon (4·2 g)=1 serve), solid fat equivalents (1 teaspoon (4·8 g)=1 serve) and alcoholic drinks (1 standard drink=1 serve).
Sociodemographic and lifestyle factors
Data on sociodemographic and economic factors, smoking status and physical activity were obtained through a self-completed questionnaire. Country of birth (COB) was grouped into three categories: (1) Australia and New Zealand; (2) Italy and Greece; and (3) other (total of thirty-eight countries). Income source was used as a proxy for personal income, assuming that age pensioners had the lowest income (age pensioners are provided with a modest pension if they are unable to support themselves) compared with those with other incomes, that is, repatriation pension, veteran’s pension, superannuation or other private income, own business/farm/partnership, wage or salary, other or any income source combination( Reference Kendig, McDonald and Piggott 4 ). Information on who is responsible for grocery shopping and cooking was obtained during the diet history interview and the data were dichotomised as self or other/assisted. Self-rated health data were obtained and dichotomised into excellent/good v. fair/poor/very poor.
Health measures and indices of interest
Number of co-morbidities and medications
Data on medical conditions were obtained from a self-reported questionnaire in which participants reported to be professionally diagnosed with any of the following diseases: diabetes, thyroid problems, osteoporosis, Paget’s disease of bone, stroke, Parkinsons disease, kidney stones, dementia, depression, epilepsy, hypertension, myocardial infarction, angina, heart failure, peripheral vascular disease, chronic obstructive pulmonary disease, liver disease, chronic kidney disease, arthritis and cancer (excluding non-melanotic skin cancer and benign tumours such as bowel polyps and meningioma). Multi-morbidity was defined as having two or more of these conditions( Reference Diederichs, Berger and Bartels 29 ). Participants were asked to bring prescription and non-prescription medications used daily or almost daily to their 5-year follow-up clinic appointment. Polypharmacy was defined as the use of five or more regular prescription medicines( Reference Gnjidic, Hilmer and Blyth 30 ).
Insulin sensitivity, lipidaemia and blood pressure
Fasting blood was collected from 663 (84 %) participants to measure circulating levels of glucose, insulin, TAG, LDL-cholesterol and HDL-cholesterol. Each of these measures was performed at the Diagnostic Pathology Unit of Concord RG Hospital, which is a National Australian Testing Authority accredited pathology service, using a MODULAR Analytics system (Roche Diagnostics). Levels of cholesterol and HDL-cholesterol were measured on a Roche Cobas 8000 analyser (Roche Diagnostics International Ltd) using a standard automated enzymatic methodology. Fasting blood samples for glucose measurement were put into fluoride-oxalate (anticoagulant) tubes. Plasma glucose was measured using the Hexokinase method. homoeostasis model assessment – insulin resistance (HOMA-IR) was calculated using HOMA calculator version 2.2.3 (©Diabetes Trials Unit, University of Oxford). The remaining 128 (16 %) participants were not fasted at the time of blood collection and so the results for these participants have not been included in the current analysis. Blood pressure was measured by trained staff according to a standardised protocol using a sphygmomanometer as previously described( Reference Hirani, Naganathan and Blyth 31 ). Participants with systolic blood pressure≥140 mmHg and diastolic≥90 mmHg were categorised as hypertensive, those with systolic blood pressure 120–139 mmHg and diastolic 80–89 mmHg were categorised as normal-high blood pressure and those with systolic blood pressure<120 mmHg and diastolic<80 mmHg were categorised as normal blood pressure( 32 ).
Anthropometry
Height and weight were measured according to a standardised protocol( Reference Bleicher, Cumming and Naganathan 33 ) and BMI was calculated as kg/m2. BMI was categorised as underweight (<22 kg/m2), normal (22–30 kg/m2) and overweight/obese (>30 kg/m2) in accordance with recent studies in older individuals (65 years and over) showing an increased risk of mortality in the lowest and highest cut-offs( Reference Bannerman, Miller and Daniels 34 – Reference Winter, MacInnis and Wattanapenpaiboon 39 ). Waist and hip circumferences were measured following a standardized protocol as recommended by the World Health Organization( 40 ). Values above 0·9 (for men) indicate an increased health risk because of the abdominal obesity( 40 ). Activity level was determined through the Physical Activity Scale for the Elderly (PASE). The PASE questionnaire includes twelve types of occupational, household and leisure related activities from previous 7-d period; scoring is calculated from weights (e.g. intensity and duration) and frequency values for each type of activity( Reference Washburn, Smith and Jette 41 ).
Frailty status
Frailty scores were determined using the five frailty components used in the Cardiovascular Health Study (CHS)( Reference Fried, Tangen and Walston 42 ). Weakness and slowness components were determined using the same criteria and the same cut-off points as in the CHS( Reference Fried, Tangen and Walston 42 ). Weight loss, exhaustion, and low activity criteria were adapted in the CHAMP study, as the exact measurements used in the CHS were not available. Weight loss was defined as current weight lower by 15 % or more than self-reported heaviest weight (or than weight at 25 years old, if missing data on heaviest weight); participants were asked ‘How much of the time during the past 4 weeks did you have a lot of energy?’( Reference Ware, Kosinski and Keller 43 ), and classified as exhausted if their response was ‘a little’ or ‘none of the time’; low activity was defined as being in the lowest quintile on the PASE (cutoff score<73)( Reference Washburn, Smith and Jette 41 ). Participants were classified as follows: frail (score≥3), pre-frail (score 1–2) and robust (score=0)( Reference Fried, Tangen and Walston 42 ).
Statistical analysis
Data on number of serves of each food group, individual DGI component scores and total DGI scores were checked for normality using graphical and statistical methods (Shapiro–Wilk test) and were found to have skewed distribution, hence, medians and ranges (minimum to maximum) were calculated. Percentages of individuals meeting each component of DGI were also calculated.
To discover factors associated with diet quality (as per DGI-2013), we used a multi-model inference procedure based on information theory( Reference Burnham and Anderson 44 ). We first implemented a global linear model (LM) using the ‘glm’ function in the base package within the statistical programming environment R version 3.3.0 for windows( 45 ), which contained DGI score as the response and all potential predictors of diet quality (age, marital status, level of education, income source, cooking responsibility, grocery shopping responsibility, smoking status, physical activity level, BMI and its quadratic effect) fitted additively. BMI was considered as both a predictor of DGI and an outcome of poor compliance to DGI. COB is a well-known factor associated with dietary patterns( Reference Gilbert and Khokhar 46 , Reference Darmon and Drewnowski 47 ) and was explored in detail separately but not included in the model for the following reasons: (1) ‘other’ category contained forty-eight countries including England/UK, China, Croatia, Hungary and Malta, for example, making this a very heterogeneous group with not many similarity in terms of dietary patterns, and (2) excluding participants from ‘other’ COB was an option we avoided as that would significantly reduce our sample and the power of our analyses. The model was standardised to the Z scale using the ‘standardise’ function in the arm package( Reference Gelman and Su 48 ). A set of candidate models was created using the ‘dredge’ function in the MuMIn package( Reference Bartoń 49 ). Models were then ranked based on Akaike information criterion with correction for small sample size (AICc). Rather than restrict our inference to that based on a single ‘best-fitting’ model, which may be subject to model-selection uncertainty and model-selection bias, we used multi-model inference( Reference Burnham and Anderson 50 ). From the set of candidate models, a top model set comprising those models with an AICc within two of the top model (that with the lowest AICc) were obtained. Model-averaged coefficients were then obtained using the ‘model.avg’ function in MuMIn. For each coefficient we also report relative importance, adjusted SE, 95 % CI (1·96×se)( Reference Nakagawa and Cuthill 51 ), and the coefficient estimates with shrinkage. R 2 for global model was calculated using the equation 10 in Nakagawa & Schielzeth( Reference Nakagawa and Schielzeth 52 ).
Chi-square and Mann–Whitney U test were used to compare DGI-2013 scores, food intake and health measures of Italian and Greek-born men v. Australian and New Zealand-born participants.
We examined whether DGI-2013 scores predicted health measures and indices by independently fitting each health measure as the response in a model with DGI score as the predictor. We explored models that were fitted both with and without socioeconomic factors that were associated with DGI-2013 (based on model averaging) and that differed between ethnic groups (Australia/New Zealand v. Italy/Greece). We used LM for continuous health measures and indices (HOMA-IR, LDL-cholesterol, HDL-cholesterol, TAG, waist:hip ratio and BMI) and quasi-poisson (log-link) generalised linear models (GLM) for all health measures and indices quantified as integers (number of co-morbidities and medications), implemented with the ‘glm’ function in the base package. For measures and indices expressed in categories (hypertension and frailty status), we used multinomial (logit-link) GLM implemented with the ‘multinom’ function in the nnet package( Reference Venables and Ripley 53 ). For all multinomial measures and indices, the healthy status (normal blood pressure and robust) was fitted as the multinomial denominator. These analyses were performed in the whole sample. However, we also explored a subset of the data including only Australian/New Zealander and Italian/Greek-born participants. For this subset of the data we fitted models with interactions (i.e. effect-modifier) between COB and DGI score to explore whether DGI differentially predicted health as a function of ethnic background. Evidence against the null hypotheses was considered statistically significant if the resulting P values were <0·05.
Results
Participants’ characteristics, food intake and Dietary Guideline Index scores
Participants’ characteristics are presented in Table 2. The median age was 80 years (74–98) and median BMI was 27·5 (15·2–43 kg/m2). The majority of participants were Australian or New Zealand-born (n 427), married and relied exclusively on the age pension as income source. Italian and Greek-born participants (n 188) were significantly less educated, more likely to live exclusively on the age pension and be married. Australian and New Zealand-born men were less likely to be smokers or former smokers, had a higher HDL-cholesterol and were more likely to rate their health as excellent/good. Overall, health measures and indices were similar in both groups (Table 2).
PASE, Physical Activity Scale for the Elderly; HOMA-IR, homoeostasis model assessment – insulin resistance.
* Overall sample included all countries of birth (Australia, New Zealand, Italy, Greece and other).
† Income source was used as a proxy of income assuming that ‘others’ have higher income than ‘pensioners only’.
‡ Other sources of income includes repatriation pension, veteran’s pension, superannuation or other private income, own business/farm/partnership, wage or salary, other or any income source combination.
§ There was a total of forty-eight countries of birth such as England/UK, China, Croatia, Hungary and Malta.
|| Hypertension was defined as systolic blood pressure≥140 mmHg and diastolic blood pressure≥90 mmHg( 32 ).
¶ Multi-morbidity was defined as having two or more of these conditions (seventeen).
** Polypharmacy was defined as the use of five or more regular prescription medicines.
Participants’ median food intake, DGI-2013 scores and proportion meeting guidelines are presented in Table 3. The average DGI-2013 score was 93·7 (54·4–121·2) with the majority of participants meeting the minimum guidelines for grains/cereals, meat and alternatives, water, alcohol and salt intake. Median dairy products and vegetable intakes were considerably below minimum guidelines. Only 1 % of the population consumed more than the recommended 2·6 litre of fluid per day. The majority of the population failed to consume enough fluid and over-consumed unsaturated fat and discretionary foods.
* P value for difference of intake between subgroups (Australia and New Zealand-born v. Italy and Greece-born participants) derived from Wilcoxon’s signed-rank test. Median may not differ between groups, however, values will be ranked differently (as per Wilcoxon’s signed-rank test method) hence significant P values for difference.
† Food variety scores were calculated based on the number of different food items consumed in a day; food item was only considered if it belonged to a core food group (grains, fruit, vegetable, protein foods and dairy products); if participant was to consume a different food item to meet their requirements of each food group, he would consume a minimum of nineteen different food items (rounded as one cannot consume half of a new food item.
‡ Includes milk alternatives, legumes, soya products, nuts and seeds.
§ Fluid intake included water and water present in milk, fruit juice, tea and coffee.
|| Number of serves of discretionary foods was determined by summing the number of serves of added sugar, solid fat equivalents and alcoholic drinks.
¶ Amount of SFA as a percentage of total energy.
** Fats naturally occurring in nuts, seeds, avocado, seafood and un-hydrogenated vegetable oils.
†† Salt intake derived from salt added before or after cooking, packaged food items and salt naturally present in food.
‡‡ Total possible score=130.
Factors associated with Dietary Guideline Index compliance
The top model set for predictors of DGI score, and associated AICc values is given in the online Supplementary Table S1. Model-averaged coefficients from these models are presented in the online Supplementary Table S2. Higher level of education, income and physical activity were associated with higher DGI scores, whereas being a smoker was associated with lower DGI scores.
Dietary Guideline Index-2013 and health measures and indices
Table 4 shows the results of analyses investigating the association between health measures and indices and DGI-2013 before and after adjustment for factors education and income. After adjustments, high DGI scores were associated with lower HDL-cholesterol, lower waist:hip ratios and lower probability of being frail. Proxies of good health such as fewer co-morbidities and medications were not associated with better compliance to the ADG.
Model 1, unadjusted; model 2, adjusted for education and income; HOMA-IR, homoeostasis model assessment – insulin resistance (high HOMA-IR values indicate low insulin sensitivity (insulin resistance)).
The influence of ethnicity on Dietary Guideline Index and health measures and indices
The DGI-2013 scores of men born in Italy and Greece was significantly lower than those born in Australia and New Zealand (Table 3). Total energy intakes were similar between the two groups of men; however, Italian and Greek-born men consumed less total energy from protein and carbohydrate but more from fat. Italian and Greek-born men had higher intake of total and MUFA fat intake, red and orange vegetables, legumes (both as vegetables and as meat alternative), refined cereals, alcohol, oil equivalents, legume protein and fresh vegetables (tomatoes, dark green, red and orange vegetables) and lower intakes of SFA, starchy vegetables, wholegrain cereals, nuts, dairy products and discretionary foods including added sugar, SFA and salt (Table 3).
In the subset analysis (Australian/New Zealander and Italian/Greek-born participants only) investigating effect of COB on the relationship between DGI and health measures and indices, we found significant interaction between COB and DGI scores for probability of being classified as frail (P=0·01), HOMA-IR (P=0·005), number of co-morbidities (P=0·039) and medications (P=0·005) (Table 5). For those born in Australia and New Zealand, higher DGI-2013 scores were associated with lower HOMA-IR, number of co-morbidities, medications and lower probability of being classified as frail whereas for Italian and Greek-born men, increasing DGI scores had the opposite effect (Table 5, Fig. 1).
COB, country of birth; HOMA-IR, homoeostasis model assessment – insulin resistance (high HOMA-IR values indicate low insulin sensitivity (insulin resistance)).
Discussion
Our study is the first to assess diet quality in men aged 74 years and older. Overall compliance to the ADG was suboptimal with at least one-third of participants not meeting the recommendation for fruit, grains and meat. Even more concerning, more than half of individuals were not meeting recommendations (under-consuming) for vegetables, dairy products and alternatives and fluid but over-consuming added sugar, unsaturated fat and SFA fat, and discretionary foods. The main factors associated with DGI-2013 were education, income, smoking status and physical activity level.
Several studies have found an association between lower income and poorer dietary quality( Reference McNaughton, Ball and Crawford 25 , Reference Du, Mroz and Zhai 54 – Reference Aggarwal, Monsivais and Cook 57 ). Nutritious and healthy foods tend to be more expensive( Reference Andrieu, Darmon and Drewnowski 58 , Reference Agarwal, Chadha and Tandon 59 ) whereas energy dense and nutrition poor diets tend to be cheaper( Reference Drewnowski 60 ), therefore cost may be a large barrier for older adults – particularly those living on the Age Pension – when choosing and purchasing foods. Furthermore, processed foods are cheaper and more palatable making them more attractive to older individuals who tend to have poor gustatory function( Reference Sergi, Bano and Pizzato 61 ) and, among older men, limited cooking facilities and/or ability( Reference Dean, Raats and Grunert 62 ). Education and nutritional knowledge tend to correlate( Reference Maindal, Toft and Lauritzen 63 ), therefore it is not surprising that less educated individuals are more likely to have poor compliance to the ADG. Similarly, poor health behaviours’ such as low physical activity and smoking are often associated with poor nutritional habits( Reference Thorpe, Milte and Crawford 18 , Reference Morabia and Wynder 64 – Reference Dallongeville, Marécaux and Fruchart 66 ). Morabia & Wynder( Reference Morabia and Wynder 64 ) investigated the association between dietary intake and smoking status of 7860 subjects aged 25–74 years and found that smokers consumed less fruit and vegetables, more alcohol and coffee than never smokers; male smokers consumed more meat and less cereals than those who were never smokers.
Compared with the general population of similar age and sex, that is, the latest nationally representative Australian Health Survey (AHS)( 67 ), CHAMP participants’ intakes of vegetable, meat and alternatives, dairy products and alternatives and fluid intakes were higher, however this difference may be because of the difference in dietary assessment method used in the two studies (AHS used 24-h recall v. DHQ in current study). Regarding higher vegetable intake in CHAMP participants, one explanation for the difference between AHS and the present study results may be related to the large proportion of CHAMP participants with a Mediterranean background – known to consume more vegetables( Reference Sofi, Cesari and Abbate 7 ). Income may also play a role in explaining some of the differences in intake between the two studies; foods high in protein (e.g. dairy products and meat) tend to be more expensive than carbohydrate rich foods( Reference Brooks, Simpson and Raubenheimer 68 ), and given that at least 40 % of CHAMP participants are likely to be on higher income (as per income source), one can assume that they have access to high protein foods.
Two of the ADG are known to have a direct effect on HDL-cholesterol levels: limiting consumption of SFA and consuming a small allowance of unsaturated fat( 17 ). In the present study, we found that high DGI scores (i.e. better compliance with ADG) were associated with lower HDL-cholesterol levels. One potential reason for this may be the limitation in unsaturated fat.
Although one would expect that better compliance with dietary guidelines would result in healthier HDL-cholesterol levels (i.e. higher levels), other non-dietary factors such as obesity and smoking status, presence of the metabolic syndrome, hypertriacylglycerolaemia and even socioeconomic status may also influence HDL-cholesterol levels( Reference Vergeer, Holleboom and Kastelein 69 ).
In this study, we found that high DGI scores were not associated with some indicators of better overall health such as fewer co-morbidities or number of medications. There are a number of potential explanations for this: first, ageing in itself is an important factor in the development of some of the health issues common in older age; second, some factors such as genetics, for example, have an impact on the relationship between nutrition and health measures and indices but cannot be accounted for. Similarly, although we have adjusted for a number of factors known to have an impact in those associations, there may still be some factors not yet identified that may confound these associations.
A recent systematic review involving both longitudinal and cross-sectional data, showed that better diet – as measured by a variety of dietary assessments and dietary indices – was associated with successful ageing as defined by better quality of life as well as good mental and physical health( Reference Milte and McNaughton 5 ). DGI is based on ADG, which provides evidence based guidelines on food types and quantities that are associated with a reduction in morbidity and mortality( 17 ). It follows that compliance with the dietary guidelines therefore, should result in better health measures and indices( Reference Hung, Joshipura and Jiang 70 – Reference Seal and Brownlee 72 ). In our study we found that overall, although not always statistically significant, good compliance to the DGI was associated with better health measures and indices.
Evidence suggests that dietary preferences established in younger ages can influence food choices in later life( 73 ); therefore, it is likely that older individuals will follow the same dietary patterns as those established in their earlier age and COB. This may explain the observed higher intake of alcohol and unsaturated fats as well as significantly higher intake of legumes and non-starchy vegetables (Mediterranean dietary pattern) amongst Italian/Greek-born men. Surprisingly, Italian/Greek-born participants with better compliance to the ADG had a tendency to poorer health measures and indices, in particular number of co-morbidities. Mediterranean dietary patterns have been shown to have beneficial effects on several age-related health outcomes( Reference Sofi, Cesari and Abbate 7 ) and this may explain, at least in part, why Italian and Greek-born men have similar health to Australian and New Zealand-born men despite lower DGI scores. This also suggests that a dietary index developed for the general population may not be suitable for or accepted by those from diverse ethnic backgrounds. The poorer health in Italian and Greek migrants with greater compliance to the ADG could reflect the loss of the protective Mediterranean dietary pattern in this group but could equally be the result of reverse causality where men have changed their diets to conform more closely to the ADG in response to ill health.
Strengths and limitations
One of the main strengths of this study was the use of a validated dietitian-administered diet history questionnaire to assess dietary intake of its participants. DHQ is a retrospective method is particularly indicated for older people because their diets tend to be consistent over long periods of time, it does not rely on short-term memory and it uses a much more interactive approach than other methods( Reference Hankin 12 , Reference Visser, De Groot and Deurenberg 74 – Reference Van Staveren, Burema and Livingstone 76 ). Furthermore, diet histories have low respondent burden, which may improve response rates among older people and they require no literacy or numeracy skills from participants( Reference Margetts and Nelson 77 – Reference Willett 79 ), making them suitable for participants from culturally and linguistically diverse backgrounds. Another advantage of using data derived from diet history questionnaires is that no pre-established serves and frequency are used, that gives a good notion of variety of food consumed and also permits us to proportionally score guidelines related to moderate and limited intake. The present study also had some limitations. First, we used data from a cross-sectional observational study, which precludes the investigation of causal mechanisms. Furthermore, when assessing these findings, it is important to take into account important ‘survival effects’ – that is, individuals with poor nutrition tend to be less likely to live to the advanced ages examined in this study( Reference Kendig, Browning and Thomas 3 ). Second, as in most studies on nutritional epidemiology, diet was self-reported and measurement bias may be present; however, measurement bias is likely to have been non-differential with regards to measures and indices and so will have led to underestimation of associations, rather than causing spurious associations. Similarly, self-reported food intake may have been influenced by participants’ desire to gain approval from interviewer/researcher( Reference Hebert, Clemow and Pbert 80 ) based on what is believed to be ‘healthy’. Finally, adaptation of DGI-2013 was necessary regarding salt and animal fat trimming; this resulted in a DGI composed of a mix of food and nutrient-based DGI score which may not be as practical in a clinical setting as a food-based dietary index.
Conclusion
In conclusion, this cross-sectional study has demonstrated that the diet of Australian men aged≥74 years is suboptimal according to ADG. However, for participants with a Mediterranean background, having poor compliance to ADG was not associated with poorer health. These findings highlight the need for development of dietary guidelines that are more acknowledge and encourage dietary patterns from culturally diverse groups. Further investigation is required to confirm these findings, particularly in longitudinal studies.
Acknowledgements
The authors thank all the CHAMP’s staff and participants for their contributions and Sydney Medical School Foundation for their support.
The CHAMP study is funded by the National Health and Medical Research Council (301916) and the Ageing And Alzheimers Institute.
The CHAMP study was designed by R. G. C., V. N., D. J. H., M. J. S., D. G. L. C., L. M. W. and F. M. B.. R. V. R., V. H. and A. K. G. developed the concepts for this paper. R. V. R. and V. H. designed the protocol for diet history. R. V. R. collected and trained the staff for nutritional data collection, coded foods, conducted all the data analyses and wrote the first draft of the manuscript. A. M. S. oversaw statistical analyses. V. H., A. M. S., A. K. G., R. G. C., F. M. B., V. N., L. M. W., D. J. H., M. J. S., S. J. S., F. S., M. A.-F. and D. G. L. C. collaborated in writing. All authors reviewed and approved the final version of the manuscript. All authors had primary responsibility for final content.
None of the authors has any conflicts of interest to declare.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114517001738