Background/aims To evaluate the associations of dietary consumption with the 10-year incidence of diabetic retinopathy (DR) progression in working-aged Australians with diabetes.
Methods We obtained longitudinal data of all diabetic subjects aged 45–65 years from the baseline of the 45 and Up Study and linked this data with Medicare Benefits Schedule (MBS) and Pharmaceutical Benefits Scheme claims until 2016. Retinal photocoagulation (RPC), as determined based on the MBS data, was used as a proxy measure of DR progression. Dietary measurements were assessed via self-reported consumption of meat, dairy products, whole-meal bread, breakfast cereal, vegetables, fruit and fruit juice using a self-administered questionnaire at baseline. Cox regression was used to assess the association between dietary consumption and incident RPC during the follow-up period.
Results A total of 8122 participants were included in the current analysis with a mean age of 57.2±5.2 years. During a mean follow-up of 8.6 years, 314 participants (3.8% of baseline) received RPC. Higher consumption of cheese and whole-meal bread was associated with a lower risk of incident RPC, with the HRs of the highest quartiles versus the lowest being 0.58 (95% CI 0.41 to 0.83; test for trend, p=0.007) and 0.64 (0.46 to 0.89; p=0.04), respectively. Body mass index, insulin treatment and gender were significant modifiers for the association between cheese/whole-meal bread and RPC.
Conclusion Consumption of cheese and whole-meal bread could reduce the risk of DR progression among the working-aged Australian population with diabetes.
- public health
- diabetic retinopathy
Statistics from Altmetric.com
Diabetes is a major public health problem globally, with 451 million people worldwide estimated to have diabetes in 2017. This figure is expected to rise by 50% by the year 2045 to 693 million.1 Diabetic retinopathy (DR), the most common diabetic microvascular complication, is a leading cause of irreversible blindness and imparts a substantial economic burden on the working-aged population.2
Diet plays an important role in prevention and management of diabetes.3 4 Though DR occurs in 35.4% of patients with diabetes,5 little is known about the role of diet in DR onset and progression.6 7 Most previous studies investigating dietary factors and DR focus on particular macronutrients such as carbohydrates,8 fatty acids,9 monounsaturated10 and polyunsaturated fatty acids11 or micronutrients including vitamins.12 Due to the complexity of the biochemical interaction among numerous nutrients in foods, clinical evidence based on food groups may be more practical for the clinical management of patients with diabetes. In recent decades, there is some evidence that higher consumption of particular food groups such as fruit13–15 and fish16 17 may have a potential protective effect on DR onset and progression. However, associations between other food groups and DR, such as dairy products, meat (both are the primary source of protein and fat) and grains (the main source of carbohydrates), have not been thoroughly explored. This information may be useful to inform specific dietary recommendations to delay the onset or progression of DR.
In this study, we aim to conduct a comprehensive assessment of associations between dietary consumption of meat, dairy product, grain, fruit and vegetables, and the risk of DR progression in a large cohort study of working-aged Australians with diabetes.
Materials and methods
Participants in this study were enrolled from the Sax Institute’s 45 and Up Study, which is the largest prospective Australian cohort study conducted in the state of New South Wales (NSW). A total of 266 896 residents aged 45 years and older (18% of invitees), representing more than 10% of the NSW population in this age group, received mailed-based questionnaire survey and written informed consent at baseline 2006–2009. A wide range of socioeconomic status, health conditions and health-related lifestyles were chaptered in the baseline questionnaire, which is available online (http://www.saxinstitute.org.au/our-work/45-up-study/questionnaires/). The study methodology has been described in detail elsewhere.18 In brief, medical information of the participants was tracked through two of Australia’s universal healthcare databases, Medicare Benefits Schedule (MBS) and Pharmaceutical Benefits Scheme (PBS), by the Sax Institute using a unique identifier provided by the Department of Human Services. The MBS database includes all diagnosed tests and treatments, while the PBS database includes medication-related records that participants claimed. Both participant MBS and PBS data captured between January 2004 and December 2016 were included in this study. Ethics approval for the 45 and Up Study was granted by the University of New South Wales Human Research Ethics Committee.
The present study included all working-aged patients with diabetes from the 45 and Up Study at baseline, and exclusion criteria included (1) aged 65 years and over; (2) choose ‘no’ in question no. 24 of the baseline questionnaire: “Has a doctor EVER told you that you have diabetes?”, and without any records of diabetic medication claims based on the PBS database at baseline, which had been proved to be a satisfied complementary way for identifying diabetic patient19; (3) with gestational diabetes, defined as no diabetes medication record after the last childbirth but report a history of diabetes in the questionnaire; (4) a history of vitrectomy prior to baseline examination; (5) a history of retinal photocoagulation (RPC) treatment prior to baseline examination; (6) an invalid age of diabetes onset: reported age of onset older than age at baseline; (7) invalid body mass index (BMI) value, <15 kg/m2 or >50 kg/m220; and (8) missing data for any dietary question. Figure 1 summarises the process of participant selection.
In the current study, incident RPC during the follow-up period based on the MBS database (MBS code: 42809) was taken as a surrogate for DR progression. The date of the first RPC treatment was used in the analysis if the patient had multiple records.
Dietary variables in this study were obtained by self-reported questionnaire using short validated dietary questions at baseline survey, which were frequently used in health monitoring and surveillance to assess the consumption of main daily food.21 Briefly, meat consumption was reported as times per week for red meat, poultry and processed meat, respectively. Consumption of fish/seafood and cheese were also reported as times per week. Milk consumption was reported as the type of milk, including ‘whole milk’, ‘reduced fat milk’, ‘skim milk’, ‘soy milk’, ‘other milk’ or ‘I didn’t drink milk’. The number of slices of whole-meal bread (including brown bread, multigrain or rye bread) and bowls of breakfast cereal consumed per week were also recorded. Fruit and vegetable consumption were both reported as the number of servings each day. A serving of fruit was defined as one medium or two small pieces of fresh fruit, or one cup of diced fruit, and a serving of vegetables was defined as half a cup of cooked vegetables or one cup of salad. Consumption of fruit juice was recorded as the number of glasses per day. The healthy diet scores (HDS) in the current study was assessed and calculated based on a seven-question quiz concerning the number of healthy food groups according to the Australian Dietary Guideline. The quiz is available at https://www.eatforhealth.gov.au/guidelines (Australian Dietary Guidelines—Summary, page 6).22 One point was given for each food group. There are seven food groups including vegetables (at least 5 servings/week), fruit (at least 2 servings/week), dairy product (eat mostly reduced fat milk, skim milk or soy milk), cereals (eat mostly whole-grain cereals), processed meats (once a week or less) and alcohol (no more than 14 servings/week consumption). A higher score indicated a wider variety of recommended food consumption and healthier diet pattern.
The following baseline variables derived from the baseline questionnaire (self-report) or the PBS database were included as covariates in the current study: demographic factors (age, gender, educational level, household income per year), health status (BMI, history of hypertension and cardiovascular disease (CVD), family history of diabetes, insulin treatment, diabetes duration) and lifestyle factors (smoking, alcohol drinking status and amount of physical activity). BMI was calculated as the self-reported weight (kg) divided by self-reported height squared (m2) and further divided into four groups according to the WHO definition: underweight (<18.5 kg/m2), normal (≥18.5–24.9 kg/m2), overweight (≥25–29.9 kg/m2) and obese (≥30.0 kg/m2). Presence of CVD was defined as any history of stroke or heart disease indicated in question no. 24 in the baseline questionnaire. The duration of diabetes was calculated as the difference between the baseline age and the reported age at diabetes diagnosis for participants enrolled based on the questionnaire, while for participants enrolled based on the PBS database, the duration was calculated as age at baseline minus the age at first prescription of diabetic medication. The diabetes duration was divided into four groups for analysis: 0–5, ≥5–10, ≥10–20 and ≥20 years. Insulin treatment was defined as those who had records of insulin medication in the PBS before baseline. The metabolic equivalent (MET) intensity level number of weekly sessions of physical activity was calculated by computational formula W+M+2V, where W represents times of continuously walking at least 10 min, M represents times of moderate activity and V represents times of vigorous activities.23 The amount of physical activity was classified into four categories based on the above MET-adjusted sessions per week: <5, ≥5–9, ≥9–14 and ≥14. History of hypertension, as well as smoking and drinking status, were analysed as dichotomous variables.
All analyses were performed using the SAS software (V.9.4; SAS Institute, Cary, North Carolina, USA). χ2 test for categorical variables was used to examine the difference of baseline characteristics among participants with different quintiles (Q) of the HDS. The associations between various diet categories and the incidence of RPC were analysed using the Cox regression model. The HR and 95% CIs were estimated for the outcome in comparison with a reference category of patients with the first quartile or quintile. The following models were used: (1) model 1 adjusted for age and gender; (2) model 2 further adjusted for household income, educational level, BMI, history of hypertension, CVD, family history of diabetes, insulin treatment, diabetes duration, current smoker, alcohol drinking and physical activity; (3) model 3 further adjusted for all dietary covariates in addition to model 2, including red meat, poultry, processed meat, fish/seafood, fruit, fruit juice and vegetables. We also tested the association between potential dietary factors with incident RPC in model 3 stratified by BMI (<25 kg/m2, ≥25 kg/m2), insulin treatment (yes/no) and gender (male/female). All p values were two-sided and a p value of <0.05 was considered statistically significant.
Figure 1 depicts a flowchart of the selection process for participants. Among 8122 diabetic participants, 314 (3.8%) had received RPC over an average of 8.6 years of follow-up. The mean (±SD) age of participants at baseline was 57.2±5.2 years, and 44.3% were women.
Table 1 describes the baseline participant characteristics by quintiles of HDS based on Australian dietary guidelines. HDS was significantly higher in elderly participants (χ 2 test, p<0.001), women (p<0.001), those with higher household income (p=0.002), higher level of education (p=0.02), insulin treatment (p=0.02) and those with higher levels of physical activity (p<0.001). A history of CVD (p=0.03) and unhealthy behaviours, such as smoking (p<0.001) and alcohol drinking (p<0.001), were inversely associated with HDS.
The association between each food group and incident RPC is presented in table 2. After adjusting for age, gender and lifestyle variables, consumption of red meat (p for trend <0.001), poultry meat (p for trend=0.001), cheese (p for trend=0.004) and whole-meal bread (p for trend=0.04) were inversely associated with incident RPC. Further adjustment for all food variables in model 3 revealed that the highest quartile consumption of cheese and whole-meal bread was associated with a lower risk of RPC during the follow-up, and the HR and 95% CI were 0.58, 0.41 to 0.83 (p for trend=0.007) and 0.64, 0.46 to 0.89 (p for trend=0.04), respectively. We observed a significant trend in lower incident RPC with higher red meat consumption (p for trend=0.002); however, the association was not significant for each subgroup comparison. No significant association was observed between incident RPC and any of the following variables: consumption of processed meat, fish and seafood, any kind of milk, breakfast cereal, fruit or fruit juice, and vegetables.
Figure 2 shows the association between cheese/whole-meal bread and incident RPC stratified by BMI, insulin treatment and gender. Compared with participants who had never undergone insulin treatment, those who had received insulin treatment could benefit significantly from higher consumption of cheese and whole-meal bread in reducing their risk of incident RPC (both with p for interaction <0.001). The favourable effects of cheese consumption on reducing the risk of incident RPC were found in participants with a BMI of ≥25 kg/m2 (Q4 vs Q1, HR 0.62, 95% CI 0.42 to 0.92; p for interaction=0.01). In addition, men could benefit significantly more from cheese consumption (p for interaction=0.011) and women could benefit more from whole-meal bread consumption (p for interaction=0.044). Associations between whole-meal bread/cheese consumption and baseline characteristics/other food groups could be found in the online supplementary file 1.
In addition to medical treatment, nutritional management is also essential in the prevention and management of diabetes and its complications.6 7 In the current study, with a mean follow-up period of more than 8 years, we report that participants with diabetes who consumed higher amounts of cheese and whole-meal bread had a significantly lower risk of DR progression. Meanwhile, consumption of meat, fish and other seafood, milk, breakfast cereal, vegetables, fruit or fruit juice were not related to DR progression. To the best of our knowledge, our study provides the most comprehensive analysis of the association between dietary food groups and DR progression to date. Results of our study could help to guide clinical management and formulate dietary recommendations for patients with DR.
Dairy food is an essential part of the daily diet, but its association with DR progression has been scarcely investigated.6 The results of the current study suggest that the risk of DR progression could be reduced by 40% for subjects who consume cheese ≥4 times per week. A similar association has been reported between cheese and incident diabetes.24 25 In addition, the Mediterranean diet which includes proportionally high consumption of dairy products, including cheese, has consistently been reported to be associated with a lower risk of diabetes, supporting our finding.26 However, we did not find a significant association between milk consumption and incident RPC, which is consistent with a report regarding the association between milk/cheese and diabetes.27 The influence of different dairy products on diabetes has been summarised in a recent meta-analysis, where the authors reported that yoghurt was the strongest protective food within the dairy food group.28 A possible explanation for the discrepancy between the effects of cheese/yoghurt and milk may be related to the acidified and/or coagulation procedure caused by the enzyme rennet during the production of cheese/yoghurt. However, before any significant conclusions can be drawn relating to the mediating pathway, the potential beneficial elements produced during the process of making cheese and yoghurt need to be further investigated. In the present study, the protective effect of cheese was more significant in subjects with higher BMI, which may due to the effect of cheese in preventing body weight elevation,29 though not consistently reported.30 Rideout and colleagues reported that low-fat dairy products may also improve insulin resistance in healthy adults,31 implying an interaction between dairy products and insulin metabolism which may also partly explain the stronger protective effect of cheese in reducing DR progression risk among participants who had received insulin treatment in our analysis. Gender was also found to be a potential modifier for the association between cheese and DR progression, which requires further investigation.
In the present study, we identified an inverse association between whole-meal bread consumption and DR progression. That is, compared with those who consumed ≤3 slices of whole-meal bread/week, subjects who consumed ≥15 slices/week (about 2 slices/day) had a nearly 40% reduced risk of DR progression. We are unaware of any previous studies investigating this association, and our results are supported by numerous previous reports in patients and animals documenting the effect of whole-meal diet on diabetes control. As described by Singhal and Kaushik, a possible explanation for the protective effect of whole-meal bread on diabetes is related to its composition of unrefined cereal grain, such as rye, barley and oats that may delay gastric emptying of starch, lower fasting plasma glucose and postprandial glucose response, lower mean glycaemic response area and improve the insulin response area.32 It is also known that consuming whole-meal bread/grain as the main intake of carbohydrates is essential to a low glycaemic index (GI) diet pattern which has been proven to be beneficial for glycaemic control.33 Stratified analysis showed that participants who were overweight and obese could benefit more from whole-meal bread, which may be attributed to that whole-meal diet improving plasma glucose control and insulin concentration,34 and also decreasing body fat.35 However, in the present analysis, breakfast cereal was not associated with incident RPC. We speculate that this finding may be attributed to the fact that many breakfast cereals contain processed grain with a high GI (eg, cornflakes and rice bubbles),36 and therefore the beneficial effect on glycaemic or weight control is limited. The gender difference in benefit from whole-meal bread may due to the difference of consumption, which requires further study.
Even though read meat was reported as a risk factor for the development of diabetes in previous studies, it was not associated with the risk of DR progression after adjusting for other confounders in this study, which may be due to that over 75% of participants consumed red meat no more than three times a week. Poultry intake may have a potential protective role in the progression of DR. Further researches are needed to better understand the association and underlying mechanisms. To date, only one study has reported a significant relationship between fish and DR, with Millen and colleagues suggesting that only dark fish (or oily fish) may have a protective effect on DR onset,16 but this is not the case for other kinds of fish. In a previous study, a reduced risk of DR was observed in a cohort with a higher intake of polyunsaturated fatty acids, which mainly derived from dark fish.11 Therefore, the failure to observe an association between fish and DR progression in the present study may be attributable to the fact that we could not stratify our analysis according to fish type, rather we examined associations of ‘any’ fish intake.
Vegetable and fruits
A diet pattern rich in fruit is widely recommended in the clinical guidelines for diabetes in many countries,37 even without reliable evidence.38 Unlike two previous reports from Asian populations, we did not observe a significant protective effect of fruit consumption on DR progression.14 15 A possible explanation may be the different methods employed between studies for outcome assessment (Japan: any stages of DR; China: hospitalisation records of DR) and fruit consumption (Japan: quantity in grams per day; China: number of days eating fruit per week). Some evidence has suggested that most of the vegetables had a protective effect on DR prevention owing to a low GI and rich in antioxidants and dietary fibre,6 but the absence of an association may be due to the potato was included as a kind of vegetable in the survey, which contains relatively less protective nutrients mentioned above.39
Strengths and limitations
Strengths of our study include a large sample size with a long period of follow-up, and we included multiple food groups and a wide range of related lifestyle factors in the analysis. Several limitations should also be noted. First, we are unable to exclude patients with diabetes who had undergone RPC prior to 2004 as records before this time were unavailable. While this may have led to selection bias, the prevalence of PDR (estimated by the RPC records) at baseline in the present study of 2.0% is comparable with that of a previous national population-based report in Australian adults.40 Thus, we expect the bias caused by missed cases to be minimal. Second, only the participants treated in private hospitals or clinics could be tracked by MBS data. The bias caused by patients who received RPC in public hospitals might be limited as the average annual incidence of RPC in our study was similar to a previous report.41 Third, as dietary patterns vary among regions globally, the potential benefit of cheese and whole-meal bread may be limited in some regions where people seldom consume these foods. Fourth, we used the RPC as a surrogate for DR progression, which may lead to bias by other diseases (eg, ischaemic retinal vein occlusion (RVO)) which require RPC. However, considering a low prevalence of RVO40 with a small proportion of the ischaemic type,42 the bias would be very limited. Lastly, food frequency was reported instead of specific quantities due to the difficulty of quantitative measurement in such a large cohort. Also, data for energy intake and GI which are important confounding factors in nutritional research were unavailable in this study; BMI and physical activity were considered in the fully adjusted model to minimise this limitation.
In summary, we provide novel evidence that higher cheese and whole-meal bread consumption may reduce the risk of DR progression among Australian adults. Further interventional studies are warranted to confirm these findings and better inform the clinical management of DR.
This research was completed using data collected through the 45 and Up Study (www.saxinstitute.org.au) and supplied by the Department of Human Services (DHS). The 45 and Up Study is managed by the Sax Institute in collaboration with major partner Cancer Council NSW; and partners: the National Heart Foundation of Australia (NSW Division); NSW Ministry of Health; NSW Government Family and Community Services—Ageing, Carers and the Disability Council NSW; and the Australian Red Cross Blood Service. We thank the many thousands of people participating in the 45 and Up Study.
Correction notice This paper has been amended since it was published Online First. The contributorship statement has been updated.
Contributors XY and XH designed the study, drafted and revised the paper, they have made equal contributions. CW designed the study and revised the paper. SK revised the paper. XS analysed the data and revised the manuscript draft. LZ designed the study and coordinated the analysis. MH initiated the project, designed the study and revised the manuscript. LZ and MH have made equal contributions.
Funding This study was funded by National Natural Science Foundation of China (81420108008).
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval The Royal Victorian Eye and Ear Hospital Human Research Ethics Committee (HREC) (HREC no. 17/1330HS).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.