A Prospective Study of the Association Between Quantity and Variety of Fruit and Vegetable Intake and Incident Type 2 Diabetes
OBJECTIVE The association between quantity of fruit and vegetable (F&V) intake and risk of type 2 diabetes (T2D) is not clear, and the relationship with variety of intake is unknown. The current study examined the association of both quantity and variety of F&V intake and risk of T2D.
RESEARCH DESIGN AND METHODS We examined the 11-year incidence of T2D in relation to quantity and variety of fruit, vegetables, and combined F&V intake in a case-cohort study of 3,704 participants (n = 653 diabetes cases) nested within the European Prospective Investigation into Cancer and Nutrition-Norfolk study, who completed 7-day prospective food diaries. Variety of intake was derived from the total number of different items consumed in a 1-week period. Multivariable, Prentice-weighted Cox regression was used to estimate hazard ratios (HRs) and 95% CIs.
RESULTS A greater quantity of combined F&V intake was associated with 21% lower hazard of T2D (HR 0.79 [95% CI 0.62–1.00]) comparing extreme tertiles, in adjusted analyses including variety. Separately, quantity of vegetable intake (0.76 [0.60–0.97]), but not fruit, was inversely associated with T2D in adjusted analysis. Greater variety in fruit (0.70 [0.53–0.91]), vegetable (0.77 [0.61–0.98]), and combined F&V (0.61 [0.48–0.78]) intake was associated with a lower hazard of T2D, independent of known confounders and quantity of intake comparing extreme tertiles.
CONCLUSIONS These findings suggest that a diet characterized by a greater quantity of vegetables and a greater variety of both F&V intake is associated with a reduced risk of T2D.
Type 2 diabetes (T2D) is one of the most common noncommunicable diseases worldwide and is a leading cause of premature mortality (1), morbidity (2), and health care expenditure (3). Lifestyle intervention trials that include dietary changes have been shown to be effective in preventing the development of T2D (4). However, it is still largely unknown which aspects of the diet confer this beneficial effect.
A World Health Organization expert consultation recommended a minimum intake of 400 g or five portions (based on an average portion weighing 80 g) of combined fruits and vegetables (F&V) per day for the prevention of several major noncommunicable diseases, including T2D (5). In addition, the five-a-day program in the U.K. and similar programs in other countries (e.g., the U.S.) (6) recommend consuming a variety of different F&V, thereby ensuring adequate intake of micronutrients, dietary fiber, and a multitude of other important bioactive compounds (7). Yet, the specific role of variety in F&V intake and T2D has not been examined. Additionally, as studies have generally used a food frequency questionnaire (FFQ) to assess F&V intake, which is suitable for ranking individuals according to their relative but not absolute intake (8), there is also an absence of research on the importance of meeting the five-a-day quantity recommendation for F&V intake. The use of a prospective food diary offers a more precise measure of dietary intake and can overcome some of the limitations of the FFQ (9), but so far no studies have used this dietary assessment method in relation to T2D risk.
In order to develop effective dietary public health strategies for T2D prevention, it is essential to clarify the contribution of both quantity and variety of F&V intake to T2D risk. The aim of this study was therefore to evaluate the association between the quantity and variety of fruit, vegetables, and combined F&V consumption, as assessed using a prospective 7-day food diary, and incident T2D.
RESEARCH DESIGN AND METHODS
The Norfolk component of the European Prospective Investigation into Cancer and Nutrition (EPIC-Norfolk) study recruited 25,639 men and women aged 40–79 years at baseline in 1993–1997. The EPIC-Norfolk study was initiated to investigate the relationship between diet and cancer but has since broadened its scope to include a range of chronic diseases, including T2D. The recruitment procedures, collection of questionnaire data, and anthropometric and dietary measures have been described in detail elsewhere (10,11). In brief, participants residing in Norfolk, England, were recruited from age-sex registers of general practices and attended a baseline health check. Follow-up of participants constituted a postal questionnaire at 18 months, a second health check in 1998–2000, and a further postal questionnaire in 2002–2004.
From the 25,639 participants in EPIC-Norfolk at baseline, we ascertained incident cases of T2D (n = 892) and selected a random subcohort of 4,000 participants. This subcohort was representative of the entire EPIC-Norfolk cohort in terms of age, BMI, education level, physical activity level, smoking status, and total energy intake (data not shown). Among the subcohort, there were 143 individuals who developed incident T2D during follow-up. Of the 4,749 participants, we excluded those with unknown diabetes status (n = 1) or prevalent diabetes at baseline (n = 121), those with fewer than 7 days of diary data (n = 435) or who did not return a diary (n = 15), or those with missing information on potential confounding variables (n = 73). Participants with prevalent myocardial infarction, stroke, or cancer were also excluded (n = 400). The final sample for analysis consisted of 653 incident T2D cases and a subcohort of 3,166 individuals (including 115 incident T2D cases). All volunteers gave written informed consent, and the study was approved by the Norwich district ethics committee.
We ascertained incident T2D cases by self-report of doctor-diagnosed diabetes from three follow-up health and lifestyle questionnaires, i.e., answering “yes” to “Has a doctor ever told you that you have diabetes?” or diabetes medication that was self-reported or brought to the second health check. In addition, external sources of information through record linkage included listing of any EPIC-Norfolk participant in the general practice diabetes register, local hospital diabetes register, hospital admissions data with screening for any diabetes-related admissions among study participants, and Office of National Statistics mortality data with coding for diabetes. Participants who gave a self-report of history of diabetes that could not be confirmed against any other sources of ascertainment were not considered as a confirmed case of T2D. Follow-up was censored at the date of diagnosis of T2D, 31 July 2006, or the date of death, whichever came first.
Assessment of diet and lifestyle variables
At the baseline medical examination, participants were instructed by trained interviewers on how to complete the 7-day food diary (11,12). The food diary consisted of 45 color pages containing food portion photographs and detailed instructions on how to record and describe preparation methods and quantities of foods eaten at main meals, snacks, and between meals. Completed diaries were returned by post to the coordinating center at the University of Cambridge. The food diary has been validated with weighed food records, 24-h urine collections, and blood biomarkers (13).
Intake of F&V (including tinned and dried) was calculated from food diary data to give average daily quantity of intake for each participant. In order to precisely quantify the actual intake of F&V consumed alone or from dishes (in accordance with the U.K. five-a-day public health guidelines) (14), all recorded foods and dishes were disaggregated into their component parts. Composite dishes containing fruits and/or vegetables included homemade and shop-bought desserts, vegetable bakes, stews, pies, and soups, for example. The F&V quantity and type was derived for the composite dishes by using recipes from McCance and Widdowson as previously described (12) and by using ingredients listed on the packages of products and ready-made meals. Potatoes were not included as a vegetable because they differ from vegetables regarding energy and carbohydrate content and are frequently used as a substitute for cereals (15). F&V juices were also not included because they differ from their source of origin in terms of food matrix and fiber content, and as such may be dissimilarly associated with diabetes (16). Variety of fruit, vegetables, and combined F&V intake was derived by calculating the total number of different items consumed at least once in a 1-week period, irrespective of quantity of intake. The groupings of items included 58 different fruit items (range 0–58), 59 different vegetable items (range 0–59), and hence a total of 117 different F&V items consumed over a 1-week period, as recorded in the 7-day food diary.
At recruitment, participants completed a detailed health and lifestyle questionnaire. Participants self-reported their education level (low, O level, A level, or degree), occupational social class (manual or nonmanual), smoking status (current, former, or never), and baseline history of myocardial infarction, stroke, and cancer (yes or no). Area deprivation was assessed from residential postal codes using the Townsend Deprivation Index, which provides a material measure of deprivation and disadvantage based on unemployment, car ownership, home ownership and household overcrowding. A higher Townsend Index score indicates greater deprivation (17). A validated, four-point physical activity index was derived, incorporating occupational and leisure-time components of physical activity (18). Trained nurses measured height, weight, and waist circumference according to standardized protocols (10). BMI was calculated as weight divided by height (kg/m2). Venous blood samples were taken by trained study nurses. Hemoglobin A1c (HbA1c) was measured halfway through the baseline health check (1995–1997) and was available in approximately half of the EPIC-Norfolk cohort. HbA1c was measured using high-performance liquid chromatography on a Bio-Rad Diamat (Bio-Rad, Richmond, CA), on a sample of EDTA-anticoagulated blood.
Baseline characteristics were summarized by tertiles of combined F&V quantity and variety among the subcohort participants, using means with SDs, medians with interquartile ranges (IQRs), or frequencies (where appropriate).
Multivariable, Prentice-weighted Cox regression (19) was used to estimate the associations between quantity and variety of fruit, vegetables, and combined F&V intake and hazard of diabetes, with intake defined as tertiles (with the lowest tertile as the reference category). To check the proportional hazards assumption of the models, interactions between quantity and variety of fruit, vegetables, and combined F&V intake and current age (i.e., the underlying timescale) were tested. The proportional hazards assumption was not violated for quantity and variety of fruit, vegetables, or combined F&V intake (all P values ≥0.32). Hazard ratios (HRs) and 95% CIs were estimated using the following modeling strategy. Age was used as the underlying timescale in all models. Model 1 was adjusted for sex. In Model 2, we additionally adjusted for BMI (continuous), waist circumference (continuous), education level (low, O level, A level, or degree), Townsend Deprivation Index (continuous), occupational social class (manual or nonmanual), physical activity level (inactive, moderately inactive, moderately active, or active), smoking status (current, former, or never), family history of diabetes (yes or no), total energy intake (continuous), and season of diary completion (December, January, February = winter; March, April, May = spring; June, July, August = summer; and September, October, November = autumn). In Model 3, in order to estimate the association between quantity of F&V consumption and hazard of diabetes independent of the effect of variety, we additionally adjusted for variety of F&V intake and vice versa for the analysis of variety in intake. We examined multicolinearity in Model 3 using the variance inflation factor.
In sensitivity analyses, the association between F&V quantity and variety and the hazard of diabetes was also investigated by including other potentially confounding variables in Model 3, including hypertension (yes or no), dairy intake (continuous), total fiber intake (continuous), red and processed meat intake (continuous), and percentage energy from carbohydrate (continuous), protein (continuous), fat (continuous), and alcohol intake (continuous). Analyses were also repeated after additionally excluding participants who 1) developed diabetes within the first 2 years of follow-up (n = 26), 2) had a baseline HbA1c level ≥6.5% (n = 15) in the subsample with HbA1c data available (n = 1,333), and 3) were in the top and bottom 1% of the ratio of energy intake to energy expenditure. Multiplicative interaction terms were added to Model 3 for quantity and variety of combined F&V intake to examine effect modification by sex, age (<60 or ≥60 years), BMI (normal weight <25 kg/m2, overweight/obese ≥25 kg/m2), and smoking status (never smoker or ever smoker) by using the Wald test. Additionally, spline regression was used to demonstrate the continuous association between quantity and variety of combined F&V intake and the HR (95% CI) of diabetes with knots placed at quartiles of the distribution (20).
All statistical analyses were performed using Stata/SE 11.1 (Stata-Corp, College Station, TX). Statistical significance was set at P < 0.05.
The median (IQR) duration of follow-up was 10.9 (9.8–11.8) years. The median quantity of combined F&V intake was 3.7 (2.5–5.0) portions per day and the mean (SD) variety of combined F&V intake was 11.7 (3.9) items/week. Fewer than 26% of study participants reported consuming at least five portions of F&V per day. There was nearly a threefold difference in quantity and in excess of a twofold difference in variety of combined F&V consumption between the highest versus lowest tertiles of F&V intake (Table 1). Participants who consumed higher quantities and a greater variety of combined F&V had more favorable lifestyle, anthropometric, and dietary profiles. Baseline characteristics by tertiles of quantity and variety of fruit and vegetable intake separately showed similar results (data not shown). The Pearson correlation coefficient between quantity of fruit and quantity of vegetable intake was 0.29, and between variety of fruit and variety of vegetable intake was 0.30. Quantity of combined F&V intake was strongly positively correlated with variety of combined F&V intake (0.60).
As shown in Table 2, quantity of fruit, vegetables, and combined F&V intake were all inversely associated with incident T2D (Model 1). Further adjustment did not appreciably alter the HRs (Model 2). After additionally adjusting for the effects of variety of intake, an inverse association with quantity of vegetable intake remained, but the associations for quantity of fruit and quantity of combined F&V intake with T2D were attenuated, such that fruit was no longer associated and F&V intake was borderline inversely associated with T2D. Further adjustment for hypertension, dairy intake, total fiber intake, percentage energy from carbohydrate, protein, fat, and alcohol intake, and red and processed meat intake did not change our results (data not shown). Those meeting the recommendation to consume at least five portions of F&V per day did not differ from those not meeting this recommendation in hazard of T2D, excluding and including variety of intake (HR 0.85 [95% CI 0.70–1.02] and 0.98 [0.80–1.21], respectively). As shown in Table 3, greater variety in fruit, vegetables, and combined F&V intake was inversely associated with incident T2D in adjusted analyses and also when accounting for quantity of intake. The relative reduction in the hazard of T2D with every additional two-item increase in F&V variety per week was 8% (0.92 [0.87–0.97]). The mean estimated variance inflation factor was <1.9 (<1.6 for both F&V quantity and F&V variety), indicating that colinearity of the variables included in Model 3 was low.
Figure 1 shows the continuous association between quantity and variety of combined F&V intake and T2D. As shown, the HR for T2D decreased between an intake of ∼3.5–7.0 portions per day (Fig. 1A), and this association was largely unchanged after accounting for the effects of variety of F&V (Fig. 1B). For variety, consuming ≥12 different F&V items per week was associated with a decreased HR of T2D (Fig. 1C), and this association was largely unaffected after accounting for the effects of quantity of F&V intake (Fig. 1D). The percentage of participants achieving a quantity of >3.5 portions per day was 53.2%, and the percentage achieving a variety of >12 different F&V items per week was 40.3%.
Results were unaffected in sensitivity analyses after excluding participants who 1) developed T2D within the first 2 years of follow-up, 2) had a baseline HbA1c level of ≥6.5% in the subsample with HbA1c data, and 3) were in the top and bottom 1% of energy intake to energy expenditure (data not shown). We found no evidence of interaction between either quantity or variety in F&V intake with sex, age, BMI, or smoking status and incident diabetes (P > 0.10).
In this prospective study of nearly 4,000 men and women with dietary information from prospective 7-day food diaries, we observed that a greater quantity of vegetable intake and a greater variety of fruit, vegetables, and combined F&V intake may independently be beneficial for reducing the risk of T2D. After accounting for potential confounding factors and the effects of quantity of intake, each different additional two item per week increase in variety of F&V intake was associated with an 8% reduction in the incidence of T2D.
Previous epidemiological studies have reported inconsistent findings for an association between quantity of F&V intake and risk of T2D. Two separate meta-analyses have reported no overall association between fruit, vegetables, and combined F&V intake and diabetes risk, although there was significant heterogeneity of association between the included studies (21,22). The meta-analysis by Carter et al. (21) did however find a significant inverse association between green leafy vegetable intake and risk of T2D (HR for highest vs. lowest intake group 0.86 [95% CI 0.77–0.97]). Although low heterogeneity in F&V consumption may be one explanation for the null findings in some study populations, our current results suggest that it may also be due to differences in the assessment methods used for measuring F&V intake. Most epidemiological studies have used an FFQ to assess quantity of F&V intake. Although FFQs can be used to rank individuals according to their relative intake (8), they are less suitable for the assessment of absolute intake (11,23), which they tend to overestimate. For example, in the EPIC-Norfolk study, mean consumption of F&V was much higher when assessed by FFQ (6.5 portions per day) than by a food diary (3.8 portions per day) (11). For this reason, FFQs are not ideal for examining adherence to, or for informing, public health guidelines. Furthermore, FFQs are based on perceptions of habitual intake, whereas food diaries are based on self-report of foods and amounts actually consumed in real time (8). Additionally, because FFQs contain only a limited list of precoded food items, which tend to be grouped together, unlike the food diary, which is open ended, they may not be as suitable as food diaries for assessing variety of food intake. Despite the fact that variety in F&V intake has been advocated by many national and international bodies (5–7), no studies that we are aware of have explored associations between variety in intake and risk of T2D. Our current findings suggest that quantity (at least 3.5 portions of F&V per day) and variety (at least 12 different F&V items per week) in F&V intake are both inversely and independently associated with T2D. However, only ∼50% and 40% of the participants reported meeting these thresholds for quantity and variety of intake, respectively.
There are several unique strengths of our study, including the large sample size, prospective study design, use of a 7-day prospective food diary with disaggregated F&V data, thorough assessment of new cases of T2D with self-report information supplemented by external sources, and comprehensive information on covariates, thus minimizing sources of bias and confounding. Another strength of our study was that we had HbA1c data available on a subsample of participants and were thus able to demonstrate that our findings were unlikely to have been influenced by the presence of previously undiagnosed cases of T2D at baseline. However, several potential limitations of our study merit discussion. First, because of the observational nature of the study, we cannot exclude the possibility of residual confounding or confounding by unmeasured factors. Second, we used baseline dietary consumption data to characterize individuals and did not take into account possible misclassification with respect to changes in consumption patterns over time. However, as this type of misclassification is likely to be nondifferential, the effect would be to attenuate the observed HRs toward the null, suggesting that the true associations between quantity and variety in F&V intake may be stronger than reported in the current study. We were also not able to adjust for lifestyle factors (e.g., smoking and physical activity) that may have changed during follow-up. Finally, our population is predominantly of European-Caucasian origin (99.1%) and middle aged. Thus, the generalizability of our findings to other populations may be limited. Nevertheless, in comparison with the general population of England, EPIC-Norfolk participants are comparable with respect to characteristics including anthropometry, blood pressure, and lipids (10).
The biological mechanisms for the inverse associations of F&V intake and diabetes risk are not clear. Our findings suggest that F&V may be inversely associated with diabetes through two distinct but complementary pathways. A plausible biological mechanism to explain the beneficial effect of quantity of F&V intake on T2D is via the low energy and high fiber content of F&V, and as such their ability to reduce the overall energy content of the diet. It has previously been demonstrated that those who consume the highest quantity of F&V, in comparison with low consumers, have a lower risk of weight gain (24,25), a major risk factor for diabetes (26). A decreased risk of T2D with increasing quantities of vegetable intake in particular may be explained by the fact that vegetables are generally consumed with other foods as part of a meal and therefore may displace or buffer the harmful effects of deleterious foods from the diet, such as energy-dense foods or foods that increase the risk of T2D. Alternatively, higher consumption of specific vegetables, particularly green leafy vegetables, might reduce the risk of T2D due to the presence of relatively high concentrations of potentially beneficial bioactive compounds (21). The biological mechanisms for the inverse associations of variety of F&V intake with T2D are not clear but may be attributable to individual or combined effects of the many different bioactive phytochemicals contained in F&V (e.g., vitamin C and carotenoids) (27,28). Thus, consumption of a wide variety of F&V will increase the likelihood of consuming these phytochemicals. As the current study was not designed to examine mechanisms of association, future studies will be required to investigate this further.
In conclusion, using the prospective 7-day food diary to assess F&V intake, we found that a greater variety of fruit, vegetables, and combined F&V intake is associated with a reduced risk of T2D, whereas for quantity, only greater vegetable, but not fruit intake, was associated with a reduced risk. These findings support current public health recommendations encouraging consumption of F&V as part of a balanced diet and place particular emphasis on the important and independent role that both quantity and variety in F&V intake may play in helping to prevent the development of T2D.
This study was supported by grants from the Medical Research Council, the Food Standards Agency, Cancer Research UK, and the British Heart Foundation.
No potential conflicts of interest relevant to this article were reported.
The sponsors did not participate in the design or conduct of this study; in the collection, management, analysis, or interpretation of data; in the writing of the manuscript; or in the preparation, review, approval, or decision to submit the manuscript for publication.
A.J.C. conceived and designed the study, analyzed and interpreted the data, drafted the manuscript, and had full access to all the data in the study and takes responsibility for the accuracy of the data analysis. S.J.S. analyzed and interpreted the data. M.A.H.L. analyzed data and contributed to the discussion. R.N.L. acquired the data. K.-T.K. obtained funding and acquired the data. N.J.W. obtained funding, acquired the data, and conceived and designed the study. N.G.F. conceived and designed the study and analyzed and interpreted the data. All authors critically revised the manuscript for important intellectual content and approved the final version. N.G.F., N.J.W., and K.-T.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
The authors thank the EPIC-Norfolk participants and the EPIC-Norfolk team for their contributions. The authors also thank Amit Bhaniani (University of Cambridge) for help with data and Anna Rickard and Soren Brage (Institute of Metabolic Science) for helpful discussion in earlier stages of this work.
- Received December 8, 2011.
- Accepted February 16, 2012.
- © 2012 by the American Diabetes Association.
Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.