The consumption of different foods was studied for their ability to predict type II diabetes mellitus.
The study design was a cohort study, based on the Finnish Mobile Clinic Health Examination Survey.
A total of 30 communities from different parts of Finland.
A total of 4304 men and women, 40–69 y of age and free of diabetes at baseline in 1967–1972 and followed up for incidence of diabetes medication during 23 y (383 incident cases).
Higher intakes of green vegetables, fruit and berries, oil and margarine, and poultry were found to predict a reduced risk of type II diabetes. The relative risks of developing type II diabetes between the extreme quartiles of the intakes were 0.69 (95% confidence interval (CI)=0.50–0.93; P for trend (P)=0.02) for green vegetables, 0.69 (CI=0.51–0.92; P=0.03) for fruit and berries, 0.71 (CI=0.52–0.98; P=0.01) for margarine and oil, and 0.71 (CI=0.54–0.94; P=0.01) for poultry.
The results suggest that prevention of type II diabetes might be aided by consumption of certain foods that are rich in nutrients with hypothesized health benefits.
Recent studies have shown that the choice of foods plays a role in diabetes prevention. Consumption of fruit and vegetables (Snowdon & Phillips, 1985; Colditz et al, 1992; Feskens et al, 1995; Ford & Mokdad, 2001) and whole grains (Liu et al, 2000; Meyer et al, 2000; Fung et al, 2002; Montonen et al, 2003) have shown inverse associations with the risk of diabetes. The associations are independent of obesity, a major risk factor for diabetes. These findings have been attributed to free radical scavenging properties of carotenoids, vitamin C, vitamin E, and flavonoids (Gordon, 1996; Halliwell, 1999), to relief of the postprandial glucose load by soluble fiber (Slavin et al, 1999), or to avoidance of hypomagnesemia (Paolisso et al, 1990). Fruit and vegetables also contain other potential compounds, such as phytates or isoflavones, which may have additive or synergistic effects (Liu, 2003).
The relation between processed meat consumption and the development of diabetes has been addressed in a number of studies (Snowdon & Phillips, 1985; Colditz et al, 1992; van Dam et al, 2002b; Schulze et al, 2003). The alleged mediators include saturated fat, cholesterol, nitrites, and advanced glycation end-products (Peppa et al, 2002; van Dam et al, 2002b; Schulze et al, 2003). The type and amount of fat consumed may also play a role (Costacou & Mayer-Davis, 2003). High fat intake may contribute to the development of diabetes via obesity. Findings of several studies suggest that higher intake of polyunsaturated fat and possibly long-chain n-3 fatty acids could be beneficial in prevention of type II diabetes, whereas higher intake of saturated fat and trans-fat may be detrimental (Hu et al, 2001b). Other dietary factors that have been related to reduced risk of type II diabetes include coffee (van Dam & Feskens, 2002; Tuomilehto et al, 2004), dairy products (Pereira et al, 2002), and vitamin D (Baynes et al, 1997).
Besides the studies focusing on individual foods or nutrients, other studies have recently followed a dietary pattern approach and have found certain dietary patterns to be associated with the incidence of diabetes (van Dam et al, 2002a; Montonen et al, 2004a) or biomarkers of diabetes development (Fung et al, 2001). The advantage of the dietary pattern approach is that interactions between potentially effective nutrients are automatically included in the effect estimate. On the other hand, the question is still open of whether interactions between nutrients, rather than a simple sum of the effect of individual components, contribute to predictive value in the dietary pattern approach. Thus far, comprehensive studies on the association between intakes of individual foods and diabetes risk are few. In this follow-up study the relationships between intakes of a wide range of different foods on diabetes incidence was investigated to establish their potential in predicting type II diabetes. The excess risk associated with a previously defined dietary pattern score (Montonen et al, 2004a) was also evaluated through controlling for the significant dietary variables.
Subjects and methods
The Finnish Mobile Clinic Health Examination Survey
The Finnish Mobile Clinic Health Examination Survey carried out health examinations in various parts of Finland during 1966–1972. The selection and characteristics of the population examined have been described elsewhere (Aromaa, 1981; Reunanen et al, 1983; Knekt, 1988). A dietary history interview of 10 054 citizens 15 y of age or more was included in the study from 1967 (Järvinen, 1996). The study population comprised 4304 men and women 40–69 y of age. It excluded other age groups, people who reported a daily energy intake less than 800 or more than 6000 kcal, pregnant women, and people with a history of diabetes.
All participants completed a self-administered questionnaire that was checked at the baseline. The questionnaire yielded information on occupation, current pregnancy, babies born with birthweight over 4500 g, previous and current illnesses, consumption of medicines, general health status, and health-related habits, such as smoking. Occupation was grouped into nine categories according to the Nordic Standard Classification of occupations (Brockington, 1967). The subjects were classified according to smoking status as follows: never-smoked, exsmokers, smokers of pipes or cigars only, smokers of less than 15 cigarettes per day, and smokers of 15 or more cigarettes per day (Knekt, 1988). Three choices were given for self-rated general health: excellent, fairly good, poor health. Body weight and height were measured, and body mass index calculated (kg/m2). Casual blood pressure was measured using the auscultatory method. Four hypertension categories were formed on the basis of systolic blood pressure (SBP) and diastolic blood pressure (DBP) and antihypertensive medication (Aromaa, 1981). Subjects with SBP ≥170 mmHg and DPB ≥100 mmHg and subjects using antihypertensive medication were considered to be definitely hypertensive. Subjects with SBP≥160 mmHg and DBP≥95 mmHg but not defined as hypertensive were considered to have mild hypertension, and those with SBP<140 mmHg and DBP<90 mmHg were considered normotensive. All subjects with intermediate values were considered to have borderline hypertension. Serum cholesterol concentration was determined with an autoanalyzer modification of the Liebermann–Burchard reaction (Huang et al, 1961).
Known cases of diabetes were identified from information given by the participants. A glucose tolerance test was carried out to diagnose new diabetes at baseline using diagnostic criteria of the World Health Organization (WHO, 1985). Previously known or new diabetics were excluded from the analysis.
Assessment of diet
Total habitual food consumption during the previous year was estimated using a dietary history interview (Järvinen, 1996). Trained interviewers conducted these structured interviews using a questionnaire form listing over 100 food items and mixed dishes common in the Finnish diet. Several questions were open-ended, giving the respondent an opportunity to provide more detailed answers and any further explanations during the interview. Food models were used as an aid in determining portion sizes. The consumption of foods was estimated per day, week, month, or year according to the choice of the respondent. Individual consumption of food items was converted to grams per day. Energy intake was calculated on the basis of the intake of protein, fat, and available carbohydrate.
Reproducibility of the dietary history method was estimated after 4–8 months (93 individuals) and 4–7 y (1844 individuals) by repeating the interview. The reproducibility of the dietary interviews completed 4–8 months apart was good, but long-term consistency 4–7 y apart was rather poor (Järvinen et al, 1993).
Determination of diabetes incidence
During a 23-y follow-up, a total of 164 male and 219 female incident type II diabetes cases were identified from the Social Insurance Institution's nationwide register of persons receiving drug reimbursement (Reunanen et al, 1983). According to the Finnish sickness insurance legislation, diabetics needing drug treatment are allowed certain drugs free of charge. In order to get this drug allowance a certificate must be obtained from the physician in charge describing the diagnostic criteria applied when the diabetes was diagnosed. The certificate is accepted after checks have been made by special advisers at the Social Insurance Institution. The participants in the present study were identified in the Institution's register on basis of the unique social security codes assigned for each Finnish citizen. The follow-up refers to the period of observation from the baseline examination until onset of the disease, or until death or the end of the observation period (late 1995), whichever came first.
The relative risks (RR) of developing type II diabetes with 95% confidence intervals (CI) between quartiles of foods were calculated using Cox's model (Cox, 1972). The data for poultry and canned or frozen fish were divided into consumers (>0 g/day) and nonconsumers, because the number of consumers was low. Energy intake and other potential confounding or effect-modifying factors (age, sex, body mass index, smoking, family history of diabetes, and geographic area) were entered in the model. Tests for trends through the quartiles were carried out based on a likelihood ratio test, treating all variables in the model as continuous by entering the quartile numbers.
In further analyses we studied the relative importance of the previously identified dietary patterns, the ‘prudent’ pattern (characterized by fruit and vegetables) and the ‘conservative’ pattern (characterized by butter, potatoes, red meat, and whole milk) (Montonen et al, 2004a). To identify dietary patterns, the principal component method with Varimax rotation in the factor analysis was applied (SAS/STAT, 1989; Hu et al, 1999). Two factors with eigenvalues greater than 2.5 were retained and other factors with eigenvalues less than 1.5 discarded based on the results of a scree test and interpretability of the factors (Kim & Mueller, 1978). The factor scores were used to rank participants according to the degree to which they conformed to each dietary pattern. Dietary pattern scores were adjusted for total energy intake using the residual method (Willett, 1998; van Dam et al, 2002a) and categorized into quartiles. To study the possible additional effect of the score variables, besides the individual dietary variables, the dietary variables were stepwise added to a model including the pattern score variables.
Baseline characteristics of the study population according to diabetes occurrence during follow-up are presented in Table 1. Persons who developed diabetes during the follow-up had a higher body mass index, were older, more likely to be hypertensive or female, less likely to be smokers, and more often had a family history of diabetes.
Consumption of vegetables (especially green vegetables), fruit and berries, margarine and oil, and poultry was associated with a reduced risk of diabetes after adjustment for age, sex, energy intake, and body mass index (Table 2). Further adjustment for geographic area, smoking, and family history of diabetes tended to attenuate the association observed. However, the inverse association between intakes of green vegetables, fruit and berries, margarine and oil, and poultry and diabetes risk remained significant. RR between the highest and lowest quartiles of intakes were 0.69 (95% CI)=0.50–0.93; P for trend (P)=0.02) for green vegetables, 0.69 (CI=0.51–0.92; P=0.03) for fruit and berries, 0.71 (CI=0.52–0.98; P=0.01) for margarine and oil, and 0.71 (CI=0.54–0.94; P=0.01) for poultry (poultry intake was divided into nonconsumers and consumers). Further adjustment for hypertension and serum cholesterol did not notably alter the result (data not shown); neither did adjustment for occupation or self-rated health (data not shown). Intake of potato was associated with increased risk of type II diabetes (RR=1.42; CI=1.02–1.98; P=0.03), whereas processed meat and red meat consumption showed no association (Table 2). No significant interaction between food intake and age, sex, body mass index, or smoking was found (data not shown).
The conservative pattern was positively associated and the prudent pattern inversely associated with diabetes occurrence when adjusted for age, sex, body mass index, energy intake, smoking, family history of diabetes, and geographic area (Table 3). To explore the extent to which the individual foods of the prudent and conservative patterns accounted for the effects observed, we evaluated the contribution of the components which were associated with diabetes risk by adding them to the model. Stepwise addition of each component of the prudent pattern to the model attenuated the association between the prudent pattern and diabetes risk (Table 3). In the final model, prediction of the prudent dietary pattern was entirely explained by fruit and berries, green vegetables, margarine and oil, and poultry (RR=1.05; CI=0.67–1.63; P=0.75). Adjustment for intakes of potato, processed meat, and butter had only a weak effect on the relation between the conservative pattern and diabetes risk (RR=1.35; CI=0.97–1.85; P=0.08).
In the present study, consumption of vegetables, fruit and berries, margarine and oil, and poultry was inversely associated with risk of type II diabetes. These results are in line with the results of previous epidemiological studies which have shown inverse associations between the consumption of fruit and vegetables and the risk of diabetes or 2-h postload glucose concentrations (Snowdon & Phillips, 1985; Colditz et al, 1992; Feskens et al, 1995; Ford & Mokdad, 2001).
Fruit and berries and vegetables are rich sources of antioxidant compounds such as carotenoids, vitamin C, vitamin E and flavonoids, and of fiber. Carotenoids, vitamin C, and vitamin E may have a protective effect against development of diabetes by relieving oxidative stress that interferes with the glucose uptake by cells (Halliwell & Gutteridge, 1989; Gordon, 1996). In several epidemiological studies, intakes or serum levels of vitamin C, vitamin E, or carotenoids have been inversely associated with diabetes (Chatterjee & Banerjee, 1979; Sinclair et al, 1994; Feskens et al, 1995; Salonen et al, 1995; Abahusain et al, 1999; Ford et al, 1999; Knekt et al, 1999; Will et al, 1999; Polidori et al, 2000; Montonen et al, 2004b). However, the findings on the relation between antioxidants and glucose metabolism are not consistent (Sanchez-Lugo et al, 1997; Reunanen et al, 1998; Liu et al, 1999). Moreover, flavonoids may provide protection against chronic diseases by their free radical scavenging properties. In recent analysis of our data, intake of flavonoids was found to be related to reduced type II diabetes risk (Knekt et al, 2002). It has also been hypothesized that soluble fiber from fruit and vegetables inhibits the postprandial glucose load by delaying the absorption of carbohydrates and may thus inhibit diabetes development (Anderson et al, 1979). Previous epidemiological studies have shown an inverse association between dietary fiber intake and diabetes incidence (Salmeron et al, 1997; Meyer et al, 2000; Hu et al, 2001a; Salmeron et al, 2001). Fiber specifically derived from vegetables or fruit has, however, not been associated with diabetes risk (Colditz et al, 1992; Montonen et al, 2003).
Margarine and oil intake was associated with a reduced diabetes risk in our study, which is in line with the results of previous epidemiological studies that have found an association between the intake of vegetable fat and decreased diabetes risk (Colditz et al, 1992; Vessby et al, 1994; Meyer et al, 2001; Salmeron et al, 2001). The type of dietary fat may play a role in diabetes development by modifying the fatty acid composition of the phospholipid cell membrane, which may affect insulin receptor properties and glucose transportation (Storlien et al, 1991). However, individuals preferring diets high in margarine and oil may be more health conscious, in which case the reduced diabetes risk among them could equally well be due to healthier lifestyles.
The CARDIA study showed an inverse association between intake of dairy products and development of insulin resistance in young adults (Pereira et al, 2002). The association was independent of the hypothesized biological mediators, that is, calcium, potassium, and magnesium. In the present study, intake of regular dairy products suggested an inverse but nonsignificant tendency in accordance with the finding by Pereira et al (2002).
In previous epidemiological studies, meat intake, especially processed meat, has been associated with increased risk of diabetes (Snowdon & Phillips, 1985; Colditz et al, 1992; van Dam et al, 2002b; Schulze et al, 2003). It has been hypothesized that the association is mediated by the beta-cell toxic effect of nitrosoamines or by saturated fat, or advanced glycation end-products (Peppa et al, 2002; van Dam et al, 2002b). The role of the processed meat intake was not supported by the present study. However, the intake of poultry was found to be inversely associated with the risk of diabetes.
Potato intake was associated with an increased risk of diabetes. Abundant potato intake is probably linked to diabetes development through a mechanism related to increased postprandial glucose levels, although the glycemic index of the potato depends largely on the strain and the method used in preparation (Foster-Powell & Miller, 1995).
Previously reported results of the Health Professionals Follow-up Study showed an inverse relation between the prudent dietary pattern (rich in vegetables, fruit, fish, poultry, and whole grains) and a reduced risk of type II diabetes (van Dam et al, 2002a). A prudent dietary pattern has also been cross-sectionally associated with lower insulin concentrations among 466 men (Fung et al, 2001). In a British cross-sectional study, a dietary pattern characterized by high consumption of fruit and vegetables and low consumption of processed meat and fried foods was inversely associated with the prevalence of type II diabetes (Williams et al, 2000).
The dietary pattern analysis closely reflects the real-world observations, accounting for potential additive effects and interactions between nutrients and potential nondietary factors associated with food intake. In the present study, however, the relation between the prudent pattern and diabetes risk was completely explained by certain dietary components, suggesting that the pattern actually summarizes the effects of individual foods or food groups. On the other hand, adjustment for dietary components did not alter the association between the conservative pattern and diabetes risk. These results suggest that potential nondietary factors beyond the prudent pattern possibly have a similar effect in the analyses of individual foods, although the predictive significance of the conservative pattern was independent of the dietary factors. It is also possible that foods in the conservative pattern, but not in the prudent one interact with each other to get the total effect.
Measurement of dietary data requires an appropriate instrument to collect the information on the use of food items of interest. In the present study, the dietary data were collected using a 12-month dietary history interview applied to a nationwide sample. Although the dietary history interview is considered relatively accurate, certain inaccuracies in the method tend to affect the associations observed between dietary exposure and outcome (Willett, 1998). In the present study, all the interviewers were trained nutrition professionals and a preformed questionnaire was used to minimize differences between interviewers. The questions were open-ended and offered opportunities to add further details to the answers (Järvinen, 1996). Likewise, food models were used to reduce errors of recall in regard to portion size. In general, the short-term repeatability of the dietary history method was good, but long-term reliability was rather poor due to changes in food consumption (Järvinen et al, 1993).
The information on incident diabetes was obtained from a nationwide register of drug reimbursements (Reunanen et al, 1983). The register does not include diabetics undergoing dietary therapy only. Persons with dietary therapy or unidentified persons with diabetes may have led to conservative estimates of the strength of the associations between food intake and diabetes risk. False-positive cases were eliminated by the centralized double-checking of the evidence on diabetes by special advisers at the Social Insurance Institution. Unfortunately, valid data on physical activity was not available. Since persons with healthier diets may be physically more active than other persons, the lack of physical activity data in particular may have confounded the results. It is, however, possible that physical activity has in part been accounted for indirectly by adjusting for body mass index and energy intake. Another methodological issue is that numerous statistical tests were carried out and accordingly the probability of chance findings was increased.
In the present study, adjustment for occupation or self-rated general health did not notably alter the result, suggesting that the associations found are not completely explained by nondietary lifestyle factors. Further studies are needed to discover whether the association between diet and diabetes is mediated by hypothesized nutrients or by lifestyle and sociodemographic factors related to dietary patterns.
In summary, the results of the present study suggest an inverse association between intakes of vegetables, fruit and berries, margarine and oil and poultry, and risk of type II diabetes. It seems conceivable that the beneficial effect of the prudent dietary pattern in prevention of type II diabetes is derived from consumption of these foods.