Population-Based Incidence of Type 2 Diabetes in Northern Spain: The Asturias Study

The Asturias Study is a prospective population-based survey of diabetes and cardiovascular risk factors. The baseline examination was carried out during 1998–1999 to determine the prevalence of type 2 diabetes and pre-diabetes in the Principality of Asturias in northern Spain. The population of Asturias is 1,073,761 and mostly Caucasian. Approximately half of the population lives in urban areas. A two-step sampling technique was used. Fifteen basic health areas were selected at random from among the 76 in Asturias with a probability proportional to the number of health cards of users aged between 30 and 75 years. A computer program was then used to randomly select 125 individuals in each basic health area. The final sample size selected was 1,875; 87 individuals were excluded for various reasons (type 1 diabetes, pregnancy, severe disease, hospitalization, or use of diabetogenic drugs). Another 162 individuals were excluded because data needed to contact them were missing. The final sample was composed of 1,626 individuals, of whom 1,034 (63.6%) responded. The results showed that 11.3% of all participants had diabetes, 8.9% had isolated IGT, 4.1% had isolated IFG, and 3.5% had combined IFG-IGT.

Between November 2004 and October 2005 the original participants were invited to participate in a follow-up examination. Vital status and current residency of all individuals were obtained from their health service identification card. Of the original cohort, 42 individuals had died and 19 had left Asturias before the follow-up started. Thirty other individuals were excluded (because of pregnancy, severe disease, hospitalization, or use of diabetogenic drugs). Of the remaining 943 individuals, 700 participated (74.2%). The present study includes only those individuals who did not have diabetes at the baseline.

The study was approved by the local ethics committee. All participants gave informed consent.

All examinations and analyses were performed at the patients’ local health centers by an endocrinologist and a trained nurse. Information on demographic data, smoking habits, physical activity, socioeconomic position, and a family history of diabetes was obtained by questionnaire. Smoking history was derived from yes/no responses to the questions “Have you ever smoked at least 1 cigarette per day for as long as 1 year?” and “Do you smoke cigarettes now?” Information was used to classify subjects as “current,” “former,” or “never smokers.” Self-reported leisure-time physical activity was classified into three categories: “low,” almost completely inactive; “moderate,” some physical activity >1 h/day, e.g., walking, gardening, dancing, or bicycling; or “high,” vigorous physical activity >3 h/week, e.g., running, swimming, ball games, or competitive sports. A family history of diabetes was coded as positive when the subject had one or more first-degree relatives with the disease. The area of residence was classified according to the municipal population: “rural,” <10,000 inhabitants; “medium,” 10,000–50,000 inhabitants; or “urban,” >50,000 inhabitants. Occupation, work status (active/unemployed/retired), education level (none/basic/high school/college), and marital status (single/married/separated/widowed) were also assessed. Medical records were reviewed to investigate previous diseases or medication.

Height, weight, and BMI (weight in kilograms divided by the square of height in meters) were measured with the subject wearing light clothing and without shoes. Blood pressure was measured with a digital sphygmomanometer (OMROM MX3) after several minutes in a seated position; the mean of two measurements taken 1–2 min apart was used for analysis.

All participants without known diabetes underwent an oral glucose tolerance test (OGTT), both at baseline and at the follow-up, obtaining fasting and 2-h venous samples following the recommendations of the World Health Organization (20). The samples were centrifuged in situ using a portable centrifuge. A portable refrigerator containing the samples was taken daily to the Biochemical Laboratory of the Central Hospital of Asturias. Glucose was determined by the hexokinase enzymatic method (Hitachi 747). The glucometabolic state was classified according to the World Health Organization 1999 criteria: IFG, fasting plasma glucose (FPG) 110–125 mg/dl; IGT, 2-h plasma glucose 140–199 mg/dl; and type 2 diabetes, FPG ≥126 mg/dl and/or 2-h plasma glucose ≥200 mg/dl (20). Additional laboratory measurements included total cholesterol, HDL cholesterol, triglycerides (colorimetric method), LDL cholesterol (Friedewald formula), and A1C (high-performance liquid chromatography, Adams HA-8160). C-reactive protein, insulin, and microalbuminuria were only measured at the follow-up examination and were not included in this analysis.

The incidence rate of diabetes in the cohort was calculated, assuming a constant rate of disease over the follow-up and expressed per 1,000 person-years. Adjusted incidence was estimated by the direct method using the total population in Asturias aged 30–75 years according to 1998 census. The strength of association between clinical categories and the development of diabetes was measured by calculating age- and sex-adjusted odds ratios (ORs) by univariate logistic regression analyses. The variables age, sex, education level, family history of diabetes, physical activity, diastolic blood pressure, total cholesterol, HDL cholesterol, triglycerides, BMI, FPG, and 2-h plasma glucose were included in a stepwise, multivariate, logistic regression analysis to study their independent prognostic significance. All statistical analyses were performed with SPSS 12.0 and Epibasic 1.0. Reported P values are based on two-sided tests with a cutoff for statistical significance of 0.05.

Of the original participants, 905 did not have diabetes at baseline. Of these, 24 had died, 15 had left Asturias before the reevaluation started, and 26 others were excluded (because of pregnancy, severe disease, hospitalization, or use of diabetogenic drugs). Of the remaining 840 individuals invited to take part in the follow-up study, 634 participated (75.5%). Table 1 shows that there were virtually no differences between participants and nonparticipants for most baseline parameters. Four individuals were excluded from the final analyses because of lack of OGTT data. Thus, the study population included 630 subjects (56.2% female and 43.8% male) with an age range of 30 to 75 years at baseline. The mean follow-up was 6.3 years (range 5.9–6.8 years).

In the 630 individuals who completed the study, we found 44 new cases of diabetes (7%). This figure represents a crude incidence of diabetes of 11.2 cases/1,000 person-years (95% CI 8.1–14.8) and an incidence adjusted for the age and sex structure of the Asturian population in this age range of 10.8 cases/1,000 person-years (95% CI 7.8–14.6).

Table 2 includes the incidence rates and ORs of various risk factors involved in the etiopathogenesis of type 2 diabetes. The incidence was slightly higher in men than in women and rose with increasing age. We found no significant risk for diabetes according to the area of residence. For socioeconomic position, rather than occupation or work status, which had no associated risk (data not shown), education level was strongly associated with the development of type 2 diabetes, the risk being more than threefold higher in individuals who did not have a high school diploma than for individuals who had received higher education. The risk for diabetes was virtually the same in current smokers as in nonsmokers. Comparison of current and former smokers versus never smokers showed similar results (data not shown). Other variables such as a family history of diabetes, low physical activity, systolic hypertension, diastolic hypertension, obesity, hypertriglyceridemia, and low HDL cholesterol were also associated with the development of diabetes. Both IFG and IGT were highly associated with diabetes. These two groups overlapped, with the risk being moderate in individuals with isolated IFG or isolated IGT and extremely high in individuals with combined IFG-IGT (Table 3).

Stepwise multiple logistic regression analysis showed that, together with FPG and 2-h plasma glucose, which were the strongest predictors of diabetes, triglycerides and BMI were also independently associated with progression to diabetes (Table 4). In a parallel analysis with a fixed model including the same variables, only FPG and 2-h plasma glucose remained significantly predictive of diabetes (both P < 0.001).

In this prospective study, we report the incidence rates of type 2 diabetes and evaluate various risk factors for the development of diabetes in a representative sample of the general population of a whole province in northern Spain, including residents in both urban and rural areas. Cases of diabetes were determined on the basis of an OGTT both at baseline and at follow-up and are thus an approximation of the total figures for diabetes (known + unknown). Although the participation at the follow-up of the original cohort was not complete (75.5%), very few differences existed between participants and nonparticipants, thereby minimizing any possible selection bias.

The incidence rate of type 2 diabetes has not been clearly defined for the European population, mainly because data are lacking. Some studies that determined cases of diabetes based on questionnaires, review of medical records, or use of drugs showed results ranging from 1 to 5 cases/1,000 person-years (21–26). These figures are clearly underestimates, as the studies only included persons who already had a diagnosis of diabetes. Studies based on fieldwork with repeated glucose analysis data are fewer. The only previous reference in Spain is the Lejona Study, in which a cohort of 594 individuals without diabetes at baseline were reevaluated after a follow-up of 10 years. The incidence of diabetes was 8.2 cases/1,000 person-years (27). In Naples, 560 employees of a telephone company (aged 40–59 years) without diabetes were prospectively studied after an initial cross-sectional study. Over a follow-up of 11.5 years, 54 had progressed to diabetes. However, only FPG was determined at the follow-up study (28). In Ely, U.K., a longitudinal study included 937 individuals without diabetes with a mean follow-up of 4.5 years and reported an incidence of 6.2 cases/1,000 person-years (29). The Hoorn Study in Holland evaluated 1,342 nondiabetic individuals at baseline and after a mean of 6.4 years, finding 133 new cases of diabetes, which represents an approximate incidence of 15 cases/1,000 person-years. However, the high age level of the cohort (50–75 years) could account for this high incidence of diabetes (30). In the Bruneck Study (northern Italy), a total of 837 individuals without diabetes were included. After a follow-up of 10 years, 64 had progressed to diabetes. The incidence adjusted to the age and sex structure of Bruneck was 7.6 cases/1,000 person-years. Only FPG was determined on reevaluation and, based on a first cutoff made after 5 years of follow-up, the authors estimated that if they had done an OGTT in this last cutoff, the incidence would have been ∼30% greater (∼1% annually) (31). This theoretical figure would be similar to that of our study, in which we found 10.8 new cases of diabetes/1,000 person-years. If we assume these data for the overall population of Asturias aged ≥30 years (762,822 inhabitants according to the census of 2004), we would have the alarming figure of ∼7,000 new cases of type 2 diabetes every year in our community.

As in other studies (31,32), both IFG and IGT, as well as family history of diabetes, low physical activity, and core components of the metabolic syndrome such as obesity, hypertension, hypertriglyceridemia, and low HDL cholesterol, were important individual risk factors for development of diabetes in our study. Interestingly, low education level was an even stronger predictor of diabetes than other clinical or biological variables, supporting previous reports that highlight its importance in the development of diabetes (33,34).

After introducing all these variables into a multivariate model, both FPG and 2-h plasma glucose showed the strongest association with diabetes. Of the two, FPG conferred the highest risk, this being a common finding in most studies (29–31). However, it should be stressed that the diabetes risk profile can best be stratified by adding the 2-h plasma glucose measure to FPG by performing an OGTT, because their effects are cumulative. This observation is not surprising, considering that the metabolic determinants of both variables are different: having a normal FPG requires adequate basal insulin secretion and appropriate hepatic insulin sensitivity to control hepatic glucose output, whereas maintaining a normal postload glucose level after an OGTT requires an adequate β-cell secretory response as well as adequate hepatic and especially peripheral insulin sensitivity. Abnormalities of these metabolic functions characterize IFG and IGT, and the worst insulin resistance–insulin secretion patterns are seen in the IFG-IGT group (35,36). In our study, subjects with isolated IGT had a 50% lower risk than those with isolated IFG, but those with both IFG-IGT combined had a more than fivefold higher risk than those with just IFG alone. Performing an OGTT enables identification of these very high-risk persons in whom preventive measures should be intensified. It should also be remembered that 2-h plasma glucose is a better predictor of death and cardiovascular disease than is FPG (37), such that performing an OGTT would also provide a better definition of cardiovascular risk than FPG alone.

Triglycerides and BMI were also strongly predictive of diabetes, and, in addition to glucose measurements, they could be useful in defining high-risk individuals. A family history of diabetes, although one of the strongest known risk factors for development of diabetes, failed to enter the model. This variable, though, like all self-reported data could be misclassified.

In summary, this 6-year, prospective study carried out in a low-risk Spanish Caucasian population showed an incidence rate of type 2 diabetes of 10.8 cases/1,000 person-years. FPG and 2-h plasma glucose were highly predictive of diabetes, and their effects were additive. Including both glucose measurements best defined the diabetes risk in this population.

The fieldwork for the baseline examination of the Asturias Study (1998–1999) was undertaken by P.B. and Elena Téllez and was supported by a grant from the Fondo de Investigación Sanitaria (98/0462). The fieldwork for the follow-up examination of the Asturias Study (2004–2005) was done by S.V. and Karmele Molinuevo and was supported by grants from the Consejería de Salud del Principado de Asturias and Pfizer Laboratories.

We thank all the volunteers who participated in the two studies, without whose help they would have been impossible to perform. We also thank the primary care teams of the selected areas, Ismael Huerta, Shura Rozada, Enrique Bustillo, and the Endocrinology Group of Asturias.