Study design

The Chilean Ministry of Health (MINSAL) has conducted two nationwide cross-sectional Health Surveys (NHS). The first NHS was completed in 2003, and included the screening of people 17 years and older. These people were recruited by using a stratified random sample representing the adult population, considering their socioeconomic status, urban/rural residence and educational level[9].

The survey conducted during 2009–2010 was designed to follow up some of the health problems included in the NHS 2003, but also incorporated new diseases, conditions or chronic health problems, risk factors and issues related to perceived health status in the sampled population. A total of 13 conditions previously evaluated in 2003 (high blood pressure, dyslipidaemia, nutritional status, diabetes mellitus, smoking, metabolic syndrome, cardiovascular risk, sedentary lifestyles, musculoskeletal symptoms, renal function, chronic respiratory symptoms, cognitive impairment of the elderly and B and C hepatitis virus) were also evaluated in 2010[10]. This second Health Survey used a complex sampling representative of the Chilean population (15 years and older) and is the data used in this publication to measure population impact. (http://epi.minsal.cl/bases-de-datos/)

Sampling and sample size

A total of 5,293 individuals were included in the analysis with more than 8 hours of fasting to measure their glycemia levels.

The sampling frame was build from the 2002 Population and Housing Census. This cross-sectional study used a complex sample design (multistage stratified cluster sample of households) with national, regional and rural/urban area representation. The target population was adults aged older than or equal to 15 years old. The survey had a response rate of 85%, the rejection rate was 12% (n 391), and 632 subjects were excluded because physical examination was not performed. Thus, 5,293 individuals were finally interviewed. Nurses performed clinical measurements and tests to 5,043 participants and 4,956 people accepted laboratory tests (blood and urine samples). The sample loss was 28% and included refusal, inability to contact and other random loss.

The sample was designed with over-representation of the elderly and people living in areas other than the Metropolitan Region (capital region), including individuals living in rural areas, in order to increase the sample design efficiency and homogenizing the precision of the estimators.

Pregnant women and persons with violent behaviour were excluded from the random selection.

Laboratory analysis

Specialists from the ‘Catholic University of Chile’ conducted all laboratory analysis and interpretation of clinical tests. Once the survey was finished, the test results were sent to each participant, along with health recommendations according to their results. In addition, local epidemiologists provided information to each participant according to their risk level. The survey incorporated quality control processes at different stages, including a National Training Program for interviewers, an interviewer manual and the use of electronic devices to obtain automated information. These provided an interim standardization in addition to high quality laboratory analysis techniques to ensure essential methodological survey procedures.

Data processing and statistical analysis

The Ministry of Health provided a complete database with the cases and variables incorporated in the survey. Data consistency was reviewed through the analysis of determined variables distribution.

Shapiro-Wilk test was used to determine if the variables presented Gaussian distribution. Statistical tests used to evaluate differences among groups depended on the variable distribution (parametric and nonparametric tests).

Accordingly to the health-related definitions used by the National Health Survey, the variables incorporated in this article are:

• T2DM (if any of these two criteria were met, a respondent was deemed to have T2DM):
  • T2DM self-report (not during pregnancy).
  • Fasting glycaemia ≥126 mg/dL.
• Smoking status (self-reported):
  • non-smoker (never smoked),
  • former smoker (≥ 6 months of quitting tobacco)
  • and current smoker (daily and occasional smoker, and < 6 months of quitting tobacco smoke).
• Obesity: Individuals with a Body Mass Index ≥ 30 Kg/M².
• Excessive alcohol consumption: Determined by using the Abnormal Alcohol Consumption Scale ≥ 2 points[11]. This Scale has been validated in Chile and it’s widely used in the country (this cutting point is referred to abnormal alcohol consumption at some time in life)[10].
• Sedentary lifestyle: Less than 30 minutes of physical activity, a minimum of 3 times a week.

Crude and Adjusted (for the risk factors sedentary lifestyle, obesity and alcohol consumption) Odds Ratio to T2DM and its corresponding 95% confidence interval (95% CI) were estimated through logistic regressions. Attributable fractions (AF) are interpreted as the proportion of disease risk that could be prevented if the exposure was eliminated. This has a practical value in public health prevention policies, especially when the exposure is modifiable[12]. It was calculated using the following formula: AF = P(D)-(D|Ē)/P(D), where P(D) is the probability of disease, and P(D|Ē) is the probability of disease not exposed to the risk factor under evaluation. Population attributable fractions (PAF) are calculated using Levin’s formula[13].

It’s important to acknowledge some considerations about the use of these risk measures in cross-sectional studies. In first place, the risk measure is calculated from the odds ratio rather than from the risk ratio[14]. Also, as the analysis rely on prevalence data to study the association between disease and a risk factor, the following minimal assumptions are fulfilled[15]: T2DM does not lead to a change in smoking habits; smoking is not a prognostic factor of the outcome (T2DM duration is independent of the smoking status); and smoking status information is relevant for the T2DM (adequate time frame).

The analysis was performed using STATA Software, version 12[16].

Ethical aspects

The Catholic University of Chile School of Medicine Research Ethics Committee granted its ethical approval. Patients were invited to participate in the study. Before their incorporation, the objectives of the study, measurements and risks were informed to each patient, if they agreed with the protocol, an informed consent were signed for each participant.

Similar studies

Similar studies using the same methods to evaluate the association between risk factors and diabetes found no significant evidence of alcohol being one of the risk factors for type 2 diabetes, furthermore, moderate alcohol consumption has been described as protective for T2DM[5]. The proposed explanation of this protection is that moderate alcohol consumption protects cardiovascular health, especially in men when consuming 22g/per day of alcohol (RR 0.87 CI 0.76 to 1.00). Meanwhile in women, consumption of 24 g/per day of alcohol is even more protective (0.60 CI 0.52 to 0.69), when compared to lifetime abstainers[5].

Based on an analysis of BMI and overall mortality, a collaborative study notes that, in both male and female, mortality was lower at a BMI between 22.5–25 kg/m²[3]. Above this range, positive associations for different specific causes of mortality were recorded. The absolute excess risk attributable to BMI and smoking is additive, i.e., for each increment of 5 kg/m² units of BMI there is an average increase of 30% in overall mortality (hazard ratio per 5 kg/m²: 1, 29 (95% CI 1.27 to 1.32)), 40% for cardiovascular related mortality (HR 1.41 CI 1.37 to 1.45), 60–120% for diabetic cause of mortality (HRs 2.16 CI 1.89–2.46), kidney (1.59 CI 1.27 to 1.99) and liver (1.82 CI 1.59 to 2.09), 10% for mortality from cancer (HR 1.10 CI 1.06 to 1.15), 20% for respiratory causes (HR 1.20 CI 1.07 to 1.34) and 20% to other causes of mortality (1.20 CI 1.16 to 1.25)[3].

A recent study estimated the population attributable fraction of physical inactivity. Life tables were used to estimate the gain in life expectancy of the population. Globally, it was estimated that physical inactivity causes 6% (range 3.2% in East Asia to 7.8% in Eastern Mediterranean) of the burden of disease related to coronary heart disease, 7% (3.9–9.6%) of T2DM, 10% (5.6 to 14.1) of breast cancer and 10% (5.7 to 13.8) of colon cancer. It was estimated that avoiding physical inactivity could increase the world’s population life expectancy in 0.68 years (range 0.41 to 0.95)[4].

Biological basis

Over the past decade, several data showed the role of regular exercise in the prevention of T2DM, as well as it’s beneficial effects on glycemic homeostasis [17]. Recent evidence showed a reduced exercise capacity in patients with T2DM compared with non-diabetic individuals[18]. This phenomenon can be explained because insulin stimulates the muscle’s glucose uptake, which is responsible for disposing 80% to 90% of the consumed glucose load[19]. The resultant hyperglycaemia presents a stimulus to the beta cells, which secretes large amounts of insulin after meals, and it’s directly involved in the generation of insulin resistance and diabetes[20]. From a biochemical standpoint, evidence has shown that intramuscular nonoptimal lipid metabolism provide the substrate to metabolite formation associated with the development of insulin resistance through different pathways[21]. In this context, it has been proposed that sedentarism is associated with mitochondrial dysfunction in the skeletal muscle [22].

Several evidence showed an association between smoking and insulin resistance[23,24], the sub clinical condition prior to the development of T2DM [25]. In 1992 Facchini et al.[26], studied the effects of smoking over glycemic homeostasis, comparing the response of 20 healthy individuals (mean age of 39 years old) with a smoker group. The smoker group had significant higher insulinaemia. After that, Attval et al.[27], compared the peripheral insulin sensibility when smoking, finding that peripheral glucose uptake decreases.

The increase in Insulinaemia would be mediated mainly by the ability of nicotine to induce a chronic increment of insulin antagonist in plasma, such as catecholamines, cortisol and growth hormone (GH)[28–32]. An increase of catecholamines reduces the peripheral sensitivity to insulin [33,34] and the secretion of this hormone[35]. Chronic plasmatic cortisol elevation has been associated with insulin resistance, independently of obesity[36] through a mechanism that would impair β-pancreatic function and peripheral tissue insulin sensitivity[37]. Growth hormone increase induces insulin resistance through altered hepatic metabolism, decreasing peripheral recruitment of glucose and β-pancreatic effects[38].

Several epidemiological studies indicate that central obesity is an important risk factor for T2DM[39]. Adipose tissue is an active secretory organ whose secretion profile drastically changes with overweight and obesity, increasing the circulating concentrations of adipokines like leptin, or resistin [21]. Furthermore, adipose tissue macrophages starts to secrete inflammatory cytokines such as TNF α[40]. An increase in circulating levels of these adipocyte- and macrophage-derived factors in obesity leads to a chronic low-grade inflammatory state that has been linked to the development of insulin resistance and T2DM[22].

On the other hand, Obesity-related accumulation of ectopic fat in key insulin-sensitive organs (e.g., skeletal muscle and viscera) causes changes in the insulin-signaling pathways[41]. Liver steatosis is an important trigger of insulin resistance and pre-diabetes, suggesting that accumulation of intrahepatic fat is more harmful than the accumulation of ectopic fat elsewhere in the body[42]. A study[43] reported that accumulation of liver fat might affect β-cell compensation for insulin resistance.

Despite that epidemiological data found that alcohol consumption reduces the incidence of T2DM, literature also suggest that binge drinking seems to increase the incidence of this disease[44]. However, these findings seems not to be applicable to some populations in Asia, where available data suggest that alcohol intake may be a risk factor for T2DM mellitus for Japanese[45]. Chronic ethanol consumption may produce steatohepatitis [46], promoting the development of T2DM through a liver dependent pathway.

Study limitations

A group of people did not accept to provide blood samples, which were necessary to detect T2DM patients without self-report, according to the study definitions: therefore, it’s possible that T2DM prevalence has been underestimated.

Finally, as in all cross-sectional study, it’s not possible to affirm a causal association between T2DM and the researched risk factors because all of them were measured at the same time. However, our findings could help estimate at a population level the impact of public health policies to prevent T2DM. In Chile’s case, where NCDs and their risk factors are on the rise and a high impact of these risk factors have been observed on diabetes, it’s urgent to implement interventions to tackle them.