The Impact of Dietary and Metabolic Risk Factors on Cardiovascular Diseases and Type 2 Diabetes Mortality in Brazil

Background Trends in food availability and metabolic risk factors in Brazil suggest a shift toward unhealthy dietary patterns and increased cardiometabolic disease risk, yet little is known about the impact of dietary and metabolic risk factors on cardiometabolic mortality in Brazil. Methods Based on data from Global Burden of Disease (GBD) Study, we used comparative risk assessment to estimate the burden of 11 dietary and 4 metabolic risk factors on mortality due to cardiovascular diseases and diabetes in Brazil in 2010. Information on national diets and metabolic risks were obtained from the Brazilian Household Budget Survey, the Food and Agriculture Organization database, and large observational studies including Brazilian adults. Relative risks for each risk factor were obtained from meta-analyses of randomized trials or prospective cohort studies; and disease-specific mortality from the GBD 2010 database. We quantified uncertainty using probabilistic simulation analyses, incorporating uncertainty in dietary and metabolic data and relative risks by age and sex. Robustness of findings was evaluated by sensitivity to varying feasible optimal levels of each risk factor. Results In 2010, high systolic blood pressure (SBP) and suboptimal diet were the largest contributors to cardiometabolic deaths in Brazil, responsible for 214,263 deaths (95% uncertainty interval [UI]: 195,073 to 233,936) and 202,949 deaths (95% UI: 194,322 to 211,747), respectively. Among individual dietary factors, low intakes of fruits and whole grains and high intakes of sodium were the largest contributors to cardiometabolic deaths. For premature cardiometabolic deaths (before age 70 years, representing 40% of cardiometabolic deaths), the leading risk factors were suboptimal diet (104,169 deaths; 95% UI: 99,964 to 108,002), high SBP (98,923 deaths; 95%UI: 92,912 to 104,609) and high body-mass index (BMI) (42,643 deaths; 95%UI: 40,161 to 45,111). Conclusion suboptimal diet, high SBP, and high BMI are major causes of cardiometabolic death in Brazil, informing priorities for policy initiatives.


Introduction
Brazil is undergoing rapid economic, demographic and behavioral transitions that are producing major impacts on health. Cardiometabolic diseases including cardiovascular disease (CVD) and type 2 diabetes are the major causes of mortality, leading to nearly 35% of deaths among Brazilian adults [1]. National trends show increasing prevalence of suboptimal levels of wellestablished cardiometabolic risk factors over last three decades. Trends in national food availability and consumption demonstrate shifts toward unhealthy dietary patterns characterized by higher intakes of processed foods, refined grains, and sugar sweetened beverages; and lower intakes of healthy, minimally processed foods such as fish, fruits, and vegetables [2,3]. At the same time, the prevalence of adverse metabolic risk factors such as obesity are rising, with a 2-6 fold increase in prevalence of overweight and obesity among adults over the last four decades [4]. Approximately one-third of Brazilian adults have dyslipidemia [5], and about onethird have hypertension [6]. Clearly, the prevalence of suboptimal levels of chronic disease risk factors, particularly those related to poor diet and metabolic health, is substantial, leading to increasing burden on cardiometabolic mortality among Brazilian adults.
To design evidence-informed policies to address this problem, there is a need for clear, consistent data on burdens due to specific risk factors. Yet, the impact of major dietary and metabolic risk factors on cardiometabolic mortality in Brazil according to sex and age has not been clearly established. To address these crucial questions, we used data from the 2010 Global Burden of Diseases, Injuries, and Risk Factors (GBD) Study to evaluate the impact of 11 dietary and 4 metabolic risk factors on cardiometabolic mortality in Brazil in 2010 by sex and age.

Methods
The 2010 GBD Study is an international collaborative effort since 1990 to produce comprehensive and comparable estimates of the burdens of diseases, injuries and risk factors in 187 countries and territories [7,8]. Details on methods and standardized data collection protocol have been previously described [9,10]. We conducted a population-level comparative risk assessment analysis within Brazil for 15 dietary and metabolic risk factors to quantify the overall burdens of disease due to each factor by age and sex.

Assessment of Risk Factors
A total of 15 major modifiable dietary and metabolic risk factors (11 dietary and 4 metabolic factors) were selected based on probable or convincing evidence for causal effects on coronary heart disease (CHD), stroke or diabetes [11][12][13][14][15]. Mean and standard deviation of dietary exposures by age and sex were obtained from the Brazilian National Dietary Survey (NDS), the dietary component of the 2008-09 Household Budget Survey (POF), which included approximately 55,000 households nationwide. A subset including 25% of the households included in the POF was randomly selected for the dietary survey [16,17]. In total 34,003 Brazilian adults and children !10y of age participated in the NDS, providing two days of dietary records (nonresponse rate:19%). These data were combined with food disappearance information from the Food and Agriculture Organization database to provide national dietary estimates. Data sources for dietary and metabolic exposures from nationally representative surveys and/or epidemiological studies are shown in Table 1.
We evaluated national distributions of body-mass index (BMI), total cholesterol, fasting plasma glucose, and systolic blood pressure (SBP) as previously described [11][12][13][14]. Information on exposure levels of metabolic risk factors was obtained from 13 observational studies including Brazilian adults conducted between 1990 and 2008 (Table 1). Age-integrating Bayesian hierarchical modeling methods were used to derive estimates of mean levels of each risk factor and their 95% uncertainty interval (UI) separately by age and sex subgroups. For each exposure, the primary model inputs were nationally representative survey data and exposure levels from epidemiologic studies for each sex and age group, data on the numbers of subjects in each stratum; survey-level indicator covariates for sampling representativeness, food disappearance information from the Food and Agriculture Organization, and country, region (21 regions), and superregion (7 groupings of regions) random effects. This model leverages the interrelationshps between all data to inform final estimates, while appropriately allowing strongest input to Brazilspecific data. [15,[24][25][26]. Since information on the inherent correlation between age and sex groups resulting from the same underlying Bayesian models was not available, we have not generated p-values for comparison of exposure estimates across subgroups. 95% UIs can be used for informal inference on comparisons between groups: i.e., if the mean value of one estimate is outside the 95% UI of another, these could be considered nominally significantly different.

Etiological Effects of Risk Factors on Mortality
The relative risks (RR) of CHD, stroke, and diabetes mortality or incidence for each dietary and metabolic risk factor were obtained from published systematic reviews and meta-analyses of randomized controlled trials or prospective cohort studies [27]. When no recent systematic review or meta-analysis were available, de novo meta-analyses were conducted [28]. For these analyses, we used evidence available through November 2014. Most observational studies included had adjusted for relevant potential confounders such as age, sex, race, socioeconomic status, education, physical activity, smoking and alcohol consumption. Etiological effects of metabolic risk factors were estimated using pooled analyses of large cohort projects [29]. For each risk factor-disease pair, we used consistent age-varying distributions of RRs as previously described, incorporating smaller relative effects with increasing age [29]. RRs for both sexes combined was used for men and women, since results from meta-analyses indicated no material differences in RRs by sex [29].

Optimal Risk Factor Levels
In order to estimate deaths due to suboptimal distributions of risk factors, the optimal distribution was defined as the exposure associated with the lowest possible disease risk that was feasible and observed in some populations, based on current evidence from epidemiologic studies [12,15,23,29] (Table 1). Optimal intake distribution for dietary factors with protective effects (e.g. fruits, vegetables, whole grains) was defined as the intakes to which beneficial effects may plausibly continue based on evidence from current epidemiological studies. For  [21,22] IHD, ischemic stroke, DM, hypertensive heart disease IHD: ischemic heart disease, CVD: cardiovascular disease, DM: diabetes mellitus; BMI: body-mass index; PUFA: polyunsaturated fatty acids 1 data sources for risk factor exposures from nationally representative surveys and/or epidemiological studies conducted in Brazil. 2 or dietary risks, a population SD of 10% of the mean was utilized, while for metabolic risks, the population SD of the optimal distribution was determined using a regression evaluating the mean to SD relationship of the corresponding risk factor [23]. 3 Cardiometabolic diseases with convincing or probable evidence of an etiologic association with risk factors of interest [9]. 4 High sodium intake was included as a risk factor for aortic aneurism, rheumatic heart disease and inflammatory heart disease based on evidence on the influence of elevated blood pressure on mortality from (not onset or incidence of) these outcomes. In other words, higher BP increases risk of death once these conditions have occurred 5 as a replacement for saturated fat. metabolic risk factors and dietary factors with harmful effects, optimal levels were defined as the exposure levels associated with the lowest level of harm seen in observational, and when relevant randomized trials as long as the selected level was also observed at the population level [29]. The standard deviation of the optimal levels was determined using the mean to SD regression co-efficient [8,15,23].

Estimating Mortality Attributable to Risk Factors
The population-attributable fraction PAF was estimated as: RRðxÞPðxÞdx where x is the exposure level, P(x) is the actual distribution of exposure in the population, P 0 (x) is the optimal exposure distribution; RR(x) is the relative risk of mortality at exposure level x, and m is the maximum exposure level. We calculated the number disease-specific deaths attributable to a risk factor by multiplying its PAF by total disease-specific mortality.
Due to joint distributions, multicausality, interaction, and because the effects of some risk factors are partly mediated through other risk factors, the number of deaths attributable to different risk factors cannot be summed. For example, the burden due to high SBP is partly mediated by high sodium intake and low intakes of fruits and vegetables; whereas much of the burden due to high BMI is mediated by high SBP, cholesterol, and glucose. Thus, the numbers of deaths attributable to each risk factor should be considered the total numbers of deaths due to this factor, including its upstream determinants and downstream mediators.

Uncertainty and Sensitivity Analyses
We quantified statistical uncertainty using probabilistic simulation analyses. Using the Monte Carlo simulations, we drew 1000 observations (unbiased random samples) from the normal distribution of exposure, optimal intake levels, and mortality, as well as a log-normal distribution of disease-specific RRs. We reported 95% UI for deaths related to a risk factor based on the 2.5 th and 97.5 th percentile of the resulting distributions of attributable deaths. These UI include all sources of uncertainty from each component of the analysis, including sampling error and uncertainties associated with model parameters [31]. In sensitivity analyses, we evaluated different optimal levels for each metabolic risk factor. All analyses were conducted using R version 2.15.0.

Dietary consumption
In 2010, the mean consumption of all evaluated healthful foods by Brazilian adults was below optimal levels, while intakes of unhealthful dietary factors were higher than optimal ( Table 2). The largest differences between mean intakes and optimal levels were observed for fruits, whole grains, and nuts, with mean national intakes only between 8% to 30% of optimal. Among unhealthy dietary factors, mean trans-fat intake exceeded the optimal level by 200%, while sodium consumption was over 100% greater than the WHO recommendation of 2.0g/ day. In contrast, intakes of fish and vegetables/beans were each closer to optimal levels. Compared to middle-aged and older adults (45y+), younger adults showed lower consumption of healthful foods such as fruits, vegetables and fish, and higher consumption of sugar-sweetened beverages, red and processed meats. Similarly, women exhibited generally healthier dietary habits than men ( Table 2).
The impact on suboptimal risk factors on cardiometabolic mortality     Figs 3 and 4). The second leading risk factor for early death was high SBP, followed by high BMI, which was the fourth leading risk factor for mortality in all age groups. The relative greater impact of high BMI on premature mortality was mostly due to suboptimal BMI among adults between 45 and 69 years of age.

Sensitivity Analysis
In sensitivity analysis, increasing the optimal level of high SBP from 115 to 120 mmHg decreased the number of SBP-attributable cardiometabolic deaths by about 9% (S1 Table).
Lowering the optimal levels of BMI (from 23 to 21 kg/m 2 ), fasting plasma glucose (from 5.3 to 4.9 mmol/l), and total cholesterol (from 4.1 to 3.9 mmol/l) increased their respective attributable mortality by 20%-30%. Due to joint distributions, multicausality, interaction, and because the effects of some risk factors are partly mediated through other risk factors, the number of deaths attributable to different risk factors cannot be summed. For example, part of the burden due to high BP is due to high sodium intake and low intakes of fruits and vegetables; whereas much of the burden due to high BMI is mediated by high BP, cholesterol, and glucose. Thus, the numbers of deaths attributable to each risk factor should be considered the total numbers of deaths due to this factor, including its upstream determinants and downstream mediators. 2 Total cardiometabolic deaths due to CHD (ICD-10 codes I20-I25), ischemic stroke (I63, I65-I67, I69.3), hemorrhagic/other non-ischemic stroke (I60-62, I69.0-2), hypertensive heart disease (I11-I13), aortic aneurysm (I71), rheumatic heart disease (I01, I02.0, I05-I09), inflammatory heart disease (I33, I42), other CVDs, and diabetes mellitus (E10-E14). 3. Burden from suboptimal diet was calculated by computing the combined population attributable fraction for dietary risk factors assuming that the contribution of each component is multiplicative. 4 Based on the number of people in the same stratum of the population (e.g. men, 25-44y etc). 5 Based on the total number of deaths in the same stratum of the population (e.g. men, 25-44y etc). 6 only mediated effects through blood pressure. doi:10.1371/journal.pone.0151503.t004 The Impact of Diet and Metabolic Risk Factors on Mortality in Brazil

Discussion
We have used up-to-date data, consistent and comparable methodology to estimate the impact of 11 dietary and 4 metabolic risk factors on cardiometabolic mortality in Brazil in 2010. Suboptimal diet and high SBP were the largest contributors to cardiometabolic death in Brazil. Among individual dietary factors, low intakes of fruits and whole grains and high intakes of sodium were the major risk factors for cardiometabolic death. Our dietary consumption estimates are consistent with recent studies reporting lower household availability and mean intakes of fruits, rice, beans, nuts, and fish, and relatively high consumption of processed meat, refined sugar, sugar sweetened beverages, and frozen meals in Brazil [2,32,33]. Previous studies reported a more than doubling in consumption of highly processed foods such as sausage, frozen meals and soft drinks in the past 3 decades, while energy from traditional foods such as rice and beans decreased between 10-20% in the same period [3]. Overall, our findings show the gap between current and optimal consumption was highest among younger adults (<45y). If these poor habits continue as these younger Brazilians age, the burden of cardiometabolic deaths due to suboptimal diet is likely to increase dramatically over time.
Although reductions in smoking and improved health care access have contributed to a nearly 30% reduction in CVD mortality between 1996 and 2007 in Brazil [5], our results The Impact of Diet and Metabolic Risk Factors on Mortality in Brazil indicate that the population burdens of major metabolic risk factors remain high, especially as it relates to premature mortality. About 60% of premature cardiometabolic deaths were attributable to high SBP, while 23% were due to high fasting plasma glucose, which suggests the need for appropriate measures to reduce the prevalence of these important metabolic risk factors in the country. Recent reports from the Longitudinal Study of Adult Health, a cohort including 15,000 Brazilian civil servants aged 35-74 years, found that about 50% of participants with diabetes had not been previously diagnosed or received treatment [34]; nearly 1 of 5 participants with measured hypertension had not been previously diagnosed; and about 47% of participants with measured hypertension had not received treatment for this condition [35]. In our investigation, the large burdens of BP highlight the relevance of excess sodium consumption among Brazilian adults and the potential population benefits of sodium reduction programs. Better BP screening and control with medications, and lifestyle treatment of those with prevalent diabetes and pre-diabetes, are also strongly indicated.
The overall adverse metabolic profile may be explained in part by the high prevalence of overweight and obesity, which was the third leading factor for premature cardiometabolic mortality in the country. Currently nearly 50% of Brazilian adults are overweight and 15% are obese [16,36]. Importantly, obesity rates in the past four decades have been greater among  individuals in lower income groups [37]. Excess adiposity increases blood pressure, dyslipidemia, as well as risk of diabetes, IHD, and stroke [38]. It has been well established that substantial reduction in cardiometabolic risk through adherence to a healthy lifestyle including healthy diet and regular exercise. Our findings emphasize the need to improve diet through increasing consumption of healthy foods including fruits, whole grains nuts and beans, and lowering intakes of unhealthy food components such sodium and processed meats, which will help reduce BMI and blood pressure levels in Brazil. Brazil has had success at implementing large-scale measures that have improved public health. For example, national immunization programs for yellow fever, small pox and poliomyelitis have nearly eradicated these conditions [39,40]. In their place, cardiometabolic diseases including CHD, stroke, and diabetes are now the leading causes of mortality and the greatest Due to joint distributions, multicausality, interaction, and because the effects of some risk factors are partly mediated through other risk factors, the number of deaths attributable to different risk factors cannot be summed. For example, part of the burden due to high BP is due to high sodium intake and low intakes of fruits and vegetables; whereas much of the burden due to high BMI is mediated by high BP, cholesterol, and glucose. Thus, the numbers of deaths attributable to each risk factor should be considered the total numbers of deaths due to this factor, including its upstream determinants and downstream mediators. 2 Total cardiometabolic deaths due to CHD (ICD-10 codes I20-I25), ischemic stroke (I63, I65-I67, I69.3), hemorrhagic/other non-ischemic stroke (I60-62, I69.0-2), hypertensive heart disease (I11-I13), aortic aneurysm (I71), rheumatic heart disease (I01, I02.0, I05-I09), inflammatory heart disease (I33, I42), other CVDs, and diabetes mellitus (E10-E14). 3. Burden from suboptimal diet was calculated by computing the combined population attributable fraction for dietary risk factors assuming that the contribution of each component is multiplicative. 4 Based on the number of people in the same stratum of the population (e.g. men, 25-44y etc) 5 Based on the total number of deaths in the same stratum of the population (e.g. men, 25-44y etc). 6 only mediated effects through blood pressure.
doi:10.1371/journal.pone.0151503.t005 public health challenges. Launched by the Brazilian Ministry of Health in 2011, the Strategic Action Plan to Tackle non-Communicable Chronic Diseases set 12 national goals for 2022, including a 2% yearly reduction in premature mortality due to non-Communicable Chronic Diseases, as well as improvements to national levels of three dietary targets (fruits, leafy vegetables, sodium) and one metabolic risk factor (BMI), in addition to increasing leisure-time physical activity [41]. Our data suggest that substantial reduction in premature mortality could be achieved by expanding the national goals to include other dietary factors such as whole grains, nuts, beans, and fish, as well as lowering processed meats and trans-fats. Our findings also emphasize the need for additional concerted efforts to improve risk factors among young adults and males, especially. Our study has several strengths. We used consistent and comparable methods to estimate the impact of key dietary and metabolic risk factors on cardiometabolic mortality in Brazil by age and sex, informing national priorities for prevention strategies. We used up-to-date evidence on effect sizes of diet-disease associations based on meta-analyses. We evaluated sensitivity of our findings to alternative optimal levels of dietary and metabolic risk factors. We accounted for potential age-and sex-differences in the current levels of risk factors, the relative magnitude of their effects cardiometabolic mortality, as well as cause-specific mortality. The Impact of Diet and Metabolic Risk Factors on Mortality in Brazil Potential limitations should be considered. Although we used RRs for associations of dietary and metabolic risk factors from meta-analyses of observational studies adjusting for relevant confounders, residual confounding may be present. To evaluate the magnitude of potential bias on measures of associations, particularly due to residual confounding by dietary factors, a sensitivity analysis was previously performed comparing GBD estimates to RRs from randomized controlled feeding trials evaluating effects of overall dietary patterns on CVD risk factors (DASH, DASH-sodium, OmniHeart) [27]. There was no substantial differences between effect size estimates used in this analysis compared to those of feeding trials, which indicates that our RRs were not substantially affected by correlation of dietary factors [27]. We used the joint population attributable fraction to estimate the number of cardiometabolic deaths attributable to the combined effect of dietary factors and assumed the intakes of individual dietary factors were independent within each age-sex stratum. This could overestimate the cardiometabolic burden of suboptimal diet. On the other hand, the cardiometabolic burden due to suboptimal diet might be underestimated due to measurement error and over-adjustment in the source studies for potential mediators (e.g., blood pressure, BMI, blood lipids). Finally, we focused on cardiometabolic mortality, and dietary and metabolic risk factors could increase the risk of The Impact of Diet and Metabolic Risk Factors on Mortality in Brazil other diseases, such as cancer. Thus, our estimates are a conservative estimate of the overall burden of disease due to these risk factors in Brazil.
In conclusion, our estimates of the impact of suboptimal dietary and metabolic risk factors on cardiometabolic mortality in Brazil in 2010 indicate that the major risk factors were suboptimal diet, particularly low fruits and whole grains and high sodium, high SBP, high fasting plasma glucose and high BMI. Our findings inform priorities for policy measures to reduce CVD and diabetes mortality in Brazilian adults.
Supporting Information S1 Table. Cardiometabolic deaths (95% UI) attributable to suboptimal risk factors in Brazil in 2010, estimated using lower reference levels. (DOCX)