Risk factors for nutrition-related chronic disease among adults in Indonesia

Objective To conduct a secondary data analysis detailing the associations between sociodemographic and behavioral factors and nutrition-related chronic disease. Methods These analyses utilized 2014 data from the Indonesian Family Life Survey, a home-based survey that collected socioeconomic, dietary intake, physical activity, and biological data among adults. We explored four outcomes in relation to sociodemographic and behavioral determinants: 1) hypertension, 2) elevated high-sensitivity c-reactive protein (hs-CRP), and 3) central obesity, as these are critical metabolic determinants in the progression to cardiovascular disease, and 4) type 2 diabetes. Hypertension was defined as systolic blood pressure ≥140 mm or diastolic blood pressure ≥ 90mm or current use of antihypertensive medication. Elevated hs-CRP was defined as hs-CRP >3 mg/dL. Central obesity was defined as waist circumference ≥ 90 cm if male and waist circumference ≥ 80 cm if female, which are specific to South Asia. Type 2 diabetes was defined as glycated hemoglobin ≥ 6.5%. We employed separate gender-stratified multivariate logistic regression models to test the associations between sociodemographic and behavioral determinants and each nutrition-related chronic disease outcome. All analyses employed sampling weights, which account for the survey design. Results In 2014, about 30% of adults were hypertensive and one-fifth had elevated hs-CRP. Approximately 70% of women had central obesity and 11.6% of women and 8.9% of men had diabetes. Older-age was consistently associated with nutrition-related chronic disease and being overweight was associated with hypertension, elevated hs-CRP, and type 2 diabetes. Regularly consuming instant noodles (women) and soda (men) were associated with elevated hs-CRP and soda consumption was associated with central obesity among men. Conclusions Large segments of the adult population in Indonesia now have or are at risk for non-communicable disease. Our analyses provide preliminary empirical evidence that interventions that target healthful food intake (e.g. reduce the intake of ultra-processed foods) should be considered and that the reduction of overweight is critical for preventing chronic diseases in Indonesia.

Introduction outcomes in relation to sociodemographic and behavioral determinants: 1) hypertension, 2) elevated hs-CRP, and 3) central obesity, as these are critical metabolic determinants in the progression to CVD, and 4) type 2 diabetes. We believe this evidence will be useful in informing policies and programs that aim to reduce NCDs in Indonesia [21].

Survey design and study population
These analyses utilized the Indonesian Family Life Survey (IFLS), an ongoing longitudinal, home-based study that was initiated in 1993 [32]. There have been four subsequent survey rounds (1997,2000,2007,2014). The original, multi-stage sampling frame was based on households from 13 out of 27 provinces, which represented 83% of the Indonesian population in 1993 [32]. For these analyses, we utilized 2014 data from adults aged above 19 years, due to data availability (described below). In Indonesia, eight new provinces have been created since 1999, thus, in 2014, 24 of 34 provinces were represented.
Among the original 33,081 household members enrolled in the IFLS, about one-third, (11,040) were found in their original IFLS households in 2014, approximately 9,000 were found elsewhere and about 4,500 had died. The recontact rate (including deaths) in 2014 among individuals enrolled in 1993 was 76%. Over the course of the survey, 11,889 (54%) responded in all survey waves [32].

Survey questions & measurements
Data were collected on household-and individual-level characteristics, as well as diet, physical activity, and health. The household questionnaire was completed by the head of household and recorded information on household size, physical infrastructure, access to sanitation, area of residence (urban/rural), and food expenditures. While the topics covered in the individuallevel adult questionnaire were wide-ranging, these analyses utilized information on gender, age, educational attainment, marital status, employment, and smoking status. Adults were asked whether they ate each food type "in the last week" and as appropriate, the frequency of consumption (number of days). Self-reported consumption of dietary staples (e.g. rice, eggs, meat, green leafy vegetables, sweet potatoes) was recorded, and in 2014, consumption of some ultra-processed foods, including instant noodles, fast food, soft drinks, and fried snacks, were also recorded. Ultra-processed foods are typically "ready-to-consume" and are entirely or mostly made from industrial ingredients and additives, not foods [33]. In addition to being energy-dense, they are also characteristically high in fat, sugar, and/or salt (unlike rice, which would be considered energy-dense). Consumption of ultra-processed foods was of interest in these analyses as they have become more common worldwide [33], their availability and consumption increases as countries undergo their nutrition transition, and their consumption has been associated with poor chronic-disease related health [34][35][36][37] and weight gain in one experimental study [37]. Adults in this sample also self-reported any vigorous physical activity, moderate physical activity, and/or walking (during the last week) using a modified version of the International Survey on Physical Activities. Vigorous activity was defined as any activities that make you breathe much harder than normal (e.g. heavy lifting). Moderate activity was defined as any activities that make you breathe somewhat harder than normal (e.g. carrying light loads). Walking included walking from place to place, walking for recreation, and walking at work and at home.
All health measurements were collected by trained enumerators, using a rigorous research protocol, whereby measurements were supervised and subject to quality control procedures [32]. Height was measured to the nearest millimeter using a Seca plastic height board. Weight was measured to the nearest one-tenth of a kilogram using a Camry model EB1003 scale. Waist circumference (for adults aged � 40 years) was measured to the nearest millimeter with a tape measure. Blood pressure was taken three times, on alternate arms from a seated position, using an Omron meter, HEM-720. Large cuffs were available as needed.
In 2007 (IFLS wave 4), dried blood spots (DBS) were collected among a random sample of individuals from wave 1 of the IFLS (1993). Those same respondents were re-contacted in 2014 to continue to collect DBS. A finger prick was taken and blood drops drawn for measurement of hs-CRP and glycated hemoglobin (HbA 1c ). Prior to the finger prick, hand warmers were used to increase blood flow. The first drops of blood were used with the hemoglobin and secondary drops were put onto Whatman 903 Protein Saver Cards [32]. Cards were allowed to dry for at least 4 hours and stored with a desiccant to keep samples dry. Samples were kept cool, mailed back to headquarters in Yogyakarta, and then, stored at-40 Celsius, until assayed [38]. The assay used to measure hs-CRP in 2007 was no longer on the market when the most recent IFLS wave was fielded [38]; in 2014, the hs-CRP enzyme immunoassay kit was manufactured by Percipio Biosciences (Catalog Number 11190) [39]. The HbA 1c assay was based on a validated protocol, described by Hu and colleagues [40]. Validation samples, for both hs-CRP and HbA 1C , were provided by the University of Washington (Seattle, U.S.) and the University of Southern California/University of California Los Angeles' Center on Biodemography and Population Health (Los Angeles, U.S.) prepared additional blood spots that were used as controls for the assays [39]. Samples were analyzed at the University of Washington and in Indonesia. Using regression-based methods, DBS results were converted to plasma-equivalent values for hs-CRP and whole blood equivalent values for HbA 1c , based on repeated measurements of the validation samples [39]. Approximately, 6,300 adults aged � 19 years (~2,800 men and~3,500 women) had usable CRP or HbA 1c data and sample weights. Subjects with incomplete data and pregnant women were excluded from these analyses.

Statistical analysis
We explored associations between risk factors and four dependent variables: hypertension, elevated hs-CRP, central obesity, and type 2 diabetes (henceforth referred to as diabetes). Systolic and diastolic blood pressure were based on the average of three measurements. Hypertension was defined as systolic blood pressure �140 mm or diastolic blood pressure � 90 mm or current use of antihypertensive medication [41,42]. Elevated hs-CRP was defined as hs-CRP >3 mg/dL, based on increased risk for CVD [31]. Central obesity was defined as waist circumference � 90 cm if male and waist circumference � 80 cm if female, which are specific to populations in South Asia [43]. Clinical guidelines define diabetes as glycated hemoglobin (HbA 1C ) � 6.5% [44].
Prior literature, along with the availability of data in the IFLS, were used to identify the independent variables in regression models. Consumption of ultra-processed foods (versus not consuming), physical activity (versus no activity), smoking (versus not smoking) and urban residence (versus rural) were modeled as binary variables. Appropriate cutoffs were applied to create dichotomous or categorical variables for age, education, employment, family size, and food expenditures. A composite of household wealth was created using principal component analysis, using following variables: type of floor material, type of toilet, type of cooking fuel and ownership of assets including: land, livestock, vehicle(s), household appliances, furniture and utensils, jewelry, and monetary savings. Wealth was divided into quintiles, based on the distribution of the data. In models exploring hypertension, elevated hs-CRP, and diabetes as the outcome, overweight/obesity (henceforth referred to as overweight) was included as an independent variable and was defined as BMI � 25 kg/m 2 .
We first employed separate, gender-stratified univariate logistic regression models to investigate risk factors and their association with each outcome (S1-S4 Tables). Independent variables that were statistically significantly associated with the dependent variable in univariate models were then included in multivariable logistic regression models. Age, urban/rural residence, and wealth were included in the multivariable models irrespective of statistical significance.
In sensitivity analyses we defined overweight as BMI � 23 kg/m 2 , because a substantial proportion of Asian people have a higher risk of diabetes and CVD at lower BMIs [45] (S5-S7 Tables). We assessed multicollinearity in all models. Alpha was set to 0.05. All analyses employed sampling weights, which account both for sample attrition from 1993 to 2014, and account for the original sample design, which included an over-sample of urban areas; thus, this sample is representative of the 2014 Indonesian population in those provinces. Analyses were performed using Stata 15.1 (StataCorp LP, College Station, TX). Survey participants provided written informed consent. The data used for this study are retrospective and the authors did not have access to any identifying information. As such, ethical approval was not required.

Results
A majority of the sample (N = 15,648 women, N = 13,567 men) was middle-aged (30-59 years) and about half had a primary level of education (Table 1). Nearly all women were nonsmokers (96.3%), compared to 32.8% of men. About one-third of women were not working (37.1%), compared to 10.2% of men. Half of the adults resided in urban areas and slightly larger proportions resided in lower (versus higher) wealth households.

Central obesity
Older women and those with higher educational attainment had higher odds of central obesity (

Discussion
Large segments of the adult population in Indonesia now have or are at risk for NCDs. Key metabolic determinants in the progression to CVD, including hypertension, elevated hs-CRP,  and central obesity, as well as type 2 diabetes, were prevalent. About 20% of adults were hypertensive in 2014. At the same time, one-fifth of the adult population had elevated hs-CRP. Of particular concern were the very high prevalence of central obesity among women (~70%) and, compared to global estimates, a relatively high prevalence of diabetes among women: 11.5%. Older-age and being overweight were consistently associated with chronic diseaserelated health. Urban area residence (versus rural) was associated with elevated hs-CRP (women) and central obesity (women and men). Consuming instant noodles (women) and soda (men) were associated with elevated hs-CRP and soda consumption was associated with higher central obesity among men. On the contrary, vigorous physical activity was associated with lower odds of central obesity among men. Higher odds of hypertension were observed among older and overweight adults in this sample. Higher education was associated with lower odds of hypertension among women, as was being employed (versus non-employed) among both women and men. These findings are generally consistent with prior studies conducted in Indonesia [46][47][48][49][50][51] and Asia [52], which suggest social determinants play a role in the development of hypertension.
Being overweight, living in urban area, and an unhealthful diet were associated with higher odds of elevated hs-CRP, in this sample. These factors are all inter-related; urban-area residence results in an increased consumption of unhealthy foods and decreased physical activity, which are associated with higher overweight [36, [53][54][55]. In turn, being overweight is a wellestablished risk factor for elevated hs-CRP [56]. For example, in a seminal U.S.-based study, obese men were 2-times as likely and obese women 6-times as likely to have elevated CRP levels [56]. Relatively few studies have explored the determinants of elevated hs-CRP in lowerincome countries, but a strong association between obesity and hs-CRP has also been reported among adults in Jamaica [57] and among Filipinos [58]. Both hypertension and elevated hs-CRP were associated with CVD [59,60] and elevated hs-CRP is associated with higher diabetes risk [61].
Similar to prior studies in Asia [62][63][64], the prevalence of central obesity was very high among women (~70%), compared to only 24% among men. In multivariable models, age, education, and urban residence were associated with higher odds of central obesity, while  agriculture-based employment and smoking were associated with lower odds of central obesity. One other Indonesian study also suggested that physical labor was related to lower central obesity among older-aged adults [63] and agriculture-based employment was associated with lower odds of overweight among women across 40 lower-income countries [65]. Central obesity is associated with higher mortality risk [66] and higher CVD among Asians [67]. Notably, Asians tend to develop CVD and diabetes with a lower BMI than individuals in other regions, due in part to their propensity to store adipose tissue around their waist [45,68]. Therefore, it is especially important to consider the prevalence of central obesity in this population, which could suggest that the risk for NCDs among women is more widespread than previously thought [69,70]. Findings from these analyses also indicate larger portions of women (versus men) have diabetes; this is similar to the 2013 National Health Survey, which found 10.1% of women and 7.3% of men were diabetic [16,71]. Prevalence estimates from this sample are higher than estimates from the NCD Risk Factor Collaboration study for women, which reported that the global prevalence of diabetes was 7.9% [72]. High diabetes prevalence in Indonesia suggests that there may be Southeast Asia-specific risk factors to consider. In addition to the aforementioned risk of diabetes with lower BMIs, physiological responses to food could vary between populations. For example, the glucose response in Asians is higher than Caucasians in the U.S. [73]. In this sample, older age, being overweight, and higher education (among men) were associated with higher odds of diabetes. This is consistent with prior studies that show that age induces a decrease of insulin sensitivity and/or insulin resistance [74] and that age is associated with a higher odds of diabetes prevalence, using regression-based methods [52,75]. This is also consistent with 2007 data, which found that age and obesity were associated with diabetes in Indonesia [76]. In multivariable models, higher education has also previously been associated with diabetes in Asia [52].
Consuming instant noodles and soda were associated with elevated hs-CRP and central obesity. Some associations between an unhealthy diet and hs-CRP and central obesity, as well as the consistent associations between overweight and hypertension, elevated hs-CRP, and type 2 diabetes in these analyses, suggests that the nutrition transition has played a major role in these trends. There have been marked changes in Indonesia's food system and a reduction in energy expenditure. More calories are now available, with an estimated 20% coming from fat [70], in part due to the ubiquitous use of vegetable-based cooking oils, that are high in trans-and saturated fats. Higher caloric intake has been associated with weight gain and the consumption of trans fats, which has been associated with increased cardiometabolic risk (51) and insulin resistance (52). Individual-level data from this sample echo trends from macrolevel data, as about 60% of adults in this sample reported regularly consuming fried snacks. In addition, polished rice and refined wheat, which are high glycemic index foods, form the basis of many Indonesians' diets; consuming high glycemic index foods is associated with a 2-fold increased risk of diabetes [77]. In addition, by one account, Indonesians are consuming 15 grams of salt per day, which is about 3-times the recommended amount [78], and strongly related to hypertension risk [24,25].
Although there were not consistent associations between physical activity and NCD-related outcomes, vigorous physical activity was associated with lower odds of central obesity among men. Some prior studies from Indonesia have also reported no associations between physical activity and hypertension [50] and between activity and central obesity, in samples that pool men and women [79]. Contrary to findings reported in these analyses, results from a robust meta-analysis, which largely includes data from high-income countries, reports that physical activity is strongly related to hypertension [80]. Physical activity has also been associated with insulin sensitivity in Indonesia [76] and reduced incidence of diabetes [81][82][83] in prior studies. Overall, the literature and findings from this study suggest that interventions that promote healthful eating and physical activity, in order to reduce the prevalence of NCDs, warrant consideration in Indonesia. Notably, very few respondents reported taking medication for hypertension or diabetes. This could suggest that these adults are undiagnosed and/or do not have regular access to health care [84]. One prior study suggests that approximately 70% of all diabetes cases in Indonesia remain undiagnosed [85]. The Government has undertaken several WHO-recommended initiatives designed to strengthen the primary healthcare system, including implementing coverage and service delivery reforms under Jaminan Kesehatan Nasional, the national health insurance plan [86]. But current evidence suggests that prevention and management of diet-related NCDs remain a challenge [21,[86][87][88][89][90].
Study limitations warrant consideration. Due to our study design, we cannot establish temporality or rule out unmeasured or residual confounding. Diet and physical activity data are self-reported and only capture a limited picture of the consumption of ultra-processed foods and physical activities. Results for hs-CRP and HbA 1C are estimated using DBS, therefore, these results may not be comparable to those that that involve assay results from venous blood. In addition, we examine the risk factors for a single outcome in separate models, and this does not provide information on risk factors for co-morbidities (e.g. co-existing hypertension and diabetes).

Conclusions
We find a relatively high prevalence of nutrition-related chronic diseases among Indonesians. Moreover, overweight was consistently associated with hypertension, hs-CRP, and diabetes. These findings provide preliminary empirical evidence that interventions that target healthful food intake (e.g. reduce the intake of ultra-processed foods) and physical activity should be considered and that the reduction of overweight is critical for preventing chronic diseases in Indonesia. Future studies should investigate risk factors for co-morbidities.
Supporting information S1