点击显示 收起
1 From the Department of Society, Human Development and Health, Harvard School of Public Health, Boston, MA (SVS), and the Department of Social Medicine, University of Bristol, Bristol, United Kingdom (GDS)
2 SVS is supported by the National Institutes of Health Career Development Award (NHLBI 1 K25 HL081275-01). 3 Address reprint requests and correspondence to SV Subramanian, 677 Huntington Avenue, KRESGE Building, 7th floor, Boston, MA 02115. E-mail: svsubram{at}hsph.harvard.edu.
ABSTRACT
Background: Little systematic evidence exists for the relation between socioeconomic position and nutritional status in countries experiencing the simultaneous presence of under- and overnutrition.
Objective: We investigated the socioeconomic distribution of nutritional status in India and whether state-level macroeconomic factors modify the relation between socioeconomic position and nutritional status.
Design: Our analysis was based on a nationally representative sample of 77 220 women from India, with multiple categories of body mass index (BMI; in kg/m2) as the outcome, namely, <18.5 (underweight), 23–24.9 (pre-overweight), 25–29.9 (overweight), or 30 (obese), with 18.5–22.9 as the reference category.
Results: In adjusted models, being underweight was inversely related to socioeconomic position, whereas socioeconomic position was positively related to being pre-overweight, overweight, and obese, and the socioeconomic gradient was most marked for obesity. State-level measures of affluence did not modify the positive association between socioeconomic position and categories of overweight. The risk of underweight was lower in affluent states, but this was seen mainly in women of high socioeconomic position.
Conclusions: Undernutrition and overnutrition are epidemics of the impoverished and the affluent, respectively, in India, and this association is consistent at the individual and ecologic levels. Policies should focus on the complex patterns of social distribution of both under- and overnutrition in the Indian context.
Key Words: Nutritional transition • underweight • overweight • socioeconomic position • India
INTRODUCTION
Fifteen years ago, an exhaustive review of the evidence for the relation between socioeconomic position and obesity in adult women concluded that socioeconomic circumstances are consistently and positively associated with obesity in developing countries (1). The positive relation between obesity and socioeconomic position in developing countries stood in sharp contrast with the inverse association observed in developed countries, where the prevalence of obesity was higher among women from low socioeconomic groups (1). When the evidence was reviewed 15 y later, it was concluded that obesity cannot be considered only as a disease of materially advantaged groups and that the burden of obesity generally shifts toward poorer groups as countries improve their level of economic development (1, 2). Developing countries, meanwhile, also have a substantial prevalence of undernutrition (3), and it is well known that chronic energy deficiency is a risk factor for adult low productivity, morbidity, and mortality (4-9), with chronic undernutrition among women additionally being a major risk factor for adverse birth outcomes for their children (10). The increasing evidence for early developmental origins of adult disease and links between both maternal undernutrition and overnutrition with adverse long-term consequences in terms of obesity, type 2 diabetes, and cardiovascular disease among their offspring (11-13) necessitate population-based assessments of the patterns, distribution, and determinants of undernutrition and overnutrition among women of childbearing age. Using body mass index (BMI) as a marker for nutritional status, we ascertained the shape of the relation between individual socioeconomic position and nutritional status among women in India and the extent to which state-level differences in macroeconomic factors modify the association between individual socioeconomic position and nutritional status.
SUBJECTS AND METHODS
Data
The analyses are based on the representative, cross-sectional 1998–1999 Indian National Family Health Survey of 90 303 women in 26 Indian states (14). The survey covering various demographic and health aspects of women aged between 15 and 49 y was conducted in 1 of the 18 Indian languages in the respondents' homes and had high response rates (14). Women were geo-coded to the primary sampling unit, district, and state to which they belonged. The primary sampling units (hereafter referred to as local areas) were villages or groups of villages in rural areas and wards or municipal localities in urban areas. After we restricted our sample to complete cases for the outcome and predictors considered for the analysis, to women who were not pregnant, and to women who were not attending school, the final analytic sample consisted of 77 220 women. Descriptive characteristics of the sample for variables considered for the study, tabulated across 7 categories of BMI, are shown in Table 1.
View this table:
TABLE 1. Descriptive information for the 1998–1999 Indian National Family Health Survey sample showing the sample size and the distribution of women across 7 categories of BMI by different predictor variables
Main outcome measure
We used BMI, calculated as weight in kilograms divided by height in meters squared (kg/m2), as the outcome for this study. Weight was measured by using a solar-powered scale with an accuracy of ±100 g, and height was measured with an adjustable wooden measuring board that is designed to provide accurate measurements (to the nearest 0.1 cm) in the context of a developing-country field situation (15). According to World Health Organization conventions (16), the following BMI cutoffs were adopted: <16 (severe underweight), 16–16.9 (moderate underweight), 17.0–18.49 (mild underweight), 18.5–22.9 (normal weight), 23–24.9 (which we refer to as "at risk of overweight" or "pre-overweight"), 25–29.9 (overweight), and 30 (obese). Given the identification of a BMI of 23 as a public health cutoff for risk of obesity in Asian populations (16) and the emerging evidence suggesting that lower cutoffs are appropriate for populations from the Indian subcontinent (17, 18), we narrowed the normal BMI range of 18.5–24.9 to 18.5–22.9, thus identifying 3 grades of under- and overnutrition (Figure 1). Of all women, 47% were in the normal BMI range. Six percent and 8% of the women were severely and moderately underweight, and 19% were mildly underweight. Obesity prevalence was 3%, whereas 10% and 9% were overweight or at risk of overweight, respectively.
FIGURE 1.. Distribution of the population of adult women in India (n = 77 220) across 7 different strata of body mass index (in kg/m2).
Indicators of individual socioeconomic position
We considered 5 measures of socioeconomic position: standard of living, caste, education, occupation, and living environment. Standard of living was defined in terms of household assets and material possessions. and these have been shown to be reliable and valid measures of household material well-being (19). Each woman was assigned a standard-of-living score that was based on a linear combination of the scores for different items that were recorded for the household in which the woman resided and that were weighted according to a proportionate possession weighting procedure (20, 21). The weighted scores were divided into quintiles for the analytic models.
Caste was based on the women's self-identification as belonging to scheduled caste, scheduled tribe, other backward class, other caste, or no caste group. Scheduled tribe and scheduled caste are the most socially disadvantaged groups. Scheduled caste consists of castes that are lowest in the traditional Hindu caste hierarchy (22) and as a consequence that experience intense social and economic segregation and disadvantage. Scheduled tribes comprise 700 tribes who tend to be geographically isolated with limited economic and social interaction with the rest of the population. "Other backward class" is a diverse collection of intermediate castes that were considered low in the traditional caste hierarchy but are clearly above scheduled castes. "Other caste" is thus a default residual group that enjoys higher status in the caste hierarchy. We classified groups for whom caste may not always be applicable (eg, Muslims, Christians, or Buddhists) and participants who did not report any caste affiliation in the survey as "no caste."
Women's educational status was measured as years of schooling. We adopted cutoffs for years of schooling that were based on typical education benchmarks: 0 y (illiterate), 1–5 y (primary), 6–8 y (secondary), 9–12 y (higher), 13–15 y (college), and >15 y (postgraduate).
Women's current occupation was defined as being currently engaged in nonmanual work (eg, professional and managerial positions, clerical or sales, or generally employed in the service sector), skilled or unskilled manual work (including paid household or domestic work), or agricultural work either as an employee or as an owner or not currently participating in the labor force (including those not seeking work, such as homemakers).
Living environment was characterized according to whether the household in which the women resided was located in a large city (population 1 million), small city (population 100 000–1 million), town (population 100 000), or village or rural area. Covariates for this study included individual age, religion, tobacco and alcohol use, parity, self-reported current morbidities associated with asthma and malaria, and whether the respondent was receiving treatment for tuberculosis.
Macroeconomic indicators
We considered 2 state-level macroeconomic exposures: 1) the 1997–1998 per capita net state domestic product [PCSDP; in Indian Rupees (INR)], and 2) the proportion of the population living below the poverty line in 1999–2000. These data were obtained from the 2001 National Human Development Report published by the Government of India (see Tables 2. 1 on page 146 and 2.21 on page 166 in reference 23). Data on PCSDP were not available for Mizoram; therefore, we imputed the overall Indian average to Mizoram. For the states of Sikkim and Nagaland, we used the 1993–1994 PCSDP for the missing data for 1997–1998 (23). The all-India PCSDP in 1998–1999 was 2840 INR [with a range of 1126 (Bihar) to 6478 (Delhi) INR], with an SD of 1323 INR. A higher PCSDP is typically associated with higher levels of economic development, and this measure is the subnational equivalent of the measure per capita gross national product often used in international comparative studies. In 1999–2000, 26% of the Indian population had income levels below the official poverty line, with substantial statewide variation from a low of 3.5% (Jammu and Kashmir) to a high of 47% (Orissa), with an SD of 12% (23). The 2 macroeconomic measures had a negative correlation of –0.56 (P = 0.003).
View this table:
TABLE 2. Age-adjusted and mutually adjusted odds ratios and 95% CIs for indicators of socioeconomic position from the fixed part of an unordered multinomial multivariate model [using a BMI (in kg/m2) between 18.5 and 22.9 as the reference] for underweight and 3 grades of overweight, with adjustment for random effects associated with primary sampling units, districts, and states1
Statistical analysis
Given the multilevel structure of the sample with an explicit interest in modeling the effects of state-level exposures and with the outcome consisting of multiple categories, a multilevel multinomial modeling approach was adopted (24-26). Formally, yijkl is the categorical outcome with t categories for woman i in local area j in district k and state l. We denote the probability of being in category s as follows:
RESULTS
The age-adjusted and mutually adjusted odds ratios (ORs) and 95% CIs for the 4 indicators of socioeconomic position and urban-rural status are shown in Table 2. We provide the mutually adjusted ORs and 95% CIs for the covariates in Supplemental Table 1 (see the current issue online at www.ajcn.org).
BMI < 18.5 (underweight)
In the adjusted models, the prevalence of underweight in women in the bottom quintile of standard of living was substantially greater than in those in the top quintile (OR: 1.96, 95% CI: 1.83, 2.11). The risk of being underweight increased systematically with decreases in standard of living. Women in the scheduled caste and other backward class groups were more likely to be underweight than women in the other caste group, whereas differentials for the remaining groups were small. Education was also associated with the risk of being underweight in a graded fashion. Compared with those not in the labor force, women engaged in agricultural or manual work were more likely to be underweight, whereas those in nonmanual jobs had a decreased risk of being underweight. Women living in rural areas were more likely to be underweight than were women living in large cities. No substantial differences in the risk of being underweight were observed between women in large cities and those in small cities or towns. The adjusted effect estimates for each of the 5 indicators of socioeconomic position were substantially attenuated when compared with effect sizes obtained from models adjusted for age only (Table 2).
BMI 23–24.9 (pre-overweight), 25–29.9 (overweight), and 30 (obese)
The associations of the pre-overweight, overweight, and obese categories with the 5 individual measures of socioeconomic position were in a similar direction and became more marked for obesity. For pre-overweight, a more than 2-fold gradient existed across standard-of-living index quintiles; for obesity, the gradient was 7-fold; and the magnitude of the gradient for overweight was intermediate between these. Education, caste, occupation, and urban or rural residence showed patterns similar to those observed for the standard-of-living index, with greater differentials for obesity than for pre-overweight and with overweight showing intermediate-magnitude gradients. The effect of individual socioeconomic position indicators on the 3 grades of overnutrition was considerably attenuated in adjusted models compared with the results of the age-adjusted models, but clear and substantial effects remained (Table 2).
Effect of macroeconomic factors on overweight, obesity, and underweight
With adjustment for only age, a 1-SD increase in PCSDP increased the OR for overweight and obesity by 1.76 (95% CI: 1.40, 2.21) and 1.79 (95% CI: 1.43, 2.24), respectively. Additional adjustment for individual covariates and measures of socioeconomic position attenuated the ORs for overweight and obesity to 1.27 (95% CI: 1.09, 1.48) and 1.34 (95% CI: 1.14, 1.57), respectively. Conversely, the ORs for overweight and obesity decreased by 0.78 (95% CI: 0.69, 0.89) and 0.73 (95% CI: 0.64, 0.83), respectively, for a 1-SD increase in state poverty in adjusted models. In models adjusted for age only, the effect of state poverty on overweight and obesity was considerably stronger. The 2 state-level macroeconomic exposures, however, were not predictive of the risk of being underweight.
Plotted in Figure 2 are the predicted probabilities of being obese or underweight by PCSDP for the household quintiles of standard-of-living index, derived from additional tests of cross-level interaction between state-level macroeconomic exposures and household quintiles of standard-of-living index. The patterns associated with state poverty were similar to those observed for PCSDP, and the interaction patterns observed for overweight were similar to those observed for obesity. There is little evidence that PCSDP, qualitatively or quantitatively, modifies the relation between socioeconomic position and BMI, as has been suggested in cross-national studies (2, 28, 29). Thus, increases in the PCSDP seem to increase the risk of being obese for women in all quintiles of the standard-of-living index, except for women in the lowest quintile, for whom the risk of being obese was almost similar at low and high levels of PCSDP. Meanwhile, the PCSDP does seem to modify the relation between individual standard-of-living index and the risk of being underweight, with higher levels of PCSDP increasing the risk of being underweight for women in the lowest quintile of standard of living and decreasing the risk for those in the highest quintile.
FIGURE 2.. Plots of predicted probabilities of being (A) obese or (B) underweight by per capita net state domestic product for quintiles of the household standard-of-living index.
DISCUSSION
Our findings relate to women aged 15–49 y but have the advantage of being derived from the most recent representative population samples for all of India. There is a clear socioeconomic distribution underlying patterns of nutritional status, with women in low socioeconomic position experiencing a greater risk of underweight, and those in high socioeconomic position experiencing the greatest risk of being pre-overweight, overweight, and obese. This pattern is consistent with prior studies showing similar social patterning of nutritional status in women and men (30-32). The observed socioeconomic gradients in nutritional status provide clues to the factors that may explain this pattern.
First, the standard-of-living index, which is directly related to the amount of disposable household income available for food, was the single measure most strongly associated with undernutrition and overnutrition. This suggests, perhaps unsurprisingly, that higher expenditure on food is related to greater weight. Second, manual and agricultural work, which are crude indicators of physical activity, were related to both undernutrition and overnutrition in the expected direction, with agricultural and manual workers having a higher prevalence of underweight and a lower prevalence of overweight. A higher prevalence of obesity for women in more favorable social circumstances, besides being attributed to the ability to obtain more-than-adequate food supplies, has also been postulated to be linked to cultural norms that may favor fat body shapes (1). Motivated by this observation made >15 y ago (1), we calibrated an interaction effect between individual standard-of-living index and women's age for risk of being obese to test whether younger women with high socioeconomic position had a lowered risk of being obese. We found no evidence that the risk of obesity might be lower among younger women in the top quintile of standard-of-living index. The risk of being obese is highest for the better-off women at every age. It seems that the higher prevalence of overnutrition in higher socioeconomic groups can partially be explained by the possible indifference to body shapes, even among the younger age groups. Data from India also show that higher-income groups consume a diet containing 32% of energy from fat, compared with 17% in lower-income groups (33). These differences may contribute to explaining the strong observed association between high socioeconomic position and overweight. We should note, however, that given the cross-sectional nature of the present study, it is impossible to directly assess how the relation between socioeconomic position and BMI may change over time (34).
The positive association between socioeconomic position and nutritional status was observed regardless of the state levels of macroeconomic development, which were measured through either PCSDP or poverty concentration. It has been suggested that the burden of obesity shifts toward low socioeconomic groups as the country's per capita gross national product (PCGNP) increases, with the crossover in the socioeconomic distribution occurring at a PCGNP US$2500 (2). This cutoff was not adjusted for purchasing power parity, which is a method of measuring the relative purchasing power of different countries' currencies over the same types of goods and services (2). According to the World Bank (35), India with its PCGNP of US$440 is a "low-income country" (ie, PCGNP US$760), and adjusted for purchasing power parity, the Indian PCGNP is US$2060. Consequently, at the national level, India has yet to reach the stage of economic development at which the burden of overnutrition shifts toward lower socioeconomic groups, and our finding that it is women with higher socioeconomic position who are at the greatest risk of being overweight is consistent with this perspective. Notably, however, there was a 6-fold difference in the PCSDP within India, which provides an opportunity to test whether in economically better-off states the beginnings of a crossover of the burden of obesity to lower socioeconomic groups is detectable. Our findings did not find support for this hypothesis. Determinants of the relation between overall economic development and socioeconomic gradients in obesity may be ones that differ systematically between countries but not between states in India.
A key (data) limitation of our study relates to the use of BMI as the only measure of overweight. Because of differences in body frame sizes and body proportions, at any given BMI, Indians may have a higher proportion of body fat and thus an elevated risk of some of the long-term consequences of obesity, in particular, diabetes and cardiovascular disease (36, 37). Thus, a given BMI may confer a greater risk of obesity-related diseases such as diabetes and cardiovascular disease among Indians than in the populations in which the BMI standards were initially developed (38). Waist circumference (18) and waist-to-hip ratios (39), consequently, have been suggested as better markers of obesity. Recent studies have shown that South Asians have the poorest correlation between waist circumference and BMI when comparing them against Europeans, Chinese, and Aboriginal persons, although the correlation is still substantial (17). Furthermore, at any given waist circumference, South Asians also have more metabolic abnormalities than do Europeans (17, 18). Thus, South Asians not only tend to gain more abdominal obesity, but their cutoff for elevated waist circumference should also be lowered (18). Indeed, the relevance of BMI as a measure of obesity has been called into question on the basis of an international case-control study (39). However, note that BMI is a strong predictor of waist circumference (40), and these critiques of BMI, although central to analyses dealing exclusively with obesity, are not particularly applicable to the present study, in which our focus was on the entire nutritional spectrum, including underweight. However, it is possible that the social patterning of BMI will not adequately reflect the social patterning of body fat, such that persons in less favored socioeconomic groups may have a higher proportion of body fat at a given BMI, as has been shown in populations from developed countries (41).
Although these limitations inhibit the ability of our study to estimate the true burden of chronic diseases and mortality associated with overnutrition in Indian women, the BMI of women may have an intrinsic relevance because of the impact of women's health on the health of their offspring. Women of low prepregnancy BMIs have babies with lower birth weights (10), and the evidence that low birth weight and low maternal BMI are associated with an increased risk of adulthood chronic disease among offspring is consistent and universal (42). As Osmani and Sen (13) point out, sex inequality can contribute to poor health within a society through such interuterine mechanisms. Indeed, the coexistence of undernutrition and overnutrition in India also means that the consequences of maternal obesity—high birth weight and higher risk of diabetes among offspring—will increasingly be seen (42).
Undernutrition and overnutrition remain epidemics of the impoverished and the affluent, respectively, in India. This is true at the individual level and at the ecologic level. The nutritional status of individuals and societies, at a given point in time, is likely to reflect the cumulative synergy between physiologic endowments and the social environment. Detecting the particular socioeconomic distribution of nutritional status is likely to provide an important evidence base for developing and targeting interventions to counter the twin problems of undernutrition and overnutrition. The positive association between nutritional status and socioeconomic position, a likely characteristic at early stages of socioeconomic and nutritional transition, holds promise for successful mitigation of the twin problem compared with a situation in which undernutrition and overnutrition are both concentrated in the lower socioeconomic groups.
ACKNOWLEDGMENTS
We thank Shailen Nandy for preparing an earlier version of the data set used in this study. We acknowledge Macro International (www.measuredhs.com) for providing us access to the 1998–1999 Indian National Family Health Survey data.
SVS conceived the study, analyzed and interpreted the data, and wrote and edited the manuscript. GDS contributed to the conceptualization of the study, interpretation of the results, and editing of the manuscript. The authors had no conflicts of interest to declare.
REFERENCES