Patterns, distribution, and determinants of under- and overnutrition:
a population-based study of women in India1⫺3
S V Subramanian and George Davey Smith
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 populationbased 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
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. Email: svsubram@hsph.harvard.edu.
Received January 20, 2006.
Accepted for publication April 20, 2006.
633
Am J Clin Nutr 2006;84:633– 40. Printed in USA. © 2006 American Society for Nutrition
Supplemental Material can be found at:
http://www.ajcn.org/cgi/content/full/84/3/633/DC1
Downloaded from www.ajcn.org by on September 24, 2009
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, 쏝18.5 (underweight), 23–24.9 (pre-overweight), 25–29.9 (overweight), or 욷30
(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.
Am J Clin Nutr
2006;84:633– 40.
634
SUBRAMANIAN AND DAVEY SMITH
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.
Main outcome measure
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 앒700 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
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 Ҁ0.56 (P ҃ 0.003).
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:
共s兲
␲ijkl
⫽ Pr共yijkl ⫽ s兲
(1)
In a multinomial logistic model, one of the outcome categories is
taken as the reference category, just as the category coded as 0 is
usually taken as the reference category in the more commonly
used binary response models. Using the BMI category of 18.5–
22.9 as the reference, we estimated a set of t Ҁ 1 logistic regressions for the underweight and 3 overweight categories in which
each of the categories was contrasted with the reference category.
Downloaded from www.ajcn.org by on September 24, 2009
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 앐100 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: 쏝16 (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 욷30
(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.
typical education benchmarks: 0 y (illiterate), 1–5 y (primary),
6 – 8 y (secondary), 9 –12 y (higher), 13–15 y (college), and 쏜15
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 욷 1 million), small city (population 100 000 –1
million), town (population 울 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.
635
SOCIOECONOMIC POSITION AND NUTRITION IN INDIA
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
BMI (kg/m2)
Variables
Sample size
쏝16
16 –16.9
17–18.49
n (%)
1,2
23–24.9
25–29.9
쏜30
n (%)
4958 (6.4)
12 284 (15.9)
15 339 (19.9)
14 124 (18.3)
12 535 (16.2)
10 115 (13.1)
7865 (10.2)
270 (5.4)
725 (5.9)
871 (5.7)
757 (5.4)
668 (5.3)
620 (6.1)
513 (6.5)
451 (9.1)
1168 (9.5)
1320 (8.6)
1045 (7.4)
818 (6.5)
673 (6.7)
525 (6.7)
1233 (24.9)
2895 (23.6)
3202 (20.9)
2521 (17.8)
1951 (15.6)
1500 (14.8)
1035 (13.2)
2744 (55.3)
6408 (52.2)
7464 (48.7)
6534 (46.3)
5531 (44.1)
4213 (41.7)
3208 (40.8)
180 (3.6)
641 (5.2)
1240 (8.1)
1381 (9.8)
1405 (11.2)
1116 (11.0)
914 (11.6)
76 (1.5)
396 (3.2)
1027 (6.7)
1481 (10.5)
1652 (13.2)
1539 (15.2)
1235 (15.7)
4 (0.1)
51 (0.4)
215 (1.4)
405 (2.9)
510 (4.1)
454 (4.5)
435 (5.5)
60 311 (78.1)
8809 (11.4)
4281 (5.6)
1880 (2.4)
1939 (2.5)
3687 (6.1)
508 (5.8)
113 (2.6)
52 (2.8)
64 (3.3)
5008 (8.3)
636 (7.2)
195 (4.6)
83 (4.4)
78 (4.0)
11 863 (19.7)
1532 (17.4)
547 (12.8)
181 (9.6)
214 (11.0)
27 861 (46.2)
4064 (46.1)
2381 (55.6)
731 (38.9)
1065 (54.9)
5103 (8.5)
828 (9.4)
469 (11.0)
254 (13.5)
223 (11.5)
5343 (8.9)
952 (10.8)
472 (11.0)
413 (22.0)
226 (11.7)
1446 (2.4)
289 (3.3)
104 (2.4)
166 (8.8)
69 (3.6)
13 055 (16.9)
9220 (11.9)
22 601 (29.3)
31 930 (41.4)
414 (0.5)
980 (7.5)
523 (5.7)
1369 (6.1)
1524 (4.8)
28 (6.8)
1300 (10.0)
738 (8.0)
1899 (8.4)
2007 (6.3)
56 (13.5)
2928 (22.4)
1889 (20.5)
4466 (19.8)
4948 (15.5)
106 (25.6)
6033 (46.2)
5028 (54.5)
10 694 (47.3)
14 158 (44.3)
189 (45.7)
873 (6.7)
598 (6.5)
1893 (8.4)
3498 (11.0)
15 (3.6)
776 (5.9)
373 (4.0)
1844 (8.2)
4396 (13.8)
17 (4.1)
165 (1.3)
71 (0.8)
436 (1.9)
1399 (4.4)
3 (0.7)
957 (1.2)
2840 (3.7)
12 265 (15.9)
9668 (12.5)
12 977 (16.8)
38 513 (49.9)
15 (1.6)
44 (1.5)
370 (3.0)
420 (4.3)
754 (5.8)
2821 (7.3)
22 (2.3)
86 (3.0)
541 (4.4)
614 (6.4)
992 (7.6)
3745 (9.7)
53 (5.5)
195 (6.9)
1475 (12.0)
1501 (15.5)
2421 (18.7)
8692 (22.6)
336 (35.1)
1060 (37.3)
5437 (44.3)
4502 (46.6)
6113 (47.1)
18 654 (48.4)
164 (17.1)
463 (16.3)
1653 (13.5)
1080 (11.2)
1169 (9.0)
2348 (6.1)
268 (28.0)
742 (26.1)
2134 (17.4)
1191 (12.3)
1219 (9.4)
1852 (4.8)
19 059 (24.7)
17 036 (22.1)
15 754 (20.4)
12 900 (16.7)
12 471 (16.1)
425 (2.2)
771 (4.5)
1018 (6.5)
1031 (8.0)
1179 (9.5)
695 (3.6)
1066 (6.3)
1324 (8.4)
1387 (10.8)
1528 (12.3)
1892 (9.9)
2668 (15.7)
3347 (21.2)
3052 (23.7)
3378 (27.1)
7712 (40.5)
8368 (49.1)
7922 (50.3)
6349 (49.2)
5751 (46.1)
2817 (14.8)
1828 (10.7)
1158 (7.4)
653 (5.1)
421 (3.4)
4123 (21.6)
1889 (11.1)
839 (5.3)
368 (2.9)
187 (1.5)
1395 (7.3)
446 (2.6)
146 (0.9)
60 (0.5)
27 (0.2)
8620 (11.1)
4995 (6.5)
10 799 (14.0)
52 806 (68.4)
251 (2.9)
225 (4.5)
442 (4.1)
3506 (6.6)
315 (3.7)
272 (5.4)
598 (5.5)
4815 (9.1)
845 (9.8)
592 (11.9)
1521 (14.1)
11 379 (21.5)
3314 (38.4)
2010 (40.2)
4696 (43.5)
26 082 (49.4)
1289 (15.0)
678 (13.6)
1363 (12.6)
3547 (6.7)
1917 (22.2)
910 (18.2)
1676 (15.5)
2903 (5.5)
689 (8.0)
308 (6.2)
503 (4.7)
574 (1.1)
48 160 (62.4)
4433 (5.7)
17 758 (23.0)
6869 (8.9)
2400 (5.0)
135 (3.0)
1425 (8.0)
464 (6.8)
3330 (6.9)
155 (3.5)
1883 (10.6)
632 (9.2)
8294 (17.2)
460 (10.4)
4203 (23.7)
1380 (20.1)
22 299 (46.3)
1857 (41.9)
8768 (49.4)
3178 (46.3)
4755 (9.9)
680 (15.3)
896 (5.0)
546 (7.9)
5470 (11.4)
885 (20.0)
506 (2.8)
545 (7.9)
1612 (3.3)
261 (5.9)
77 (0.4)
124 (1.8)
67 927 (88.0)
9293 (12.0)
3606 (5.3)
818 (8.8)
5148 (7.6)
852 (9.2)
12 377 (18.2)
1960 (21.1)
31 658 (46.6)
4444 (47.8)
6287 (9.3)
590 (6.3)
6883 (10.1)
523 (5.6)
1968 (2.9)
106 (1.1)
75 012 (97.1)
2208 (2.9)
4296 (5.7)
128 (5.8)
5855 (7.8)
145 (6.6)
13 920 (18.6)
417 (18.9)
34 827 (46.4)
1275 (57.7)
6734 (9.0)
143 (6.5)
7326 (9.8)
80 (3.6)
2054 (2.7)
20 (0.9)
75 139 (97.3)
2081 (2.7)
4228 (5.6)
196 (9.4)
5825 (7.8)
175 (8.4)
13 872 (18.5)
465 (22.3)
35 071 (46.7)
1031 (49.5)
6758 (9.0)
119 (5.7)
7326 (9.7)
80 (3.8)
2059 (2.7)
15 (0.7)
76 782 (99.4)
438 (0.6)
4345 (5.7)
79 (18.0)
5931 (7.7)
69 (15.8)
14 239 (18.5)
98 (22.4)
35 939 (46.8)
163 (37.2)
6863 (8.9)
14 (3.2)
7394 (9.6)
12 (2.7)
2071 (2.7)
3 (0.7)
74 076 (95.9)
3144 (4.1)
77 220 (100.0)
4147 (5.6)
277 (8.8)
4424 (5.7)
5675 (7.7)
325 (10.3)
6000 (7.8)
13 656 (18.4)
681 (21.7)
14 337 (18.6)
34 618 (46.7)
1484 (47.2)
36 102 (46.8)
6678 (9.0)
199 (6.3)
6877 (8.9)
7265 (9.8)
141 (4.5)
7406 (9.6)
2037 (2.7)
37 (1.2)
2074 (2.7)
Chi-square test for a cross-tabulation between each variable and the 7 categories of BMI: 1 P ҃ 0.001, 2 P ҃ 0.000.
99 (10.3)
250 (8.8)
655 (5.3)
360 (3.7)
309 (2.4)
401 (1.0)
Downloaded from www.ajcn.org by on September 24, 2009
Age (y)1
15–19
20–24
25–29
30–34
35–39
40–44
45–49
Religion1
Hindu
Muslim
Christian
Sikh
Other
Caste2
Scheduled caste
Scheduled tribe
Other backward class
Other caste
No caste
Education (y)1
쏜15
13–15
9–12
6–8
1–5
None
Household standard of
living1
Top quintile
Fourth quintile
Third quintile
Second quintile
Bottom quintile
Living environment1
Large city
Small city
Town
Rural area
Occupation1
Not working
Nonmanual work
Agricultural work
Manual work
Tobacco chewing1
No
Yes
Drink1
No
Yes
Tobacco smoking1
No
Yes
Tuberculosis treatment1
No
Yes
Malaria1
No
Yes
Total
18.5–22.9
636
SUBRAMANIAN AND DAVEY SMITH
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).
Then, a multilevel multinomial logistic regression model with
logit link was written as follows:
共s兲
共t兲
共s兲
log共␲ijkl
/␲ijkl
兲 ⫽ ␤共s兲X ⫹ 共ujkl
⫹ vkl共s兲 ⫹ f l共s兲兲
(2)
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:
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,
Downloaded from www.ajcn.org by on September 24, 2009
where s ҃ 1, . . . t Ҁ 1. A separate intercept and slope parameter
was estimated for the underweight and 3 overweight categories,
as indicated by the s superscripts. The notation ␤(s) represents the
fixed part of the model and is interpreted as the effect of a 1-unit
increase in X (the set of predictor variables) on the log odds of
being in category s (ie, the underweight or one of the 3 overweight categories) rather than category t (the normal category).
For presentation and discussion, we used exp(␤(s), which is the
effect of a 1-unit increase in X on the odds of being in category s
rather than category t. The terms inside the brackets in equation
2 represent the random effects associated with primary sampling
units, districts, and states, respectively, which are assumed to be
normally distributed with mean 0 and variances ␴u(s), ␴v(s), and
␴f(s). The random effects are specific to each of the contrasted
categories, as indicated by the s superscript, because different
unobserved factors at each level may affect each contrast. We
allow for the possible correlation in the random effects at each
level across different contrasts. Regression and variance parameters are based on penalized quasi-likelihood estimation, with
second-order Taylor series linearization (24, 27). We also calibrated a model with 6 categorical contrasts (shown in Figure 1)
by using the normal BMI range (18.5–22.9) as the reference.
However, because we did not find substantial differences in the
patterning of the socioeconomic exposures between the different
grades of underweight, we collapsed the different grades of underweight to 1 category of BMI 쏝 18.5.
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).
637
SOCIOECONOMIC POSITION AND NUTRITION IN INDIA
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
BMI 23–24.9
BMI 25–29.9
BMI 욷30
1.00
1.43 (1.35, 1.51)
1.94 (1.83, 2.06)
2.15 (2.02, 2.29)
2.45 (2.30, 2.61)
쏝0.0001
1.00
0.62 (0.58, 0.66)
0.42 (0.39, 0.46)
0.30 (0.27, 0.32)
0.20 (0.18, 0.22)
쏝0.0001
1.00
0.48 (0.45, 0.51)
0.26 (0.24, 0.28)
0.15 (0.14, 0.17)
0.09 (0.08, 0.11)
쏝0.0001
1.00
0.31 (0.28, 0.34)
0.11 (0.09, 0.13)
0.06 (0.04, 0.07)
0.03 (0.02, 0.04)
쏝0.0001
1.00
1.29 (1.22, 1.37)
1.66 (1.55, 1.76)
1.78 (1.66, 1.90)
1.96 (1.83, 2.11)
쏝0.0001
1.00
0.78 (0.73, 0.83)
0.66 (0.61, 0.72)
0.53 (0.48, 0.59)
0.41 (0.36, 0.46)
쏝0.0001
1.00
0.63 (0.59, 0.68)
0.43 (0.40, 0.48)
0.31 (0.27, 0.35)
0.22 (0.19, 0.26)
쏝0.0001
1.00
0.52 (0.46, 0.59)
0.30 (0.25, 0.37)
0.22 (0.17, 0.29)
0.14 (0.09, 0.21)
쏝0.0001
1.00
1.35 (1.29, 1.42)
1.15 (1.08, 1.24)
1.14 (1.09, 1.19)
1.34 (1.08, 1.66)
쏝0.0001
1.00
0.58 (0.54, 0.63)
0.52 (0.47, 0.57)
0.71 (0.67, 0.75)
0.33 (0.20, 0.55)
쏝0.0001
1.00
0.48 (0.44, 0.52)
0.41 (0.36, 0.47)
0.66 (0.61, 0.70)
0.51 (0.32, 0.81)
쏝0.0001
1.00
0.28 (0.24, 0.33)
0.16 (0.12, 0.20)
0.41 (0.37, 0.46)
0.17 (0.05, 0.54)
쏝0.0001
1.00
1.14 (1.08, 1.21)
0.97 (0.90, 1.04)
1.05 (1.00, 1.10)
1.16 (0.94, 1.44)
쏝0.0001
1.00
0.90 (0.83, 0.98)
0.85 (0.76, 0.95)
0.96 (0.91, 1.02)
0.61 (0.36, 1.02)
0.006
1.00
0.80 (0.73, 0.88)
0.76 (0.65, 0.87)
0.87 (0.81, 0.93)
0.84 (0.51, 1.36)
쏝0.0001
1.00
0.67 (0.56, 0.80)
0.52 (0.40, 0.69)
0.77 (0.69, 0.86)
0.54 (0.17, 1.70)
쏝0.0001
1.00
1.13 (0.87, 1.46)
1.49 (1.17, 1.89)
1.88 (1.48, 2.38)
2.23 (1.76, 2.82)
2.61 (2.06, 3.30)
쏝0.0001
1.00
1.03 (0.85, 1.26)
0.74 (0.62, 0.89)
0.58 (0.49, 0.70)
0.41 (0.34, 0.49)
0.26 (0.22, 0.31)
쏝0.0001
1.00
1.06 (0.89, 1.26)
0.71 (0.61, 0.84)
0.51 (0.43, 0.60)
0.35 (0.29, 0.41)
0.19 (0.16, 0.22)
쏝0.0001
1.00
0.91 (0.71, 1.16)
0.53 (0.43, 0.67)
0.34 (0.27, 0.43)
0.17 (0.13, 0.22)
0.07 (0.06, 0.09)
쏝0.0001
1.00
1.05 (0.80, 1.36)
1.17 (0.92, 1.49)
1.28 (1.00, 1.63)
1.36 (1.06, 1.74)
1.38 (1.08, 1.77)
쏝0.0001
1.00
0.98 (0.81, 1.20)
0.96 (0.80, 1.15)
0.92 (0.76, 1.11)
0.80 (0.66, 0.97)
0.68 (0.56, 0.83)
쏝0.0001
1.00
0.94 (0.79, 1.12)
0.85 (0.72, 0.99)
0.76 (0.64, 0.90)
0.65 (0.55, 0.77)
0.51 (0.43, 0.61)
쏝0.0001
1.00
0.83 (0.65, 1.07)
0.77 (0.61, 0.97)
0.75 (0.59, 0.97)
0.59 (0.46, 0.76)
0.43 (0.33, 0.56)
쏝0.0001
1.00
0.76 (0.70, 0.83)
1.24 (1.19, 1.30)
1.22 (1.15, 1.30)
쏝0.0001
1.00
1.45 (1.33, 1.59)
0.43 (0.40, 0.46)
0.70 (0.64, 0.77)
쏝0.0001
1.00
1.27 (1.16, 1.38)
0.27 (0.25, 0.30)
0.54 (0.49, 0.60)
쏝0.0001
1.00
1.43 (1.25, 1.63)
0.10 (0.08, 0.13)
0.43 (0.36, 0.52)
쏝0.0001
1.00
0.85 (0.77, 0.93)
1.06 (1.01, 1.11)
1.06 (1.00, 1.13)
쏝0.0001
1.00
1.11 (1.01, 1.22)
0.75 (0.69, 0.82)
0.85 (0.77, 0.94)
쏝0.0001
1.00
1.00 (0.91, 1.09)
0.53 (0.47, 0.59)
0.75 (0.68, 0.83)
쏝0.0001
1.00
0.87 (0.75, 1.01)
0.43 (0.34, 0.55)
0.64 (0.53, 0.78)
쏝0.0001
1.00
1.09 (0.93, 1.26)
1.20 (1.05, 1.37)
1.66 (1.48, 1.87)
쏝0.0001
1.001
0.84 (0.76, 0.92)
0.71 (0.66, 0.78)
0.35 (0.32, 0.37)
쏝0.0001
1.00
0.89 (0.74, 1.06)
0.63 (0.54, 0.74)
0.21 (0.18, 0.25)
쏝0.0001
1.00
0.72 (0.62, 0.82)
0.48 (0.42, 0.54)
0.11 (0.10, 0.12)
쏝0.0001
(Continued)
Downloaded from www.ajcn.org by on September 24, 2009
Household standard of living
Unadjusted
Top quintile (reference)
Fourth quintile
Third quintile
Second quintile
Bottom quintile
P
Adjusted
Top quintile (reference)
Fourth quintile
Third quintile
Second quintile
Bottom quintile
P
Caste
Unadjusted
Other caste (reference)
Scheduled caste
Scheduled tribe
Other backward class
No caste
P
Adjusted
Other caste (reference)
Scheduled caste
Scheduled tribe
Other backward caste
No caste
P
Education (y)
Unadjusted
쏜15 (reference)
13–15
9–12
6–8
1–5
0
P
Adjusted
쏜15 (reference)
13–15
9–12
6–8
1–5
0
P
Occupation
Unadjusted
Homemaker (reference)
Nonmanual
Agricultural
Manual
P
Adjusted
Homemaker (reference)
Nonmanual
Agricultural
Manual
P
Living environment
Unadjusted
Large city (reference)
Small city
Town
Rural area
P
BMI 쏝18.5
638
SUBRAMANIAN AND DAVEY SMITH
TABLE 2 (Continued)
Adjusted
Large city (reference)
Small city
Town
Rural area
P
BMI 쏝18.5
BMI 23–24.9
BMI 25–29.9
BMI 욷30
1.00
1.01 (0.87, 1.18)
1.06 (0.93, 1.20)
1.17 (1.04, 1.32)
0.0003
1.00
0.91 (0.82, 1.00)
0.84 (0.77, 0.91)
0.61 (0.57, 0.66)
쏝0.0001
1.00
0.92 (0.79, 1.07)
0.79 (0.70, 0.90)
0.49 (0.44, 0.56)
쏝0.0001
1.00
0.80 (0.70, 0.92)
0.63 (0.56, 0.71)
0.30 (0.26, 0.34)
쏝0.0001
1
In addition to individual age, mutually adjusted models included religion; tobacco and alcohol use; morbidities associated with asthma, malaria, and
tuberculosis; and parity.
Downloaded from www.ajcn.org by on September 24, 2009
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 crosslevel 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 standardof-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.
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.
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.
SOCIOECONOMIC POSITION AND NUTRITION IN INDIA
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.
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,
Downloaded from www.ajcn.org by on September 24, 2009
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 쏜15 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 욷US$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 울 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
639
640
SUBRAMANIAN AND DAVEY SMITH
interpretation of the results, and editing of the manuscript. The authors had no
conflicts of interest to declare.
REFERENCES
Downloaded from www.ajcn.org by on September 24, 2009
1. Sobal J, Stunkard AJ. Socioeconomic status and obesity: a review of the
literature. Psychol Bull 1989;105:260 –75.
2. Monteiro CA, Moura EC, Conde WL, Popkin BM. Socioeconomic status and obesity in adult populations of developing countries: a review.
Bull World Health Organ 2004;82:940 – 6.
3. Svedberg P. Poverty and undernutrition: theory, measurement and policy. New Delhi, India: Oxford University Press, 2000.
4. Ferro-Luzzi A, Sette S, Franklin M, James WPT. A simplified approach
to assessing adult energy deficiency. Eur J Clin Nutr 1992;46:173– 86.
5. WHO. Physical status: the use and interpretation of anthropometry.
Geneva, Switzerland: World Health Organization, 1995.
6. Naidu NA, Rao PN. Body mass index: a measure of nutritional status in
Indian population. Eur J Clin Nutr 1994;48:S131– 40.
7. Body mass index and mortality rates—a 10 year perspective study.
Hyderabad, India: National Institute of Nutrition, 1990.
8. Hosegood V, Campbell OMR. Body Mass index, height, weight, arm
circumference and mortality in rural Bangladeshi women: a 19-y longitudinal study. Am J Clin Nutr 2003;77:341–7.
9. Thomas D, Frankenberg E. Health, nutrition and economic prosperity: a
micro-economic perspective. Los Angeles, CA: California Center for
Population Research, University of California, 2002.
10. Kusin J, Kardjati S, Renqvist U. Maternal body mass index: the functional significance during reproduction. Eur J Nutr 1994;48:56 – 67.
11. Barker DJP. The effects of nutrition of the foetus and neonate on cardiovascular disease in later life. Proc Nutr Soc 1992;51:135– 44.
12. Caballero B. Obesity in developing countries: biological and ecological
factors. J Nutr 2001;131:866S–70S.
13. Osmani S, Sen A. The hidden penalties of gender inequality: fetal origins
of ill-health. Econ Hum Biol 2003;1:105–21.
14. IIPS. National Family Health Survey 1998-99. Mumbai, India: International Institute of Population Sciences, 2000.
15. Mishra V. Effect of obesity on asthma among adult Indian women. Int J
Obes 2004;28:1048 –58.
16. WHO. Expert Consultation. Appropriate body mass index for Asian
populations and its implications for policy and intervention strategies.
Lancet 2004;363:157– 63.
17. Razak F, Anand S, Vuksan V, et al. Ethnic differences in the relationships between obesity and glucose-metabolic abnormalities: a crosssectional population-based study. Int J Obes 2005;29:656 – 67.
18. Misra A, Vikram NK, Gupta R, Pandey RM, Wasir JS, Gupta VP. Waist
circumference cutoff points and action levels for Asian Indians for identification of abdominal obesity. Int J Obes 2006;30:106 –11.
19. Filmer D, Pritchett L. Estimating wealth effects without expenditure
data— or tears: an application to educational enrollments in states of
India. Demography 2001;37:155–74.
20. Hallerod B. Poor Swedes, poor Britons: a comparative analysis of relative deprivation. In: Andres A, ed. Empirical poverty research in a
comparative perspective. Aldershot, United Kingdom: Ashgate, 1998.
21. Subramanian SV, Nandy S, Kelly M, Gordon D, Davey Smith G. Patterns and distribution of tobacco consumption in India: cross-sectional
multilevel evidence from the 1998 –99 National Family Health Survey.
BMJ 2004;328:801– 6.
22. Chitnis S. Definition of the terms scheduled castes and scheduled tribes:
a crisis of ambivalence. In: Pai Panandiker VA, ed. The politics of
backwardness: reservation policy in India. New Delhi, India: Centre for
Policy Research, 1997.
23. National human development report 2001. New Delhi, India: Planning
Commission, Government of India, 2002.
24. Goldstein H. Multilevel statistical models. 3rd ed. London, United Kingdom: Arnold, 2003.
25. Subramanian SV. The relevance of multilevel statistical models for
identifying causal neighborhood effects. Soc Sci Med 2004;58:1961–7.
26. Subramanian SV, Jones K, Duncan C. Multilevel methods for public
health research. In: Kawachi I, Berkman LF, eds. Neighborhoods and
health. New York: Oxford University Press, 2003:65–111.
27. Rasbash J, Steele F, Browne W, Prosser B. A user’s guide to MLwiN
version 2.0. London, United Kingdom: Centre for Multilevel Modelling,
Institute of Education, 2004.
28. Monteiro CA, Conde WL, Lu B, Popkin BM. Obesity and inequities in
health in the developing world. Int J Obes 2004;28:1181– 6.
29. Mendez MA, Monteiro CA, Popkin BM. Overweight exceeds underweight among women in most developing countries. Am J Clin Nutr
2005;81:714 –21.
30. Reddy NR. Body mass index and its association with socioeconomic and
behavioural variables among socioeconomically heterogeneous populations of Andhra Pradesh, India. Hum Biol 1998;5:901–17.
31. Dhurandhar NV, Kulkarni PR. Prevalence of obesity in Bombay. Int J
Obes 1992;16:367–75.
32. Shukla HC, Gupta PC, Mehta HC, Hebert JR. Descriptive epidemiology
of body mass index of an urban adult population in western India. J
Epidemiol Community Health 2002;56:876 – 80.
33. Shetty PS. Nutrition transition in India. Public Health Nutr 2002;5:
175– 82.
34. Ball K, Crawford D. Socioeconomic status and weight change in adults:
a review. Soc Sci Med 2005;60:1987–2010.
35. World Bank. DEPweb: GNP per capita, Asia (South and East) and the
Pacific. Internet: http://www.worldbank.org/depweb/english/modules/
economic/gnp/datasia.html (accessed 23 November 2005).
36. Yajnik CS. Obesity epidemic in India: intrauterine origins? Proc Nutr
Soc 2004;63:387–96.
37. Misra A, Vikram NK. Insulin resistance syndrome (metabolic syndrome) and obesity in Asian Indians: evidence and implications. Nutrition 2004;20:482–91.
38. Snehalatha C, Viswanathan V, Ramachandran A. Cutoff values for normal anthropometric variables in Asian Indian adults. Diabetes Care
2003;26:1380 – 4.
39. Yusuf S, Hawken S, Ôunpuu S, et al. Obesity and the risk of myocardial
infarction in 27 000 participants from 52 countries: a case-control study.
Lancet 2005;366:1640 –9.
40. Molarius A, Seidell JC, Sans S, Tuomilehto J, Kuulasmaa K. Waist and
hip circumferences, and waist-hip ratio in 19 populations of the WHO
MONICA Project. Int J Obes Relat Metab Disord 1999;23:116 –25.
41. Ness AR, Leary S, Reilly J, et al and the ALSPAC study team. The social
patterning of fat and lean mass in a contemporary cohort of children. Int
J Paediatr Obes 2006;1:56 – 61.
42. Gluckman P, Hanson M. The fetal matrix: evolution, development and
disease. Cambridge, United Kingdom: Cambridge University Press,
2004.