Table 10. Mean height, weight, BMI and CED according to height class
Height class percentile 
Height (cm) 
Weight (kg) 
BMI 

n 
Mean 
SD 
Mean 
SD 
Mean 
CV 
CED (BMI <18.5) % 

Men 

Short 10 
943 
151.1 
3.03 
43.0 
5.94 
18.8 
13.5 
49.7 
Average 4060 
1888 
162.7 
0.95 
50.3 
6.99 
19.0 
13.8 
47.5 
Tall 90 
944 
174.0 
1.41 
57.5 
8.30 
19.0 
13.6 
49.5 
Women 

Short 10 
1190 
140.2 
2.76 
37.6 
6.55 
19.1 
16.9 
50.8 
Average 4060 
2381 
150.4 
0.81 
43.2 
6.58 
19.1 
14.9 
49.2 
Tall 90 
1191 
160.8 
2.71 
48.7 
7.67 
18.9 
15.7 
52.2 
Table 11. Maternal parameters according to BMI status
BMI status 

<16 
1617 
1718.5 
18.520 
2025 
2530 
>30 

CEDIII 
CED II 
CED I 
(normal) 
normal) 
(obese) 
(obese) 
Total 

Sample size 

n 
81 
133 
460 
553 
717 
68 
5 
2017 
% 
4.0 
6.6 
22.8 
27.4 
35.5 
24.0 0.2 
100.0 

Mother's age (years) 

Mean 
24.0 
23.8 
23.6 
23.8 
24.6 
26.2 
28.8 
24.1 
CV % 
21.6 
20.9 
19.8 
18.0 
18.9 
20.3 
25.7 
19.3 
Gravida 

Mean 
2.6 
2.6 
2.4 
2.5 
2.6 
2.7 
3.2 
2.5 
CV % 
65.8 
59.4 
57.7 
54.9 
55.1 
57.7 
86.7 
56.6 
Haemoglobin 

Mean (g%) 
10.3 
10.5 
10.8 
10.9 
10.9 
11.3 
9.9 
10.9 
CV % 
23.2 
23.3 
26.4 
21.8 
28.8 
19.0 
11.1 
25.6 
Mother's height (cm) 

Mean 
151.9 
151.4 
151.2 
151.2 
151.0 
151.5 
151.7 
151.1 
CV % 
3.7 
3.9 
3.6 
3.9 
3.5 
3.3 
2.2 
3.7 
Mother's weight (kg) 

Mean 
35.4 
38.1 
40.9 
44.1 
49.6 
60.6 
75.5 
45.2 
CV % 
8.8 
8.1 
7.7 
8.1 
9.2 
8.4 
11.6 
14.7 
Birth weight (g) 

Mean 
2510 
2573 
2653 
2771 
2812 
2972 
2972 
2742 
CV % 
17.3 
20.3 
17.5 
20.1 
18.6 
19.2 
17.3 
19.2 
Low birth weight (<2500g) % 
53.1 
41.4 
35.9 
27.7 
26.4 
14.7 
20.0 
30.5 
Odds ratio of LBW 
3.16 
1.97 
1.56 
1.07 
1.0 
2.08 
1.73 
2.02 
Correlation coefficient between:
BMI and weight = 0.8763, P < 0.001.
BMI and height = 0.0035; NS.
Weight and height = 0.4743; P < 0.001.
BMI and confounding variables for LBW
Studies on LBW and its determinants suggest that the age of mother and parity are two confounding variables. In Tables 12 and 13, results of analyses using MantelHaenszel (^{2}test (e.g. Test, 1990) are presented to assess the effect of maternal BMI on LBW controlling for the confounding factors of maternal age and parity. The weighted mean odds ratio was 1.76 with 95% confidence interval of 1.432.20. The test for interaction was undertaken to assess the risk of maternal age at different levels of BMI. The (^{2}test showed a nonsignificant value, indicating that the difference within the BMI classifications between the age groups was only due to chance errors. This suggests that CED of mothers is itself a definite risk factor for LBW; maternal age does not seem drastically to alter the relationship between BMI and LBW.
The odds ratios in respect of LBW and BMI given in Table 13 showed only a parity level, e.g. >6 emerged as a confounding risk in the present analyses. To investigate the relationship between birth weight as the dependent variable in a dichotomous form (i.e. LBW as present or not) and other variables, e.g. anthropometry and socioeconomic status as independent variables, a logistic regression analysis was performed.
This finding apparently contradicts the earlier observations that the height did not vary much between different BMI groups. We believe that height exerts an influence on LBW in communities where chronic undernutrition is primary and widely prevalent with a retardation in growth in the community. In other words, the shortness of women in this community, perhaps unlike the Chinese and Japanese experience, is not due to genetic factors but is due to nutritional constraints. This means that it is remediable: if nutritional constraints are removed and Indian girls are allowed to achieve their full growth potential then height should cease to be a risk factor.
Table 12. Percentage low birth weight by age and BMI and their odds ratios (OR)
Age (years) 
BMI 
OR 
log OR 
Var (log) OR 

<18.5 
<18.5 

n 
% 
n 
% 

<18 
60 
51.7 
95 
38.9 
1.6757 
0.5162 
0.1110 
1921 
142 
43.7 
239 
34.3 
1.4838 
0.3946 
0.0472 
2224 
117 
33.3 
237 
22.4 
1.7358 
0.5515 
0.0628 
2527 
124 
30.6 
300 
17.3 
2.1073 
0.7454 
0.09612 
2830 
77 
35.1 
219 
23.7 
1.7342 
0.5506 
0.0823 
>31 
45 
42.2 
124 
24.2 
2.2897 
0.8284 
0.1351 
Weighted mean OR = 1.76; 959/0 confidence
interval 1.43,2.2.
Test for inequality of odds ratios: c^{2} (5 d.f. = 1.69 (NS).
Table 13. Percentage low birth weight by parity and BMI and their odds ratios (OR)
Parity 
BMI 
OR 
log OR 
Var (log OR) 

<18.5 
³18.5 

n 
% 
n 
% 

1 
139 
48.2 
238 
36.1 
1.6447 
0.4976 
0.0470 
2 
169 
39.6 
208 
40.9 
1.7233 
0.5442 
0.0410 
3 
116 
30.2 
260 
18.1 
1.9582 
0.6720 
0.0669 
45 
102 
29.4 
282 
23.0 
1.3910 
0.3300 
0.0672 
>6 
37 
48.6 
125 
18.4 
4.2014 
1.4354 
0.1615 
Weighted mean OR = 1.74; 95%
confidence interval 1.44,2.3.
Test for inequality of odds ratios: c^{2} (4 d.f.) = 5.76 (P < 0.05).
BMI and mortality
Results of a study relating mortality and BMI status, reported earlier by NIN (NIN Report, 198990), are depicted in Fig. 3. It clearly shows that mortality steadily increased from 12 per 1000 population with 'normal' BMI class with BMI >18.5 to 33 per 1000 population in grade III CED.
BMI and childhood malnutrition
For studying the relationship between BMI and childhood malnutrition, the data from a longitudinal study of 23 years (196583) conducted at the NIN were used (Satyanarayana, personal communication). Results are depicted in Fig. 4. The curves on the left are frequency distributions of the BMI values of adult males and females (aged 2023 years) who were malnourished (on a weightforage basis as described by Gomez et al. (1956) at age 5 years. The adult mean BMI for the formerly undernourished group is ~16.0 for the wellnourished group in childhood it is ~21. The 2 SD value of the wellnourished group is ~ 17.8, while the 2 SD of the undernourished group is ~18.5. The overlap of the two distributions is ~10%, which is within the range of a normal distribution. A look at the figures in Table 14, which provide BMI data from British army recruits and from affluent Indians together with the longitudinal data of the affluent children, suggests that the 2 SD values of the three groups vary within a narrow range and have a mean of 18.0. Therefore, on this basis we propose that a cutoff point of 18.0 would be more appropriate to distinguish the groups of normal adults who are energy deficient (CED).