Previous - Next
Introduction
Nutritional anthropometric indices and their relationship to BMR
Do population groups in developing countries in the tropics have lower BMRs?
Low BMIs, BMRs and energy requirements
Changes in body weights and stature and their influence on BMI and energy requirements
References
Discussion
P. S. Shetty1,2, M. J. Soares1
and W. P. T. James3
1Nutrition Research Centre, St John's Medical College, Bangalore, India; 2Department of Public Health & Policy, London School of Hygiene and Tropical Medicine, London; and 3Rowett Research Institute, Greenburn Road, Bucksburn, Aberdeen, UK
Correspondence to: Professor W. R T.
James.
Two significant developments over the last decade
have influenced our understanding of energy requirements of humans and their implications
in arriving at the numbers of individuals in population groups worldwide who are
undernourished and do not receive adequate levels on a daily basis. The first major
development has been the recommendation of the FAD/WHO/UNU Expert Consultation on Energy
and Protein Requirements (1985) to (a) rely, as a matter of principle, on estimates of
energy expenditure (actual or desirable) to arrive at estimates of energy requirements,
and (b) to use the basal metabolic rate (BMR) factorial approach for the assessment of
total energy expenditure of individuals, communities and population groups. The second
advance, more recently made, has been the suggestion that nutritional anthropometric
measures, more specifically the use of body mass index (BMI) could be a simple, reliable
and easily obtainable objective anthropometric criterion for both the definition and
diagnosis as well as an estimate of the severity of undernutrition or chronic energy
deficiency (CED) in adults (James et al., 1988).
This paper will discuss the
relationships between BMI and BMR and their implications in arriving at estimates of
energy requirements of adults both in developed and developing countries. It will also
compare the relationships of estimates of BMR derived from other nutritional
anthropometric variables such as body weight and height. Evaluation of the lower BMRs of
populations in tropical regions - and in particular of those who are likely to be
chronically undernourished - and their relationship to BMI will be carried out. Whether
all individuals below an acceptable cut-off for BMI (i.e. BMI <18.5) are in fact
suffering from CED and the implications for estimating energy requirements will also be
explored. Finally, we will consider the effect in the age profile of populations and the
impact of secular changes in stature in communities worldwide on the BMR and energy
requirements.
Several anthropometric parameters, such as body
weight and height, as well as their transformations, such as BMI, show associations with
BMR. The extensive analysis by Schofield, Schofield & James (1985) seemed to indicate
that when BMR was plotted against weight (or height) the relationship appeared to be
quadratic, cubic or of a more complicated form although there was a strong linear
component. François (1981) had earlier dealt with a similar problem by computing
semi-logarithmic regression equations split into four weight groups thus representing the
relationship by a series of straight lines fitted along a full curve. It is now generally
accepted that weight provides the best predictor for BMR (Schofield et al., 1985;
Soares & Shetty, 1988) since the two are linearly related on regression analysis.
However, the addition of terms for height and BMI to such equations does not improve the
quality of prediction.
More recent analyses of a reasonably large database on BMRs measured prospectively and carefully collected from well-nourished, adult, male Indians within age ranges 18-30 years suggested that height and BMI contribute roughly in equal measures to variations in BMR (Soares & Shetty, 1988). The correlation coefficients in the series reported by Soares & Shetty (1988) for multiple linear regressions using BMR as the dependent variable were as follows: weight alone, multiple r = 0.80 (r2 = 0.64); weight + age, multiple r = 0.81 (r2 = 0.65); weight + BMI + age, multiple r = 0.81 (r2 = 0.65). The addition of other variables makes hardly any difference to the strong correlations that body weight has with BMR. However, in the 30-60 year age group of adult males the inclusion of age increased the explained variance in BMR by 5.3% (Soares, Francis & Shetty, 1993) which is in contrast to Schofield's observation that age made little difference to the final prediction equation relating BMR to body weight for males >18 years old. A careful examination of the analysis of data by Rand (1982) confirms, for both males and females separately, that single equations could be successfully fitted using a single independent anthropometric variable (or its transformation such as log of weight or square of log of weight) and that the addition of several variables such as age or height added nothing more to the prediction other than a small increase in noise.
Comparison of the correlation coefficients obtained by both Schofield and by Soares & Shetty (1988) shows that BMR has a stronger correlation with body weight than with any other nutritional anthropometric index used as a single independent variable. In Schofield's analysis (1985) the r value for body weight was 0.875 for males (n = 4809) and r = 0.851 for females (n = 2364). The correlations for height alone were r = 0.87 and r = 0.886 for similar numbers of males and females respectively. It is important to point out that the relationships are not perfect and the high correlations obtained with height are largely the result of the contributions of large numbers in the database.
The analysis by Soares & Shetty (1988) demonstrated an r = 0.80 for body weight accounting for 64% of the variance while BMI had an r = 0.64 and height an r = 0.57, thus explaining only 41% and 33% of the variance respectively. This association between BMI and BMR, seen in both well-nourished adults and those with low BMIs in developing countries, has also been reported in well-nourished populations and those with increasing degrees of obesity in the West (Garrow et al., 1988).
BMI is thus not as useful a
predictor of BMR as body weight and so is not the most useful index for predicting the BMR
of individuals or population groups when applying the factorial method to estimate human
energy requirements. However, since BMI has approximately the same value for short, medium
height and tall individuals, being independent of stature (Khosla & Lowe, 1967), the
index simplifies the approach to estimating the acceptable or optimum body weight for that
height among groups of different heights in a population. BMR, and consequently total
energy expenditure, can be predicted from the derived optimum body weight, and thus BMI
may be a useful addition in arriving at desirable levels of energy requirements of
individuals or populations (Shetty & James, 1994).
At the behest of those international agencies (FAO,
WHO & UNU) responsible for the Expert Consultation on Energy and Protein Requirements
(1985), Schofield et al. (1985) derived predictive equations for BMR of both sexes
based on body weight as the most suitable nutritional anthropometric predictor. BMR
measurements of Indians constituted one of the largest single ethnic groups from the
tropics. During these computations, Schofield made the observation that when the measured
BMRs of Indians in this vast database were compared with the predicted value derived from
equations obtained after excluding Indians there was a significant lack of fit. The
actual, measured BMRs of Indians were 11.2% and 9.9% lower for males and females
respectively compared with the data on Europeans and North Americans. A plot of BMR vs
body weight and BMR/weight vs body weight revealed that even for the same body weight,
Indians on average had lower BMRs. Lack of sufficient data at that stage precluded any
extension of this analysis to other Asiatic groups, although an earlier analysis by
Quenouille et al. (1957) had suggested the same. A more recent analysis by Henry
& Rees (1988) also provided further evidence to support these observations and implied
that this phenomenon of a lower BMR in tropical populations was not unique to Indians but
was found in other Asiatic groups as well.
Obviously, this observation has enormous implications for the estimates of energy requirements (based on the BMR factorial method) of population groups in developing countries, these being located mostly in the tropical belt. The bulk of the Indian data used by Schofield were reported over 30-40 years ago (Mason & Benedict, 1931; Niyogi, Patwardhan & Modecai, 1939; Sokhey & Malandkar, 1939), but were clearly derived from a series of meticulous studies. Soares & Shetty (1988) therefore reexamined the issue in current population groups in India to see if the conclusions still held and if so what the basis for the difference in BMR might be. From prospective measurements of BMRs of well-nourished young adult Indian males (18-29+ years) it was evident that the entire group of men had mean BMRs 9.3% lower than those predicted by Schofield's equation for non-Indians. Schofield's predictive equation, derived from age-matched European or North American subjects only, was BMR (in MJ. d-1) = 0.0582 × body weight (kg) + 3.2399 for males aged 18 to 29 years.
However, the percentage deviation of our Indian data from this particular BMR predictive equation of Schofield varied among the subgroups from the same ethnic sample. The better nourished from upper socio-economic groups (in both urban and rural areas) had BMRs only 5.5-5.7% less than the Schofield's European and American values. The age-matched individuals from poor socio-economic groups, who were likely to be undernourished, deviated by 12.7%. When comparisons were made between the measured BMRs and the Schofield predicted BMRs over 5 kg body weight ranges, deviations from the predictive equation showed a curvilinear pattern decreasing to a minimum (<5%) as body weight increased. The deviations were greater again over a range of body weights from <45 to >65 kg (Soares & Shetty, 1988). This pattern was then seen to be true for the BMR data collected over 30 years ago from the same ethnic groups (Niyogi et al., 1939; Sokhey & Malandkar, 1939) when these were reexamined.
In the 55-60 kg body weight range for adults - this being the currently accepted ideal weight for the Reference Indian male for estimating energy requirements (ICMR Expert Group, 1990) - the deviations from those predicted by Schofield's equations were small and amounted to 4.6% in the recent study (Soares & Shetty, 1988) and to 6.8% in the age-matched males of the 1939 study (Niyogi et al., 1939; Sokhey & Malandkar, 1939). The relationship between body weight and BMR is not necessarily one of simple linearity as commented upon earlier by François (1981). Nevertheless, the correlations between body weight and BMR are good. The differences may be accounted for by differences in the body composition affecting not only the ratio of fat to fat-free mass (FFM) but also differences in the contribution of muscle and visceral tissues within the FFM (Shetty, 1993).
Similar deviations between the observed Indian data and predicted BMRs, using Schofield's equations, are seen as the BMIs change from BMI 14 to 25. The lowest deviation (-6.5%) is seen in the BMI range of 18-20 which can be considered as normal for India (Soares & Shetty, 1988). Comparisons of three separate data sets of BMR from the Indian subcontinent obtained prospectively in recent years (Srikantia, 1985; McNeil et al., 1987; Soares & Shetty, 1988) not only demonstrate a concurrence between these three sets but also that all three data sets deviate by <2.7% from that predicted by the equations of Henry & Rees (1991) for tropical people. However, the observed BMRs in these three separate sets of data deviate by up to >7.5% from those predicted by Schofield's equations (Soares, Francis & Shetty, 1993).
Climate and environmental factors
may account for these 5-6% differences between the observed BMRs of well-nourished Indians
and that predicted from equations based on North American or European data. Several
reports support the observations that Europeans in tropical climates have BMRs which are
about 5% lower than their counterparts in Europe (Mason, 1934; McGregor & Loh, 1941;
Munro, 1950; Quenouille et al., 1957). Asians living in the UK have BMRs comparable
to their Western counterparts (Mahadeva, 1954; Henry, Piggott & Emery, 1987). A
greater degree of muscle relaxation during BMR measurements by Asians may also contribute
to the lowered BMR and it is also likely that differences in body composition in the
population groups in the tropics may also contribute to this difference.