Contents - Previous - Next

Body mass index and illness in rural Sarawak


S. S. Strickland and S. J. Ulijaszek

London School of Hygiene & Tropical Medicine, Keppel Street, London WC1E 7HT; and Department of Biological Anthropology, Downing Street, Cambridge CB2 3DZ, UK

Correspondence to: S.S. Strickland

This paper reports findings of a cross-sectional anthropometric survey of 447 men and 564 women aged >18 years of the Iban tribe in rural Sarawak. The percentage of each approximately 10 year cohort with body mass index (BMI: kg/m2) <18.5 increased progressively with age >40 years in both sexes. BMI and arm-muscle plus bone cross-sectional area (AMBA) were influenced significantly by reported single symptoms in men and two concurrent symptoms in women. Sensitivity of BMI <18.5 for identifying those reporting symptoms of illness was 21% in men and 29% in women. Independently of illness effects, values of stature, BMI and AMBA in men, and stature alone in women, were significantly greater in those farming plots with the shortest time left fallow, suggesting their comparative success. The BMI of men may reflect primarily their economic activities, while that of women their experience of illness; the AMBA may also be a better overall index of health than BMI in this population.


Several attempts have been made to establish cut-off points of the body mass index (BMI: kg/m2) appropriate for identifying acute or chronic low energy nutritional status in adults. Thus, FAO/WHO/UNU (1985) proposed 20.1 in men and 18.7 in women; and on the basis of cadaver studies, Henry (1990) has postulated lower limits to human survival of 13.0 in males and 12.0 in females. Across the approximately intervening range, James, Ferro-Luzzi & Waterlow (1988) and Ferro-Luzzi et al. (1992) have proposed a graded classification of chronic energy deficiency (CED) using BMI values <18.5 for both sexes, with or without an estimate of physical activity level expressed as a multiple of the basal metabolic rate.

However, it remains unclear to what extent decreasing values of BMI are associated with progressive functional impairments, whether in terms of morbidity, or ratings of physical performance or social economic success. Further, their relationship to physical deficiencies, which are general consequences of ageing rather than energy nutrition, does not appear to have been investigated systematically.

This paper examines relationships between BMI and illness using data from a harvest-season survey of rural Sarawakis belonging to the Iban tribe. It is important at the outset to stress that two general ideas of 'illness' are intended in this context. The first is the individual's perception of his own physical condition, indicated by self-reported symptoms. Without examination, this perception is to be distinguished from clinically identified morbidity; but both are counterparts to the concept of health as the absence of illness or disease (Lewis, 1953, 1975) and are used as such in this paper.

The second notion of 'illness' is more ambitious and complex. This is the counterpart to the concept of population health as an ecologically or demographically 'sustainable' state, which King (1990) has argued should guide public health policy. King's argument is that the maintenance of the capacity of the ecosystem to support life in quantity and variety ('sustainability') should be embodied in a concept of population health, which is at risk from an excessive population increase. Thus he defines health, expanding on the WHO charter, as 'a sustainable of complete physical, mental and social well-being, and not merely the absence of disease and infirmity'.

It follows from King's definition that any unsustainable state of health is in itself unhealthy, regardless of the physical characteristics of the population. However, in assessing the functional significance of the BMI in the present case, it is useful to exploit the idea of ecological sustainability since the Iban practice land-extensive, forest-fallow or 'shifting' agriculture. Populations engaged predominantly in this form of subsistence characteristically live at low densities, since failure to sustain long fallows between periods of cultivation would entail a risk of progressive soil erosion and irreversible degradation of the habitat (Freeman, 1955). Concern over the implications of persistent population growth for forest ecosystems has therefore led many observers to see shifting agriculture as a problem, sometimes formulated as a cycle (Fig. 1) of which the dynamics resemble those of the 'demographic trap' outlined by King (1990). Thus, claims that shifting agriculture 'almost inevitably leads to undernutrition and what could even be loosely termed as semi-starvation' (Hatch & Lim, 1978), particularly where either land law or commercial timber extraction entails restrictions on migration or access to mature forest (Hong, 1987), suggest that the ecological status of the forest farm could be used to indicate the sustainability of the subsistence pattern.

In assessing relationships between the BMI and reported morbidity, it is therefore possible to explore the hypothesis that low BMI is associated with farms left fallow for short periods. Thus BMI may reflect ecological stress or interact with the effects of illness. By assessing both relationships, the functional value of BMI as an index of ecological stress can be investigated at a local level.


The Sarawaki populations of the neighbouring Districts of Song and Kanowit were chosen for survey because these areas had been reported to differ significantly in the prevalence of childhood undernutrition. The Department of Medical Services' clinical nutritional surveillance system indicated for children aged <6 years that, between 1983 and 1987, approximately 50-70% of those in Song and 20-40% of those in Kanowit were <81% weight-for-age according to Singaporean reference values (Department of Medical Services, 1988). Were the pattern of undernutrition in children to be reflected in that of adult body size, then the two Districts could also be expected to contrast anthropometrically in older age groups and to show a range of variation in body size suitable for analysis of low BMI. In this paper, only the adult data are reported.

Anthropometric measurements included weight, height, mid-upper arm circumference, and skinfold thicknesses at the biceps and triceps sites. Beam-balance scales were used to weigh lightly clothed subjects. A Harpenden anthropometer, steel measuring tape, and Holtain calipers were used for the measurement of stature, arm circumference, and skinfold thicknesses according to methods described in Weiner & Lourie (1981).

As a measure of lean tissue, the cross-sectional area of arm muscle plus bone (AMBA) was calculated using the formula

where C is mid-upper arm circumference (cm), and S is the mean of triceps and biceps skinfold thicknesses (mm) (Norgan & Jones, 1990).

Fig. 1. The problem of shifting cultivation.

Perceived illness was assessed by simple interview in the Iban language. Individuals were asked whether, on the day preceding the measurement, they had experienced diarrhoea (cirit), coughing (batok), fever (angat), or forced recumbence (enda'tau' bejalai 'not able to walk', excluding those unable through injury to stand for measurement). About 19% of adults reported combinations of symptoms in these terms, giving rise to cells too small for statistical treatment. Subjects were therefore classified for analysis by the reported presence or absence of any symptom, and by the number of symptoms. No estimates of duration were obtained.

Age was established as far as possible by identity cards or birth certificates. However, registration has been systematically administered only since the late 1940s; and it is acknowledged that ages of adults born before then may be accurate perhaps to +5 years. Of the 379 women aged <50 years, 9.0% claimed to be pregnant and have been excluded from the analysis; a sex ratio of 0.79 suggests that the study population is somewhat biased, because of migrant labour and probable differential mortality.

Subsistence patterns in rural Sarawak involve a complex of farming, horticulture, hunting, fishing, gathering, and variable amounts of what is usually migratory wage-labour; the proximity to downriver market centres is economically significant. In the sampling procedure, longhouse villages were therefore selected randomly from official District lists until it was judged that about 2000 individuals of all ages could be measured. Figure 2 shows that the distribution of longhouses includes both relatively remote upriver villages and some close to the District centres bordering the Rejang River, which is the principal arterial waterway of the State.

Local and seasonal variation make data on annual average incomes from all sources difficult to collect or to interpret in cross-sectional surveys. Thus, classification of the sample households by relative social economic standing was not attempted. However, for those practicing shifting agriculture, the type of farm made during the season preceding the measurement was established from interview, and described by the quality of the fallow vegetation from which it was cleared. In order to obtain reasonable sample sizes, four derivative categories were used:

0. no forest fallow farm;
1. kerukoh and kakah, less than 1 year fallow;
2. dijap and temuda', approximately 1-4 years fallow;
3. damun and pengerang, estimated to be 5-30 years fallow.

Fields made from pampas grass (Imperata cylindrica), usually a consequence of repeated short fallow farming, were few and have been included under category 1. The estimates of fallow duration have been derived from the definitions of these terms given by Freeman (1955) and Richards (1981). These therefore depend upon the Iban's judgement of the size and composition of forest regrowth, and are not botanically defined.

Quantitative analysis used the SPSS/PC+ statistics package to enable analysis of variance, linear regression, multiple range and c2 tests on the data. The level of statistical significance was set at 5%.

Contents - Previous - Next