Contents - Previous - Next

Results

General findings

There were no differences in BMI between Districts, and so the data were pooled. Table 1 shows the general physical characteristics of the measured sample divided into approximately 10 year age cohorts. There was an overall decline in the mean averages of these variables with age, which was more marked for weight and BMI in women than in men, and more marked for weight than height in both sexes. The BMI values suggest a threshold before declining from about age 50 in men and 40 in women. The range of variation was larger in women than in men, as has been reported elsewhere (Casey et al., 1992).

A result of the age-related pattern is the finding that in both sexes there was an increase with age in the percentage of each cohort with low values of the BMI after age 40 years (Fig. 3 a, b). This resembles the general pattern of ageing which can be inferred from data for several non-Western groups, including East African pastoralists (Little, Galvin & Mugambi, 1983) and Javanese farmers (Bailey, 1962) and low income Brazilians (Vasconcellos, 1994). It is the converse of that reported in Western populations, which characteristically gain weight with age (Waaler, 1984; Casey et al., 1992); and this is perhaps a consequence of progressively falling levels of physical activity related to retirement thresholds. This decline in physical activity is less pronounced or absent in non-Western populations. There is no evidence that obligatory underfeeding by the elderly, reported in studies in Java (Bailey, 1961), Mali and elsewhere, has contributed to this pattern in the present case.

Table 1. Body size of Iban adults by age cohort (mean ± SD)

Age	Height (cm)	Weight (kg)	BMI (kg/m2)	AMBA (cm2)	n
Men
18-	159.5 ± 5.6	54.5 ± 6.4	21.4 ± 2.1	50.5 ± 7.9	111
30-	159.5 ± 5.6	54.5 ± 7.0	21.4 ± 2.3	52.1 ± 9.7	75
40-	157.6 ± 5.2	52.7 ± 7.6	21.2 ± 2.7	51.1 ± 10.4	95
50-	156.1 ± 5.2	50.2 ± 8.2	20.6 ± 3.1	46.3 ± 10.3	60
60-	155.4 ± 5.1	47.7 ± 5.7	19.7 ± 2.0	43.8 ± 8.2	71
70-	151.1 ± 13.8	45.5 ± 10.8	19.6 ± 3.2	40.3 ± 11.4	35
Women
18-	149.0 ± 4.5	50.7 ± 7.9	22.8 ± 3.3	41.7 ± 11.3	129
30-	148.6 ± 4.4	51.2 ± 9.2	23.2 ± 3.7	44.3 ± 9.8	107
40-	148.0 ± 4.7	47.6 ± 9.1	21.7 ± 3.6	41.8 ± 8.9	108
50-	146.9 ± 4.9	45.4 ± 8.0	21.0 ± 3.4	41.0 ± 9.5	91
60-	145.3 ± 5.9	41.8 ± 8.1	19.7 ± 3.3	36.6 ± 8.5	84
70-	143.0 ± 7.5	37.4 ± 7.9	18.2 ± 3.4	33.2 ± 8.7	45

BMI = body mass index; AMBA = arm-muscle plus bone cross-sectional area.

The relationship of BMI to illness

The point prevalences of reported symptoms of illness varied between the sexes with age group (Fig. 4). Men aged <40 years reported significantly more illness than did women, although in the older group the frequency distribution of the numbers of symptoms did not differ between the sexes.

In men aged <40 years, the BMI and the AMBA were significantly lower in those reporting single symptoms than in asymptomatic individuals (Table 2). This was also true for women aged <40 years reporting two concurrent symptoms. However, it is not possible to argue causality on the basis of cross-sectional data, although this is critical for understanding the functional significance of BMI (James et al., 1988). Nevertheless, in the older age group, examination of the statistical effects of age and illness on BMI and AMBA in both sexes by two-way Anova suggested (Table 3) that the number of symptoms was significantly associated with BMI and AMBA independently of the effects of age in men; but there was no evidence for a significant effect of number of symptoms on these index in women. Thus the effects of illness and ageing were additive in men and, although women were more resilient than men to effects of illness, the women were affected to a greater degree by ageing itself.

Table 2. Body mass index (BMI) and arm muscle plus bone cross-sectional area (AMBA) by age group and number of reported symptoms (mean ~ SD)

	Men				Women
	18-40 years		40+ years		18-40 years		40+ years
	BMI	AMBA	BMI	AMBA	BMI	AMBA	BMI	AMBA
Asymptomatic	21.4 ± 2.1	50.9 ± 8.2	21.1 ± 3.0	49.5 ± 12.3	23.5 ± 3.4	44.4 ± 11.9	20.8 ± 3.1	40.6 ± 9.7
n	96		107		157		128
Symptoms: 1	21.6 ± 2.5	51.6 ± 7.5	19.7 ± 2.4*	42.8 ± 7.7*	22.3 ± 3.5	40.7 ± 6.9	20.5 ± 3.9	38.4 ± 9.1
n	63		83		54		129
Symptoms: 2	21.2 ± 1.9	51.5 ± 13.6	20.6 ± 2.3	46.8 ± 8.4	21.1 ± 2.5*	37.1 ± 6.4	19.7 ± 3.8	37.0 ± 8.7
n	22		53		22		59
Symptoms: 3+	20.5 ± 1.4	48.2 ± 4.4	19.8 ± 2.9	45.5 ± 12.8	23.3 ± 5.3	43.3 ± 10.1	21.2 ± 4.6	39.5 ± 12.1
n	5		18		3		12
F	0.46	0.28	4.63**	6.73***	4.23**	4.11**	1.49	2.37

*Significantly lower than asymptomatic group by Scheffé's multiple range test.
**P <0.01; ***P <0.001 by analysis of variance across symptom groups.

Table 3. F statistics for two-way analysis of variance of body mass index (BMI) and arm-muscle plus bone cross-sectional area (AMBA) by age group (40-, 50-, 60-, 70- years) and number of symptoms in Iban adults

	No. of symptoms Age group Interaction
Men (n = 261)
BMI	4.14**	4.19	0.85
AMBA	5.19**	7.44***	0.86
Women (n = 328)
BMI	0.66	9.48***	0.66
AMBA	0.59	9.05***	1.13

**P <0.01; ***P <0.001. Source: Strickland ± Ulijaszek (1993).

Good screening indicators have high sensitivity at high levels of specificity (Habicht, 1980); and when the cut-off point of BMI = 18.5 was examined in relation to the presence or absence of symptoms (Table 4), the sensitivity of BMI for identifying the ill was approximately 21% in men and 29% in women, with specificities respectively 90% and 87%. These values approximate, or are poorer than, those of the QUAC-stick and MUAC used in predicting mortality risk in children (Trowbridge & Sommer, 1981). Reducing the cut-off point to BMI = 17.0 resulted in sensitivities of 7% in men and 13% in women, and specificities of ±97% in both sexes. It could be concluded from this that the value of the BMI for identifying those reporting symptoms of poor health is likely to be limited, and that the higher cut-off point is better than the lower one because of the combined effects of sex and age.

Table 4. Sensitivity and specificity of the body mass index (BMI) for identifying individuals reporting symptoms

	BMI range
	³ 18.5	<18.5	³ 17.0	<17.0
Men (n = 447)
Asymptomatic	182	21	198	5
Symptomatic	193	51	228	16
Sensitivity	20.9%		6.6%
Specificity	89.7%		97.5%
Women (N = 564)
Asymptomatic	248	37	275	10
Symptomatic	198	81	242	37
Sensitivity	29.0%		13.3%
Specificity	87.0%		96.5%

The effects of the length of the fallow period

In order to obtain reasonable sample sizes, the anthropometric values are shown in Table 5 grouped by fallow class and the presence or absence or any reported symptoms rather than being grouped by the number of symptoms. After removing statistically the linear effects of age, two-way analysis of variance indicated that there were significant effects of fallow class on the anthropometric values for men. The strongest effects were on weight and the BMI with short fallow periods relating to greater, not lower, weights as predicted by some authors. In men, fallow class and illness independently affected AMBA; but illness did not affect the other anthropometric variables significantly. It is surprising that stature should also link with this overall pattern, although the level of statistical significance was low. These findings suggest that men engaged in short fallow farming may have benefited by comparison with those on long fallow farms.

Among women, height was the only measure related significantly to fallow class with women on average taller in those with farms using short fallow periods. Effects of illness on the BMI and, more significantly, on AMBA were independent of type of farm. Thus the overall pattern among women was distinct from that in men. Neither in women nor in men were any interaction effects observed. The data for both sexes could be interpreted as showing that AMBA may be a more sensitive functional index of ill health than the BMI.

Table 5a. Anthropometry of Iban men by fallow category and presence or absence of symptoms of illness (mean ± SD)

	Height (m)		Weigh (kg)		BMI (kg/m2)		AMBA (cm2)		n
Symptoms	Absent	Present	Absent	Present	Absent	Present	Absent	Present	Absent	Present
Fallow
0	1.580 ± 0.054	1.579 ± 0.064	53.7 ± 7.5	51.4 ± 7.3	21.5 ± 2.8	20.6 ± 2.3	50.2 ± 10.0	47.1 ± 8.9	54	70
1	1.588 ± 0.055	1.569 ± 0.057	55.4 ± 9.4	52.0 ± 7.7	21.9 ± 3.3	21.1 ± 2.9	53.2 ± 12.9	48.4 ± 10.0	50	58
2	1.595 ± 0.049	1.568 ± 0.052	52.3 ± 5.7	49.5 ± 8.3	20.5 ± 1.7	2.01 ± 2.8	50.3 ± 8 6	45 0 ± 8.3	40	47
3	1.554 ± 0.119	1.562 ± 0.058	50.9 ± 8.9	49.7 ± 6.9	20.9 ± 2.1	20.3 ± 2.0	47.6 ± 9 4	47 3 ± 10.5	59	69
F for
symptomaticity		0.02		3.37		3.25		5.54*
fallow		2.60*		4.60**		4.79**		2.89*
interaction		1.31		0.94		0.66		1.23
Covariate age:
B		-0 33		-0.36		-0.24		-0.31
t		-7.38***		-9.19***		-5.20***		-82***

*P <0.05; **P <0.01; ***P <0.001.

Table 5b. Anthropometry of Iban women by fallow category and presence or absence of symptoms of illness (mean ± SD)

	Height (m)		Weight (kg)		BMI (kg/m2)		AMBA (cm2)		n
Symptoms	Absent	Present	Absent	Present	Absent	Present	Absent	Present	Absent	Present
Fallow
0	1.482 ± 0.049	1.472 ± 0.055	50.9 ± 9 5	45.4 ± 9.2	23.1 ± 3.8	20.9 ± 3 7	44.6 ± 11.3	38.4 ± 8.1	68	82
1	1.479 ± 0.049	1.483 ± 0.049	48.3 ± 7.4	46.5 ± 9.3	22.1 ± 3.1	21.1 ± 3.7	41.7 ± 7.7	39.6 ± 9.1	80	65
2	1.479 ± 0.043	1.478 ± 0.059	48.2 ± 9.5	44.9 ± 9.4	22.0 ± 3.9	20.5 ± 3 5	43.5 ± 15.8	37.6 ± 7.1	56	54
3	1.461 ± 0.063	1.459 ± 0.056	47.3 ± 9.0	44.3 ± 10.8	22.0 ± 3.5	20.7 ± 4.3	41.7 ± 9.7	38.5 ± 9.8	81	78
F for
symptomaticity		3.67		1.53		4 95*		11.44***
fallow		4.53**		1.08		0.64		0.22
interaction		0.54		1.26		0.83		1.77
Covariate age:
B		-0.32		-0.41		-0 35		0.21
t		-8.03***		-10.58***		-8.80***		-5.14***

*P <0.05; **P <0.01; ***P <0.001.

Contents - Previous - Next