Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

TABLE 4. Communality values and Pearson correlation coefficients of variables after controlling for age, children under four years old (fit: 46.3%)

See notes to table3. AGE =age of the child.

The effect of sex on the association between dead siblings and nutrition status

The sex of the 62 children under four years old was not associated with differences in their Z scores. Of the 37 boys, 13 had a total of 17 older siblings (6 male, 11 female) who had died before the age of five. Of the 25 girls, 10 had a total of 14 older sibling deaths (12 male, 2 female). Of seven index children with two or more older sibling deaths, the deaths were all male in three cases, all female in two, and male and female in two.

Given these skewed sex distributions, the interrelations between the sex and nutrition status of index children were explored. Boys had a significantly better ZHDCRC if one or both older dead siblings were female: mean ZHDCRC +0.45 compared with -1.16 and -1.77 for male deaths and no deaths among siblings respectively (one-way: F ratio = 11.2, p<.OOOl). Differences between the ZHDCRC of boys with one and those with two female sibling deaths were not significant. In girls, the ZHDCRC was not related to the number or sex of the dead siblings. The association between the number of older sibling deaths and HAZ in the index children could not be explained by differences in sex.

FIG. 2. Means (± standard errors) for Z scores of head circumference for age in 62 children under four years old, by the number of deaths of older siblings (under five years old, including stillbirths; the "two-deaths" group includes one child with three sibling deaths).

The Z scores were calculated from a sex- and age-matched Dutch lowland reference population [30], with Z score = 0 for the median of the reference population. The difference between groups marked with an asterisk (*) is significant (p<.05; one-way [Tukey]).

Discussion

We conclude that in this stunted and underweight highland population, children under four years old had better nutrition status (evidenced by higher Z scores for height and head circumference for age) if older siblings had died before the age of five. There was socio-economic factors (the source of drinking water, the education of the father) associated with nutrition status as well, but they could not explain this effect. A similar effect could not be demonstrated for deaths among younger siblings. The effect of sibling deaths on the nutrition status of children under age four was strongest for the head circumference of boys, and in this case was associated with one or more deaths among older sisters. For height no association with sex was found.

Several observations suggest a causal relationship between deaths among older siblings and the improved nutrition status of surviving children under four. The association with older siblings (in contrast with younger ones) suggests a time relation between the siblings' death and the improved nutrition status of the index child. The number of deaths and the index child's nutrition status suggest a dose-response relationship. The positive association between the number of sibling deaths and height for age, and the fact that the associations between infant or perinatal deaths and the nutrition status of the surviving siblings are weaker than those for child deaths, further support this.

FIG. 3. Means (+ standard errors) for height-for-age Z scores in 62 children under four years old, by the number of deaths of older siblings (as in fig. 2).

The Z scores were calculated from a sex- and age-matched United States reference population [29], with Z score = 0 for the median of the reference population. Significance as in fig. 2.

The validity of the mortality data should be discussed. Non-response in the survey was low, and anthropometry was performed on all but two of the children under 14 years old in the study households. The boy: girl ratios in children under 14 and under four years of age were 1.17 and 1.48 respectively. It is unlikely that mistakes occurred in the assignment of sex to living children. Verifications of birth dates did not show bias in the age of offspring by sex as reported by parents.

The validity and consistency of the mortality data in the study population are discussed elsewhere [35]; from differences between the male: female birth and death ratios of children with different birth orders, and between different age groups, we inferred that underreporting of female neonatal deaths with birth order 3 or more occurred. Bolton [36] hypothesized that, among other factors, a cultural bias in favour of males and preferential female infanticide was a likely explanation for the high male: female ratios in Andean children. No cultural anthropological fieldwork was conducted, and the contribution of these factors to differences in the nutrition status of the children in the study households cannot be explored further here.

Selection of groups, such as differences in death rates in the study households with severely malnourished children, could pose another explanation of the results. No longitudinal data on growth and mortality in the children in the study households are available, and bias as a result of differences in the time of dying in malnourished children thus cannot be ruled out. If we eliminate the Z scores of the 8 children with the lowest Z scores from the 39 children without sibling deaths (simulating mortality due to severe malnutrition at a rate of more than 200 per 1,000 in this group), the null hypothesis that no relation exists between older dead siblings and nutrition status in the remaining 54 children still has to be rejected for ZHDCRC (one-way, p < .01). One-way analyses indicate a tendency to improved height and weight in the children with sibling deaths when a similar elimination of 8 of 39 children without dead siblings was carried out for these variables (one-way for HAZ, p = . 10).

Genetic factors and diseases (e.g., infection, iodine deficiency) could be put forward as alternative explanations for the results of our study. However. we do not conceive of any theory of disease that would result in the death of one child and better nutrition status for another child in the same household. Recently, an increased risk of death from measles has been reported for children with a sibling of the opposite sex [37]; however, measles is commonly associated with malnutrition in children [38].

We have no reason to believe that underreporting, selection of group, diseases or systematic inaccuracy of measurements due to scalp oedema or differences in the thickness of the skull and hair can explain the findings concerning head circumference. As head circumference has been shown to correlate with brain volume [39], with brain weight, protein, and DNA [40], and with cholesterol [41] in healthy and severely malnourished children, nutritional effects seem a much more likely explanation.

Physiologically, a causal relationship between dead siblings and the nutrition status of surviving children mediated by nutritional factors can be postulated. In the present study this relationship cannot be explained simply by the number of mouths to feed given a fixed amount of food, as household size was not correlated with the nutrition status of young children. The death of older siblings at a young age could indirectly benefit the foetus and newborn child by an improved nutrition status (i.e., decreased depletion of body stores) of the mother. In our study, data on birth weights were lacking, so this issue cannot be explored further. The physical growth of children in the United States living at high altitude is affected by altitude independently of nutritional factors [5, 42].

In the present study, mean Z scores for the head circumference, height, and weight of children under four years old with two deceased older siblings were not significantly less than those of the well-nourished United States children in the reference population. The number of children in this category was small (seven), and difference in sex between the index children and deceased siblings could be related to these observations. Nevertheless, these findings are in accordance with the view that marginal nutrition is the primary stressor for human growth retardation in the Andes of southern Peru [7, 8].

There is a close connection between food production and food consumption in the subsistence economy of Andean peasants, and the presence or absence of an extra young child affects this balance. Mothers caring for several very young children are less mobile, and their ability to work in agriculture and livestock raising is curbed. If no older children are contributing to the economy of the nuclear family, food production may be less in a situation of increased needs for the household as a whole. Ethnographic research has shown that heavily burdened mothers in peasant societies, having less opportunity to breast-feed, will wean their children at an earlier age [43].

The death of one or more children in peasant households in malnourished populations could alleviate the nutritional situation of children born later through several mechanisms, and might have a permanent effect on their growth. In the present study, greater perinatal and infant mortality in Aymara than in Quechua families could be related to a significantly greater mean ZHDCRC (a measure of brain size) in the Aymara than in the Quechua children. There were differences between the Aymara and the Quechua in the safety of drinking water, and these were associated with ZHDCRC as well. In the children under 14 years old, social factors (the education of the father, birth order, the amount of land held) and environmental factors (crowding, water supply) were associated with nutrition status. These factors, among others, could explain the deterioration of nutrition status with age, and may confound an association, if any, between the death of older siblings and the nutrition status of surviving children after the age of four.

The evidence for the impact of dead siblings on the nutrition status of surviving young children in this highland population adds a new element to current theories on the causes of malnutrition in young children and supports the importance of nutrition. The hypothesis that child mortality is related to improved nutrition status in surviving children has potentially important implications for health care policy. Reduction of high rates of perinatal, infant, and child mortality is a priority in many developing countries and is not generally thought to be related to decreased growth in surviving children. The impact of the composition of the peasant household, the sex distribution of offspring, intrahousehold food production and distribution, the energy expenditure of household members, and the nutrition status of women in their reproductive cycle on the relation between sibling deaths and the nutrition status of surviving children in different ecosystems in developing countries needs further study.

Appendix

The acronym PRINCALS stands for a non-linear principal components analysis through optimal scaling by means of alternating least squares [44]. For the data in the present study, comparisons of linear multiple regression analyses with PRINCALS showed a much higher fit for the latter type of analysis. There were differences in the levels of measurement and distributions of the variables, but these were accounted for by resealing. Requantifications of categories after optimal resealing were used to calculate a matrix of Pearson correlation coefficients among the variables.

In SPSS-X* output of PRINCALS, the principal components and component loadings of variables are not optimally projected in a horizontal or vertical dimension. Therefore, orthogonal rotation is required. The transformation used in the present study was a varimax rotation, carried out as follows: If

• f_ij is the matrix of the component loadings of i = 1. . . m components with j = 1 . . . n variables,
• h² is the communality of a variable, that is, the squared component loadings summed over i = 1. . .m components, and
• h_j² is the communalities for variables j= 1. . . n, then the algorithm

is maximized.

After rotation, pairs of components are identified by horizontal and vertical dimensions. Eigenvalues for components can be recalculated. It should be noted that the communalities do not change under orthogonal rotation, as component loadings are kept normalized with respect to the communalities. Variables with component loadings between -0.31 and +0.31 on any dimension explain less than 10% of its variance and are not considered strong determinants of that dimension in the present study.

Acknowledgements

The authors thank J. Laura for field assistance, and R. Th. J. Buve, H. W. A. Voorhoeve, W. F. L. Buschkens, and J. D. Speckmann for their comments on the manuscript.

References

1. de Meet K. Mortality in children among the Aymara Indians of southern Peru. Soc Sci Med 1988;26:253-58.

2. Frisancho AR. Growth and morphology at high altitude. In: Baker PT, Little MA, eds. Man in the Andes. Philadelphia, Pa, USA: Dowden, Hutchinson and Ross, 1976:180-207.

3. Beall CM, Baker PT, Baker TS, Haas JD. The effect of high altitude on adolescent growth in southern Peru. Hum Biol 1977;49:109-24.

4. Hoff C. Altitudinal variations in the physical growth and development of Peruvian Quechua. Homo 1974 ;24:8799.

5. Yip R. Binkin, NJ, Trowbridge FL. Altitude and childhood growth. J Pediatr 1988;113:486-89.

6. Gonzales G. Crespo-Retes 1, Rodríquez L, Vega L, Guerra-Garcia R. Secular change in growth of native children and adolescents at high altitude, Huancayo, Peru. Am J Phys Anthrop 1984;64:47-51.

7. Leonard WR. Nutritional determinants of high-altitude growth in Nuñoa, Peru. Am J Phys Anthrop 1989; 80:34152.

8. Leonard WR, Leatherman TL, Carey JW, Thomas RB. Contributions of nutrition versus hypoxia to growth in rural Andean populations. Am J Hum Biol 1990;2:61326.

9. Scrimshaw NS, Taylor CE, Gordon JE. Interactions of nutrition and infection. Geneva: WHO, 1968.

10. Suskind RM. Characteristics and causation of proteincalorie malnutrition in the infant and preschool child. In: Greene LS, ed. Malnutrition, behavior and social organization. New York: Academic Press, 1977:117.

11. Sommer A, Lowenstein MS. Nutritional status and mortality: a prospective validation of the QUAC stick. Am J Clin Nutr 1975;28:287.

12. Kielman AA, McCord C. Weight-for-age as an index of risk of death in children. Lancet 1978;1:1247.

13. Chen LC, Alauddin Chowdhury AKM, Huffman SL. Anthropometric assessment of energy-protein malnutrition and subsequent risk of mortality among preschool aged children. Am J Clin Nutr 1980;33:1836-45.

14. Unda de Morales E. Patrones socio-culturales que influyen en la salud del niño del altiplano. Puno, Peru: Instituto de Investigaciones pare el Desarrollo Social del Altiplano, 1982.

15. Morales R. Desarrollo y pobreza en Bolivia. La Paz. Bolivia: Editorial Mundy Color, 1984.

16. Ryan JG, Bidinger PD, Prahlad Rao N. Pushpamma P. The determinants of individual diets and nutritional status in six villages in southern India. Patancheru, Andra Pradesh: International Crops Research Institute of the Semi-arid Tropics, 1984.

17. Desai P. Standard KL, Miall WE. Socio-economic and cultural influences on child growth in Jamaica. J Biosoc Sci 1970;2:133-43.

18. Melville B. Williams M, Francis V, Lawrence O. Collins L. Determinants of childhood malnutrition in Jamaica. Food Nutr Bull 1988;10(1):43-48.

19. Victora CG, Vaughan JP, Kirkwood BR, Martines JC, Barcelos LB. Risk factors for malnutrition in Brazilian children: the role of social and environmental variables. Bull WHO 1986;64:299309.

20. Marchione TJ. Factors associated with malnutrition in children of western Jamaica. In: Jerome M, Pelto G. Kandel R. eds. Nutritional anthropology. New York: Redgrave Press, 1980:263-66.

21. Martorell R. Leslie J. Moock P. Characteristics and determinants of child nutritional status in Nepal. Am J Clin Nutr 1984;39:74-86.

22. Melville BF. Are land availability and cropping pattern critical factors in determining nutritional status? Food Nutr Bull 1988;10(2):44-48.

23. Food and Agricultural Organization. Integrating nutrition into agricultural and rural development projects. Rome: FAO, 1984.

24. Rosenhouse S. Identifying the poor. Washington, DC: World Bank, 1989.

25. Glewwe P. The distribution of welfare in Peru in 198586. Washington, DC: World Bank, 1988.

26. Mendizabal G. Rosello C. Extension of health service coverage in Puno, Peru. Bull Pan Am Health Org 1981 ;15: 121-30.

27. de Meer K, Bergman R. Kusner JS, Voorhoeve HWA. Differences in physical growth of Aymara and Quechua children living at high altitude in Peru. Am J Phys

Anthro 1993;90:59-75.

28. Jelliffe DB. The assessment of the nutritional status of the community. Geneva: WHO, 1966.

29. Sullivan KM, Gorstein J. ANTHRO documentation: software for calculating pediatric anthropometry. Geneva: WHO, Atlanta, Ga, USA: Centers for Disease Control, 1990.

30. Gerver WJM. Measurement of the body proportions in children. PhD thesis, University of Groningen, Groningen, Netherlands, 1988.

31. Voorhoeve HWA. A new reference for the mid-upper arm circumference? J Trop Pediatr 1990;36:256-62.

32. World Health Organization. International classification of diseases. Geneva: WHO, 1977.

33. Morris JA, Gardner MJ. Calculating confidence intervals for relative risks (odds ratios) and standardised ratios and rates. Br Med J 1988;296:1313-36.

34. SPSS-X* user's guide. Chicago, SPSS, 1988.

35. de Meer K, Bergman R. Kusner JS. Socio-cultural determinants of child mortality in southern Peru: including some methodological consideration. Soc Sci Med 1993;36:317-31.

36. Bolton R. High-altitude sex ratios: how high? Med Anthrop 1980;4:107-43.

37. Pison G. Aaby P. Knudsen K. Increased risk of death from measles in children with a sibling of opposite sex in Senegal. Br Med J 1992;304:284-87.

38. Tomkins A, Watson F. Malnutrition and infection: a review. Geneva: United Nations (ACC/SCN), 1989.

39. Bray OF. Shield WD, Wolcott GJ, Madsen JA. Occipitofrontal head circumference an accurate measure of intracranial volume. J Pediatr 1969;75:303-05.

40. Winick M, Rosso P. Head circumference and cellular growth of the brain in normal and marasmic children. J Pediatr 1969;74:774-78.

41. Rosso P. Hormazábal J. Winick M. Changes in brain weight, cholesterol, phospholipid, and DNA content in marasmic children. Am J Clin Nutr 1970;23:1275-9.

42. Yip R. Altitude and birth weight. J Pediatr 1987; 111:869-76.

43. Nerlove SB. Women's workload and infant feeding practices: a relationship with demographic implications. Ethnology 1974;13:125-214.

44. Gifi A. Non-linear multivariate analysis. Leiden, Netherlands: Department of Data Theory, University of Leiden, 1981.

Contents - Previous - Next