Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
David L. Pelletier
Abstract
The physiological synergism between malnutrition and infection has been recognized for some time, its implications have not been addressed in current child survival policies and programmes. A recent analysis of 28 epidemiologic studies of the malnutrition-mortality relationship concluded that the relationship is consistent across diverse world populations; a significant effect exists of mild to moderate malnutrition (MMM), as well as of severe malnutrition; and the effect is not simply due to confounding by socioeconomic factors or inter-current illness. In addition, evidence supports the hypothesis that malnutrition and infection have multiplicative effects on child mortality, not the additive effects implicitly assumed. An empirically based model suggests that by potentiating infection, malnutrition accounts for 56% of child deaths, 83% of which are due to MMM. These estimates are far higher than conventional figures that do not take account of potentiation and MMM.
Introduction
The documentation of the synergistic relationship between malnutrition and infection is one of the landmark publications in the recent history of nutritional sciences [1]. Drawing together extensive evidence from biomedical research, clinical research, and clinical practice, this synthesis formalized the vicious cycle view of malnutrition and infection that is now widely accepted in scientific and applied circles. Stated simply, this view holds that malnutrition adversely affects an individual's ability to resist and respond to infection, and infection adversely affects the individual's ability to use energy and nutrients from the diet. As illustrated best by measles [2], this physiological synergism can have devastating consequences for the individual.
Despite the broad recognition and acceptance of this phenomenon for over two decades, the implications of the synergism have yet to be reflected in policies designed to improve child survival in developing countries. Among the relevant questions are the following: Does the relationship between malnutrition and mortality differ across diverse populations? Is the effect of malnutrition on the risk of death limited to severe malnutrition, or is it also present in mild to moderate malnutrition (MMM)? Do malnutrition and infection have multiplicative effects on modality at the population level, as would be predicted by theory? Is it possible to estimate the percentage of child deaths due to malnutrition, to decide on the most appropriate mix of intervention strategies on global, country, and community bases? Does malnutrition have similar effects on many infectious diseases, or are its effects confined to the well-defined examples of measles and diarrhoea? Answers to these questions have profound implications for health and development policy and are now forthcoming from a synthesis and reanalysis of epidemiologic studies conducted over the past two decades.
Description of studies
A computer-assisted search of the literature, complemented by bibliographic branching, generated 28 research reports meeting the basic criteria for this meta-analysis. The criteria are that the studies were community-based in developing countries, rather than hospital-based; they employed prospective methods to relate child mortality to indicators of nutrition status; they used anthropometric methods to indicate nutrition status; and the target population was preschool children (under five years). The list of 28 reports is believed to be complete with respect to studies meeting these criteria.
The 28 reports actually refer to 21 separate studies representing populations in 11 countries. A clear bias exists in favour of Bangladesh, with 14 reports based on 7 studies. There are 11 reports from Africa (Guinea-Bissau, Senegal, Zaire, Uganda, Tanzania, Malawi, Yemen), 3 from Asia outside of Bangladesh (India, Indonesia, Papua New Guinea), and none from Latin America. All reports used one or more of the following anthropometric indicators: weight-for-age (WA), height-for-age, weight-for-height, and mid-upper arm circumference (MUAC). Further details on this sample of studies, methods, and findings are available in the full report [3].
Consistency across populations
The earliest of the prospective studies took place in India [4], Bangladesh [5,6], and Papua New Guinea [7] and established the basic finding that the risk of mortality is inversely related to anthropometric indicators of nutrition status. The generalizability of this finding was first called into question by the findings from Kasongo, Zaire [8]. The investigators found no association between anthropometric indicators and subsequent mortality. Discussion of possible population-specific relationships continues [8,9].
When the Kasongo study is examined in light of the entire set of prospective studies, it appears that the negative results are more likely due to methodological reasons. This is because more recent work in Africa found the expected inverse relationship between nutrition status and mortality, including in Guinea-Bissau [11], Senegal [12], Uganda [13, 14], Tanzania [15], and Malawi [16,17]. This is also confirmed by recent research from a different region of Zaire [18]. Close inspection of the Kasongo report [8] reveals that mortality was grossly under-enumerated (only about 20% of the expected number of deaths), and the anthropometric measurements were "obtained under conditions that are similar to operational conditions of screening in clinics" [8].
Thus, the overwhelming body of evidence supports the idea that the fundamental inverse relationship between nutrition status and mortality is consistent across populations. In fact, a surprising degree of consistency is observed even in the details of the relationship, as revealed by a subset of studies described in the following section.
Mild to moderate malnutrition
The perception that MMM may have no consequences for child mortality was created in large part by an early report [6] and the discussion that ensued. Empirically, the investigators observed elevated mortality among children with severe weight deficits (WA<65% of the international reference) but no consistent relationship above that threshold. A commentary on this finding presented results based on MUAC from an earlier Bangladesh study, which showed a sharp increase in mortality among the severely malnourished (MUAC < 12.0 cm) and a more modest elevation among children with moderate deficits (12-12.9 cm) [19]. Since then, many researchers confirmed the characteristic exponential relationship between mortality and anthropometric indicators [3], which seemingly supports the concept that anthropometric deficits are a serious concern only at the extremes of the distribution. This reinforced the theories of adaptation to malnutrition [20] and small-but-healthy [21], and seems to reinforce the widespread practice of screening for severe malnutrition in many supplementary feeding programmes.
With this historical perspective, it is of great interest to note that the results of Chen et al. [6] have not been replicated by prospective studies, including reports from the same area of Bangladesh, other areas of Bangladesh, other Asian countries, and several African countries. Figure 1 shows the results of seven other studies that employed similar methodologies and that can be compared in detail with those from Chen et al. [6]. The top panel depicts mortality in natural units, and the bottom panel depicts the log (base 10) of mortality. The dominant impression from the top panel is, indeed, the marked elevation in mortality below 60% WA, especially in Tanzania, Papua New Guinea, and Malawi, which have high mortality rates at any given WA. However, the figure also reveals a clear elevation in mortality even at moderate (60-69% WA) and mild (7079% WA) anthropometric deficits. The bottom panel accentuales this observation. It is interesting that the MUAC results described above [19] fit this pattern exactly, but at the time they were interpreted as confirming a threshold effect seen by Chen et al. [6].
These figures suggest, first, that the study by Chen et al. [6] produced unusual results for inexplicable reasons. They also reveal a clear elevation in mortality among children with MMM that is remarkably consistent across populations, albeit not as marked as that seen in severe malnutrition. It is important to note that this modest elevation in mortality results in a lower screening efficiency for this group, and one that is probably unacceptably low in practical settings. However, it still has relevance for broader pole icy formulation in light of the much higher prevalence of MMM compared with severe malnutrition. This is highlighted by the quantitative estimates of malnutrition's effect provided below.
Although the accumulated results of the malnutrition-mortality relationship are striking for their consistency, it can be hypothesized that the association is simply or largely due to statistical confounding. According to this hypothesis, malnutrition and mortality may co-occur in the same households simply because both are associated with poverty or low socio-economic status. It may be, for instance, that the malnutrition in those households is caused by poor nutrient intake and high disease exposure, whereas mortality may be caused by low immunization rates or inappropriate treatment of illness.
Another possibility is that low WA is a by-product of high disease exposure, and appears associated with mortality for this reason, but actually plays no causal role in mortality. Several studies examined the possibility of confounding. As reviewed elsewhere [3], all of them observed that a significant association between malnutrition and mortality persists even after controlling for confounding through various statistical techniques.
The multiplicative effects of malnutrition and morbidity
The foregoing sections suggest that the effects of malnutrition on mortality are consistent across populations, are found in MMM as well as severe malnutrition, and are not simply due to confounding by socio-economic factors or intercurrent illness. The eight studies represented in figure I are important for another reason. Their results confirm that physiological synergism [1] does have multiplicative effects on mortality at the population level.
As shown for six of these studies [23], a simple specification of the synergism is that exposure to disease is constant within any given population, but the fatality rate per exposure varies with the degree of malnutrition. If this is so, the risk of death for an individual child is related to the product (not the sum) of the probability of exposure to disease and the probability of being malnourished (i.e., it is a multiplicative model rather than an additive model). At the population level, it follows that the mortality rate should be related to the product of the burden of disease (exposure) and the prevalence of malnutrition. Note that one indicator of the burden of disease in a given population is simply the mortality rate among the well-nourished, because some proportion of well-nourished children will die of infections diseases at a baseline rate that is determined by the types of diseases present and the health care available for treating them.
The data shown in figure 1, especially the lower panel, conform precisely to this theoretical model of synergism. It shows that populations with high baseline mortality (mortality among the relatively well-nourished, reflecting the population's burden of disease) have a systematically higher response to malnutrition. (Terms such as "mortality response" and "increase in mortality" are used as a shorthand for attributable risk, which is the difference in mortality rates between the malnourished and the well-nourished.) For instance, the population in Punjab, India, had the lowest level of baseline mortality (2.8/1,000/year) and experienced an increase of 34/ 1,000/year in going from more than 80% to less than 60% WA. By contrast, Iringa, Tanzania, had the highest baseline mortality (23/1,000/year) and experienced an increase of 189/1,000/year in going from greater than 80% to below 60% WA. Under an additive model, Tanzania would have experienced roughly the same number of excess deaths as India (34/1,000/year) in going from the well-nourished category to the severely malnourished category. The parallelism in the eight lines reveals a consistent tendency across the studies in which the mortality response to malnutrition is proportional to the baseline mortality level. This was formally tested and confirmed [23, 24].
These results and the inferences drawn from them are important because the sample size in the only controlled intervention trial was not adequate for testing the multiplicative effects of malnutrition and morbidity [4]. Thus, the present results provide the only evidence currently available for testing this hypothesis.
These observations have important implications for conceptualizing the relationships among malnutrition, morbidity, and mortality; classifying causes of death; and planning actions to improve health and survival in developing countries. Specifically, malnutrition should not be viewed as a cause of death on its own; rather, it acts as a potentiation of existing infectious diseases, with the degree of potentiation proportional to the severity of malnutrition. Consequently, it is meaningless to ascribe a certain number of deaths to malnutrition alone, and it is grossly misleading to ascribe a certain number of deaths to infectious diseases alone, the latter being a particularly common practice.
In developing countries with high rates of malnutrition, the excessive number of deaths ascribed to diarrhoea, acute respiratory tract infections, measles, and other common infections places primacy on the proximate and clinically obvious cause, while ignoring the potentiating effects that severe malnutrition and, less obviously, MMM have on those diseases. For example, the 1993 world development report ascribes only 2.4% of disability-adjusted life years lost to protein-energy malnutrition, compared with 63% for common infectious diseases [25]. One consequence at the policy level may be the neglect of nutritional improvement as a broad strategy for reducing mortality due to infectious diseases.
Malnutrition and mortality: Quantifying the effects
Bearing in mind the potentiation paradigm of malnutrition's effects on mortality, the results shown in the lower panel of figure 1 indicate that the absolute level of child mortality can be accurately modeled simply as a function of baseline mortality (among those with WA > 80%), and the percentage of children falling in each of the grades of malnutrition below 80% of median. However, this observation has limited practical utility when stated in those terms, because most countries do not know the mortality level among those with WA above 80% of median. Thus, it would not be possible to estimate the contribution of malnutrition to child mortality in most populations.
An alternative formulation relies on the fact that the relative risk (RR) of mortality at various grades of WA can be calculated from the data in figure 1. (Relative risk is the mortality within a given grade of malnutrition divided by morbidity among the well-nourished, in this case children with WA > 80%.) The RRs are 8.4 for severe (WA < 60%), 4.6 for moderate (60-69%), and 2.4 for mild (70-79%) malnutrition. The contribution of malnutrition to child mortality through its potentiating effects on infections disease can be calculated using the standard epidemiological statistic of population attributable risk (PAR), which combines the RR estimates with estimates of the prevalence of low WA in a given population. The methodology is fully described and tested elsewhere [24]. This section presents the results when it is applied to 53 countries for which suitable anthropometric data have been published [26].
Figure 2 shows the percentage of child deaths due to the potentiating effects of malnutrition on disease in each of 53 countries. The total PAR is divided into the portion due to severe malnutrition (WA < 60%) and that due to MMM (WA 60-79%). Using the average for all 53 countries, the results indicate that 56% of all child deaths are due to the potentiating effects of malnutrition on disease, of which 83% are due to MMM. The values for any given country vary in proportion to its prevalence of low WA. Among these countries, the range for total PAR is from about 15% in Paraguay to about 85% in India. The percentage due to MMM varies from zero among several countries where severe malnutrition is extremely rare, to a high of 68% in India.
These estimates are remarkably close to those arising from the inter-American investigation of childhood mortality over two decades ago [27]. That study was based on Latin American and selected North American samples, and used clinical and verbal autopsy methods to establish a certain cause of death. As in the present study, it reported that 54% of child deaths (2-4 years) in the Latin American countries had malnutrition as an underlying or associated cause, with about 15% severe malnutrition. Among infants (0-11 months), about one-fourth to one-third of all deaths had malnutrition as an underlying or associated cause. The estimates for infants are lower than for children because a large proportion of neo-natal deaths are due to congenital and obstetrical complications. Thus, the inter-American investigation, using a different methodology than the prospective studies shown in figure 1, confirmed the quantitative estimates of the effect of malnutrition on child mortality and demonstrated that the Latin American results are similar to those in Africa and Asia. In addition, they confirmed that conventional methods for classifying cause of death [25] underestimate the importance of malnutrition by a factor of 8- to 10-fold.
Effects of malnutrition on different causes of death
The consistency in the slope of mortality on WA shown in figure 1 is striking in light of the differences in ecological circumstances and associated dis ease exposure, as well as cultural differences across studies. For instance, the Papua New Guinea research was conducted in the highlands where acute respiratory infection was a major cause of death and malaria was presumably absent; yet, it has a slope similar to those for Bangladesh, Tanzania, and Malawi where diarrhoea and malaria are combined with acute respiratory infection as major diseases. This empirical observation of relative uniformity in slope across populations suggests that malnutrition may potentiate the effects of many or all of the common infectious diseases.
FIG. 2. Deaths due to potentiating effects of malnutrition on infectious diseases (Source: ref. 30)
Somewhat more direct confirmation of this is provided by three of the prospective studies as well as the inter-American investigation. The three prospective investigations, two in Bangladesh, one in Uganda, collected verbal reports of symptoms at the time of death and thus were able to estimate the RR of death due to malnutrition for each symptomatic cause separately [3]. All three showed elevated RRs for diarrhoea and measles, the only diseases reported separately. In addition, in Uganda, the RRs were elevated for fever and acute respiratory infection. Fever is usually assumed to be due to malaria in African settings with endemic malaria, but the workers in Uganda did not collect detailed clinical data to confirm this. The two Bangladesh studies grouped fever and acute respiratory infection with other infections, and found elevated RRs for that combined category. Thus, these three studies are consistent with the evidence shown in figure 1, that malnutrition may have a potentiating effect on many or all infectious diseases.
The inter-American investigation, based on the 13 Latin American samples, found that malnutrition was an associated cause in 47% of all deaths of children under age five years (excluding neonatal deaths). It was an associated cause in roughly 60% of deaths due to diarrhoea, measles, and other infections and parasitic diseases, compared with about 32% in deaths due to respiratory disease or other causes. The latter figure was no higher than that seen in the "other" category, which represents a pseudo-control category.
Finally, recent work in Zaire challenged the notion that MMM (subclinical) is associated with elevated child mortality, and suggested that a major reason may be the uniformity with which malaria kills children regardless of nutrition status [18]. It is difficult to interpret this report, for two reasons. First, it took place in an area that has been the target of an integrated health and development programme for the past 20 years. As such, immunization levels are higher and diarrhoea is lower than in most parts of Africa, and access to curative care is presumably greater. The authors suggest that this may help exe plain the absence of an overall effect of MMM on mortality. If so, the results may have limited generalizability to areas that do not share in those characteristics.
A second difficulty relates to the analytical strategy. In contrast to other prospective studies, the authors removed from the MMM sample any children showing clinical signs of malnutrition, such as muscle wasting (by inspection or palpation) with or without loss of subcutaneous fat, visible skeletal structures, and hanging skin. Pitting oedema was also considered a clinical sign. The difficulty in interpretation arises from the fact that children with any of these signs were all considered severely malnourished, but did not come exclusively from the category with WA below 60%. Results published separately [28] showed that roughly 20% of all children below 80% WA showed these signs and were excluded from the analysis. This makes it extremely difficult to compare this with the other prospective studies. It appears that the possibility of disease-specific effects of malnutrition is an issue that deserves further evaluation, and it may be most amenable to study through case control analysis of clinical data.
Summary and policy implications
Analysis of 28 reports from 11 developing countries leads to a number of clear conclusions concerning malnutrition and child mortality. First, the inverse association between nutrition status and mortality is a consistent finding across diverse world populations, thereby contradicting the earlier suggestion that the results may be region or population specific.
Second, mortality is elevated even among children with MMM, contrary to the widely held view that the effects are confined to the severely malnourished. The association does not appear to be due simply to the confounding effects of socio-economic factors and intercurrent illness.
Next, the long-recognized physiological synergism between malnutrition and infection leads to the prediction that these two factors should have multiplicative effects on mortality at the population level. This prediction is fully consistent with the results from eight epidemiologic studies of malnutrition and mortality. Malnutrition is observed to multiply the number of deaths caused by infectious disease, rather than acting in a simple additive fashion. The effects are strong and consistent across populations.
Applying the results of these 8 studies to a larger set of 53 countries for which suitable anthropometric data exist, it is found that malnutrition, through its potentiating effects on infectious diseases, contributes to 56% of all child deaths. This is roughly 8 to 10 times higher than conventional estimates that ignore the potentiating effects of malnutrition on disease and the effects of MMM. Of the malnutrition related deaths, 83% are due to MMM as opposed to severe malnutrition, which is also much higher than commonly recognized.
Finally, the quantitative relationship between male nutrition and mortality is remarkably consistent across eight populations representing diverse ecological, disease, and cultural environments. Based on a smaller number of more detailed studies, malnutrition potentiates deaths due to several infectious diseases. However, some researchers raise doubt concerning its effects on malaria and respiratory infection, leading to the conclusion that further investigation into this question is required.
These results have a number of implications for health policy and for equity-oriented development policy more generally. The PAR estimates suggest that programmes directed at screening and treating only the severely malnourished would have the potential to prevent only about 17% of malnutrition related deaths (using the average figure for the 53 countries included here), representing only about 10% of all child deaths (i.e., 0.17 x 0.56). The actual preventable fraction in current health facilities and child survival projects is likely to be even smaller than this because existing interventions are not 100% effective. Greater effects could be achieved by pursuing policies and programmes that attempt to shift entire distribution of nutrition status, thereby improving MMM, which accounts for most of the nutrition-related deaths. In addition, because of the multiplicative effects arising from the synergism between malnutrition and morbidity, the largest effects could be expected in populations with the highest exposure to disease and/or highest prevalence of malnutrition.
The results suggest that improving the nutrition of populations is expected to reduce mortality due to several diseases simultaneously, even if exposure to those diseases remains unchanged. This is opposed to a disease-focused approach that employs separate interventions to prevent or treat each disease. Clearly, it is desirable to improve nutrition status as well as to reduce disease exposure, but the suggestive across-disease effects of improved nutrition should be taken into account when attempting to design the most cost-effective interventions in the face of resource constraints.
Finally, the focus on the malnutrition-morbidity synergism has led to neglect of the other important variable in child survival policies, namely, curative health care. Analysis of gender differentials in the effects of malnutrition on mortality in these prospective studies [29] confirmed that the mortality consequences of the powerful synergism can be ameliorated by access to health care, as seen among males in Bangladesh. However, it also reveals that failure to address the synergism, by either reducing exposure to disease or improving nutrition status, places children at high risk for mortality when access to health care is limited, as seen among females in Bangladesh. This suggests that the relative risk of mortality (and percentage of deaths) due to malnutrition's potentiating effects may be higher when access to health care is limited and lower when it is improved.
This has implications for the cost-effectiveness and sustainability of alternative approaches to reducing child mortality. Specifically, it suggests that calculations could be performed for various combinations of health care improvement, nutritional improvement, and reductions in disease exposure, to determine the most cost-effective strategy in the medium term. However, the sustainability of strategies weighted heavily toward curative care is likely to be lower than those giving more attention to reducing malnutrition and disease exposure, and such strategies would do little to improve social equity.
Acknowledgements
The author acknowledges the contributions of Drs. Jean-Pierre Habicht, Edward A. Frongillo, Jr., and Dirk Schroeder who have collaborated in the research described here. The support of the Rockefeller Foundation, the Thrasher Research Fund and UNICEF is also gratefully acknowledged.
References
1. Scrimshaw NS, Taylor CE, Gordon JE. Interactions of nutrition and infection. Geneva: WHO, 1968.
2. Morley D. Pediatric priorities in the developing world. London: Butterworth and Co., Ltd. 1973.
3. Pelletier DL. The relationship between child anthropometry and mortality in developing countries: implications for policy, programs and future research. J Nutr 1994;124(10S):2047-81.
4. Kielmann AA, McCord C. Weight-for-age as an index of risk of death in children. Lancet 1978;1:1247-50.
5. Sommer A, Loewenstein MS. Nutritional status and mortality: a prospective validation of the QUAC stack. Am J Clin Nutr 1975;28:287-92.
6. Chen LC, Chowdhury AKMA, Huffman SL. Anthropometric assessment of energy-protein malnutrition and subsequent risk of mortality among preschool aged children. Am J Clin Nutr 1980;33:183645.
7. Heywood P. The functional significance of malnutrition: growth and prospective risk of death in the highlands of Papua New Guinea. J Food Nutr 1982;39(1): 13-19.
8. Kasongo Project Team. Anthropometric assessment of young children's nutritional status as an indicator of subsequent risk of dying. J Trop Pediatr 1983;29:69-75,
9. Bairagi R. Why mortality-discriminating power of anthropometric indicators differs among populations. J Trop Pediatr 1985;31:63-4.
10. Ewbank DC, Gribble JN. Effects of health programs on child mortality in sub-Saharan Africa. Washington, DC: National Academy Press, 1993.
11. Smedman L, Sterky G. Mellander L, Wall S. Anthropometry and subsequent mortality in groups of children aged 6-59 months in Guinea-Bissau. Am J Clin Nutr 1987;46:369-73.
12. Briend A, Bari A. Critical assessment of the use of growth monitoring for identifying high risk children in primary health care programmes. Br Med J 1989;298; 1607-11.
13. Vella V, Tomkins A, Borghesi A, Migliori GB, Ndiku J. Adriko BC. Anthropometry and childhood mortality in northwest and southwest Uganda. Am J Publ Health 1993;83:1616-8.
14. Vella V, Tomkins A, Ndiku J. Marshal T. Cortinovis 1. Anthropometry as a predictor for mortality among Ugandan children, allowing for socio-economic variables Eur J Clin Nutr 1994;48(3):189-97.
15. Yambi O. Nutritional status and the risk of death: a prospective study of children six to thirty months old in Iringa region, Tanzania. Doctoral degree thesis, Cornell University, Ithaca, NY, USA, 1988.
16. Lindskog U. Lindskog P. Carstensen J. Larsson Y. Gebre-Medhin M. Childhood mortality in relation to nutritional status and water supply: a prospective study from rural Malawi. Acta Paediatr Scand 1988;77:260-8.
17. Pelletier DL, Low JW, Johnson FC, Msukwa LAH. Child anthropometry and mortality in Malawi: testing for effect modification by age and length of follow-up and confounding by socioeconomic factors. J Nutr 1994;124(10S):2082-105.
18. Van Den Broeck J. Eeckels R. Uylsteke J. Influence of nutritional status on child mortality in rural Zaire. Lancet 1993;341:1491-5.
19. Trowbridge FL, Sommer A. Nutritional anthropometry and mortality risk. Am J Clin Nutr 1981;34:2591-2.
20. Lipton M. Poverty, undernutrition and hunger. Staff working paper no. 597. Washington, DC: World Bank, 1983.
21. Seckler D. Small but healthy: a basic hypothesis in the theory, measurement and policy of malnutrition. In: Sukhatme PV, ed. Newer concepts in nutrition and their implications for policy. Pune, India: Maharashtra Association for the Cultivation of Science Research Institute, 1982.
22. Cogill B. Ranking anthropometric indicators using mortality in rural Bangladesh children. Master's degree thesis, Cornell University, Ithaca, NY, USA, 1982.
23. Pelletier DL, Frongillo EA Jr, Habicht JP. Epidemiologic evidence for a potentiating effect of malnutrition on child mortality. Am J Publ Health 1993;83:1130-3.
24. Pelletier DL, Frongillo EA, Schroeder DG, Habicht JP. A methodology for estimating the contribution of malnutrition to child mortality in developing countries. J Nutr 1994;124(10S):2106-22.
25. World Bank. World development report: investing in health. New York: Oxford University Press, 1993.
26. UNICEF. Child malnutrition: progress toward the world summit for children goal. Statistics and Monitoring Section. New York: UNICEF, 1993.
27. Puffer RC, Serrano CV. Patterns of mortality in childhood. Scientific publication no. 262. Washington, DC: Pan American Health Organization, 1973.
28. Van Den Broeck J. Assessment of child health and nutritional status in a rural tropical area. Doctoral degree thesis, Katholieke Universiteit, Leuven, Belgium, 1994.
29. Pelletier DL. The relationship between malnutrition, morbidity and child mortality in developing countries. Ithaca, NY, USA: Cornell Food and Nutrition Policy Program, 1994.
30. Pelletier DL, Frongillo EA Jr, Schroeder DG, Habicht JP. The effects of malnutrition on child mortality in developing countries. Bull WHO 1995;73(4):(in press).
![FIG. 1. Relationship between child mortality and weight-for-age as a percentage of international median. The Chen et al. study is labelled "Matlab 1d" in this figure and is based on a reanalysis by Cogill in 1982 [22]](8F163E07.GIF){kind=link}
