The Assessment and Treatment of Severe Adult Malnutrition

THE ASSESSMENT AND TREATMENT OF SEVERE ADULT MALNUTRITION

Thesis submitted to the University of

London in partial fulfilment of the degree

of MD

D f Steve Collins MB BS BSc

Address for correspondence:

Oleuffynon, Old Hall, Llanidloes Powys SY18 6PJ Wales, UK

Tel: +44 (0) 1686 413989 E-mail: [email protected] ProQuest Number: 10016057

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest.

ProQuest 10016057

Published by ProQuest LLC(2016). Copyright of the Dissertation is held by the Author.

ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 Abstract

This thesis examines the assessment and treatment of severe adult starvation during famine. The author collected data from 573, 98 and 1059 severely malnourished adults, admitted to therapeutic feeding centres in Baidoa (Somalia) during 1992-3, Ayod (Sudan) during 1993 and Melanje (Angola) during 1993-4. The data collected are unique, recording recovery from extremes of adult starvation hitherto undocumented in the medical literature. All the centres were very rudimentary in nature, with no beds, running water, electricity or equipment for special investigations. Mortality rates in the Somalia centre were 21%; war disrupted the collection of outcome data in die centres in Sudan and Angola. The thesis assesses the relative merits of Body Mass Index, Middle Upper Arm Ciccumference and clinical signs, for screening adult admissions into therapeutic feeding centres. The analysis demonstrates that the use of Body Mass Index is inappropriate for this role and the discussion proposes a combination of clinical signs and Middle Upper Arm Circumference as an alternative.

The thesis also examines the effect on rehabilitation of two diets differing primarily in their protein content (one diet with 156g protein and 16.5Mj of energy day^ , the other with Big of protein and 16.5Mj of energy day'^). Twenty five percent of oedematous patients given the lower protein diet during the initial phase of treatemnt recovered, compared with only 48% who received the higher protein diet during the initial phases of treatment. This is a three-fold decrease in mortality amongst this group of patients.

The extreme levels of social disintegration, violence and death during the fieldwork prevented the execution of prospective highly controlled research. This is always true of the extreme famines where severe acute adult malnutrition is commonest and such difficulties have discouraged scientists. This is a sad state of affairs, as the extreme human suffering that occurs during famine has not received sufficient scientific attention. There is still a great need for research to assist the assessment and treatment of severe adult malnutrition. Although the conclusions of this thesis are tentative, in the absence of other data examining screening cut-offs based upon evidence gathered during famine, these are the best we have. TABLE OF CONTENTS

Abstract______ii Table of Contents...... iii List of figures...... vü List of tables...... viii Glossary ...... ix Acknowledgements ...... xi Chapter 1 Foreword - The thesis______12 Section 1.01 Objectives ------12 Section 1.02 Organisation of this thesis ------13 Section 1.03 The originality and the contribution to medicalpractice. ------13 Chapter 2 Introduction______15 Section 2.01 Definitions ------15 2.01(a) “Acute and chronic” malnutrition...... 15 2.01(b) “Primary” malnutrition...... 17 2.01(c) Severe adult malnutrition...... 17 2.01(d) A working definition of severe adult malnutrition...... 19 Section 2.02 The epidemiology of severe adult malnutrition ------19 Section 2.03 Thefocus of emergency reliefprogrammes ------24 Section 2.04 Research into severe adult malnutrition ------25 2.04(a) Primary acute severe adult malnutrition...... 25 2.04(b) Secondary adult malnutrition ...... 27 2.04(c) Indicators for screening: a research priority ...... 28 Chapter 3 Background to the studies and core methods ______31 Section 3.01 Background ------31 3.01(a) Concern Worldwide ...... 31 3.01(b) Concern adult feeding programmes where data were collected ...... 32 Section 3.02 Methods ------36 3.02(a) Subjects...... 36 3.02(b) Measurements, equipment and precision ...... 36 3.02(c) Clinical data collection and coding ...... 37 3.02(d) Data analysis ...... 38 Section 3.03 Methodological issues ------41 3.03(a) Constraints...... 41 3.03(b) Subject selection...... 42 3.03(c) Controls...... 43 3.03(d) Measurement errors and training ...... 43 Chapter 4 Study one; Is BMI a useful tool for the assessment of severe adult malnutrition during famine?______47 Section 4.01 A im s ------47 Section 4.02 Introduction ------47

III Section 4.03 Uterature review ------48 4.03(a) BMI in the assessment of chronic malnutrition in adults ...... 48 4.03(b) BMI and body composition ...... 49 4.03(c) Determinants of BMI...... 50 4.03(d) The use of BMI to screen severely malnourished adults ...... 56 4.03(e) Previous investigations into the limits of human response to starvation ...... 57 4.03(f) Triage and BMI limits of human adaptation to starvation ...... 64 Section 4.04 methods ------65 4.04(a) Subjects...... 65

4.04(b) Treatment in the centre ...... 6 6 4.04(c) Discharge criteria ...... 67 Section 4.05 Results ------67 4.05(a) General inpatient data ...... 67

4.05(b) Age and BMI ...... 6 8 Section 4.06 BM I data. ------70 4.06(a) BMI and the risk of mortality ...... 73 4.06(b) The lowest BMI compatible with life ...... 77 4.06(c) Famine oedema and the risk of mortality ...... 78 Section 4.07 Discussion ------80 4.07(a) The BMI limit in human starvation ...... 80 4.07(b) Sex difference in adaptation ...... 82 4.07(c) Famine oedema ...... 83 4.07(d) BMI and the prediction of mortalit}" ...... 84 4.07(e) The use of BMI during severe famine ...... 85 4.07(^ Adult therapeutic care...... 87

4.07(g) Conclusions...... 8 8 Chapter 5 Study two: Clinical assessment of severe adult malnutrition ______90 Section 5.01 Intwduction ------90 Section 5.02 aim ------90 Section 5.03 Background literature ------90 Section 5.04 Subjects and methods ------91 5.04(a) Subjects...... 91 5.04(b) Methods ...... 92 Section 5.05 Results: ------95 5.05(a) Clinical signs at presentation and prognosis ...... 95 Section 5.06 Dismssion ------99 Chapter 6 Study three: MUAC and the assessment of severe adult malnutrition ______105 Section 6.01 A im ------105 Section 6.02 Intwduction ------105 Section 6.03 Backgwund literature ------106 6.03(a) Body Mass Index (BMI) ...... 106 6.03(b) Middle upper arm circumference (MUAC) ...... 106 Section 6.04 Subjects and method ------109 6.04(a) Subjects...... 109

IV 6.04(b) General methods ...... I l l 6.04(c) Analysis ...... 111 6.04(d) Results...... 112 Section 6.05 Discussion ------125 6.05(a) SH/S ratio and BMI ...... 125 6.05(b) The effect of oedema ...... 126 6.05(c) The effect of acute malnutrition...... 127 6.05(d) BMI and MUAC in extreme malnutrition ...... 131 6.05(e) The relationship between MUAC and BMI in SAM ...... 131 6.05(f) The use of MUAC in adults...... 133 Section 6.06 Conclusion ------134 Chapter 7 Study four: The dietary treatment of severe adult malnutrition ______137 Section 7.01 A im ------137 Section 7.02 Intwduction ------137 Section 7.03 Subjects and Methods ------139 7.03(a) Subjects...... 139 7.03(b) Methods ...... 139 7.03(c) The rehabilitation diets ...... 139 7.03(d) Analysis of mortality ...... 142 7.03(e) Analysis of the effect of HP and LP diets on rates of weight change and the loss of oedema during rehabilitation ...... 142 Section 7.04 Results ------144 7.04(a) General patient data ...... 144 7.04(b) Effect of diet on mortality ...... 144 7.04(c) Loss of oedema and gastro-intestinal function ...... 144 7.04(d) Weight change during rehabilitation ...... 145 Section 7.05 Discussion ------147 7.05(a) The importance of lower protein diets during the initial treatment of severe malnutrition in adults...... 147 7.05(b) Rates of weight change during the recovery phase of rehabilitation ...... 149 7.05(c) Comparison with children ...... 150 Section 7.06 Conclusions ------151 Chapter 8 Conclusions ______152 Section 8.01 Screening severely malnourished adults duringfamine reliefprogrammes ------152 8.01 (a) Admission criteria for TFCs ...... 153 8.01(b) Admission to Supplementary Feeding Programmes ...... 153 Section 8.02 The treatment of severely malnourished adults duringfamine reliefprvgrammes ------154 Section 8.03 Where do we go fwm here? future research needs: ------154 8.03(a) Defining functional cut-offs...... ;.154 8.03(b) Patterns and prevalence of adult under-nutrition during famine ...... 155 8.03(c) Practicality of measurements and calculations ...... 155 8.03(d) Adjusting for differences in body shape ...... 155 8.03(e) The use of MUAC to monitor recovery from under-nuttition ...... 155 8.03(f) More data on differentiating between secondary and primary under-nuttition ...... 156 8.03(g) The aetiology, significance and treatment of famine oedema and ascites ...... 156 Section 8.04 Commmity therapeutic can ------156 Chapter 9 Appendices ______157 Section 9.01 Appendix 1 ------157 Section 9.02 Appendix 2 ------160 Section 9.03 Appendix 3 ------161 Section 9.04 Appendix 4 ------164 9.04(a) The correction of BMI using the Cormic Index (SH/S) ...... 164 References______166 Supplementary material ______182 1. Collins, S. 1993, "The need for adult therapeutic can in emergency feediugproprammes”, JAMA, vol.270 #5,pp.637-638. ------1 8 2 2. Collins, S. 1995, "The limit of human adaptation to starvation". Nature Medicine, vol.1, no.8, pp.810^814.------182 3. Collins, S. 1996, "Using middle am circumfennce to assess severe adult malnutntion during famine", JA M A , vol.276, no.5, pp.391-395. ------182 4. Collins, S., Myatt, M., <0° Golden, B. E. 1998, "Dietary tnatment of severe malnutrition in adults", American Journal o f Clinical Nutrition, vol.68,pp. 193-199.------182 5. Collins, S. Myatt, Ai. Aug 2000, "Short term prognosis in seven adult and adolescent malnutrition. " JAMA, 284, no 5,pp 621 - 626 ------182 6. Collins, S., Dujjield, A ., Myatt, M. Aug 2000 "Assessment of nutritional status in emergency afjlicted populations -Adults", ACC/SCN, RNISsuppL ------182

VI LIST OF FIGURES

Number Page

Figure 1 ...... 20 F igure 2 ...... 23 F ig u r e s ...... 40 F igure 4 ...... 45 FIGURES...... 51 F igure 6 ...... 53 F ig u r e ?...... 54 F ig u r e s ...... 56 F igure 9 ...... 59 F igure 1 0 ...... 68 F igure 1 1 ...... 69 F igure 1 2 ...... 74 F igure 1 3 ...... 75 F igure 1 4 ...... 76 F igure 1 5 ...... 77 F igure 1 6 ...... 79 Figure 1 7 ...... 85 F igure 1 8 ...... 98 F igure 1 9 ...... 99 F igure 2 0 ...... 108 Figure 2 1 ...... 116 F igure 2 2 ...... 117 F igure 2 3 ...... 120 F igure 2 4 ...... 123 Figure 2 5 ...... 124 F igure 2 6 ...... 132 F igure 2 7 ...... 146 F igure 2 8 ...... 160 Figure 2 9 ...... 161 F igure 3 0 ...... 163 F igure 3 1 ...... 163 F igure 3 2 ...... 164

VII LIST OF TABLES

Number Page

T a b le 1 22 T a ble 2 36 T a b le 3 60 T a ble 4 61 T a b le 5 72 T a ble 6 84 T able 7 92 T ab le 8 95 T a ble 9 96 T a b le 10 114 T a b le 11 115 T a ble 12 118 T a ble 13 118 T able 14 121 T a ble 15 122 T able 16 126 T a ble 17 130 T a b le 18 130 T a ble 19 141 T a ble 20 141 T a ble 21 143 T a ble 22 157 T a b le 23 157 T a ble 24 157 T a ble 25 158 T a ble 26 159 T a ble 27 162 T a ble 28 162

VIII GLOSSARY

AED Acute energy deficiency AMA Arm muscle area BMI Body mass index CAMA Corrected arm muscle area GDC Centre for disease control CED Chronic energy deficiency DPA Dual photon absorptiometry DXA Dual-energy x-ray absorptiometry FFM Fat free mass FM Fat mass HEM High-energy milk Ht-for-Age Height for age ICRC International committee of the Red Cross IDA International Drug Agency IDECG International Dietary Energy Consultative Group IDP internally displaced person IFRC International federation of the Red Cross MDM Medecin Du Monde MPLA Movement Patriotic Liberation Angola MSF Médecins Sans Frontières MUAC Middle upper arm circumference NGO Non governmental organisation OLS Operation Lifeline Sudan PEM Protein Energy Malnutrition SAM Severe primary adult malnutrition SFC Supplementary feeding centre SH/S Sitting height : stature ratio (Cormic index) SPLA Sudan people’s liberation army TFC Therapeutic feeding centre TSF *Skin fold thickness UNITA The National Union for the Total Independence of Angola WFP World food programme Wt-fot-Ht Weight for height

ACKNOWLEDGEMENTS

I would like to tliank the Concern staff and the patients in Baidoa, Ayod and Melanje who made this study possible. Ben Rigby for his help with data coding and entry, and Michael Golden, Hannah Scrase, Jeremy Shoham, Keith Sullivan for their advice and encouragement. Special mentions must go to Barbara Golden for her indispensable advice both in the field and during the writing up; Mark Myatt for his great help in data analysis and advice about statistics and Andrew Tomkins for his advice and support as my supervisor. My thanks also to Matanta whose help and friendship were vital for the Angolan study.

I also thank Concern Worldwide who employed me for the first three years during which time the data presented in this thesis were collected.

XI Chapter 1 Foreword - The thesis

Section 1.01 Objectives

Severe adult malnutrition is commonly encountered during famine-relief operations. There have however, been few studies into the problems associated with treating severely malnourished adults during famine and consequendy, little guidance is available to field workers thus engaged. There is at present, no definition or classification of acute adult malnutrition and no specific treatment guidelines for the condition. This makes the screening and selection of admissions into tlierapeutic feeding centres and the dietary treatment of those admitted, problematic. The objective of this work is to improve the processes of patient selection into adult Therapeutic Feeding Centres and the dietary management of those admitted. The tliesis aims to:

• Clarify the definition of severe adult malnutrition and propose a working definition that can be used to guide future research

• Assess the usefulness of different indicators of adult nutritional status

• Assess the effectiveness of two dietary regimes for treating severely malnourished adults during an emergency relief programme.

Tlie core data presented were collected from severely malnourished adults during the height of famines in Somalia, South Sudan and Angola.

The thesis falls into two distinct parts. The first, comprising three studies, examines the definition of severe adult malnutrition and the selection of inpatients into adult therapeutic centres. As a starting point, the hypothesis that: A Body Mass Index (BMI) is a useful indicator for the selection and assessment of malnourished adults presenting to an adult therapeutic centre during the height of a major emergency, is examined. The first study investigates the risks of mortality associated with low BMI. It examines whether there is a lower limit of BMI below which recovery is not possible and whether there are other BMI thresholds associated with a markedly worse prognosis. The use of BMI as the core indicator in this study reflects the recognition that BMI is useful in the assessment of chronic protein energy malnutrition in adults. The second study investigates the

12 risks associated with clinical signs in severe adult malnutrition. The third study attempts to compare the efficiency of Middle Upper Arm Circumference (MUAC) as an alternative indicator of severe adult malnutrition.

The second part of the thesis examines the treatment of severe adult malnutrition during famine- relief programmes. The study was prompted by the recognition that severely malnourished adults in the Concern adult therapeutic centres, when given the standard recovery diet suffered extremely high mortality rates. The hypothesis that: A protein intake little above maintenance is more appropriate in the initial treatment of severe adult malnutrition than diets with higher levels of protein, is examined. This hypothesis reflects recent successes witli the use of lower-protein rehabilitation diets in the treatment of severely malnourished children in hospitals. The study compares mortality and rates of weight gain in two groups of patients given different levels of protein during their treatment.

Section 1.02 Organisation of this thesis

The thesis consists eight chapters. The first contains this foreword. The second presents an outline of the importance and relevance of severe acute adult malnutrition, highlighting the relative neglect of the condition during the past 45 years. The third describes the background, general methods and relevant methodological issues of relevance to this study. Chapters 5 — 8 contain the bulk of the thesis, each describing an individual study, including aims, a brief literature review, specific methods, results and a discussion. Chapters 4, 5 and 6 are concerned with the definition, classification and assessment of severely malnourished adults during famine and chapter 7 with their treatment. Chapter 9 draws together these various studies providing overall conclusions and a framework for the classification of severe adult malnutrition. Future research priorities are highhghted. Supplementary data are presented in the four appendices.

Section 1.03 The originality and the contribution to medical practice.

This is the first systematic examination of the usefulness of BMI and MUAC in adult patients suffering from severe, acute malnutrition. The findings that a BMI of less than 10 kg m^ is compatible with recovery, redefines the accepted limits of human response to starvation. Many practical and theoretical limitations of BMI as a screening tool for severely malnourished adults

13 during famine are identified. These, and the demonstration of the usefulness of clinical signs and potential for MUAC, are hitherto unrecognised and of practical significance, clearly indicating where the need for more research lies. The comparison of the effect of dietary protein levels in the treatment of severe mahiutrition under “field conditions” is unique and has not to date been repeated in either adults or children. The demonstration of the large decrease in mortality associated with a lower protein diet, given during the initial phase of rehabilitation, has important implications for all future therapeutic feeding programmes.

14 Chapter 2 Introduction

The chapter starts with a discussion of the definitions used in this thesis. It goes on to describe the epidemiology and worldwide importance of severe acute adult malnutrition. It contrasts the large scale of this problem with the low emphasis placed upon on alleviating the condition. The evidence reviewed indicates that the treatment of severely malnourished adult during emergency relief programmes is frequently insufficient. The paucity of the recent scientific literature bears witness to the lack of attention paid by the scientific community towards this issue. The epidemiology of the condition, which usually occurs in situations of war and social disintegration, is identified as an important reason behind this mismatch.

Section 2.01 Definitions

This thesis is concerned witli severe, acute, primary adult malnutrition. In contrast to cliild malnutrition where definitions of acute and severe are accepted, there is no satisfactory definition, or classification, of severe, acute, malnutrition in adults. The following four sections attempt to clarify this definition, as used in this thesis.

2.01(a) “Acute and chronic” malnutrition

In children two patterns of malnutrition, stunting and wasting, are recognised. (Waterlow 1972; Waterlow, Buzina, & et al. 1977) Different processes produce these two patterns and they are assessed using separate anthropometric indices. Acute nutritional deficit produces wasting, characterised by a reduction in weight for height (Wt-for-Ht), or middle upper arm circumference (MUAC). Prolonged nutritional insults result in stunting, characterised by a reduction in height for age. Weight for age is a composite index that reflects both wasting and stunting. (Gomez & et al. 1955) Wasting and stunting are associated with different functional consequences and consequently weight for height and height for age relate to different risks. Weight for height is a powerful discriminator of the risk of short-term mortality, whereas height for age relates to the risk of longer-term mortality. (Bakagi 1981; Briend et al. 1989) The importance of using the correct indicator and cut-off, for the correct purpose; nutrition screening in stable situations, or screening entries into feeding centres during famine for example, is stressed in the academic literature(Bakagi

15 et al. 1985; Chen, Chovvdhury, & Huffman 1980; Habicht 1980; Smedman et al. 1987) and implemented at field level. (Boelaert et al. 1995; Hakewell & Moten 1991; Young 1992)

Anthropometric assessment of adults is more problematic. Despite metabolic differences between chronic and acute malnutrition, (James, Ferro-Luzzi, & Waterlow 1988; Keys 1950a; Shetty 1984; Waterlow 1986) the absence of linear growth removes the power of a height variable to discriminate between the two. Since the Second World War, anthropometric studies of adult malnutrition have been accorded low priority. (Shetty, James, & Ferro-Luzzi 1994) In contrast to the large volume of literature dating back to the 1960s on the use of anthropometry in children, the use of anthropometry for the assessment of adult malnutrition dates from only the last decade. In the 1980s, the meetings of the International Dietary Energy Consultative Group (IDECG) took several important steps to provide a basis for contemporary studies into chronic adult malnutrition. The most important was the creation of a more precise definition of chronic adult malnutrition, called chronic energy deficiency CED. This entailed the clear differentiation of CED from acute energy deficiency (AED). CED was defined as: “A. steady state at which a person is in an energy balance although at a cost either in terms of inct'eased risk to health or as an impaimient oj functions and healtir. (James, Ferro-Luzzi, & Waterlow 1988) This clearly differentiated it from AED that was defined as: state of negative energy balance, i. e. a progressive loss of body energf\ (James, Ferro-Luzzi, & Waterlow 1988)

The differentiation of acute and chronic adult malnutrition is important, as the two conditions entail different adaptations and different risks (see 4.07(a) iv). The use of the term “energy deficiency” by the IDECG and in most of the subsequent literature on chronic adult malnutrition is however rather unhelpful in acute malnutrition. Although the oxidisation of body tissue to provide energy is an important aspect of starvation, to call the condition “energy deficiency” overly focuses attention upon this single aspect. Particularly in severe acute malnutrition, but also in chronic malnutrition, this can obscure the importance of protein catabolism and deficiencies of micronutrients and minerals. (Golden 1995) In acute severe adult malnutrition, deficits of these nutrients are likely to play very important roles in the pathogenesis of the condition (see Section 7.05 ) and for this reason the term acute malnutrition, rather than acute energy deficiency, is preferred. In the interest of compatibility with other authors however, chronic adult malnutrition is referred to as chronic energy deficiency (CED).

16 2.01 (b) ‘‘Primary” malnutrition

In this thesis, the term malnutrition is used to denote a cHnical condition comprising several overlapping syndromes. In adults, wasting, a decrease in body mass, is common to all these syndromes, whereas in children, a failure of growth is the unifying feature. In any individual, tlie most important cause is an inadequate supply of energy or other nutrients, relative to metabolic demands. The reasons for this mis-match of supply and demand can be classified as either primary or secondary. Primary malnutrition develops when nutrient intake is insufficient to cater for normal physiological needs. In children this maybe due to either, the absence of sufficient food, or to the inappropriate preparation of food, for example a diet with an insufficient nutrient density. In adults, primary malnutrition is invariably due to an absence of sufficient food. Primary adult malnutrition is the subject of this thesis.

Secondary malnutrition occurs when an underlying disease process increases metabolic demands or decreases food intake or both. Wide ranging metabolic, hormonal and cytoldne-related abnormalities are involved. These are different to that seen in primary malnutrition and are beyond the scope of this thesis.

2.01 (c) Severe adult malnutrition

In children, classifying wasting and stunting using many different anthropometric measurements has been well studied and reviews are available. (Waterlow 1972; Waterlow 1993) It is accepted that the validity and usefulness of a nutritional indicator depends on the degree to which it reflects health consequences and in children there has been much work assessing the Linkage between anthropometric indices, morbidity and mortality. The most significant of these studies have been prospective in design, involving the follow up of cohorts of children who had previously undergone anthropometric assessment. These have yielded data enabling the relationship of the different indices (mainly Wt-for-Ht, MUAC, Wt-for-Age, Ht-for-Age and growth velocity) and mortality and morbidity patterns to be assessed. (Bairagi 1981; Bairagi et al. 1985; Briend et al. 1989; Briend & Zimick 1986; Chen, Chowdhury, & Huffman 1981; Smedman et al. 1987; Vella et al. 1994) After several years of such research, the strengths and weaknesses of the different childhood nutritional indicators are reasonably understood and rational choices, matching indicator with intended use, can be made. This provides famine-relief workers with well-defined anthropometric and cUnical guidelines facilitating the process of selecting cliild patients for feeding centres. Altliough these have to be interpreted with common sense in order to match the number of

17 admissions to the available resources, the definitions of severe and moderate child malnutrition are valuable aids to the rational operation of child feeding centres.

In adults, the history of anthropometry for the classification of malnutrition is very recent. Although measurements of body weight in malnourished adults have been made for over a century, weight and height measurements have very rarely been recorded systematically. This has precluded any anthropomorphic classification of adult malnutrition. It was only at the IDECG meeting in the late 1980s, that a classification of adults CED in terms of physiological parameters, rather than socio-economic, ones was accepted (see 2.01(a) ). (James, Ferro-Luzzi, & Waterlow 1988) Subsequently anthropometric assessment of adult CED has been common and a classification based upon BMI has been proposed and to an extent validated (see 4.03(a) ). By contrast, the study of acute adult malnutrition has been neglected. To the authors knowledge, excepting one work from the IDP camps in Southern France during 1944, (Zimmer, Wed, & Dubois 1944) there has been no other systematic collection of anthropometric data from severely malnourished adults before this work. In the absence of data from severely malnourished adults, there has been a tendency to extrapolate findings from chronically malnourished adults or those with secondary malnutrition and apply them to severe adult malnutrition. (Boelaert et al. 1995; Ferro-Luzzi & James 1996; Young 1992) I his has occurred despite the general acceptance that adaptations to acute and chronic malnutrition are different. (James, Ferro-Luzzi, & Waterlow 1988) There are many reasons why such extrapolations are theoretically unsound and these are examined in 4.03(e) and 6.05(e) . Such extrapolations can be dysfunctional, serving to mislead inexperienced famine-relief workers. For example, in Baidoa during the famine in 1992 the 16 kg m ^ BMI cut-off that currently appears in two of the largest relief agency field manuals (Boelaert et al. 1995; Young 1992) would have selected almost all of the adult population, estimated at 50, 000. (CDC 1992) At that time, numbers of adult therapeutic places available in the town was 150. (Collins 1993a) Any centres attempting to follow such guidelines would have immediately filled witli admissions not requiring specialised treatment to survive. Furthermore, as the majority of severely malnourished adults are weaker they are often pushed to the back of queues. The result of following such guidelines would therefore have been a centre filled with the less severely malnourished, not requiring specialised treatment to sundve, whilst those in real need remained outside forced to fend for themselves.

In practice during emergency relief programmes, such unhelpful definitions of what constitutes severe adult malnutrition are seldom followed. The concept of severity therefore remains vague

18 and inconsistent, founded upon subjective appreciation of increased mortality risks by the admitting clinicians (see 2.04(c) ).

2.01 (d) A working definition of severe aduit mainutrition

For the purpose of this thesis, there are several important points that characterise severe adult malnutrition (SAM);

• The malnutrition is “primary” malnutrition occurring in any post-pubertal individual because

of a lack of food. It is not malnutrition secondar)^ to some other diseases, although these might complicate and exacerbate the condition.

• It is an “acute” condition where dietary intake is insufficient for metabolic demands, resulting in a continuing loss of body tissue.

• It is “severe” in that if untreated it leads quickly to death.

Beyond these rather general criteria, it is not possible at this stage of the thesis to better define severe adult malnutrition and one of the central aims of tliis work is to improve the definition and classification of the condition.

Section 2.02 The epidemiology of severe aduit mainutrition

Primary SAM in young and middle aged adults rarely exists outside of famine. (Jellife 1966) This is a very different distribution to severe childhood malnutrition and adult CED, both of which are more geographically and temporally widespread, occurring in many developing countries even during “normal” times. Child malnutrition and adult CED therefore occur in situations of greater stability, facilitating research and long term studies with adequate experimental control and follow- up. This has fostered a large and ever increasing volume of research into these subjects. By contrast, the socio/economic disintegration that provides the backdrop for severe adult malnutrition greatly impedes research (see 3.03(a) ). Controlled trials are difficult or impossible to organise and often the foUow up of even a small number of inpatients is prevented by the circumstances (see 6.04(a) i). This is an important factor explaining the limited amount of research into SAM. The small volume of research does not reflect the importance of SAM as a global health problem. Famines are still common. In 1992, at the start of this study, it was estimated

19 that up to 40 million people were at risk of starvation in African alone. (Mecgan 1992) During the two years ot fieldwork for this MD, there were severe famines in Somalia, South Sudan, and Angola. Suhsecjuently the author has worked in famines affecting millions of adults and children ill Jâheria, the Democratic People’s Republic of Korea, the Democratic Republic of the Congo, South Sudan, Burundi, Sierra J.eone and Haiti.

In Baidoa (Somalia) between August 1992 and I'cbruar)' 1993, the Somali Red Crescent Societ}' reported that adults accounted for 8 , 900 of the total 15, 105 deaths recorded by the death-truck workers responsible for collecting bodies (Somali Red Cross & Crescent 1993) (sec Figure 1).

3500 3000 I l i Ochild 2500 li □ adults 2000 II

1500

1000 .

500

Aug Sep Oct Nov Dec Jan month

Figure 1

Monthly body count in Baidoa town August 1992 - January 1993 (town population approximately 60-85, 000). Figures taken from the records of the Somali Red Crescent Society “death-truck" which collected unclaimed bodies (those not buried by relatives) for burial. The definition of child or adult used by the Somali Red Crescent workers is not known

Figure 1 shows that after September the majority of all deaths were reported to be adult. It can also be seen that the ratio of reported adult to child deaths varied over the period shown, and that children tended to die earlier on during the evolution of a famine than adults. Unfortunately,

20 workers on the death-truck were paid according to the numbers of bodies they buried, and in December, it was found that their figures were grossly inflated. Whether this distortion was taking place throughout the period of data collection is unknown. In the opinion of the author, the extremely high number of deaths before November, when rows of dead bodies lined the streets, would have resulted in less incentive for the death-truck workers to inflate figures in order to increase earnings. Given such high mortality, the relative importance of deaths spuriously added to the true total by the collectors, would also have been less during the earlier phase of data collection. Also, any exaggeration of the numbers of dead for financial gain would not be expected to demonstrate a change in the ratio of child to adult deaths as the workers were paid the same price for aU bodies. Thus, although the ICRC figures must be viewed with caution, they probably do reflect two important facts: 1. Children died before adults and 2. High rates of adult mortality persisted despite the fact that reasonable quantities of food and resources were available in the town.

The only other available data comes from two small Centre for Disease Control (CDC) mortality sun^eys carried out in the town during April and November / December 1992. (CDC 1992; Moore et al. 1993) These reported very high rates of mortality during this period, witli some of the highest mortality rates ever recorded in a civilian population. (Centre for Disease Control 1992) They included estimates that 75% of the under-five population had died by the end of November 1992. (CDC 1992; Moore et al. 1993) These data, presented in Table 1, must also be viewed with caution. The CDC study was based upon a small sample size (47 households) and had wide confidence limits around the mean mortality rates. In addition, the sampling methods used probably over-represented displaced people in the town; a group that suffered higher mortality rates than the resident population(De Waal 1993) and the retrospective methods for mortality estimations probably over-estimated the number of deaths. (Collins 1993b) Their further analysis, based upon a sub sample of only 212 adults and 19 under fives (see Table 1), came to the conclusion that there had been no improvement in the deaths rates during November. This conclusion was very different to the impressions of those of us living and working in Baidoa at tliat time, who witnessed a dramatic decrease in the numbers of dead people lying in the streets. It also contradicted the Somali Red Crescent Society data presented in Figure 1, which showed a five-fold reduction in tlie total number of bodies during November.

21 Table 1

Number of deaths and mortality rate, by age group, for the 30 days preceding the surveys (November/December 1992) and for April-October 1992 for internally displaced persons (IDPs) in Baidoa and Afgoi and resident populations in Afgoi -Somalia, 1992. Reproduced from Moore et ai (Moore et al. 1993)

Numbers Deaths April Nov April - Oct Nov - Dec 8-Month total No. death rate No. death rate No. death rate (10.000 ' dav ' ) (10.000'' dav') (10.000'' dav ' All ages 349 212 121 17.2 16 23.4 137 16.9

Aged < 5yrs 63 19 39 30.6 5 69.4 44 30.1

Despite the difficulties in interpreting these data, it appears hkely that during the autumn of 1992 there had been extremely high mortality rates with an early peak in chüd mortality followed by a relatively greater number of adult deaths. This interpretation makes theoretical sense and accords with the impressions of workers in the field. Children can withstand starvation for less time than adults, as their energy stores are smaller relative to their energy requirements. In the severe famine that occurred in Baidoa, tlie extremely high cumulative child death rate over the summer meant that by October there were fewer children left to die. At tliis time, there was still almost no provision of nutritional relief for adults and consequently their death rate was increasing. The combination of these two effects meant that the number of adult deaths in comparison to those of children became relatively greater. Similar peaks in child mortality occurring earlier in the evolution of famines and wars have been described before, in the Warsaw ghetto during 1942(Winick 1979) and during many recent refugee emergencies. (Toole & Waldman 1990)

A similar pattern of mortality was recorded in Melanje, Angola during the famine in late 1993. Data from the Concern Grave-watching programme, set up by the author in November 1993 during the first three weeks of the Concern operation in the town, (Collins 1993c) are presented Figure 2. This programme identified 21 cemeteries in the town and positioned one worker in each to record the number of people buried. These workers filled in a return classifying deaths by sex, reported age at death, and reported cause of death. A system of random checks was instituted to try to improve data quality supervision of these workers and during the first month of the programme, the author visited all but one of the sites. In most cases the reported number of deaths correlated with the numbers of recent graves. However, beyond this basic verification supervision was problematic due to transport difficulties and the close proximity of some of the cemeteries to the front line and mine fields. These data are tlierefore rough and must be interpreted with caution. The reported age is imprecise, based upon the opinions of the relatives

22 and friends burying their dead. A bias towards under-reporting child deaths is also likely, as in Melanje infants were often buried around the family yard. 1 he reported cause of death were unverified, again based upon the opinion of relatives and provide no more than a rough idea as to the proportion of the deaths overtly involving malnutrition.

' 6 0 % o

5 100

40% i

: 20%

0-9 10-19 20-29 30-39 40-49 50-59 60-69 70-79 80-89 90+

age (years) numbers of death cumulative mortality

Figure 2

Age distribution of those buried in Melanje, Angola 21/11/93 - 16/12/93 Data collect by the Concern Worldwide cemetery surveillance programme.

Despite the severe limitations, the data presented in l igure 2 provide some interesting information. As in Baidoa, the percentage of adult deaths appears to be higher than child deaths with only 23% of the deaths occurring in those reportedly aged less than 10 years and 38% in those aged less than 20 years. I’his ratio of child to adult deaths is different to the expected population pyramid, wherein approximately 15-20% of the population would be expected to be under five. (1985) Anecdotal reports, collected by the author during December 1993, indicate that the child death rate had previously been perceived by the population as much higher. This perception is supported by the estimates that in December 1993 the percentage of the population under fi\ e was very much lower than expected. (World Food Programme 1993)

Poor quality data from emergencies, as described above, is a common problem complicating the study of SAM. Despite this several re-occurring themes emerge;

23 • During severe famine, high adult mortality is a serious problem

• Adult mortality tends to occur at a later stage of the evolution of a famine than does child mortality

• In severe famine, by the time relief operations are operating effectively adult mortality often constitutes the greatest public health problem

Section 2.03 The focus of emergency relief programmes

The previous section has argued that malnourished adults constitute a large proportion of the nutritional problem during famine-relief programmes. This fact, although recently recognised in tlie literature, (Davis 1996) is not reflected in the design and orientation of famine-relief programmes.

In Baidoa the international response to the crisis started in June 1992, but because of fighting and

looting, relief was mainly limited to “wet selective feeding ” 2 until the US army inten^ention in mid December 1992. Malnourished children were rehabilitated either in supplementar)^ feeding centres providing two meals of milk, biscuits, and “Unimix”, porridge each day, or in 24 hour TFCs

serving 6 - 8 meals per day. The Supplementary Feeding Centres (SFCs) were run by a nurse and provided oral rehydration, basic medical care, and referral to the hospital, MSF Inpatient Department (IPD) or to TFCs. The TFCs, usually organised by two nurses, with the back up of a doctor on call (the author), were quasi-medical inpatient departments. In addition to the special diets and basic nursing care, these centres provided intense supervision, active oral and naso gastric or occasionally intra-venous rehydration. Basic antibiotics such as ampicillin, cotrimoxazole and metronidazole, antihelminthics, malaria treatment and prophylaxis, vitamins and minerals were also given. In sharp contrast, up until the end of October, the only service for starving adults were the ICRC community kitchens. These kitchens had no medical input and, although supposed to provide 8.4 MJ of energy (one plate of rice and beans) a day, they were plagued by looting and shortages of supplies. Even when food was available the lack of supervision and care meant that the sheer physical struggle needed to maintain a place in the queue often resulted in the thinnest going without food.

2 The distribution of pre-prepared cooked food to targeted populations

24 This division of resources between adults and children did not reflect the relative numbers of cases of adult and child starvation. In October, when the available evidence suggested that adult that approximately 75% of deaths were adult (see Section 2.02 ), there were several thousand child supplementary places and several hundred child therapeutic places available but no specialised rehabilitation facilities for adult. The famine in Somalia, although severe, was not qualitatively different from other famines. In all severe famines adults starve and die. Experienced aid workers report however, that specialised services for severely malnourished adult have not been a feature in previous famine-relief programmes. Prior to the Baidoa centre, the author can find no evidence of specialised adult therapeutic services included in famine-relief programmes since the end of the Second World War. In Melanje (Angola) during 1993, the patterns of mortality were again similar (see Section 2.02 ) with a relative excess of adult mortality at the beginning of the humanitarian intervention. In Melanje however, based upon previous experiences in Baidoa, Concern were far quicker to set up specialised adult feeding facilities and within two weeks of commencing the

programme over 1 0 0 0 adult were admitted into a specialised adult feeding centre.

The large numbers of adults treated successfully in the Concern adult centres (see Chapter 4 and Chapter 7 ), indicate that many lives might be saved if famine-relief programmes pay more attention towards adults. Yet still adult nutritional services are implemented late or ignored

completely. 4 The author considers that an important function of this thesis and the publications tliat it has given rise to, is to highlight of the need and importance of adult therapeutic care. It is hoped that this wül stimulate both more rational prioritisation of famine-relief programmes and research to facilitate this.

Section 2.04 Research into severe adult malnutrition

2.04(a) Primary acute severe adult malnutrition

Descriptions of famine and adult malnutrition exist in some of the earliest writings, (Aykroid 1974) and in “The Biology of Human Starvation”, Keys et al. list over 400 famines, dating back to 1708 BC, for which documentary evidence is available. (Keys 1950a) The history of formalised scientific

3 A mixture of Corn and Soya flours with oil, with or without the addition of minerals and vitamins (see 7.03(c) and Table 19).

4 In October 1998, the author was asked by Concern to set up adult feeding centres in Agip, South Sudan. Here, despite the extremely high mortality rates amongst adults and a large emergency famine relief intervention for the previous three months, there were no specialised adult feeding centres.

25 study into the subject dates back into the 19^'’ centuty, with detailed clinical descriptions of the victims of famine and anorexia nervosa appearing in medical journals and monographs. (Gull 1888; Porter 1899; Stephens 1895) Many studies during the two world wars provided additional data concerning the effects of severe malnutrition on adults. (Burger, Sandstead, & Drummond 1945; Debray et al. 1946; Mollinson 1946; Stapleton 1946; Winick 1979; Zimmer, Weil, & Dubois 1944) Several studies of mild to moderate malnutrition present in professional fasters, (Benedict 1907; Benedict et al.l919) German civilians after the Second World War, (McCance & Widdowson 1951) and volunteer conscientious objectors in the USA (Keys 1950a) have supplemented these data. This early research focused upon clinical aspects of malnutrition. Anthropometric assessment with measurements of weight, and occasionally height, was usually only included as a part of individual clinical case histories. Systematic records of heights and weights from large numbers of people from populations at extreme grades of emaciation are very rare as there were almost no epidemiological studies on adult populations suffering from severe acute malnutrition (see 4.03(e) ). Since the end of the Second World War, scientific attention has shifted towards the malnourished child. In the last 45 years, there have been few clinical and no epidemiological studies of severely malnourished adults.

The most important reason for this inattention is that SAM only really occurs during famine and war, usually co-existing with extreme social disintegration and armed conflict. This makes study difficult, unpredictable and dangerous. The high mortality and extreme suffering characteristic of such situations also makes research more “ethically” challenging, with a powerful compulsion to act to address the immediate circumstances, rather than study to improve efficiency at some time in the future. This, and the relative inexperience of most famine-relief volunteers (see 2.01(c) ), means that few workers devote time to research. Although understandable, this “action mentality” poses difficulties for those who wish to research in such situations as extra resources are rarely made available by programme managers.

The combination of social upheaval, armed conflict, massive need and low resources makes research difficult, and sometimes impossible. This is particularly true of prospective, randomised, blinded study designs which, to the knowledge of the author, have never been performed during famine conditions. Even epidemiological studies are often problematic and of a poor quality. For example the CDC study in Baidoa, sampling just 47 households, was the only epidemiological survey carried out in the whole of die Bay region of Somalia during one of the most severe famines ever recorded (Section 2.02 ). It is unthinkable that similar mortality rates in any stable situation would elicit such a limited scientific input.

26 Further problems for researchers are created by the physical danger of working in such environments. For example during the 1992 / 1993 famine in Somalia, violence claimed the lives of several expatriates and many local aid workers. During the past 7 years the author has been caught in crossfire, had his team kidnapped, experienced armed attacks on his house and had colleagues murdered. This is common when working in wars and famines and there is evidence of a substantially increased risk of death or serious injury amongst aid workers. (Schouten & Borgdorff 1995; Sheik et al.2000) Consequentially there is reluctance on die part of scientists to work in such places and those researchers who do are usually those with fewer family commitments and responsibilities. Such people tend to be younger and by virtue of that fact, less experienced. This inexperience, combined with difficulties in communicating with distant supervisors, promotes lower standards of research. Flowever, established academics in developed countries rarely understand such restrictions. (Caballero 1989) The difficulties of producing research that meet die high epistemological standards required during peer review for reputable scientific journals discourage publication and therefore funding of research programmes.

The result of these constraints is an almost complete absence of contemporary research into SAM. To the knowledge of the author, this work represents the first study for over 45 years attempting to research SAM during an emergency. Given the extent of the adult mortality and morbidity associated with large-scale emergencies (see Section 2.02 ) this represents a serious omission.

2.04(b) Secondary adult malnutrition

Valuable additional data has also been obtained from patients with severe malnutrition secondary to pre-existing systemic disease. (Heymsfield et al. 1984) However such research, although useful, is by itself not sufficient. There are major, clinically relevant, differences between primary and secondary malnutrition and between acute and chronic malnutrition. The processes of screening, diagnosis and management of severely malnourished adults are situation specific. For example, the instruments and techniques required for the assessment of a single patient in a western hospital differ to those required when poorly trained health workers screen thousands of people for entrance into a TFC. The quality of water and sanitation, the availability of famine foods, many of which are poisonous, the incidence of infectious disease and the irregular and variable supplies of relief food, all modify the presentation and management of severe primar}'^ adult malnutrition in ways that do not occur in secondary or chronic malnutrition. These differences necessitate research into primary malnutrition during famine.

27 2.04(c) Indicators for screening: a research priority

Section 2.02 argued that the provision of assistance to malnourished adults is one of the major tasks facing emergency famine-relief programs. At the height of a famine, when large numbers of people are competing for scarce resources, accurate targeting of emergency relief supplies is essential if impact is to be maximised. During the initial phases of famine-relief operations, resources are inevitably scarce and competition for them intense and there maybe thousands of

adults crowding for admission to a single adult therapeutic centre. 5 In dealing with such high demand for limited resources, screening admissions to feedling centres, in order to separate those who require special treatment in order to survive, from tfiie majorit}'^ who do not, is necessary. Violence is common in the pressurised atmosphere characteristic of these situations and when it occurs it is always the most severely malnourished, too weak to hold their own, who loose out. To minimise the risk of fights and riots, screening must therefojre be as efficient and fast as possible.

Screening of children is carried out using anthropometric indicators based upon measurements of

weight, height, and M UAC. (Hakewell & Moren 1991; Ville de Goyet, Seaman, & Geijer 1978) For adults, anthropometric indicators with cut-off values appropriate to severe famine do not exist. The lack of an appropriate screening tool to differentiate quicldy between those requiring specialised assistance and those who do not, poses difficulties for famine-relief workers. In their absence, admission to adult feeding centres is somewhat arbitrary, based upon the subjective impression of the admitting clinician. These staff are often junior medical or nursing personnel who, in response to appeals and press reports of tragedy, volunteer to work with an aid agency for three to six months. Often they have little or no previous experience of similar situations. A good example is that of the author, who had never treated malnomrished patients before being posted to Baidoa to provide medical and nutritional support to a feeding programme covering 450 therapeutic and over 10, 000 supplementary patients. In the absence of clear guidelines, tliis limited staff experience means that rational patient selection is not guaranteed. Although, to the authors knowledge data do not exist, this combination of inappropriate guidelines and limited staff experience must have a negative effect on efficiency. Given tire fierce competition for scarce resources during famine, ^ and the very high mortality^ amongst severely malnourished adults not treated in specialised centres, this sub-optimal efficiency must translate into many potentially avoidable deaths.

5 For example, on the day the Concern Worldwide adult feeding centre cpened in Melanje (Angola) in November 1993, in the region of 2000 adults turned up to try and gain admission. 6 See previous note.

28 The development of indicators to enable the rapid screening of adult admissions to feeding centres is therefore a research priority. This screening is a different role to that of epidemiological evaluation of malnutrition at a population level or to the monitoring of an individual’s response to treatment. The different demands placed upon an indicator in each of the roles have important consequences for the characteristics of a suitable indicator. These characteristics of indicators suitable for such differing roles have been well described by Habicht (Habicht 1980) and are briefly presented below.

i.Surveillance

In the epidemiological roles of assessment, monitoring and sun^eülance at the population level, the prevalence of malnutrition is the important parameter. The appropriateness of a nutritional indicator for such studies lies in its ability to reflect the true prevalence of malnutrition in a population, -j The objective of such epidemiological surveys is to provide knowledge in order to aid decision making at a community level and there are no immediate benefits for the individuals surveyed. (Sackett & Holland 1975) The numbers defined by the indicator are not related to the numbers requiring treatment and the cut-off points used can therefore be determined universally and do not need adjusting according to the resources available. So long as the estimated prevalence of malnutrition reflects the true prevalence in the population, a lack of sensitivity and specificity in the identification of individuals is unimportant. Errors in the determination of an individual’s nutritional status (i. e. variation not dependant on nutritional status) are not problematic, so long these are normally distributed within the population. If this is the case, the data obtained can be standardised using a correction factor based upon the mean population error. The use of such correction factors allows the prevalence in malnutrition in different populations to be compared. This is not the case for indicators used in screening individuals (see below and 4.03(c) i and 6.05(a) ).

ii.Screening

A very different set of criteria applies to the selection of an indicator for screening. Here the specificity and the sensitivity of the indicator for identifying individuals who require individual assistance, and the ease and speed of use of the indicator, are the important criteria of

7 To reflect changes in true prevalence, the sum of the sensitivity (the ability of the indicator to correctly identify individuals who are positive for the condition) and specificity (the ability of the indicator to correctly identify individuals who are negative for the condition) at the chosen cut-off point, must be greater than 1. The greater this sum the more that changes in the indicator reflect changes in the true prevalence of the disease, **80)

29 appropriateness. The purpose of screening is to select individuals at increased risk and treat them. Thus the numbers defined by the indicator are those that need immediate treatment. (Sackett & Holland 1975) In emergency relief programmes, the appropriate indicator cut-off point is often that which selects exactly the number of individuals that can be treated with the resources at hand. (Habicht 1980) Such cut-off values cannot therefore be determined universally but must be tailored to suit the resources available in each particular situation. (Habicht 1980) In practice, the cut-off levels used in screening are based upon universal standards tailored to meet specific circumstances. (Ville de Goyet, Seaman, & Geijer 1978) Non-nutritional factors that change the indicator reduce both its sensitivity and specificity. This decreases the ability of the indicator to differentiate correctly between tliose individuals that require special intervention and those that do not. In contrast to the case of epidemiological indicators, such errors cannot be corrected for using a population mean value and require the extent of that variation to be individually assessed and allowed for. In practice this is rarely feasible (see 4.03(c) i and 6.05(a) ).

Hi. Clinical monitoring

The appropriateness of a nutritional indicator for the clinical monitoring of individual patients once they have been admitted to a feeding centre is based upon criteria that are different again. The requirements of an indicator for this role is sensitivity to changes in nutritional status, ratlier than either an ability to reflect prevalence or sensitivity and specificity in the identification of individuals at risk. For example, weight is neither a good indicator of prevalence of malnutrition nor useful in screening. It is however, a good and widely used indicator of the progress of an individual once admitted to a centre.

30 Chapter 3 Background to the studies and core methods

Section 3.01 Background

3.01 (a) Concern Worldwide

The field research for this thesis was undertaken in Somalia during 1992&3, South Sudan during 1993 and Angola during 1993. During the fieldwork, the author was employed as a rehef doctor by the Irish charity Concern Worldwide. Concern Worldwide is a non-governmental, non-profit organisation, based in Dublin Ireland. Emerging out of church relief programmes during the Bifaran war in 1967, Concern has expanded and secularised and now engages in a variety of developmental and relief work. The organisational aim is to provide relief and assistance to the poorest, most disadvantaged people in the world. At the time of writing, they work in

approximately 2 0 countries, focusing upon relieving poverty through long term, sustainable programmes that improve local capacity through the facilitation and development of local resources. However, in the past few years the increased donor emphasis upon emergency response and the scale and media exposure of several humanitarian emergencies has forced Concern to become more active in emergency relief operations. In particular they have been involved in emergency responses in Somalia, Sudan, Angola, Rwanda, the Congo, Liberia, Sierra Leone, Afghanistan and the Democratic Peoples Republic of Korea.

The organisation has a strong volunteer ethos and the majority of their field staff, particularly in developmental programmes, are volunteers. These people receive only basic expenses based upon the local cost of living for their services. This was also the case for the majority of emergency staff who volunteered for 3 — 6 months and are paid living allowances plus ^500 to cover mortgage expenses etc at home. However the extreme expansion in the numbers of emergency staff required and the “rights based” approach to humanitarian assistance (see the Sphere Project(IToject team 1999) ), requiring greater professionalism, has fostered change in the organisation. Concern, along with many other rehef NGOs have been forced to develop tlieir core human resource capacity with professional workers whilst trying to maintaining the volunteer ethos. The huge elasticity in demand for staff and the absence of formal career paths have

31 hindered this process and Concern are not alone amongst aid NGOs in finding this organisational development a difficult and challenging process.

3.01(b) Concern adult feeding programmes where data were collected

I. Baidoa Somalia (November 1992 - March 1993)

Study 1 took place in Baidoa, the town at the epicentre of the 1992 famine in Somalia. The famine was caused by the civil war that followed the overthrow of the president Siad Barre in 1990. The Bay region of the country was particularly badly affected, and by the autumn of 1992, tens of tiiousands of people had migrated to Baidoa, the regional capital, in search of food. The death rate among these people was one of the highest ever recorded in a civilian population (CDC 1992; Moore et al. 1993) (see Table 1).

A large emergency food-aid programme started in Baidoa in August and rapidly expanded during the next few months. I ’he structure of tliis food distribution programme was unusual as, until the end of December 1992, the conflict prevented an adequate general-ration distribution. In the absence of a general-ration, relief agencies were forced to adopt a extensive wet-feeding strategy, distributing pre-cooked food direct to beneficiaries in selective feeding centres and kitchens. The primary beneficiaries of these programmes were children. By the peak of the programme, in

December 1993, 20, 000 children were receiving either wet “supplementary” (approximately 8 MJ day ’) or wet 24 hour therapeutic rations (> 800 kj kg ’ day ’) each day.

By contrast, no specialised supplementary or therapeutic services were available for adults until the end of October, when the Concern adult feeding centre opened. In the absence of specialised adult feeding, all adults no matter how ill, were expected to survive given a once daily wet ration from tlie ICRC adult kitchens. These kitchens had no medical input and, although supposed to provide 8.4 MJ of energy (one plate of rice and beans a day), were plagued by looting and shortages of supplies. Even when food was available, the lack of supervision and care meant that the physical struggle of maintaining a place in the queue was often too much for the most severely malnourished who frequently went without food and died.

On October 25*’’, Concerned opened Bardalli adult TFC and on November 4*, the author joined the programme as the doctor in charge of medical and nutritional care.

32 The centre consisted of nine shelters constructed out of local sticks, imported monoflex plastic sheeting roofs and mud floors. These provided room for 140 inpatients and 140 family attendants who helped with the day to day care of their relatives. Separate shelters were designated for new admissions, those with severe oedematous malnutrition, those with dysentery and those suspected of having pulmonary tuberculosis. The staff consisted of the author, up to two expatriate nurses, depending upon evacuations and sickness, three local nursing staff and approximately 30 other helpers and cooks. These staff distributed the food and oral rehydration solution (ORS), encouraged patients to eat and drink, and helped feed those who were unable to feed themselves. High levels of supervision with one staff member to eight patients were maintained during the daytime. At night, frequent gun battles and attempts at hostage taldng meant that supervision by expatriate workers was impossible.

The data included in studies 1, 3 and 4 were collected in the Baidoa centre from the beginning of November 1992 until the author’s departure in March 1993.

#7. Ayod, South Sudan (March -April 1993)

Study 2 is based upon data collected in Ayod, South Sudan, during March and April 1993. Ayod is a small town in the West Nile province of Sudan, approximately 150 miles north of Juba. Although previously a colonial administrative centre, today the town consists of grass huts with one permanent building. Since the beginning of the war between the Khartoum government in the north of the country and the Sudan people’s liberation army (SPLA) in the south, the town and surrounding area have experienced frequent, sporadic armed conflict. During early 1993, after a bloody split between the Torit (predominantly Dinka) and Nasir (predominately Nuer) factions of the SPLA, intense fighting again flared up in the area. This intra-SPLA fighting caused destruction of cattle and crops and massive displacement into the town. Although the extreme insecurity in the area precluded population estimates, it is likely that this displacement swelled the towns’ population from approximately 1000 up to 15-20, 000. (World Food Programme - Operation Lifeline Sudan 1993)

In March 1993, after the World Food Programme (WFP) opened the airstrip in the town and started general-ration distributions. Concern staff visited the town for an assessment mission. They were confronted with a very serious situation; widespread severe malnutrition in adults and children, obvious cases of bloody diarrhoea and many people obviously close to death. As the French NGO Medicine Du Monde (MDM) were in the process of starting feeding programmes

33 for the under-fives, it was decided that Concern target specialised feeding for those aged five years and over.

On the 20'*’ March, two Concern nurses arrived and started to construct a compound and two days later the author arrived to help implement inpatient care. By the end of the month, the programme was feeding 1 , 200 adults and adolescents, of which 70 were inpatients. Transport of staff and materials to and from the centre was a major constraint. Ayod was three hours flying time from the Operation Lifeline Sudan (OLS) base in northern Kenya and the airstrip was short and uneven, only accommodating planes carrying less than 7 MT. The scale of the OLS operations at the time also meant that there was great competition for planes between the various sites and in general only one aid flight was possible each day.

The logistic difficulties and the early phase of the relief operation in Ayod, meant that the conditions in the centre were far more rudimentary than in Baidoa. The centre was constructed out of grass and sticks with the use of some monoflex plastic. The food was cooked outside on open fires and the medicine given out from a cardboard box during the “ward rounds”. The medication available came from a basic IDA drug kit that contained cotrimoxazole, metronidazole and ampicillin as the only antibiotics. Despite these huge constraints, the response to treatment was usually good and within a week, the number of deaths in the centre had been reduced from several each day to almost none. Tragically, during the first week of April, the village of Ayod, including the Concern adult centre where the study was taking place, was attacked. The villagers were forced to flee and all Concern staff, including the author, were evacuated. When Concern workers were able to return to Ayod, the centre had been destroyed and the occupants had either fled or been murdered.

ill Melanje, Angola (November - December 1993)

Some of the data included in study 3 come from the Concern adult and adolescent feeding programme in Melanje, Angola. Melanje is a large provincial capital approximately 300 miles to the east of Luanda. The town has always been under the control of the government Movement Patriotic Liberation Angola (MPLA). However, after the resumption of the civil war in Angola in 1992, it was surrounded by “The National Union for the Total Independence of Angola” (UNITA) forces and was shelled frequently. This UNITA action cut off all road access and meant that the flow of almost all food from the surrounding countryside into Melanje all but ceased. Consequently, the population was forced to rely on food flown in from Luanda supplemented by

34 the little they could grow in the small amount of agricultural land within the town. In addition, the fighting caused substantial numbers of people to leave their homes and seek refuge inside the government-protected area. Consequently, the town’s population, previously estimated at 250, 000 increased to approximately 400, 000 with the influx of around 150, 000 — 200, 000 internally displaced people (IDPs). (Government of Angola 1993; World Food Programme 1993)

The fighting, displacement and land blockade meant that the population’s food stocks were extremely low and it was estimated that 85% of the population were dependent on food-aid. This all arrived by air, via the World Food Programme (WFP) air-bridge, however, shortages of fuel, planes, congestion in Luanda and poor air-traffic control in Melanje meant that the quantities provided were insufficient. In December, it was estimated that less than 40% of the minimum needs of the town were being supplied. Predictable the result was an extremely high prevalence of malnutrition, estimated in November at 34% of all children between 6 and 59 months as moderately or severely malnourished (<-2 z-scores), with 20% of tliem severely malnourished (<-3 z-scores and 14% < -2 z-scores. (World Food Programme Melanje & MS F Belgium 1993) The author arrived in November 1993, during the first week of the ficst Concern operation, with an assignment to assess the situation and initiate appropriate responses. At that time, there were no mortality statistics available and one of the first tasks was to set up a cemetery surveillance programme and attempt to obtain some basic mortality data. The initial results of tliis programme are presented in Table 2, and showed that the workers watching the graves reported that over 76% of all bodies buried in the cemetery were over ten years. At that time, almost all the targeted feeding services were targeted towards children. CARITAS operated 73 centres of wliich only 2 are for adults and the government 17 centres, all for children. Recognising the large disparity between needs and services. Concern set up the Upokha adult and adolescent supplementary/therapeutic feeding centre. The centre opened on December 7‘'’ 1993. On the first day, there were crowds of several thousand adults and adolescents attempting to gain admission. These people were screened and registered during the following week and within 7 days, the centre was providing a ration of approximately 13 MJ day ' to 1, 100 patients. Initially the centre provided only daytime care, however, the poor state of health of many of the beneficiaries meant that within a few days, night time services including food and blankets were instituted.

35 Table 2

Reported age of death amongst 811 subjects buried in Melanje between November 1993 and December 1^^ 1993. Data collected from the 21 cemeteries included in the Concern cemetery surveillance programme.

age number of, percentage cumulative (years) deaths of total deaths percentage

0-9 187 23% 23% 10-19 136 17% 40% 20-29 108 13% 53% 30-39 116 14% 68% 40-49 105 13% 80% 50-59 62 8% 88% 60-69 41 5% 93% 70-79 18 2% 95% 80-89 14 2% 97% 90+ 24 3% 100%

Section 3.02 Methods

3.02(a) Subjects

Unless stated all subjects were patients at Concern Worldwide adult feeding centres. The profile of the populations included in the individual studies are included in the appropriate sections (see 4.04(a) 5.04(a) 6.04(a) and 7.03(a) ). In addition, study three includes two groups of more normally nourished adults (see 6.04(a) ).

3.02(b) Measurements, equipment and precision

In Somalia, the weight of the patients, in rninimal clothing, was assessed using a set of standard ‘Hanson’ spring scales. The scales were calibrated in Baidoa and later upon return to London, using known weights. The results were similar at each site. In the useful range of 20 - 50 kg linear regression analysis indicated that there was a constant, linear increase in error (r = 0.99), with the true weight being expressed by the equation:

“True weight (kg) = (1.013987 * scale weight (kg)) 4- 0.707”

36 During analysis, the correction factor derived from this equation was used to correct each weight measurement used in the study. Locally constructed height boards and standard, non-stretchable tailors’ tape measures were used to measure the height of patients.

In South Sudan and Angola, the weight of the patients in minimal clothing was assessed to the nearest 100 g, using Soehnle digital electronic scales, previously calibrated against known weights in the UK. These scales proved to be accurate and required no correction factors to be applied to the weights obtained. The heights were assessed using Microtoise stadiometers, recorded to the nearest 5 mm. AU height measurements were made with the patients standing with their heels together, and legs and back as straight as possible.

MUAC was assessed on the subject’s left arm hanging down loosely by their side, mid way between the olecranon and the tip of the acromium process of the scapular. A standard nonstretchable tailor’s tape measure, read to the nearest 5 mm, was used.

3.02(c) Clinical data collection and coding.

I. Col lection

In Baidoa, upon admission, patients were registered, weighed, and had their height measured by trained local enumerators, supervised by a speciaUy trained nurse or the author. An admission BMI was calculated. A rapid clinical screen, assessing degree of pitting oedema, ascites, anaemia, level of hydration, dysentery, diarrhoea, signs of chest infection and ability to stand was performed by myself or an expatriate nurse. Individual treatment cards were then issued to each patient and this basic data recorded in them and in the centre’s register. The clinical condition and weight of each patient was monitored during rehabilitation and the outcomes recorded.

In Sudan and Angola, data was not coUected on patient’s cards but was entered directly into the centres’ register books.

if. Data coding

‘Admission BMP was a BMI calculated from weight measurements taken within two days of admission. Pitting oedema was coded using the classification of Beattie et al, (Beattie, Herbert, & Bell 1948) (0 = absent, 1 = minimal oedema on the foot or ankle that was demonstrable but not visible, 2 = visible on foot or ankle, 3 = demonstrable up to knee, 4 = demonstrable up to inguinal

37 ligament & 5 = anasarca). Pitting oedema, recorded as present but unquantified, was assumed to be moderate (grade 3). Periorbital oedema was separately graded as 0 (absent), 1 (mild), 2

(moderate) and 3 (severe), with ‘present but unquantified’ oedema assumed to be grade 2 . Ascites was graded as 0 (absent), 1 (mild), 2 (moderate) and 3 (severe), with ‘present but unquantified’ oedema assumed to be grade 2. . Severe oedematous malnutrition was defined as either pitting oedema greater than grade 2 , periorbital oedema greater than grade 1, or any ascites. Apparent dehydration was defined as sunken eyes, decreased urine output or dry mucous membranes. Missing data values were coded as not present. Many patients left the centre spontaneously, especially immediately before and during the planting season. If this defaulting was recorded on the cards and the patient had been otherwise well, these defaulters were coded as survivors. If no note was made as to the fate of the patient on the treatment card they were coded as “lost to follow up” and excluded from any survival analyses. Patients who were discharged to the MSF referral in patient department were followed and their discharge category in that centre recorded on the cards.

iii.Data entry

AU data were entered in duphcate into a PC and validated for entry error using EPI-INFO version

6 . (Dean et al. 1990)

3.02(d) Data analysis

All data entry and analysis were completed using EPI-INFO. v6.02 (Dean et al. 1990) , Logistic (Dallai 1988) and curvefit professional VI. (Hyams 1993) Graphics were presented using Microsoft excel V5. (Microsoft Corporation 1994)

(.The use of means, medians and geometric means

In describing and comparing data such as the BMI and MUAC, a combination of means, medians and geometric means are used. If the data distributions approximated to a ‘normal distribution’, population means are quoted and if the distributions were skewed, population median values were used. As the sample sizes are relatively large and the statistics used relatively robust against departures from normahty, a visual assessment of the distribution was used to assess approximation to normahty. (Kirkwood 1992) In the case of ‘lognormal’ or approximately

38 lognormal distributions, a logaritlimic transformation was used to normalise the data according to the metiiod described in Kirkwood 1992, (Kirkwood 1992) and the ‘geometric mean’ quoted.

a. Use of Odds ratios

Odds ratios were used for the presentation of the chances of mortality associated with BMI, famine oedema and clinical signs. Odds ratios were considered preferable to ‘relative risk’ as, within the cohort of patients admitted and followed up in the Baidoa centre, there were many patients that were lost to follow up. The reasons for these ‘lost to follow up’ was not known. Some patients would have been removed from the centre in order to allow them to die at home, other would have defaulted when they felt well enough to return home, especially when the planting season came. The selection of patients for whom there is data cannot therefore, be considered as random and the baseline for the study could not be the determination of exposure. Instead, the basehne for the study is the outcome, i. e. dead or alive. Those who died are equivalent to cases and those who lived equivalent to controls. The analysis of the risks of mortaUty is therefore treated as a case control study, nested inside the wider cohort study rather than a cohort study in it’s own right.

Hi. Use of logistic regression

Logistic regression analysis was used to assess the association of clinical features at presentation and dietary regimes with mortality. The use of logistic regression enabled other variables, potential confounders of the relationships of interest, to be included and controlled for in the analysis. This was performed by the “Logistic” computer package, using the multiple logistic regression model:

log (P(Y=1 I xl, . . . , xp)/P(Y=01 xl, . . . , xp)) = bO + bl*xl + . . . + bp*xp(Dallal 1988)

Where Y is a binary response variable such as died/survived and XI, . . . , Xp are explanatory variables. The explanatory variables included diet, BMI or oedema together with other potential confounders such as age, sex, date of admission, level of hydration etc. Variables that were not independently associated with mortality were eliminated from the models in both backwards and forward stepwise fashions using estimation techniques.

39 iv.Receiver operating characteristics (ROC) curves

i his thesis uses ‘receiver operating characteristics’ curves (ROC) to examine the ability of an indicator to discriminate between patients with differing prognoses. I'hese cur\es plot true positive rate (sensitivity) against the false positive rate (lOO-specificity) and are a useful means of assessing the trade off between sensitivit)' and specificit}' in a screening test. The values plotted to make the cuiwe are obtained over a range of cut-off points. An ideal indicator has a sensiti\ itv of 100% and a false positive of 0%. Such a point is at the top left-hand corner of a ROC curve. A random indicator has a straight-line plot passing through the origin with a slope of one. These basic parameters are presented in Figure 3. For assessing the performance of BMI in the prediction of mortalit}’, the cun es were constructed from the sensitivities and specificities obtained from 2*2 tables for BMI cut-off values of U) kg m “ to 13.5 kg m “.

00 60

> 40 -- ■good indicator

'random Indicator

oi'' 0 30 40 50 60 70

1 - specificity % (false positives)

Figure 3

Receiver operating characteristic curve for near ideal and random indicators. The point "X" at 100% sensitivity and 0% false positives represents the optimal performance of an ideal indicator

40 Section 3.03 Methodological issues

3.03(a) Constraints

AU of the field work for this MD took place within war zones. This placed severe constraints on the quantity and quaUty of the data coUected and the degree of control that was possible. In addition, each study site has specific constraints.

In Baidoa, the site of studies one, three and four, the frequent gun fighting, bandit attacks on expatriate staff houses, attempts at hostage taking, and mUitary occupations that occurred during the study period, disrupted data coUection. The frequency with wliich patients were weighed and the quaUty of the cUnical data that was coUected varied depending upon circumstances. For example, for much of the month of December 1992, the threat of an escalation of the war in Baidoa accompanying the American intervention forced tlie evacuation of aU but the author and tliree other of concern’s expatriate staff. The author was left as sole supervisor for tliree therapeutic and four supplementary feeding centres with a workload that did not aUow sufficient time to adequately supervise data coUection and patient care in adult TFC. The high incidence of expatriate illness and resulting evacuations on medical grounds, contributed to the problems. AU the expatriate staff working in the Baidoa adult centre, including the author, were medicaUy evacuated at one time or another during the study period. This often left the centres short on supervision and forced the author to devote extra time to activities other than data coUection.

The study in Ayod S. Sudan, originaUy aimed to investigate the use of MUAC and BMI as prognostic indicators for adults admitted to the therapeutic centre. As a result of the sacking of the viUage and murder of the inpatients (see 3.01(b) Ü), the study was terminated before foUow up data on the functional consequences associated with different indicator cut-offs could be obtained.

The Angolan data in study three, came from Melanje, a city surrounded by a hostile army and only accessible by air. The shelling of the city, particularly at night, and the threat from snipers and land mines, prevented adequate night-time supervision. The huge size of the centre,

approximately 1 , 1 0 0 adult patients, the predominance of outpatients and the insecurity situation, meant that defaulters could not be followed up. This, again, prevented any analysis of the functional consequences associated with indicator cut-offs.

41 3.03(b) Subject selection

All subjects;, were either inpatients or outpatients of adult feeding centres, in Somalia, South Sudan or Angola. As these were selected populations, with all subjects having presented and been admitted to a TFC and as such, they cannot be taken as representative of the general population. The effects of selection bias, resulting from differences in the frequency of presentation between certain sub-groups is difficult to evaluate. In Somalia, there is some evidence that men were preferentially transported to the centres (see 4.07(b) ). In Sudan and Angola, the effects of this bias remain unknown.

Selection bias therefore complicates the interpretation of many of tlie data presented here. This weakens the ability of the studies to assess the appropriateness of indicators in the screening of adult patients to TFCs during famines. To assess such screening indicators, ideally, a study relating these nutritional indices to the risk of mortality in the general population is required. Such a population-based study could establish functional cut-off points that apply to the general population. Such studies have been used in stable, development settings, such as the ICCDR studies in Bangladesh, to assess child nutritional screening indicators. However, famines always involve massive social upheaval and large numbers of afflicted people and these, combined with the armed conflict that has characterised almost all famines during the last ten years, make such broad-based population studies unfeasible. Consequently, no such studies have yet been performed for any nutritional indicator in either adults or children during the height of a famine. Given these difficulties, studying selected feeding centre populations is a reasonable alternative. It is certainly better than the almost complete absence of information on the use and suitability of adult nutritional indicators in famine, that existed when the current work was started.

The admission criteria for each centre are reported in the specific methods section of each study (4.04(a) , Section 5.03 i, Section 7.03 ). These admission criteria were necessarily variable. The demands placed upon the centres, the often-limited resources and staff, war and violence, sometimes forced the admission criteria to be tightened in order to avoid becoming swamped with admissions. The criteria quoted are the baseline criteria before any such restrictions were made. The existence and quality of alternative medical care also played an important role in patient selection, sometimes forcing the centres to admit some patients who were not primarily nutritional cases. For example, during October and November 1992, there were no alternative medical centres in Baidoa and 16 patients with BMI greater than 13.5 kg m ^ and medical rather than

8 The exception is the data from Concern workers and adult women in Somalia, included in study 3.

42 nutritional problems were admitted. Their data were excluded from analysis. This practice stopped in mid-December 1992 when a medical inpatient unit was opened in Baidoa enabhng us to refer patients with primarily medical problems. In generail, only patients with a measured BMI below 13.5 kg m^, or those with famine oedema were admitted. Less severely malnourished people were referred to supplementary kitchens. Data from patients with BMIs > 13.5 kg m'^ and no oedema were excluded from aU analysis unless otherwise stated. In Melanje and Ayod, the alternative medical care was extremely rudimentary and some medical patients were included in the centres.

The potential inclusion of some primarily medical patients in the analyses is of most relevance in studies 1 and 4 where outcome is relevant to the study aims. The data for these two smdies come from Baidoa. In this site, for four of the five months of the study, all primarily medical were referred away from the centre to the MSF Holland inpatient department in the town. During the first month when such referrals were not possible, the relative completeness of the data set, including inpatient registers and individual inpatient cards usually completed by tlie author, allowed most primarily medical cases to be excluded. For studies 2 and 3, the effect of the inclusion on medical patients is unknown. Tliis is of less importance, as outcome is not under consideration in these studies. However, it is Hkely that concurrent disease does alter the relationship between MUAC and BMI. This is discussed in 6.05(c) .

3.03(c) Controls

Study 4 aimed to assess the effectiveness of two diets in the treatment of severe adult malnutrition by comparing two groups of patients receiving different diets. There were however, many problems with the study design and implementation. In particular confounding factors surrounding allocation of patients to the diets during the dietary treatment, the use of ORS and actual dietary intake make interpretation tentative. These are discussed in Section 7.05 .

3.03(d) Measurement errors and training

Conditions in the study sites were disorganised and confused, with extremely high rates of mortality and morbidity (see 3.01(b) ). Often at the start o f programmes, there were few trained staff to cope with the very high levels of unmet need amongst the population. Workloads were therefore high and time available for supervision, low. These conditions promoted errors of

43 precision, accuracy and reproducibility in the assessment of height, weight, MUAC and clinical data. They also precluded any assessment of food intake (see 7.03(b) ).

Several attempts were made to reduce and to quantify measurement errors. In all sites the registrars responsible for weighing, measuring and assessing MUAC were hterate and were trained before and during the studies. However, the need to set up centres quickly, the high workloads, and the few experienced workers available limited their training and supervision.

In Baidoa, almost all the clinical data from the adult TFC was collected and recorded by the author, a medical doctor. In the Angolan TFC, although the author screened most of the admissions for oedema and gross nutritional status, the collection and recording of oedema and ascites was performed by local nurses. These people were trained and supervised by an experienced expatriate nurse who had also worked in the Somali centre. During the study period, Melanje was a besieged town under artillery attack, the town’s food supply had been inadequate for many months and the population had been doubled by the many malnourished people displaced from the countryside. These difficult circumstance, coupled with the need to set up nutritional centres quickly in the face of widespread severe malnutrition and death, meant it was impossible to give the local staff comprehensive training before the centre opened.

Given these constraints, problems with the quality of the data collected in all the study sites cannot be ruled out. This must make the interpretation of results more tentative. However, random errors in data collection would have been accounted for by the statistical methods used. Systematic errors were potentially far more serious and several steps were taken to avoid introducing systematic errors into the data. Systematic errors in the precision of tlie scales used in Baidoa were identified and a correction factor built into tlie analyses (see 3.02(b) ). Scale cahbration in Angola and South Sudan revealed no such systematic errors in precision.

In Somalia, an assessment was made of the accuracy and reproducibility, with which the height, weight and MUAC measurements were performed. In the TFC a composite error of reproducibility, consisting of both “intra” and “inter” observer errors was calculated by comparing repeat BMI assessments made on the same patients separated by a three-hour period. The BMI assessments were done in a blinded fashion with the registrars involved not Icnowing the results of the previous estimations. The index is composite because the registrar that made the first measurement was not recorded.

44 1 he errors recorded were normally distributed in the 107 subjects assessed, with a mean error of 0.115 kg m “ and similar magnitudes of error for each registrar ((ANOVA) b = 0.79 d. f. = 1, 106, p = 0.38). No systematic errors were identified. I’hese data are presented in Figure 4.

. 50 X)m E 40 z 30

- 0.8 -0.4 0 0.4 0.8 > 1.0 Error in BMI assessment (kg m

Figure 4

The distribution of errors of reproducibility in BMI assessm ent in 107 patients measured and weighed by the two registrars in the Baidoa adult TFC. The mean error was 0.115 kg m'^, (registrar 1 = 0.15 kg m'^ and registrar 2 = 0.06 kg m'^ (ANOVA F = 0.79 d. f. = 1, 106, p = 0.38).

MUACi in children is known to associated with inter and intra obseiwer errors in reproducibility. (McLaren 1986) In Somalia the reproducibility of the MUAC assessments performed upon the study group of Concern staff were assessed by repeating 67 of the measurements. 'I’hese showed that the mean error between the two MUAC measurements was 0.022 cm (S. D. 0.277), ecjLiivalent to a percentage error of 0.074%. 4 he mean BMI error for the same subjects was 0.014 kg m^ (n = 6 6 , S. D. 0.283), equivalent to a percentage error of 0.05%. Both the MUAC errors and the BMI errors were normally distributed. In Angola and Sudan, no such assessments of error were made.

45 i.Conclusions

Many methodological problems complicate the interpretation of the data presented in this thesis and weaken many of the conclusions drawn. The data do however provide “conjectures” as to the limits of human adaptation to starvation, the usefulness of BMI and MUAC for assessing adult malnutrition and the desirable level of protein in rehabilitation diets. These conjecmres provide a hitherto non-existent base of empirical data and being clearly stated, are amenable to the possibility of refutation. Such development of conjecture based upon empirical methods is at the centre of good scientific method. (Popper 1960) Despite the many methodological problems, they represent an advance in the subject, laying a foundation for future investigations into the management of severe adult malnutrition during famine and war.

46 Chapter 4 Study one: Is BMI a useful tool for the assessment of severe adult malnutrition during famine?

Section 4.01 Alms

To investigate whether BMI is a useful tool for the assessment of severely malnourished adults and adolescents during famine. To examine the risks of mortality associated with different levels of BMI, in patients admitted to a TFC during the height of a severe famine and explore the potential of BMI as a screening tool in these situations.

In particular, the questions of, whether a lower limit of BMI, beyond which recovery is not possible, exists and whether there are specific thresholds of BMI, below which patients require specialised rehabilitation in order to survive, are examined. Practical aspects of the use of BMI are also investigated.

Section 4.02 introduction

This study took place in the concern adult TFC in Baidoa, the town at the epicentre of the 1992 famine in Somalia. For a description of the background to the famine, centre and study see 3.01(b) i. One of the most difficult operational problems was deciding who to admit and who to turn away. Reviews of literature by Concerns Dublin office and attempts to gain advice from other nutritionists, at the time proved, unhelpful. In then field, nobody was able to provide us with realistic admission criteria for such a centre and, despite the presence in Somalia of a very many experienced famine-relief workers, nobody had experience of running such a centre. This subsequent research project has revealed that this was because our centre was probably the first such adult TFC since the liberation of the concentration camps at the end of the Second World War.

Under the extreme conditions present in Baidoa at that time (see Section 2.02 and 3.01(b) i), a useful indicator for screening adult admissions, was one that could rapidly differentiate between tliose who require treatment and those who did not. Central to this was the ability of the indicator to reflect risk of mortality. It was also important that the measurements required were easy to

47 obtain under the very difficult conditions that existed. If they were not, the high demand for the centre and the limited numbers of trained staff, would mean that the indicator would not be used. The concept of the usefulness of BMI, the subject of this study, is a combination of BMI’s ability to reflect risk of mortality in starving adults and the ease with which it can be used during emergency feeding programmes. In particular, the ability of BMI to differentiate between three grades of malnutrition; those who will survive without specialised therapeutic care; those who require specialised nutritional support to survive and those who will not survive whatever were considered important.

Three basic questions were asked:

• Is there a level of BMI below which recovery is not possible? • Is there a level of BMI below which, prognosis is worse and specialised treatment required?

• Are the measurements required feasible under the conditions that prevail in a TFC during the height of a famine?

Section 4.03 Literature review

4.03(a) BMI in the assessment of chronic malnutrition in adults

BMI (wt (kg) m^), sometimes called de Quetelets index, was originally promoted as an indicator for the assessment of obesity in adults. (Garrow & Webester 1985) In obese adults the relationship between grades of obesity, defined by BMI, and morbidity and mortality has been well investigated and the evidence for an increased risk of morbidity and mortality above a BMI 30 kg m^ is now t» generally accepted. (Byers 1996) In 1988, the use of BMI was extendedj^CED and three BMI cut off, 18.5, 17 and 16 kg m^ were proposed. (James, Ferro-Luzzi, & Waterlow 1988) This classification has since provided a useful framework for the analysis of height and weight data from malnourished adult populations. It is also attractive since it conformed to the three grades of childhood malnutrition originally proposed by Gomes et al(Gomez & et al. 1955) and the three grades of adult obesity proposed by Garrow. (Garrow 1993)

The evidence for functional consequences of these lower BMI cut-offs is however, problematic. Some of the large-scale population based studies in developing countries, have indicated that the relationship between risk of mortality and BMI is “U” or “J” shaped. In simple bivariate analysis, mortality appears to increase below a BMI of 18.5 or 19 kg m ^. (Ferro-Luzzi et al. 1992; James 1994;

48 Kushner 1993; Waaler 1984) However, when the effects of confounders, such as cigarette smoking and concurrent sub-clinical disease, are excluded from the models, this excess mortality mostly disappears. (Iribarren et al. 1995; Kushner 1993; Manson et al. 1995) This lack of evidence most probably reflects the low proportion of the population in developed countries with BMI values under 18.5 kg m^ and no concurrent illness, rather than the absence of an increased mortalit}'^ at low levels of BMI. In theory therefore, data from populations in developing countries, where a far higher proportion of the population have BMIs below 18.5 kg m'^, might be expected to provide good evidence of the functional consequences of low BMI. Such evidence is, however, scarce. Few longitudinal smdies of the relationship between low BMI and mortality or morbidity exist and the retrospective, cross sectional design, of most hinders interpretation. An exception comes from a prospective smdy of the relationship between low BMI and mortality in Indian labourers. This ten-year follow up of a cohort of 792 Indian labourers provides evidence of an increase in mortality^ in subjects with low BMIs. In this smdy, men with a BMI below 16 kg m^ exhibited mortalit}'^ rates almost three times than those with a BMI greater than 18,5 m^ (higher (CMR 32.5% compared to a CMR of 12.1%). (Reddy 1991) However, the mortality rates quoted are unadjusted for age and concurrent disease and it is not possible to differentiate between excess mortality directly attributable to CED and tliat attributable to other risk factors. Recently a 5 year follow up of a cohort of adults in Papua New Guinea reports an increased mortality and morbidity of adults with low BMI, even after the effects of concurrent disease are controlled for. Preliminary results indicate that those with CED (BMI< 18.5kg m^) had a relative risk of mortality of 1.87 and a relative risk of subsequently reporting morbidity of 3.00 when compared to the non- CED group. (Duffieid 1998)

4.03(b) BMI and body composition

The use of BMI as an indicator of adult malnutrition has arisen out of an appreciation of its usefulness in obesit)\ Consequently to date, almost all the data on the relationship between the index and body composition are from adequately nourished populations. These data have been reviewed by Morgan. (Morgan 1994a) In obesity, it has been shown that when a person puts on weight, the added tissue has a roughly constant ratio of fat to lean tissue. This results in a strong relationship between BMI and body fat content, with ccrrebMon coefficients in the region of 0.955. (Webster, Hesp, & Garrow 1984) This providers a theoretical rational for using tlie index as a screening tool for excess adiposity. (Lazarus et al. 1996) However, newer more accurate newer methods of body composition analysis such as Dual-energy X-ray Absorptiometry (DXA) or bio

49 electrical impedance, have shown that BMI cannot explain a significant proportion of tire variation in body fat content and there is a wide range of body fat percent for any given BMI. (lloubenoff, Dallai, & Wilson 1995; Smalley et al. 1990) For example, a BMI of 20 kg m ^ in an adult woman corresponds to a range of 13-32% body fat. (Hortobagyi, Israel, & O'Brien 1994) The relationship between BMI and percentage fat is also sex dependent and at any given BMI, women have a higher percentage of body fat. (Forbes 1987; Gallagher et al, 1996; Norgan 1994a; Ross et al. 1994)

The relationship between body composition and BMI in adult under-nutrition is less clear. Of the few studies available, some from developing countries have questioned the use of BMI as a classification of CED. (Immink, Flores, & Diaz 1992) O f particular interest in the current context are questions over the premise that the composition of tissue laid down during weight gain is constant. This appears not to be true at low levels of body fat. Forbes has argued that the composition of tissue lost of"gained varies according to the initial level of fatness and there is less fat in tissue change, at lower levels of fatness. Models of this effect suggest that below a body fat content of approximately 8 %, equivalent to a BMI of 18 kg m^ in men, BMI falls more quickly tlian the % fat. (Forbes 1987) Although there is some empirical data to support this model, the small selective sample of only 164 Caucasian women, is insufficient to determine how applicable the model is to men or to other ethnic groups.

In severe acute malnutrition the relationship between BMI and body fat content is important, and although disagreements continue, the availability of body fat appears to be a central limiting factor in the survival of the severely malnourished. (Forbes 1983; Kerr, Stevens, & Robinson 1978; Korcok 1981; Leiter & Marliss 1982) These data are discussed in 4.03(e) iii.

4.03(c) Determinants of BMI

In health body shape, age and sex are determinants of BMI, independent of nutritional status. Given the steep gradient of BMI cumulative frequency plots in third world adult populations, (James & Francois 1994) even small effects caused by non-nutritionally based variables, can produce substantial changes in the nutritional status of large numbers of people as defined by BMI. (Morgan 1994b)

50 i.Body shape l^MI is dependent on stature and BMI values decline as height increases. In the 1983 metropolitan Height-Weight tables, ^ the ideal ranges of body weight in women corresponds to a BMI of 21.4 - 27.4 kg m " in the shortest and 18.8 to 24.8 kg m “ in the tallest. (Kushner 1993) Inter-ethnic differences in body shape, particularly with regards to the relationship between BMI and the SH/S ratio or sitting height stature ratio (sitting height (SH) / total height (S)) are also well documented. African and Australian Aboriginal phenotypes tend to have longer legs in comparison to their trunk and therefore a lower SII/S ratio, whereas the legs of Asians and l ar I^astern populations are proportionally shorter. These differences result in worldwide variation in the SII/S ratio from 0.48 in Australian aborigines(Norgan 1994c; Morgan 1994b) up to 0.55 in the japanese. (Morgan 1994c) This has a considerable bearing on BMI which, at the extremes of the range of SH/S ratio, are ec|uivalent to a variation of almost 5 kg m^ in BMI (see Ingurc 5). I his probably explains much some of the differences of 3 kg m “ in mean adult BMI values found in populations around the world. (Iwelcth & fanner 1976)

30 1

20 X -

S' E CD 15

CD 10 ^

5 -

0 0.48 0.5 0.52 0.54 0.56

Sitting Height stature ratio

Figure 5

The effect of varying sitting height: stature ratio (SH/S) on the relationship of Body Mass Index and SH/S ratio in a 70 kg, initially 1.75 man (reproduced from Norgan 1994(Morgan I994d)

51 Correction for this variation in SH/S ratio, using a correction factor based upon die mean SH/S ratio value for the particular population, is possible. In this way, comparisons between ethnic groups can be made and BMI remains useful for population monitoring and surveillance (see 2.04(c) i and 6.05(a) ).

Considerable intra-population variation in SH/S ratio also exists. In an aboriginal population

studied by Norgan, the intra-population range for SH/S ratio was 0.41 - 0.54 (mean 0.48, S. D. ± 0.02). (Norgan 1994b) This is larger than the world wide variation in population mean values of SH/S ratio (Norgan 1994c) and is equivalent to a 4 kg m^ intra-ethnic variation in the range of BMI, dependant on shape and independent of nutritional status. This, non-nutritional variation, weakens the relationship between BMI and “true” nutritional status in any particular individual and can only be corrected for by the concurrent estimation of individual SH/S ratio values. If the individual SH/S values are not corrected this variation lowers both the sensitivity and specificity of BMI as a screening indicator of individual nutritional status. The practicalities of routinely measuring sitting height during an emergency programme are discussed in 4.07(e) .

ii.age

Adult body size, shape and composition vary with age, (Borkan & Norris 1977; Forbes & Halloran 1976; Gallagher et al. 1996; Mazariegos et al. 1994; Noppa et al. 1980; Norgan 1994a; Rolland-Cachera et al. 1991; Stoudt et al. 1965; Strickland & Ulijaszek 1992; Wang et al. 1994) and these change alter the sigmfïcance of BMI and also the ease with which the requisite height measurements can be taken. The adolescent growth spurt produces a rapid increase in length that is not matched by weight increases. This causes , high rates of change in the mean BMI with increasing age in adolescents and young adults. This may explain some of the extremely low BMIs found in anorexic patients (see 4.03(e) ii). Increased prevalence of spinal disease (predominantly osteoarthritis and osteoporosis) that decrease subjects capacity to stand straight to have their heights measured makes the application of BMI difficult or impossible in an increasingly large proportion with increasing years. (Todorovic 1998) These changes in BMI with age have been well-

documented developed country populations (see Figure 6 ). (RoUand-Cachera et al. 1991)

9 the metropolitan Helght-Weight tables are often used as the basis for the determination of "ideal" body weights based upon risk of mortality

52 (a) French Melee

32 97 30 90

7 5 -

I 3 -

0 5 10 IS 20 2S 30 3S 40 45 SO 55 60 65 70 75 60 85 90

Y mt« of Aq «

lb) French Pemelee

32 9 7 " 30 9 0 . 2 # 26 24 22 * 0 _ 20 II 16 14

1 2 10 0 5 10 IS 20 2S 30 38 40 48 80 98 10 §9 70 79 80 IS 00 Yeert of Ag#

Figure 6

BMI centiles by sex and age in the French population (llolUtid-Cachcra ct al. 1991)

In addition to the influence on body shape, age also affects body composition. Adults tend to loose h’I'M and increase I'M with age (Forbes & Rein a 1970) and these changes alter the functional significance of BMI at different ages. In people aged over 85 levels of BMI > 30 kg m “, considered as grade 1 obesity and associated with increased morbidity and mortality in younger adults, are beneficial. Conversely levels below 20 kg m ", a level where mortality is at it lowest in younger adults is associated with a considerable lower five years siuwival rate in people aged 85 and over (see Figure 7). l.ead-time and socio-economic biases contributes to this effect, however even when concurrent disease and socio-economic factors are controlled out, the effect appears to

53 remain. C^liangcs in body composition, with a decrease in body fat with age might be relevant factors (see 4.03(b) and 4.03(e) iii). b'rom fhe above discussion, it is clear that age must be taken into account when interpreting BMI and deciding if BMI is a relevant nutritional indicator, (janies & brancois 1994)

0 6

Mbots of foitow up

Figure 7 key : (all values in kg m'^)

>=2 8 ,'...... 2 6 -27.9 , ...... 24 - 25.9 , ------2 2 -23.9,— ' — ' 20 .0 -21 .9, ------< 20.0

Five year survival curves according to Body Mass Index in subjects aged 85 years and over. Reproduced from Mattila, K. et al, 1986 (Mattila, I laavisto, & Rajala 1986)

Hi. Gender

1 he same BMI cut-off points for CI3D have been proposed for both men and w^omen. (Ferro- lxv///.i et al. 1992; Janies 1994; James, Fcrro-lAizzi, & W aterlow 1988) 1 hese are however, tentative, based on limited data of the functional consequences of low BM I (see 4.03(a) ). Data from developed countries indicate that sex differences in the relationship between BMI and body

54 composition exist (see 4.03(b) ) and have been used to argue for sex specific cut-offs for obesity. (Pfortobagyi, Israel, & O'Brien 1994) In other studies on obese subjects, sex specific functional consequences of BMI cut-offs have been demonstrated with respect to risk of disease, (Delpeuch et al. 1994; Strickland & Ulijaszek 1992) levels of blood markers (Folsom et al. 1994) and significance of genetic differences. (Harris, Tambs, & Magnus 1995) These finding are however not universal. (Negri et al. 1992; Tuomilehto 1991) Given the importance of body fat during severe starvation (see 4.03(e) iii), sex differences in body fat at any given BMI are likely to result in sex specific BMI cut offs. This appears to be the case in the data reviewed by Henry (see 4.03(e) ), but data are at present scant. This study attempts to shed light on this issue.

iv.Ethnicity

Ethnic variations in body shape, particularly the SH/S ratio have been explored above. In addition to this effect, ethnic differences in body composition may mean that the nutritional significance specific BMI levels also varies between different ethnic groups. In particular, there appears to be substantial inter-ethnic variation in the percentage of body fat at any given BMI (for

examples see Figure 8 ). Survival in starvation is intimately related to body fat content (see 4.03(e) iii) and this inter ethnic variation may well result in BMI cut-offs having different functional significance for different ethnic groups. Such differences may result in different BMI cut-offs for different ethnic groups.

10 With the increasing availability of newer, highly accurate techniques for estimating body composition such as neutron activation and dual photon absorptiometry these differences are becoming ever more precisely quantified. To date all these studies using such sophisticated technology have been performed on relatively well nourished ethnic subgroups living in developed countries and now validation has been attempted upon those suffering from CED or AED in the developing world. *"’’"8 ^ang ei ai. 1994)

55 • Italien adults • Italien adults * New Guinea * New Guinea o Ethiopian & Somalis 0 Ethiopian & Somalis

- % British soldiers >34 .• 3 2 -3 4 30- 3* • 2 8 -3 0 • 2 6 -2 6 • « 2 4 -2 # • i Z * 24 *0 • 2 0 -2 2 *0 » 18-20 »c # ^ <4 # 16 18 <5— " A 14-18 0 • 0 <14 . 0 « 1 I 1 < 7 1013 16 10 2 2 2 9 2 8 3 1 3 4 16 10 22 26 26 31 34 37 40 43 fat % fat %

Figure 8

Body Mass Index and body fat in adults from different parts of the world. Data from studies on 226 Italian men aged 38 +/- 7.4 and 76 women aged 37 +/- 5.5 years; 482 Ethiopian and Somali men aged 33.4 +/- 13.1 years; 255 Ethiopian women aged 36.1 +/- 12.1 years; 18 New Guinea men aged 31.3 +/- 8.2 years and 90 women aged 29.4 V- 5 years. The mean +/- 2 S. D. . lines refer to data from 5072 UK soldiers studied by Durnin et al. . (1984) and predominantly aged 17-34 years. Most or these estimates of body fat in the Italian and New Guinea studies were from body density and the others from skinfold thickness measurements. Reproduced from James et al. 1988. ()amcs, 1'crro-Liizzi, & Waterlow 1988)

4.03(d) The use of BMI to screen severely malnourished adults

lb date all research into the use of BMI in the assessment of malnutrition has focused upon the assessment of chronic malnutrition in stable populations, d'his role is primarily that of the epidemiological surv eillance of chronic malnutrition at a population level. This is a very different role to that of screening individuals suffering from acute malnutrition, in order to regulate admissions to feeding centres. 1 he demands placed upon an indicator for surveillance are not the same as those for screening (see 2.04(c) ). 4 he common assumption, in contemporary NGC) field manuals and recent academic articles, (Boelaert ct al. 1995; Ferro-Luzzi & james 1996; Young 1992) that BMI is also an appropriate indicator for screening during famine, has not been tested. Indeed BMI cut-offs for screening adult admissions to feeding centres, extrapolated directly from the definition of severe adult CED, can be highly inappropriate. The cut-off of 16 kg m “, that appears in two of the largest relief agency field manuals, (Boelaert et al. 1995; Young 1992) would, in Baidoa during the end of 1992, have selected almost all of the adult population, estimated at 50, 000. (GDC 1992) At that time, numbers of adult therapeutic places available in Baidoa was onlv 150.

56 Obviously few out of the 50, 000 were admitted to adult TFC, however, many did survive given only a basic general-ration or one cooked meal a day. This indicates that 16 kg m^ BMI cut-off did not reflect a degree of starvation that required specialised treatment.

Such non-specific cut-offs are unhelpful for screening. The 16 kg m " threshold for severe CED was proposed for the surveillance of the nutritional status of populations and relates to the risk of long term functional consequences in the absence of intervention. (Ferro-Luzzi et al. 1992; James, Ferro-Luzzi, & Waterlow 1988; James & Shetty 1994; Naidu & Rao 1994) In famine, such long-term functional consequences are largely irrelevant and it is the ability to ascertain the immediate risks of mortality and degree of nutritional support that individuals require, that are important. There is no reason to suppose that the BMI cut-off that relates to long-term consequences is the same as one that relates to such short-term consequences. Until now, there have been no data from adults in famine available and the usefulness of BMI for screening in such circumstances remains untested. This study aims to do that.

4.03{e) Previous investigations into the iimits of human response to starvation

LEvidence from famine victims

Some of the earliest height and weight data from severely malnourished adults was recorded by doctors working for the colonial medical service in Madras during the famine of 1877. Although there is litde data from individuals, some composite data for those that died does exist. Porter reportedly recorded the weights and heights of 30 non-oedematous and 13 oedematous male subjects who died of starvation. The mean BMIs at the time of death were reported as 13.6 kg m ^ and 18 kg m'^ in the non-oedematous and oedematous subjects respectively(Porter 1899) . Interpretation of population means in the absence of information on the shape of the population distribution must however be tentative. If there is a fixed BMI threshold below which life is unsustainable, the distribution of the BMI values would be expected to be positively skewed. „ In a skewed distribution the geometric mean, or the median, are describe the population better than the mean. In a positively skewed distribution, these are lower than the mean. The mean values reported above are therefore of limited values in establishing the lower limit of BMI compatible with survival. The records do however record the interesting observation that tlie men who died

11 This was the case In the Baidoa data (see 4.06(a) ii).

57 had lost a greater proportion of their body weight than the women, with both sexes dying at a mean body weight of 77 lbs.

Several studies into adult malnutrition took place during the 1 *' world war. All appear to have been clinical studies primarily concerned with famine oedema (see review by McCance et al. 1951 and Keys 1950(Keys 1950a; McCance & Widdowson 1951) ), and height and weight data were not usually recorded. At the time of the Second World War, adult malnutrition was again common and several studies were performed. Reports from the Warsaw ghetto(Winick 1979) , France, (Debray et al. 1946; Zimmer, Weil, & Dubois 1944) the concentration & prisoner of war camps, (Leyton 1946; Lipscomb 1945; Mitchell & Black 1946; Mollinson 1946; Murray 1947; Vaughan & Pitt Rivers 1945) Holland during 1944-5, (Burger, Sandstead, & Drummond 1945) and in post war Germany (McCance & Widdowson 1951) have been studies for this review. In these studies although weight and height data were rarely collected in a systematic fashion some relevant information, particularly percentage of the original body weight that was lost, are available.

Frenchmen transferred from the Nazi concentration camps to the Salpietre hospital in Paris were frequently reported to have experienced a reduction in body weight of 30 kg approximately 40% of their normal body weight. (Debray et al. 1946) Given a mean BMI for a normal European population of 22 kg m ^ this represents levels of BMI in the region of 13 kg mA One 19 year old male is reported to have been repatriated to France from the camps with a BMI of 9.5 kg m^ (Ht 172 cm, wt 28 kg) however no note is made as to whether he survived. (Debray et al. 1946) In the internment camps of the unoccupied zone of France, weight losses of up to 49% were recorded in botli male and female internees. The distribution of weight loses were positively skewed (see Figure 9) and women exhibited greater weight loss than men (median values 25 - 30 % in women and 18-22% in men. The high prevalence of famine oedema in the camps (48% in men and 52% in women) and the failure to differentiate between oedematous and non-oedematous patients in the weight data, means that these figure are likely to underestimate the true extent of the weight losses. (Zimmer, Weil, & Dubois 1944)

58 n m ale ■ fe m a le

1 rH m an rr^m

percentage of physiological weight

Figure 9

Distribution of the percentage of physiological weight in male and female inmates in the internment camps in the occupied zones of France during 1942. N not given in the text but the population of the camps were stated elsewhere to be approximately 10, 000, divided up into individual camps of approximately 2, 500 people.

Weil, & Diihois 1944)

In Rclsen concentration canip, even greater estimates of weight loss, as a percentage of previous body weight were recorded (see 1 able 3). ILxtreme as some of these weight losses appear, the author states them to be an underestimate of the degree of emaciation seen in the concentration camp. Mollinson reports that the most severely emaciated patients had been too ill to be weighed and because of this the data in I ’able 3 were recorded some days after the start of treatment, presumably after some increase in weight had occurred. (Mollinson 1946)

59 Table 3 Loss of body weight in 10 patients liberated from Belsen concentration camp(Mollinson 1946)

case weight origional weight weight loss weight loss ka ka ka (% of ohaional weiaht) m42 35.0 80.0 45.0 56.3

f116 25.0 50.0 25.0 50.0

fx1 34.0 63.0 29.0 46.0

m37 39.0 67.0 28.0 41.8

f114 35.1 56.0 20.9 37.3

fx4 39.9 62.0 22.1 35.6

m35 32.0 50.0 18.0 36.0

f111 31.0 47.0 16.0 34.0

f124 33.2 48.0 14.8 30.8

f101 33.0 47.0 14.0 29.8

Amongst the records written by the doctors in the Warsaw ghetto during 1942 - 1944 are occasional weight and heights data. Eighteen patients aged between 16 and 42 years (mean 27 years) were recorded as having weight losses of 50% of dieir original body weight resulting in a mean weight of 32 kg. (Apfelbaum-Kowalski 1979) Unfortunately, neither the height nor the sex of the subjects are available from these records. However if the mean BMI of these subjects is assumed to have been in the middle of the normal range (18,5 - 25 kg m^ ) a 50% reduction in weight would result in a mean BMI of approximately 11 kg m^. Such an estimate seems reasonable and when combined with the mean weight of 32 kg gives a mean height of 1.7 m, a reasonable value for a mixed Caucasian population. One record where height and weight data exist is of a woman aged 30 years with a reported weight of 24 and height of 1.52, equivalent to a BMI of 10.4 kg m^ (Fliederbaum et al. 1979)

Records from the Dutch winter famine of 1944-5 contain more anthropometric data from both survivors and non-survivors. In these data, collated by Henry, (Henry 1990; Henry 1993) the mean values of BMI in both survivors (17.4 kg m^, n = 16) and non-survivors (13.3 kg m'^, n = 5), are considerably higher than those estimates from the Warsaw ghetto (see above). A possible explanation for these differences could be the low ambient temperature during the Dutch winter combined with the lack of fuel for heating. Such an explanation is consistent with tire many observations that hypothermia and an inability to control body temperature are usual in malnutrition(Keys 1950a; Winick 1979) and the reports of higher death rates from starvation that

60 occur during the winter months. (Zimmer, Weil, & Dubois 1944) 4.07(a) provides further discussion of possible reasons for these differences, pertaining to the current data.

Despite several severe post war famines, in particular in India and the African continent and the frequent studies that have taken place during these famines, height and weight data from severely malnourished adults do not appear to have been recorded.

ii.Evidence from Anorexia nervosa

Extremely low levels of BMI are relatively common in anorexia nervosa. Early studies of adolescent girls include descriptions of a BMI of 8.7 kg m^ (Stephens 1895) in a 16 year old measured and weighed at autopsy 56 hours after death, and 9.9 kg m^ (Gull 1888) in a 14 year old who survived. Tliis century several series of anorexic nervosa patients with extremely low BMI have been reported. Based upon data collected from adult women with anorexia nervosa and extreme emaciation (see Table 4), Berkman argued that a reduction of greater than 50% of the original body weight was tlie critical level below which recovery was less likely.

Table 4

Height and weight data on 4 adult female anorexic patients (LTF = lost to follow up). Adapted from Berkman et al 1947 (Berkmann, Weir, & Kepler 1947)

Case number Age Weight Height BMI Outcome vears kp m kam '^ 1 24 22.7 1.6 9.1 Lost to follow up

2 18 23.2 1.6 8.8 died

7 32 27.0 1.6 10.6 recovered

18 31 27.3 1.6 10.9 Lost to follow up

mean 26.3 25.1 1.6 9.9 -

Similar extremely low levels of BMI have been reported in several other series of patients with anorexia nervosa (see Farquharson (1941) (Farquharson 1941; Farquharson & Hyland 1938) and Ljunggren et al. (1961) (Ljunggren, Ikkos, & Luft 1961) and Bruckner et al. (1938). (Bruckner, Wies, & Lavietes 1938)

Many factors promote low values of BMI in patients with anorexia nervosa. Age is a determinant of BMI (see 4.03(c) ii) and adolescents, in whom the incidence of anorexia nervosa is highest, have lower BMIs than adults. Adolescent anorexics would therefore, be expected to survive at lower

61 levels of BMI than adults (see 4.03(c) ii). The relatively high quality, albeit energy deficient diets, characteristic of the early phases of the disease, (Beumont, Russell, & Touyz 1993) would produce some sparing of lean tissue and micronutrient reserves, (Owen et al. 1998; Sapir et al. 1972). facilitating a more complete utilisation of energy stores before death occurred (see 4.03(e) iii). The higher socio-economic conditions associated with anorexia would tend to decrease the prevalence of infectious disease. There is also some evidence that at comparable degrees of emaciation, patients with anorexia nervosa suffer less severe immunosuppression than people with other forms of protein energy malnutrition. (Bessler et al. 1993; Golla et al. 1981) A relatively reduced prevalence of infectious disease compared to famine victims would also decreasing the free radical

load thereby conserving anti-oxidants , 2 and promote survival at lower BMIs in anorexics, reducing the relative risk of mortality at any given BMI. The reported tendency towards dehydration in anorexic patients might also contribute to depressed weights in anorexic subjects. (Bruckner, Wies, & Lavietes 1938)

In 1990, CJK Henry reviewed some of the above data togetlier with some additional sources obtained from professional fasters. He concluded that a BMI of approximately 12 kg m^ was around the lowest limit compatible with life, refining the estimate with sex specific cut-offs of 13 kg m ^ for men and 11 kg m ^ for women. (Henry 1990; Henry 1993) As Henry himself points out the proposed cut-offs were tentative, based upon sparse data from diverse sources, and designed more to stimulate further research rather than provide cut-offs for use in famine-reHef programmes.

The hmit of adaptation of 13 kg m ^ for men is largely extrapolated from the data collected by Buerger et al. during the Dutch winter famine (see above). The conditions that prevailed during the time of this data collection were extremely harsh, with a cold ambient temperature, little fuel for heating and a breakdown of social services and infrastructure. It is possible that these factors prevented survival at levels of BMI that would, in less harsh circumstances, have been compatible with recovery. The figure of 11 kg m^ for women is far less clear cut and there is substantial variation in BMIs both within the group that survived and those that died. In contrast to the male data, the majority of the female subjects are anorexics nervosa sufferers living in western societies.

12 Anti-oxidant status has been shown to correlate with risk of mortality in severely malnourished children, * Ramdath 1987)

62 iii. The iower BMi iimit of human response to starvation

The above data suggest that a reduction of greater than 50% of normal body weight, equivalent to a BMI of approximately 12 kg m^, is the lowest limit of BMI compatible with Hfe. The available evidence does however contain substantial individual variation in this absolute limit. Support for limit of 50% loss of normal body weight also comes from studies of body composition in normally nourished subjects and those suffering from secondary malnutrition. Heymsfield has demonstrated that at a body weight approximately 50% below normal in men and between 50% and 60% below normal in women, body fat reserves are effectively exhausted. As weight loss proceeds towards this point, an increasing proportion of the bodies’ energy needs must come from the catabolism of fat free mass. Each gram of FFM produces approximately only 4 kj (Between 20 and 25% of FFM is protein (Heymsfield et al. 1984) ). This is much less than the 30 kJ available from each gram of fat (adipose tissue consists of 80% triacylglyceride, 18% water and 2 % protein (Heymsfield et al. 1984) ). If FFM were the sole metabolic fuel, between 0.5 - 1 kg of FFM would have to be cataboHsed each day in order to meet resting energy requirements alone. Survival times at tliis rate of FFM loss in individuals with pre-existing FFM depletion would be extremely short.

Studies of the percentage of energy derived from protein (P-ratio), support these findings. (Payne & Dugdale 1977) Although the P-ratio is usually a fixed constant in any particular individual, there is evidence of a “pre-terminal” sudden increase in protein catabolism during a prolonged fast . (Henry, Rivers, & Payne 1988) This would be expected with a switch to FFM as the sole metabolic fuel.

The sparse data from starving adults is also compatible with exhaustion of fat stores being a limiting factor in the adaptation to star\^ation. An analysis of the tirning of deaths in the Irish hunger strikers indicated that loss of fat was an important factor in the timing of their deaths. (Leiter & Marliss 1982; Leiter & Marliss 1983) Post mortems records of subjects that died of starvation or anorexia also indicate the almost complete absence of fat reserves. (Stephens 1895; Winick 1979; Zimmer, Weil, & Dubois 1944)

This model of protein and fat stores limiting adaptation to starvation is based upon data gathered from populations in developed countries. The background conditions in these countries are however very different to those seen in famine. The hygienic surroundings, high quahty foods and food supplements available and usually acute starv^ation superimposed upon well-nourished pre- morbid states, all reduce the importance of other factors that limit adaptation to starvation. In

63 famine the breakdown of environmental hygiene, contaminated food and toxic famine foods all increase exposure to infection and damage by free radicals. In addition, starvation is prolonged and superimposed upon pre-existing nutritional deficiencies and the diets eaten before and during the evolution of malnutrition frequently deficient in a whole range of nutrients. These factors would increase tlie frequency with which micro-nutrient deficiency, infection, poor anti-oxidant status and free radical damage disrupt normal adaptation before aU available fat stores were used up. Conversely, the superimposition of acute starvation upon a background of adaptation to chronic starvation might predispose individuals to greater adaptation. These issues are further discussed in 4.07(a) .

The result of these differences is that malnourished people in famine are likely to die from different causes than those starving under clinical conditions in western countries. Rather than respiratory or cardiac insufficiency caused insufficient energy or muscle as an endpoint to starvation, infection and/or metabolism dysfunction are frequently the terminal events. Given such different end-points, universal stipulation of BMI limits to starvation appears to be theoretically unrealistic. Tliis assertion is examined in the rest of tliis chapter.

4.03(f) Triage and BMI limits of human adaptation to starvation

During famine-relief programmes demand for food and basic medical attention greatly outstrips the resources available. Selective feeding centres, important interventions to reduce excess mortality, cannot be too large or crowded. If they are they cannot function effectively and case fatality rates and the lengths of time patients stay, increase. Efficient patient selection to limit numbers is therefore essential. At present, there are no tools available to screen severely malnourished adults. Such selection requires that those patients that need specialised feeding in order to survive, are differentiated from those that will survive on a general-ration and those who win not survive whatever. BMI is potentially useful at both ends of this spectrum. If there are BMI thresholds below which prognosis is considerably worse, BMI could be used for the selection of patients for specialised for therapeutic feeding. At the most severe end of the spectrum of starvation, identifying patient that are beyond help no matter what therapy is given, would also be useful. Although distasteful, resource restrictions are such that a triage tool is often valuable in

targeting scarce resources towards those for whom they can do some good. ,3 The more severe

13 In a modern world, the need for such triage, In the treatment of a slowly evolving condition that, if caught early is so easily treatable, is an obscene failure. However, the collective will of the international community to prevent famines before they happen, or even to promptly intervene before the horrors of widespread death from starvation supervene, is sadly lacking. The reality of famine relief programmes is still usually “too little too late”.

64 admissions fequife most staff, resources and time. If those for whom there was no possibility of survival could be screened out prior to admission, the efficiency with which therapeutic care could be targeted would be increased. A requirement of any indicator used for this role is a very high degree of specificity. This is essential to help ensure that only those who are beyond help are excluded form curative treatment. In the context of famine-relief programmes, accurate identification of the level of BMI below which recovery is not possible, has, therefore, more than a theoretical importance. If there are BMI thresholds related to prognosis, at either the upper or lower ends of the spectrum of severe adult malnutrition, they would be extremely useful.

The literature contains little data as to the relationship between BMI and mortality in severe acute adult malnutrition. Extrapolations from adults with CED wiU not do (see 4.03(d) ). At the most severe grades of starvation BMI data are almost non existent and at present it is not possible to specify, with any degree of certainty the BMI limits to survival (see 4.03(e) ). This study attempts to address these issues.

Section 4.04 methods

4.04(a) Subjects.

Data were collected on 573 patients with a reported age of 15 years and over, admitted to the Concern Worldwide Adult Therapeutic Centre, Baidoa, between 25''’ October 1992 and the end of March 1993. The reported ages of those admitted were between 15 and 80 years (mean 34). It is unlikely that these reported ages are accurate as people were generally unsure of their exact ages and no local calendar was available to verify the ages given. Analysis of the distribution of the age data revealed clustering around ages divisible by five, suggesting that reported ages were often given to the nearest five years. 46% of those admitted were male.

The criteria for admission varied depending upon the available space in the centre, and the existence of other medical facilities in Baidoa (see 3.03(b) )

65 4.04(b) Treatment in the centre

LSupervlslon

High levels of supervision, with staff patient ratios of 1:9 or less, particularly during the first few days after admission, were maintained. These staff distributed the food and WHO formula oral rehydration solution (ORS, composition: sodium chloride 3.5g/l, trisodium citrate 2.9g/l, potassium chloride 1.5g/l and glucose 20g/l). They encouraged patients to eat and drink and helped feed those who were unable to feed themselves.

H.Rehydratlon

Patients were orally rehydrated using the WHO formula ORS. This was prescribed upon an “ad hbitum” basis, with all patients encouraged to drink as much as possible.

ill. Dietary treatment

There were 6 - 8 meals each day. The diets used consisted of high energy milk ((HEM), made from dried skimmed milk, vegetable oil and sugar), sweet tea, rice, oil, beans, biscuits and local fruits, especially bananas. During the first few days after admission the HEM component was diluted with WHO formula ORS. Two therapeutic regimes were used in the centre. A high protein regime wherein patients were offered 156g protein and 16.5 MJ (16% of energy from protein, P:E ratio 9.5 g/MJ), and a low protein regime wherein patients were offered 81 g protein and 16.5 MJ (8.5% of energy from protein, P:E ratio 5 g/MJ). For a detailed discussion of the diets see 7.03(c)

iv.Medicai treatment

Oral antibiotics were needed by the majority of the patients upon admission, with many patients remaining on an antibiotic throughout their stay in the centre. Penicillin V or ampicillin, were the first Hne antibiotics for pulmonary infections; cotrimoxazole and metronidazole for dysenter)/^ and persistent diarrhoea. Chloramphenicol was the second line antibiotic.

66 4.04(c) Discharge criteria

Discharge criteria were predominantly clinical: Freedom from infectious disease, a good appetite, constant weight gain, and the ability to walk and care for themselves, were all necessary conditions for discharge. The presence of minimal pedal oedema did not preclude discharge so long as all the above criteria were fulfilled. Upon discharge all patients were issued with a “rehabilitation pack” containing a sheet of plastic, hoes, seeds, a machete, pots and pans, a blanket, a water containers and some cloth. Transport to the patients’ home village was provided.

For additional methods see core methods see Section 3.02

Section 4.05 Results

4.05(a) General inpatient data.

Of the 573 admissions for whom data were collected, 413 (72%) survived, 122 (21%) died, and 38 (7%) were lost to follow-up. Those lost to follow up were excluded from the analysis of mortality. Mean reported age of those that died was 36 years (95% C. I. = 33-39), slightly older than the survivors whose mean age was 33 years (95% C. I. = 31.5 - 34.5, ANOVA F = 3.1, p = 0.08 ). 46.4 % of those admitted were male and the gender composition of those that died did not differ from sunnvors (likelihood ratio test = 1.5 p = 0.2). 19 % of deaths occurred within 48 hours after admission, 48% within the first week and 27% during the second week after admission (see Figure 10). Mean time to discharge was 30.8 days, (34.6 days in oedematous patients and 30.2 days in marasmic patients).

67 "Os o XIO) 3E C

2 6 10 14 18 22 26 30 34 38 42 46 50

days elapsed after admission

Figure 10

Timing of deaths after admission. N = 122

4.05(b) Age and BMI

i.all admissions

I'hcrc was a I)i-phasic relationship between age and BMI at admission in the non-oedematous patients. Below a reported age of 25 there was a significant relationship between admission BMI

and age (regression co-efficient BMI on age - 0.24, P-coefficient 0.14, I- statistic = 7.315 d. f. 1, 120, p < 0.05). No such relationship existed in patients with a reported age of 25 and above

(regression co efficient BMI on age = 0.11, p-coefflcient 0.01, h statistic = 2.43 d. f. 1, 189, p > 0.1), see f igure 11. The relationship between age and BMI appeared to be similar for both sexes.

68 20

XX 16

E S’ 14

;XX;

t I I I 10 20 30 40 50 60 70 80 reported age (years)

Figure 11

The relationship between age and BMI on admission. Regression line “A” is a regression of admission BMI on age for patients with a reported age of less than 25 years n = 121, regression coefficient = 0.24 , slope = 0.14 , and intercept = 9.4. The regression line "B" is of admission BMI upon age for patients with a reported age of 25 years or more n = 190, regression coefficient = 0.11, slope = 0.015, and intercept = 12.55.

ii.Survivors

I'he distribution of the admission BMI values in survivors was jiositively skewed. Amongst non- oedematous suiwivors (n = 262), the median BMI at admission was 11.9 kg m ^in the young adult group. This was significantly lower than the older adult group where the median BMI on admission was 13.0 kg m “ (Kruskal-Wallis H = 26.238 d. f. = 1 p < 0.0001). Amongst the oedematous survivors (n = 51), the median admission BMI for young adults was 13.7 kg m^. This was not significantly lower than the median admission BMI of 14.6 kg m “ ^Kruskal-Wallis H =

2.05 d. f. = 1 p < 0.15) in oedematous older adults.

69 iii. Non survivors

In contrast to these survivors, the levels of BMI at admission of those who died did not differ between older and younger adults (ANOVA F = 1.74, d. f. 1, 58, p= 0.2). The following data on the relationship between BMI at presentation and death are therefore not stratified by age.

Section 4.06 BMI data.

70 Table 5 summarises anthropometric data recorded upon admission to the centre. The prevalence of some degree of famine oedema upon admission to the centre was 23.7% (136 / 573). The

prevalence of severe famine oedema upon admission was 16.1% (92/573 ) , 4 and there were no

differences in the prevalence between men and women (Yates corrected — 1.07, p = 0.85). The presence of severe famine oedema was associated with a higher BMI at admission (Kruskal- Wallis H = 29, p < 0.001). This was greater in male than in female patients (partial F statistic = 5.3 p < 0.001).

14 In 4.3% of the 573 admissions aged 15 years and over, for which data is available oedema was recorded as unquantified and therefore coded as grade 3 (see 3.02(c) )

71 Table 5

Data recorded upon admission to the Concern aduit therapeutic centre in Baidoa Somaiia, November 1992 - Aprii 1993. p < 0.001 ANOVA, **, p < 0.005 Kruskal-Wallis t , p 0.01 Kruskal-Wallis ft, p = 0.067 Yates corrected.

Admissions with marasmus or mild Admissions with moderate or severe famine oedema famine oedema

n means difference n means difference male female (male : female) male female (male : female) age, years 479 32 35 -3 92 35 37 -2

weight, kg 370 37.2 31.5 5.66* 83 42.0 33.7 8.334 *

height, m 339 1.68 1.57 0.115* 79 1.7 1.6 0.093 *

BMI survivors 263 12.9 12.7 0.2 51 15.4 13.5 1.966 t (kg/ m’^)

BMI non-survivors 37 12.3 11.5 0.839 ** 24 14.8 13.7 1.062 (kg m'^)

Case fatality rate 442 18% 21% -3% 92 48% 27% 21% tt

72 4.06(a) BMI and the risk of mortality

LAII admissions

BMI values for those that died did not differ significantly between older and younger adults. Therefore, the results for the risk of mortality associated with specific BMI cut-offs, are not stratified by age. BMI values upon admission to the Centre were bimodally distributed, depending upon the presence or absence of oedema. A peak in mortality around a BMI of 11 kg m^ corresponded to the mortality associated with marasmus, another around 15.5 kg m^ to mortalit}^ associated with oedematous malnutrition. The chances of death, related to thresholds of admission BMI, are presented in Figure 12.

73 □ marasmus or mild famine oedema 6 T

■ severe famine oedema 5 -1

"OCO j ■a O

2 -

0 ilill Ml Ml I, I <11 <11.5 <12 <12.5 <13 <13.5 <14 <14.5 <15 <15.5 <16 <16.5 <17 <17.5

Presentation BMI kg m-2

Figure 12

The chances of mortality below different thresholds of admission BMI for oedematous (N=75) and marasmic patients (N=218)

74 ii.Marasmic patients

In female marasmic patients, the level of BMI at admission reflected the chances of mortality, which increased sharply below a admission BMI threshold of 11.0 (Odds ratio = 7.88, corrected

13.8, D. h. = 1 P < 0.0005). No such threshold was observed in male patients were the slight increase in chances of mortality* with decreasing BMI was not significant (Odds ratio for each decrease of 1 kg m “ = 1.4 Likelihood ratio = 1.3 d. f. = 1 p = 0.25) (see Pigure 13).

M male

□ female

CÜ q: V) ■ O T3O

< 11 < 11.5 < 12 < 12.5 < 13 < 13.5

BMI Cutoff (kg m )

Figure 13

The risk of mortality associated with various BMI cut-offs in Marasmic patients (N males 103, females 107)

ili.Oedematous patients

1 he relationship between admission BMI and mortality in oedematous patients was very different to that seen in marasmic patients. Ingure 12 presents the chances of death for all admissions, stratified by the presence or absence of oedematous malnutrition. I he form of this relationship differed betAvecn male and female oedematous patients. Oedematous male patients suffered increased mortality at levels of admission BMI greater than 14.5 kg m “. Phis pattern was not seen in female patients (see f igure 14)

75 ■male

7 - 0 female

6 -

5 -

03 cr (/) ■O •D O 3 -

1 - In .□ I d I oiu □ <11 < 11.5 <12 < 12.5 <13 < 13.5 <14 < 14.5 <15 < 15.5 <16 < 16.5 <17 < 17.5

BMI cutoff (kg m"-2x )

Figure 14 The risk of mortality associated with admission BMI cut-offs in oedematous patients, stratified bv sex (n = 75).

76 4.06(b) The lowest BMI compatible with life.

Mgurc 15 presents data on BMI at admission amongst sur\dvors with marasmus or mild famine oedema.

4 0 T ■ o ld e r adults' age >= 25 geometric mean = 13.5 (95% cl. = 13.2- 13.8), minimum = 10.6

3 0 -- O'Young adults' age 15 - < 25 geometric mean = 12.4 (95% ci. = 12.1 - 12.7), minimum = 9.6

10 11 12 13 14 15 16 17 18 Body Mass Index (kg

Figure 15

Body Mass Index on admission of survivors with marasmus or miid famine oedema. N = 262

I he geometric mean and minimum admission BMI in the older adult group were 13.2 m'kg (95% C. I. = 12.9 - 13.5) and 10.4 ni^kg . Younger adults had BMIs approximately 1 kg m “ below those recorded in patients aged >= 25 years. Geometric mean and minimum BMIs in this group were 12.1 m'^kg^ (95% C. I. = 11.8 - 12.4) and 9.6 m^kg . Mean and minimum BMI values were similar for both male and female patients, once the severit)' of famine oedema was controlled for. Weight loss, due to loss of oedema during the initial phase of treatment, reduced many patients’ BMIs still further. The lowest BMIs recorded in suivivors were 10.1 m^kg for older, and 8.7 m"^

77 kg for younger adults. 22% (n=43) of survivors aged >=25 years, and 49% (n=59) of survivors aged between 15 and 24 years, had BMIs on admission below 12, the level previously considered

the lowest compatible with life. 6 » The survival rate amongst those admitted with BMIs below 11.0 r^kg , was 65 % for older adults (n= 23) and 82% for younger adults (n=34). Mean BMI on discharge was 15 n^kg .

4.06(c) Famine oedema and the risk of mortality.

Mortality rates were higher in patients presenting with severe famine oedema (see Table 5). This was common, seen in 16% of all admissions and 28% of those that subsequently died, and

associated with a poorer prognosis (odds ratio = 2.4, Yates corrected %^ = 11.7 p < 0.001). This

association was stronger in male than female patients (Wolfs’ — 4.59, p < 0.05). The poor prognosis seen in patients with severe famine oedema was predominantly a result of the increased mortality in patients presenting with pitting oedema. Pitting oedema was independently associated with mortality, (Odds ratio for each increase in grade = 1.5, likelihood ratio test 22.2 p < 0.001). Figure 16 shows that male patients had a poorer prognosis at severe grades of pitting oedema than female patients (odds ratio for each increase in oedema grad (male) = 1.7, likelihood ratio test 23.5 p < 0.001, (female) = 1.3, Hkeliliood ratio test 4.4 p < 0.05). Minimal pitting oedema was not associated with an increased risk of death in either sex.

78 12 T

10 -- ■ maie

n female

ro ir 6 o

2 3 4

Grade of oedema

Figure 16

Pitting oedema and the risk of mortality. Data from Bardalli therapeutic centre, Baidoa, Somalia, November 1992 April 1993. N = 534, numbers in each sub-group above each bar.

79 Section 4.07 Discussion

4.07(a) The BMI limit in human starvation.

These data show that the level of emaciation compatible with life is much lower than previously thought. The BMI data from survivors is positively skewed with the numbers surviving quickly decreased below a BMI of 12 kg m'^. The progressively increasing mortality in female marasmic patients with decreasing BMI, below 12 kg m^ (see Figure 13) explains some of this skew. However, the inability of the most severely emaciated patients to stand did not allow for BMI assessment of many of the most emaciated patients. This systematic under-sampling those at the lower end of the BMI spectrum would have introduced a bias into these data, also positively skewing the data presented in Figure 15 The true mean and minimum BMI values in the centres are therefore lower than those recorded here. There are several potential explanations for such low BMI values.

i.Variation due to body shape

Conceivably, such low BMI values could be in part due to a tall, long legged, phenotype. This phenotype occurs in some east African ethnic groups, notably Nilotic peoples, such as the Dinka of southern Sudan and the Samburu of northern Kenya and Australian Aborigines. (Eveleth & Tanner 1976; James, Ferro-Luzzi, & Waterlow 1988; Morgan 1994c; Morgan 1994b) The BMI values in Somalis, a Cushitic people, have however been recorded as lying within the normal range of 19-21. (Branca & Abdulle 1993) Complimentary BMI data, collected in the spring of 1993 from 164 of Concern’s local Somali workers in Baidoa, and 642 adult Somali women living in the villages surrounding Baidoa, also recorded mean BMIs within the normal range (mean BMI = 20.7 kg m^ and 19.9 kg m^ respectively, see Table 27. )

Although the mean population BMI for Somalis lies within the normal range, it is still conceivable that intra-population variation in BMI, independent of nutritional status, could account for the extremely low levels of BMI seen in this population. In Aborigine populations it has been demonstrated that there is large intra-population variation in the sitting height stature ratio (sitting height (SH) / total height (S) or SH/S ratio). In a group of 349 Aborigine adults the SH/S ratio was 0.48, with a standard deviation of 0.02 and range of 0.41 - 0.54. (Morgan 1994b) Such variation is equivalent to standard deviation of ± 7 cm in the height of a population with a mean

80 height of 170 cm, giving 95% confidence limits of 156 cm - 185 cm. As height is a square term in the calculation of BMI, this error is further magnified. In a population with a mean BMI of 12 kg m ^ and height of 170 cm, the 95% confidence limits of BMI dependant solely on SH/S ratio and independent of nutritional status is 10.1 kg m^ - 14.25 kg m^. No SH/S ratio data on adult Somalis appears to exist, however, there is no reason to suppose that intra-population variation in SH/S ratio should be any less than in Aborigine populations. Such variation in individual BMI could explain both the low levels of BMI in patients that survived and also the poor predictive qualities of BMI marasmic patients (see Figure 17). These issues are also discussed in 6.05(a) )

ii.Temperature

Survival at these extremely low BMIs would have been aided by the warm climate that existed in Baidoa throughout the period that the centre was operating. Much of die previous data concerning the limits of survival during famine comes from colder climates. (Henry 1990; Henry 1993) Given the inability of severely malnourished people to thermoregulate, (Burger, Sandstead, & Drummond 1945; Keys 1950a; Leyton 1946; Winick 1979) the limits of survival are likely to be lowest in warm climates. Indeed reports from the camps in the unoccupied zone of France during 1942 describe a tripling of the mortality rates during the winter months. (Zimmer, Weil, & Dubois 1944)

Hi.Dehydration

The dehydration present in many of the admissions to the centre would have reduced admission BMIs. The magnitude of this effect cannot be accurately estimated, as the clinical evaluation of the grade of dehydration in the severely malnourished is difficult. The normal signs of dehydration, such as sunken eyes, decreased skin elasticity, changes in heart rate and reduced peripheral perfusion, can all occur in severely malnourished independently of the state of hydration. It is likely however, that the reduction of admission BMIs due to dehydration was not a major effect, as the inability of most severely dehydrated people to stand, meant that they did not have their BMIs estimated immediately upon admission.

iv.Adaptation to starvation

Previous exposure to CED may have facilitated survival at such low BMIs. Habituation to CED has also been described in Indian labourers who, with a mean BMI of 16.6, were able to function as wage earning labourers. (Shetty 1984) By contrast, similar BMIs, rapidly induced in 32

81 previously well nourished volunteers during the Minnesota experiments, resulted in extremely poor physical and psychological states. (Keys 1950a) The precise mechanism for such adaptation remains unknown and several studies have failed to show much evidence of metabolic adaptation. One possible mechanism is that the acute reduction in the numbers of muscle cells and the increased the ratio of collagen in lean tissue induced by and previous exposure to under-nutrition to some extent persists after recovery. (Stini 1973) Such changes would reduce the resting metabolic requirements of an individual thereby adapting them to future periods of under nutrition. The absence of any way of assessing body composition during this study means such discussions necessarily fall largely outside the scope of this thesis.

v.The availability of small amounts of food

Ih e gradual nature of the reduction in food intake over a period of two years and die sporadic availability of some food would have helped to spare protein reserves in some of our patients. (Sapir et al. 1972) It has been demonstrated that even a small amount of carbohydrate, in the order of 7.5 g kg \ will, by reducing the requirements for short chain carbon intermediaries for anaplerosis, reduce nitrogen loss. (Owen et al. 1998) Such protein sparing would help promote complete exhaustion of fat reserves before FFM became incompatible with life.

4.07(b) Sex difference In adaptation.

The results of this study indicate that men and women responded differently to stannation. Women with famine oedema had a better prognosis than men with equivalent grades of oedema (see Figure 16), whereas severe marasmus appeared to mark a worse prognosis in women than in men (see Figure 13). No differences in admission BMIs, the minimum BMIs in those who survived, or tlie frequency of famine oedema were observed (see 1.01(a) ). These data should be interpreted with caution. The inpatients at the centre were a selected population, people who had either, walked, been carried by friends and family, or been transported to the centre by aid workers. The greater numbers of male medical admissions, many of whom could not walk far, is possibly a reflection of an increased tendency to carry male patients to the centre. If there was such a selection bias, with male moribund patients more likely to present than female ones, the effect would be expected to be greatest at the lower levels of BMI, where the inability to walk was more usual. Consequently, mean female BMI could have been artificially raised and sexual differences in the lower Limit of BMI compatible with recovery obscured.

82 The evidence that women withstand starvation better than men has been summarised by Keys (1950). The finding in this study, that the presence of famine oedema is associated with a better prognosis in women, is in keeping with previous observ^ations. (Gopalan & Ktiishnaswamy 1975; Keys 1946; Knack 1950; Widdowson 1976) Data on gender differences in the prevalence of famine oedema appear equivocal. Amongst more than 3000 inmates in the internment camps of France during 1942, 48 percent of men and 52% of women were affected by famine oedema, although ascites only occurred in the men. (Zimmer, Weil, & Dubois 1944) In other studies, famine oedema occurs less frequently in women. (Gopalan & Krishnaswamy 1975; McCance & Widdowson 1951; Widdowson 1976)

Whether the lowest level of BMI compatible with Hfe, differs between men and women is also unclear. Previous anthropometric data are contradictory on this point. Mean weights of people that died during the Madras famine of 1877, were reported as 77 pounds for both men and women (quoted in Aykroid (Aykroid 1974) ). On the other hand, Henry, reviewing data largely collected during the Dutch famine of 1944-5, concluded that the minimum BMI compatible with hfe is lower in women (see 4.03(e) ). It does seem likely that women are more efficient at sparing protein during starvation than men. At any given body weight, women have a higher percentage of body fat and it has been demonstrated that the loss of body protein is inversely proportional to body fat content . (Forbes 1987) This finding is supported by examination of the percentage of energy derived from protein during starvation, the p ratio. (Payne & Dugdale 1977) C. J. K. Henry has provided evidence that during prolonged complete starvation, this ratio is a fixed constant for any given individual. (Henrj^ Rivers, & Payne 1988) Although similar p ratio estimations during starvation do not appear to have been made for women, it has been demonstrated that in startled rats the p ratio is lower in the females than in the males. (Widdowson 1976) The reduction in the sexual dimorphism in upper arm muscle, that occurs during prolonged human under-nutrition, (Stini 1973) further supports the hypothesis that during starvation women tend to spare protein more efficiently than men.

4.07(c) Famine oedema

The prevalence of famine oedema as a complication of severe adult malnutrition shows great world-wide variation and the 23.5% of subjects with famine oedema, observed in this study, is

common (see Table 6 ). This is similar to the variation seen child oedematous malnutrition in where the prevalence of kwashiorkor or marasmic kwashiorkor, varies between 0 % and over 90%. (Ifekwunigwe 1971; Waterlow 1993) In the three Concern TFCs in Baidoa during 1992 and early

83 1993, the famine oedema was much less than in the adults, (child prevalence = 5.7% (25/411) chi square = 24.7 Odds Ratio 3.1, P < 0.0001 (Collins 1993d) ). This accords with other reports describing a higher occurrence of oedematous malnutrition as age increases. (Gopalan & Kjrishnaswamy 1975; Ramalingaswami et al. 1971) In this study the incidence of severe famine oedema was similar in both sexes and was associated with a markedly increased risk of death, especially in tlie male patients (see 4.06(c) ). Ascites occurred less commonly that pitting oedema and was most often seen in the male subjects. The prevalence of ascites in severe adult malnutrition also appears to be variable. In some famines, the condition has been reported to be ahnost as common as pitting oedema, (Gopalan et al. 1952; Ramalingaswami et al. 1971) in others ascites is only seen in a small proportion of adults, predominantly the male patients, (Zimmer, Wed, & Dubois 1944) or not at all. (Gopalan et al. 1952) The functional significance of ascites is unclear in this study; in bivariate analysis it was associated with a marked increase in the risk of mortality in male patients, however the significance of this association disappeared when pitting oedema was added into the logistic models (see also 6.05(b) ).

Table 6

Variation in the prevalence of famine oedema during severe adult malnutrition

Prevalence of famine Study Location Date of study oedem a (%)

Collins Somalia 1992 23.7

Collins Angola 1993 90

Coliins South Sudan 1993 0

Zimmer et al. France 1942 50

Mollinson Germany 1945 typical finding

McCance et al. Germany 1946 common

Debray France 1945 100

Kevs et al. USA 1945 87

4.07(d) BMI and the prediction of mortality

BMI was related to the risk of short-term mortality in female but not in male marasmic patients. Figure 17 illustrates the receiver-operating characteristics for BMI in the prediction of mortality in female and male patients. It can be seen that, at low levels of BMI, the prediction of female mortality is reasonable, with 56% of deaths correctly predicted and only 14% of survivors

84 incorrectly identified at point “X”. I'his point corresponds to tlie threshold of 11 kg n i“, below which there is a marked increased in the risk of death (see f igure 13). 1 he prediction of death in adults and adolescents by BMI was broadly similar (see f igure 29). In male patients, BMI is little better at predicting death than a random indicator, fo r a fuller discussion of the reasons behind these problems see Chapter 5 and Chapter 6 (see also 4.07(a) i).

100%

P------

80%

1 70% +

Q. x:(/) 60% s T3 50%

l! 40% I o — 1 o (U cn 2 30% all marasmics S 0) Q. 20% female marasmics

“ male marasmics 1 0 %

0% 0% 20% 40% 60% 60% 100%

100 - specificity (percentage of the population selected)

Figure 17

The receiver operating characteristics (ROC) for BMI in predicting short term mortality in male (n = 99) and female (n = 108) patients.

4.07(e) The use of BMI during severe famine.

I'his section discusses the practical aspects of using BMI during famine-relief programmes. I he height and weight measurements required to assess BMI, were problematic during famine, as the

85 absence of chair, bed or stretcher-scales in these situations meant that patients had to be able to stand. The assessment of height was particularly difficult, requiring the subject to stand up straight, a central problem with the use of BMI in the elderly and deformed. (Steele & Chenier 1990) In this study, many of the most severely malnourished could not stand on admission, resulting in approximately 11% of all admissions and 33% of those who subsequently died, not having their BMI assessed on admission. These findings are not unique and several other studies reported that gross weakness, (Winick 1979) flexor contractions, (Mollinson 1946) or scoliosis (Zimmer, Weil, & Dubois 1944) prevented measurements of weight or height in a substantial proportion of the most severe cases. Recognition of this problem has prompted enquiries into other indicators of stature and researchers have demonstrated a good relationship between arm span, femur length and stature. (Feldesman 1992; Feldesman, Kleckner, & Lundy 1990; Reeves, Varakamin, & Henry 1996; Steele & Chenier 1990; Steele & Mattox 1987) However, there are several problems with the use of such proxies in the severely malnourished. Different regression equations must be used in people of different ethnic origin because of the substantial ethnic differences in the relationship between arm span(Reeves, Varakamin, & Flenf)^ 1996; Steele & Mattox 1987) or femur length(Feldesman, Kleckner, & Lundy 1990) and height. The additional mathematics required is problematic during emergency relief operations, where time and train staff are scarce. There is also significant error involved in the estimation of stature at the individual level, where the standard error of the estimate of height from arm span appears to be in the order of 2.5 - 3.8 cm. (Steele & Chenier 1990) The squaring of

the height element in BMI magnifies these differences: For example, taldng a stature of 1. 6 m and weight of 60 kg and a SEE of height from arm span of 3.00 cm, the 95% confidence limits around the estimated BMI would be 21.8 - 25.2 kg m^. This variation would be a function of the height estimation and independent of nutritional status. Such problems may be surmountable and arm span could turn out to be a more stable measurement than height. At present these data do not exist and there is a need for research into the use of such proxies for stature, in the assessment of severe adult malnutrition.

The high incidence of adult famine oedema (see 4.07(c) ) is a serious drawback to the usefulness of BMI as an indicator in severe famine. The association between famine oedema and a poor prognosis on the one hand (see Figure 16), and increased weight on the other results in patients with severe famine oedema often having a poorer prognosis the higher the admission BMI. This is the opposite to the situation in marasmic patients (see Figure 12) and greatly reduces the ability of BMI to reflect severity of malnutrition or risk of mortality in patients with severe famine oedema.

86 Even in non-oedematous populations where the majority of patients can stand the height and weight measurements required in order to calculate BMI are time consuming and difficult to obtain from the most severely emaciated people. During the early phase of an emergency, when there are large numbers of people seeking treatment, this extra time greatly reduces the attractiveness of BMI as a screening tool. In the assessment of the use of BMI screening in such situations such the ease of use is extremely important. In this study, the difficulties in obtaining such measurements under the pressurised conditions that occur during the height of a famine resulted in BMI assessment frequently being omitted, especially in the more severe cases. Individual sitting height measurements required to correct BMI measurements for intra-population variation in the SH/S ratio (see 4.03(c) i), are slow and relatively difficult to perform. These would not be practical in the majority of famine-relief programmes.

One of the main conclusions from this study is that there is a need for alternative indicators for the screening and assessment of adult malnutrition. The possibiht}^ that clinical signs upon presentation may provide good screening tools for malnutrition has been investigated in children but not in adults. (McLaren, Pellett, & Read 1967; McLaren & Shirajian 1969) Chapter 5 examines the use of such models in adults under famine conditions. In children assessment of MUAC, a quick but sometimes less precise, index of nutritional status (Shakir 1975; Velzeboer et al. 1984) is often used during these early phases. There is also some research to show that MUAC maybe a useful nutritional indicator in adults, (Friedman 1991; Gassull et al. 1984; Hey ms field et al. 1984; James et al. 1994; Ohlson et al. 1956) however its use emergency feeding programmes has not yet been evaluated. Chapter 6 examines the use of MUAC for the assessment of severely malnourished adults under famine conditions.

4.07(f) Adult therapeutic care.

The provision of food and medical care in a protected inpatient environment was instrumental in the survival of those admitted. It is Likely that very few would have survived outside the centre, forced to queue for food in the general kitchens. However, since the concentration camps were liberated at the end of the Second World War, adult therapeutic centres have not been a feature of emergency relief programmes. The relatively low case fatality rates, even at extreme grades of emaciation, and the fast rates of rehabilitation reported here, support the feasibility and utility of adult therapeutic centres. These should constitute a standard part of emergency feeding programmes. There is a need for more research into indicators useful in the screening and assessment of admissions to such centres.

87 4.07(g) Conclusions

BMI has been found to have many drawbacks limiting its usefulness in screening mahiourished adults during famine. Under these conditions, BMI assessment is slow and the measurements required cannot be obtained in many patients. The index is also confounded by famine oedema, commonly seen in starving adults. Ethnic origin, age, sex and ambient temperature all affect the sigmficance of BMI values. In particular, the large variation in BMI due to phenot)pe, independent of nutritional status, reduces its usefulness. In the absence of SH/S ratio estimations, the receiver-operating characteristic of BMI in the prediction of mortality, are poor.

The BMI limit below which individual survival is impossible is also dependent upon many factors. Important among these are; the availabilit)'^ of sources of carbohydrate, previous exposure to periods of chronic food shortage, ambient temperature, level of hydration and phenotype. In this study adults survived with admission BMI of less than 11 and adolescents with BMIs less than 10. Even using very low cut-offs values, little above these minimum values, the specificity of BMI in prediction of was only 80%. This is far too low to make the indicator useful for triage. These data strongly support the view that it is not possible to stipulate a minimum BMI value below which recovery is impossible. BMI cannot therefore, be used as a triage tool to identify patients who should be given palliative rather than curative care.

The results of this study indicate a need for research to evaluate alternative indicators, such as MUAC or clinical signs, for the screening and assessment of adult malnutrition. Clinical signs are investigated in Chapter 5 and MUAC in Chapter 6 .

At these extremes of emaciation, the ability to obtain food and medical care passively is instrumental in survival, and it is very unlikely that most of these people would have survived outside a therapeutic centre. However since the concentration camps were liberated at the end of the Second World War, adult therapeutic centres have not been a feature of emergency relief programmes. The comparatively high survival rate and the reasonably fast turn around of patients at tlie centre, achieved in the very difficult circumstances, support the feasibility and usefulness of such centres. These should be included more frequently in emergency feeding programmes. Concern has since opened other adult therapeutic centres in south Sudan and Angola with similar success.

BMI was not shown to be a useful indicator for the assessment of SAM. At the centre we did not manage to record BMIs in 40% of all those who subsequently died and even when it was possible

88 to obtain an admission BMI, this correlated poorly with short-term mortality. At the lowest end of the BMI spectrum practical constraints meant that only the fittest patients, those strong enough to stand and be measured, had BMIs recorded, whilst at the higher end BMI was falsely raised by famine oedema. Consequently screening cut-offs based on BMI exhibited low sensitivity at the high specificity required from an efficient indicator.

89 Chapters Study two: Clinical assessment of severe adult malnutrition

Section 5.01 Introduction

The previous study described many problems associated with the use of BMI in the assessment of severely malnourished adults during famine-relief programmes. This study examines the prediction of short-term mortality (mortality occurring within a therapeutic treatment centre) amongst severely malnourished adult and adolescent inpatients. It examines whether clinical signs at admission, either alone or used as an adjunct to BMI, may be useful in assessing severe adult and adolescent malnutrition under famine conditions.

Section 5.02 aim

1. To assess the performance of clinical models in the prediction of mortality occurring in severely malnourished inpatients in an adult/adolescent TFC.

2. To examine whether the predictive power of BMI in this role can be improved by the addition of clinical variables.

Section 5.03 Background iiterature

Much of the relevant literature is discussed in Section 4.03 and Section 6.03 . This section reviews only the literature relevant to the use of clinical models in malnutrition.

i.The effects of infection

For many years, it has been recognised that malnutrition increases both susceptibility to and the severity of infection. Vitamin, mineral and other dietary deficiencies, depressed cell mediated and humeral immunity, gastric acidity, mucosal integrity^ and altered flora are all known to increase susceptibility to infection. (Tomkins & Watson 1989; Waterlow 1993) The situation in famine is

90 usually worsened by tlie breakdown in public health infrastructure and the congregation of displaced people in crowded and unhygienic conditions. (De Waal 1989) This combination of poor public health environment and immunosuppression means that in famine it is infection rather than absolute loss of fat or fat free mass that kills people. Tliis is different from the situation in developed countries where exhaustion of fat or fat free mass is more often the terminal event. (Korcok 1981; Lei ter & Marks s 1982)

The strong association between infection and mortality means that clinical signs are likely to be useful prognostic indicators. Clinical models for tire prediction of mortality, useful in the screening of admission to chüd feeding centres, have been proposed. (McLaren, Pellett, & Read 1967) Although reported as effective in identifying children at a high risk of mortality, (McLaren & Shirajian 1969) these models have been criticised because of interactions between the features used, such as oedema and hypoprotinaemia, were not taken into account. (Waterlow 1972) In adults, the use of clinical models to assess nutritional status appears to have been restricted to well nourished surgical patients. (Baker et al. 1982) To date, similar assessments have not been made in severely malnourished adults or during famines.

Extreme conditions often characterise famine relief programs. The levels of need are very high and trained staff, equipment and buildings always scarce. Sometimes crowds of several thousand people may gather, attempting to gain admission to centres. Such centres may readily become over-filled, disorganised and dysfunctional if patient selection and prioritisation is not efficient. Tools to assess patients must be quick, simple and reliable, even when used by minimally trained staff, who maybe barely literate. This study investigates the effectiveness of two assessment tools, BMI and models using readily ascertainable cknical signs, during the height of a major famine.

Section 5.04 Subjects and methods

5.04(a) Subjects

This retrospective cohort study involved clinical record data of 393 patients admitted to the Concern Worldwide Adult Therapeutic Centre, Baidoa, Somalia, between November 4, 1992, and March 15, 1993 (see Chapter 3 :3.01(b) i. Section 3.02 . Criteria for inclusion in this retrospective analysis were: a BMI at admission of 13.5 kg/m2 or less or any signs of oedematous malnutrition; a reported age of 15 years or older; the presence of an inpatient treatment and clinical data card.

91 and no record that the patient was admitted because of medical criteria alone. Of these 393 patients, 10 (2.5%) were excluded owing to the absence of outcome data. 383 patient who met the admission criteria and for whom outcome data was available were included in the study. The mean reported age of these 383 subjects was 33.0 with no significant age difference between the sexes {t — 1.32, d. f. = 380 p = 0.19. 48.4 % of those admitted were male. Mean BMI at admission was 12.7 kg m^ (SD 2.0); mean weight was 34 kg (SD 7.2). Mean admission BMI of sur\dvors was 12.6 kg m^ and 12.9 kg m^ for non-survivors ( see Table 7).

Table 7

Characteristics of the study popuiation

Category All patients Survivors Non-survivors

N 383 292 91 Mean = 33.0 Mean = 32.7 Mean = 33.8 SD = 16.0 SD= 14.8 SD = 16.5 Age (self-reported in years) Minimum = 15.0 Minimum = 15.0 Minimum = 15.0 Maximum = 89.0 Maximum = 89.0 Maximum = 80.0 Males =185 Males = 140 Males = 45 Sex Females = 197 Females = 152 Females = 45 Missing = 1 Missing = 0 Missing = 1 Mean = 12.7 Mean = 12.6 Mean = 12.9 SD = 2.0 SD = 2.0 SD = 1.9 BMI at admission (kg m-2 Minimum = 9.0 Minimum = 9.0 Minimum = 9.3 Maximum = 24.0 Maximum = 24.0 Maximum = 17.6 Missing = 87 Missing = 48 Missing = 39

Mean = 34.0 Mean = 34.0 Mean = 34.1 SD = 7.2 SD = 7.2 SD = 7.4 Body weight (kg) Minimum = 19.2 Minimum = 21.0 Minimum = 19.2 Maximum = 59.5 Maximum = 59.5 Maximum = 53.4 Missing = 42 Missing = 21 Missing = 21

5.04(b) Methods

Details of the centre are presented in 3.01(b) i, and details of the supervision, rehydration, medical treatment and discharge criteria in 4.04(b) . Methodological problems are described in Section T03).

92 LData collection

Details of data collection and coding are presented in 3.02(c) . During this study, the level of supervision of data collection was variable. Occasionally, expatriate access to the Centre was restricted for part of a day due to the gun fighting, bandit attacks, attempts at hostage-taking, and military occupations that occurred in Baidoa during the study period. In addition, the author (SC) was absent from the Centre due to illness requiring medical evacuation and for rest and recuperation, for approximately three weeks during the nineteen-week study period. During this time data collection was supervised by an experienced expatriate nurse.

il.Data analysis

Bivariate (i. e. two-by-two table) and multivariate logistic regression analysis were used to assess tlie association of clinical features at presentation with mortality in the centre. Variables that were not independently associated witli mortality were eliminated from the logistic models in a backwards stepwise fashion using estimation techniques. The odds ratios of the three signs independently associated with death (severe oedema, inability to stand, and apparent dehydration) were used to construct a predictive model. If a patient exliibited none of the signs, their score was zero. If they exhibited one sign, their score was the odds ratio associated with that sign. If more than one clinical sign was present, their score was the sum of the odds ratios associated with each sign. The sensitivity and specificity of predicted mortality at score inten^^als of 0.5 were calculated and receiver operator characteristic (ROC) curves plotted. ROC cuiwes plot sensitivity against

1 0 0 -specificity and are a useful means of assessing the abTity of an indicator to discriminate between healthy and diseased persons or, in this case, between patients with differing prognoses (see 3.02(d) iv). A simpler model, using a count of clinical signs, was also constructed. In this model, the score for each patient was the number of three relevant clinical signs exhibited.

The association between BMI and survival was explored in a systematic manner. A series of ‘indicator’ variables was created using successively higher BMI cut-points at intervals of 0.5 kg m^ (i. e. variables were created to indicate whether an individual BMI was below 10.0, 10.5, 11.0, 11.5, 12.0, 12.5, and so on). Each indicator was tabulated against survival. The indicator with the most significant positive association (BMI < 11.0 kg m^, odds ratio = 2.44, 95% C. I. = 1.11-5.32, Yates corrected chi-square = 5.22, p < 0.05) was then included in the predictive model.

A simulation exercise was undertaken to validate the methods used to construct the models. Each run of the simulation involved splitting the dataset into 2 parts by randomly selecting (using a

93 pseudo mndom number generator) approximately one half of the available records as a training dataset and using the other half of the data as a validation dataset. The odds ratios and uncorrected chi-squares for the association between predictor variables and death were calculated using the training dataset. Sums of odds ratios and sign-counting models were then constructed using those variables with significant positive associations with death, defined as any variable having an odds ratio of greater than 1.0 and an uncorrected chi-square of greater than 3.84 (p <= 0.05). Models constructed using these variables were then tested on the validation dataset. The simulation was run1 0 0 0 times.

All data entry and analysis were performed using Epilnfo, (Dean et al. 1990) Logistic(Dallal 1988) and Microsoft Excel. (Microsoft Corporation 1994) (see 3.02(d) ).

94 Section 5.05 Results:

5.05(a) Clinical signs at presentation and prognosis

91 of the 383 patients (23.8%) included in the study died. Median time to death was 8 days (range 1-53 days).

Table 8

Variables Independently associated with survival

Category Survivors Non-survivors Unadjusted odds ratio (95% Cl) N 292 91 Inability to stand 7% 19% 2.96 (1.40- 6.26) t

Dehydration 22% 43% 2.73 (1.60 - 4.66) t

Severe oedema 19% 36% 2.45 (1.41 - 4.27) *

Dysentery 29% 38% 1.50 (0.88 - 2.53)

Anaemia 21% 25% 1.28 (0.71 -2.31)

Diarrhoea 34% 37% 1.18 (0.70 -1.99)

Chest infection 58% 57% 0.97 (0.58 -1.61)

t Corrected. Chi -square == 9. OOr P = 0. 005 f C orrected Chi-squ are == 15. 01. P < 0. 001

* Corrected Chi -square == 10. 94, p = 0. 001

Table 8 presents the clinical sign variables that were independently associated with survival and the ORs associated with these variables. Severe oedema, inability to stand, and apparent dehydration were indepcndcndy associated with mortality. Ascites was not associated with survival

Table 9 presents the sensitivity and specificity for the prediction of death in aU patients (n = 383) for all combinations of clinical models created using oedema, inability to stand and apparent dehydration. The most discriminatory model (i. e. the model with the greatest sum of specificity and sensitivity) was that based on the presence of any one of these three signs. This model predicted 77% of deaths at a specificity of 59%.

95 Tables

The prediction of death amongst all patients (n = 383) by clinical signs

Numbers of deaths Number of survivors Sign(s) Sensitivity Specificity Accuracy PPVNPV predicted to die predicted to die

Single

De 39 63 43% 78% 70% 38% 81% Od 33 55 36% 81% 70% 38% 80% Ns 17 21 19% 93% 75% 45% 79%

Combined using AND

Od & Ns 3 3 3% 99% 76% 50% 77% De & Ns 8 4 9% 99% 77% 67% 78% De & Od 9 12 10% 96% 75% 43% 77% De & Od & Ns 1 1 1% 100% 76% 50% 76% Combined using OR

Od or Ns 47 73 52% 75% 69% 39% 83% De or Ns 48 80 53% 73% 68% 38% 83% De or Od 63 106 69% 64% 65% 37% 87% De or Od or Ns (any one sign) 70 121 77% 59% 63% 37% 89%

Any two signs 18 17 20% 94% 77% 51% 79%

De = Dehydration, Od = Oedema, Ns = Inability to stand

96 Figure 18 shows the ROC curves for the prediction of short-term mortality using the sum of odds ratios clinical model and BMI. The prediction of death by the sum of odds ratios of severe oedema, inability to stand and apparent dehydration performed at the highest level of sensitivity and specificity. Prediction of death using BMI alone in these patients (n = 383) yielded results that were worse than random. When BMI was used in a sub-population, selected to include only marasmic patients who could stand (n = 218), performance was similar to the clinical model, predicting 46% of deaths at a specificity of 85%. When use of BMI was further restricted to include only marasmic females who could stand (n = 108), BMI performed better than the clinical model, predicting 56% of deaths at a specificity of 87%.

97 100%

iJ> ■H > •H 4-) •H dCO 0 ) (0 40%-

0 % 20% 40% 60% 80% 100%

100 - specificity

----BMI in marasmic patients — o— BMI in edematous patients

— BMI in all patients — O— Sum of odds ratios in all patients

■H BMI in female marasmic patients Random indicator

Figure 18

ROC curves for the prediction of mortality by the stepped-sum clinical model and BMI

I'hc RU ( 2 cunc for the model constructed by counting clinical signs was identical to that produced using the sum of the odds ratios clinical model. Addition of a BMI marginally improved these models. ROC cuiwes for these models are presented in Figure 19.

98 100 %

80% I

60% >. >

40% -■^■-Sign count

-X— Sum of odds ratio 2 0 %

- -A-- Sum of odds ratio with BMI < 11.0

0 %

0 % 10 % 2 0 % 30% 40% 50% 60% 70% 80% 90% 100 %

100-speci ficity

Figure 19

The prediction of short-term mortality by counting clinical signs with and without a BMI threshold (<= 11.5 kg m'^), in 383 adult and adolescent inpatients.

Section 5.06 Discussion

1 he risks of mortalit\^ associated with apparent dehydration, inability to stand and famine oedema were indejoendent of each other. A patient presenting with more than one of these signs therefore had a risk of mortality etjuivalcnt to the sum of the risks of all the signs that were present. A scoring system based upon the sum of the odds ratios of these three signs, predicted 77% of deaths at a specificity^ of 59%. A simplified model, using only the presence or absence of the 3 clinical signs, predicted deaths in a near identical fashion (see ffigure 19). The three clinical signs used were quick and easy to elicit in all patients and the application of the model involved only counting. (Consequently the model can be readily taught to local workers and is suitable for use in the field, even during the early phases of a famine relief programme.

I’hc capacity of BMI to predict death was limited. Only in female marasmic patients who could stand, did BMI perform better than the clinical models. When non-marasmic patients and those

99 unable to stand were included (i. e. the entire patient cohort), the prediction of mortality using BMI was worse than random (see Figure 18). The assessment of BMI was also difficult and time- consuming. Dehydration present among patients admitted to the centre reduced admission BMIs. Because apparent dehydration was also a sign of poor prognosis, its presence tended to make BMI a better predictor of death in the centre. The inability of most severely dehydrated people to stand meant that they could not have their BMI estimated within 2 days of admission and were therefore not included in the BMI prognostic models. It is therefore likely that the reduction of admission BMIs due to dehydration was not a major effect.

The presence of severe oedema was associated with a poor prognosis and increased admission BMI. This confounding meant that the ability of BMI to predict death among patients with oedema was worse than random. Because severe oedema was common, BMI was not useful as a prognostic indicator when applied to the entire centre population

These results indicate that at the extremes of emaciation seen in adult therapeutic feeding centres, clinical illness is a better indicator of prognosis than the degree of wasting as defined by BMI. A model using the presence or absence of readily discernible clinical signs is useful in identifying patients at high risk of death. The high-risk patients identified by these models can then be moved to specialised areas within feeding centres with the most motivated staff, more medical supervision and if necessary higher care giver : patient ratios.

This is a different role to that of screening admissions for entry into feeding centres, a research priority in severe adult/adolescent malnutrition. (Collins 1993a) Rather than identifying inpatients with a poor prognosis, a screening indicator must differentiate between those in the general population who are likely to respond to treatment if admitted to a feeding centre, but who wiU die if not admitted. The specificity and sensitivity requirements for screening are different to those required for prognostic indicators. To be useful in famine relief programs, where there are often large crowds of people attempting to gain entry into feeding centres, a screening indicator must have a high specificity. This is not the prime requirement for a prognostic indicator. This study has demonstrated that the sign-count model is sensitive for predicting death within a therapeutic feeding centre. It has not assessed its potential for use as a screening tool to exclude individuals from admission to feeding centres

This potential use for screening admissions to therapeutic feeding centres needs to be investigated further. Although the model is somewhat non-specific for identification of patients who would die despite having been given treatment in the centre, it is likely that in the absence of treatment,

100 many more of the individuals identified would have died. The model might therefore be more specific if used for screening. Despite this effect, it is likely that the relatively low specificity of the clinical model proposed here means that on its own it is not well adapted for screening. To increase the model's specificity for identifying malnourished individuals at greater risk of dying, an indicator of nutritional status would need to be added to the model or combined with it in a two stage screening procedure to improve specificity. MUAC shown in Chapter 6 to be a useful indicator of adult nutritional status might be suitable (see pages 105-134). Addition of MUAC would differentiate between those with clinical illness but no malnutrition, better treated in medical units, from those with both illness and malnutrition, best treated in specialised feeding centres.

By weighting the model appropriately, having an MUAC below a certain threshold could be made a necessary prerequisite for selection. This would ensure a high prevalence of malnutrition in the population assessed by the sign-count part of the model

Within this selected population, the specificity of the clinical signs for identifying death among persons who did not receive treatment would be high, fulfilling the requirement for screening.

Use of this single model would be equivalent to application of a 2 -stage screening procedure.

Such a two-stage system is similar to that employed by the author when confronted by large crowds of malnourished adults. In such situations, a rapid visual inspection and a very brief clinical examination are under-taken outside the centre to assess the fat in the upper arms. This is followed by more involved anthropometric and clinical inspection inside. In this “ad hoc” system, the initial screen outside the centre increases the prevalence of those requiring urgent assistance that are allowed through the gates to be assessed by a more time consuming clinical examination. In this selected population, the higher prevalence of life-threatening malnutrition makes the relatively low specificity of the clinical screen less important. The sign count model proposed here, combined with an MUAC cut-off, would formalise this method, making it more useful to workers with less experience. The threshold of both MUAC and clinical score chosen as a screening cut-off will depend upon the context and the relative balance of resources and needs. (Habicht 1980)

Emergency relief programs must ensure that the majority of a population has access to the minimum requirement to maintain life. This requires prioritising lower-input interventions with a large coverage of the vulnerable population over high-input services treating a relatively few. , providing at least 8786 kj from grains, legumes, and vegetable oil; adequate water; sanitation; basic

101 health care; and dry supplementary feeding, must, therefore, form the basis of any relief program. (Project team 1999) Therapeutic feeding centres are only efficient and effective if these basic pre requisites are in place. This hierarchy of interventions constitutes a form of triage. During the initial stages of famine, need wiU usually outstrip the available resources and the focus should be on ensuring that the general ration is adequate before targeted feeding centres are established. At this stage, those too malnourished to survive on this basic support will die. As resources increase it becomes possible to undertake therapeutic feeding, although initially clinical triage may be necessary. For example, in Wau town in South Sudan during June - August 1998, over 100 therapeutic centres would have been required to treat the estimated 16, 0 0 0 children adults requiring therapeutic feeding. (Buchanan-Smith et al. 1999) At that time, there was only one a single 24-hour therapeutic feeding centre with a maximum capacity of 400 patients. In this context, given the limited means available, selecting only those patients with a good chance of survival would have optimised the efficient use of resources. In this study, the combination of two relevant clinical signs predicted death at sensitivities of greater than 95%. In Wau, the combination of these two signs would have been a useful triage tool determining the bottom end of the spectrum of severity that the centre would admit. Relief programs are rarely totally overwhelmed and it is important that the existence of a triage tool does not undermine international will to provide sufficient humanitarian assistance during emergencies.

Screening and prognostic indicators are also different to markers of improvement during treatment. In this study, maintenance of fluid balance and hydration, disappearance of oedema and a steady increase in weight in the absence of increasing oedema, were all markers of improvement (see Section 4.05 , Section 4.06 and Section 7.04 )

Ideally, a study evaluating the relationship between clinical signs and the risk of mortaHt)^ in a general population suffering from famine is required to establish relevant clinical screening criteria. However, conducting such broad-based population studies is probably not feasible during famine and none have yet been undertaken in these conditions, for any nutritional indicator, in either adults or children. Given these difficulties, evaluating prognostic models in selected feeding centre populations is a feasible alternative. It is certainly better than the current absence of information on prognostic indicators in famine.

In conclusion, counting simple clinical signs is a useful method of assessing prognosis in severe adult and adolescent malnutrition. The addition of an anthropometric indicator may improve the performance of tliese models. However, difficulties in obtaining BMI estimations and its

102 inappropriateness in many cases make it unlikely to be useful at the height of a famine. MUAC might prove to be more useful in this context and a combination of MUAC and clinical models might also be useful in screening admissions to therapeutic feeding centres. More work on identifying relevant clinical signs and combining them with MUAC in such models is required.

103 104 Chapter 6 Study three: MUAC and the assessment of severe adult malnutrition

Section 6.01 Aim

To compare the use of BMI and MUAC in the assessment of SAM.

Section 6.02 introduction

The results of study 1 indicated that there are many problems involved in the use of BMI for the assessment of severely malnourished adult during famines (see Section 4.07 ). This study attempted to investigate the use of MUAC assessment as an alternative to BMI. It took place in Ayod, a small village in South Sudan, during March - April 1993 and Melanje, a regional capital in Angola, during November-December 1993. Supplementary data from Somali women attending village food distributions in the Bay region of Somalia, during 1993, are also included. The aim was to examine the potential usefulness of MUAC and BMI in the assessment of severely malnourished adults during the height of a famine. The aim was for this to be achieved by comparing estimations of MUAC with those of BMI and relating each to short term outcomes such as mortality and length of stay in the centres. Methodological issues surrounding this aim are discussed in Section 3.03 . Unfortunately in both sites it was impossible to evaluate short-term outcome and only the comparisons of MUAC and BMI are presented here. In South Sudan, during the second week of April SPLA forcers loyal to John Garang attacked Ayod, forcing the villagers to flee and Concern to evacuate all staff. Tragically when Concern workers were able to return to Ayod, the centre had been destroyed and the occupants had either fled or been murdered. In the Angola study, the majority of patients were outpatients and there was a high incidence of self-discharges. The size of the city and the insecurity during the study period precluded foUow-up of these patients once they ceased to attend the centre.

105 Section 6.03 Background literature

This section reviews the evidence indicating that MUAC might be a useful indicator of adult malnutrition during famine.

6.03(a) Body Mass Index (BMI)

For a discussion of the Literature relevant to the use of BMI in the assessment of acute malnutrition, see Section 4.03 .

6.03(b) Middle upper arm circumference (MUAC)

I.ChildMUAC

In children, MUAC has been demonstrated as useful in the assessment of nutritional status. (Briend & et al. 1986; Burgess & Burgess 1969; Jelliffe & Jellife 1969; Shakir 1975; Velzeboer et al. 1984; WHO 1986) It is efficient at predicting mortality and, in some studies, MUAC alone (Alam,

Wojtyniak, & Rahaman 1989; Briend & Zimick 1986; Vella et al. 1994) or MUAC for age, (Chen, Chowdhury, & Huffman 1980) predicted death better than all other anthropometric indicators.

This advantage of MUAC was greatest when the period of follow up was short. (Briend, Wojtyniak,

& Rowland 1987) Debate continues over whether the increased sensitivit)^ and specificity of MUAC in predicting death is due to the confounding effects of age as MUAC is known to preferentially select younger children. It is thus possible that the higher mortalit)^ in younger children from non-nutritional causes is the factor behind MUAC predictive qualities. However, the inclusion of age in logistic models used to evaluate the predictive power of MUAC, made little difference to its performance, (Alam, Wojtjmiak, & Rahaman 1989; Briend & Zimick 1986; Vella et al. 1994) indicating that preferential section of younger children is probably not of prime importance. The measurement of MUAC is easy to perform even on the most debilitated children and measurement techniques can be taught readily to minimally trained health workers. (Velzeboer et al. 1983) It is thus potentially suited to screening admissions to feeding centres during emergency aid operations. However, the use of MUAC in emergencies is still controversial, and disagreement over the preferential selection of younger children, the levels of cut-offs used, the efficiency of a two phase screening process and poor reproducibility in the measurement continue. (Bern &

Nathanail 1995; Lindtjorn 1985; McLaren 1986; Ross et al. 1990; Ross, Berry, & Taylor 1986)

106 Consequently, some humanitarian relief agencies remain sceptical about the use of MUAC in emergencies, (Diskette 1995)

UAdultMUAC

At present during famines, MUAC is only recommended for use in cliildren between 1 and 5 years of age and is not used routinely in the assessment of adult malnutrition. (Boelaert et al. 1995; Hakewell & Moren 1991; Young 1992) However, measurements of adults mid-arm circumference are known to reflect changes in adult body weight. (Ohlson et al. 1956) Estimates of arm muscle area (AMA) or corrected arm muscle area (CAMA), corrected for humerus cross-sectional area,

can also be derived from MUAC and measurements of triceps skin-fold thickness (TSF). , 5 These have been incorporated into diagnostic schemes for adult malnutrition in hospitals(Gassull et al. 1984) and used as prognostic indicators in the elderly and cancer patients. (Friedman 1991; Ileymsfield et al. 1984) In the elderly, CAMA area is reportedly a more sensitive indicator of changing nutritional status than BMI(Potter et al. 1995) and a CAMA threshold of <= 16.9 cm^ has been shown to sensitively predict impending death. (Friedman 1991) Heymsfield demonstrated that in secondary malnutrition once AMi\ fell to 9-11 cm^ death was imminent. Post mortem examination of these patients, revealed no other immediate reasons for death other than profound wasting. (Heymsfield et al. 1984) This AMA level therefore appears to represent a threshold, below which the amount of FFM is incompatible with life.

15 Using the equation: CAMA (cm^) = [AMC {cm )Y /(4;r - lOcm^) , where AMC = MUAC— TtTSF dîcyn’sficld, McManus, & Smith 1982)

107 □ METASTAHC CARCINOMA OF THE PANCREAS. 36 cf ■ METASTATIC BREAST CARCINOMA. 42 9 O METASTATIC OROPHARYNGEAL CARCINOMA. 80

o c/3

OC <

12 16 20 24 28 32 36 40 44 48 TIME (WEEKS)

Figure 20

The relationship between AMA and point of death in secondary malnutrition. Left and right ordinates also plot data from normal ^ and malnourished O m en (left) and women (right). Reproduced from Heymsfield 1984.

(I Icymsficld ct al. 1984)

Lhifortunatcly, these studies did not include data on the relationship between MUA(] and mortality and, although AMA and CAMA are derived from MUAC, there are important differences. MUy\C is determined by several components; arm muscle, subcutaneous fat, the humerus, neuro vascular bundle and skin. Although bone demineralisation and a decrease in density do occur in prolonged severe staiwation percentage losses are much less than weight. It is therefore reasonable to assume that the cross-sectional area of the humerus remains relatively constant even during severe malnutrition. I'he percentage reduction of skin and the neuro-vascular bundle, consisting of a large proportion of collagen, is also less than that of fat or muscle protein. (Picou, I lalliday, & Carrow 1966) I’he determinants of MUAC that change during starvation are therefore arm muscle and sub-cutaneous fat. These have both been identified as the most important determinants of sur\ ival in Pb>M (see section 4.03(e) iii). Phis central role of fat and muscle in determining sundval

108 during starvation might mean that MUAC, a reflection of both, is more functionally relevant than CAMA, which reflects muscle mass alone. The relationship between peripheral subcutaneous fat and total body fat is however, not simple. The distribution of fat in the body varies with sex, ethnicity and age, (Bishop, Bowen, & Ritchey 1981; Heymsfield et al. 1984; Wang et al. 1994) and during under-nutrition, the amounts of subcutaneous and central fat do not change in direct proportion to each other. As starvation progresses however, the utilisation of subcutaneous fat in severe adult malnutrition (see section 4.03(e) iii) would be expected to increase the correlation between MUAC and CAMA.

iii.Sexual dimorphism in arm circumference

In normally nourished North American adults, sexual dimorphism in MUAC, triceps sltin fold thickness and AMA exist: men have larger AMA and MUAC and women larger TSF. (Bishop, Bowen, & Ritchey 1981) Tliis dimorphism appears around puberty(Stini 1973) and has been attributed to differences in androgenic and oestrogenic steroids increasing the bulk of skeletal muscle in men and subcutaneous fat in women. (Stini 1973; Widdowson 1976) As the TSF is larger in women the degree of sexual dimorpltism in AMA is greater than in MUAC. The extent of sexual dimorphism in both MUAC and ARIA however, appears to be variable, dependant on cultureQames et ai. 1994) and nutritional status. (Stini 1973) Dimorphism in MUAC can actually disappear in cultures where a disproportionately large physical work burden is placed upon women, particularly if this is during adolescence. (Dettwyler 1992)

Section 6.04 Subjects and method

6.04(a) Subjects

A group of severely malnourished adults in South Sudan with a low prevalence of famine oedema was chosen as a study population. The absence of oedema in this population minimised the confounding the relationship between BMI and MUAC by oedema which has previously been shown to increase body weight independently of changes in Fat and Fat Free Mass (see Section 4.07 ). The other study group is malnourished adults from Angolan with the majority of subjects suffering from oedematous malnutrition. This group was chosen in recognition of the importance of famine oedema as a complication of severe adult malnutrition. (Debray et al. 1946; Keys 1950a; McCance & Widdowson 1951; Winick 1979; Zimmer, Weil, & Dubois 1944)

109 Two groups of more normally nourished adults from Somalia are also included in the analyses. One study group was obtained by taking a systematic random sample of adult Somali women attending the Concern Worldwide general-ration distributions in the Bay region of Somalia during 1993. The other is the Concern Worldwide staff in Somalia during 1993.

i.South Sudan

During April 1993, BMI and MUAC were assessed in 98 severely malnourished Nuer adult inpatients at the Concern Worldwide feeding centre in Ayod, south Sudan. The population was suffering from acute malnutrition resulting from escalation in the war between opposing factions of the Sudanese Peoples Liberation Army (SPLA). This acute malnutrition was superimposed upon a background of chronic malnutrition, a result of the long-term war between the SPLA and the Sudanese government forces. Twenty five years of war had caused the death of almost all the cattle, the traditional source of livelihood for the Nuer. Superimposed upon this background of chronic mahiutrition, patients were suffering from acute, non-oedematous malnutrition, complicated by a variety of infections commonly seen during famine, of which dysentery, respiratory infection, helminthiasis and malaria were the commonest. An epidemic of Kalar Azar was reportedly also present in the area (personal communication A. Davis, MSF Holland). All patients who could stand were measured, weighed, and had their MUAC assessed during the first week after their admission, using standard techniques (see 3.02(b) ). The sacldng of Ayod and the destruction of the centre with the murder of many of the inpatients in early April, ended the study and all patients were lost to follow up.

iLAngola

During November and December 1993, MUAC and BMI data were collected from adult and adolescent patients at the Concern Worldwide feeding centre (Angola (1), N =769) and Caritas adult feeding centre supported by Concern Worldwide (Angola (2), N=290), in Melanje, Angola. The city’s population was suffering from acute malnutrition caused by several months of a total ground blockade of the government held town by UNITA guerrillas. Subjects were severely malnourished, suffering from acutely oedematous or non-oedematous malnutrition complicated by a variety of infections of which helminthiasis and respiratory tract infection were the commonest. The incidence of dysentery was extremely low. The majorit}^ of subjects were treated as outpatients with only those who were unable to return home at night sleeping in the centre. Tliis

110 made follow up difficult, as most undertook took self-discharge and were lost to follow up once they felt better.

Hi. Soma lia

During 1993 MUAC and BMI data were collected from a sample of 642 rural Somali women attending village level general-ration distributions in the Bay region of Somalia. In addition, similar data were collected from all of Concern Worldwide’s 145 local employees in regional capital, Baidoa^ The sample of rural women was selected using a systematic random sampling technique, during Concern Worldwide general-ration distributions in the villages during July 1993. This distribution began in January 1993 and provided a monthly, family ration of approximately 15 kg of maize per person, delivered to the women of each family. Standard techniques were used to assess MUAC and BMI (see 3.02(b) ).

6.04(b) General methods

See Chapter 3

6.04(c) Analysis

Sensitivity/specificity analysis, based upon 2*2 tables, and linear regression techniques were used to compare MUAC and BMI data. To make the results comparable with previous literature, the relationships between MUAC and BMI are presented using the regression of MUAC on BMI (regressions derived from MUAC = b * BMI + c (b = slope and c = intercept). Multiple linear regression techniques using EPI-Info(Dean et al. 1990) , according to the method of Draper and Smith, (Draper & Smith 1981) were used to examine the relationship between BMI and MUAC whilst controlling for confounders such as oedema, sex, ethnic group and presence of acute malnutrition. Before the regression analysis was used to compare the Sudanese, Angolan and Somali data, outlying data points were removed. Outliers were defined as those with a mean BMI or MUAC greater than four standard deviations away from their respective population means. Only one data point from the Angolan data was excluded according to these criteria. Subsequently, after tlie initial linear regression of the groups of data, data points having (residuals) ^ greater than four standard deviations away from the mean of the (residual) ^ were also excluded

111 from further analysis. This excluded 1.13% of the 1, 944 data points from further analysis, The degree to which the two indices identified the same individuals in the Sudanese data was analysed in terms of the sensitivitj’^ and specificity of MUAC in identifying those individuals below specific BMI cut-offs. In this analysis BMI was the dependent variable and the regression Hne BMI = b * MUAC + c was used. Integer values of BMI between 16 and 10 were assigned as cut-off points and the corresponding mean MUAC values predicted from the regression line. The choice of BMI as the reference indicator was somewhat arbitrary as, although an accepted indicator of adult CED, BMI has not yet been shown to be a criterion standard for the assessment of acute adult malnutrition (see Section 4.03 ), Data were analysed using EPI-Info, (Dean et al. 1990) Microsoft Excel(Microsoft Corporation 1994) and Curvefit Professional. (Hyams 1993) See also 3.02(d) .

6.04(d) Results

i.MUAC and BMI in adult marasmus

dhe MUAC and BMI values recorded from 98 marasmic patients in the South Sudan centre are presented in Table 10. Mean BMI values were higher for male than female patients (mean (kg m ^); male 13.3, female 12.6, (ANOVA) F = 3.26 d. f. = 1, 96, p = 0.07). This result is of borderline significance. The mean MUAC was significantly higher for male than female patients (mean (mm): male 188, female 179, (ANOVA) F = 4.24 d. f. = 1, 96, p < 0.05).

BMI measurements were difficult to perform on the patients who experienced problems in standing. Many patients were very slow to step onto the scales and had great difficulty in standing up straight enough to be measured. Other inpatients could not stand at all and BMI measurements were not obtained. These patients were excluded from the study. MUAC measurements were much easier to perform on all subjects and could be taken with the patient standing, sitting or, in extreme cases, lying.

Figure 21 presents a scatter diagram of MUAC and BMI measurements with the male and female regression lines for MUAC on BMI. There was a highly significant correlation between MUAC and BMI, with a correlation coefficient of 0.88 (95% C. I. = 0.82 - 0.92, F = 319, d. f. = 1 , 96, p < 0.001). Correlation coefficients were similar for male and female patients, male = 0.79 (95% C. I. = 0.64 - 0.88, F = 65, d. f. = 1, 39, p < 0.001), female = 0.92, (95% C. I. = 0.87 - 0.95, F=

16 Two of the 98 in the Sudanese sample, two of the 87 non oedematous Angolan, three of the 241 mildly oedematous Angolan, three of the 731 oedematous Angolan subjects, eight of the 642 Somali women and two of the 145 Concern

112 303, d. f. — 1, 55, p < 0.001). There was no significant difference between the sexes in the relationship expressed by the regression line of MUAC on BMI.

The proportion of the population below various cut-off values and an analysis of the sensitivity, specificity, and predictive power of MUAC with respect to BMI are presented in Table 11.

The proportions defined by the two indicators were similar, although at the lower end of the BMI range MUAC tended to define a smaller proportion of the population than the comparable BMI cut-offs. The sensitivity-specificity analysis indicated that the individuals identified by the two indices were also similar, particularly within the BMI range of 16 - 13 kg m^. The positive predictive value for the prediction of BMI by MUAC remained high for aU the cut-off values. Below an MUAC of 18.5 cm (equivalent to a BMI of 13 kg m^), the relative proportion of the population defined by MUAC and the sensitivity of MUAC in predicting BMI feU. This cut-off point corresponded to a position on the receiver operating characteristic (ROC) curv e for the prediction of BMI by MUAC, where sensitivity was maximised without appreciable loss in specificity (see Figure 22).

Somali workers, were excluded.

113 Table 10

BMI and MUAC values in 98 severely malnourished adults belonging to the Nuer tribe in South Sudan

BMI MUAC kg/(m^) (cm)

sex n median minimum lower upper normal values median minimum lower upper normal values quartile quartile (mean ± SD) t quartile quartile (mean +/- SD) t

male 41 13.4 10.2 11.8 14.4 20.0 ± 2.2 19.0 15.0 17.8 20.0 25.6 ± 2.4

female 57 12.5 9.3 11.3 13.9 20.4 ± 2.7 17.5 12.5 16.0 19.5 24.8 ± 2.6

t Normal values based upon a sample of 5669 adults from developing countries a #n-

114 Table 11

Comparison of the sensitivity and specificity of MUAC and BMI cut-offs in 98 severely malnourished Nuer adults from south Sudan. The sensitivity, specificity and predictive values for the MUAC cut-offs are relative to the BMI cut-offs. ■-sensmvny-cji ----- “TTwgeti vtrpi eoiuiive numbers proportion numbers proportion MUAC cutoff specificity of MUAC positive predictive value of MUAC cut below BMI cut below BMI cut MUAC cut below MUAC below MUAC relative to BMI cut cut-off relative to value of MUAC cut-off off relative to BMI BMI cut-off off off off cut-off cut-off off BMI cut-off relative to BMI cut-off cut-off (kg/m2) (n) (%) (cm) (n) (%) (%) (%) (%) (%)

16 92 94 22.5 96 98 100 33 96 100

15 86 88 21.0 82 84 91 67 95 50

14 71 72 20.0 68 69 87 78 91 70

13 48 49 18.5 45 46 83 90 89 85

12 34 35 17.0 26 27 71 97 92 86

11 14 14 15.5 8 8 50 99 88 92

10 2 2 14.0 2 2 100 100 100 100

115 32 O M ale

O l'em ale 30

Maie - répression M UAC on H Ml slope (1.86. inlercepi 7..16

l einale - repression MUAC on BMI slope 1.16, inlercepi 3.28

■°" ■ 'Jam es cl al - maie 26 ■ "James cl al - l'emale

oo m

O □ O D O O 6 o □ 18 i

& 6

12 I

10 12 14 16 18 24 Hoily mass iiulc.x (kp m ')

Figure 21

Scattergram illustrating the relationship between MUAC and BMI measurements in the S. Sudan study population, taken at the height of a severe famine (n = 98). Included for comparison are data from more normally nourished populations in developing countries. (|amcs et al. 1994) There was a close correlation between MUAC and BMI in the Sudanese population. The slope of the regression lines of MUAC on BMI in these severely malnourished subjects were greater than those recorded in the better-nourished populations. At the most severe levels of emaciation, this difference is of practical significance, equivalent to a difference of 1 - 2.5 cm in MUAC.

116 100

g- 60

> 50

0 10 20 30 40 50 60 1-specificity (%)

Figure 22

Receiver operating characteristic curve (ROC) for the prediction of BMI by MUAC in the S Sudanese data population (n=98). The point "X”, beyond which any increase in sensitivity is accompanied by a marked decrease in specificity, corresponds to a BMI cut-off of 13 kg m'^.

a. MU AC and BMI and oedematous adult malnutrition

1 he incidence of famine oedema in the Angolan subjects was high; 90% of subjects present with some pitting oedema and over 50% with moderate or severe pitting oedema of grade 3 or more, lable 12 & Fable 13 present one-way analyses of variance, comparing mean BMI and MUAC values in 1056 of these subjects, grouped according to their grade of oedema. Fable 12 demonstrates that increased severity of famine oedema (from grade 0 up to grade 5), were associated with a progressive elevation in BMI (F statistic = 20.25, p < 0.001). By contrast. Fable 13 demonstrates that MUACi did not vary according to the grade of famine oedema (see, F statistic = 1.68, p = 0.14).

117 Table 12

Analysis of variance for the relationship between grade of famine oedema and BMI in 1059 patients attending adult feeding centres in Melanje Angola.

oedema grade n Total Mean BMI Variance Std. Dev. beattie ka m kp m kp m ^ kp m 0 101 1609 15.9 5.9 2.44 1 102 1648 16.2 5.8 2.42 2 160 2676 16.7 5.3 2.30 3 489 8551 17.5 6.1 2.47 4 159 2856 18.0 6.0 2.45 5 49 924 18.9 5.3 2.30 oedema grade Minimum 25%ile Median 75%ile Maximum Mode beattie kp m kp m kp m kp m 0 11.6 14.2 15.7 17.2 25.4 16.1 1 11.3 14.6 15.9 17.6 25.6 15.4 2 10.8 15.2 16.6 18.1 22.6 17.4 3 9.4 15.9 17.5 19.2 25.1 17.5 4 12.4 16.1 17.9 19.3 24.5 18.0 5 13.3 17.6 19.0 20.6 24.2 19.2

analysis of variance (ANOVA for normally distributed data) Variation S S df MS F statistic p-value Between 577 5 115.3 19.6 <0.0001 Within 6203 1054 5.885 Total 6780 1059

Table 13

Analysis of variance for the relationship between grade of famine oedema and MUAC in 1059 patients attending adult feeding centres In Melanje Angola. oedema grade n Total Mean MUAC Variance Std. Dev. beattie cm cm cm cm 0 101 1993 19.7 5.9 2.42 1 102 1992 19.5 5.2 2.27 2 160 3175 19.8 5.0 2.24 3 489 9797 20.0 5.3 2.30 4 159 3113 19.6 4.9 2.22 5 49 977 19.9 6.3 2.50 oedema grade Minimum 25%ile Median 75%ile Maximum Mode beattie cm cm cm cm 0 15 17.8 19.6 21.1 27.0 19.8 1 15.2 18 19.4 20.8 28.8 19.2 2 15.6 17.8 19.6 21.2 27.2 21.0 3 14.2 18.4 20.0 21.6 33.6 18.0 4 13.8 18 19.6 21.0 27.2 17.0 5 10.8 18.5 19.9 21.8 24.4 19.6

analysis of variance (ANOVA for normally distributed data) Variation SS df MS F statistic p-value Between 41 5 8.1 1.54 0.17 Within 5577 1054 5.3 Total 5618 1059

118 Simple linear regression of MUAC on BMI, using the Angolan data, stratified according to the

degree of oedema, indicated that the P coefficient (slope) and the correlation co-efficient (r),

decreased as the grade of oedema increased (see Figure 23 and Table 14). Both the r and p coefficients were smaller in the Angolan data, even when the non-oedematous group Angolan

subjects were compared to the South Sudanese subjects (comparison of P; partial F statistic 9.0, d. f. 2, 182, p< 0.001; comparison of r: z = 4.22, p < 0.001,^). The linear regression lines did however converge around a BMI of 11.5 — 12.5 kg m'^ (see Figure 23).

17 Method for the comparison of correlation coefficients described in Clarke and Cooke, p 333 - ^^*2)

119 o 24 o * *

c? +

§ 0 < 1 18 ^

16 X South Sudanese

X ° Angolan - no oedema

X I ♦ Angolan - mild oedema

^ Angola - moderate & severe oedema

12 * ‘ ‘ I I 8 10 12 14 16 18 20 22 24 26

BMI (kg m

Figure 23

Simple linear regression models showing the relationship between MUAC and BMI in the South Sudanese non- oedematous patients to those seen in the Angolan oedematous patients stratified by degree of oedema present. Sudanese subjects: n =96, r = 0.91, /J -coefficient = 1.12, intercept = 3.9; Non oedematous Angolan subjects: n = 85, r = 0.71, /^-coefficient = 0.74, intercept = 7.9; Mildly oedematous Angolan subjects: n = 238, r = 0.69 coefficient = 0.64, intercept = 9.1; Moderately and severely oedematous Angolan subjects: n = 727, r = 0.61, f3 ■ coefficient = 0.54, intercept = 10.5.

The relationship between the MUAC and BMI in the Sudanese and Angolan data was examined further using multiple linear regression techniques that included the variables BMI, oedema, sex ascites and ethnic origin in models. The regression model is presented in lable 14. It demonstrates that the presence of pitting oedema was the most significant of the factors influencing the relationship between BMI and MUA(i (partial f statistic = 46.21, p < O.OOl). 1 he

120 presence of ascites and differences in the gender composition of the two samples also exerted an influence over the relationship (ascites partial f statistic = 3.90, p < 0.005; sex partial f statistic 11.2, p < 0.001). Ethnic differences between the South Sudan (Nilotic) group and the Angolan (Bantu) group were not significant (partial f statistic 0.62, p > 0.05).

Source df Sum of Squares Mean Square F-statistic

Regression 5 2471.77 494.35 157.21

Residuals 1150 3616.21 3.14 Total 1155 6087.98

Lower 95% Upper 95% Variable Mean B -coefficient S.E. Partial F-test C.l. C.l.

BMI 16.87 0.57 0.53 0.62 0.022 663.98

oedema 2.39 -0.31 -0.40 -0.22 0.045 46.21

sex 1.35 -0.38 -0.60 -0.15 0.113 11.12

ascites 0.08 -0.38 -0.76 0.00 0.192 3.90 data set^ 1.92 -0.18 -0.63 0.27 0.229 0.62

Y-lntercept 11.69

^ 1 = Sudanese, 2 = Angolan

Table 14 Multi-linear regression model for the relationship between MUAC and BMI, with MUAC as the dependant variable amongst 1155 malnourished Sudanese and Angolan subjects.

Hi. The relationship between BMI, MUAC and nutritional status

The subjects admitted into the S. Sudan and Angolan feeding centres were clinically all severely wasted. The very low mean MUAC and BMI values of these groups accord with this clinical observation. The Somali subjects, on the other hand, were clinically better nourished and in both cases the population means of BMI and MUAC were within the normal range (see Section 9.02 ). This difference in nutritional state between the Sudanese/Angolan subjects compared to the Somali subjects was used to examine whether the relationship between MUAC and BMI differed according to nutritional state. A multiple linear regression model, including variables for clinical nutritional status (defined by population group), oedema, sex and ethnic origin indicated that the relationship between MUAC and BMI might be different between normally nourished and malnourished populations. In the severely malnourished, the slope (P - coefficient) of the regression line MUAC on BMI was significandy greater than in the better nourished population

121 (partial F statistic = 92, d. f. 5, 1939, p < 0.001, see Table 15). These results concord with MUAC and BMI data from other normally nourished populations in developing countries (Figure 24).

Table 15

Multi-linear regression modei demonstrating the effect of malnutrition (defined clinically), on the relationship between MUAC and BMI in 1939 subjects from Somalia, Sudan and Angola. All P values are < 0.001.

Source d.f. Sum of Squares Mean Square F-statistic

Regression 5 14749.19 2949.84 911.48

Residuals 1934 6259.04 3.24 Total 1939 21008.23

Lower Upper Variable Mean 8 -coefficient Std. Error Partial F-test 95% C.l. 95% C.l.

BMI 18.14 0.62 0.59 0.65 0.017 1332.10

clinical nutritional status * 0.60 -2.72 -3.28 -2.17 0.284 92.03 oedema 1.43 -0.36 -0.44 -0.28 0.041 75.79

sex 1.57 -0.56 -0.75 -0.36 0.099 31.43 data set^ 3.24 -0.22 -0.39 -0.06 0.084 7.01

Y-lntercept 14.08

^ Data sets coded as: 1 — Sudanese, 2 & 3 — Angolan, 4 & 5 — Somali * Data sets coded as: 1 = Sudanese & Angolan, 2= Somali

122 33 - South Sudanese - no oedem a (very severe malnutrition)

Angolan - no oedem a (severe malnutrition)

^ Somali women (reasonably nourished)

+ Concern's Somali workers (reasonably nourished)

27 'James et al. Male (reasonably nourished)

James et al. Women (reasonably nourished) 25 -

21 -

13 11 13 15 17 19 21 23 25 27 29 31 BMI (kg m "

Figure 24 Linear regression lines for MUAC on BMI data in severely malnourished and adequately malnourished populations. Angolan data from the Concern Worldwide and Caritas adult feeding centres in Melanje Angola. Included for comparison are data presented by James et al. from normally nourished populations in developing countries. (James et al. 1994)

123 1 he MLh\(> data were noii-normally distributed with a skew to the right (see Jdgure 30 in the appendix). Only six of the 1, 940 cases fell below 14.5 cm and examination of the regression on MUAC on BMI for all data, revealed a threshold around 14.5 — 15.5 cm, below which values appeared not to fall. 1 his threshold meant that when a simple linear regression model for MUAC on BMI was fitted the residuals were not randomly distributed. A curvilinear, 4''’ degree

polynomial model (MUAC = a + Pi BMI + P 2 BM1 ^ + PjBMI^), fitted the data more closely, yielding residuals that were randomly distributed (sec f igure 25). A ver)' similar model provided the best fit in the non-oedematous data (see 1 figure 26).

1 o < Z) 20

10 1111 9 11 13 15 17 19 21 23 25 27 29 BMI (kg m^)

Figure 25

Polynomial regression model MUAC =39.7 + (-5.9 * BMI) + (0.50 * BMI)^ +-0.016* BMI)^ + (0.0.0002 * BM // describing the relationship between MUAC and BMI In all subjects (r = 0.75, Std error for the predicted y value 2.20 n = 1940).

124 Section 6.05 Discussion

6.05(a) SH/S ratio and BMI

Many of the patients described above presented with extremely severe emaciation. Thirty five percent (34/98) of the Sudanese sample had a BMI below 12 kg the level previously considered to be the lowest limit compatible with recovery. (James, Ferro-Luzzi, & Waterlow 1988) The lowest BMI recorded in the centre was that of a woman who was still able to walk to the weigliing scales with a BMI of 9.3 kg m ^. Other patients were too weak to stand and were not weighed or measured. If the BMI could have been estimated in these patients, even lower values are likely to have been recorded.

The Nuer, a Nilotic race, have a phenotype with long legs relative to stature and consequently have low values for the SH/S ratio ((see 4.03(c) i). In an extensive review of anthropometric data from around the world, the Dinka and the Samburu, both Nilotic peoples from sub-Saharan Africa, were the only populations to have mean BMIs below 18.5 kg m ^. (Eveleth & Tanner 1976) This phenotype is ahnost certainly one of the reason why the BMIs found in the study population were so low.

SH/S ratio was not recorded in any of the centres during this study. The pressurised circumstances that surround emergency famine-relief operations mean that such measurements are unlikely to be routinely feasible. In the absence of these data, individual BMI values cannot be standardised for individual SH/S ratio. This detracts from BMI’s usefulness in the routine assessment and screening of individual admissions to feeding centres.

Using an estimated mean population SH/S ratio of 0.48 and the correction equation proposed by Norgan,g the mean BMI in the Sudanese sample increased by 2 kg m ^ to a value of 15.2 kg m^. The minimum corrected BMI values at the centre were 11.6 kg m^ and 12.5 kg m^ for female and male patients respectively (see Table 16). Interestingly the corrected rninimum BMI values are similar to the values of 13 kg m ^ for male and 11 kg m^, suggested by Henry as the limits of BMI in starvation. (Henry 1990; Henry 1993)

18 This correction is based upon on the linear regression model of: BMI = 57.2 * SH/S - 7. 4 and the equation BMUtd = BMIo. 52 + (BMlob - BMIes) where BMUtd = standardised BMI, BMI0.52 = estimated BMI at SH/S of 0.52, BMlob = actual BMI, and BMIes = estimated BMI at actual SH/S (see Section 9 .0 4 )

125 Table 16

BMI in the Sudanese centre, corrected mean SH/S ratio according to the method of Norgan (see footnote 18) using a SH/H value of 0.48 for the Nuer phenotype and 0.52 for the “normal”phenotype. (Norgan 1994c; Norgan 1994b)

______BMI (kg m'^) ______Sex n Minimum 25%ile Mean Median 75%ile Maximum

male 41 12.48 14.08 15.59 15.68 16.68 19.28

female 57 11.58 13.58 14.93 14.78 16.18 19.98

6.05(b) The effect of oedema

Clinically the Angolan population was malnourished. The severit}" of malnutrition was however, due to the increases in weight caused by the presence of famine oedema BMI did not reflect this (see Table 12). MUAC values were less affected by famine oedema (see Table 13) although in extreme grades of famine oedema the upper arm can become swollen. This was rare and overall MUAC appeared to be a more stable reflection of decreased body mass in these subjects. Consequently, the gradient (P co-efficient) of the regression of MUAC on BMI was less with increasing severity of oedema (see Figure 23).

The gradient of the MUAC on BMI regression hne in Angolan patients with no signs of famine oedema was less tiian a similar regression using the Sudanese data. The most liltely explanation for this is that the diagnosis of famine oedema, especially in it’s early, “pre-pitting” stage(Waterlow 1993) can be problematic and even amongst trained doctors estimations of the amount of oedema fluid are very variable. (Golden 1999) It is therefore possible that many mild cases of oedema were wrongly classified as oedema free. Ascites also poses diagnostic difficulties and several litres of free fluid in the peritoneum can be missed, especially by workers inexperienced in clinical examination and the diagnosis of ascites. These issues are discussed in 3.03(d) . The convergence of the Angolan regression Hnes around a BMI of 13-14 kg m^ (MUAC of 17 — 18 cm) is compatible with this explanation. At extremes of emaciation, the minimum amount of lean tissue necessary to sustain life is present and the weight of excess extra cellular oedema fluid must approach zero. Consequently, the confounding effect of oedema upon the relationship between MUAC and BMI would be expected to decrease with increasing emaciation and the regression lines would converge as in Figure 23. Although the point of convergence with the Sudanese

126 regression line appears to occur at a lower BMI, correcting the Sudanese data for SH/S ratio (see 6.05(a) ) would make the convergence points closer.

6.05(c) The effect of acute malnutrition

The presence of malnutrition appeared to exert a significant influence on the relationship between MUAC and BMI (see Table 15). At severe grades of emaciation, actual MUAC was 2 - 2.5 cm lower than the theoretical value predicted by the regression of MUAC on BMI in several better nourished populations (see Figure 21 & Figure 24). There are several possible reasons for this finding. It is possible that the difference is due to ethnic differences in body shape. The common regression lines published by James et al. presented in these figures, were based upon data from

many different populations , 9 and contain substantial racial variation. In some of the contributory data sets (Ethiopian, Somali, and Chinese female), the relationship between BMI and MUAC is similar to that seen in the Sudanese data (see Figure 31). The differences demonstrated between the severely and less severely malnourished groups might also be an artefact resulting from deficiencies in the study design. Although the effect of ethnicity did not appear to be significant in the multivariate model presented in lable 15, the ‘malnutrition’ and ‘ethnic’ variables were not independent. Thus, ethic variation in the relationship between MUAC and BMI cannot be ruled out adequately by this study. It might also be that workers responsible for measuring MUAC in the different studies systematically recorded MUAC in shghtly different ways. There are however, several physiological reasons why the relationship between BMI and MUAC should vary with the presence of acute malnutrition. During acute energy deficiency, tissue with a low metabolic rate, such as sub-subcutaneous fat and skeletal muscle, is preferentially catabolised. (Heymsfield et al. 1984; Waterlow 1986) The high rates of infection in acute malnutrition during famine, (Ibmkins & Watson 1989) would also increase catabolism of skeletal (peripheral) muscle. (Broom 1993; Heyms field et al. 1984) The reduction in physical work using the arms that occurs during famine would further decrease arm muscle. The resultant accelerated loss of upper arm tissue in comparison to other tissues would reduce MUAC faster than the BMI in the acutely malnourished when compared to better nourished populations. The slope of the MUAC on BMI regression line would therefore be increased and its position shifted to the right, as malnutrition increased. This is seen in the South Sudanese data (see Figure 21).

19 Somalia, Papua New Guinea, India, China, Senegal, Zimbabwe, Mali and Ethiopia

127 An important deficiency in these studies was the failure to record individual SH/S values. However, it is unlikely that the differences in the relationship between BMI and MUAC in severely malnourished and more normally nourished populations demonstrated here can be explained by differences in SH/S ratio. The long legged phenotype of the Sudanese population would be associated with a low SH/S ratio. Therefore observed BMI values would tend to be lower than if they had been standardised to a SH/S ratio of 0,52. The use of un-standardised BMI values would therefore mean that the MUAC on BMI regression line would lie more to the left, the reverse to tliat seen in Figure 21. Standardisation of the BMI values to a SH/S ratio of 0.52 would tend to shift the MUAC on BMI regression line to the right, increasing the differences between the severely malnourished and more normally nourished populations.

In the absence of other data from SAM in famine, it is parsimonious to conclude that the effect seen in Figure 21 & Figure 24 represent a physiological change rather than, or in addition, to any etlinic differences. This has important implications for the development of MUAC cut-offs for use in famine. If the relationship between MUAC and BMI does change in acute malnutrition, MUAC cut-offs appropriate to famine must be based upon data from acutely malnourished populations. Since 1992, when the attention of the scientific community was re drawn towards severe adults malnutrition in famines, (Collins 1993a; Collins 1995) attempts have been made to extrapolate from non-acutely malnourished populations to create indicators for use in famine-relief programmes. In 1996, Ferro-Luzzi and James adjusted their theoretical estimation of the lowest BMI compatible with life down from 12 kg m^ to 10 kg m^ [sc 192] in order to take into account the extremely low BMIs being observed in Somalia during the famine of 1992. (Ferro-Luzzi & James 1996) They proposed two new BMI and three adult MUAC cut-offs for use during famine (see Table 17). (Ferro-Luzzi & James 1996) These MUAC cut-offs were based upon MUAC Z- scores, derived from MUAC measurements on 5669 adults living in developing countries (see Table 10). Implicit in their use of Z-scores is the assumption that the distribution of MUAC values is the same in normally nourished and acutely malnourished populations. This assumption is not correct if acute malnutrition is accompanied by a physiological change altering the relationship between MUAC and tissue mass.

An increased slope of the MUAC on BMI regression line in acute malnutrition would mean that MUAC cut-offs derived from normal populations using Z-scores, would over-estimating levels of malnutrition in famine populations. The result would be a reduced specificity (and increased sensitivity) in their identification of “true” malnutrition giving a less discriminatory ROC curv'^e. Comparing the sensitivity, specificity and predictive power of such theoretical MUAC cut-offs in

128 identifying BMI thresholds, with MUAC cut-offs derived from acutely malnourished subjects, gives a magnitude of this error. This comparison again assumes BMI to be a criterion standard for adult AED (see 6.04(c) ) and require a famine population with no oedema. Table 18 shows the result of this comparison. See also Figure 32 in the appendix that presents the ROC of the different MUAC indicators using the data from South Sudan.

This analysis suggests that MUAC cut-offs, derived by extrapolation from normally nourished adult populations, introduce substantial error into the assessment of SAM. A cut-off based upon die -2 Z-score had a specificity of 54% in die identification of a BMI threshold of 13 kg m ^, compared to a specificity of 90% for the 18.5 cm cut-off derived from acutely malnourished adult populations. If used for screening this fall in specificity would be translated into a very substantial increase in the workload of a TFC. In the absence of a sound theoretical or empirical basis, such an increase in caseload would appear to be unwarranted.

129 Table 17

MUAC values proposed by Ferro-Luzzi and James for the assessm ent of adult malnutrition during fam/nes(Ferro-Luzzi & James 1996)

MUAC Z-score MUAC values (mm) diagnostic catagory corresponding BMI (kg m'^) grade of malnutrition men______women men wom en

< -1 < 230 < 220 undernourished < 17 < 17 3

< -2 <200 <190 severe wasting < 13 < 13 4

< -3 <170 <160 extreme wasting < 10 < 10 5

Table 18

The prediction of BMI by MUAC cut-off derived from non-famine populations and severely malnourished populations.

source of indicator MUAC cut-off BMI cut-off numbers below numbers below sensitivity of specificity of PPV of MUAC cut-off (cm) (kg m-2) MUAC cut-off BMI cut-off MUAC cut-off MUAC cut-off cut-off relative to BMI relative to BMI relative to BMI (%) (%) (%)

normal population 20.0 13.0 68 48 94 54 66 James et al. 17.0 10.0 26 2 100 75 8 severely 18.5 13.0 45 48 83 90 89 malnourished population 14.5 10.0 2 2 100 100 100

130 6.05(d) BMI and MUAC in extreme malnutrition

During starvation, the total amount of collagen and skeletal mass in the body remains reasonably constant. (Heymsfield et al. 1984; Picou, Halliday, & Garrow 1966) Therefore, in the pre-terminal stages of extreme emaciation as peripheral stores of fat and muscle non coUagenous protein become exhausted, elements such as skin, bone and the neuro-vascular bundle, form an increasingly large proportion of the arm volume. This exhaustion of peripheral stores is accompanied by a relative increase in the catabolism of central, visceral protein stores and a decrease in catabolism of peripheral stores (see 4.03(e) iii). These changes would reduce the rate of decrease in MUAC compared to BMI, causing the regression line of MUAC on BMI to level off. The drop in the sensitivity of MUAC when compared to BMI, that occurred below an MUAC of 18.5 cm (see Table 11), might be a reflection of such a change at the lower end of the BMI and MUAC ranges. The near absence adult MUACs lower than 14 to 15 cm is also evidence in support this change in metabolism. The requirement to be able to stand as a criterion for entry into the study would have resulted in a selection bias partially obscuring this effect. Only those with sufficient peripheral muscle mass remaining to enable them to stand would have been selected. Although fewer than 10 patients were excluded from the Sudanese data in this way, many of them would have been those who had exhausted their peripheral stores. The number of exclusion in the Angolan data is not known. The drop in the sensitivit}^ of MUAC at extreme grades of emaciation would therefore most likely have been greater had the whole population of the centres been measured. Tlie paucity of data at extremely low BMI and MUAC values means that empirical support of this phenomena is sketchy. At present the relative importance of these effects is unknown and more work is required relating MUAC cut-offs to risk of short term functional consequences before MUAC cut-offs can be more firmly stipulated

6.05(e) The relationship between MUAC and BMI in SAM

The opposing effects of increased peripheral catabolism during the evolution of severe acute adult malnutrition and decreased peripheral catabolism in the terminal stages means that a straight linear model of tlie relationship between MUAC and BMI is not likely appropriate. When fitted such a model yields residuals that are not normally distributed. An “S” shaped polynomial regression model describing the relationship between BMI and MUAC is theoretically more appropriate. Such a model, where the slope of the regression line of MUAC on BMI is increased over the BMI range 19-14 and levels out below 13 kg m^ and above 19 kg m^ is shown in Figure 25. The

131 correlation co efficient for this model is not significantly greater than a linear model the residuals are randomly distributed. y\ very similar model, but with a greater increase in slope between a BMI of 19-13 kg m^, also provide the best fit if the oedema tous patients are excluded from the analysis (see f igure 26).

31 I I- + + +■»«■ + 29 I . + ++ +.* #++4 27 E 25 O < Z) 23 21

17 + + + + 15 Ù * normal obese

13 I I Î 9 11 13 15 17 19 21 23 25 27 29

BMI (kg m-2)

Figure 26

Polynomial regression model MUAC =38.2 + (-6.4 * BMI) + (0.63 * BMI)^ + -0 .0 2 * BMI)^ + (0.0.0003 * 6 M // describing the relationship between MUAC and BMI in all non-oedematous subjects (r = 0.82, Std error for the predicted y value = 1.87 n = 971). fhe levelling out of the regression line at the extremes of emaciation indicates a point where peripheral stores are exhausted and central catabolism of protein provides the main energy source. J'his level of MLb\C may therefore represents a functional threshold below which there is an increase in the catabolism of visceral protein from vital organs such as the heart and other vital structures. This point jrrobably corresponds to the pre-terminal increased in the P-ratio. (I Icnry,

132 Rivers, & Payne 1988) The increase in visceral catabolism vvül decrease the function of vital organs and increase the necessity for special therapeutic diets. Such severely emaciated patients are unlikely to sur\tive given only a general-ration alone. The point where the regression line flattens is therefore likely to represent the threshold at wliich therapeutic feeding is necessary (see Chapter 7

)•

6.05(f) The use of MUAC in adults

In children, the use of MUAC is associated with two problems; the preferential selection of younger children as malnourished(Bern & Nathanail 1995; Smedman et al. 1987; Sommer & Loewenstein 1975) and a lack of reproducibility in MUAC measurements, (McLaren 1986) The preferential selection of younger children as malnourished results from the gradual increase in the normal MUAC between the ages of 1 and 5. (Ross et al. 1990) In older children and adolescents, the rapidly changing patterns of skeletal muscle and subcutaneous fat are likely to exaggerate this problem and it is likely that age specific MUAC cut-offs will be required. In adults the use of MUAC, is not subject to errors caused by growth, but can be affected by the redistribution of subcutaneous fat towards central areas of the body with ageing. (Euronut SENECA investigators 1991) Problems with the reproducibility of MUAC measurements are potentially more serious obstacles to the use of MUAC in adults. As in children, inter- and intra- obsen'^er errors in MUAC measurements may occur. These did not appear to be great problems with these data (see 3.03(d) ) and it maybe that the larger dimension of the adult arm reduces the relative importance of these errors. The development of coloured, non-numeric, MUAC bands reflecting threshold values of MUAC with a change of colour would further reduce these problems by remove numerical errors. Given the ease with which MUAC measurements can be performed it would be feasible to refer any patients found to have an MUAC within a few mm either side of the threshold to a more experienced worker for verification. This area for referral could be designated on the band as a different coloured zone. The width of this zone would need to be based upon a more detailed examination of errors in the evaluation of adult MUAC by minimally trained workers. These problems are being investigated using data from several Concern Worldwide adult therapeutic centres.

Assessment of adult nutritional status using MUAC requires no equipment apart from a tape measure. As the index is the actual measurement itself, mathematical manipulation of the measurements obtained is not necessar)\ By contrast, assessment of BMI requires a height board and weighing scales that often require frequent re-calibration in the hot dusty environments in

133 which famines usually occur. In addition, the two measurements obtained must be mathematically transformed before being useful. All these process are prone to error, especially during the liighly pressurised circumstances characteristic of the early phases of famine-relief programs. The ease with which MUAC can be assessed therefore suit it for nutritional screening during the height of an emergency where time and skilled personnel are at a premium.

Section 6.06 Conclusion

There is a great need for a practical indicator with which to assess adult malnutrition, especially during the initial phases of emergency relief operations during famine. This study indicates that MUAC is well suited to such a role and its use has several important advantages over the use of BMI. However, the inability to follow up patients in all study sites and the resulting absence of outcome data prevented any analysis of the relationship between MUAC and mortality. This is a major limitation in this study and in the absence of these data, functional MUAC cut-offs based on risk of mortality cannot be presented with any degree of certainty.

At present, there is insufficient data available to stipulate with any degree of certainty functional MUAC cut-off differentiating malnourished adults that require specialised therapeutic care from those who can survive without it. This study does demonstrate that it is inappropriate to derive such thresholds by extrapolation from data recorded in normally nourished populations. There is a need for more field studies during emergency famine-relief operations, to evaluate the power of adult MUAC to predict mortalit}^ in different famine affected populations.

In severe adult malnutrition oedema is common, associated with a poor prognosis and confounds anthropometric indicators dependant on weight. At present, there are no clear guidelines as to how to assess oedema. By analysis oedema using readily definable anatomical landmarks this smdy goes some way to providing a useful classification of the condition. However there are still problems over the way that oedema is assessed in practice, particularly how much pressure should be applied and for how long. There is a need for further research to asses this.

At the moment, there is an urgent need for indicators of SAM for use in famine relief programmes. The recent increased interest in severely malnourished adults has resulted in many centres being opened in recent famines. Many of these are operating without guidelines, or with guidelines based upon the BMI indicators for chronic adult malnutrition or MUAC cut-offs derived from normally malnourished populations. Given this situation, the author feels that there

134 is a need for tentative recommendations for screening indicators, based upon what little data there is from severely malnourished adults during famine and taking due regard of theoretical considerations. Although far from perfect these are likely to be better than either an absence of guidelines or those derived without reference to data from famine. The “best guess” indicators from this study are presented in Chapter 8 .

135 136 Chapter 7 Study four: The dietary treatment of severe adult malnutrition

Section 7.01 Aim

To examine the effect of dietary protein intake on the rehabilitation of severely malnourished adults

Section 7.02 Introduction

The difficulties of feeding severely malnourished adults, especially those with oedematous malnutrition, have long been recognised. (Digby 1878) During the first half of this century, these problems received considerable attention from the scientific community and, generally, diets containing relatively high levels of protein were recommended. (Beattie, Herbert, & Bell 1948; Burger, Sandstead, & Drummond 1945; Leyton 1946; Lipscomb 1945; McCance & Widdowson 1951; Mollinson 1946) In the early 1950s, scientific attention shifted towards malnourished children in whom the quantity and quality of dietary protein required for successful rehabilitation became the focus of much research. (Waterlow 1961) In particular, the importance of liver pathology in kwashiorkor was recognised. (Waterlow 1948) Recently, it was demonstrated in children with kwashiorkor, that mortality was rninimal when diets containing only maintenance levels of protein ( < 1 g kg'May'^) were given. (Golden 1993; Golden 1996) Up to the time of the Somali famine in 1992, these diets had only been tested in specialised hospital units and no field research had been performed during an emergency relief program. The difficult circumstances that exist during famines generally prevent the execution of classical, scientificaUy rigorous research. This has resulted in an absence of field research since the late 1940s when the concentration camps in Europe and the Far East were liberated. In its absence, results from studies performed in less severely malnourished subjects, usually in hospital settings, have been extrapolated to the very different circumstances found during war and famine. Thus, it was shown that adult subjects with BMI of 17-18 kg m‘^ respond well to diets with P:E ratios greater than 19%. (Barac-Nieto et al. 1979) As a result, such high protein diets have continued to be recommended for the management of severe adult malnutrition. They were used in all Concern Worldwide TFCs during the Somali relief operation

137 in 1992-1993. However, the adults admitted to these TFCs generally had BMI of 10-13 kg m^, far less than those that had been studied, and often, they had oedema (see Chapter 4 ). It was observed that many of these patients, particularly those with oedema, were refusing the high protein diets. Thus, the hypothesis was advanced that the high protein diets were deleterious during the initial phase of rehabilitation. The aim of this study was to compare the immediate and short-term effects of a lower protein diet with those of the conventional higher protein diet during nutritional rehabilitation of severely malnourished adults admitted to a TFC during the emergency relief program in Baidoa.

138 Section 7.03 Subjects and Methods

7.03(a) Subjects

Data were collected on 573 patients with a reported age of 15 years and over, admitted to the Concern Worldwide Adult Therapeutic Centre, Baidoa, between 25* October 1992 and the end of March 1993. The reported ages of those admitted were between 15 and 80 years (mean 34). It is unlikely that these reported ages are accurate as people were generally unsure of their exact age and no local calendar was available to verify the ages given. Analysis of the distribution of the age data reveals clustering around ages divisible by five, suggesting that reported ages were often given to the nearest five years. 46% of those admitted were male.

The criteria for admission varied depending upon the available space in the centre, and the existence of other medical facilities in Baidoa. In general, only patients with a BMI below 13.5 kg m " or patients assessed as very severely malnourished by the clinician (SC. ) and too ill to be weighed and measured on admission, or patients with oedematous malnutrition were admitted. Patients with primarily medical problem were referred to the MSF Holland inpatient unit and the less severely malnourished people to supplementary kitchens (see Chapter 3 for a fuller discussion of subject selection).

7.03(b) Methods

Details of the centre are presented in 3.01(b) i, and details of the supervision, rehydration, medical treatment and discharge criteria in 4.04(b) .

7.03(c) The rehabilitation diets

During their stay in the centre, patients received 6 - 8 meals each day. For the first month of operation, the standard higher protein diet (HP diet), used in aU the other TFCs in Baidoa, was the only diet available. This consisted of ‘recovery milk’ [King’s Food (Ermelo, Holland), a blend of dried skim milk (DSM), vegetable oil, vitamins and minerals], UNIMIX (a blend of Soya flour, oü and sugar), rice, beans and BP5 biscuits (Compact, Bergen, Norway) (Table 19). The recovery mük and BP5 biscuits were premixed fortified foods, designed especially for famine-relief and

139 marketed in Europe: the UNIMIX was a premixed fortified food made up for UNICEF in various factories in Africa. Patients on this diet were offered, on average, 158 g protein and 16.2 MJ per day (P:E ratio 16.4%), including 70 g oil and 137 g of lactose. The cost of this diet was approximately US $1.90 person^ day \ During the first few days after admission, the milk component was diluted to half strength with the WHO formula ORS containing sodium chloride 3.5g 1', trisodium citrate 2.9g 1% potassium chloride l.Sg 1^ and glucose 20g l \ This reduced the total daily average offered to 129 g protein and 14 MJ (P:E ratio 15.5%). From 5 December 1992, a lower protein diet (LP diet), mixed on site from basic commodities, was also available. The LP diet consisted of ‘high energy milk’, a blend of DSM, vegetable oil and sugar, together with bananas, white rice and sweet tea (Table 20). Patients on this diet were offered, on average, 82 g protein and 16.1 MJ per day (P:E ratio 8.5%) including 152 g of oil and 95 g of lactose. The cost of this diet was approximately $1.00 person^ day \ During the first few days after admission, the milk component of this diet was diluted to one-third strength with WHO formula ORS. This reduced the total daily average offered to 35 g protein and 10.9 MJ (P:E ratio 5.0%). Initially, the potassium and sodium contents of the two diets were similar. However, during January 1993, bags of the individual salts mineral became available. The author tlien used cooking scales to measure out appropriate weights to dilute in water to make a concentrated solution. For CuCl 2 , the crude scales made this an inaccurate process. From then on, and patients on the LP diet received approximately (per kg body wt d^ ) 1 mmol KCL, 1 mmol tripotassium citrate, 0.4-0 . 8 mmol MgSo^, 0.31 mmol zinc acetate and 0.003 mmol CuClg, raising the amount of potassium they received by 2 mmol kg body wt ’ d \ For those patients who required rehydration, usually only during the initial phase of treatment, WHO formula ORS was used. It was not feasible to quantify the amount of ORS and, therefore, the additional sodium and potassium consumed over these periods.

On 5 December 1992, the LP diet was first offered to 11 oedematous patients, who had not responded to the HP diet. The positive response in these patients to the change of diet was so dramatic that it was soon considered unethical to use the HP diet in the treatment of oedematous patients. Thus, from 7 December 1992, patients presenting with oedema and those whom the clinician (SC. ) considered very iU or moribund, were offered the LP diet during the initial phase of treatment. The HP diet continued to be used from admission for the less severe cases and for many of the patients during the recovery phase of rehabilitation. The onset of this recovery phase was defined clinically in oedematous patients, by a return of appetite and a substantial loss of oedema and / or ascites. As loss of oedema was accompanied by loss of weight, this was accepted

140 as a sign of a positive response to treatment. (Golden 1993) In marasmic patients, the recovery phase was heralded by a return of appetite and accompanied by steady weight gain for at least three days.

Table 19

The higher-protein diet (HP), containing 16.4% of the energy from protein

Ingredient Quantity Protein Energy Sodium P otassium

g/d g/d kJ/d mmol/d mmol/d

R ecovery milk 252 55 4276 42 71

UNIMIX 100 12 1680 0 14

R ice 200 12 2940 0 8

B ean s 130 33 1966 1 46

BPS 275 46 5290 2 18

Table 20

The lower-protein diet (LP), containing 8.5% of the energy from protein

Ingredient Quantity Protein Energy Sodium

g/d g/d kJ/d mmol/d

DSM 180 65 2646 44

Oil 150 0 5733 0

Sugar 160 0 2688 0

Rice 200 12 2940 0

Bananas 500 5 2100 0

Total - 82 16107 44

141 7.03(d) Analysis of mortality.

O f the 573 patients for which admission data were collected, 16 were admitted primarily on medical grounds, and there was insufficient dietary information for 70 others. These 8 6 patients were excluded from the analysis of mortality. For analysis, patients were assigned to either the HP group (n = 343) or the LP group (n = 144), according to the diet they received during the initial period of their stay in the centre. Thirteen patients were wrongly allocated to diets by junior staff upon admission. I rectified these mistakes within 3 days and allocated the patients to the correct diet. For the analysis, the diet group of these patients was that of the corrected diet. Twenty- seven patients (aU from the marasmic group) were lost to follow up and were excluded from further analysis of mortality associated with the two diets, leaving 377 patents with marasmus or müd oedema and 83 patients with moderate to severe oedematous malnutrition. These data are presented in Table 21. Two by two contingency tables and multiple logistic regression were used to assess the effect of the 2 diets and the mineral supplement on mortality. This allowed us to control for the potential confounding effects of morbidity (oedema, lower respiratory tract infection, dysentery, dehydration, and anaemia) and other variables (age, sex, and time since opening of the centre) upon mortality. The time variable was included in the analysis in order to control for the effect of any time bias in the data, as it was conceivable that patient care, independent of the introduction of the LP diet, improved over time. Variables not independently associated with mortality were excluded from models in a stepwise fashion using estimation techniques.

7.03(e) Analysis of the effect of HP and LP diets on rates of weight change and the loss of oedema during rehabilitation.

The rate of weight change (g kg'May'^) for each consecutive 3 day period was calculated for each patient. A matched analysis of the rate of weight change during the first fifteen days of treatment on each diet, (using a paired t-test), was performed on those oedematous patients who changed from HP to LP diets after they had been in the centre for more than 3 days and for whom there were sufficient data (n = 7).

142 Table 21

Outcome data associated with the use of the higher (HP) and lower (LP) protein diets

All patients Oedematous patients Marasmic patients Total Died Survived Lost to follow up Total Died Survived Total Died Survived HP diet 343 78 240 25 27 14 13 291 64 227

LPdiet 144 25 117 2 56 14 42 86 11 75

Total 487 103 357 27 83 28 55 377 75 302

143 Section 7.04 Results

7.04(a) General patient data.

The characteristics of the patients who survived and those who dies has been reported in Section 4.05

7.04(b) Effect of diet on mortality.

Mortality was lower in patients with oedematous malnutrition who received the LP diet during the

initial phase of treatment (crude OR 0.31, 95% C. I. 0.10, 0.90; Yate’s corrected 4.74, p < 0.05). This effect remained after adjusting for confounding variables using logistic regression (adjusted OR 0.31, 95% C. I. 0.12, 0.81, likelihood ratio statistic 5.74, p < 0.05). This is equivalent to a three fold reduction in mortality. No such difference was observed in marasmic patients

(crude OR 0.52, 95% C. I. 0.24, 1.08; Yates corrected X^ 2.97, p = 0.08: adjusted OR 0.87, 95% C. 1. 0.38, 1.97; likelihood ratio statistic 0.10, p = 0.74). Analysis of the effects of the mineral mix in patients receiving the LP diet did not demonstrate any differences in mortality before and after addition of the mineral supplement.

7.04(c) Loss Of oedema and gastro-lntestinal function.

The appetite of many of the oedematous patients appeared to be very poor on the HP diet. Upon changing to the LP diet, an increase in appetite was reported by many of these patients, including those who had suffered from persistent oedema lasting several weeks on the HP diet. In the 7 patients for whom matched rate of weight change data were available, the rate of loss of oedema, (reflected as weight loss), accelerated after transfer to the LP diet (rate of weight change, g kg'May' \ mean, S. D. ): HP +6.3, 12.1; LP -7.2, 18.5; difference = 13.5, paired t = 3.18, p < 0.05. Associated with the increased appetite were episodes of watery diarrhoea. This was occasionally very severe, resulting in the rapid appearance of intravascular hypovolaemia. It responded to dilution of the mdk element of the diet with WHO formula ORS during the first few days of treatment. The number of days and the degree to which the milk was diluted were tailored for each patient according to the severity of the diarrhoea, clinical signs of intravascular hypovolaemia

144 (heart rate, peripheral perfusion, jugular venous pressure, hydration of mucous membranes) and the response to previous dilutions. On occasion, it was necessary to dilute the milk down to one- ninth strength for several days in order to reduce diarrhoea and maintain the appropriate intravascular volume. Once the diarrhoea resolved, the mük concentration was gradually increased to full strength over a week.

7.04(d) Weight change during rehabiiitation.

The mean rates of weight change for oedematous and marasmic patients during the initial month of treatment on each diet are shown in Figure 27. During the first 9 days, oedematous patients on the LP diet tended to lose weight as they lost oedema. By contrast, oedematous patients on the HP diet tended to gain weight. After this period the situation reversed and those on the LP diet started to gain weight whüst those receiving the HP diet stayed at the same, or gradually lost, weight. Marasmic adults gained weight similarly on both diets during this initial period. During the recovery phase (days 16-60 after starting treatment), mean rates of weight change were similar in both marasmic and formerly oedematous patients (mean, S. D. : marasmic, 6.1, 5.3 g kg'May" formerly oedematous, 5.1, 4.2 g kg'May"^) and for the HP and LP diets (mean, S. D. : HP, 5.9, 5.6 g kg'May'LP, 5.6, 4.8 g kg'May'^)

145 la. Edematous lb. Marasmic 10

>1 •s

x: -2 X -2 O' o (U3 3 -4 -4 o o

-6 -6

-10 -10 0 3 6 9 12 15 18 21 24 21 30 33 36 0 3 6 9 12 15 18 21 24 27 30 33 36

Days from start of diet Days from start of diet

Figure 27

Rates of weight change in oedematous (n = 45) and marasmic (n=219) patients receiving the lower-protein (if) and higher-protein ( 0) diets during the first 33 days of treatment. Note error bars are +/- 1 S. E. of the mean.

146 Section 7.05 Discussion

7.05(a) The importance of lower protein diets during the initial treatment of severe malnutrition in adults.

In this study, oedematous patients receiving the HP diet suffered a three-fold higher mortality than those receiving the LP diet. Many of these patients also experienced prolonged anorexia and persistence of oedema. The LP diet was associated with lower mortality and accelerated loss of oedema. These differences appear likely to have been due to the different amounts of dietary protein offered during the initial phase of therapy, however problems with the study design and implementation means that such conclusions cannot be categorical.

Other potential explanations must also be examined. The study compared groups of patients differing primarily according to the diet they received. However, patients were not randomly allocated to the diets. The HP diet was started earlier than the lower protein diet and within a week after its introduction, the LP diet was preferential allocated to patients considered by the author to be more severe. It is possible that the differences in outcome that have been ascribed to the differences in diet were, instead, due to differences related to the timing of the diets or to the subjects assigned to each diet. The many factors that changed with time such as improvements in the training of the staff, changes in the general condition of admissions as the famine progressed, and changes in resources for the centre, might have contributed to the differences observed. Although during analysis logistic models, containing a time variable, attempted to allow for these biases, such statistical techniques are poor substitutes for rigorous study design. The non-random allocation of patients to the diets however is likely to have obscured the reported improvements

associated with the LP diet. During the first 6 weeks of the study, all patients received the HP diet: during the next 16 weeks, those patients assigned to the LP diet were oedematous or otherwise very Ü1 or moribund were preferentially allocated to the LP diet by the author. On clinical grounds, these patients were expected to have a higher mortality than those assigned to, or ahcady recovering satisfactorily on the HP diet. However, the study has shown that the patients on the LP diet had a three-fold lower mortality and a more rapid loss of oedema than those on the HP diet. Thus, it is possible that, had patient allocation been randomised, the difference in outcome would have been in the same direction but even greater. This is compelling evidence that the difference in outcome was due to the difference in diet.

147 The diets, however, differed in more than their protein content. Lower energy intake, particularly energy derived from carbohydrate, during the initial treatment with the LP diet, is a possible explanation of the differences in outcome. This might be expected to lower mortality by reducing the incidence of ‘refeeding syndrome’. (Weinsier & Krumdieck 1980) Although the aim was to offer diets of similar energy content (see Tables 1 Sc 2), the LP diet had a greater proportion of its energy derived from fat and was usually given in a more dilute form during the first few days of treatment. This dilution substantially reduced the energy being offered to patients during that time. Although the ‘recovery milk’ element of the HP diet was also diluted during the first few days of treatment, this dilution caused less of a decrease in the energy content of the HP diet as a whole. These differences are, however, unlikely to explain the higher mortality on the HP diet as, owing to persistent anorexia, most of the oedematous patients receiving the HP diet consumed only a small fraction of the food offered. Their food intakes, though not formally measured, appeared to be very low, making ‘refeeding syndrome’ unlikely. The marked increase in appetite and thus, presumed increase in food intake, observed in these patients when offered the LP diet, is likely to have more than compensated for the decreased energy content of the diluted diet. Indeed, it is likely that energy intake during the initial phase of treatment on the LP diet was higher than on the HP diet. This would accord with documented experiences in feeding severely malnourished children in Jamaica. (Golden 1982)

Immediately after the change to the LP diet, most of the oedematous patients who had been in the centre for some weeks developed watery diarrhoea. This diarrhoea appeared to be the ‘refeeding diarrhoea’ frequently described in new arrivals at feeding centres, and generally regarded as the response of a starving and atrophic intestine to the initial réintroduction of food. (Roediger 1986) The absence of ‘refeeding diarrhoea’ amongst oedematous patients receiving the HP diet and its development when these patients were transferred to the LP diet, supports the observation that the previous intake of the HP diet in these patients was very low.

Alternatively, other differences in the composition of the two diets could have contributed to the difference in mortality. The HP diet had more lactose that might have increased mortality. This is also unlikely as the lower incidence of diarrhoea on the HP diet indicates that lactose intolerance was not a major problem. Although the two diets contained similar amounts of sodium it is also possible that if the critically iU patients ate only the milk elements of the diets sodium intake would have been much on the HP diet. This again is unlikely, as the large quantities of ORS added whilst diluting the LP diet (see 7.03(c) ), would have increased the sodium intake in the patients receiving the LP diet.

148 In marasmic patients, the diet given during the initial phase of treatment did not affect mortality or initial rate of weight change. This suggests that the dietary differences interacted with metabolic differences between marasmus and oedematous malnutrition. One of the main metabolic differences is in liver function. In this study, few of the marasmic but many of the oedematous patients had clinical signs consistent with liver failure. These included petechial rash and jaundice as well as anorexia, oedema and ascites. Thus, it is possible that there was no difference in outcome associated with diet in marasmic patients because their Hver function was better preserved. If correct, this again suggests that it was the protein content of the diet that was of importance.

The study failed to demonstrate any effect of the mineral supplements in patients given the LP diet. This may have been due to the study design whose aim was not to test their effect. However, it is also possible that the mineral intake from the LP diet alone was adequate. This is likely to be true for potassium, whose dietary intake was usually at least 2 mmol kg'May"’, but unlikely to be so for the other minerals.

Although recovery from severe adult malnutrition has been systematically studied only rarely, there have been reports of similar poor responses to HP diets wherein oedema was slow to disappear, and sometimes even appeared during treatment (26-28). Ex-inmates of Belsen concentration camp, who experienced degrees of weight loss comparable to patients in Baidoa (mean weight loss 38?"o corresponding to a BMI of 12 to 14 kg m'^) suffered frequently from famine oedema and ascites. (Mollinson 1946) During rehabilitation on a diet containing 64.8 g protein and 3.4 MJ (P:E ratio 32.4%), famine oedema often appeared or increased. (Lipscomb 1945) In these patients, use of casein hydrolysates was associated with increased mortality. (Vaughan & Pitt Rivers 1945) Similar accounts come from the Dutch famine of 1944-5, where the recommended rehabilitation diet contained 300g of protein and 13.4 MJ (P;E ratio 37%). (Burger, Sandstead, & Drummond 1945)

7.05(b) Rates of weight change during the recovery phase of rehabilitation.

The extreme levels of disruption and insecurity present in Baidoa during the time of this study made the operation of the centre very difficult. The mean rates of weight gain, of 5 - 6 g kg'’ day'’ during the recovery phase of rehabilitation, probably represent the lower end of the spectrum of reasonable rates of weight gain in adults recovering from severe malnutrition. Severely malnourished ex-inmates of the Sanbostel concentration camp receiving 31.5 MJ and 297g protein day '(P:E ratio 15.8%) gained around 7 g kg ’d ’ during the recovery phase of their treatment.

149 (IVIurray 1947) In less severely malnourished patients, rates of weight gain appear to be lower. In the Minnesota experiment, 32 volunteers with mild to moderate malnutrition, receiving 10-14 MJ and 75 -lOOg protein day'^ (P;E ratios 10.6 - 14.4%) gained only 1.85 g kg'May'h (Keys 1950b)

7.05(c) Comparison with children.

The patterns of presentation and recovery in severely malnourished adults are similar to those seen in children. In children, hypoalbuminaemia is evidence of a poor prognosis. (Vis 1985) Although oedema, in the absence of hypoalbuminaemia, is not necessarily a poor prognostic indicator, (Waterlow 1993) in practice, the frequent coexistence of the two makes oedema a useful prognostic marker. In Baidoa, oedema in adults was associated with a much poorer prognosis (see 4.06(c) ). However, it is possible that this finding is not universal. In a Concern TFC in Melanje, Angola during 1993 and 1994, 90% of adults admitted suffered from oedema but this did not appear to be associated with such a poor prognosis (unpublished, Collins 1993). More work is needed on prognostic indicators in severe adult malnutrition.

The maximum rates of weight gain, typically 10 - 20 g kg^ day % recorded in children recovering from severe malnutrition, (Ashworth 1969; Waterlow 1993; Waterlow, Golden, & Patrick 1978) are considerably higher than the rates of weight gain reported here. However, the pattern of recovery is similar. Initially, on a low protein, maintenance energy intake, oedematous children often lose oedema within a week. (Golden 1993) Appetite returns and this, together with loss of oedema, heralds the recovery phase. Amongst our oedematous adult patients, the rates of oedema loss were variable and often much slower. On the LP diet, some patients lost most of their visible oedema and ascites within a few days, the rapid loss generally accompanied by watery diarrhoea. In these patients, care had to be taken to avoid intravascular hypovolaemia. In the absence of guidelines for adults, the author considered a loss of approximately 0.25 to 0.5 litres day % equivalent to a 0.25 to 0.5 kg day^ weight loss. Regulation of the rate of oedema loss was achieved by diluting the high-energy milk with ORS to an extent dictated by the severity of diarrhoea. In other patients, particularly those on the HP diet, the rate of loss of oedema was much slower and pedal oedema or ascites persisted for weeks. Persistent anorexia and debility accompanied this. These patients generally responded rapidly to introduction of the LP diet. The response involved marked increase in appetite and general well being and loss of oedema and ascites. However, minor grades of oedema sometimes persisted for weeks, even after they had recovered much of their original body mass. When oedema increased even on the LP diet the patient usually died.

150 Section 7.06 Conclusions

Due to the extreme conditions experienced in the TFC operating in a war zone, the subjects in this study were not randomised to the 2 diets, nor were their dietary intakes estimated. During most of tlie study, those with a worse prognosis on admission received the LP diet. Notwithstanding, on the LP diet, oedematous adults suffered lower rates of mortality and lost oedema more quickly than those receiving a more conventional HP diet. In marasmic adults, there was no difference in mortaht}'^ between patients on LP and HP diets. During the recovery phase, there were no differences in rates of weight gain between the LP and HP diet groups.

Thus, it appears that, compared with the conventional HP diet, the LP diet was more effective in the treatment of oedematous adults and as effective in the treatment of marasmic adults. The LP diet usually needed to be diluted during the initial phase of treatment. It was cheaper and more easily obtained than the specially manufactured famine-reHef foods of the HP diet. This study suggests that such a diet, based on ntilk, oü, sugar and locally available foods, with a relatively low P:E ratio, should be offered to all severely malnourished adults in both the initial and recovery phases of rehabilitation. By using locally available produce, it may also provide some stimulation of local economy. It should also ease the organisational difficulties involved in the provision of food items during emergency feeding operations. However, more research on treatment protocols for use in severe adult malnutrition is still needed.

151 Chapters Conclusions

This section pulls together the lessons from this study and presents tentative recommendations for the screening and treatment of adults during famine. The many methodological problems with this research means that it provides no definitive answers to the problems faced by humanitarian workers trying to admit and treat severely malnourished adults during famine. However, it does provide starting points for future research into the assessment and treatment of the condition. For the first time, it provides tentative recommendations for screening severely malnourished admissions into feeding centres, based on data gathered from severely malnourished adults during famine, rather than extrapolated from non-famine populations. This represents an advance from the very limited state of knowledge and screening guidelines that existed when this work began in

1992.

Section 8.01 Screening severeiy mainourished aduits during famine relief programmes

MUAC in combination with clinical signs should be used to screen adult entrance into feeding centres, using the classification shown below. BMI is inappropriate for this purpose, as it is affected by oedema and body shape and is difficult to measure. In any particular situation, workers should only use these suggested criteria as a starting point and adapt them to situation- specific factors.

152 8.01 (a) Admission criteria for TFCs

Admission criteria into adult therapeutic feeding centres should be based upon:

1. MUAC <160 mm irrespective of clinical signs or

2. MUAC 161-185 mm plus one of the following:

♦ Bilateral pitting oedema (Beattie grade 3 or worse) ♦ Inability to stand ♦ Apparent dehydration ♦ Famine oedema (Beattie grade 3 or worse). This must be assessed by a clinician to exclude other causes of pitting oedema. Additional social factors can be included in the model. The relative weighting of these must be determined locally; for example whether you need one, two or three additional social factors to tip the balance in favour of therapeutic rather than supplementary care. Relevant social factors could include the following:

♦ Access to food (quantity and quality) ♦ Distance from centres ♦ Presence / absence of carers ♦ Shelter ♦ Dependants ♦ Cooking utensils

In any particular situation, workers should take these suggested standards as the starting point and adapt them according to situation-specific factors.

8.01 (b) Admission to Suppiementary Feeding Programmes

Admission to adult supplementary feeding centres should be based upon the following cut-off:

1. M U A C 161 -185 mm and no relevant signs or few relevant social criteria.

153 Workers should again take these suggested standards as the starting point and adapt them according to situation-specific factors

Section 8.02 The treatment of severely malnourished aduits during famine relief programmes

The findings of this study indicate that a diet with a relatively low protein : energy ratio should be offered to all adults in therapeutic feeding centres. In this study a diet with a protein : energy ratio of 8.5%, diluted with ORS in cases of diarrhoea, proved more effective in treating severely malnourished adults than one with a protein : energy ratio of 16.4%. Since 1992, when the study was undertaken, similar lower protein diets have become commercially available and are now used as standard procedure during relief operations. In particular the Formula 75 (F75) developed for phase 1 feeding for the first few days after admission and formula 100 (F100)(Briend & Golden 1993; Golden 1997) for the rapid growth phase of treatment. F75 has a protein : energy ratio of approximately 5%, an osmolarity of 333 mOsmol/1 and a low lactose content. In addition, it is fortified with a mineral mix similar to that used in this study. It therefore appears appropriate for the phase one treatment of severely malnourished adults and I have more recently used it for that purpose with good effect. It is recommended that these diets be used in the treatment of severe adult malnutrition.

Section 8.03 Where do we go from here? future research needs:

8.03(a) Defining functionai cut-offs

Longitudinal studies should be undertaken to determine whether individuals falling below specific cut-off points for BMI, and MUAC have elevated morbidity or mortality, poor pregnancy outcome, decreased work ability or physical performance measures. Such studies must be conducted in a variety of situations with varying levels of adult under-nutrition. During emergencies, such studies are problematic. Ideally, studies relating indicators to the risk of mortality in a general population are required to establish functional cut-off points that could be used for the screening of adult admissions to feeding centres. However, famines always involve massive social upheaval and large numbers of afflicted people. These factors, combined with the

154 armed conflict that has characterised almost all famines during the last ten years, make such broad- based population studies unfeasible. Consequently, no such studies have yet been performed for any nutritional indicator in either adults or children during the height of a famine. Given these difficulties, relating indicators to mortality in a selected feeding centre population may be all that is reasonably possible. The selection bias involved in such studies must be acknowledged and results interpreted with caution.

8.03(b) Patterns and prevalence of adult under-nutrition during famine

Data on the prevalence and patterns of adult under-nutrition during famines may be useful in defining cut-offs. Data on different ethnic groups and settings would be useful.

8.03(c) Practicality of measurements and calculations

The practicality of obtaining various measures should be explored in field situations. Survey organisers should assess the ease of training survey workers in measuring MUAC, weight, and height, as well as assessing inter- and intra- observer variability in these measurements when measuring adults.

8.03(d) Adjusting for differences in body shape

Surveys undertaken in a variety of populations should explore the utüity of adjusting indices using weight and height for differences in body shape by using the SH/S Index or other indicators of body shape.

8.03(e) The use of MUAC to monitor recovery from under-nutrition

The changes in MUAC and weight during recovery from under-nutrition should be compared with a view to establishing whether MUAC is useful in this role and if it is, estabhshing relevant MUAC discharge criteria.

155 8.03(f) More data on differentiating between secondary and primary undernutrition

Examination of adult failure-to-thrive in centres should include attempts to identify measurements or signs that predict failure-to-thrive in feeding centres.

8.03(g) The aetioiogy, significance and treatment of famine oedema and ascites

More information on the significance and prevalence of famine oedema and ascites is required. It is recommended that surveys assess subjects for these symptoms.

Section 8.04 Community therapeutic care

The present focus on therapeutic feeding centres (TFC) as the sole mode of treating severely malnourished people during famine is inappropriate and often counter-productive. The centralised and resource-intensive TFC model of care is maladaptive. A TFCs huge requirement for resources, skilled staff and imported therapeutic products makes it very expensive and highly dependent upon external support. The centralised approach to care and high staff requirements undermine local health infrastructure, disempower communities and promote the congregation of people. Congregation of severely malnourished patients inside the centres promotes centre- acquired infection, a major problem in many TFCs. The congregation of communities around TFCs promotes breakdowns in public health, an important cause of mortahty and morbidity during famine. Vitally, during severe famine when severe adult malnutrition is prevalent, the TFC model of care cannot attain adequate coverage of the target population in a timely manner.

During the past three years Ready to Use Therapeutic Foods, nutritional equivalent to the lower protein diet research here, have become available. This opens the possibility of effective outpatient care for the majority of severely malnourished. Workers must research mechanisms of delivering these products to those who require them in an appropriate and cost-effective manner and the possibility of making Ready to Use Therapeutic Foods locally. Drawing upon mother to mother educational and mobihsational techniques, common in development projects might be an appropriate method to promote compliance with therapeutic regimes in the community.

156 Chapters Appendices

Section 9.01 Appendix 1

Table 22

BMI at admissions in non-oedematous patients who survived stratified by age.

age 1 1 Mean BMI Median BMI Mode BMI Variance Standard (}cars) kg kg m ^ kg m “ Deviation

157 13.3 13.0 11.4 3.24 1.80

< 105 1 2 . 2 11.9 1 2 . 0 2.56 1.60

Table 23

BMi at admissions in oedematous patients who survived stratified by age. { Age n Mean BMI Median BMI Mode BMI Variance Standard (years) kg kg kg m “ Deviation

+ 35 14 . 6 14 .1 14 . 4 7 .7 2.77

16 13. 6 13.0 12.7 13.2 3.63

Table 24

BMi at admissions of patients who died stratified by age

age n Mean BMI Median BMI Mode BMI Variance Standard (years) kgm'^ kg kgm“ Deviation

39 13.3 13.5 10.9 3.6 1.90

< 2 ; 2 1 12.7 1 2 . 1 10.7 3.6 1.90

157 LThe risk of death In oedematous patient stratified by sex

Table 25

Male oedematous patients admission BMI died survived Total kg m -

9 0 0 0

1 0 0 0 0

1 1 1 0 1

1 2 0 1 1

13 0 5 5

14 4 4 8

15 8 3 1 1

16 1 2 3

17 1 2 3

18 1 3 4

2 0 0 1 1

2 1 0 1 1

23 0 0 0

24 0 0 0

Total 16 2 2 38

158 Table 26

Female oedematous patients admission BMI died survived Total kg m -

9 0 1 1

10 0 3 3

11 1 3 4

12 1 3 4

13 1 7 8

14 2 5 7

15 1 4 5

16 2 1 3

17 0 0 0

18 0 0 0

20 0 0 0

21 0 0 0

23 0 1 1

24 0 1 1 total 8 29 37

159 Section 9.02 Appendix 2

100% 1

80%

70%

60%

> w i

30%

0% 0% 1 0 % 20% 30% 40% 50% 60% 70% 80% 90% 100% 100 - specificity ■clinical stepped sum model -“ BMI in marasmic patients

BMI in oedematous patients BMI in all patients

BMI in adult marasmic patients BMI in marasmic women

Figure 28

ROC curves for the prediction of mortality in the centre by the stepped sum clinical model in all admissions (n = 383), BMI in marasmic admissions (n = 218), BMI in marasmic adults (n = 112), BMI in oedematous admissions (n = 55) and BMI in all admissions (n = 383).

160 100%

90%

80% ü

1 70%

60%

50% ‘

II0 3 o 40% ■ “o H—marasmic adults 30% •

marasmic adolescents 20% •

0% 0%20% 40% 60% 80% 100%

100 - specificity (percentage of the population selected)

Figure 29

ROC curve for the prediction of short-term mortality by BMI in adolescents (n = 95) and adults ( n = 112)

Section 9.03 Appendix 3

1 he nieaii BMI value iii the Sudanese subjects was 3.0 kg m ” lower than in the Angolan and 7.1 kg in ■ lower than the weighted mean for the Somali subjects. The difference in weighted mean BMI between the Angolan and Somali subjects was 4.0 kg m “. The pattern in the MUAC data was similar, however the difference in mean MUAC values between the Sudanese and Angolans was less marked (1.3 cm) and that between the Angolan and Somali data greater (4.7 cm).

BMI, MUAC, sample size and analysis of variance of the BMI and MUAC data from 971 non- oedematous adult and adolescent subjects are presented in Table 27 and Table 28.

161 Table 27

BMI values in the different data set. 1 = S. Sudanese, 2 = non oedematous Angolan, 3 = Concern’s Somali workers, 4 = Somali women

data set Obs. Total Mean Variance Std. Dev. kgm'^ kgm'^ kg m'^ kg m'^ 1 96 1248 13.0 3.1 1.75 2 88 1408 16.0 6.1 2.46 3 145 3000 20.7 9.2 3.03 4 642 12757 19.9 5.0 2.2

data set Minimum 25%ile Median 75%ile Maximum Mode kpm'^ kpm'^ kgm'^ kgm'^ kgm'^ kpm'^ 1 10.1 11.6 13.1 14.2 17.7 13.7 2 11.6 14.3 15.9 17.2 25.4 16.1 3 13.5 18.8 20.2 22.3 29.9 21.8 4 10.8 18.4 19.8 21.2 29.0 20.8

Table 28

in the different data set. 1 = S. Sudanese, 2 = jnon oedematous Angolan, 3 = Concern lomali women

data set Obs. Total Mean Variance Std. Dev. cm cm cm cm 1 96 1768 18.4 4.2 2.06 2 88 1739 19.7 5.9 2.43 3 145 3733 25.7 7.3 2.71 4 642 15445 24.1 5.6 2.4

data set Minimum 25%ile Median 75%ile Maximum Mode cm cm cm cm cm cm 1 14.5 16.8 18.5 20.0 23.0 19.0 2 15.0 17.8 19.6 21.1 27.0 19.8 3 19.0 24.0 25.5 27.0 37.0 24.0 4 15.0 22.5 24.0 25.5 33.3 24.5

162 250

200

150

100

12 14 16 18 20 22 24 26 28 30 32 34 MUAC (cm)

Figure 30

Distribution of MUAC values in all patients Included in this work (n = 1940)

19 James et al. - Ethiopian women

James at al. Somali women 17

16 James et al. Chinese women

15 - - o - -South Sudanese (Nuer) women 14

14 15 16 B M I (k g /m

Figure 31

Ethnic variation in the relationship between MUAC and BMI. A comparison of the data from South Sudanese women included in this work, with data from populations not suffering from famine, reproduced from the work of James et al. (James et al. 1994)

163 14.5 cm & BMI 17 cm & BMI 10 100 *

X 20 cm & BMI 13

90 -

• 18.5cm & BMI 13 > :-Ew 80 - c — X— MUAC cut-off derived 0) (/) from normal populations

- - ♦ - - MUAC cut-offs derived 70 - from famine populations

0 10 20 30 40 50 60

1- specificity (%)

Figure 32

ROC for MUAC indictors in the identification of BMI thresholds (based upon data in Table 18)

Section 9.04 Appendix 4

9.04(a) The correction of BMI using the Cormic Index (SH/S)

I'o standardise BMI to take into account changes in SH/S ratio use the equations below to calculate BMI standardised to the actual SH/S ratio for the population under study.

Male subjects - BMI = 0.78(S11/S)-18.43 heinale subjects - BMI = I.19(SH/S)-40.34

Note: SH/S ratios should be expressed as a percentage

I he observed BMls can then be standardised to a SH/S ratio of 0.52 by adding the differences between the obser\"ed BMI and BMI standardised for the population SH/S ratio to a BMI standardised to 0.52 using the equation below:

BMlSstd=BMlo,52 + (BMlob-BMbs),

Where BMl^^j= standardised BMI,

164 BMI,, 52 — estimated BMI at SH/S of 0.52

B]\II,,y = actual BMI

BMI^^ = estimated BMI at actual SH/S

i. Exam pies

A Male population “A” has a mean BMI of 18.5 kg m^ and a mean SH/S ratio of 50%. The

BMI,, 52= 0.78*52-18.43 = 22.13. The BMI,, = 0.78*50-18.43 = 20.57. Therefore the BMI,^j = 22.13 + (18.5 - 20.57) = 20.06kg m^

A Female population “A” has a mean BMI of 17.0 kg m^ and a mean SH/S ratio of 54%. The BM1„52= 1.19 *52-40.34 = 23.92. The BMI,, = 1.19*54-40.34 = 21.54. Therefore the BMI,,j = 21.54 + (17.0 - 23.92) = 14.62 kg m"

165 References

19^5, Epidemiology and the control of disease in warm climate countries, 2 edn, Churchill Livingstone.

Alam, N., Wojtyniak, B., & Rahaman, M. M, 1989, "Anthropometric indicator and risk of éedXh", American Journal of Clinical Nutrition, vol.49, pp .884-888.

Apfelbaum-Kowalski, E. 1979, "Pathophysiology of the circulatory system in hunger disease," in Hunger disease, M. Winick, ed., John Wiley & sons. New York.

Ashworth, A. 1969, "Growth rates in children recovering from protein-calorie m a ln u tr itio n " , Journal of nutrition, \o\.22), pp.835-845.

Aykroid, W. R. 1974, The conquest of famine Chatto & Windus, London.

Bairagi, R. 1981, "On validity of some anthropometric indicators as predictors of mortality", A/wenca/i Journal of Clinical Nutrition, vol.34, pp.2592-2594.

Bairagi, R., Chowdhury, M., Kim, Y., & Curlin, G. 1985, "Alternative anthropometric indicators of mortality", American Journal of Clinical Nutrition, vol.42, pp.296-306.

Baker, J. P., Detsky, A. S., Wesson, D. E., Wolman, S. L., Stewart, S., Whitewell, J., Langer, B., & Jeejeebhoy, K. N. J 982, "A comparison of clinical judgement and objective measurements", Afew England Journal of Medicine, vol. APRIL 22, pp.969-972.

Barac-Nieto, M., Spurr, G. B., Lotero, H., Maksud, M. G., & Dahners, H. W. 1979, "Body composition during nutritional repletion in severely undernourished men", American Journal of Clinical Nutrition, vol.32, pp.981- 999.

Beattie, J., Herbert, P. H., & Bell, D. J. 1948, "Famine oeàtma", British Journal of nutrition, vol.2, pp .47-65.

Benedict, P. G.1907, The influence o f inanition on metabolism, 1 edn, Carnegie Institution of Washington, Washington D.C.

Benedict, F. G., Miles, W. R., Roth, P., & Smith, H. M.1919, Human vitality and efficiency under prolonged restricted diet Washington Publications, Washington.

Berkmann, J. M., Weir, J. P., & Kepler, E. J. 1947, "Clinical observations on starvation oedema, serum protein and the effects of forced feeding in anorexia nervosa". Gastroenterology, vol.9, pp.357-390.

166 Bern, C. & Nathanail, L. 1995, "Is mid-upper arm circumference a useful tool for screening in emergency sellings?", Lance/, vol.345, pp .631-633.

Bessler, H., Karp, L., Nolli, 1., Apler, A., Tyano, S., Djaldelli, M., Weizman, & R. 1993, "Cytokine production in anorexia nervosa". Clinical Neuropharmacology, vol.l6,no.3, pp.237-243.

Beumont, P. J. V., Russell, J. D., & Touyz, S. w. 1993, "Treatment of anorexia nervosa", Lance/, vol.341, pp.1635-1640.

Bishop, C. W., Bowen, P. E., & Ritchey, S. J. 1981, "Norms for nutritional assessment of American adults by upper arm anthropometry", Amencan Journal of Clinical Nutrition, vol.34, pp .2530-2539.

Boelaert, M., Davis, A., Le Lin, B., Michelet, M., Ritmeijer, K., Van Der Kam, S., & Vautier, F. 1995,Nutrition guidelines, 1 edn, Médecins sans frontières, Paris.

Borkan, G. A. & Norris, A. H. 1977, "Fat redistribution and changing body dimensions of the adult male". Human Biology, vol.49, pp.495-514.

Branca, F. & Abdulle, A. 1993, "Famine in Somalia",Lance/, vol.341, p.1478.

Bricnd, A. & et al. 1986, "Usefulness of nutritional indices and classification in predicting death of malnourished children", J5M7, vol.293, pp .373-376.

Bricnd, A. & Golden, M. H. N. 1993, "Treatment of malnutrition in refugee camps", Lance/, vol.342, pp.360-360.

Briend, A., Hasan, Kh. Z., Aziz, K. M. A., Hoque, B. A., & Henry, F. J. 1989, "Measuring change in nutritional status: a comparison of different anthropometric indices and the sample size required", Ear.J.C/m./Vn/r, vol.43, pp .769-778.

Briend, A., Wojtyniak, B., & Rowland, M. G. M. 1987, "Arm circumference and other Factors in children at high risk of death in rural Bangladesh", Lance/, vol. Sept.26, pp.725-727.

Briend, A. & Zimick, S. 1986, "Validation of arm circumference as an index of risk in 1 to 4 year olds",iVM/r./?e5., vol.6, pp.249-261.

Broom, J. 1993, "Sepsis and trauma," in Human nutrition and dietetics, 9 edn, J. S. Garrow & W. P. T. James, eds., Churchill Livingstone, London, pp.456-464.

Bruckner, W. J., Wies, C. H., & Lavietes, P. H. 1938, "Anorexia nervosa and pituitary cachexia", American Journal of Medical Science, vol. 196, pp .663-673.

167 Buchanan-Smith, M., Collins, S., Dammers, C., & Wekesa, F. 1999, Evaluation of Danish Humanitarian assistance to Sudan 1992 -1998, ODI, London. Report

Burger, G. C. E,, Sandstead, H. R., & Drummond, J. 1945, "Starvation in western Holland: 19A5",Lancet pp.282- 283.

Burgess, H. J. L. & Burgess, A. P. 1969, "The arm circumference as a public health index of protein-calorie malnutrition of early childhood, (ii) A modified standard for mid upper arm circumference in young children", J.Trop.Pediatr., vo\.15, pp.189-193.

Byers, T. 1996, "Body weight and mortality", New England Journal of Medicine, vol.333,no.ll, pp.723-724.

Caballero, B. 1989, "Optimal amount of dietary protein for treating adult malnutrition", Awm'can Journal of Clinical Nutrition, vol.68, pp.10-11.

CDC 1992, "Population based mortality assessment: Baidoa and Afgoi, Somalia 1992",M.M.W.R.,vo\Al, pp.913-917.

Centre for Disease Control 1992, "Famine affected, refugee and displaced populations: recommendations for public health issues",M.M.IT./?., vol.41.

Chen, L. C., Chowdhury, A., & Huffman, S. L. 1980, "Anthropometric assessment of energy-protein malnutrition and subsequent risk of mortality among pre-school children ",y4mer/cfln Journal of Clinical Nutrition, vol.33, pp.1836-1845.

Chen, L. C., Chowdhury, A., & Huffman, S. L. 1981, "The use of anthropometry for nutritional surveillance in mortality control programmes", Awencfln Journal of Clinical Nutrition, \o\34, pp .2596-2599. j I Clarke, G. M. & Cooke, D. 1982, A basic course in statistics, 2 edn, Edward Arnold, London.

I Collins, S. Data from Concern adult feeding programme, Melanje, Angola 1992-3. 1993. Unpublished Work

Collins, S. Data from Concern child therapeutic centres in Baidoa Somalia 1992 - 3. 1993.1993. Unpublished Work

Collins, S. 1993b, "Famine in Somalia",Lancet, vol341,no. June 5, pp.1478-1479.

Collins, S. 1993a, "The need for adult therapeutic care in emergency feeding programmes", JAMA, vol.270 #5, pp .637-638.

168 Collins, S. 1995, "The limit of human adaptation to starvation",iVflfwre Medicine, vol.l,no.8, pp.810-814.

Dallai, G. E. 1988, "Logistic: A logistic regression programme for the IBM PC,", The American statistician, vol.42, p.272.

Davis, A. P. 1996, "Targeting the vulnerable in emergency situations: who is vulnerable?",Lancer, vol.348, pp.868-871.

De Waal, A. 1989, Fa/nine that kills: Darfur Sudan, 1984-1985 (Oxford Studies in African Affairs), 1 edn. Clarendon Press, Oxford.

De Waal, A. 1993, "Famine in Somalia",Lancer, vol341,p.l479.

Dean, A. 0., Dean, J. A., Burton, A. H., & Dicker, R. C. Epi Info version 6: A word processing, database and statistics programme for epidemiology on microcomputers. 1990. Stone mountain, USD Incorporated. Computer Program

Debray, C., Zarazovitch, M., Ranson, B., Jacquemin, J., Robert, J., & Siraga, M. 1946, "Contribution a 1'etude de la pathologie des déportés", Sem.Hop.Pamno.22, pp.863-870.

Delpeuch, P., Cornu, A., Massamba, J. P., Traissac, P., & Maire, B. 1994, "Is body mass index sensitively related to socio-economic status and to economic adjustment? A case study from the Congo",EurJ.Clin.Nutr.,\o\A8 Suppl 3, pp. S141-7.

Detlwyler, K. A. 1992, "Nutritional status of adults in rural MaXi",AmJ.PhysAnthropol.no.8?>, pp.309-321.

Digby, W.1878, The famine campaign in southern India Longmans, Green & Co., London.

Diskette, P. 1995. Personal Communication

Draper & Smith 1981, Applied regression analysis, 2 edn, John Wiley and sons. New York.

Duffield, A. 1998, Chronic adult malnutrition . Thesis

Euronut SENECA investigators 1991, "Nutritional status and anthropometry",Far J.C//n.A^arr., vol.45,no. Suppl 3,pp.31-42.

Eveleth, P. B. & Tanner, J. M. 1976, Worldwide variation in human growth, 1 edn, Cambridge University press, Cambridge.

169 Farquharson, R. F. 1941, "Anorexia nervosa",///mow MedicalJournal, vol.80, p. 193.

Farquharson, R. F. & Hyland, H. H, 1938, "Anorexia nervosa: a metabolic disorder of psychological origin", X4A//4no.Ill, pp.1085-1092.

Feldesman, M. R. 1992, "Femur/stature ratio and estimates of stature in children", A/nenco/r Journal of Physical Anthropology , vol.87,no.4, pp.447-459.

Feldesman, M. R., Kleckner, J. G., & Lundy, J. K. 1990, "Femur/stature ratio and estimates of stature in mid- and late- Pleistocene fossil hominids",Awmcan Journal of Physical Anthropology, vol.83,no.3, pp .359-372.

Ferro-Luzzi, A. & James, W. P. T. 1996, "Adult malnutrition: simple assessment techniques for use in emergencies", 5n7/5/i Journal of nutrition, vol.75, pp.3-10.

Ferro-Luzzi, A., Setti, S., Franklin, M., & James, W. P. T. 1992, "A simplified approach to assessing adult chronic energy deficiency", £wr.J.C/in.//utr., vol.46, pp.173-186.

Fliederbaum, J., Heller, A., Zweibaum, K., & Zarchi, J. 1979, "Clinical aspects of hunger disease," in Hunger disease, M. Winick, ed., John Wiley & sons, New York, pp.12-43.

Folsom, A. R., Li, Y., Rao, X., Gen, R., Zhang, K., Liu, X., He, L., Irving, S., & Dennis, B. H. 1994, "Body mass, fat distribution and cardiovascular risk factors in a lean population of south Chim.", Journal of Clinical Epidemiology, vol.47,no.2, pp.173-181.

Forbes, G. B. 1983, "Survival during fasting [letter]", JAMA, vol.250,no.8, pp.1026-1026.

Forbes, G. B. 1987, "Lean body mass-body fat interrelationships in humans. ,Nutrition Reviews, \o\AS, pp .225-231.

Forbes, G. B. & Halloran, E. 1976, "The adult decline in lean body ma&s", Human Biology, vol.48,no.l, pp.l62- 173.

Forbes, G. B. & Reina, J. C. 1970, "Adult lean body mass declines with age: some longitudinal observations". Metabolism: Clinical & Experimental, vol.l9,no.9, pp.653-663.

Friedman, P. 1991, "Severe arm-muscle wasting: Sign of lethal malnutrition in sick elderly mQn",Nutrition, vol.7 #3, pp.223-226.

170 Gallagher, D., Visser, M., Sepulveda, D., Pierson, R. N., Harris, T., & Heymsfield, S. B. 1996, "How useful is body mass index for comparison of body fatness across age, sex, and ethnic groups?", Awencan Journal of Epidemiology, vol,143,no3, pp.228-239.

Garrow, J. S. 1993, "Human nutrition and dietetics," J. S. Garrow & W. P. T. James, eds..

Garrow, J. S. & Webester, J. 1985, "Quetelet's index (w/h2) as a measure of id\r\tss",IntJ.Obesity, vol.9, pp.l47- 153.

Gassull, M. A., Cabre, E., Vilar, A., Alastrue, A., & Montserrat, A. 1984, "Protein energy malnutrition: An integral approach and a simple new classification". Human Nutrition: Clinical Nutrition, vol.38C, pp.419-431.

Golden, B. E. Primary protein energy malnutrition. Garrow, J. S. and James, W. P. T. Human nutrition and dietetics.9[30], 440-455.1993. London, Churchill Livingstone. Data File

Golden, M. H. N. 1982, "Protein deficiency, energy deficiency and the oedema of malnutrition",vol.l, pp.1261-1265.

Golden, M. H. N. Protein requirements. 1995. Personal Communication

Golden, M. H. N. Severe Malnutrition. Weatherall, D. J., Ledington, J. G. G., and Warrell, D. A. The Oxford Textbook of Medicine (3rd ed.). [10.3], 1278-1296.1996. Oxford, Oxford University Press. Data File

Golden, M. H. N. 1997, "WHO manual on the treatment of severe malnutrition," in WHO manual on the treatment of severe malnutrition, WHO, Geneva.

Golden, M. H. N. 1999. Personal Communication

Golden, M. H. N. & Ramdath, D. 1987, "Free radicals in the pathogenesis of kwashiorkor", of the nutritional society, vol.46, pp53-68.

Golla, J. A., Larson, L. A., Anderson, C. F., Lucas, A. R., Wilson, W. R., & Tomasi, T. B., Jr. 1981, "An immunological assessment of patients with anorexia nervosa", Awencan Journal of Clinical Nutrition, vol.34/10.12, pp.2756-2762.

Gomez, F. & et al. 1955, "Malnutrition in infancy and childhood with special reference to kwashiorkor," in Advances in paediatrics, 7 edn, S. LEVIN, ed.. Year book publishers. New York, pp.131-131.

Gopalan, C. & Krishnaswamy, K. 1975, "Famine Oedema", Progress in food and nutrition science, vol.l #3, pp .207-224.

171 Gopalan, C., Venkatachalam, P. S., Someswara Rao, K,, & Menon, P. S. 1952, "Studies on nutritional oedema. Clinical, biochemical and pathological observations",/nrf/an Journal of Medical Sciences, vol.6,no.5, pp .277-295.

Government of Angola, M. p. o. 1993. Unpublished Work

Gull, W.1888, "Anorexia nervosa",La/icef, vol.l,no. June 30, pp.516-517.

Habicht, J. P. 1980, "Some characteristics of indicators of nutritional status for use in screening and surveillance", American Journal of Clinical Nutrition, vol33, pp 331-535.

Hakewell, P. & Moren, A. 1991, "Monitoring and evaluation of relief programmes", Tropical Doctor, vol. supplement 1, pp .24-28.

Harris, J. R., Tambs, K., & Magnus, P. 1995, "Sex-specific effects for body mass index in the new Norwegian twin panel". Genetic Epidemiology, vol.l2,no.3, pp.251-265.

Henry, C. J. K. 1990, "Body Mass Index and the limits of human survival",Æ'm/'.J.C/m v o l . 44, pp.329-335.

Henry, C. J. K. 1993, "Variability in adult body size: Uses in defining the limits of human survival," in Anthropometry: the individual and population, S. Ulijaszek, ed., Cambridge University Press, Cambridge, pp.117- 129.

Henry, C. J. K., Rivers, J. P. W., & Payne, P. R. 1988, "Protein and energy metabolism in starvation reconsidered", EwrJ.C/m.VMfr., vol.42, pp .543-549.

Heymsfield, S. B., McManus, C. B., Seitz, S. B., Nixon, D. W., & Smith Andrews, J. Anthropometric assessment of adult protein-energy malnutrition. Wright, R. A. and Heymsfield, S. B. [3], 27-82.1984. Boston, Blackwell scientific publications. Inc. Data File

Heymsfield, S. B., McManus, C. B., & Smith, M. F. 1982, "Anthropometric measurements of arm muscle mass: revised equations for calculating bone-free arm muscle area", American Journal of Clinical Nutrition, vol36, p.680.

Hortobagyi, T., Israel, R. G., & O'Brien, K. F. 1994, "Sensitivity and specificity of Quetelets index to assess obesity in men and women", Ewr_/.C/m.JVwm., vol.48, pp369-375.

Hyams, D. Curvexpert. [1.34]. 1993. Computer Program

Ifekwunigwe, A. E. 1971, "Recent field experiences in eastern Nigeria", Symposia of the Swedish nutrition foundation, vol.9, pp.144-154.

172 Immink, M. D., Flores, R., & Diaz, E. 0 . 1992, "Body mass index, body composition and the chronic energy deficiency classification of rural adult populations in Guatemala",vol.46,no.6, pp.419-427.

Iribarren, C., Sharpe, D. S., Burchfiel, C. M., & Petrovich, H. 1995, "Association of weight loss and weight fluctuation with mortality among Japanese American mtn", New England Journal of Medicine, vol.333, pp.686- 692.

James, W. P. T. 1994, "Introduction: The challenge of adult chronic energy deficiency ", EwnJ.C/m.JVwfr., vol.48,no. Suppl., p. S1-S9.

James, W. P. T., Ferro-Luzzi, A., & Waterlow, J. C. 1988, "Definition of chronic energy deficiency in adults. Report of a working party of the International Dietary Energy Consultative Giou^",EurJ,Clin.Nutr., vol.42,no.l2, pp .969-981.

James, W. P. T. & Francois, P. J. 1994, "The choice of cut off points for distinguishing normal body weight or "chronic energy deficiency" in didnlXs",EurJ.Clin.Nutr.,vo\AS^o. Suppl., pp. S179-184.

James, W. P. T., Mascie-Taylor, C. G. N., Morgan, N. G., Bistrian, N. G., Shetty, P. S., & Ferro-Luzzi, A. 1994, "The value of arm circumference in assessing chronic energy deficiency in Third World Q.àvXXs.",EurJ.Clin.Nutr., vol .49,no.l2, pp.883-894.

James, W. P. T. & Shetty, P. S. 199A,Body Mass Index: A measure of chronic energy deficiency in adults FAQ, Rome.

Jellife, D. B. 1966, The assessment o f the nutritional status of the community WHO, Geneva.

Jelliffe, E. & Jellife, D. B. 1969, "The arm circumference as a public health index of protein calorie malnutrition of early childhood", J.Trop.Pediatr.,vo\. December, pp.179-188.

Kerr, D. S., Stevens, M. C., & Robinson, H. M. 1978, "Fasting metabolism in infants. I. Effect of severe undemutrition on energy and protein utilisation". Metabolism: Clinical & Experimental, vol.27,no.4, pp.411-435.

Keys, A. 1946, "Human starvation and its consequences". Jour Am.Diet Assoc., vol 22, pp .582-587.

Keys, A. Diets for rehabilitation. Keys, A. The biology of human starvation.1 [50], 1057-1069.1950. Minnesota, The university of Minnesota press. Data File

Keys, A. 1950a, The biology of human starvation, 1 edn, Minnesota press, Minnesota.

Kirkwood, B. 1992, Essentials of medical statistics, 1 edn, Blackwell scientific publications, Oxford.

173 Knack 1950, "Quoted in The Biology of human starvation," A. Keys, ed..

Korcok, M. 1981, "Hunger strikers may have died of fat, not protein loss", JAMA, vol.246,no.l7, pp.1878-1879.

Kushner, R. F. 1993, "Body weight and mortality",iVutnïio/i Reviews, vol.51,no.5, pp.127-136.

Lazarus, R., Baur, L., Webb, K., & Blyth, F. 1996, "Body mass index in screening for adiposity in children and adolescents: systematic evaluation using receiver operating characteristic curves", American Journal of Clinical Nutrition, wo\.63,no A, pp .500-506.

Leiter, L. A. & Marliss, E. B. 1982, "Survival during fasting may depend upon fat as well as protein stores", JAMA, vol.248, pp .2306-2307.

Leiter, L. A. & Marliss, E. B. 1983, "Survival during fasting", JAMA, vol.250,no.8, p.1026.

Leyton, G. B. 1946, "The effects of slow starvation",Lance? pp.73-79.

Lindtjorn, B. 1985, "Measuring acute malnutrition: a need to redefine cut-off points for arm circumference?". Lancet, vol. ii, pp.1229-1230.

Lipscomb, F. M. 1945, "Medical aspects of Belsen concentration camp ", Lance?, vol. 1, pp .313 -315.

Ljunggren, H., Ikkos, D., & Luft, R. 1961, "Basal metabolic rates in obese and anorexic women", British Journal of nutrition no.15, pp .21-34.

Manson, J. E., Willett, W. C., Stampfer, M. J., Colditz, G. A., Hunter, D. J., Hankinson, SE, Hennekens, C. H., & Speizer, F. E. 1995, "Body weight and mortality among women [see commenis]", New England Journal of Medicine, vol.333,no.ll, pp.677-685.

Mattila, K., Haavisto, M., & Rajala, S. 1986, "Body mass index and mortality in the elderly",.SMJ, vol.292, pp .867-868.

Mazariegos, M., Wang, Z. M., Gallagher, D., Baumgartner, R. N., Allison, D. B., Wang, J, Pierson, R. N., Jr., & Heymsfield, S. B. 1994, "Differences between young and old females in the five levels of body composition and their relevance to the two-compartment chemical model", Journal of Gerontology, vol .49,no 5, pp. M201-8.

McCance, R. A. & Widdowson, E. M. 195\, Studies in undernutrition, Wuppertal 1946-9,1 edn, Medical Research Council, London.

McLaren, D. S. 1986, "Measuring malnutrition in famines (letter)",La/ice?, vol. ii,pp.152-152.

174 McLaren, D. S., Pellett, P. L., & Read, W. W. C. 1967, "A simple scoring system for classifying the severe forms of protein-calorie malnutrition of early childhood". Lancet, vol. March 11, pp.533-535.

McLaren, D. S. & Shirajian, E. 1969, "Short term prognosis in protein calorie malnutrition", Ammcan Journal of Clinical Nutrition, vo\22, pp.863-870.

Mecgan, M. K. 1992, "Rethinking famine relief", La/icef, vol.340,no.8830, pp.1292-1293.

Microsoft Corporation. Excel. Microsoft Excel[4]. 1994. Redmond, Washington, Microsoft corporation. 1994. Computer Program

Mitchell, J. B. & Black, J. A. 1946, "Malnutrition in released prisoners of war and internees at Singapore", Lance? no. Dec 14, pp.855-862.

Mollinson, P. L. 1946, "Observation on cases of starvation in Belsen",5MJ, vol. Jan 5, pp.4-8.

Moore, P. S., Marfin, A. A., Quenemoen, L. E., Gessner, B. D., Ayub, Y. S., Miller, D. S., Sullivan, K. M., & Toole, M. J. 1993, "Mortality rates in displaced and resident populations of central Somalia during 1992 famine". Lancet, vol.341, pp.935-938.

Murray, R. 0 . 1947, "Recovery from starvation",Lance? pp.507-511.

Naidu, A. N. & Rao, N. P. 1994, "Body mass index; a measure of the nutritional status in Indian populations. [Review]",L'ar.T.C/m.Aa?/'., vol.48 Suppl 3, pp. S131-40.

Negri, E., Santoro, L., DAvanzo, B., Nobili, A., & La Vecchia, C. 1992, "Body mass and acute myocardial infarction.", Preven?/ve Medicine, vol.21,no.3, pp.292-301.

Noppa, H., Anderson, M., Bengtsson, C., Bruce, A., & Isaksson, B. 1980, "Longitudinal studies on anthropometric data and body composition. The population study of women in Goteborg, Sweden", Awencan Journal of Clinical Nutrition, \o\3?>, pp.155-162.

Morgan, N. G. 1994b, "Body Mass Index and nutritional status: the effects of adjusting body mass index for the relative sitting height on estimates of the prevalence of chronic energy deficiency, overweight and obesity". As/an and Pacific Studies, vol. supplement.

Morgan, M. G. 1994c, "Interpretation of low body mass indices: Australian abongin&s", American Journal of Physical Anthropology, vol.94,no .2, pp .229-237.

175 Morgan, N. G. 1994a, "Population differences in body composition in relation to the body mass index. [Review]", Eur.J.Clin.Nutr.,vo\AS, Suppl.,no.3, p. sl0-s25.

Morgan, M. G. 1994d, "Relative sitting height and the interpretation of body mass index", Anna /5 of Human Biology, vol.21,no.1, pp.79-82.

Ohlson, M. A., Biester, A., Brewer, W. D., Hawthorne, B. E., & Hutchinson, M. B. 1956, "Anthropometry and nutritional status of adult women", T/nmanBiology, vol.28, pp.189-202.

Owen, O. E., Smalley, K. J., D'Alessio, D. A., Mozzoli, M., & Dawson, E. K. 1998, "Protein, fat and carbohydrate requirements during starvation: anaplerosis and cataplerosis", Am mcanJoarna/ of Clinical Nutrition, vol.68,no.1, pp.12-34.

Payne, P. R. & Dugdale, A. E. 1977, "Patterns of lean and fat deposition in adults", A^arare, vol.266, pp.349-351.

Picou, D., Halliday, D., & Garrow, J. S. 1966, "Total body protein, collagen and non collagen protein in infantile protein malnutrition". Clinical Science, vol.30, pp.345-351.

Popper, K. 1960, Conjecture and refutation Routledge, London.

Porter, A.1899, "Diseases of the Madras famine 1877-79. Quoted in The Biology of Human Starvation, Keys et al. 1950)," Madras.

Potter, J., Klipstein, K., Reilly, J. J., & Roberts, M. 1995, "The nutritional status and clinical course of acute admissions to a geriatric nmC, Age & Ageing, vol.24,no.2, pp.131-136.

Project team 1999, The Sphere project, 1 edn. The Sphere Project, Geneva.

Ramalingaswami, V., Deo, M. G., Guleria, J. S., Malhotra, K. K., Sood, S. K., Prakash, O., & Sinha, R%. V. M. 1971, "Studies of the Bihar famine of 1966 - 67", Symposia of the Swedish nutrition foundation , vol.9, pp.94-110.

Reddy, V. 1991, "Body mass index and mortality xdXes", Nutritional News, National Institute of Nutrition, Hyderabad, vol.12, pp.1-1.

Reeves, S. L., Varakamin, C., & Henry, C. J. K. 1996, "The relationship between arm span measurements and height with special reference to gender and tihniciiy”,EurJ.Clin.Nutr.,vo\50, p.398.

Roediger, W. E. W. 1986, "Metabolic basis of starvation diarrhoea: implications for treatment", La/iccr, vol.l, pp.1082-1084.

176 Rolland-Cachera, M. F., Cole, T. J., Sempra, M., Rossignol, G., & Charraud, A. 1991, "Body Mass Index

variations: centiles from birth to 87 y&2irs",EurJ.Clin.Nutr.,\o\A5, pp.13-21.

Ross, D. A., Berry, A. M., & Taylor, N. 1986, "Measuring malnutrition in famines ( l e t t e r ) vol. i, pp.151- 151.

Ross, D. A., Taylor, N., Hayes, R., & Mclean, M. 1990, "Measuring malnutrition in famines; are weight for height and arm circumference interchangeable?",/nrJ.£p/i/.,vol.l9: 3, pp.636-645.

Ross, R., Shaw, K. D., Rissanen, J., Martel, Y., de Guise, J., & Avruch, L. 1994, "Sex differences in lean and adipose tissue distribution by magnetic resonance imaging: anthropometric relationships", Ammca/r Journal of Clinical Nutrition, vol.59,no.6, pp.1277-1285.

Roubenoff, R., Dallai, 0 . E., & Wilson, P. W. 1995, "Predicting body fatness: the body mass index vs. estimation by bioelectrical impedance [published erratum appears in Am J Public Health 1995 Aug;85(8 Pt 1):1063]", American Journal of Public Health, vol.85,no.5, pp .726-728.

Sackett, D. L. & Holland, W. W. 1975, "Controversy in the detection of disease", Ta/icef, vol. August 23, pp.357- 359.

Sapir, D. G., Owen, O. E., Cheng, J. T., Ginsberg, R., & Boden, G. 1972, "The effects of carbohydrates on ammonium and ketoacid excretion during starvation", J.C//n./nve5ï.,vol51,pp2093-2102.

Schouten, E. J. & Borgdorff, M. W. 1995, "Increased mortality among Dutch development workers",5M7, vol.311,no.7016, pp.1343-1344.

Shakir, A. 1975, "Arm circumference in the surveillance of protein energy malnutrition in Baghdad", Awer/can Journal of Clinical Nutrition, vol.26, pp .661-665.

Sheik, M., Gutierrez, M. I., Bolton, P., Spiegel, P., Thieren, M., & Burnham, G.2000, "Deaths among humanitarian workers",BAfJ, vol221,no.7254, pp.166-168.

Shetty, P. S. 1984, "Adaptive changes in basal metabolic rate and lean body mass in chronic undemutrition". Human Nutrition: Clinical Nutrition, vol.38C, pp.443-451.

Shetty, P. S., James, W. P. T., & Ferro-Luzzi, A. 1994, "Malnutrition and the immune response.1. Malnutrition in the community: recent concepts". Transactions of the Royal Society of Tropical Medicine and Hygiene, vol .88, pp.612-614.

177 Smalley, K. J., Knerr, A. N., Kendrick, Z. V., Colliver, J. A., & Owen, O. E. 1990, "Reassessment of body mass indices", Amenca/z Journal of Clinical Nutrition, vol52,no.3, pp .405-408.

Smedman, L., Sterky, G., Mellander, L., & Wall, S. 1987, "Anthropometry and subsequent mortality in groups of children aged 6-59 months in Guinea-Bissau", A/nenca/z Journal of Clinical Nutrition, vol .46, pp369-373.

Somali Red Cross & Crescent. Figures for the number of unclaimed bodies buried in Baidoa 1992-1993. 1993. Unpublished Work

Sommer, A. & Loewenstein, M. S. 1975, "Nutritional status and mortality: a prospective validation of the QUAC stick.", A/zzenca/zJournal of Clinical Nutrition, vol.28, pp .287-292.

Stapleton, T. 1946, "Oedema in recovered prisoners of war",Lancet, vol.l, pp.850-850.

Steele, M. P. & Chenier, T. C. 1990, "Arm-span, height, and age in black and white women", Annals o f Human Biology, vol.l7,no.6, pp.533-541.

Steele, M. F. & Mattox, J. W. 1987, "Correlation of arm-span and height in young women of two races". Annals of Human Biology, vol. 14,no3, pp.445-447.

Stephens, L.1895, "Case of anorexia nervosa. Necropsy",Lancet, vol.l, pp.31-32.

Stini, W. A. 1973, "Reduced sexual dimorphism in upper arm circumference associated with a protein deficit diet in a South American population ", Journal of Physical Anthropology ,vo\36,pp341-352.

Stoudt, H. W., Damon, A., Mcfarland, R., & Roberts, J. Weight height and selected body dimensions of adults (1960 - 62). 76-1074.1965. Washington, US Gov. printing office. Health statistics series. Serial (Book, Monograph)

Strickland, S. & Ulijaszek, J. 1992, "Body mass index, ageing and differential reported morbidity in rural Sarawak", EurJ.Clin.Nutr.,volA l, pp.9-19.

Tomkins, A. & Watson, F. 1989, Malnutrition and infection. A review, WHO, Geneva. Report

Toole, M. J. & Waldman, R, 1990, "Prevention of excess mortality in refugee and displaced populations in developing countries",TAMA, vol.263, pp.3296-3302.

Tuomilehto, J. 1991, "Body mass index and prognosis in elderly hypertensive patients: a report from the European Working Party on High Blood Pressure in the Elderly", American Journal of Medicine,vo\.90,no.3A,pp.3AS-AlS.

178 Vaughan, J. & Pitt Rivers, R. 1945, "The value of hydrolysates in the treatment of severe starvation", Proc.R.Soc.Med.,\o\3^, p.395.

Vella, V., Tomkins, A,, Ndiku, J., Marshal, T., & Cortinovis, 1.1994, "Anthropometry as a predictor for mortality among Ugandan children, allowing for socio-economic variables", , vol.48.

Velzeboer, M. 1., Selwyn, B. J., Sargent, S., Pollitt, E., & Delgado, H. 1983, "The use of arm circumference in simplified screening for acute malnutrition by minimally trained health workers", J.Trop.Pec/mrr., vol.29, pp.l59- 166.

Velzeboer, M. 1., Selwyn, B. J., Sargent, S., Pollitt, E., & Delgado, H. 1984, "Evaluation of arm circumference as a public health index of protein calorie malnutrition in early childhood",J.Trop.Pediatr. ,wo\. ??, p. ??-??

Vera Todorovic. Performing some simple anthropometric measurements on residents in nursing homes. 1998. Unpublished Work

Ville de Goyet, C., Seaman, J., & Geijer, U. 1978, The management of nutritional emergencies in large populations World Health Organisation, Geneva.

Vis, H. L. 1985, "On the treatment of certain forms of protein energy-malnutrition in childhood with respect to fatal complications (an example from rural Central Afnca).", Annales Nestle, vol.43, pp.19-30.

Waaler, h. 1984, "Height, weight and mortality. The Norwegian experience",Acta.Med.Scand.,wo\.619,wo. Suppl., pp.1-59.

Wang, J., Russell-Aulet, M., Mazariegos, M., & et al. 1992, "Body fat by dual photon absorptiometry (DPA): comparisons with traditional methods in Asians, blacks and Caucasians", Awencan journal of Human Biology, vol.4,pp.501-510.

Wang, J,, Thornton, J. C., Russell, M., Burastero, S., Heymsfield, S. B., & Pierson, R. 1994, "Asians have lower body mass index (BMI) but higher percent body fat than do whites; comparisons of anthropometric measurements", Awencan Journal of Clinical Nutrition, vol.60, pp.23-28.

Waterlow, J. C. 1948, "Fatty liver disease in the British West Indies", MRC special reports, vol.263.

Waterlow, J. C. 1961, "The rate of recovery of malnourished infants in relation to the protein and calorie levels of

the diet.",J.7>'o/7.Re^/iafA'., vol.7,pp.16-22.

Waterlow, J. C. 1972, "Classification and definition of Protein Calorie malnutrition", RM7, vol.3,pp.566-569.

179 Waterlow, J. C. 1986, "Metabolic adaptation to low intakes of energy and T^roXtm",Ann.Rev.Nutr.,\o\.6, pp.495- 526.

Waterlow, J. C. 1993,Protein energy malnutrition, 1 edn, Edward Arnold, London.

Waterlow, J. C., Buzina, R., & et al. 1977, The presentation and the use of height and weight data for comparing the nutritional status of groups of children under the age of 10 years. Bulletin of the WHO, Geneva, 55. Report

Waterlow, J. C., Golden, M. H. N., & Patrick, J. Protein-energy malnutrition: treatment. Dickerson, J. W. T. and Lee, H. A. Nutrition in the clinical management of disease. 49-71.1978. London, Edward Arnold. Data File

Webster, J. D., Hesp, R., & Garrow, J. S. 1984, "The composition of excess weight in obese women measured by body density, total body water and total body potassium.", Human Nutrition: Clinical Nutrition, vol.38,no. C, pp .299-306.

Weinsier, R. L. & Krumdieck, C. L. 1980, "Death resulting from overzealous total parenteral nutrition: the refeeding syndrome rtvisiitù", American Journal of Clinical Nutrition, vol.34, pp.393-399.

WHO 1986, "Use an interpretation of anthropometric indicators of nutritional status", WHO Bulletin, vol.64 (6), pp.929-941.

Widdowson, E. M. 1976, "The responses of the sexes to nutritional slr&ss", Proceedings of the nutritional society, vol.35,pp.l75-180.

Winick, M. 1979, Hunger Disease Wiley-Interscience, New York.

World Food Programme - Operation Lifeline Sudan. Food distribution figures for Ayod, March 1993. 1993. Unpublished Work

World Food Programme Melanje & MSF Belgium, M. Under five nutritional survey in Melanje, November 1993. 1993. Unpublished Work

World Food Programme, M. Background data for Melanje, Angola - December 1993, Melanje. 1993. Personal Communication

Young, H. 1992, Food scarcity and famine: assessment and response Oxfam, Oxford.

Zimmer, R., Weil, J., & Dubois, M. 1944, "The nutritional situation in the camps of the unoccupied zone of France in 1941 and 1942 and its cor\st(\\\encts", New England Journal vol.230,pp.303-314.

180 181 Supplementary material

1. Collins, S. 1993, "The need for adult therapeutic care in emergency feeding programmes", vol.270 #5, pp.637-638.

2. Collins, S. 1995, "The limit of human adaptation to starvation". Nature Medicine^ vol.l, no.8, pp.810-814.

3. Collins, S. 1996, "Using middle arm circumference to assess severe adult malnutrition during famine", JAMA, vol.276, no.5, pp.391-395.

4. CoUins, S., Myatt, M., & Golden, B. E. 1998, "Dietary treatment of severe malnutrition in adults", American Journal o f Clinical Nutrition^ vol.68, pp. 193-199.

5. Collins, S. & Myatt, M. Aug 2000, “Short term prognosis in severe adult and adolescent malnutrition.” 284, no 5, pp 621 — 626

6. Collins, S., Duffield, A., & Myatt, M. Aug 2000 “Assessment of nutritional status in emergency afflicted populations - Adults”, ACC/SCN, RNIS suppl

182 The limit of human adaptation to starvation

Steve Collins

Reprinted from Nature Medicine, Vol. 1, No. 8, 1995 ARTICLES

The limit of human adaptation to starvation

St e v e C ollins

Concern Worldwide, Camden Street, Dublin 2, Ireland and Centre for International Child Health, Institute of Child Health, 30 Guilford Street, London WCIN lEH, UK Correspondence should be addressed to S.C., Oleuffynnon, Old Hall, Llanidloes, Powys SY18 6PJ

During the height of the 1992-93 famine in Somalia, data were collected from 573 inpatients at the Concern Worldwide Adult Therapeutic Centre in Baidoa, the town at the epicentre of the disaster. These data indicate that a body mass index (BMI, body weight in kilograms divided by height in metres squared) of less than 10 kg m ^ can be compatible with life, so long as specialized care is provided. Such low levels of BMI may be explained, in part, by the high ambient temperature, the tall phenotype of the Somalis, the gradual reduction in food intake and previous exposure to chronic energy deficiency. Famine oedema occurred with the same prevalence in male and female patients, but male patients had more severe oedema and a poorer prognosis at any given degree of severity. Survival from these extremes of emaciation has never before been recorded, and many of the BMI values documented here are below the level of 12, previously thought to mark the limit of human adaptation to starvation.

Severe adult malnutrition is commonly encountered during Centre in Baidoa began operating in October 1992, after it be famine emergency relief programmes'. As adult energy require came apparent that three-quarters of all deaths occurring in the ments are proportionally less than those in children, severe town were adults or adolescents®. malnutrition and death among adults occurs later than the peak in child mortality^. Often, by the time an emergency relief opera General inpatient data tion is up and running, many of the children have already died^ Of the 573 admissions from whom data were collected, 413 and starving adults constitute a large proportion of the popula (72%) survived, 122 (21%) died, and 38 (7%) were lost to follow- tion with nutritional problems^ \ Although there has been much up. Mean reported age of those that died was 36 years (95% interest in adult malnutrition, simple measurements such as confidence interval (Cl) = 33-39), slightly older than the sur height and weight have rarely been recorded systematically®. In vivors whose mean age was 33 years (95% Cl = 31.5-34.5; studies such as those of Keys et alJ and McCance et al^ where analysis of variance (ANOVA) F = 3.1, P = 0.08 ). The gender height and weight data were recorded, the degree of malnutri composition of those that died did not differ from survivors tion investigated was not severe. There is no classification of (likelihood ratio test - 1.5, P = 0.2). Nineteen percent of deaths severe adult malnutrition. Equivalent classifications in children occurred within 48 hours after admission, 48% within the first (weight for height and mid upper arm circumference) are funda week and 27% during the second week after admission (see Fig. mental to the assessment and screening of admissions to child 1). Mean time to discharge was 30.8 days, (34.6 days in oedema- feeding centres® '®, and the lack of such a classification in adults tous patients and 30.2 days in marasmic (extreme malnutrition hampers effective patient selection and assessment. This is espe with wasting of subcutaneous tissues and muscles) patients). cially difficult if the volunteers working in famine relief programmes are young, first-time recruits with little or no expe BMI data rience in similar situations. Body mass index (BMI, body weight Anthropometric data were recorded on admission to the centre in kilograms divided by height in metres squared) is a useful in (Table 1). The presence of severe famine oedema increased BMI dicator of chronic energy deficiency in adults""'®. Potentially it is at admission (Kruskall-Wallis statistic H = 29, P < 0.001). This in also useful in the assessment of severe acute adult energy defi crease was greater in male than in female patients (partial f ciency. This article presents an assessment of BMI during a severe statistic = 5.3, P < 0.001). BMI values for those that died were bi- famine. modally distributed depending on the presence or absence o f This study took place in Baidoa, the town at the epicentre of oedema. A peak in mortality around a BMI of 11 corresponded to the 1992 famine in Somalia. The famine was caused by the civil the mortality associated with marasmus, another around 15.5, to war that followed the overthrow of the president, Siad Barre, in mortality associated with oedematous malnutrition. In maras 1990. The Bay region of the country was particularly badly af mic patients, the level of BMI at admission was a reasonable fected, and by the autumn of 1992 tens of thousands of people predictor of mortality in the centre. This was not the case in p a had migrated to Baidoa, the regional capital, in search of food. tients with famine oedema (see Fig. 2). The association between The death rate among these displaced people was one of the BMI and risk of mortality in marasmics took different forms in highest ever recorded in a civilian population®. It was estimated male and female patients. In female patients the risk of mortality that by November 1992 up to 75% of the displaced children increased sharply below a BMI threshold of 11.0 (odds under the age of 5 had died®. The Concern Adult Therapeutic ratio = 9.0, likelihood ratio statistit = 7.9, P < 0.005). In male pa-

810 NATURE MEDICINE, VOLUME 1, NUMBER 8 , A U G U S T 19‘)S ARTICLES

tients no such threshold was observed, the risk of mortality grad Table 1 Data recorded on admission to Concern Adult ually increased as BMI declined (odds ratio for each decrease of 1 Therapeutic Center in Baidoa, Somalia kg m'^ = 2.3 likelihood statistic 10.8, P < 0.005). Admissions with marasmus or Admissions with severe famine mild famine oedema oedema

The lowest BMI com patible with life M eans Difference Means Difference (m ale: (m ale: The values for BMI at admission of survivors with marasmus or fem ale m ale fem ale fem ale) fem ale) mild famine oedema are presented (Fig. 3). Below a reported age of 25 there was a significant relationship between admission BMI Age, years 4 7 9 32 35 -3 92 35 37 -2 and age (regression coefficient of BMI on age = 0.22, P-coeffi- W eight, kg 370 37 .2 31.5 5 .6 * 83 4 2 .0 3 3 .7 8 .3 * cient 0.14, F statistic = 5.0, P < 0.05). No such relationship H eig h t (m ) 339 1 .6 8 1 .5 7 0 .1 2 * 79 1.67 1 .58 0 .0 9 * existed in patients with a reported age of 25 and above. These B M I survivors, 262 12.9 12.8 0.1 51 15.4 13.5 2 t data are therefore divided into older adults, aged >25 years, and kg m ' younger adults, aged 15-24 years. I'he geometric mean and min BMI non 36 13.0 11.5 1 .5 Î 24 14.8 1 3.7 1.1 imum admission BMI in the older adult group were 13.2 (95% survivors, kg m ' Cl = 12.9-13.5) and 10.4. Younger adults had BMI values ap Case fatality rate 442 18% 21% -3% 92 48% 27% 21 %§ proximately 1 kg m ^ below those recorded in patients aged >25 years. Geometric mean and minimum BMI values in this group *P < 0.001 ANOVA. were 42.1 (95% Cl = 11.8-12.4) and 9.6. Mean and minimum fP < 0.005 Kruskall-Wallis. BMI values were similar for male and female patients, once they $P = 0.01 Kruskall-Wallis. were adjusted for the severity of famine oedema. Weight loss, §P = 0.067 Yates corrected. due to loss of oedema during the initial phase of treatment, re duced many patients' BMI values still further. The lowest BMI usefulness in emergency situations. The time required to obtain values recorded in survivors were 10.1 for older, and 8.7 for the necessary height and weight measurements was much younger adults. Among these adults, 22% (/; = 43) of survivors greater in adults than in young children. During the early phase aged > 25 years and 49% {n = 59) of survivors aged between 15 of an emergency, when there are large numbers of people seek and 24 years had BMI values on admission below 12, the level ing treatment, this extra time effectively prevents the use of BMI previously considered the lowest compatible with life"'\ The sur as a screening tool. In children, assessment of mid upper arm cir vival rate among those admitted with BMI values below 11.0, cumference (MUAC), a quick, but sometimes less precise, index was 65% for older adults (n = 23) and 82% for younger adults of nutritional status’ ", is often used during these early phases. (» = 34). Mean BMI on discharge was 15. There is some research to show that MUAC may be a useful nu tritional indicator in adults'^"*; however, its use for emergency Famine oedema feeding programmes has not yet been evaluated. Mortality rates were higher in patients presenting with severe The inability of many of the sickest patients to stand up famine oedema (Table 1). This was common, seen in 16% of all straight to have their height measured also limited the usefulness admissions and 28% of those that subsequently died, and was as of BMI. These problems, similar to those met during the assess sociated with a poorer prognosis (odds ratio = 2.4, Yates' ment of prisoners liberated from Belsen concentration camp", corrected = 11.7, P< 0.001). This association was stronger in resulted in approximately 11% of all admissions and 33% of male than female patients (Wolfs' x’ = 4.59, P < 0.05). The poor those who subsequently died not having their BMI assessed on prognosis seen in patents with severe famine oedema was pre admission. The use of a proxy for height, such as arm span, dominantly a result of the increased mortality in patients would improve this. However, the high prevalence of famine presenting with pitting oedema, where the oedematous tissue re oedema with its poor prognosis would still result in underesti tains for a time an indentation produced by pressure. Pitting mating the degree of emaciation based on BMI in many of the oedema was independently associated with mortality (odds ratio most severe cases. The incidence of famine oedema recorded for each increase in grade = 1.5, likelihood ratio test = 22.2, P < here is not unique. In Belsen, whereas only 6% of patients had 0.001). Male patients had a poorer prognosis at severe grades of moderate or severe famine oedema at admission, many more de pitting oedema than female patients as shown in Fig. 4 (odds veloped it upon treatment’". In Melanje, Angola, 90% of 1,100 ratio (male) = 1.7, likelihood ratio test 23.5, P< 0.001; odds ratio (female) = 1.3, likelihood ratio test = 4.4, P < 0.05). Minimal pit 25 ting oedema was not associated with an increased risk of death in either sex.

Discussion BMI as an indicator for use in severe famine BMI was relatively good for indicating the risk of short-term mortality in marasmic patients. The BMI levels used to define differing grades of chronic energy deficiency (CED) were, how ever, not relevant to severe famine. Most of the displaced adults living in Baidoa during the autumn of 1992 had BMI values

below 16 kg m'\ the upper limit defining severe CED". Such a 2 6 10 14 18 22 26 30 34 38 42 46 50 cut-off was therefore of little use in differentiating between those Days elapsed after admission who needed therapeutic feeding and those who would survive without it. BMI also had practical limitations that reduced its Fig. 1 Timing of the deaths after admission.

NATURE MEDICINE, VOLUME 1, NUMBER 8, AUGUST 1995 ARTICLES

adults admissions to the Concern Worldwide centre suffered from famine oedema (S.C., unpublished data). There is a need for research to evaluate the use of an alternative indicator, such as MUAC or clinical signs, for the screening and assessment of adult malnutrition.

The BMI limit in human starvation These data show that the level of emaciation compatible with life is much lower than previously thought. Conceivably the low BMI values presented here could be in part due to a tall, long- legged phenotype. This phenotype occurs in some east African <11 < 11.5 <12 < 12.5 < 13 < 13.5 <14 < 14.5 <15 < 15.5 <16 < 16.5 <17 < 17.5 ' races, notably Nilotic peoples, such as the Dinka of southern BMI cut-off Sudan and the Samburu of northern Kenya’^ '’. The BMI values in Somalis, a Cushitic people, however, have been recorded as lying ■ Marasmus or mild famine oedema ■ Severe famine oedema within the normal range of 19-21 (ref. 20). Complementary BMI data, collected in the spring of 1993 from 164 of Concern's local Fig. 2 The risk of m ortality associated w ith various BMI cut-offs. Somali workers in Baidoa, and 326 adult Somali women living in the villages surrounding Baidoa, also recorded mean BMI values within the normal range (mean BMI = 20.8, 95% Cl = 20.3-21.3 was such a difference, the BMI data would have a sexual bias, and mean BMI = 19.5, 95% Cl = 19.2-19.7, respectively, S.C., un greatest at the lower levels of BMI where the inability to walk was published data). Nilotic people can survive at even lower BMI more frequent. As a result mean female BMI could have been ar values. In a Concern Adult Therapeutic Centre in Ayod, southern tificially raised and any sexual differences obscured. Previous Sudan, BMI values as low as 9.3 were recorded in adults who anthropometric data are contradictory on this point. Mean were still able to walk (S.C., manuscript in preparation). weights of people that died during the Madras famine of 1877, Survival at these extremely low BMI values would have been were 77 pounds for both men and women (quoted in Aykroitf'). aided by the warm climate that existed in Baidoa throughout the On the other hand, Henry, reviewing data largely collected dur period that the centre was operating. Much of the previous data ing the Dutch famine of 1945, concluded that the minimum concerning the limits of survival during famine comes from BMI compatible with life is markedly lower in women''. More re colder climates*’^'. Given the inability of severely malnourished search is needed before realistic anthropometric cut-offs for use people to regulate their body temperature^^“ “, the limits of sur in severe adult malnutrition are established. vival are likely to be lowest in warm climates. The gradual nature of the reduction in food intake over a pe Adult therapeutic care riod of two years, as well as previous exposure to CED, may also The provision of food and medical care in a protected inpatient have facilitated survival at such low BMI values. Habituation to environment was instrumental in the survival of those admitted. CED has been described in Indian labourers who, with a mean It is likely that very few would have survived outside the centre, BMI of 16.6, were able to function reasonably normally^\ By forced to queue for food in the general kitchens. However, since contrast, similar BMI values, rapidly induced in 32 previously the concentration camps were liberated at the end of the Second well-nourished volunteers, resulted in extremely poor physical World War, adult therapeutic centres have not been a feature of and psychological states. The inability of many of the most ema emergency relief programmes. The relatively low case fatality ciated patients to stand did not allow for BMI assessment. The rates, even at extreme grades of emaciation, as well as the fast mean and minimum BMI values recorded here are therefore rates of rehabilitation reported here, support the feasibility and higher than the true values at the centre. utility of adult therapeutic centres. These should constitute a standard part of emergency feeding programmes. There is a need Sex difference in adaptation for more research into indicators useful in the screening and as Sex differences in the extent and prognosis associated with sessment of admissions to such centres. famine oedema were observed. Male patients suffered from more severe grades of oedema and had a worse prognosis at any given grade of severity. This is in keeping with previous observations 35 that famine oedema occurs less frequently in women“'”“. These 30 differences in prognosis will need to be taken into account in the 25 creation of guidelines for the assessment and screening of severe 20 adult malnutrition. 15

Whether the lowest level of BMI compatible with life varies be 10 tween men and women is less clear. Once the data were adjusted 5 to allow for severity of famine oedema, BMI values appeared sim 0 ilar for male and female patients. However, this should be 9 10 11 12 1314 15 16 17 18 interpreted with caution. The inpatients at the centre were a se Body mass index (BMI) lected population, people who had walked, been carried by ■ Older adults' age >25 'Young adults' age 15-25 friends and family, or been transported to the centre by aid workers. The greater numbers of male medical admissions, many of whom could not walk far, is possibly a reflection of an in Fig. 3 Body mass index on admission (survivors with marasmus or creased tendency to carry male patients to the centre. If there mild famine oedema).

NATURE MEDICINE, VOLUME 1, NUMBER 8, AUGUST 1995 T L ARTICLES

Methods 12.00 Subjects. Data were collected on 573 patients aged 15 years and over, admitted to the Concern Worldwide Adult Therapeutic Centre, 10.00 Baidoa, between 25 October 1992 and the end of March 1993. 8.00 They were aged between 15 and 80 years (mean 34) and 46% were male. On admission, patients were registered and weighed and had 6.00 their height measured by trained local enumerators, supervised by a 4.00 nurse or doctor. A rapid clinical screen, assessing degree of pitting oedema, ascites, anaemia, level of hydration, dysentery, diarrhoea, 2.00 17 14 signs of chest infection and ability to stand was performed by the 0.00 author or an expatriate nurse. Individual treatment cards were then 2 3 issued to each patient. The clinical condition and weight of each pa Grade of oedema tient was monitored during rehabilitation, and the outcomes were 0 Male female recorded. The criteria for admission varied depending on the avail able space in the centre and the existence of other medical facilities in Baidoa. During October and November 1992 there were no alter Fig. 4 Pitting o ed em a and th e risk of mortality. native medical centres in Baidoa, and therefore, we had to admit a small number of patients whose problems were primarily medical. From mid-December 1992 on, however, MSF (The Netherlands) op fectious disease, a good appetite, constant weight gain, and the abil erated a medical inpatient unit in Baidoa allowing us to refer such ity to walk and care for themselves were all necessary conditions for patients. In general only patients with a BMI below 1 3.5, or those discharge. The presence of minimal pedal oedema did not preclude with famine oedema were admitted. Less severely malnourished discharge so long as all the above criteria were fulfilled. Upon dis people were referred to supplementary kitchens. charge all patients were issued a 'rehabilitation pack' containing a Treatment at the centre. High levels of supervision, with staff pa sheet of plastic, hoes, seeds, a machete, pots and pans, a blanket, tient ratios of 1:8 or less, particularly during the first few days after water containers and some cloth. Transport to the patients' home admission, were maintained. These staff distributed the food and village was provided. oral rehydration solution (ORS), encouraged patients to eat and Equipment. The centre consisted of nine stick shelters with plastic drink and helped feed those who were unable to feed themselves. sheeting roofs and mud floors. These provided room for 140 inpa Patients were orally rehydrated using the WHO formula ORS tients and 140 family attendants who helped with the day-to-day (sodium chloride 3.5 g I ’, trisodium citrate 2.9 g I ', potassium chlo care of their relatives. Separate shelters were designated for new ad ride 1.5 g I ’ and glucose 20 g I '). Six to eight meals were offered missions, those with severe oedematous malnutrition, those with each day. The diets consisted of high-energy milk (HEM), made dysentery and those suspected of having pulmonary tuberculosis. A from dried skimmed milk, vegetable oil and sugar), sweet tea, rice, set of standard Hanson spring scales and locally constructed height oil, beans, biscuits and local fruits, especially bananas. During the boards were used to weigh and measure the height of patients. The first few days after adm ission, th e HEM c o m p o n en t w as diluted w ith scales were calibrated in Baidoa and London using known weights. ORS. This reduced the incidence and severity of the watery diar The results were similar at each site: In the useful range of 20-50 kg rhoea and vomiting that is often associated with the early stages of there was a linear increase in error with the true weight being ex refeeding severely malnourished people. The degree to which the pressed by the equation, true weight = (1.013987 x scale weight) + HEM was diluted was tailored according to the severity of the diar 0.707, obtained using linear regression techniques. The correction rhoea. factors derived from this equation were used to correct each weight Two therapeutic regimes were used in the centre. A high-protein measurement. regime wherein patients were offered 156 g protein and 16.5 Mj Data coding and analysis. Data were transferred from the inpatient (16% of energy from protein, P:E ratio 9.5 g per Mj), and a low-pro- record cards onto specially designed coding forms, entered in dupli tein regime wherein patients were offered 81 g protein and 16.5 Mj cate into a PC, and validated for entry error. Pitting oedema was (8.5% of energy from protein, P:E ratio 5 g per Mj). In oedematous coded using the classification of Beattie et (0 = absent, 1 = mini patients the high-protein diet was commonly associated with pro mal oedema on the foot or ankle that was demonstrable but not longed anorexia, increased or persistent oedema, and occasional visible, 2 = visible on foot or ankle, 3 = demonstrable up to knee, instances of sudden death. The introduction of the low-protein diet 4 = demonstrable up to inguinal ligament, and 5 = anasarca). Pitting in December 1992 reduced these complications dramatically. This oedema, recorded as present but unquantified, was assumed to be effect was so marked that subsequently all oedematous patients moderate (grade 3). Periorbital oedema was coded as 1 (mild), 2 were given the low-protein diet during the initial phase of their (moderate) and 3 (severe), with 'present but unquantified' again treatment. Data collected at the centre has supported these obser designated as moderate (grade 2). Ascites was similarly coded. vations, demonstrating a reduction to one-fourth in mortality Severe oedematous malnutrition was defined as either pitting amongst the severe oedematous patients who received the low-pro oedema greater than grade 2, periorbital oedema greater than tein diet (S.C., manuscript in preparation). grade 1, or any ascites. Missing data values were coded as not pre Oral antibiotics were needed by the majority of the patients upon sent. Logistic regression analysis was used to assess the association of admission, with many patients receiving an antibiotic throughout clinical features at presentation with mortality. Variables that were their stay in the centre. Penicillin V or ampicillin were the first-line not independently associated with mortality were eliminated from antibiotics for pulmonary infections; cotrimoxazole and metronida the models in a backwards stepwise fashion using estimation tech zole for dysenteries and persistent diarrhoeas. Chloramphenicol was niques. All data entry and analysis were completed using Epi-lnfo. the second-line antibiotic. v6.02 (ref. 29) and Logistic’®. Discharge criteria were predominantly clinical: freedom from in- Problems. The frequency with which patients were weighed and the

NATURE MEDICINE, VOLUME 1, NUMBER 8, AUGUST 1995 ARTICLES

collection of clinical data varied. This was due to the frequent gun 11. Ferro-Luzzi, A., Setti, S., Franklin, M. 6c James, W.P.T. A simplified approach to fighting, bandit attacks on expatriate staff houses, attempts at assessing adult chronic energy deficiency. Eur. J. Clin. Nutr. 46,173-186 (1992). 12. James, W.P.T., Ferro-Luzzi, A. 6c Waterlow, J.C. Definition of chronic energy hostage taking, and military occupations that occurred in Baidoa malnutrition in adults. Eur. ]. Clin. Nutr. 42, 969-981 (1988). during the study period. The high incidence of expatriate illness and 13. James, W.P.T. 6c Shetty, P.S. Bcxly mass index: An objective measure for the es timation of chronic energy deficiency in adults. (World Health Organization, resulting evacuations on medical grounds, contributed to these Geneva, 1992). problem s. 14. Ross, A.R., Taylor, N., Hayes, R. 6c Mclean, M. Measuring malnutrition in famines: are weight for height and arm circumference interchangeable? i«(./. Epidemiol. 19 (3), 636-645 (1990). Acknowledgements 15. Friedman, P. Severe arm-muscle wasting: Sign of lethal malnutrition in sick el I would like to thank the Concern staff in Baidoa, who made this study derly men. Nutrition 7 (3), 223-226 (1991). possible; Mark Myatt for his considerable help with data analysis; Barbara 16. James, W.P.T. 6c Norgan, N.G. The value of arm circumference in assessing chronic energy deficiency in Third World adults. Eur. /. Clin. Nutr. 49, 883-894 (1994). Golden, Mike Golden, Sally Grantham-McGreggor, Ben Rigby, Jeremy 17. Mollinson, P.L. Observation on cases of starvation in Belsen. Br. Med.f. 5 (Jan), Shoham, Keith Sullivan and Andrew Tomkins for their advice and 4-8 (1946). encouragement. 18. Lipscomb, P.M. Medical aspects of Belsen concentration camp. Lancet 1, 313-315 (1945). 19. Eveleth, P.B. 6c Tanner, J.M. Worldwide Variation in Human Growth (Cambridge RECEIVED 2 8 FEBRUARY; ACCEPTED 16 JUNE 1 9 9 5 Univ. Press, Cambridge, 1976). 20. Branca, F. 6c Abdulle, A. Famine in Somaiia. Lancet 341,1478 (1993). 1. Collins, S.R. The need for adult therapeutic care in emergency feeding pro 21. Henry, C.J.K. Variability in adult body size: Uses in defining the limits o f grammes. JAMA 270, 637-638 (1993). human survival, in Anthropometry: The Individual and Population. (Cambridge 2. Winick, M. Hunger Disease. (Wiley-Interscience, New York, New York, 1979). Univ. Press, Cambridge, 1993). 3. Moore, P.S., et al. Mortality rates in displaced and resident populations of cen 22. Burger, G.C.E., Sandstead, H.R. 8c Drummond, J. Starvation in western Holland: tral Somalia during 1992 famine. Lancet 341, 935-938 (1993). 1945. Lancet 2S2-283 (1945). 4. Centers for Disease Control. Population based mortality assessment: Baidoa 23. Leyton, G.B. The effects of slow starvation. Lancet 1, 73-79 (1946). and Afgoi, Somalia 1992. Morbid. Mortal. Weekly Rep. 41, 913-917 (1992). 24. Shetty, P.S. Adaptive changes in basal metabolic rate and lean body mass in 5. Somali Red Cross & Red Crescent: Figures for the number of unclaimed bodies chronic undernutrition. Hum. Nutr. Clin. Nutr. 41C, 287-290 (1984). buried in Baidoa 1992-1993 (1993). 25. Keys, A. Human starvation and its consequences. /. Am. diet. Assoc. 22, 582-587 6. Henry, C.J.K. Body mass index and the limits of human survival. Eur. J. Clin. (1946). Nutr. 44, 329-335 (1990). 26. Gopalan, C. 6c Krishnaswamy, K. Famine oedema. Progr. Food Nutr. Sci. 1 (3), 7. Keys, A. The Biology of Human Starvation. (Univ. Minnesota Press, Minneapolis, 207-224 (1975). Minnesota, 1950). 27. Aykroid, W.R. The Conquest o f Famine. (Chatto &c Windus, London, 1974). 8. McCance, R.A. 6c Widdowson, E.M. Studies in undernutrition, Wuppertal 28. Beattie, J., Herbert, P.H. 6c Beli, D.J. Famine oedema. Br. J. Nutr. 2, 47-65 (1948). 1946-9. (Medical Research Council, London, 1951). 29. Dean, A.G., Dean, J.A., Burton, A.H. 6c Dicker, R.C. Epi Info version 6: A word 9. Ville de Goyet, C., Seaman, J. 6c Geijer, U. The management of nutritional processing, database and statistics programme for epidemiology on microcom emergencies in large populations. (World Health Organization, Geneva, 1978). puters. (USD Inc., Stone Mountain, Georgia, 1990). 10. World Health Organization. Use and interpretation of anthropometric Indica 30. Dallai, G.E. Logistic: A logistic regression programme for the IBM PC. Am. tors of nutritional status. WHO Bull. 64 (6), 929-941 (1986). Statistic. 42, 272 (1988).

NATURE MEDICINE, VOLUME 1, NUMBER 8, AUGUST 1995 Letter From Baidoa

The Need for Adult Therapeutic Care in Emergency Feeding Programs Lessons From Somalia

Severe adult malnutrition has been neglected by both the centers that served 6 to 8 meals per day. The supplementary medical profession and the agencies providing relief to famine feeding centers were generally medically supervised, pro victims. As adults are less vulnerable to the effects of lack of viding oral rehydration, basic medical care, and referrals. The food in the short term, it has been tacitly assumed that their therapeutic centers, usually organized by two non-Somali rehabilitation requires nothing more than a basic ration. This nurses, with the backup of a physician on call, were quasi- assumption continues despite several post-World War II medical inpatient departments. In addition to the special articles relating the importance of appropriate diets and care diets, these centers provided intense nursing supervision; in the treatment of severely emaciated adults (for example, active oral, nasogastric, or occasionally intravenous rehydra Lipscomb,^ 1945). Thus, while severely malnourished chil tion; basic antibiotics, such as ampicillin, eo-trimoxazole, and dren are provided with specifically designed rehabilitation metronidazole; antihelminthics; malaria treatment and pro diets with medical and social care, such measures are not phylaxis; vitamins; and basic nursing care. usually taken to help adults or adolescents. In sharp contrast, up until the end of October, the only The high adult death rate seen in Baidoa, Somalia, during services for the starving adults were community kitchens. the autumn of 1992 calls for a reconsideration of the ways in These kitchens had no medical input, and although they were which emergency aid is targeted. Adult deaths accounted for supposed to provide about 8.4 MJ of energy in the form of one 8900 of the total 15105 deaths recorded by the Somali Red plate of rice and beans a day, they were plagued by looting Crescent Society in Baidoa between August 1992 and Feb and shortages of supplies. Even when food was available, the ruary 1993. More than 4200 of those adult deaths occurred lack of supervision and care meant that the sheer physical after September 1992, despite the fact that reasonable quan struggle needed to maintain a place in the queue often re tities of food and resources were available. These figures are sulted in the thinnest going without. This problem was par from the records of the Somali Red Crescent Society “death tially addressed at a meeting of the nongovernmental orga truck,” which collected unclaimed bodies (those not buried by nizations in Baidoa at the end of October and special “vul relatives) for burial. Unfortunately, the workers on the truck nerable” kitchens were started. Scouts were appointed to were paid according to the numbers of bodies they buried, identify the most vulnerable people who were then provided and in December it was found that their figures were grossly with oral rehydration and basic medical care in addition to the inflated. Whether this distortion was taking place throughout usual meal of rice and beans. the collection period is unknown. However, as the numbers In October 1992, Concern Worldwide, a nongovernmental recorded for the period before November are consistent with relief and development agency, opened Bardalli Adult Ther those obtained by a team from the Centers for Disease Con apeutic Center, situated in Baidoa town in the Bay region of trol and Prevention,^’^ it seems more probable that the ex central Somalia. The center consisted of nine stick shelters aggeration became greater as the actual number of deaths with mud floors providing room for 140 severely malnour and the earnings of the body collectors decreased. Although ished adults and adolescents and approximately 140 addi the mortality figures have to be viewed with caution, they do tional family attendants. All of the first 650 adult and ado illustrate the point that a large percentage of all deaths were lescent admissions were seriously ill at presentation, and it adult and that this high adult mortality rate persisted in the is likely that none would have survived given a basic ration face of a massive international relief operation. of rice and beans alone. Simple measures rehabilitated most The majority of the international response to the crisis in very quickly, with an average stay of 25 days. Baidoa started in June 1992, but because of fighting and The majority were treated with only active oral rehydra looting, relief was mainly limited to “wet selective feeding” tion, basic antibiotics, and a therapeutic diet. The patients (the distribution of prepared food to targeted populations) were generally from the Bay region of Somalia, an area that until the US Army intervention in mid December. Malnour up until 1990 had been agriculturally rich, providing surplus ished children were rehabilitated either in supplementary food for other regions in Somalia. The patients were agri feeding centers that provided two meals of milk, biscuits, and culturalists and pastoralists who had been robbed of their “unimix” porridge each day, or in 24-hour therapeutic feeding crops, animals, and stocks of food during the previous 2 years of war. Many had walked scores of miles in search of food, and all were chronically malnourished having existed on virtually Edited by Annette Flanagin, RN, MA, Associate Senior Editor. Reprint requests to Concern Worldwide, 248-250 Lavender Hill, London, England nothing apart from local roots, leaves, and grasses for months. SW11 1LJ (Dr Collins). On admission most of the patients were extremely dehy

JAMA, August 4, 1993—Vol 270, No. 5 Letter From Baidoa 6 3 7

Reprinted from JAMA ® The Journal of the American Medical Association Augusl4, 1993-Volume 270 Copyright 1993, American Medical Association drated. Diarrhea was universal; 30% had dysentery; 40% had Whereas the majority of the children with kwashiorkor rap pulmonary infections; and 30% had edema often combined idly improved on the relatively high protein diets generally with ascites. Penicillin V or ampicillin were the first-line used in Baidoa, many of the edematous adults given the same antibiotics for pulmonary infections; co-trimoxazole and met diet did not respond well, often increasing their edema and ronidazole were administered for dysentery and diarrhea; dying unexpectedly. These observations are consistent with and chloramphenicol was the general second-line antibiotic. records of the refeeding of malnourished adults in Holland The majority of patients received one of these antibiotics on during 1944® and from Belsen concentration camp in 1945' admission, and many remained on an antibiotic throughout where relatively high protein diets were used. their stay in the center. In addition to these drug regimens, During the last 35 y ears the lack of attention paid to the prob the segregation of the patients with tuberculosis and infec lems of severely malnourished adults and adolescents by the tive diarrhea fi-om the rest of the patients appeared to reduce medical profession and rehef community is astonishing. Despite the rate of acquired infections. Although data fi*om the center much early work on adult starvation,^®"'^ there have been vir are still being analyzed, the average body mass index (BMI, tually no further descriptions of the forms of adult malnutrition ie, weight[kg]/height[m^]) was approximately 13, significant or protocols for treatment since the 1950s. Although weight ly below the 16 that has been proposed as the indicator of the deficits of 40% are common in famines,'® the most recent BMI most severe grade of adult malnutrition.^’® Other adult pa scale for adult malnutrition defines the category of g rea test risk tients survived with BMIs below 10, refuting the currently as a BMI below 16, equivalent to a weight deficit of only 20%.^’’ accepted view that a “BMI below 12 is the absolute lower This is substantially higher than the BMI of 14 used as a dis limit compatible with life.”® charge criterion at Bardalli. There are no recommended thresh Given the number of lives saved (413 of 596 patients for olds for screening techniques (such as mid upper-arm circum whom outcome is now known) by a comparatively small amount ferences) that can be used for anybody older than age 10 years, of input, we believe adult therapeutic units should be included and adolescents are ignored in all scales. more frequently in famine relief programs. Currently, pro This lack of attention led to many deaths in Somalia. A viders sponsored by Concern Worldwide are treating many reconsideration of the importance of adult therapeutic care in chronically malnourished adults and adolescents in a thera emergency feeding programs and research into the problems peutic center in Ayod, South Sudan. At the Ayod center, the encountered therein is urgently needed. Hopefully the work average BMI of the first 100 inpatients was below 12, and at Bardalli will rekindle interest in this neglected sphere of adults still walking with BMIs as low as 9.2 have been ad human suffering. When data analysis is complete, it should be mitted. possible to provide adult anthropometric scales of risk that The treatment of adult malnutrition is not the same as that are applicable to famine situations, together with basic pro of children because of differing nutritional requirements and tocols for treating the different forms of adult malnutrition. the practical problems associated with care. For example, However, treating severe adult emaciation is a difficult task, anorexia, although common in both children and adults, pre and there remains much work to be done. sented far more of a problem among the adults treated in The famine in Somalia, although extremely severe, was not Baidoa. People were suspicious of nasogastric tubes, which qualitatively different fi"om other famines, such as Ethiopia they would generally not allow to remain in place for more in 1984 or the current situation in South Sudan. In all major than a few hours. Whereas with small children it was practical famines, adults die as well, often leaving orphans requiring to introduce a small bore nasogastric tube several times a day long-term care behind them. Many of these people can be even if the patient resisted. The adults would not tolerate saved with some redirection of resources toward adult ther multiple insertions. apeutic units. In saving adults and adolescents, we are saving The precise cause of anorexia in malnutrition is unknown; those who grow the food, create the wealth, and care for the however, it would appear that a combination of several fac children. tors is involved. Infections (eg, amebiasis and shigella); par Steve Collins, M B B S , BSc asitism (eg, giardiasis and small bowel bacterial overgrowth); Concern Worldwide and deficiencies of specific nutrients, such as zinc and mag London, England nesium, with the resulting surfeit of other nutrients that 1. Lipscomb F. Medical aspects of Belsen concentration camp. Lancet. 1945;1:3I3- 315. cannot be adequately metabolized or excreted are all likely to 2. CDC population-based mortality assessment: Baidoa and Afgoi, Somalia, 1992. play a role.^ Depression, a very common presenting com MMWR Morh Mortal Wkly Rep. 1992;41:913-917. 3. Moore P , Marfin A, Quenemoen L, et al. Mortality rates in displaced and resident plaint, is also almost always a factor. populations of central Somalia during the 1992 famine. Lancet. 1993;341:935-938. Other problems seen in these malnourished adults included 4. Ferro-Luzzi A, Sette S, Franklin M, James W. A simplified approach to a s s e s s i n g adult chronic energy deficiency. Eur J Clin Nutr. 1992;46:173-186. lower-limb wasting and contractures, caused by months of 5. Shetty P, James W. Body Mass Lndex: An Objective Measure for theE s tim a tm squatting. These often required extensive physiotherapy, de of Chronic Energy Deficiency in Adults. Geneva, Switzerland: World Health Orga nization. In press. laying the discharge of otherwise well people. 6. James WIT, Ferro-Luzzi A, Waterlow J. Definition of chronic energy malnutri Nutritional edema was twice as common in adults as in tion in adults. Eur J Clin Nutr. 1988;42:969-981. 7. Golden MHN. The nature of nutritional deficiency in relation to growth failure and children, present to some degree in 30% of the admissions. poverty. Acta Paediatr Scand. 1991;374(suppl):95-110. 8. Golden B. Primary Protein-Energy Malnutrition: Human Nutrition and Dietet This regularly coexisted with ascites (not usually found in ics. New York, NY: Churchill Livingstone Inc. In press. children with kwashiorkor) and was associated with a poorer 9. Drummond J. Starvation in western Holland. Lancet. 1945;1:282-283. 10. Beattie J, Herbert P, Bell D. Famine oedema. Br J Nutr. 1948;2:47-65, prognosis. In the early stages of recovery these edematous 11. Keys A, Brozek J, Henschel A, Mickelsen 0, Taylor H. The Biology of Humm people needed a diet with a relatively low protein to energy Starvation, L and LL. Minneapolis: University of Minnesota Press; 1950. 12. McCance R. Studies on Undemutrition, Wuppertal, 19L6-1H9. London. En ratio of 5 to 1 g/MJ. Interestingly, although the diet used was gland: Her Majesty’s Stationary Office; 1951. Medical Research Council special re originally designed for severely malnourished children,® the port series 275. 13. Gopalan C, Krishnaswamy K. Famine oedema. Prog Food Nutr Sci 1975;1:207- protein content proved to be far more critical in the adults. 224.

6 3 8 JAMA, August 4, 1993—Vol 270, No. 5 Letter From

Printed and Published in the United States of America r /

ASSESSMENT OF NUTRITIONAL STATUS IN EMERGENCY-AFFECTED POPULATIONS

July 2000

Steve Collins, Ai-abella PuffieM & Mai-k Myatt This report is issueci on the general responsibility o f the Secretariat o f the UN ACC/Sub-Comm ittee on Nutrition; the material it contains shoul4 not be regarded as necessarily endorsed by, or reflecting the official position o f the ACC/SCN and its UN member agencies. The designations employed and the presentation o f material in this publication do not imply the expression o f any opinion whatsoever on the part o f the ACC/SCN o f its UN member agencies concerning the legal status o f any country, territory, city or area o f its authorities, or concerning the delimitation o f its frontiers or boundaries.

This raport was written by 6teva Collins, Valid International Arabella Duffield, ACC/5CN Secretariat and Mark Myatt, Institute of Ophthalmology, London

We are grateful to the following reviewers for their comments:

Monika Blosnner (WHO), Tim Cole (institute o f Child Health), Mary Corbett (Concern), Mercedes de Onis (WHO), Mike Golden (University o f Aberdeen), Saskia van der Kam (Médecins Sans Frontières), Carlos Navarro-Colorado (Action Contre La Faim), Nick Norgan (University of Loughborough), Claudine Prudhon (Action Contre La Faim), Simon Strickland (London School of Hygiene and Tropical Medicine), Stanley Ulijaszek (Oxford University).

if you would like to request any o f the ACC/Sub-Committee on Nutrition's publications please contact:

ACC/Sub-Committee on Nutrition, 20, avenue Appia, 1211 Geneva 27, SWITZERLAND Tel: +(41 2 2 ) 791 0 4 56, Fax: +(41 2 2 ) 798 88 91, E-mail: [email protected], Web; http://acc.unsystem.org/scn/

if you have any encjuities about this supplement please contact either the ACC/SCN on

Steve Collins, Valid International, Oleuffynon, Old Hall, Llanidloes, Powys, SY18 6PJ, UK E-mail: [email protected]

Funding support is gratefully acknowledged from Cl DA, PFA (Ireland), NORAP, DFIP (UK), UNHCR, UNICEF, WFP and Valid International. This report was made possible through the support provided by the Food and Nutrition Assistance (FA Nta) Project by the Office o f Health and Nutrition Bureau for Global Programs, Field Support and Research at the U.S. Agency for International Development, under the terms o f Cooperative Agreement No. HNE-A -Q O -98-00046-00 awarded to the Academy for Educational Development (AED). The opinions expressed herein are those ofthe authors and do not necessarily reflect the views o f the U.S. Agency for International Development. Sum m ary

This article describes simple techniques suitable inappropriate for this purpose as it is affected by For the assessment o f the nutritional status o f oedema and body shape and difficult to measure. adults aged 20-60 years in emergency-affected In any particular situation, workers should only populations. The BMI (Body Mass Index), MUAC use these suggested criteria as a starting point and (Mi<\-uppe)r Arm Circumference), and clinical adapt them to situation-specific factors. models are assessed for their usefulness in determining the prevalence of chronic Admission criteria into adult therapeutic feeding undemutrition in adults at the population level, centres should be based upon the following cut- and also for screening severely undernourished off^i adults for entrance to feeding clinics. □ MUAC < 160 mm irrespective o f clinical No consensus on a definitive method to assess signs adult undemutrition has been reached; more □ MUAC 161-185 mm plus one of the research is required to do this. This article makes following: only preliminary recommendations. -Bilateral pitting oedema (Beattie grade 3 or worse) Su r v eys A N D p o p u iv \ t io n le v e l assessments o f -Inability to stand CHRONIC UNDERNUTRITION -Apparent dehydration

The BMI may be used to estimate the prevalence □ Famine oedema (Beattie grade 3 or o f chronic undemutrition in a population survey, worse) alone as assessed by a clinician to using the classification system shown below. exclude other causes.

The classification o f categories o f chronic Additional social factors can be included in the undernutrition using the BMI model. The relative weighting of these must be BMI Ckssific^iion ofchtonic determined locally; for example whether you ancfctwdght categories (k g nrr^) need one, two or three additional social factors to Normal >=18,5 tip the balance in favour o f therapeutic rather Grade 1 17.0-18.4 than supplementary care. Relevant social fstctojrs Grade II 16.0-16.9, Grade III <=15.9 could include the following;

The BMI is known to vary with age and body □ Access to food (quantity and quality) shape. In order to account for changes in body a Distance from centres shape the Cormic Index (sitting height/standing a Presence /absence o f carers height) must be taken into account, and a shelter standardised for, when comparing the BMI of a Dependants different populations. a Cooking utensils

ML/AC may also be used to assess the prevalence Admission to adult supplementary feeding o f chronic undemutrition at the population level. centres should be based upon the following cut o ff Sc r e e n in g sever ely undernourished a d u l t s

No single definition or classification of acute MUAC 161-185 mm and no relevant signs or few adult undemutrition has been universally relevant social criteria. accepted, but interim techniques may be recommended until further research clarifies In any particular situation, workers should take criteria. We suggest that the ML/AC in these suggested standards as the starting point combination with clinical signs should be used to and adapt them according to situation-specific screen adult entrance into feeding centres, using factors. the classification shown below. The BMI is Signalling the neccj fot- nom s i^n4 st^nct^tcjs. The 5CN will identify for the attention oftechnical agencies or other bodies critical a^eas where norms and stands Ms ai^e missing or out-of-d ^ te and holding programmes back. This includes (especially) identifying knowledge gaps and significant areas in dispute or controversy; as well as identifying areas requiring operational research, and facilitating this work.

Extract from the ACC/SCN's Strategic Plan, April 2 0 0 0 CONTENTS

Introduction ...... 1 Acute and chronic undernutrition ...... 1 Measuring nutritional status in adults ...... 2 Weight 2 Body Mass Index ...... 2 Theoretical problems with BMI...... 3 Practical problems with BMI...... 5 Summary — the use o f BMI ...... 8 Mid-Upper Arm Circumference ...... 8 Theoretical problems with MUAC...... 9 Practical problems with M UAC...... 9 Summary — the use of MUAC ...... 9 Combination measures...... 10 Clinical signs for screening acute undernutrition ...... 10 Famine oedema ...... 10 Clinical m od els...... 11 Activities o f daily living ...... 12 Summary — the use o f clinical models ...... 12 Primary and secondary undemutrition ...... 12 General summary — suggested indicators for assessing adult undernutrition in the field...... 13 w here do we go from here^ Future research n eed s ...... 14 References 16 Appendix 1: Basic defi'ntions...... 19

List o f Tables, Figures and Boxes Box 1 The correction o f BMI using the Cormic Index ...... 4 Table 1 The classification o f categories o f uri(\ernub\i\on...... '...... 3 Table 2 Incidence o f famine o ed em a ...... 6 Table 3 Suggested MUAC cu t-off points for moderate and severe adult undernutrition oedema...... 10 Table 4 Classification o f famine oedema based on the Beattie classification ...... 11 Table 5 The CHANCES screening model for acute adult undemutrition during fam ine ...... 11 Table 6 How BMI, MUAC, and CHANCES meet the criteria o f an ideal index o f nutritional status in famines ...... 13 Table 7 The applicability o f BMI, MUIAC, Weight, and CHANCES in different situations ...... 14 Figure 1 The effect o f varying sitting height/stature ratio CSH/S)on BMI ...... 3 Figure 2 Pitting oedema and the odds o f mortality ...... 7 Figure 3 The odds o f mortality below different thresholds o f admission BMI o f oedematous and marasmic patients admitted to a therapeutic centre mortality ...... 7 Assessment o f Nutritional Status o f Adults in Emergency-Affected Popuktions

INTRODUCTION complex task and is not described in detail here 2. In addition, the article does not consider the assessment of obesity, micronutrient deficiencies This supplement has been produced in response to or pregnancy. the increasing number of reports on adult nutrition surveys received by the RNIS and the concomitant Appendix one provides some basic definitions of interest in the subject shown by the readership. We terms that are employed in the discussion of adult hope that it will help the readers of the RNIS to nutritional assessment. interpret the results of the adult nutritional surveys reported. ACl/TE AND CHRONIC VNDERNUTRITION

This article describes simple techniques suitable There are two main patterns of undemutrition found in children. These are stunting and wasting . 43 for the assessment of adult nutritional status in emergency-affected populations. We do not intend Different processes produce these two patterns and they are assessed using separate it as a comprehensive review of all aspects of anthropometric indices. In children, acute assessing adult nutritional status, but as a guide to nutritional deficit and/or disease (such as techniques useful in the field. diarrhoea) produce wasting, characterised by a reduction in weight-for-height or arm During famine-relief operations workers are circumference, or both. Prolonged nutritional deficit increasingly recognising and treating severe adult and/or disease result in stunting, characterised by undemutrition. There have, however, been few a reduction in height-for-age ^ Wasting and studies to investigate the problems associated with stunting are associated with different functional screening and treating severely undernourished consequences. Weight-for-height is a powerful adults during famine and consequently, little predictor of short-term mortality, as is the mid guidance is available to field workers. There is at upper arm circumference (MUAC). Height-for-age present, no universally accepted definition or predicts longer-term mortality classification of acute adult undemutrition and no specific treatment guidelines for the condition. Thus the screening and selection of admissions into The nutritional assessment of adults is more problematic. Despite metabolic differences therapeutic feeding centres and the dietary between chronic and acute undemutrition 12.13. treatment of those admitted becomes problematic. Since 1992 however, there have been several the absence of linear growth removes the power of a height variable to discriminate between the two advances made in these areas and this article main pattems of undemutrition. In 1988, the attempts to pull some of these together. We also Intemational Dietary Energy Consultative Group make some recommendations as to the techniques suitable for the assessment of adult nutritional proposed a definition of chronic adult undemutrition calling it 'chronic energy deficiency’ (CED), clearly status under different circumstances. differentiating it from ‘acute energy deficiency’ All of the indicators described in this article attempt (AED). to assess adult nutritional status. It is important to realise that, to date, no consensus on a definitive Chronic Energy Deficiency CCED) was method has been reached; more research is 4efine4 as: required to achieve this. None of the anthropometric indices described below can be "X steady st^fG ^t which ^ person is in considered to be a gold standard, although body ^n energy b^knce although ^t cost mass index (BMI) has often, mistakenly, been either in terms o f increased risk to treated as if it were. health or ^s «?/? impairment o f This supplement focuses on the nutritional functions 3nd health" assessment of adults between twenty and sixty years of age. The assessment of older persons is a RNIS Supplement — July2 0 0 0

Acute Energy Deficiency CAED) was 4efinecl MEASURING NUTRITIONAL 35; STATUS IN ADULTS

"A st^te o f negative enei-gy b^knce. An ideal index of nutritional status, for any age I.e. ptogtessive loss ofhocfy enetgy*^* group, should meet the following requirements:

The differentiation of acute and chronic adult □ it must be correlated with body fat and undemutrition is important because the two protein stores. conditions entail different adaptations and have □ It must be correlated with health or different functional consequences. For example, functional outcomes. habituation to CED has been described in Indian □ It must be simple to obtain and interpret in labourers who, with a mean BMI of 16.6kg m-^, the field. It must also be accurate (close to were able to function reasonably normally By the true value), valid (represent what it is contrast, similar BMIs, rapidly induced in 32 thought to represent) and precise previously well-nourished volunteers, resulted in (repeatable). extremely poor physical and psychological states^^. Differentiating between these two types of In addition to these requirements, as adult height is undemutrition may be difficult with a one-off largely determined by an individual's genotype and measurement. They can be distinguished using a childhood nutritional experience it follows that if series of measurements taken over'time, but in an index is to reflect current nutritional status in practice, this option is often not available. As acute adults it must be independent of height. undemutrition wastes peripheral body tissues faster than central tissues, it may be possible to compare two different body measurements in order WEIGHT to differentiate between these two forms of The use of weight alone to assess nutritional status undemutrition. It may also be that adaptation below should be limited to monitoring purposes because it a certain threshold for each measurement is is confounded by height. Weight is appropriate for impossible and hence those falling below that monitoring the progress of patients suffering from threshold must have acute undemutrition. long-term morbidity, recovering from disease or However, at present there are few data available surgery, or during nutritional rehabilitation within a with which to examine these problems. therapeutic feeding centre.

We feel that the term “energy deficiency" is unhelpful when applied to undemutrition because it BODY MASS INDEX obscures the importance of protein catabolism, deficiencies of vitamins and minerals. For this The body mass index (BMI) is calculated from reason, we prefer the term “undemutrition" rather weight and height measurements using the formula than “energy deficiency". BMI = weight (in kg) divided by height (in m ^). The BMI was first introduced by Quetlet in order to Primary undemutrition develops when nutrient eliminate the confounding effects of height on intake is insufficient to provide for normal weight. In normal adults, the ratio of the weight to physiological needs. In adults, primary the square of height is roughly constant, and a undemutrition is invariably due to a lack of food. person with a low BMI is underweight for their Secondary undemutrition occurs when an height BMI reflects protein and fat reserves, underlying disease process (for example, which in turn reflect functional reserves including HIV/AIDS, TB and cancer) increases metabolic the ability to survive nutritional deficit and some demands and/or decreases food intake or diseases. utilisation. The treatment of primary and secondary undemutrition may be quite different. BMI may be appropriate for population-level assessments of chronic undemutrition. In 1988, researchers proposed the use of BMI to define and diagnose chronic undemutrition^^. This AnthtopometHc Assessment ofthe Nutrition Status ofAcjults in Emetgency-AfectecJ Popuktions classification provides a useful framework for the However, there are several difficulties associated analysis of height and weight data from chronically with the use of BMI as an anthropometric index. undernourished adult populations. These difficulties can broadly be separated into theoretical and practical problems. Table 1 The classification o f categories Theoretical Problems with BMI o f chronic unclernutrltion Classification o f BMi Bocly sh^pe - Many factors other than nutritional uncjenveight categories (kg m~^) status determine BMI. Most important of these is Normal >=18.5 body shape, in particular the ratio of leg-length to Grade 1 17.0-18.4 trunk-length, sometimes called the sitting-height to Grade 11 16.0-16.9 standing height ratio (SH/S) or Cormic index. This index varies both between populations and within Grade III <=15.9 populations. These differences result in world-wide ( ici^pted from Ferro-Luzzi et

Figure 1 The effect o f varying sitting height: stature ratio CSH/S) on BMI for a 70 kg, initially 1.75 metre male Cfrom Norgan 1994^*)

2 3 -

2 2 -

2 0 -

0 .5 5 0 .5 6

Sitting Height to Standing Ratio RNIS Supplement — July2 0 0 0

There is also considerable within-population variation in the Cormic Index. In one aboriginal population, the within-population range for the BOX1 Cormic Index was 0.41 - 0.54 23. This is greater than the worid-wide variation in Cormic Index 22 THE CORRECTION OF BMI USING and equivalent to more than 10 kg m-2 variation in THE C orm ic Inpex CSH/S) BMI. dependant upon shape alone.

Sitting height can be measured by sitting the In order to standardise BMI to take into account person on a straight-backed chair with a height changes in SH/S ratio we recommending using the board strapped to the back. This measurement is equations below to calculate BMI standardised to then used to correct BMI by applying a correction the actual SH/S ratio for the population under study. factor based on a linear regression model (see box Male subjects • BMI = 0.78(SH/S)-18.43 1) 23. Female subjects - BMI = 1.19(SH/S)-40.34 Comparisons of BMI between different populations Note: SH/S ratios should be expressed as a can be made using a correction factor based upon percentage the mean Cormic Index for each population. Such corrections should always be made when BMI is The observed BMIs can then be standardised to a used to compare the nutritional status of different SH/S ratio of 0.52 by adding the differences populations. between the observed BMI and BMI standardised for the population SH/S ratio to a BMI standardised Follow-up surveys for the comparison of within- to 0.52 using the equation below: population data (for example, before and after an BMISstd~BMI o.s 2 + (BMIob-BMIts), intervention, surveillance by repeated surveys etc) do not require the Cormic Index correction. Where BMIstd= standardisedBMI, BMIo.52 = estimated BMI at SH/S of 0.52 If BMI is being used to assess an individual for BMlob = actual BMI undemutrition the estimation of the individual's BMIes= estimated BMI at actual SH/S Cormic Index should be used as a correction factor. Without this correction the sensitivity and Examples specificity of BMI as a screening indicator may be 1. A Male population “A” has a mean BMI of 18.5 low 22.24 During emergencies, especially at the kg m-2 and a mean SH/S ratio of 50%. The height of a famine relief program, when there are BMio.52= 0.78*52-18.43 = 22.13. The BMIes = large numbers of people competing for relatively 0.78*50-18.43 = 20.57. Therefore the BMIstu = scarce resources, there is almost never sufficient 22.13 + (18.5 - 20.57) = 20.06kg m -2 time or staff to perform this standardisation. We 2. A Female population "A" has a mean BMI of therefore, feel that BMI is inappropriate for this role. 17.0 kg m-2 and a mean SH/S ratio of 54%. The BMI o.52= 1.19 *52-40.34 = 23.92. The BMIes = Age - Adult body size, shape and composition 1.19*54-40.34 = 21.54. Therefore the BMIstu = 21.54 + (17.0 - 23.92) = 14.62 kg m-2 vary with age.2s> 26. 27. 28 . 29 . 30. 31, 32. 33.34 Adults tend to loose fat free mass (PPM) and increase fat mass (PM) with age These changes may alter the functional significance of BMI at different ages. These are ad-hoc modifications to the standard Some NGOs use different cut-off points for older cut-off points for the use of BMI to assess adults when admitting individuals to a feeding undemutrition. As yet there are no published data programme. For example, for adults aged 50y+ that support the use of distinct cut-off points for Action Contre La Paim (ACP) admits adults to different age groups. Many adults in the developing therapeutic feeding centres and supplementary world do not know their exact age and it may, feeding centres using the cut-offs of 15kg m-2 and therefore, be difficult to differentiate the diagnosis 16 k g m - 2 respectively, but admits those aged less of nutritional status according to age in an than 50 at 16kg m -2 and 17kg m-2. emergency situation. It may be useful, however, to Anthtopometnc Assessment ofthe Nuttition Status o f Adults in Emetgency-Affected Populations separate age groups when presenting the results of account the Somali long-legged phenotype an an adult nutritional survey using BMI in a non- important factor explaining the very low BMIs emergency setting, where age can be ascertained observed Thus the cut-off values they propose using instruments such as local event diaries which are probably too low. In our experience a BMI of 10 would seldom be available in an emergency kg m-2 after standardisation to a SH/S ratio of 0.52 situation. is probably not compatible with life. One of 13 kg m-2, probably represents a degree of emaciation The increasing prevalence of kyphosis and where peripheral stores have already been scoliosis with age further necessitates the use of exhausted with a corresponding increase in central proxies for height when assessing the nutritional catabolism. This level is therefore probably status of older adults^ inappropriately low to be used as a cut-off for admission into an adult therapeutic centre. Chtonic ^nd ^cute undernuttition- A great deal of research has focused on use of BMI for the P r a c t i c a l P r o b l e m s w i t h b m i assessment of chronic undemutrition in stable Difficulties in obtaining the component populations. This role is primarily that of prevalence me^sutes o f BMI duHng fkmine - The height estimation, providing information useful in planning and weight measurements required to assess BMI at a population level. This is a different role to that are often difficult to obtain during famine. Chair or of screening individuals who may be suffering from bed-scales are usually unavailable and thus acute undemutrition in order to regulate admissions patients must be able to stand in order to be to feeding centres. The common assumption in weighed. Usually, many of the most severely contemporary NGO field manuals and recent undernourished adults requiring admission to academic articles 37. 38 that BMI is also an therapeutic feeding centres cannot stand at all and appropriate indicator for screening during famine, BMI cannot be estimated. In addition, many studies has not been tested. have reported that gross weakness 42. 43^ fexor contractions 43, or scoliosis 44 are common. These BMI cut-off points for screening adult admissions to prevent many patients standing straight enough for feeding centres, extrapolated directly from CED, accurate height estimation. As height is a squared may be inappropriate. The cut-off point of 16 kg m^, term, these errors are magnified in BMI calculation. that indicates severe chronic undemutrition does not necessarily reflect the degree of acute If BMI is to be used during an emergency there is a undemutrition that requires specialised treatment need to obtain robust, reliable and precise scales 39.40. During a famine, there is intense competition that can withstand repeated measurements under for entry into feeding centres and it is important that dry, dusty and hot conditions. These may be screening indicators are specific, only selecting expensive. those who would die if not given specialised treatment. Difficulties in the calculation o f BMI - Even in non-famine situations the calculation of BMI and As adults are usually the primary caregivers and Cormic Index may be unfamiliar to field workers income eamers in a household, it is also important and therefore difficult. AGP have developed tables not to admit those who do not need therapeutic of weight-for-height that show BMI ranges (like treatment into a centre as this may have a negative those used for children) that may reduce this affect on the rest of the household. difficulty.

In 1996, Ferro-Luzzi and James ^ adjusted their theoretical estimation of the lowest BMI compatible Difficulties in obtaining the component with life down from 12 kg m ^ in order to account for me^sutes o f BMi in eldetly ^nd handicapped the extremely low BMIs being observed in Somalia p e o p le - As adults become older, spinal disease during the famine there in 1992. They created two (predominantly osteoarthritis and osteoporosis) new BMI cut-offs of <13 kg m-2 and <10 kg m-2, affects an increasing proportion of people. These denoting severe wasting and extreme wasting conditions affect the ability to stand straight and respectively. These values did not take into make the accurate measurement of height RNIS Supplement — July2 0 0 0 impossible. BMI, based on height cannot therefore the frequent co-existence of pitting oedema and be used in older adults Recognition of this ascites means that oedema fluid can often account problem has prompted research into the use of for over 10% of body weight. Famine oedema is proxy measures of height. Researchers have also associated with poor prognosis (see Figure 2). shown a good relationship between arm-span, Consequently, patients with severe famine oedema demi-span, femur length, knee height and height often have a poorer prognosis the higher their 47. 48. 49. 50 These proxies are converted to admission BMI, the opposite of the situation in estimates of height using correction factors derived marasmic patients (see Figure 3). BMI is therefore, from regression equations. As the relationship not an appropriate indicator for people suffering between proxies and height has been shown to from famine oedema. This may be corrected by vary between ethnic groups and by age, different using a modified screening criteria (i.e. BMI below correction factors should be applied to different a cut-off point OR the presence of oedema). populations Suitable population-specific However, as the presence of oedema, particularly correction factors to apply to proxy measures of in older adults, may not always be indicative of height are usually unavailable in emergencies. undemutrition, it will be necessary to train field workers to differentiate between the causes of In elderly individuals, there are no viable oedema in adults. Alternatively, adults presenting alternatives to estimating height from arm span or with oedema will have to be referred to a clinician demi-span. It should be recognised, however, that who is able to differentiate between the types of at the individual level, there is significant error oedema. involved in the estimation of height using correction factors based on population means. For example, As the prevalence of famine oedema is frequently the standard error of the estimate of height from high during emergencies (see Table 2), the inability arm span is reported to be between 2.5 and of BMI to assess oedematous adults limits the 3.8cm^. The squaring of the height element in usefulness of BMI as a screening tool to assess calculating BMI magnifies these differences. acute adult undemutrition.

Famine oedema - Adult nutritional oedema is common during famine and its presence increases weight, producing an upward bias in BMI. In adults

Table 2 Indcjence o f famine oecjema

Study Location Ye^r Prevalence o f famine oedema Collins Somalia 1992 24.% Collins Angola 1995 90% Collins Sucian 1995 0% Zimmer et gl. France 194.2 50% Mollinson Germany 194.5 typical finding McCanceetal. Germany 194.6 common Debray France 194.5 100% Keys etal.* U.S.A. 194.5 87%

in seiectej population ofpteviously well-nouHsbecl Ametic^n volunteers st^rvecj uncjer expérimental conditions Anthropométrie Assessmeni o fthe Nutrition Status ofAcju/ts in Emergency-Afectecj Populations

Figure 2 Pitting oedema the oclds of mortality, b^secl upon the grac|ing system

clescribecl in Tqble 4 adaptedfrom Coiims

# rouilel

Grade of Oedema

Figure 3 The ocjcjs o f mortality below (different thresholcjs o f admission BMI for oedematous CN=75) and marasmic CN=218) patients admitted to a therapeutic centre in Baidoa, Somali^ during 1992/3 c^d^pted ftomcoiiins w s p

Body Mass Index (kg m at Presentation R.NI5 Supplement — July2 0 0 0

Sum m ary - th e use o f BMI levels of cut-off points used, the efficiency of a two- In our opinion, the many problems with the use of phase screening process and poor reproducibility BMI for screening acutely undernourished adults in the measurement continue 52, ea. 64, 65, 66 admissions to feeding programmes during famine Consequently, some humanitarian relief agencies relief programmes make the indicator inappropriate remain sceptical about the use of MUAC in for this role. emergencies 57.

BMI combined with an assessment of the At present during emergencies, MUAC is only prevalence of famine oedema is an appropriate recommended for use with children between one indicator for popuiation-level assessment of chronic and five years of age 58. 37, 30 \{ js, however, undemutrition. These data can be categorised increasingly being used to assess adult according to the classification given in Table 1 undemutrition during famine 59, 70, 71 above. Such surveys should also assess MUAC. Measurements of adult MUAC have long been known to reflect changes in adult body weight ^2, If BMI survey data are used to compare BMIs and the major determinants of MUAC, arm muscle between populations, estimates should be and sub-cutaneous fat, are both important corrected by standardisation to a SH/S ratio of 0.52 determinants of survival in starvation 73. 74,42 As using the mean SH/S ratio for the specific MUAC is less affected than BMI by the localised populations being studied. accumulation of excess fluid (pedal oedema, periorbital oedema, ascites) common in famine, it is Inside feeding centres it is useful to assess a likely to prove to be a more sensitive index of standardised BMI on each patient admitted tissue atrophy than low body weight. It is also (standardised using individualised SH/S ratios). relatively independent of height 7°.

The use of MUAC has not been evaluated as a /VIID-iypPERARM prognostic indicator. However, estimates of arm muscle area (AMA) or corrected arm muscle area CIRCb/MFERENCE CMUAC) (CAMA), corrected for humerus cross-sectional area, have been incorporated into diagnostic MUAC is the circumference of the left upper arm, schemes for adult undemutrition in hospitals 75 and measured at the mid-point between the tip of the used as prognostic indicators in the elderly and in shoulder and the tip of the elbow (olecranon cancer patients 76.77_ However, it is unlikeiy that process and the acromium). In children, MUAC is CAMA or AMA will be of use in emergency useful for the assessment of nutritional status ^2. assessments as both require accurate measures of 53,54.55 56 |j jg good at predicting mortality and in skin-fold thickness that would be hard to obtain some studies, MUAC alone % 59 or MUAC for given the rush and pressure of an emergency age predicted death in children better than any operation. other anthropometric indicator. This advantage of MUAC was greatest when the period of follow-up Ferro-Luzzi and James ^ have proposed MUAC was short cut-off points for use in screening acute adult undemutrition. They base these on extrapolation The MUAC measurement requires little equipment from more normally nourished populations in and is easy to perform even on the most debilitated developing countries, without reference to data individuals. Although it is important to give workers from acutely undemourished adults during famine. training in how to take the measurement in order to Although there is some evidence that the reduce inter- and intra- observer error, the undemourished category may be associated with technique can be readily taught to minimally trained increased morbidity in chronically undemourished health workers It is thus potentially suited to populationsi5, we doubt whether the criteria screening admissions to feeding centres during proposed are appropriate for screening acutely emergencies. The use of MUAC in emergencies is, undemourished adults. however, still controversial, and disagreement over the preferential selection of younger children, the Data from famines suggest that the relationship between MUAC and BMI is not constant during Anthropométrie Assessment o f the Nutrition Status o f Adults in Emergency-Affected Populations acute undemutrition and that an accelerated loss of Practical Problems w ith m i/a c peripheral tissue during acute undemutrition has a Measurement error - In children, the use of relatively greater depressing effect on MUAC than MUAC is associated with two problems; the upon BMI These data also suggest that during preferential selection of younger children as acute undemutrition the differences in MUAC undemourished ^ and a lack of reproducibility between men and women become less in MUAC measurements Problems with the pronounced, a finding supported by previous reproducibility of MUAC measurements are observations in more normally nourished potentially a more serious obstacle to the use of population MUAC in adults. As in children, both inter- and intra- observer errors in MUAC measurements may It is likely, therefore, that in populations suffering occur. The importance of these errors needs to be from famine, MUAC cut-off points denoting investigated, but it is likely that the larger moderate to severe undemutrition should be dimension of the adult arm will reduce the relative adjusted. Values of 185 mm denoting moderate importance of such errors. The development of undemutrition and 160 mm denoting severe colour-banded numeric MUAC bands reflecting undemutrition in both sexes have been proposed threshold values of MUAC with a change of colour and used in famines Given that there are would further reduce these problems by removing different cross-sectional humerus bone areas in numerical errors. Given the ease with which MUAC men versus women it is unclear whether measurements can be performed it would be common cut-off points for both sexes will prove feasible to refer any patients found to have a appropriate. MUAC within a few millimetres on either side of the threshold (designated on the band as a different Theoretical Problems w ith MUAC coloured zone) to a more experienced worker for Uck o f d^t^ upon which to decide useful verification or for further (e.g. clinical) assessment cut-off points - There are insufficient data as part of a two-stage screening process. The available correlating MUAC with mortality and other width of this zone should be based upon a more functional measures in adults. Cut-off points based detailed examination of errors in the evaluation of on risk of mortality cannot, therefore, be presented adult MUAC by minimally trained workers. Colour- with any degree of certainty. There exists a need banded MUAC measurement straps are already in for more field studies during emergency famine- use with children. relief operations to evaluate the power of MUAC to predict adult mortality in different famine affected The assessment of adult nutritional status using populations. MUAC requires no equipment apart from a tape measure. As the index is the actual measurement A g e - The use of MUAC in adults may be affected itself, mathematical manipulation of the by the redistribution of subcutaneous fat towards measurement obtained is not necessary. The ease central areas of the body during ageing In with which MUAC can be assessed make it older children and adolescents, the rapidly suitable for nutritional screening during the height changing pattems of skeletal muscle and of an emergency where time and skilled personnel subcutaneous fat are also likely to be a problem. are at a premium. Age specific MUAC cut-off points may be required for older children, adolescents, and the elderly. Sl/MMARY - THE USE OF MIVAC In our opinion, MUAC is an appropriate indicator for E th n icity - Ethnic differences in MUAC have not the assessment of acute adult undemutrition. The been sufficiently studied to determine whether a indicator is useful for both screening acute adult single cut-off point for MUAC could be used for all undemutrition and for estimating prevalence of ethnic groups. undemutrition at a population level. We suggest, that until more data are available, the following cut off points are used for both sexes for screening adult admissions to feeding centres:* K RNIS Supplément — Ju ly2 0 0 0 Table 3 Suggesteci MUAC cut-ofFpoints people. This is different from the situation in for moderate and severe acute industrialised countries where exhaustion of fat or fat-free mass is more often the terminal event ^3. adult undernutrition Therefore, the clinical signs of infection or metabolic dysfunction are likely to be useful Level ofundetnutrition MUAC Cmm) prognostic indicators. This possibility has been Mocfergte <185 investigated in children but rarely in adults 88. Severe <160 Although reported as effective in identifying children at a high risk of mortality 8®, these models At present there is insufficient data to assess the have been criticised because the interactions usefulness of MUAC as a tool with which to monitor between the features used, such as oedema and treatment in adult feeding centres. More data is hypoprotinaemia, were not taken into account required to assess this role. FAMINE OEDEMA In both children and adults, famine oedema has COMBINATION MEASURES long been recognised as an important sign relating Some authors have proposed the combined use of to the severity of undemutrition (see above). In BMI and MUAC, suggesting that such a adults, famine oedema is common (see Table 2) classification might prove better able to and usually (but not always) related to a poor discriminate between at-risk individuals and those prognosis (see Figure 2) 89. 39,92 For this reason who are thin but not at risk Given the relative the presence of famine oedema is usually used as strengths of the two indicators, it is probable that an indicator of severe undemutrition. It is important this combination will only be appropriate for the to note, however, that the prognostic significance of population assessment of CED. However, we see adult famine oedema varies according to the little advantage of a combination measure over the context and in some occasions, the sign is of less use of BMI alone. use as an indicator of severity.

Famine oedema in adults should be diagnosed in a similar way to that in children, using firm pressure CLINICAL SIGNS FOR applied over a bony prominence for approximately SCREENING ACUTE 3 seconds and assessing whether an indentation remains after the pressure is removed. The UNDERNUTRITION severity of oedema should be graded using the system devised by Beattie during the Second For many years, it has been recognised that World War (see Table 4) ®9. in our experience undemutrition increases both susceptibility to, and pitting oedema of grade 3 and above are often the severity of, infection. Vitamin, mineral and other associated with a markedly worse prognosis dietary deficiencies, depressed cell mediated and particularly if they occur in male patients 8®. Lesser humeral immunity, gastric acidity, mucosal integrity grades of oedema rarely appear to be clinically and altered flora are all known to increase relevant. susceptibility to infection and the ability of an individual to utilise their energy and protein It is important to note that oedema in adults may be reserves 84 The situation in an emergency is induced by reasons other than undemutrition usually made worse by the breakdown in public including cardiac, vascular, renal and hepatic health infrastructure and the congregation of disease. Differentiating between nutritional oedema displaced people in crowded and unhygienic and oedema secondary to other causes can be conditions This combination of poor public difficult and usually requires clinical expertise. health environment and immunosuppression means that in famine it is usually infection combined with metabolic dysfunction rather than absolute loss of fat or fat-free mass that kills

fO Anthropométrie Assessment ofthe Nutrition Status o f Adults in Emergency-Affected Popuktions

Table 4 Classification o f famine oecjema basecj on the Beattie classification ^

Cr^de Extent o f oedema 0 absent 1 minimal oedema on the foot or ankle that was demonstrable but not obvious 2 obvious oedema on foot or ankle 3 oedema demonstrable up to knee 4 oedema demonstrable up to inguinal ligament 5 total body oedema (anasarca)______

Features o f ^mine oedema not included in the Beattie ckssifcation ______Ascites in isolation probably not a useful indicator o f the severity o f primary undernutrition. Prognosis relates to extent o f accompanying oedema- Often occurs in secondary to disease (especially TB) Peri-orbital in isolation does not appear to reflect a poor prognosis Scrotal probably the result o f ascitic fluid tracking downwards under the influence o f gravity, in isolation does not appear to reflect a poor prognosis hydroarthrosis significance unknown ______

Clinical models basic clinical model. This allows the model to In adults, until recently the use of clinical models to differentiate between those with clinical illness but assess nutritional status appears to have been restricted to the nutritional assessment of surgical no undemutrition, better treated in medical units, patients Since 1992, similar assessments have from those with both illness and undemutrition, been made amongst severely undemourished adult best treated in specialised feeding centres. A inpatients in several therapeutic feeding centre combination model, “The Concern Health and during different famines. A model using three Nutrition Evaluation Score” (CHANCES) has been clinical signs: apparent dehydration, oedema and used in Ajiep in South Sudan during 1998 (see inability to stand has proved useful in predicting Table 5). In addition to the basic signs prognosis among adult patients ^ 2. These three demonstrated in Table 5, additional relevant clinical signs were far better at predicting mortality criteria, in particularly social criteria such as than BMI, were easy to elicit and the model only presence of a carer or distance away from feeding involves counting. centres can also be added to the model.

To be useful in screening admissions to therapeutic Preliminary indications are that this model feeding centres during famine rather than performed well ^3. predicting prognosis in those already admitted, an indicator of nutritional status must be added to this

Table 5 The CHANCES screening mociel for acute adult undernutrition during famine

Category Relevant Action MUAC Cmm) clinical signs

Normal Do not admit >185 + /-

Moderate undemutrition Supplementary feeding 160 -185 - Severe undemutrition Therapeutic feeding 160 -185 + Severe undemutrition Therapeutic feeding <160 + /- ^ An ^dultpresenting with bil^ter^loedema (Beattiegr^de 3 or more), but not low MUAC, should be referred to 9 clinician in order to assess whether s/he h^s nutritional oedema, ifthe oedema is nutritional s/he should be admitted to the Therapeutic Feeding Centre. RN15 Supplement — Ju/y2 0 0 0

Additional social factors can be included in the ACTIVITIES OF PAILY LIVING model. The relative weighting of these; for example whether you need one, two or three additional Some workers have suggested that measuring social factors to tip the balance in favour of functional ability rather than anthropometry may therapeutic rather than supplementary care must provide a useful screening tool. Functional ability is be determined locally. Relevant social factors could usually measured using scores derived from the include the following: answers to a set of several related questions. Such an instrument should be able to differentiate □ Access to food (quantity and quality) between those who can and cannot care for □ Distance from centres themselves. Valid and reliable sets of questions □ Presence /absence of carers can, however, be time consuming to develop and □ Shelter test. Currently available instruments (e.g. those □ Dependants used for needs assessment in the elderly) may be □ Cooking utensils appropriate in emergency situations but still require field testing during famine and identification of valid and reliable cut-off points. The CHANCES model Admission to adult supplementary feeding centres creates a composite function/clinical score, with should be based upon the following cut-off: function being assessed solely by ability of an individual to stand. As in the CHANCES model, an indicator of nutritional status would need to be MUAC < 161-185 mm and no relevant signs or few included to allow the instrument to differentiate relevant social criteria. between those with clinical illness but no undemutrition, best treated in medical units, and In any particular situation, workers should take those with undemutrition, best treated in specialist these suggested standards as the starting point feeding centres. and adapt them according to situation-specific factors.

Su m m a r y - the use of clinical models It is important to note that the CHANCES model presented here screens adults inurgent need of In our opinion, the combination of MUAC and nutritional support. If in a particular situation, the clinical signs, based upon the CHANCES clinical needs are such that workers have to make the model, is the method of choice for screening CHANCES model more stringent in order to avoid acutely undemourished adult admissions into being overwhelmed by admissions it is essential feeding centres. that they call for assistance and additional resources. In such situations vigorous advocacy is Admission criteria into adult therapeutic feeding essential to publicise the extent of the crisis and centres should be based upon the following cut call for help. offs: PRIMARY A N D SECONDARY □ MUAC < 160 mm irrespective of clinical signs UNDERNl/TRITION □ MUAC < 161-185 mm plus one of the following: An important problem in assessing adult -Bilateral pitting oedema (Beattie undemutrition during famine is the inability to grade 3 or wome) differentiate between primary and secondary -Inability to stand undemutrition. At present, there are no one-off -Apparent dehydration measurements that can do this. In practice □ Famine oedema (Beattie grade 3 or admission into selective feeding programmes worse) alone as assessed by a clinician to should be based on the CHANCES criteria, exclude other causes. irrespective of whether the adult is suffering from primary or secondary undemutrition. Those with malnutrition secondary to infection (for example. 12 AnihropometHc Assessment ofthe Nuttition Status o f Adults in Emetgency-Affected Popuktions

TB or HIV) will fail to respond adequately to between primary and secondary undemutrition. To treatment. Adult selective feeding programmes date, no work has been undertaken to investigate must therefore be designed in a way that allows for this possibility. these people to be referred to other more appropriate support or treatment programmes. The G eneral sum m ary - Suggested indicators design of these is beyond the scope of this article. FOR ASSESSING ADULT UNDERNUTRlTlON IN THE FIELD In future, as catabolic hormones produced in Earlier in this report we defined the criteria of an response to infection have a greater influence on ideal index of nutritional status. Table 6 shows how peripheral energy and protein stores “ it might be BMI, MUAC, and CHANCES meet these criteria. that in secondary undemutrition resulting from infection the MUAC is more depressed than BMI. This might lead to the possibility of using a combination of the two indicators to differentiate

Table 6 How BMI, MUAC, and CHANCES meet the cnteria o f an Icjeai index o f nutritional status in famines

CHtem BMI MUACCHANCES

I n4epen4ent of height Yes Yes Yes

Correlgte4 with bo4y Yes Yes Yes energy stores Upwar4ly biase4 by oe4ema.

Correlate4 with health / Yes in chronic Yes functional outcomes ? acute Insufficient data.

Simple to obtain an4 N o Yes Yes interpret Measurements 4ifficult Proposes clinical signs in patients unable to are easy to elicit. stan4 9n4 those with musculo-skeletal problems.

Accurate ? Yes Yes Nee4s correction for bo4y shape using extra measurement o f SH/S an4 further arithmetic.

Vali4 Yes Yes Not in the presence o f oedema.

Precise Yes measurement errors in See MUAC. children. Insufficient data in adults.

13 RNIS Supplement — July2 0 0 0

numbers of afflicted people. These factors, In table 7 we present some interim combined with the armed conflict that has recommendations for techniques that may be used, characterised almost all famines during the last ten under different circumstances for the assessment years, make such broad-based population studies of adult nutritional status. These recommendations unfeasible. Consequently, no such studies have yet are preliminary and there is a need for further been performed for any nutritional indicator in research to clarify the criteria. It must also be noted either adults or children during the height of a that, in emergency relief programmes, the famine. Given these difficulties, relating indicators appropriate indicator cut-off point (screening level) to mortality in a selected feeding centre population is that which selects the number of individuals that may be all that is reasonably possible. The can be treated with the resources at hand In selection bias involved in such studies must be reality, such cut-off point values often cannot be acknowledged and results interpreted with caution. determined universally but must be tailored to suit the resources available in each particular situation Patterns ^n4 prevalence o f adult The choice of underlying body measurement undemutrition during farnlne - Data on the may also be determined by available equipment prevalence and pattems of adult undemutrition (e.g. scales may not be available to measure during famines may be useful in defining cut-offs. weight). Data on different ethnic groups and settings would be useful.

Table 7 The applicability o f BMI, Ml/AC, Weight, an4 CHANCES in different situations

BMI MUAC W eight " CHANCES Chronic undemutrition Acute undernutrition Monitoring nutritional rehabilitation progressive illness / post operative recovery)______

WHERE DO WE GO FROM Practicality o f measurements and HERE/ FITTVRE RESEARCH calculations - The practicality of obtaining various measures should be explored in field situations. NEEDS: Survey organisers should assess the ease of training survey workers in measuring MUAC, Defining functional cut-offs - Longitudinal weight, and height, as well as assessing inter- and studies should be undertaken to determine whether intra- observer variability in these measurements individuals falling below specific cut-off points for when measuring adults. BMI, and MUAC have elevated morbidity or mortality, poor pregnancy outcome, decreased Adjusting for differences In body shape - work ability or physical performance measures. Surveys undertaken in a variety of populations Such studies must be conducted in a variety of should explore the utility of adjusting indices using situations with different levels of undemutrition weight and height for differences in body shape by among adults. During emergencies, such studies using the SH/S Index or other indicators of body are problematic. Ideally studies relating indicators shape. to the risk of mortality in a general population are required to establish functional cut-off points that The use o f MUAC to monitor recovery could be used for the screening of adult admissions from undemutrition. The changes in MUAC to feeding centres. However, famines always and weight during recovery from undemutrition involve massive social upheaval and large should be compared with a view to establishing ~ü AnthtopometHc Assessment o f the Nuttition Status o f Adults in Emetgency-Affecte4 Populations whether MUAC is useful in this role and if it is, or signs that predict failure-to-thrive in feeding establishing relevant MUAC discharge criteria. centres.

Mote cf^t^ on cjiffetenti^ting between f^e aetiology, significance ^nd tte^tment o f second^ty ^nd ptim^ty undctnuftition - fknnine oedema ^nd ascites - More information Examination of adult failure-to-thrive in centres on the significance and prevalence of famine should include attempts to identify measurements oedema and ascites is required. It is recommended that surveys assess subjects for these symptoms.

15 R.N15 Supplement — July2 0 0 0

REFERENCES

1. ACC/SCN (1999) Nutrition and Healthy Ageing. SCA/ News 19. ACC/SCN, Geneva. 2. Ismail S, Manandhar M (1999) defter Nutrition for Older People: Assessment and Action. HelpAge International, London. 3. Waterlow JC, Buzina R, Keller W, Lane JM, Nichaman M2, Tanner JM. (1977) The presentation and use of height and weight data for comparing the nutritional status of groups of children under the age of 10 years. Bulletin of the World Health Organization 55(4):489-498. 4. Waterlow JC. (1972) Classification and definition of Protein Calorie malnutrition. British Medlcal Joumal 3:566- 569. 5. Beaton G, Kelly A, Kevany, Martorell R, Mason J. (1990) Appropriate uses of anthropometric indices in children. Nutrition Policy Discussion Paper. ACC/SCN, Geneva 6. Briend A, Hasan KhZ, Aziz KMA, Hoque BA, Henry FJ. (1989) Measuring change in nutritional status: a comparison of different anthropometric indices and the sample size required. European Joumal of Clinical Nutrition 43:769-778. 7. Bairagi R. (1981) On validity of some anthropometric indicators as predictors of mortality. American Joumal of ainlcal Nutrition 34:2592-2594. 8. Bairagi R, Chowdhury M, Kim Y, Curlin G. (1985) Alternative anthropometric indicators of mortality. American Joumal of Clinical Nutrition 42:296-306. 9. Smedman L, Sterky G, Meliander L, Wall S. (1987) Anthropometry and subsequent mortality in groups of children aged 6-59 months in Guinea-Bissau. American Joumal of Clinical Nutrition 46:369-373. 10. Chen LC, Chowdhury A, Huffman SL. (1980) Anthropometric assessment of energy-protein malnutrition and subsequent risk of mortality among preschool children. American Joumal of Clinical Nutrition 33:1836-1845. 11. Shetty PS. (1984) Adaptive changes in basal metabolic rate and lean body mass in chronic undemutrition. Human NutÀlon. Clinical Nutrition 38C:443-451. 12. Keys A. (1950) The biology of human starvation, led. Minnesota Press, Minnesota. 13. Waterlow JC. (1986) Metabolic adaptation to low intakes of energy and protein. Annual Review of Nutrition 6:495- 526. 14. James WPT, Ferro-Luzzi A, Waterlow JC. (1988) Definition of chronic energy deficiency in adults. Report of a working party of the International Dietary Energy Consultative Group. European Joumal of Clinical Nutrition 42(12):969-981. 15. Eveleth PB, Tanner JM. (1976) Worldwide variation In human growth, led. Cambridge University Press, Cambridge. 16. Quetelet LA. (1871) Anthropométrie ou mesure des différentes facultés de l'homme. C. Muqvariot, Brussels. 17. Ferro-Luzzi A, Sette S, Franklin M, James WPT. (1992) A simplified approach to assessing adult chronic energy deficiency.European Joumal of Clinical Nutrition 46:173-186. 18. Reddy V. (1991) Body mass index and mortality rates. Nutrition News, 12. National Institute of Nutrition, Hyderabad. 19. Duffieid A. (1998) Anthropometry, morbidity and mortality in rural savannah. PhD dissertation, London School of Hygiene and Tropical Medicine. 20. Prentice AM, Whitehead RG, Watkinson M, Lamb WH, Cole TJ. (1983) Prenatal dietary supplementation of African women and birth weight. Lancet 1:489-492. 21. Norgan NG. (1990) Body mass index and body energy stores in developing countries.European Joumal of Clinical Nutrition 44(SI):79-84. 22. Norgan NG. (1994) Interpretation of low body mass indices: Australian aborigines. American Joumal of Physical | Anthropology 94(2):229-237. | 23. Norgan NG. (1994) Body Mass Index and nutritional status: the effects of adjusting body mass index for the | relative sitting height on estimates of the prevalence of chronic energy deficiency, overweight and obesity. Aslan I and Pacific Studies (SuppI). | 24. Norgan NG. (1994) Relative sitting height and the interpretation of body mass index. Annals of Human Biology I 21(1):79-82. I 25. Forbes GB, Haiioran E. (1976) The adult decline in lean body mass. Human Biology 48(1 ):162-173. | 26. Stoudt HW, Damon A, Mcfariand R, Roberts J. (1965) Weight height and selected body dimensions of adults ^ (1960 - 62). Health Statistics Series 76-1074, US Government Printing Office, Washington. | 27. Borkan GA, Norris AH. (1977) Fat redistribution and changing body dimensions of the adult male. Human Biology | 49:495-514. { 28. Noppa H. Anderson M, Bengtsson C, Bruce A, Isaksson B. (1980) Longditudinal studies on anthropometric data i and body composition. The population study of women in Goteborg, Sweden.American Joumal of Clinical Nufrif/on 33:155-162. : fà Anthtopometnc Assessment o f the Nuttition Status ofAciults in Emetgency-Affecteci Popuktions

29. Wang J, Thornton JC, Russell M, Burastero S, Heymsfield SB, Pierson R. (1994) Asians have lower body mass index (BMI) but higher percent body fat than do whites: comparisons of anthropometric measurements. American Joumal of Clinical Nutrition 60:23-28. 30. Strickland S, Ulijaszek J. (1992) Body mass Index, ageing and differential reported morbidity in rural Sarawak. European Journal of Clinical Nutrition 47:9-19. 31. Rolland-Cachera MF, Cole TJ, Sempre M, Rossignol C, Charraud A. (1991) Body Mass Index variations: centiles from birth to 87 years. European Joumal of Clinical Nutrition 45:13-21. 32. Norgan NG. (1994) Population differences in body composition in relation to the body mass index. Review. European Joumal of Clinical Nutrition 48(s3):s10-s25. 33. Mazariegos M, Wang ZM, Gallagher D, Baumgartner RN, Allison DB, Wang J, Pierson RN Jr, Heymsfield SB. (1994) Differences between young and old females in the five levels of body composition and ttieir relevance to the two-compartment chemical model. Joumal of Geronfo/ogy 49(5):M201-208. 34. Gallagher D, Visser M, Sepulveda D, Pierson RN, Harris T, Heymsfield SB. (1996) How useful is body mass index for comparison of body fatness across age, sex, and ethnic groups? American Joumal of Epidemiology 143(3):228-239. 35. Forbes GB, Reina JC. (1970) Adult lean body mass declines with age: some longitudinal observations. Metabolism: Clinical and Experimental 19(9):653-663. 36. Ferro-Luzzi A, James WPT. (1996) Adult malnutrition: simple assessment techniques for use in emergencies. British Joumal of Nutrition 75:3-10. 37. Young H. (1992) Food scarcity and famine: assessment and response. Oxfam, Oxford. 38. Boelaert M, Davis A, Le Lin B, Michelet M, Ritmeijer K, Van Der Kam S et al. (1995) Nutrition guidelines, led. Paris: Médecins Sans Frontières, Paris. 39. Collins S. (1995) The limit of human adaptation to starvation. Wafum Medicine 1(8):810-814. 40. Collins S. (1993) The need for adult therapeutic care in emergency feeding programmes. The Joumal of the American Medical Association 270(5):637-638. 41. Hiernaux J. La diversité humaine en Afrique subsaharienne. Etudes ethnologiques 2000. 42. Winick M. (1979) Hunger Disease. Wiley-Interscience, New York. 43. Mollinson PL. (1946) Observation on cases of starvation in Belsen. British Medical Joumal Jan 5:4-8. 44. Zimmer R, Weil J, Dubois M. (1944) The nutritional situation in the camps of the unoccupied zone of France in 1941 and 1942 and its consequences. New England Joumal of Medicine 230:303-314. 45. Todorovic V. (1998) Performing some simple anthropometric measurements on residents in nursing homes. 46. Steele MF, Chenier TC. (1990) Arm-span, height, and age in black and white women. Annals of Human Biology 17(6):533-541. 47. Steele MF, Mattox JW. (1987) Correlation of arm-span and height in young women of two races. Annals of Human Biology H{5)M5-M7. 48. Reeves SL, Varakamin C, Henry CJK. (1996) The relationship between arm span measurements and height with Special reference to gender and ethnicity.European Joumal of Clinical Nutrition 50:398. 49. Feldesman MR. (1992) Femur/stature ratio and estimates of stature in children. American Joumal of Physical Anthropology ;87(4):447-459. 50. Feldesman MR, Kleckner JG, Lundy JK. (1990) Femur/stature ratio and estimates of stature in mid- and late- Pleistocene fossil hominids.American Joumal of Physical Anthropology B3{2):359-372. 51. WHO. (1986) Use an interpretation of anthropometric indicators of nutritional status. Bulletin of the World Health Organ/zaf/on 64(6):929-941. 52. Shakir A. (1975) Arm circumference in the surveillance of protein energy malnutrition in Baghdad. American Joumal of Clinical Nutrition 26:661-665. 53. Velzeboer Ml, Selwyn BJ, Sargent S, Pollitt E, Delgado H. (1984) Evaluation of arm circumference as a public health index of protein calorie malnutrition in early childhood. Joumal of Tropical Pediatrics. 54. Burgess HJL, Burgess AP. (1969) The arm circumference as a public health index of protein-calorie malnutrition of early childhood, (ii) A modified standard for mid upper arm circumference in young children.Joumal of Tmplcal Ped/africs 15:189-193. 55. Briend A, Dykewicz C, Graven K, Mazumder RN, Wojtyniak B, Bennish M. (1986) Usefulness of nutritional indices and classification in predicting death of malnourished children.British Medical Joumal 293(6543):373- 376. 56. Jelliffe E, Jellife DB. (1969) The arm circumference as a public health Index of protein calorie malnutrition of early childhood.Joumal of Tropical Pediatrics December: 179-188. 57. Briend A, Zimick S. (1986) Validation of arm circumference as an index of risk in 1 to 4 year olds. Nutrition Research 1986;6:249-61. 58. Alam N, Wojtyniak B, Rahaman MM. (1989) Anthropometric indicator and risk of death. American Joumal of Clinical Nutrition 49:884-888. 59. Vella V, Tomkins A, Ndiku J, Marshal T, Cortinovis I. (1994) Anthropometry as a predictor for mortaiity among Ugandan children, allowing for socio-economic variables. European Joumal of Clinical Nutrition 48. 60. Briend A, Wojtyniak B, Rowland MGM. (1987) Arm circumference and other factors in children at high risk of death in rural Bangladesh. Lancet Sept. 26:725-727. RNI5 Suppfemeni — July2000

61. Velzeboer Ml, Selwyn BJ, Sargent S, Pollitt E, Delgado H. (1983) The use of arm circumference in simplified screening for acute malnutrition by minimally trained health workers. Joumal of Tropical Pediatrics 29:159-166. 62. Lindtjom 8. (1985) Measuring acute malnutrition: a need to redefine cut-off points for arm circumference? Lancet ii:1229-1230. 63. Ross DA, Berry AM, Taylor N. (1986) Measuring malnutrition in famines (letter). Lancet i:151. 64. Ross DA, Taylor N, Hayes R, Mclean M. (1990) Measuring malnutrition in famines: are weight for height and arm circumference interchangeable? Intemational Joumal of Epidemiology 19(3):636-645. 65. McLaren DS. (1986) Measuring malnutrition in famines (letter). Lancet ii:152. 66. Bern C, Nathanail L. (1995) Is mid-upper arm circumference a useful tool for screening in emergencysettings? Lancef 345:631-633. 67. Diskette, P. (1995) Personal communication. 68. Hakewell P, Moren A. (1991) Monitoring and evaluation of relief programmes. Tropical Doctor Supp11:24-28. 69. Rodrigues VC, Rao RSP, Lena A. (1994) Utility of arm circumference as an instrument to identify women at nutritional risk. Tropical Doctor 24:164-166. 70. Olukoya AA. (1990) Identification of underweight women by measurement of arm circumference. Intemational Joumal of Gynecology and Obstetrics 31:231 -235. 71. Collins S. (1996) Using middle arm circumference to assess severe adult malnutrition during famine.The Joumal of the American Medical Association 276(5):391-395. 72. Ohison MA, Biester A, Brewer WD, Hawthorne BE, Hutchinson MB. (1956) Anthropometry and nutritional status of adult women. Annals of Human Biology 28:89-202. 73. Leiter LA, Marliss EB. (1982) Survival during fasting may depend upon fat as well as protein stores. The Joumal of the American Medical Association 248:2306-2307. 74. Leiter LA, Marliss EB. (1983) Survival during fasting. The Joumal of the American Medical Association 250(8):1026. 75. Gassull MA, Cabre E, Vilar A, Alastrue A, Montserrat A. (1984) Protein energy malnutrition: An integral approach and a simple new classification. Human Nutrition. Clinical Nutrition 38C:419-431. 76. Friedman P. (1991) Severe arm-muscle wasting: Sign of lethal malnutrition in sick elderly men. Nutrition 7(3):223- 226. 77. Heymsfield SB, McManus CB, Seitz SB, Nixon DW, Smith Andrews J. (1984) Anthropometric assessment of adult protein-energy malnutrition. Wright, R. A. and Heymsfield, S. B. (3), 27-82. Blackwell Scientific Publications, Inc, Boston. 78. Stini WA. (1973) Reduced sexual dimorphism in upper arm circumference associated with a protein deficit diet in a South American population. American Joumal of Physical Anthropology ZB:ZA^-252. 79. Norgan N (2000) Personal communication 80. Euronut SENECA investigators. (1991) Nutritional status and anthropometry. European Joumal of Clinical Nutrition 45{S3):ZU2. 81. Sommer A, Loewenstein MS. (1975) Nutritional status and mortality: a prospective validation of the QUAC stick. American Joumal of Clinical Nutrition 28:287-292. 82. James WPT, Macsie-Taylor CGN, Norgan NG, Bistriaw BR, Shetty PS, Ferro-Luzzi A. (1994) The value of arm circumference measurements in assessing chronic energy deficiency in third world adults.European Joumal of Clinical Nutrition m S Z -m . 83. Tomkins A, Watson F, (1989) Malnutrition and infection. A Review. NPD 5, ACC/SCN Geneva. 84. Waterlow JC. (1992) Protein energy malnutrition. 1 ed. Edward Arnold, London. 85. De Waal A. Famine that kills: Darfur Sudan, 1984-1985 (Oxford Studies in African Affairs), led. Clarendon Press, Oxford. 86. Korcok M. (1981) Hunger strikers may have died of fat, not protein loss. The Joumal of the American Medical Association 246(17):1878-1879. 87. McLaren DS, Pellett PL, Read WWC. (1967) A simple scoring system for classifying the severe forms of protein- calorie malnutrition of early childhood. Lancet March 11:533-535. 88. McLaren DS, Shirajian E. (1969) Short term prognosis in protein calorie malnutrition. American Joumal of Clinical Nutrition 22:863-870. 89. Beattie J, Herbert PH, Bell DJ. (1948) Famine Oedema. British Joumal of Nutrition 2:47-65. 90. Collins S, Myatt M, Golden B. (1998) Dietary treatment of severe malnutrition in adults. American Joumal of Clinical Nutrition 68:193-9. 91. Baker JP, Detsky AS, Wesson DE, Wolman SL, Stewart S, Whitewell J. et al. (1982) A comparison of clinical judgement and objective measurements. New England Joumal of Medicine April 22:969-972. 92. Collins S, Myatt M. (2000) Short term prognosis in severe adult and adolescent malnutrition. The Joumal of the American Medical Association. (In press). 93. Salama P, Collins S. (2000) Unpublished data from Ajep, south Sudan 94. Habicht JP.(1980) Some characteristics of indicators of nutritional status for use in screening and surveillance, American Joumal of Clinical Nutrition 33:531-535. 95. Last JM. A dictionary of epidemiology. 2ed. OUP; Oxford. 96. Sackett DL, Holland WW. (1975) Controversy in the detection of disease. Lancet August 23:357-359. AnthtopometHc Assessment ofthe Nuttition Status ofAciults in Emetgency-Affectecj Popuktions

APPENDIX 1: BASIC DEFINITIONS

W a s t i n g In c i d e n c e Low weight-for-height, usually defined as less than The number of new cases of undemutrition in a -2SDs of the NCHS/WHO reference median value, defined population within a specified period of or sometimes 80% of the NCHS/WHO reference timers median of weight-for-height. At present, no clear definition of wasting in adults is available. P r e v a l e n c e The total number of undemourished individuals in a STUNTING given population at any one timers Low height-for-age, usually defined as less than -2SDs of the NCHS/WHO reference median value, In d i c a t o r or sometimes 80% of the NCHS/WHO reference An indicator is a variable taking one of two possible median of height-for-age. values, one indicating the presence of a condition, the other indicating absence of the condition In m \ the context of assessing nutritional status this will Body mass index (weight/height^). An index of usually mean whether or not an individual is above protein and fat stores. or below a pre-defined value of a particular body measurement or combination of measurements FAMINE OEDEMA (e.g. MUAC, weight, BMI). A clear idea of the Bilateral dependant pitting oedema resulting from purpose of the assessment being undertaken undemutrition. In both adults and children this is an should inform the choice of body measurement important sign of severe undemutrition, carrying a used and the cut-off points that are applied to it. high mortality risk. Se n s it iv it y MUAC Sensitivity is the number of people correctly The mid upper arm circumference, measured on a identified by an indicator as being undemourished straight left arm (in right handed people) mid way divided by the total number of undemourished between the tip of the shoulder (acromium) and the people measured. A sensitive indicator will identify tip of the elbow (olecranon). An index of peripheral a large proportion of the undemourished people protein and fat stores. This is an accepted measure measured. of acute undemutrition. S p e c if ic it y

A s c it e s Specificity is the number of people correctly The accumulation of serous fluid in the peritoneal identified by an indicator as non-undemourished cavity. divided by the total number of non-undemourished people measured. A specific indicator will correctly CORMIC INDEX identify a large proportion of the non- undemourished people measured. The ratio of leg-length to trunk-length, sometimes called the sitting height to standing height ratio (SH/S).

fp RNIS Supplement — July2000

Surveillance SCREENING Ongoing scrutiny, generally using methods In the context of nutrition, the purpose of screening distinguished by their practicability, uniformity, and is to select individuals at increased risk of morbidity frequently their rapidity, rather than by complete and mortality who are likely to respond to treatment accuracy. Its main purpose is to detect changes in and to treat them. The numbers selected by the trend or distribution in order to initiate investigative indicator used are those that need treatment In or control measures The appropriateness of a emergency relief programmes, the appropriate nutritional indicator for surveillance liesjn its ability indicator cut-off point (screening level) is that which to reflect the true incidence of undemutrition in a selects the number of individuals that can be population. The principal objective of a surveillance treated with the resources at hand Such cut-off system is to provide information in order to aid point values cannot therefore be determined decision making at a community level and there universally but must be tailored to suit the may be no immediate benefits for the individuals resources available in each particular situation surveyed Surveillance normally estimates The choice of underlying body measurement will incidence rather than prevalence. often be determined by available equipment (e.g. scales may not be available to measure weight). N u t r it io n a l s u r v e y A survey which examines the presence or absence C l in ic a l m o n it o r in g of undemutrition in each member of a The performance and analysis of routine representative sample of a population at one measurement aimed at detecting changes in the particular time. A survey estimates prevalence nutritional status of an individual. rather than incidence

2 0 UNITED NATIONS Administrative Committee on Coordination SUB-COMMITTEE ON NUTRITION THE UN SYSTEMS FORUM FOR NUTRITION

VISION AND ,MANDATE

Our long-run vision is o f g worl4 in which malnutrition is no longer g human development constraint. This is possible, but to achieve it will reguire decisive action at country level, supported by a coherent and co-ordinated international strategy, foun(ie(\ on human rights and providing a framework for action throughout the UN and international development finance system, implemented in close partnership with NGOs, bilgtergls and governments, Nutrition needs to be made a key development priority, recognized as vital to the achievement o f other social and economic goals. Good nutrition under normal conditions contributes to the prevention and mitigation o f death and malnutrition in emergency situations. Good nutrition facilitates the prompt return to conditions favouring development following disasters.

The mandate o f the ACC/SCN is to raise awareness o f nutrition problems and mobilize commitment to solve them — at global, regional and national levels; to refine the direction, increase the scale and strengthen the coherence and impact o f actions against malnutrition world wide; and to prom ote c o operation amongst UN agencies and partner organizations in sixpport o f national efforts to end malnutrition in this generation.

Three main areas for action have been identified: Ci) Promote o f harmonized approaches among the UN agencies, and between the UN agencies and governmental and non-governmentgl partners, for greater overall impact on malnutrition. Cii) Review the UN system response to malnutrition overall, monitor resource allocation and collate information on trends and achievements reported to specific UN bodies. Ciii) Advocate and mobilize to raise awareness o f nutrition issues at global, regional and country levels and mobilize accelerated action against malnutrition. These three functions are all vital and o f equal importance and can be seen as a triangle, one dependent on the other.

The UN members of the ACC/SCN are the FAO, IAEA, IFAD, ILO, UN, UNAIDS, UNDP, UNEP, UNESCO, UNFPA, UNHCHR, UNHCR, UNICEF, UNRISD, UNU, WFP, WHO and the World Bank. The ADB and IFPRI are also part o f this group. From the outset, representatives o f bilateral donor agencies and NGOs have participated actively in SCN activities. The Secretariat is hosted by WHO in Geneva.

The SCN undertakes a range o f activities to meet its mandate. Annual meetings have representation from those mentioned above as well as academia — a one-day Symposium is held during the annual meeting, focussing on a subject o f current importance for policy. The SCN convenes working groups on specialized areas o f nutrition; currently there are nine working groups in areas ranging from foetal and infant malnutrition, nutrition o f the school aged child, and household food security to capacity building .

The SCN's reports on the world nutrition situation, published every two to three years, are authoritative sources o f information to guide the international community in its nutrition work. Nutrition Policy Papers and the SCN News summarise current knowledge on selected topics. Quarterly bulletins on the nutritional status o f refugees and displaced persons are also published in collaboration with a large network o f NGOs. GENERAL SUMMARY - SUGGESTED INDICATORS FOR ASSESSING ADULT UNDERNUTRITION IN THE FIELD

SL/R VEYS AN P POPULA TION LEVEL ASSESSMENTS OE CHRONIC UNPERNUTRinON

1. BMI 2. MUAC

if the survey results are usect to compare BMI between populations then BMI must be standardised for the average population Cormic Index (sitting height/standing height)

THERAPEl/riC FEEDING CENTRES

CHANCES mo4cl:

1. M U A C < 1 6 0 m m a lo n e 2. MUAC < 161-185 mm plus one of the following: - O edem a - Inability to stand - Apparent dehydration 3. Famine oedema (Beattie grade 3 or worse) alone as assessed by a clinician to exclude o th e r causes

Additional social factors can be included in the model. The relative weighting o f these, for example whether you need one, two or three additional social factors to tip the balance in f^vouf o f therapeutic rather than supplementary care must be determined locally. Relevant social factors couId include the following: - Access to food (quality and quantity) - Distance from centres - Presence/absence o f carers - shelter - Dependants - Cooking utensils

SUPPLEMENTARY FEEDING CENTRES

MUAC < 161-185mm and no relevant clinic signs or few relevant social criteria

In any particular situation, workers should take these suggested standards as the starting point and adapt them according to situation-specific factors. u Preliminary Communication

Using Middle Upper Arm Circumference to Assess Severe Adult Malnutrition During Famine

Steve Collins, MB, BS

Objectives.—To examine the use of middle upper arm circumference mea ment from the majority who would is surement (MUAC [cm]) and body mass index measurement (BMI [kg/m^]) in the central to effective targeting. In chil screening of severely malnourished adults during famine. dren, this is carried out using anthro Design.—Nonrandomized cohort study, correlating measurements of MUAC pometric indicators based on measure and BMI. ! ments of weight, height, and middle upper arm circumference (MUAC).^® Setting.—The Concern Worldwide adult feeding center in the village of Ayod in For adults, anthropometric indicators south Sudan. The area had experienced several years of war, leading to severe of acute energy deficiency with cutoff famine during early 1993. values appropriate to severe famine do Participants.—A total of 98 adult inpatients belonging to the Nuer tribe. Criteria not exist. During the initial phases of for entry into the study were prior admission to the feeding center and the ability to famine relief operations, there may be stand and have a BMI measured. many hundreds of adults crowding for Main Outcome Measures.—A comparison of the ease of use of MUAC and BMI admission to an adult therapeutic cen assessments, and a correlation of MUAC and BMI measurements. ter. In a Concern Worldwide adult feed Results,—An MUAC measurement was easier to perform on severely malnour ing center in Angola during 1993, more ished adults than BMI assessment. For MUAC, the patient could be standing, sit than 300 people were screened per day by a single expatriate worker and his ting, or, in extreme cases, lying. For BMI, patients were required to stand. Measur team of local staff. The lack of an ap ing BMI requires a height board, weighing scales, and mathematical calculations; propriate screening tool to differentiate to measure MUAC, only a tape measure is required. A correlation between mea quickly those who will not live without surements of MUAC and BMI was demonstrated (r=0.8B; 95% confidence inter specialized assistance from those who val, 0.82-0.92; P<.001). The proportions of the population and the actual individu will poses difficulties for famine relief als identified as malnourished by the 2 indicators were similar. workers. Conclusions.—The MUAC measurement reflects adult nutritional status as de This is especially true if the volun fined by BMI. During famine, MUAC may be better suited to screening admissions teers working in famine relief programs to adult feeding centers than BMI, Studies to assess the capacity of MUAC cutoffs are young, first-time recruits with little to predict mortality In severe adult malnutrition are needed. or no experience in similar situations. JMIA. 1996;276:391-395 These problems can lead to turning away adults who require specialized therapy or admitting those not requiring such attention. This results in unnecessary SEVERE ACUTE adult malnutrition a high mortality.®-® Delivering assistance adult mortality, inefficient operation of is common during famine.’ Without spe to these malnourished adults is there adult feeding centers, and suboptimal cialized treatment, it is associated with fore one of the major tasks facing emer allocation of resources. gency relief programs. During the height Body mass index (BMI [kg/m®]) is an of a famine when large numbers of people accepted indicator of chronic energy de From Concern Worldwide, Dublin, Ireland, and the are competing for scarce resources, ac ficiency (CED) in adults* that has been Center for International Child Health, Institute of Child curate targeting of emergency food sup shown to reflect risk of mortality in non Health, London, England. plies is vital. Screening admissions to famine situations.^ In the absence of data Corresponding author: Steve Collins, MB, BS, Ole- uffynon. Old Hall, Llanidloes, Powys SY18 6PJ, Wales feeding centers to separate those who from acutely malnourished adults, it has (e-mail; [email protected]). would not survive without special treat been assumed that BMI is also an ap-

JAMA August 7. 1996-Vol 276, No. 5 Assessing Severe Adult Malnutrition—Collins 391 propriate indicator for use during fam 'assessment of severely malnourished the Concern Worldwide feeding center ine.”’® However, there are many prob adults during the height of a famine and in Ayod during April 1993. Patients were lems associated with the use of BMI in the relationship between MUAC and suffering from acute nonedematous mal the assessment of acute adult malnutri BMI. This was to be achieved by com nutrition, complicated by a variety of tion. The cutoff values applicable to adult paring estimations of MUAC with those infections commonly seen during fam CED are too high to be appropriate in of BMI and relating each to short-term ine, of which dysentery, respiratory in famine situations, when the majority of outcomes such as mortality and length of fection, helminthiasis, and malaria were adults can have BMIs below 16 kg/m*, stay. Ideally, a study relating MUAC to the most common. An epidemic of vis the upper limit defining severe CED.”’'” the risk of mortality in the general popu ceral leishmaniasis (kala-azar) was re Such a cutoff is therefore of little use in lation is requh ed to establish functional portedly also present in the area (oral differentiating between those who need MUAC cutoff points that could be used communication, A. Davis, MSc, Médecins therapeutic feeding and those who would for the screening of adult admissions to Sans Frontières, Holland, 1995). survive without it. feeding centei*s. However, famines always During the first week after their ad The height and weight measurements involve massive social upheaval and large mission, all patients who could stand required to calculate BMI aie time-con numbers of afflicted people. These fac were measured, weighed, and had their suming and difficult to obtain from the tors, combined with the armed conflict MUAC assessed. The MUAC was as most severely emaciated people who of that has characterized almost all famines sessed on the left arm midway between ten cannot sit up unaided. In practice, during the last 10 yeai’s, make such broad- the olecranon and the tip of the acro this results in BMI assessment fi-equently based population studies unfeasible. As a mium process of the scapula, using a being omitted. Famine edema, seen in up consequence, no such studies have yet standard, nonstretchable tailor’s tape to 90% of adults suffering from severe been perfonned for any nutritional indi measure, read to the nearest 5 mm. The acute malnutrition,'" '” also confounds es cator in either adults or children duiing weight of the patients in minimal cloth timations of BMI. Famine edema is a poor the height of a famine. ing was assessed to the nearest 100 g prognostic sign in adult malnutrition and Given these difficulties, relating using Soehnle digital electronic scales, is associated with a markedly increased ^ MUAC to moi'tality in a selected feeding- previously calibrated against known relative risk of mortality. It also increases center population is a reasonable alter weights in the United Kingdom. The body weight.'® As a consequence, in those native, better than an absence of infor height was assessed using Microtoise patients with severe famine edema, the mation on the use and suitability of adult stadiometers to the nearest 5 mm. relative risk of mortality increases as BMI nutritional indicators in famine. Data Height measurements were made with increases, the opposite of that seen among obtained from such selective studies are the patients standing with their heels marasmic patients. This increase nullifies useful in providing a general idea of suit- together and their legs and back as the ability of BMI to reflect severity of |able cutoff values for screening admis straight as possible. All patients were malnutrition or risk of mortality in pa sions into feeding centers during fam lost to follow-up. tients with severe famine edema. Thus, ine. In addition, they provide information To allow for comparisons with previ BMI may be an inappropriate measure about anthropometric thresholds requir ous studies, the relationship between for the majority of a famine-afflicted popu ing various levels of intervention once a MUAC and BMI is presented using the lation. person is admitted to a feeding center regression line MUAC=pBMI+c 0 = In children, MUAC has been shown and background information against slope and c=intercept). The degree to to be useful in the assessment of nutri which the perioimance of differing cura which the 2 indices identified the same tional status.'”'” The measurement is tive regimens can be judged. individuals, however, was analyzed in easy to perform even on the most de ^ The Ayod, south Sudan, study popu terms of the sensitivity and specificity of bilitated children, is efficient at predict lation was chosen for this study based MUAC in identifying those individuals ing mortality,'”'” and can be readily on the investigator’s observations of a below specific BMI cutoffs. In this analy taught to minimally trained health work high prevalence of severe malnutrition sis, BMI was the dependent variable, and ers.'” It is well suited to screening ad and very low prevalence of famine the regression line BMI=pMUAC+c was missions of children to feeding centers edema. A low prevalence of famine used. Integer values of BMI between 16 during the initial phase of an emergency edema was considered desirable as and 10 were assigned as cutoff points, aid operation. At present during fam edema is a potential confounder of the and the corresponding MUAC values pre ines, MUAC is only recommended for relationship between BMI and MUAC. dicted from the regression line. Body use in children between 1 year and 5 ' In this study, it was not possible to mass index was taken as the reference years of age””’” and is not used in the evaluate short-term outcomes and their indicator because it is the more accepted assessment of adult malnutrition. How relation to the nutritional indicators. The indicator of adult nutritional status. How ever, measurements of arm circumfer reason for this failure was that during ever, since BMI has not yet been shown ence in adults have long been known to the second week of April the village of to be a criterion standard for the assess reflect changes in body weight.*® More Ayod, including the Concern Worldwide ment of acute malnutrition in adults, this recently, such measurements have been adult center where the study was tak choice was somewhat arbitrary. Data used in the diagnosis of adult malnutri ing place, was attacked. The villagers were analyzed using Epi Info*” and Mi tion in hospitals*' and as a prognostic were forced to flee, and all Concern staff crosoft Excel.*® indicator in the elderly or in those with were evacuated. Tragically, when Con cancer.**’*” A reasonably close relation cern workers were able to return to RESULTS ship between adult MUAC and BMI has Ayod, the center had been destroyed, The MUAC and BMI values recorded been demonstrated in noimal adult popu and the occupants had either fled or been in the center are presented in Table 1. lations from a number of developing murdered. Mean BMI values (kg/m*) were higher countries,*^ but to date no such com for male than female patients (mean, parisons have been made during severe METHOD male 13.3 kg/m*, female 12.6 kg/m*; analy famine. I Body mass index and MUAC were sis of variance (ANDVA) F =3.26; The aim of this study was to examine assessed in 98 severely malnourished d/=l,96; P=.07). This result is of bor the potential usefulness of MUAC in the adult inpatients from the Nuer tribe at derline significance. Mean MUAC val-

392 JAMA, August 7 , 199B—Vol 276, No. 5 /Assessing Severe Adult Malnutrition—Collins Table 1 — BMI and MUAC Data (n=98)*

BMI, kg/m* MUAC, cm 1 1 ( I Normal Normal Lower Upper Values, Lower Upper Values, Sex No. Median Minimum Quartile Quartile Mean+SDt Median Minimum Quartile Quartile Mean+SDt Male 41 13.4 10.2 11.8 14.4 20.0+2 19 19 12.5 17.8 20 25.6+2.4 Female 57 12.5 9.3 11.3 13.9 20.4+2.65 17.5 15 16 19.5 24.8+2.59

*BMI Indicates txjdy mass index; and MUAC. middle upper arm circumference. fNormal values based on a sample of 5669 adults from developing countries.’*

ues (cm) were also higher for male than female patients (mean, male 18.8 cm, fe male 17.9 cm; ANOVA F =4.24; rf/= 1,96; 0 Male O Female — Male, Regression MUAC on — Female, Regression MUAC on P<.05). BMI Slope 0.86, Intercept 7.36 BMI Slope 1.16, Intercept 3.28 The height and weight measurements Male, James et aP* Female, James et aP* required for BMI assessment were dif ficult to perform on the patients who had difficulty in standing. Many patients in 24 -, cluded in this study were very slow to step onto the scales and had great diffi 0 22 - culty in standing up straight enough to be 1 measured. The MUAC measurements 2 0 - were easier to perform on the subjects I and could be taken with the patient stand -OCD O ing, sitting, or, in extreme cases, l>dng. I 18 - Ü Figure 1 presents a scatter diagi am of MUAC and BMI measurements with the < 16 - □ O male and female regiession lines for & MUAC compared with BMI in the Ayod Q. study population and other more nonually nounshed populations in developing coun tries. The con elation coefficient between MUAC and BMI was 0.88 (95% confi dence interval [Cl], 0.82-0.92; F=319; d f =1,96; P<.001 ). Correlation coefficients 5 7 911 13 15 17 19 were similar for male and female patients, Body Mass Index, kg/m® men=0.79 (95% Cl, 0.64-0.88; F =65; df=\,99\ P<.001), women=0.92 (95% Cl, Figure 1.— Scattergram illustrating the relationship between middle upper arm circumference (MUAC) and 0.87-0.95; F =303; # 1 ,5 5 ; P<.001). There body mass index (BMI) measurements in the Ayod study population, taken at the height of a severe famine was no significant difference between men (n=98), and a comparison between these data and other data from more normally nourished populations In developing countries.** There was a close correlation between MUAC and BMI in the Ayod population. and women in the relationship expressed The regression line of MUAC on BMI in this population differed from that recorded in the better nourished by the regression line of MUAC on BMI. populations. At the most severe levels of emaciation (ie, MUAC < 16 cm and BMI <11 kg/m*), this differ The proportion of the population be ence is of practical significance, equivalent to a change of 1 to 2.5 cm In MUAC. low various cutoff values and an analy sis of the sensitivity, specificity, and pre sitivity was maximized without appre levels of emaciation, this difference is of dictive power of MUAC with respect to ciable loss in specificity (Figure 2). practical significance and is equivalent BMI are presented in Table 2. The pro to a change of 1 cm to 2.5 cm in MUAC portions defined by the 2 indicators were COMMENT (Figure 1). similar, although at the lower end of the The patients in this study presented Currently, no other data are available BMI range, MUAC tended to define a with extremely severe emaciation. from severely malnourished adults, and slightly smaller proportion of the popu Thirty-five percent (34/98) of the pa speculation as to the cause of this dif lation than the comparable BMI cutoffs. tients had a BMI below 12 kg/m^, the ference must be tentative. Some of the The sensitivity-specificity analysis in level previously considered to be the difference may be due to variation be dicated that the individuals identified lowest limit compatible with recov- tween racial groups. The common re by the 2 indices were also similar, par gj.y 27,28 These extremely low BMI mea gression lines proposed by James et aF ticularly within the BMI range of 16 surements, although in part due to the are based on data from Somalia, Papua kg/m^ to 13 kg/m^. The positive predic tall, long-legged, Nilotic phenotype of N ew Guinea, India, China, Senegal, Zim tive value for the pi ediction of BMI by the Nuer people, are similar to recent babwe, Mali, and Ethiopia, which con MUAC remained high for all the cutoff observations from Somalia.’-*® tain significant racial variation. In some values. Below a MUAC of 18.5 cm (BMI Body mass index and MUAC were of the data sets that contribute to the of 13 kg/m^), the relative proportion of closely related in both male and female common regression lines, such as those the population defined by MUAC, and patients. Differences between the male for Ethiopian, Somali, and Chinese wom the sensitivity of MUAC in predicting and female regression lines were not en, the relationship between BMI and BMI, fell. This cutoff point corresponded signifidant. The regiession line of MUAC MUAC is much more similar to that to a position on the relative operating compared with BMI did differ from that seen in this study. characteristic (ROC) curve for the pre recorded in better nourished populations Several other reasons may explain diction of BMI by MUAC, where sen by James et al.^ At the most severe why the relationship between MUAC

JAMA, August 7, 1996— Vol 276, No. 5 Assessing Severe Adult Malnutrition— Collins 3 9 3 ■i, •

Table 2.—Comparison of the Sensitivity and Specificity of MUAC Cutoffs Against Reference BMI Cutoffs in (98)*

Positive Negative BMI Sensitivity Specificity Predictive Predictive Cutoff, Below BMI Below BMI MUAC , Below MUAC Below MUAC of MUAC of MUAC Value of MUAC Value of MUAC kg/m* Cutoff, No. Cutoff, % Cutoff, cm Cutoff, No. Cutoff, % C utoff,t % Cu1off,t % Cutoff,* % Cutoff,* % 16 92 94 22.5 96 98 100 33 96 100 15 86 1 88 21.0 82 84 91 67 95 50 14 71 72 20.0 68 69 87 78 91 70 13 48 49 18.5 45 46 83 90 89 85 12 34 35 17.0 26 27 71 97 92 86 11 14 14 15.5 8 8 50 99 88 92 10 2 2 14.5 2 2 100 100 100 100

*MUAC indicates middle upper arm circumference: and BMI, body mass Index. tThe specificity and sensitivity of the MUAC cutoff are relative to the BMI cutoff. ilh e positive and negative predictive values of the MUAC cutoff are relative to the BMI cutoff.

tral, predominantly protein, stores. tional screening during the height of an 1001 These changes would tend to reduce the emergency where time and skilled per rate of decrease in MUAC compared sonnel are at a premium. 90- with BMI and revert the MUAC on BMI In children, MUAC is associated with 80- regi ession line back toward that seen in 2 problems: the preferential selection of better-nourished populations. The drop younger children as malnourished^'-^ and 70- in the sensitivity of MUAC when com- a lack of reproducibility in MUAC mea Ipared with BMI that occurred below an surements.™ The preferential selection MUAC of 18.5 cm (Table 2) might be a of younger children as malnourished re reflection of such a change. The require sults from the gradual increase in the ment to be able to stand as a criterion normal MUAC between the ages of 1 40- for entry into this study would have year and 5 years.™ In older children and resulted in a selection bias within the adolescents, the rapidly changing pat 30- sample population partially obscuring terns of skeletal muscle and subcutane

2 0 - this effect. Only those with sufficient ous fat are likely to exaggerate this prob peripheral muscle mass remaining to en lem, and it is unlikely that MUA(] will 1 0 - able them to stand would have been se be of much use in these groups. In adults, lected. Although fewer than 10 patients the use of MUAC is not subject to such 10 20 30 40 50 60 70 were excluded from the study because errors caused by growth, but can be 1-Specificity, % they could not stand, many of them would affected by the redistribution of subcu have been those who had exhausted their taneous fat toward central areas of the Figure 2.—Relative operating characteristic curve peripheral stores. The drop in the sen body with aging.^" Potentially more se for the prediction of body mass index (BMI) by sitivity of MUAC at extreme grades of rious obstacles to the use of MUAC in middle upper arm circumference in the Ayod study emaciation would most likely have been adults are the problems with the repro population (n=98). The point “X," beyond which any greater had the whole population of the ducibility of MUAC measurements. As increase in sensitivity is accompanied by a marked decrease in specificity, corresponds to a BMI cutoff center been measured. At present the when MUAC is measured in children, value of 13 kg/m’. relative importance of these effects is interobserver and intraobser\^er errors unknown, and more work is required in MUAC measurements may occur. and BMI might differ between a famine- before more precise MUAC cutoffs can However, the larger dimension of the affected population and other better- be derived. adult arm is likely to reduce the relative nourished populations. For example dur Assessment of adult nutritional sta importance of these errors. The devel ing starvation, tissue with a low meta tus using MUAC requires no equipment opment of colored, nonnumeric MUAC bolic rate, such as subcutaneous fat and apart from a tape measure. As the index bands reflecting threshold values of skeletal muscle, is preferentially catabo- is the actual measurement itself, there MUAC with a change of color would lized.^’® This would tend to decrease is also no need for mathematical ma further reduce these problems by elimi the MUAC at a greater rate than the nipulation of the measurements ob nating numerical errors. Given the ease BMI, changing the MUAC on BMI re tained. In contrast, assessment of BMI with which MUAC measurements can gression line toward that seen in Figure requires a height board and weighing be performed, it would be feasible to 1. The increased catabolism of skeletal scales that often need recalibration in refer any patients found to have an muscle caused by the higher rates of the hot dusty environments in which MUAC within 0.5 cm either side of the infection in acute malnutrition™ would famines usually occur. In addition, the 2 threshold, a zone marked by a different further increase this effect, as would measurements obtained must be math color, to a more experienced worker for the reduction in physical work using the ematically transfoiTned before being use verification. These problems are cur arms, that occurs during famine. Con ful. All these process are prone to error, rently being investigated using data versely, as the severity of malnutrition especially during the highly pressurized from several Concern Worldwide adult progresses, elements such as bone and circumstances characteristic of the early therapeutic centers. the neurovascular bundle, the volumes phases of famine relief programs. At the The results of this study indicate that of which change little during acute star height of a famine, there are always MUAC is potentially useful for differ vation, form an increasingly large pro many people in need of treatment and entiating adults who require emergency portion of the arm volume. Peripheral few facilities or staff with which to treat feeding from those who are likely to energy and protein stores also become them. The ease with which MUAC can survive without it. The destruction of exhausted, and there is change to cen be assessed, therefore, suit it for nutri the Ayod center and the death and dis

3 9 4 JAMA, August 7, 1996—Vol 276, No. 5 Assessing Severe Adult Malnutrition—Collins persion of its inpatients prevented any 1 ceral protein from vital organs such as several advantages over the use of BMI. analysis of the relationship between'; the heart. At this degree of emaciation, Insufficient data are currently available MUAC and mortality. This is a major ) patients are unlikely to survive given to stipulate the fimctional MUAC cutoff limitation in this study, and in the ab-1 only a general ration, and specialized value that differentiates malnourished sence of these data, functional MUAC. nutritional and medical support is re adults that require specialized therapeu cutoff values based on risk of mortality j quired (S.C., M. Myatt, B.E. Golden, MD, tic care from those who can survive with cannot be presented. However, the drop ; unpublished data, 1995). out it. To that end, additional research in sensitivity of MUAC below a level of is required to evaluate the power of adult 18.5 cm is likely to reflect the change CONCLUSION MUAC to predict mortality in different from peripheral to central protein stores ! During famine, a practical indicator famine-affected populations. at a point when peripheral energy and with which to assess adult malnutrition protein stores have become exhausted. is needed, especially during the initial I would like to thank Anita Ennis, Fr Jack Finu- cane, and Roberta Gordon, without whose help this This level of MUAC may represent a phases of emergency relief operations. work would not be possible, and Mark Myatt, Han functional threshold below which there i This study indicates that MUAC is well nah Scrase, MSc, and Andrew Tomkins, FRCP, for is an increase in the catabolism of vis-1 suited to such a role, and its use has their advice and encouragement.

References 1. Collins S. The need for adult therapeutic care in Delgado H. Evaluation of arm circumference as a of lethal malnutrition in sick elderly men. Nutri- emergency feeding programs. JAMA. 1993;270;637- ■public health index of protein calorie malnutrition tion. 1991;7:223-226. 638. in early childhood. J Trop Pediatr. 1983;29:135-144. 24. James WPT, Taylor-Mascie GCN, Norgan NG, 2. Centers for Disease Control. Population based 15. Shakir A. Arm circumference in the surveil et al. The value of arm circumference in assessing mortality assessment: Baidoa and Afgoi, Somalia lance of protein energy malnutrition in Baghdad. chronic energy deficiency in Third World adults. 1992. MMWR Morb Mortal Wkly Rep. 1992;41:913- Am J Clin Nutr. 1975;26:661-665. Eur J Clin Nutr. 1994;48:883-894. 917. , 16. Bairagi R. On validity of some anthropometric 25. Dean AG, Dean JA, Burton AH, Dicker RC. 3. Collins S. Famine in Somalia. Lancet. 1993;341: indicators as predictors of mortality. Am J Clin Epi Info Version 6: A Word Processing, Database 1478-1479. Nutr. 1981;34:2592-2594. and Statistics Program for Epidemiology on Mi 4. Ville de Goyet C, Seaman J, Geijer U. The Man 17. Briend A, Wojtyniak B, Rowland MGM. Arm crocomputers. Stone Mountain, Ga: USD Incorpo agement of Nutritional Emergencies in Large Popu circumference and other factors in children at high rated; 1990. lations. Geneva, Switzerland: World Health Orga risk of death In rural Bangladesh. Lancet 1987;2; 26. Microsoft Excel Version 5. Redmond, Wash: nization; 1978. 725-727. Microsoft Corporation; 1995. 5. Hakewell P, Moren A. Monitoring and evalua 18. Vella V, Tomkins A, Ndiku J, Marshal T, Corti 27. Henry CJK. Body mass index and the limits of tion of relief programs. Trop Doct 1991;l(suppl): novis I. Anthropometry as a predictor for mortality human survival. Eur J Clin Nutr. 1990;44:329-335. 24-28. among Ugandan children, allowing for socio-eco 28. James WPT, Ferro-Luzzi A, Waterlow JC. Defi 6. FeiTO-Luzzi A, Setti S, Franklin M, James WPT. nomic vaiiables. Eur J Clin Nutr. 1994;48:189-197. nition of chronic energy malnutrition in adults. Eur A simplified approach to assessing adult chronic en 19. Velzeboer MI, Selwyn BJ, Sargent S, Pollitt E, J Clin Nutr. 1988;42:969-981. ergy deficiency. Eur J Clin Nutr. 1992;46:173-186. Delgado H. The use of arm circumference in sim 29. Waterlow JC. Metabolic adaptation to low in 7. Reddy V. Body mass index and mortality rates. plified screening for acute malnutrition by mini takes of energy and protein. Annu Rev Nutr. 1986; Nutr Netos. 1991;12:1. mally trained health workers. J Trop Pediatr. 1983; 6:495-526. 8. Boelaert M, Davis A, Le Lin B, et al. Nutrition 29:159-166. 30. Broom J. Sepsis and trauma. In: Garrow JS, Guidelines. Paris, France: Médecins Sans Fron 20. Ohison MA, Biester A, Brewer WD, Hawthorne James WPT, eds. Human Nutrition and Dietetics. tières; 1995. BE, Hutchinson MB. Anthropometry and nutri London, England: Churchill Livingstone Inc; 1993: 9. Young H. Food Scnmty and Famine: Assess tional status of adult women. Hum Biol. 1956;28: 456-464. ment and Response. Oxford, England: Oxfam; 1992. 189-202. 81. Bern C, Nathanail L. Is mid-upper arm cir 10. Collins S. The limit of human adaptation to 21. Gassull MA, Cabre E, Vilar A, Alastrue A, cumference a useful tool for screening in emer starvation. Nat Med. 1995;1:810-814. Montserrat A. Protein energy malnutrition: an in gency settings? Lancet. 1995;345:631-633. 11. Burger GCE, Sandstead HR, Drummond J. tegral approach and a simple new classification. 32. Ross DA, Taylor N, Hayes R, Mclean M. Mea Starvation in western Holland: 1945. Lancet. 1945; Hum Nutr Clin Nutr. 1984;38C:419-431. suring malnutrition in famines: are weight for height 1:282-283. 22. Heymsfield SB, McManus CB, Seitz SB, Nixon and arm circumference interchangeable? Int J Epi 12. McCance RA, Widdowson EM. Studies in Un DW, Smith Andrews J. Anthropometric assessment demiol. 1990;19:636-645. demutrition; Wuppertal 1H6-9. London, England: of adult protein-energy malnutrition. In: Wright 33. McLaren DS. Measuring malnutrition in fam Medical Research Council; 1951. RA, Heymsfield SB, eds. Nntritiomil Assessment ines. Lancet. 1986;2:152. 13. Lipscomb FM. Medical aspects of Belsen con Boston, Mass: Blackwell Scientific Publications Inc; 34. Euronut SENECA Investigators. Nutritional centration camp. Lancet. 1945;1:313-315. 1984:27-82. status and anthropometry. Eur J Clin Nutr. 1991; 14. Velzeboer MI, Selwyn BJ, Sargent 8, Pollitt E, 23. Friedman P. Severe arm-muscle wasting: sign 45:31-42.

JAMA, August 7 ,1 9 9 6 —Vol 276, No. 5 Assessing Severe Adult Malnutrition—Collins 395 : ' I ; ' . ■ '

/ , LETTER FROM SOMALIA

Short-term Prognosis in Severe Adult and Adolescent Malnutrition During Famine Use of a Simple Prognostic Model Based on Counting Clinical Signs

Steve Collins, MBBS Context In the setting of famine, infection is likely to cause mortality among se Mark Myatt, BA verely malnourished persons. Although clinical signs are likely to be useful prognostic indicators in this setting, use of a clinical assessment model has not been studied. o r m a n y y e a r s , i t h a s b e e n r e c - Objective To examine the use of clinical signs in the prediction of short-term mor ognized that malnutrition tality in severely malnourished adults and adolescents during famine. increases both susceptibility to Design Retrospective cohort study. and severity of infection. Vita Fmin, mineral, and other dietary defiSetting Concern Worldwide Adult Therapeutic Feeding Center in Baidoa, Somalia. ciencies, depressed cell-mediated and Patients Data from the clinical records of 383 adult and adolescent inpatients ad humeral immunity, gastric acidity, mitted to the center between November 1992 through March 1993 who were aged mucosal integrity, and altered flora are 15 years or older and had a body mass index (BMI) of 13.5 kg/m^ or less or any signs of edematous malnutrition. all known to increase susceptibility to infection. The situation in famine is AAain Outcom e M easures Association of mortality with presence or absence of 8 usually worsened by a breakdown in clinical signs (edema, hydration, ascites, dysentery, diarrhea, anemia, chest infection, and ability to stand) and BMI at admission, and sensitivity and specificity of models public health infrastructure and con including a count of clinical signs and BMI in the prediction of mortality at the center. gregation of displaced people in crowded and unhygienic conditions.^ In situa Results Ninety-one patients (23.8%) died, with a median time to death of 8 days tions of malnutrition in developed coun from admission. Of the 8 clinical signs, severe edema (unadjusted odds ratio [OR], 2.45; 95% confidence interval [Cl], 1.41 -4.27), apparent dehydration (unadjusted OR, tries, exhaustion of fat or fat-free mass 2.73; 95% Cl, 1.60-4.66), and inability to stand (unadjusted OR, 2.96; 95% Cl, is most often the terminal event,^ but 1.40-6.26) were independently associated with mortality. The most useful clinical the combination of poor public health model was that based on the presence of any 1 of these 3 signs, with a sensitivity of environment and immunosuppression 77% and a specificity of 59%. Ability of admission BMI to predict mortality was less in famine means that it is infection, rather than random. than absolute loss of fat or fat-free mass, Conclusions Models based on clinical signs predicted death better than BMI. Simple that kills people. counts of clinical signs performed as well as more complex models based on the ad The strong association between infec dition of ORs. Counting relevant clinical signs is an easy and effective prognostic tool tion and mortality indicates that clini in severe adult and adolescent malnutrition during famine; however, it is not sensitive cal signs are likely to be useful prognos enough for use as a screening tool. tic indicators.’ Clinical models for JAMA. 2000;284:621-626 www.jama.com prediction of mortality, useful in the screening of admission to child feeding tional status appears to have been Author Affiliations: Concern Worldwide, Dublin, Ire centers, have been proposed.® Although restricted to well-nourished surgical land (Dr Collins): Center for Intemational Child Health, Institute of Child Health (Dr Collins) and the Interna reported to be effective in identifying patients.® To date, similar assessments tional Centre for Eye Health, Institute of Ophthalmol children at high risk of mortality,^ these have not been made in severely mal ogy, University College London (Mr Myatt), Lon models have been criticized because nourished adults or during famines. don, England; and Valid International Ltd, Llanidloes, Wales (Dr Collins). interactions between the features used, Extreme conditions often character Corresponding Author and Reprints: Steve Collins, such as edema and hypoproteinemia, ize famine relief programs. Levels of MBBS, Oleuffynon, Old Hall, Llanidloes, PowysSYIS 6PJ, Wales (e-mail: [email protected]). were not taken into account.® In adults, need are high and trained staff, equip Section Editor: Annette Flanagin, RN, MA, Manag the use of clinical models to assess nutri- ment, and buildings are scarce. Some- ing Senior Editor.

JAAAA, August 2, 2000—Vol 284, No. 5 621 SHORT-TERM PROGNOSIS IN MALNUTRITION DURING FAMINE

Medical Treatment proti Table 1. Characteristics of the Study Population • Oral antibiotics were needed by the ma long All Patients Survivors Nonsurvivors tent Variables (N = 383) (n = 292) (n = 91) jority of patients on admission, with sudd Self-reported age, mean 33.0 (16.0) [15.0-89.0] 32.7 (14.8) [15.0-89.0] 33.8 (16.5) [15.0-80.0] many patients continuing to receive an (SD) [range], y antibiotic throughout their stay in the prot Sex center. Penicillin V and ampicillin were thes No. 382 292 90 the first-line antibiotics for pulmo effe Male, No. (%) 185(48.4) 140(47.9) 45 (50.0) nary infection, as were trimethoprim- quei Female, No. (%) 197(51.6) 152(52.1) 45 (50.0) sulfamethoxazole and metronidazole give; BMI, kg/m® tial ; No. 296 244 52 for dysentery and persistent diarrhea. atth Mean (SD) [range] 12.7 (2.0) [9.0-24.0] 12.6 (2.0) [9.0-24.0] 12.9 (1.9) [9.3-17.6] Chloramphenicol was the only second- Weight, kg line antibiotic used. vatii No. 341 271 70 Discharge criteria were predomi tion Mean (SD) [range] 34.0 (7.2) [19.2-59.5] 34.0 (7.2) [21.0-59.5] 34.1 (7.4) [19.2-53.4] nantly clinical. Freedom from infec sev( *No. for sex, BMI (body mass Index), and weight indicates how many persons for whom data were availatjie. tious disease, a good appetite, constant proi weight gain, and abihty to walk and care times, crowds of several thousand weight was 34 (7.2) kg. Mean admis for oneself were all necessary condi Dat people may gather, attempting to gain sion BMI of survivors was 12.6 kg/m^ vs tions for discharge. The presence of On admission to centers. Such centers may 12.9 kg/m^ for nonsurvivors. minimal pedal edema did not preclude tere become overfilled, disorganized, and discharge as long as these criteria were me; dysfunctional if patient selection and Center, Staff, and Equipment met. On discharge, transport to pa tors prioritization is not efficient. Tools to The therapeutic feeding center con tients’ home villages was provided. nui assess patients must be quick, simple, sisted of 9 stick shelters with plastic col and reliable, even when used by mini sheeting roofs and mud floors. Sepa Dietary Treatment mai mally trained staff, who may be barely rate shelters were designated for new Patients were orally rehydrated using me literate. This study investigates the effec admissions, those with edematous mal the World Health Organization for adi tiveness of 2 assessment tools, body nutrition, those with dysentery, and mula for ORS (sodium chloride, 3.5 g/L; we: mass index (BMI) and models using those suspected of having pulmonary trisodium citrate, 2.9 g/L; potassium sio: readily ascertainable chnical signs, dur tuberculosis. Beds were not available chloride, 1.5 g/L; and glucose, 20 g/L). dej ing the height of a major famine. and all patients had to lie on plastic All patients were given 6 to 8 meals per tio: sheets on compacted mud floors. day. Diets consisted of high-energy milk of I METHODS The center was staffed by 1 expatri (made from dried skimmed milk, veg wa Subjects ate physician, 1 to 2 expatriate nurses, etable oil, and sugar), sweet tea, rice, This retrospective cohort study involved 3 local nurses, and approximately 30 oil, beans, biscuits, and local fruits, es cU; clinical record data of393 patients admit other helpers and cooks. High levels of pecially bananas. During the first few me ted to the Concern Worldwide Adult supervision were maintained, with staff- days after admission, the high-energy cei Therapeutic Center, Baidoa, Somalia, patient ratios of 1:8 or less, particularly milk component was diluted with ORS up between November 4,1992, and March during the first few days after admis to reduce the incidence and severity of tie 15,1993. Criteria for inclusion in this sion. These staff members distributed watery diarrhea and vomiting that is of retrospective analysis were a BMI at food and oral rehydration solution ten associated with the early stages of re admission of 13.5 kg/m^ or less or any (ORS), encouraged patients to eat and refeeding severely malnourished per in signs of edematous malnutrition, a drink, and helped feed those who were sons. The degree to which the high- so reported age of 15 years or older, the unable to feed themselves. energy milk was diluted was tailored ac er presence of an inpatient treatment and A set of standard Hanson spring cording to severity of diarrhea. cl clinical data card, and no record that the scales and a locally constructed height Two therapeutic diets were used in the se patient was admitted because of medi board were used to weigh and mea center. For the higher-protein diet, Û cal criteria alone. Of these 393 patients, sure patients. The scales were cali ■patients were offered 156 g of protein 2 10 were excluded owing to the absence brated in London, England, and Baidoa and 16.5 XICP kj (16% of energy from St of outcome data. The mean reported age using known weights. Results were protein; protein-energy ratio, 9.5 mg/kj t( of the remaining 383 subjects was 33.0 similar at both sites. In the useful range per day) each day; for the lower- ti years, with no significant age differ of 20 to 50 kg, there was a linear in protein diet, patients were offered 81 g b ence between sexes (t3so= 1.32; P = .19). crease in error (r=0.99). A correction of protein and 16.5 X 10^ kJ (8.5% of e Of those admitted, 48.4% were men factor derived using ordinary least energy from protein; protein-energy c (T a b l e 1). Mean (SD) BMI at admis squares linear regression was used to ratio, 5 mg/kJ per day) each day. In V sion was 12.7 (2.0) kg/m^; mean (SD) correct each weight measurement. patients with edema, the higher- e

622 JAAAA, August 2, 2000—Vol 284, No. 5 m m m Ê

SHORT-TERM PROGNOSIS IN MALNUTRITION DURING FAMINE

protein diet was associated with pro (grade 2). Ascites was similarly coded. patient was the number of 3 relevant longed anorexia, increased or persis Severe edematous malnutrition was clinical signs exhibited. tent edema, and occasional instances of defined as either pitting edema greater The association between BMI and sudden death. Introduction of the lower- than grade 2, periorbital edema greater survival was explored in a systematic protein diet in December 1992 reduced than grade 1, or any ascites. Apparent manner. A series of indicator variables these complications dramatically. This dehydration was defined as sunken eyes, was created using successively higher effect was so marked that subse- decreased urine output, or dry mucous BMI cut-points at intervals of 0.5 kg/m^ quendy, all patients with edema were membranes. Diarrhea (profuse, fre (ie, variables were created to indicate given the low-protein diet during the ini quent, watery stool) and dysentery (pro whether an individual BMI was <10.0, tial phase of treatment. Data collected fuse, frequent, watery stool with blood 10.5,11.0,11.5,12.0,12.5, etc). Each at the center have supported these obser and with or without pus) were coded as indicator was tabulated against sur vations, demonstrating a 4-fold reduc present or absent. Missing data values vival. The indicator with the most sig tion in mortality among patients with were coded as not present. nificant positive association (BMI <11.0 severe edema who received the low- The level of supervision of data col kg/m^; OR, 2.44; 95% confidence in protein diet. lection was variable. Occasionally, ex terval, 1.11-5.32; Yates-corrected patriate access to the center by the su X^i=5.22; P<.05) was then included in Data Collection and Coding pervising clinician and nurses was the predictive model. On admission, patients were regis restricted for part of a day because of A simulation exercise was under tered and weighed and had their height gun fighting, bandit attacks, attempts taken to validate the methods used to measured by trained local enumera at hostage taking, and military occu construct the models. Each run of the tors, supervised by a specially trained pations that occurred in Baidoa dur simulation involved splitting the data set nurse or the clinician responsible for ing the study period. into 2 parts by randomly selecting (us collecting clinical data (S.C.). Body ing a pseudorandom number genera mass index was calculated from weight Data Analysis tor) approximately half of the available measurements taken within 2 days of Bivariate and multivariate logistic regres records as a training data set and using admission; 97.6% (n = 289) of BMIs sion analyses were used to assess the as the other half of the data as a validation were calculated on the day of admis sociation of clinical features at presen data set. The ORs and uncorrected for sion. A rapid clinical screen, assessing tation with mortality in the center. each association between predictor vari degree of pitting edema, ascites, hydra Variables that were not independendy as ables and death were calculated using the tion, dysentery, diarrhea, anemia, signs sociated with mortality were elimi training data set. Sums of ORs and sign- of chest infection, and ability to stand, nated fi’om the logistic models in a back count models were then constructed us was performed by either the supervis ward-stepwise fashion using estimation ing variables with significant positive as ing clinician or a nurse. W eight and/or techniques. Odds ratios (ORs) of the 3 sociations with death, defined as any clinical condition of each patient was signs independently associated with variable having an OR of greater than monitored daily. Outcome (death in the death were used to construct a predic unity and an uncorrected x^ of greater center, discharge ahve, or lost to follow- tive model. If a patient exhibited none than 3.84 (ie, P:^.05 for tables with 1 df). up) was recorded on individual pa of these signs, the score was 0. If they ex Models constructed using these vari tient cards and in the center’s register. hibited 1 sign, their score was the OR as ables were then tested on the valida Data were transferred from inpatient sociated with that sign. If more than 1 tion data set. The simulation was run record cards onto specially designed cod clinical sign was present, their score was 1000 times. All data entry and analyses ing forms, double-entered into a per the sum of the ORs associated with each were performed using Epi Info,^^ Logis- sonal computer, and validated for entry sign. The sensitivity and specificity of tic,^^ and Microsoft Excel. error. Pitting edema was coded using the predicted mortality at score intervals of classification of Beattie et aP^ (0=ab 0.5 were calculated and receiver operat RESULTS sent; 1 = minimal edema on foot or ankle ing characteristic (ROC) curves were Ninety-one (23.8%) of the 383 pa that is demonstrable but not visible; plotted. Receiver operating characteris tients died before discharge. Median 2=visible on foot or ankle; 3=demon tic curves plot sensitivity against time to death was 8 days (range, 1-53 strable up to knee; 4=demonstrable up 100-specificity and are a useful means days). T a b le 2 shows the clinical sign to inguinal ligament; 5=anasarca). Pit of assessing the abihty of an indicator to variables that were independently as ting edema that was recorded as present discriminate between healthy and dis sociated with survival and the ORs as but unquantified was assumed to be mod eased persons or, in this case, between sociated with these variables. Ascites erate (grade 3). Periorbital edema was patients with differing outcomes (sur was not associated with survival. Se coded as 1 (mild), 2 (moderate), or 3 (se vival or death). A simpler model using vere edema, inability to stand, and ap vere), with present but unquantified a count of clinical signs was also con parent dehydration were indepen edema again assumed to be moderate structed. In this model, the score for each dently associated with mortality.

JAA/\A, August 2, 2000—Vol 284, No. 5 623 SHORT-TERM PROGNOSIS IN MALNUTRITION DURING FAMINE

T a b le 3 presents the sensitivity and (n=218), performance was similar to ter than random. The mean area under mode specificity for prediction of death in all the clinical model, predicting 46% of the ROC curves for these 1000 simula and t 383 patients for all combinations of deaths at a specificity of 85%. When use tions was 63.42% (SD, 4.23%). cludei clinical models created using edema, in of BMI was further restricted to in predii ability to stand, and apparent dehydra clude only marasmic women who could worst tion. The most discriminatory model stand (n= 108), BMI performed better COMMENT ment (ie, the model with the greatest sum of than the clinical model, predicting 56% The risks of mortality associated with ap const specificity and sensitivity) was that of deaths at a specificity of 87%. parent dehydration, inability to stand, pane based on the presence of any 1 of these The ROC curve for the model con and severe edema were independent of ducet 3 signs. This model predicted 77% of structed by counting clinical signs was each other. A patient presenting with enld deaths at a specificity of 59%. identical to that produced using the more than 1 of these signs therefore had prog] a risk of mortaUty equivalent to the sum F ig u r e 1 shows the ROC curves for sum-of-ORs chnical model. Addition of BMI the prediction of short-term mortality a BMI marginally improved these mod of the risks of all the signs that were pres cente using the sum-of-ORs clinical model els. ROC curves for these models are ent. A scoring system based on the sum hydn of the ORs predicted 77% of deaths at a and BMI. The prediction of death by the presented in FIGURE 2. couli sum of the ORs of severe edema, in Nine hundred ninety-three runs specificity of 59%. A simplified model, withi ability to stand, and apparent dehydra (99.3%) of the simulation of the sum- using only the presence or absence of 3 there tion performed at the highest level of of-ORs clinical model yielded a predic clinical signs—severe edema, inability nosti sensitivity and specificity. Prediction of tive model better than random. The to stand, and apparent dehydration— then death using BMI alone in these pa mean area under the ROC curves for predicted deaths in a near identical fash dehy tients yielded worse-than-random re these 1000 simulations was 63.25% (SD, ion (Figure 2). Use of this sign-count Pi sults. When BMI was used in a sub 4.39%). Nine hundred ninety-seven runs model required only a count of rel date population selected to include only (99.7%) of the sign-count clinical model evant clinical signs, not calculation of creaî marasmic patients who could stand and BMI yielded a predictive model bet ORs, which would be difficult in the ingi field. Eliciting the 3 clinical signs used prec in the model was quick and easy in all edei Table 2. Clinical Signs Associated, With Survival in Severe Malnutrition (N = 383) patients and required no equipment. caus Survivors, % Nonsurvivors, % Unadjusted Odds Ratio Consequently, the sign-count model can was Variables (n = 292) (n = 91} {95% Confidence Interval) be readily taught to local workers and tor 1 Inability to stand 7 19 2.96 (1.40-6.26) is suitable for use as a prognostic model pop Dehydration 22 43 2.73 (1.60-4.66) in the field, even during the early phases T Severe edema 19 36 2.45 (1.41-4.27) of a famine relief program. tren Dysentery 38 29 1.50(0.88-2.53) The capacity of BMI to predict death ther Anemia 21 25 1.28(0.71-2.31) was limited. Only among female ma nesf Diarrhea 34 37 1.18(0.70-1.99) rasmic patients who could stand did thar Chest infection 58 57 0.97 (0.58-1.61) BMI perform better than the clinical Am cen Table 3. Prediction of Mortality by 3 Clinical Signs (N = 383)* tifyi No. of No. of higl Nonsurvivors Survivors Signs Predicted to Die Predicted to Die Sensitivity, % Specificity, % Accuracy, % PPV, % NPV, % nos Single cial Dehydration 39 63 43 78 70 38 81 exit Edema 33 55 36 81 70 38 80 mo Inability to stand 17 21 19 93 75 45 79 cal Combined using “and” can Edema and inability to stand 3 3 3 99 76 50 77 ent Dehydration and inability to stand 8 4 9 99 77 67 78 sio: Dehydration and edema 9 12 10 96 75 43 77 res All 3 signs 1 1 1 76 50 76 100 ad( Combined using “or” Edema or inability to stand 47 73 52 75 69 39 83 ide Dehydration or inability to stand 48 80 53 73 68 38 83 no: Dehydration or edema 63 106 69 64 65 37 87 tif) Any 1 sign 70 121 77 59 63 37 ' 89 wb Any 2 signs 18 17 20 94 77 51 79 if a *PPV indicates positive predictive value; NPV. negative predictive value. wi

624 JAMA, August 2, 2000—Vol 284, No. 5 SHORT-TERM PROGNOSIS IN MALNUTRITION DURING FAMINE models. When nonmarasmic patients and sensitivity requirements for screen similar to that used by 1 of us (S.C.) and those unable to stand were in ing are different than those required for when confronted with large crowds of cluded (ie, the entire patient cohort), the prognostic indicators. To be useful in malnourished adults. In such situa prediction of mortality using BMI was famine relief programs, where there are tions, a rapid visual inspection and a very worse than random (Figure I). Assess often large crowds of people attempt brief clinical examination are under ment of BMI was also difficult and time- ing to gain entry into feeding centers, a taken to assess fat in the upper arms out- consuming. Dehydration present among screening indicator must have a high patients admitted to the center re specificity. This is not the prime require Figure 1. Receiver Operating Characteristic duced admission BMIs. Because appar ment for a prognostic indicator. We have Curves for Prediction of Mortality by the ent dehydration was also a sign of poor demonstrated that the sign-count model Sum of Odds Ratios for 3 Clinical Signs and prognosis, its presence tended to make is sensitive for predicting death within Body Mass Index (BMI) BMI a better predictor of death in the a therapeutic feeding center. We have not o B M I In Ail P atients — R a n d o m Indicator center. The inability of most severely de assessed its potential for use as a screen I o BMI in Marasmic a BMI in Edematous hydrated people to stand meant that they ing tool to exclude individuals from ad Patients P atients a B M I in F em ale ' Sum of Odds Ratios In could not have their BMI estimated mission to feeding centers. Marasmic Patients All P atients within 2 days of admission and were This potential use for screening ad Specificity, % therefore not included in the BMI prog missions to therapeutic feeding centers too 80 60 40 20 nostic models. It is therefore likely that needs to be investigated further. Al 100 j- the reduction of admission BMIs due to though the model is somewhat nonspe dehydration was not a major effect. cific for identification of patients who Presence of severe edema was asso would die despite having been given ciated with a poor prognosis and in treatment in the center, it is likely that creased admission BMI. This confound in the absence of treatment, many more ing meant that the ability of BMI to of the individuals identified would have predict death among patients with died. The model might therefore be more edema was worse than random. Be specific if used for screening. To in cause severe edema was common, BMI crease the model’s specificity for iden was not useful as a prognostic indica tifying malnourished individuals at tor when applied to the entire center greater risk of dying, an indicator of nu 200 4060 80 100 population (Figure 1). tritional status would need to be added 100-Specificity, % These results indicate that at the ex to the model. Middle upper arm circum tremes of emaciation observed in adult ference (MUAC), recently shown to be The 3 clinical signs are severe edema, inability to stand, and apparent dehydration. therapeutic feeding centers, clinical ill a potentially useful indicator of acute ness is a better indicator of prognosis adult malnutrition,'^ might be suitable. than degree of wasting as defined by BMI. Addition of MUAC would differentiate Figure 2. Prediction of Short-term Mortality A model using the presence of readily dis between those with clinical illness but by Counting 3 Clinical Signs With and W ith ou t a BMI Threshold cernible clinical signs is useful in iden no malnutrition, who would be better tifying patients at high risk of death. The treated in medical units, from those with Specificity, % high-risk patients identified by this prog both illness and malnutrition, who 100 60 40 20 1 0 0 - nostic model can then be moved to spe would be best treated in specialized feed cialized feeding centers or areas within ing centers. By weighting the model ap existing feeding centers with the most propriately, having an MUAC below a motivated staff, a higher level of medi certain threshold could be made a nec cal supervision, and, if necessary, higher essary prerequisite for selection. This 60- caregiver-patient ratios. This is a differ would ensure a high prevalence of mal ent role than that of screening admis nutrition in the population assessed by 40 o— Sign C o u n t sions for entry into feeding centers, a the sign-count part of the model. Sign Count With research priority in severe adult/ Within this selected population, the 20 BMI Thresfiold Û— Sum of Odds Ratios adolescent malnutrition.^^ Rather than specificity of the clinical signs for iden - ■ Sum of Odds Ratios identifying inpatients with a poor prog tifying death among persons who did not With BMI Threshold nosis, a screening indicator must iden receive treatment would be high, fulfill 20 40 60 80 tify persons in the general population ing the requirement for screening. Use 100-Specificity, % who are likely to respond to treatment of this single model would be equiva The 3 clinical signs are severe edema, inability to stand, if admitted to a feeding center but who lent to application of a 2-stage screen and apparent dehydration. A body mass index (BMI) threshold of 11.5 kg/m^ or less was used. will die if not admitted. The specificity ing procedure. Such a 2-stage system is

JAMA, August 2. 2000— Vol 284, No. 5 625 SHORT-TERM PROGNOSIS IN MALNUTRITION DURING FAMINE side of the center, followed by more in requiring therapeutic feeding.^® At that treme conditions, such as the famine in volved anthropometric and clinical time, there was only a single 24-hour Somalia, make BMI inappropriate for inspection inside the center. In this ad therapeutic feeding center vrith a maxi such a role. Middle upper arm circum hoc system, the initial screen outside of mum capacity of 400 patients. In this ference might prove to be more useful the center increases the prevalence of se context, given the hmited means avail in this context. More work on identify vere malnutrition requiring urgent as able, selecting only those patients with ing relevant clinical signs and combin sistance among persons who are al a good chance of survival would have op ing them with MUAC in such models for lowed through the gates to be assessed timized the efficient use of resources. In use in screening is required. by a more time-consuming clinical ex this study, the combination of 2 rel Funding/Support This work was supported by Valid amination. In this selected population, evant clinical signs predicted death at International. the higher prevalence of life-threaten sensitivities of greater than 95%. In Wau, Acknowledgment: We thank the staff and patients in the Baidoa Concern Worldwide Adult Therapeutic ing malnutrition makes the relatively low the combination of these 2 signs would Center and Jack Finucane, Ben Rigby, BA, Hannah specificity of the chnical screen less im have been a useful triage tool to deter Scrase, MSc, Keith Sullivan, PhD, and Andrew Tomkins, portant. The sign-count model pro mine the low end of the spectrum of mal MBBS, FRCP, for their advice and encouragement. posed here, combined with an MUAC nutrition severity among patients whom cutoff, would formalize this method, the center would admit. REFERENCES making it more useful to workers with Reliefprograms are rarely totally over 1. Tomkins A, Watson F. Malnutrition and Infec tion: A Review. Geneva, Switzerland; World Health less experience. The threshold of both whelmed, and it is important that the ex Organization: 1989. MUAC and the chnical score chosen as istence of a triage tool does not under 2. Waterlow JC. Protein Energy M alnutrition. Lon don, England: Edward Arnold; 1993. a screening cutoff would depend on the mine international will to provide suf 3. De\Na3iA.FamineThatKilis:DarfurSudan, 1984- context and the relative balance of re ficient humanitarian assistance during 1985 (OxfordStudies in African Affairs). Oxford, En gland; Clarendon Press; 1989. sources and needs. emergencies. Screening and prognostic 4. Leiter LA, Marliss EB. Survival during fasting may Emergency relief programs must en indicators are also different than mark depend upon fat as well as protein stores. VA/WA. 1982; sure that the majority of a population has ers of improvement during treatment. 248:2306-2307. 5. Korcok M. Hunger strikers may have died of fat, access to the minimum requirement to In this study, maintenance of fluid bal not protein loss. JAMA. 1981;246:1878-1879. maintain life. This requires prioritizing ance and hydration, disappearance of 6. McLaren DS, Pellett PL, Read WW. A simple scor ing system for classifying the severe forms of protein- lower-input interventions with a large edema, and a steady increase in weight calorie malnutrition of early childhood. Lancet. 1967; coverage of the vulnerable population in the absence of increasing edema were 1:533-535. ^ 7. McLaren 05, Shirajian E. Short-term prognosis in above high-input services treating rela all markers of improvement. This has protein calorie malnutrition. Am J Clin Nutr. 1969; tively few. Access to a life-sustaining gen been reported previously. 22:863-870. eral ration, providing at least 8786 kj Ideally, a study evaluating the rela 8. Waterlow JC. Classification and definition of pro tein calorie malnutrition. BMJ. 1972;3:566-569. from grains, legumes, and vegetable oil; tionship between clinical signs and risk 9. Baker JP, Detsky AS, Wesson DE, et al. A compari adequate water; sanitation; basic health of mortahty in a general population ex son of clinical judgement and objective measure ments. N EnglJ Med. 1982;306:969-972. care; and dry supplementary feeding, periencing famine is required to estab 10. Collins S, Myatt M, Golden BE. Dietary treat must, therefore, form the basis of any re lish relevant clinical screening crite ment of severe malnutrition in adults. Am J Clin Nutr. 1998;68:193-199. lief program. “ Therapeutic feeding cen ria. However, conducting such broad- 11. Beattie J, Herbert PH, Bell DJ. Famine oedema. ters are efficient and effective only if these based population studies is probably not BrJ Nutr. 1948;2:47-65. 12. Dean AG, Dean JA, Burton AH, Dicker RC. Epi basic prerequisites are in place. This hi feasible during famine and none have Info Version 6: A Word Processing Database and Sta- erarchy of interventions constitutes a yet been undertaken in such condi tistlcs Program for Epidemiology on Microcomput form of triage. During the initial stages tions. Given these difficulties, evaluat ers. Stone Mountain, Ga: USD Inc; 1990. 13. Dallai GE. Logistic: a logistic regression program of famine, need will usually outstrip the ing prognostic models in selected feed for the IBM PC. Am Stat 1988:42:272. available resources and the focus should ing center populations is a feasible 14. Microsoft Corporation. M icrosoft Excel [com puter program]. Version 4. Redmond, Wash: Micro be on ensuring that the general ration alternative. It is certainly better than the soft Corp; 1994. is adequate before targeted feeding cen current absence of information on prog 15. CollinsS.Theneed for adult therapeutic care in emer gency feeding programs. JAMA. 1993;270:637-638. ters are established. At this stage, those nostic indicators in famine. 16. Collins S. Using middle arm circumference to as too malnourished to survive with this In conclusion, counting simple clini sess severe adult malnutrition during famine. JAMA. basic support will die. As resources in 1996;276:391-395. cal signs is a useful method of assess 17. Habicht JP. Some characteristics of indicators of crease, it will become possible to under ing prognosis in severe adult and ado nutritional status for use in screening and surveil take therapeutic feeding, although, ini lescent malnutrition during famine. The lance. Am J Clin Nutr. 1980;33:531-535. 18. Humanitarian charter and standards in disaster re tially, clinical triage may be necessary. addition of an anthropometric indica sponse. In; Minimum Standards in Nutrition. Geneva, For example, in the town of Wau in tor is likely to make these models use Switzerland: Sphere Project; 1999:chap 3. 19. Buchanan-Smith M, Collins S, Dammers C, Wekesa southern Sudan duringjune through Au ful for screening admissions of adults to F. Evaluation of Danish Humanitarian Assistance to gust 1998, more than 100 therapeutic feeding centers. The poor association of Sudan 1992-1998. Copenhagen: Danish Ministry of Foreign Affairs; 1999. centers would have been required to treat BMI with prognosis and difficulties in 20. Collins S. The limit of human adaptation to star the estimated 16000 children and adults obtaining BMI estimations during ex vation. WatMed. 1995;1:810-814.

626 JAMA, August 2, 2000—Vol 284, No. 5 Dietary treatm ent of severe malnutrition in adults^

)teve Collins, Mark Mvatt, and Barbara Golden

See corresponding editorial on page 10.

ABSTRACT The aim of this study was to compare the attention from the scientific community and, generally, diets effects of two diets, differing primarily in protein content, on the containing relatively high amounts of protein were recom nutritional rehabilitation of severely malnourished adults. The mended (7-12). In the early 1950s, scientific attention shifted study took place in the Concern Worldwide Adult Therapeutic toward malnourished children, in whom the quantity and quality !{Feeding Centre in Baidoa, the town at the epicenter of the 1992 of dietary protein required for successful rehabilitation became il famine in Somalia. The response to treatment in 573 patients the focus of much research (13, 14). In particular, the importance admitted to the center between November 1992 and March 1993 of liver pathology in kwashiorkor was recognized (15). Recently, was studied. Mortality, appetite, rates of edema loss, and weight it was shown in children with kwashiorkor that mortality was gain in 2 groups of patients receiving either a higher-protein minimal when diets containing only maintenance levels of pro (16.4% of energy from protein) or lower-protein (8.5% of ener tein (< 1 g kg~’ -d"') were given (16, 17). Up to the time of the igy from protein) diet were compared. Among edematous Somali famine in 1992, however, these diets had been tested 1 patients, the use of the lower-protein diet during the initial phase only in specialized hospital units and no field research had been I of treatment was associated with a threefold decrease in mortal- performed during an emergency relief program. I ity (P <0.05) and accelerated resolution of edema {P < 0.05). The difficult circumstances that exist during famines generally Among marasmic patients, no differences in mortality or rate of prevent the execution of classical, scientifically rigorous I weight gain were observed. The large reduction in mortality research. This has resulted in an absence of field research since associated with the use of the lower-protein diet in edematous the late 1940s when the concentration camps in Europe and the j patients appeared to be due to the lower amount of dietary pro- Far East were liberated. In the absence of field research, results I tein. However, differences in the 2 diets other than or in addition from studies performed in less severely malnourished subjects, to the protein content may have contributed. Notwithstanding, usually in hospital settings, have been extrapolated to the very the data obtained suggest strongly that severely malnourished different circumstances found during war and famine. Thus, it adults, particularly those with edema, recover more successfully was shown that adult subjects with body mass indexes (BMIs, in with a diet of lower protein content than usually recommended. kg/m^) of 17-18 respond well to diets with protein-to-energy The lower-protein diet used in this study was much cheaper and ratios (P:Es) > 19% (18). As a result, such high-protein diets have more easily obtained than the conventional higher-protein diets continued to be recommended for the management of severe in Baidoa. Am J Clin Nutr 1998;68:193-9. adult malnutrition. Such diets were used in all Concern World wide therapeutic feeding centers (TFCs) during the Somali relief KEY WORDS Malnutrition, refeeding, protein-energy mal operation in 1992-1993. However, the adults admitted to these nutrition, PEM, edematous malnutrition, dietary protein, Soma TFCs generally had BMIs of 10-13, far less than those that had lia, marasmus, famine been studied. They also often had edema (19). It was observed that many of these patients, particularly those with edema, were refusing the high-protein diets. Thus, the hypothesis was INTRODUCTION advanced that the high-protein diets were deleterious during the Severely malnourished adults are encountered frequently dur initial phase of rehabilitation. Therefore, the aim of this study was ing emergency famine relief programs (1). Because adult energy to compare the immediate and short-term effects of a lower-pro- requirements are proportionately less than those of children, the peak incidence of severe malnutrition and death in adults gener ' From Concern Worldwide, Dublin, and the Centre for International Child ally occurs later than in children (2). Often, by the time an emer Health, Institute of Child Health, London; Brixton Heath, Llanidloes, United Kingdom; and the Department of Child Health, University of Aberdeen. Med gency relief operation is up and running, many of the children ical School, Aberdeen, United Kingdom. have already died (3) and malnourished adults constitute a large ^ Address reprint requests to S Collins, Oleuffynon, Old Hall, proportion of the nutritional problem (4, 5). The difficulties of Llanidloes, Powys SYI8 6PJ, Wales, United Kingdom. E-mail: steve@ feeding severely malnourished adults, especially those with ede concern 1 .demon.co.uk. matous malnutrition, have long been recognized (6). During the Received July 30, 1997. first half of this century, these problems received considerable Accepted for publication February 9, 1998.

1 tein diet with those of the conventional higher-protein diet during tuberculosis. The center was staffed by 1-3 expatriate staff nutritional rehabilitation of severely malnourished adults admit members, 3 local nursing staff members, and =30 other helpers ted to a TFC during the emergency relief program in Baidoa. and cooks. These staff members distributed the food and oral rehydration solution (ORS), encouraged patients to eat and drink, and helped feed those who were unable to feed them SUBJECTS AND METHODS selves. High levels of supervision, with one staff member for All adults admitted to the Concern Worldwide Adult Thera less than 9 patients, were maintained during the daytime, partic peutic Feeding Centre in Baidoa between 25 October 1992 and ularly during the first few days after admission. At night, fre 30 March 1993 were studied. The center was established accord quent gun battles and attempts at hostage-taking meant that ing to internationally accepted guidelines and operated accord supervision by expatriate workers was impossible. ing to the principles of best clinical practice given the prevailing A set of standard spring scales and a locally constructed circumstances. The setting and the conditions, that is, an emer height board were used to weigh and measure patients. The gency feeding center in a war zone, did not allow us to carry out scales were calibrated in London and Baidoa with known a formally designed study and the number of observations made weights. The results were similar at both sites. In the useful and the degree of control possible were constrained by the cir range of 20-50 kg, there was a linear increase in error (r = 0.99) cumstances. The criteria for admission varied depending on the with the true weight being expressed by the following equation: space available in the center and the existence of other medical true weight = (1.013987 X scale weight) + 0.707. The correction facilities in Baidoa. In general, only patients with a BMI < 13.5, factor derived from this equation was used to correct each those assessed as very severely malnourished by the clinician weight measurement. (SC), those too ill to be weighed and measured on admission, or those with edematous malnutrition were admitted to the center. Medical treatment in the center Less severely malnourished persons were referred to supplemen Oral antibiotics were given to most patients at admission, with tary kitchens. During October and November 1992, there were many continuing to receive an antibiotic throughout their stay. no alternative medical centers in Baidoa and 16 patients with Penicillin V, ampicillin, cotrimoxazole, and metronidazole were BMIs > 13.5 and medical rather than nutritional problems were first-line antibiotics and chloramphenicol was the second-line admitted. This practice stopped in mid-December 1992 when a antibiotic. Discharge criteria were predominantly clinical. medical inpatient unit was opened in Baidoa. Absence of clinical evidence of infection, a good appetite, con Five hundred seventy-three patients were admitted to the cen stant weight gain for a minimum of 3 d, and the ability to walk ter. They were aged between 15 and 80 y (median: 30 y) and and care for oneself were all necessary conditions for discharge. 46% were male. The age distributions of male and female The presence of minimal pedal edema did not preclude discharge. patients were similar. On admission, patients were registered, Diets were weighed, and had their height measured by trained local assistants, supervised by a specially trained nurse or the clinician During their stay in the center, patients received 6-8 meals (SC). A rapid clinical screen, assessing degree of pitting edema, each day. For the first month of operation, the standard higher- ascites, hydration, dysentery, diarrhea, anemia, signs of chest protein diet (HP diet) used in all of the other TFCs in Baidoa was infection, and ability to stand was performed by either the clini the only diet available. This diet consisted of recovery milk cian (SC) or the nurse. The weight or clinical condition or both [King’s Food (Ermelo, Holland), a blend of dried skim milk, veg of each patient were monitored daily during rehabilitation and etable oil, vitamins, and minerals], UNIMIX (a blend of soya outcome was recorded. Lack of trained staff and an extremely flour, oil, and sugar), rice, beans, and BP5 biscuits (Compact, high patient load prevented detailed recording of individual food Bergen, Norway) (Table 1). The recovery milk and BP5 biscuits intake. Pitting edema was graded by using the classification of were premixed fortified foods, designed especially for famine Beattie et al (12) as grade 0 (absent), 1 (minimal edema on the relief and marketed in Europe; the UNIMIX was a premixed, for foot or ankle, demonstrable but not visible), 2 (visible on foot or tified food made up for UNICEF in various factories in Africa. ankle), 3 (demonstrable up to knee), 4 (demonstrable up to Patients receiving this diet were offered, on average, 158 g pro inguinal ligament), and 5 (anasarca). Pitting edema that was tein/d and 16.2 MJ/d (P:E, 16.4%), including 70 g oil and 137 g recorded as present but unquantified was assumed to be moder lactose. The cost of this diet was =US$1.90 per person daily. ate (grade 3); unquantified edema was recorded in 4.3% of patients. Periorbital edema was separately graded as 0 (absent), TABLE 1 1 (mild), 2 (moderate), and 3 (severe), with present but unquan The higher-protein diet^ tified edema assumed to be grade 2. Ascites was similarly graded. Edematous malnutrition was defined as either pitting Ingredient Quantity Protein Energy Sodium Potassium edema greater than grade 2, periorbital edema greater than grade gid gId kJ/d mmolld mmol/d 1, or any ascites. Missing data values were coded as absent and Recovery milk 252 55 4276 42 71 excluded from further analysis. Data were analyzed by using EPI UNIMIX 100 12 1680 0 14 INFO (20), LOGISTIC (21), and Microsoft EXCEL (22). Rice 200 12 2940 0 8 1 46 The center, the staff, and the equipment Beans 130 33 1966 BP5 biscuits 275 46 5290 2 18 The center consisted of nine stick shelters with roofs made of Total — 158 16151 45 156 plastic sheeting and mud floors. Separate shelters were desig ^ The higher-protein diet contained 16.4% of energy from protein. nated for new admissions, those with edematous malnutrition, UNIMIX was made up for UNICEF in various factories in Africa. BP5 bis those with dysentery, and those suspected of having pulmonary cuits were manufactured by Compact (Bergen, Norway).

■'1 DIETARY TREATMENT OF SEVERE MALNUTRITION 195

TABLE 2 return of appetite and a substantial loss of edema or ascites or The lower-protein diet' both. Because loss of edema was accompanied by loss of weight, weight loss was accepted as a sign of a positive response to treat Ingredient Quantity Protein Energy Sodium Potassium ment (16). In marasmic patients, the recovery phase was defined s/d gid UId mmol/d mmolld by good appetite and steady weight gain for >3 d. Dried skim milk ISO 65 2646 44 73 Oil 150 0 5733 0 0 Analysis of mortality Sugar 160 0 2688 0 0 Of the 573 patients for whom admission data were collected. Rice 200 12 2940 0 8 16 were admitted primarily on medical grounds and we did not Bananas 500 5 2100 0 51 have sufficient dietary information for 70. These 86 patients were Total — 82 16107 44 132 excluded from the analysis of mortality. For analysis, patients ‘ The lower-protein diet contained 8.5% of energy from protein. were assigned to either the HP group (n = 343) or the LP group {n = 144) according to the diet they received during their initial During the first few days after admission, the milk component of period of stay in the center. Thirteen patients were wrongly allo this diet was diluted to half strength with the World Health cated to diets by junior staff on admission. These mistakes were Organization (WHO) formula ORS (WHO, Geneva), containing rectified within 3 d by the clinician (SC) and the patients were 3.5 g NaCl/L, 2.9 g trisodium citrate/L, 1.5 g KCl/L. and 20 g assigned to the correct diet. For the analysis, the diet group of giucose/L. This reduced the total daily average offered to 129 g these patients was that of the corrected diet. Twenty-seven protein and 14 MJ (P:E, 15.5%). patients (all from the marasmic group) were lost to follow-up and From 5 December 1992, a lower-protein diet (LP diet) mixed were excluded from further analysis of mortality associated with on site from basic commodities was also available. The LP diet the two diets, leaving 377 patients with marasmus or mild edema consisted of “high-energy milk.” a blend of dried skim milk, and 83 patients with moderate to severe edematous malnutrition. vegetable oil, and sugar, together with bananas, white rice, and These data are presented in Table 3. sweet tea (Table 2). Patients receiving this diet were offered, on Two-by-two contingency tables and multiple logistic regres average. 82 g protein and 16.1 MJ/d (P:E, 8.5%), including 152 sion were used to assess the effect of the 2 diets and the mineral g oil and 95 g lactose. The cost of this diet was =US$1.00 per supplement on mortality. This allowed us to control for the person daily. During the first few days after admission, the milk potential confounding effects of morbidity (edema, lower respi component of this diet was diluted to one-third strength with the ratory tract infection, dysentery, dehydration, and anemia) and WHO formula ORS. This reduced the total daily average offered other variables (age, sex, and time since opening of the center) to 35 g protein and 10.9 MJ (P;E. 5.0%). Initially, the potassium on mortality. The time variable was included in the analysis to and sodium contents of the two diets were similar. However, dur control for the effect of any time bias in the data, as it was con ing January 1993 mineral supplements became available and ceivable that patient care, independent of the introduction of the patients receiving the LP diet received (per kg body wt-d“‘) 1 LP diet, improved over time. Variables not independently asso mmol KCl, 1 mmol tripotassium citrate, 0.4-0.8 mmol MgSO^, ciated with mortality were excluded from models in a stepwise 0.031 mmol zinc acetate, and 0.003 mmol CuCL, raising the fashion using estimation techniques. amount of potassium they received by 2 mmol ■ kg body Analysis of effects of HP and LP diets on rates of weight wt“‘ -d"*. For those patients who required rehydration, usually change and loss of edema during rehabilitation only during the initial phase of treatment, the WHO formula ORS was used. It was not feasible to quantify the amount of The rate of weight change (g • kg“* d"') for each consecutive ORS and, therefore, the additional sodium and potassium con 3-d period was calculated for each patient. In those edematous sumed during these periods. patients who changed from the HP to the LP diet after they had On 5 December 1992, the LP diet was first offered to 11 ede been in the center for > 3 d and for whom there were sufficient matous patients who had not responded to the HP diet. The pos data (rt = 7), a matched analysis using a paired ( test o f the rate itive response in these patients to the change of diet was so dra of weight change during the first 15 d of treatment on each diet matic that it was soon considered unethical to use the HP diet in was performed. the treatment of edematous patients. Thus, from 7 December 1992 on, patients with edema and those whom the clinician (SC) considered very ill or moribund were offered the LP diet during RESULTS the initial phase of treatment. The HP diet continued to be used General patient data from admission in the less severe cases and for many patients during the recovery phase of rehabilitation. The onset of this The degree of emaciation and clinical condition of the 573 patients recovery phase was defined clinically in edematous patients by a at the center were reported elsewhere (19). Mean BMI on admission

TABLE 3 Outcome data associated with use of higher-protein (HP) and lower-protein (LP) diets

All patients Edematous patients Marasmic patients Total Died Survived Lost to follow-up Total Died Survived Total Died Survived

HP diet 343 78 240 25 27 14 13 291 64 227 LP diet 144 25 117 2 56 14 42 86 11 75 Total 487 103 357 27 83 28 55 377 75 302

i m m s s f 196 COLLINS ET AL

Edematous 10 10

8 8

6 6

4 4

2 2

0 0

•2 •2

-6 6

•8 -8

-10 •10 03 6 9 12 15 18 21 24 27 30 33 36 0 6 9 12 15 18 21 24 27 30 33 363

Days from start of diet Days from start of diet

FIGURE 1. Mean (± 1 SEM) rate of weight change in edematous (n = 45) and marasmic (n = 219) patients receiving the lower-protein (■ ) and higher-protein ( • ) diets during the first 33 d of treatment. was 13.1 (95% Cl: 12.9, 13.3); mean body weight was 35 kg. Four weight loss, accelerated after transfer to the LP diet [mean hundred thirteen (72%) of these patients survived, 122 (21%) died, (± SD) rate of weight change: 6.3 ± 12.1 and —7.2 ± 18.5 and 38 (7%) were lost to follow-up. Edematous malnutrition was com g kg"' j-i with the HP and LP diets, respectively; difference mon and was associated with increased mortality. It was present in = 13.5; paired t = 3.18, P <0.05]. The appetite of many of the 16% of all admissions and 28% of those who subsequently died. Case edematous patients appeared to be poor with the HP diet. When fatality rates were 37% for edematous malnutrition and 20% for these patients were switched to the LP diet, an increase in marasmus or mild edema [odds ratio (OR): 2.4; 95% Cl: 1.4, 3.9; appetite was observed in many, including those who had suf Yates’s corrected chi-square: 11.7, P <0.001]. The prognosis in ede fered from persistent edema lasting several weeks while they matous malnutrition was worse for male than for female patients (male received the HP diet. Associated with the increased appetite OR: 4.1, 95% Cl: 1.9, 8.7; female OR: 1.4, 95% Cl: 0.6, 2.9; Woolf’s were episodes of watery diarrhea. These were occasionally test for the heterogeneity of odds ratios: 4.59, P <0.05). Nineteen per severe, resulting in the rapid appearance of intravascular hypo cent of deaths occurred within 48 h of admission. 48% within the first volemia. Patients with watery diarrhea responded to dilution of week, and 27% during the second week after admission. Median time the milk element of the diet with the WHO formula ORS dur to discharge for all patients was 28 d (35 d in edematous patients, 27 ing the first few days of treatment. The number of days and the d in marasmic patients). Mean BMI at the time of discharge was 15.0. degree to which the milk was diluted were tailored for each patient according to the severity of the diarrhea, clinical signs Effect of HP and LP diets on mortality of intravascular hypovolemia (heart rate, peripheral perfusion, Mortality was lower in patients with edematous malnutrition jugular venous pressure, and hydration of mucous membranes), who received the LP diet during the initial phase of treatment and the response to previous dilutions. On occasion, it was (crude OR: 0.31; 95% Cl: 0.10, 0.90; Yates’s corrected chi- necessary to dilute the milk to one-ninth strength for several square: 4.74, P < 0.05) than in those who received the HP diet. days to reduce diarrhea and maintain the appropriate intravas This effect remained after confounding variables were adjusted cular volume. Once the diarrhea resolved, the milk concentra for by logistic regression (adjusted OR: 0.31; 95% Cl: 0.12, 0.81; tion was gradually increased to full strength over 1 wk. likelihood ratio statistic: 5.74, P <0.05). This is equivalent to a Weight change during rehabilitation threefold reduction in mortality. No such difference was observed in marasmic patients (crude OR: 0.52; 95% Cl: 0.24, 1.08; The mean rates of weight change for edematous and marasmic Yates’s corrected chi-square: 2.97, P = 0.08, and adjusted OR: patients during the initial month of treatment on each diet are shown 0.87; 95% Cl: 0.38, 1.97; likelihood ratio statistic: 0.10, P = in Figure 1. During the first 9 d, edematous patients receiving the 0.74). Analysis of the effects of the mineral mix in patients LP diet tended to lose weight as they lost edema. By contrast, ede receiving the LP diet did not uncover any significant differences matous patients receiving the HP diet tended to gain weight. After in mortality before and after addition of the mineral supplement. this period the situation reversed and those receiving the LP diet started to gain weight whereas those receiving the HP diet stayed at Effect of HP and LP diets on loss of edema and the same weight or gradually lost weight. Marasmic adults gained gastrointestinal function weight similarly with both diets during this initial period. In the 7 patients for whom matched data for rate of weight During the recovery phase (16-60 d after the start of. treat change were available, the rate of loss of edema, reflected as ment), mean (± SD) rates of weight change were similar in both DIETARY TREATMENT OF SEVERE MALNUTRITION 197 marasmic and formerly edematous patients (rate of weight pensated for the decreased energy content of the diluted diet. change: 6.1 ±5.3 and 5.1 ±4.2 g-kg“*-d“', respectively) and for Indeed, it is likely that energy intake during the initial phase of the HP and LP diets (rate of weight change: 5.9 ± 5.6 and 5.6 ± treatment with the LP diet was higher than with the HP diet. This 4.8 g • kg~* • respectively). agrees with documented experiences in feeding severely mal nourished children in Jamaica (24). Immediately after the change to the LP diet, most of the ede DISCUSSION matous patients who had been in the center for some weeks Importance of lower-protein diets during the initial developed watery diarrhea. This diarrhea appeared to be the treatment of severe malnutrition in adults refeeding diarrhea frequently described in new arrivals at feed ing centers and generally regarded as the response of a starving In this study, mortality was threefold higher in edematous and atrophic intestine to the initial réintroduction of food (25). patients receiving the HP diet than in those receiving the LP diet. The absence of refeeding diarrhea among the edematous patients Many of these patients also experienced prolonged anorexia and receiving the HP diet and its development when these patients persistent edema. The LP diet was associated with lower mortal were transferred to the LP diet supports the observation that ity and accelerated loss of edema. These differences appear intake of the HP diet in these patients was low. likely to have been due to the different amounts of dietary pro Other differences in composition of the two diets could also tein offered during the initial phase of therapy. have contributed to the difference in mortality. The HP diet had Other potential explanations, however, must also be exam more lactose, which could possibly have increased mortality. ined. The study compared groups of patients differing primarily This is also unlikely, though, because the lower incidence of according to the diet they received. However, allocation of diarrhea among patients receiving the HP diet indicates that lac patients to the diets could not be randomized and it is possible tose intolerance was not a major problem. that the differences in outcome ascribed to differences in diet In marasmic patients, the diet given during the initial phase of were instead due to differences in time since the center opened treatment did not affect mortality or initial rate of weight change. or to differences in the subject groups assigned to each diet. In This suggests that the dietary differences interacted with meta the analysis of mortality, the use of multiple logistic regression bolic differences between marasmus and edematous malnutrition. allowed us to control for differences in time. With respect to sub One of the main metabolic differences is in liver function. In this ject differences, during the first 6 wk of the study, all patients study, few of the marasmic but many of the edematous patients received the HP diet; during the next 16 wk, those patients had clinical signs consistent with liver failure. These included assigned to the LP diet were edematous or otherwise very ill or petechial rash and jaundice as well as anorexia, edema, and moribund. On clinical grounds, these patients were expected to ascites. Thus, it is possible that there was no difference in out have a higher mortality than those assigned to or already recov come associated with diet in marasmic patients because their liver ering satisfactorily on the HP diet. However, the study showed function was better preserved. If correct, this again suggests that that patients receiving the LP diet had a threefold lower rate of it was the protein content of the diet that was of importance. mortality and a more rapid loss of edema than patients receiving The study failed to show any effect of the mineral supple the HP diet. Thus, it is possible that had we randomly assigned ments in patients given the LP diet. This may have been due to patients the difference in outcome would have been in the same the study design, the aim of which was not to test the effect of direction but even greater. This is compelling evidence that the these supplements. However, it is also possible that the mineral difference in outcome was due to the difference in diet. intake from the LP diet alone was adequate. This is likely to be The diets, however, differed in more than their protein con true for potassium, dietary intake of which was usually tents and it is also probable that intake of the two diets differed. > 2 mmol • kg“* • d~ but unlikely to be so for the other minerals. Lower energy intake, particularly energy derived from carbohy Although recovery from severe adult malnutrition has been drate, during the initial treatment with the LP diet is a possible systematically studied only rarely, there have been reports of explanation of the differences in outcome. Lower energy intake similar poor responses to HP diets wherein edema was slow to might be expected to lower mortality by reducing the incidence disappear and sometimes even appeared during treatment of refeeding syndrome (23). Although the aim was to offer diets (26-28). Ex-inmates of Belsen concentration camp, who experi of similar energy contents (Tables 1 and 2), the LP diet had a enced degrees of weight loss comparable with those in patients greater proportion o f its energy derived from fat and was usually in Baidoa (mean weight loss of 38%, corresponding to a BMI of given in a more dilute form during the first few days of treat 12-14) frequently suffered from famine edema and ascites (9). ment. This dilution substantially reduced the energy being During rehabilitation with a diet containing 64.8 g protein and offered to patients during that time. Although the recovery milk 3.4 MJ (P:E, 32.4%), famine edema often appeared or increased element of the HP diet was also diluted during the first few days (10). In these patients, use of casein hydrolysates was associated of treatment, this dilution caused less of a decrease in the energy with increased mortality (29). Similar accounts were reported content of the HP diet as a whole. These differences, however, from the Dutch famine of 1945, for which the recommended are unlikely to explain the higher mortality associated with the rehabilitation diet contained 300 g protein and 13.4 MJ HP diet because as a result of persistent anorexia most of the (P:E, 37%) (8). edematous patients receiving the HP diet consumed only a small Rates of weight change during the recovery phase of fraction of the food offered. Their food intakes, although not for rehabilitation mally measured, appeared to be very low, making refeeding syn drome unlikely. The marked increase in appetite, and thus pre The extreme levels of disruption and insecurity present in sumed increase in food intake, observed in these patients when Baidoa during the time of this study made the operation of the they were offered the LP diet is likely to have more than com center difficult. As a result, the mean rates of weight gain 198 COLLINS ETAL

(5-6 g-kg" ' •(!“’) during the recovery phase of rehabilitation mortality between patients receiving LP and HP diets. During the probably represent the lower end of the spectrum of reasonable recovery phase, there were no differences in rates of weight gain rates of weight gain in adults recovering from severe malnutri between the LP and HP diet groups. tion. Severely malnourished ex-inmates of the Sanbostel con Thus, it appears that, compared with the conventional HP diet, centration camp receiving 31.5 MJ and 297 g protein/d (P;E, the LP diet was more effective in the treatment of edematous 15.8%) gained « 7 g-kg“L(j-> during the recovery phase of their adults and as effective in the treatment of marasmic adults. The treatment (30). In less severely malnourished patients, rates of LP diet usually needed to be diluted during the initial phase of weight gain appear to be lower. In the Minnesota experiment, 32 treatment. It was cheaper and more easily obtained than the spe volunteers with mild to moderate malnutrition who received cially manufactured famine relief foods of the HP diet. We sug 10-14 MJ and 75-100 g protein/d (P:E, 10.6-14.4%) gained only gest that such a diet, based on milk, oil, sugar, and locally avail 1.85 g kg-' , d '(31). able foods, with a relatively low P:E, should be offered to all severely malnourished adults in both the initial and recovery Comparison with children phases of rehabilitation. By using locally available produce, use The patterns of presentation and recovery in severely malnour of such diets may stimulate the local economy. Their use should ished adults are similar to those in children. In children, hypoalbu- also ease the organizational difficulties involved in the provision minemia is evidence of a poor prognosis (32). Although edema, in of food items during emergency feeding operations. More the absence of hypoalbuminemia, is not necessarily an indicator of research on treatment protocols for use in severe adult malnutri a poor prognosis (14), in practice, the frequent coexistence of the tion is needed. 13 two makes edema a useful prognostic marker. In Baidoa, edema in adults was associated with a much poorer prognosis (19). However, We thank the Concern Worldwide staff and the patients in Baidoa, who it is possible that this finding is not universal. In a Concern World made this study possible; Ben Rigby, for his help with data coding and entry; and Michael Golden, Hannah Scrase, Jeremy Shoham, Keith Sullivan, and wide TFC in Melange, Angola, during 1993 and 1994, 90% of Andrew Tomkins, for their advice and encouragement. adults admitted suffered from edema but this was not associated with such a poor prognosis (S Collins, unpublished observations, REFERENCES 1993). More work is needed on prognostic indicators in severe adult malnutrition. 1. Collins S. The need for adult therapeutic care in emergency feeding The maximum rates of weight gain, typically 10-20 g- kg“' • d“', programs. JAMA 1993;270:637-8. recorded in children recovering from severe malnutrition (14, 33, 2. Winick M. Hunger disease. New York: Wiley-Interscience, 1979. 34) are considerably higher than the rates of weight gain reported 3. Moore PS, Marfin AA, Quenemoen LE, et al. Mortality rates in dis here. However, the pattern of recovery is similar. Initially, with a placed and resident populations of central Somalia during 1992 low-protein, maintenance energy intake, edematous children ofteri famine. Lancet 1993;341:935-8. lose edema within 1 wk (16). Appetite returns and this, together 4. Population based mortality assessment: Baidoa and Afgoi, Somalia 1992. MMWR Morb Mortal Wkly Rep 1992;41:913-7. with loss of edema, heralds the recovery phase. Among our edema 5. Collins S. Famine in Somalia. Lancet 1993;341:1478-9. tous adult patients, the rates of edema loss were variable and often 6. Digby W. The famine campaign in southern India. London: Long much slower. With the LP diet, some patients lost most of their vis mans, Green & Co, 1878. ible edema and ascites within a few days, the rapid loss generally 7. McCance RA, Widdowson EM. Studies in undemutrition, Wuppertal being accompanied by watery diarrhea. In these patients, care had 1946-9. London: Medical Research Council, 1951. to be taken to avoid intravascular hypovolemia. In the absence of 8. Burger GCE. Sandstead HR, Drummond J. Starvation in western guidelines for adults, we aimed at a loss of =0.25-0.5 L/d, equiva Holland: 1945. Lancet 1945;2:282-3. lent to a weight loss of 0.25-0.5 kg/d. Regulation of the rate of 9. Mollinson PL. Observation on cases of starvation in Belsen. Br Med J 1946;5:4-8. edema loss was achieved by diluting the high-energy milk with 10. Lipscomb PM. Medical aspects of Belsen concentration camp. ORS to an extent dictated by the severity of diarrhea. In other Lancet 1945;2:313-5. patients, particularly those receiving the HP diet, the rate of loss of 11. Leyton GB. The effects of slow starvation. Lancet 1946;2:73-9. edema was much slower and pedal edema or ascites persisted for 12. Beattie J, Herbert PH, Bell DJ. Famine edema. Br J Nutr weeks. This was accompanied by persistent anorexia and debility. 1948;2:47-65. These patients generally responded rapidly to introduction of the LP 13. Waterlow JC. The rate of recovery of malnourished infants in rela diet. The response involved marked increase in appetite and general tion to the protein and calorie levels of the diet. J Trop Pediatr 1961;7:16-22. well-being and loss of edema and ascites. However, minor grades of 14. Waterlow JC. Protein energy malnutrition. London: Edward Arnold, edema sometimes persisted for weeks, even after the patients had 1993. recovered much of their original body mass. When edema increased 15. Waterlow JC. Fatty liver disease in the British West Indies. Med Res even with the LP diet, the patient usually died. Counc Spec Rep Ser (Lond) 1948;263. 16. Golden BE. Primary protein energy malnutrition. In: Garrow JS, Conclusions James WPT, eds. Human nutrition and dietetics. 9th ed. London: Because of the extreme conditions in this TFC, which was Churchill Livingstone, 1993:440-55. 17. Golden MHN. Severe malnutrition. In: WeatheraU DJ, Ledington operating in a war zone, the subjects in this study were not ran JGG, Warrell DA, eds. The Oxford textbook of medicine. 3rd ed. domly assigned to the 2 diets, nor were their dietary intakes esti Oxford, United Kingdom: Oxford University Press, 1996:1278-96. mated. During most of the study, those with a worse prognosis 18. Barac-Nieto M, Spurr GB, Lotero H, Maksud MG, Dahners HW. on admission received the LP diet. Notwithstanding, with the LP Body composition during nutritional repletion of severely under diet, edematous adults suffered lower rates of mortality and lost nourished men. Am J Clin Nutr 1979;32:981-91. edema more quickly than did those receiving a more conven 19. Collins S. The limit of human adaptation to starvation. Nat Med tional HP diet. In marasmic adults, there was no difference in 1995;1:810-4. DIETARY TREATMENT OF SEVERE MALNUTRITION 199

20. Dean AG, Dean JA, Burton AH, Dicker RC. EPI INFO version 6: a 27. Henschel A, Mickelsen O, Taylor HL, Keys A. Plasma volume and the word processing, database and statistics program for epidemiology thiocyanate space in famine edema and recovery. Am J Physiol on microcomputers. Stone Mountain, GA: USD Incorporated, 1990. , 1947;150:170-80. 21. Dallai GE. Logistic: a logistic regression program for the IBM PC. 28. Stapleton T. Edema in recovered prisoners of war. Lancet 1946;1:850-1. Am Stat 1988;42:272. 29. Vaughan J, Pitt Rivers R. The value of hydrolysates in the treatment of 22. Microsoft Corporation. Excel. Redmond, WA: Microsoft Corpora severe starvation. Proc R Soc Med 1945;38:395. tion, 1994. 30. Murray RO. Recovery from starvation. Lancet 1947;1:507-11. 23. Weinsier RL, Krumdieck CL. Death resulting from overzealous total 31. Keys A. Body weight. In: Keys A, ed. The biology of human starvation. parenteral nutrition: the refeeding syndrome revisited. Am J Clin Minneapolis: University of Minnesota Press, 1950:84-129. Nutr 1981;34:393-9. 32. Vis HL. On the treatment of certain forms of protein energy-malnutri- 24. Golden MH. Protein deficiency, energy deficiency and the edema of üon in childhood with respect to fatal complications (an example from malnutrition. Lancet 1982;1:1261-5. rural Central Africa). Ann Nestle [Fr] 1985;43:19-30. 25. Roediger WEW. Metabolic basis of starvation diarrhea: implications 33. Ashworth A. Growth rates in children recovering from protein-calorie for treatment. Lancet 1986;1:1082-4. malnutrition. BrJ Nutr 1969;23:835-45. 26. Widdowson EM. The response to unlimited food. In: McCance RA, 34. Waterlow JC, Golden MHN, Patrick J. Protein-energy malnutrition: Widdowson EM, eds. Studies of undemutrition, Wuppertal 1946-9. treatment. In: Dickerson JWT, Lee HA, eds. Nutrition in the clinical London: Medical Research Council, 1951:313-47. management of disease. London: Edward Arnold, 1978:49-71.