THE IMPORTANCE OF BIRTH IN KATHMANDU, .

Thesis submitted by

Dr Matthew Edward Ellis Institute of Child Health

towards the degree of

Doctor of Philosophy

University of London.

1999 ProQuest Number: 10797847

All rights reserved

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 com plete 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 10797847

Published by ProQuest LLC(2018). Copyright of the Dissertation is held by the Author.

All rights reserved. This work is protected against unauthorized copying under Title 17, United States C ode Microform Edition © ProQuest LLC.

ProQuest LLC. 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106- 1346 Institute of Child Health University of London 30 Guilford Street, London, WCIN !EH 2

Abstract Background Birth asphyxia is considered to be an important cause of and neurodevelopmental impairment but epidemiological data from low income countries is scarce. Objectives To describe the prevalence, risk factors and outcome of birth asphyxia in the main maternity hospital in Kathmandu, Nepal where more than 50% of local mothers deliver.

Methods The prevalence of birth asphyxia Design: Prospective descriptive cross-sectional survey Subjects: 14,771 births Outcomes: Fresh (SB) of 2000 g or more Neonatal encephalopathy (NE) in term Perinatal Risk factors for neonatal encephalopathy and fresh in term infants Design: Retrospective case control study Subjects: 131 NE cases 1107 fresh stillbirths of 2000g or more / 635 controls Outcomes following neonatal encephalopathy Design: Prospective cohort study Subjects: 131 NE cases / 208 controls Outcomes: Death / Neurodevelopmental impairment at one year Results The prevalence of birth asphyxia The prevalence of fresh stillbirth and NE (as measures of birth asphyxia in term infants) were 8.5 per 1000 total births (95% Cl 7.1-10.1) and 6.4 per 1000 live births (95% Cl 5.2-7.8) respectively. Birth asphyxia in term infants accounts for 24% (95% Cl 21-28%) of the total perinatal mortality of 44 per 1000 total births in the study population. Evidence of intrapartum compromise was more likely during the labours of fresh stillbirths (76%) and NE infants (68%) than of controls (12%).

Risk factors for neonatal encephalopathy and fresh stillbirth Multivariate analysis found significant (p<0.05) independent risk factors to be a) for NE: low maternal educational status, low maternal height, previous neonatal death, lack of antenatal care, prolonged rupture of membranes and/or smelly liquor, 3 non-cephalic presentation, exposure to syntocinon, large head size and low infant weight. b) for fresh SB: low maternal educational status, multiparity, previous stillbirth, prolonged rupture of membranes and/or smelly liquor, non-cephalic presentation, induction with syntocinon and low infant weight.

Outcomes following neonatal encephalopathy 102 NE cases (78%) and 106 controls (51%) were followed up to one year of age. The relative risk of death (95% Confidence Interval) for mild, moderate and severe NE compared to controls was 4.5 (1.4-15), 7.7 (2.6-22) and 25.6 (9.8-67). The excess risk of death was limited to the neonatal period for all grades of NE. Of 27 survivors of moderate NE 16 (59%[95% Cl 39-78]) developed major neurodevelopmental impairment by one year of age. The prevalence of disability at one year of age following birth asphyxia at term in this population is no more than 1 per 1000 birth cohort.

Conclusions Birth asphyxia is an important cause of perinatal mortality in low income countries but accounts for less childhood disability than has previously been estimated. This study failed to identify a simple and affordable public health measure likely to have a major impact on the prevalence of birth asphyxia. The induction of delivery using syntocinon in the absence of continuous fetal and uterine monitoring may be dangerous to the . Available resources are best aimed at the extension of obstetric services to reach more precisely targeted risk groups at the outset of labour. Closer attention to the management of intrapartum meconium both in the uterus prior to delivery and in the newborn airway immediately after delivery may help to improve outcome. Improvements in perinatal care in low income countries are unlikely to have a major impact on the prevalence of childhood disability. Acknowledgements Dr Anthony Costello conceived this project, formulated the early hypotheses and study design and drew up the funding proposal. As my principal supervisor he guided my development as a research fellow in international child health and ensured the experience was not only stimulating but also enjoyable. Dr Dharma Manandhar, in whose neonatal unit the study was conducted, made the whole project possible with wise advice at all stages of the study. He shared many insights concerning the provision of appropriate neonatal care and the wider social and cultural issues facing Nepal. Dr Ruth Gilbert encouraged my early exposure to epidemiological methods and contributed to study design. Professor John Wyatt assisted with the design of the neonatal examination protocols and made helpful suggestions throughout the study. Hospital Directors Drs Laxmi Shrestha and Saraswati Padye supported the project whilst Dr Kasturi Malla contributed helpful obstetric advice. Dr Nilu Manandhar, my principal clinical counterpart in Nepal, and Dr Urmila Sakya helped pilot the initial protocols and with other junior colleagues ensured data collection ran smoothly. Drs Prakash and Laxman Shrestha performed neurodevelopmental screening with unfailing patience. Ms Puma Shrestha and Bimala Manandhar helped with data entry and Meena, Reena, Tara and Madan were tireless in their efforts in tracking down patients. Colleagues at the Centre for International Child Health, London provided stimulating discussion whilst Juana Willumson ensured all the blood samples reached London. Dr John Land arranged the assays of the blood samples. Dr Liz Paul reviewed the statistical procedures. Colleagues both in Ealing and Bristol were generous in permitting occasional space from clinical duties whilst writing up the study. The study was performed under the auspices of the Maternal and Infant Research Activities (MIRA) non governmental organisation registered in Kathmandu. The project was funded throughout by the Wellcome Trust. During the project period MIRA also received support from the Overseas Development Agency (now the Department for International Development) of the UK government. I was given official support by the Ministry of Health of His Majesty's Govemement of Nepal. Throughout the four years Alison and I lived in Nepal we enjoyed the goodwill and assistance of innumerable Nepali men, women and childen. The births of our daughters, Harriet and Kate, and the of Mabel, my mother, and Jon, my eldest brother, all occurred whilst this work was in progress. Shortly after completion my father George died. Nepal has helped me to understand a little more of the wheel of life and death. This work is intended to nudge that wheel towards life in some small way in Nepal. It is dedicated to my family, Nepali friends and colleagues and to the memoiy of my parents, George and Mabel. ma samjhanchu, namaste Bristol, 1999. 5

Contents Chapter Pages Abstract 2-3 Acknowledgements 4 Contents 5-6 List of Tables 7-10 List of Figures 11 Introduction 12-13 1. Birth asphyxia 14-46 Biology of birth asphyxia 14-22 Clinical recognition of asphyxia 22-34 Birth asphyxia and 35-40 Risk factors associated with birth asphyxia 40-44 Investigations for birth asphyxia 45-46 Conclusion 46 2. International perinatal epidemiology 47-62 Perinatal mortality and birth asphyxia 47-48 Neonatal morbidity 49 Risk factors for birth asphyxia in low income countries 50-51 Outcome studies 51 Maternity service delivery 51-55 Intrapartum management 56-57 Industrialised country comparative data 58-60 Study hypotheses 61-62 3. Study design, setting and methodology 63-93 Study design 63-66 Setting 66-71 Methodology 71 -90 Analysis 90-93 Ethical approval, informed consent and competing interests 93 4. Prevalence of fresh stillbirth and neonatal encephalopathy 94-98 Perinatal vital statistics 94 Stillbirth rates 94-95 Neonatal encephalopathy prevalence 95 Neonatal clinical characteristics 95 Low Apgar scores as a predictor of neonatal encephalopathy 96 Adverse intrapartum events and birth outcome 96-97 Grading and early neonatal outcome of neonatal encephalopathy 97 6 Contents (cont.) Perinatal mortality and birth asphyxia 97-98 Key findings 98 5. Risk factors for neonatal encephalopathy and fresh 99*117 stillbirth in term infants Comparison of control groups 99 Clinical risk factors 99-102 Risk factors for neonatal encephalopathy 102-104 Risk factors for fresh stillbith 105-106 Subgroup analysis 106-107 Laboratory data 108-114 Analysis of neonatal factors affecting grade of outcome 114-115 Seasonality 116 Key findings 117 6. Neurodevelopmental outcome of neonatal encephalopathy 118=142 at one year Clinical characteristics of the infant population 118-123 Clinical outcome 123-132 Anthropometry 132-134 Cranial ultrasonography 135-136 Predictors of major neurodevelopmental impairment at 1 year 137 Lost to follow up 137-140 Estimated contribution of birth asphyxia to childhood disability 141 7. Implications for interventions in low income countries 142-160 Principal findings 142-144 Limitations 145-150 Generalisability 150-152 Implications 152-158 Future research priorities in low income settings 158-159 Conclusions 159-160 Bibliography 161-187 Appendix 1 Neonatal neurological examination protocol 188-190 Appendix 2 Neonatal neurological examination data form 191 Appendix 3 Perinatal clinical data form 192 Appendix 4 NE management protocol 193-194 Appendix 5 Parametric data distribution 195-199 Appendix 6 Reviews of low cost predictors of impairment in NE infants 200-202 Appendix 7 Publications and Presentations 203-205 7

List of Tables Page Table 1. Review of major studies of birth asphyxia in which early neurological 28 signs have been entry criteria or the subject of the study. Table 2. The New York post-asphyxial score. 30 Table 3. Risk factors shown to be significantly associated on univariate 40 analysis (p<0.05) with birth asphyxia (variously defined) in fo. Table 4. Comparison of several markers of birth asphyxia with a measure of 45 umbilical artery base concentration as a gold standard. Table 5. Review of 20 perinatal mortality studies published over the last 15 48 years which have assessed the relative importance of birth asphyxia in developing countries. Table 6. Comparison of low income studies of neonatal encephalopathy. 49 Table 7. International comparison of the prevalence and early outcome of 58 neonatal encephalopathy in high income countries. Table 8. Sample size calculations showing detectable odds ratios for exposures 64 of potential interest based on different sample size and control:case scenarios. Table 9. Operational definitions for the clinical grading of neonatal 73 encephalopathy. Table 10. Inter-rater agreement for neonatal encephalopathy grade. 75 Table 11. Clinical characteristics of 11 infants excluded with evidence of early 83 neonatal encephalopathy secondary to perinatal sepsis. Table 12. Classification of cerebral palsy. 89 Table 13. Criteria used in the definition of definite and possible intrapartum 91 suboptimality. Table 14. Perinatal vital statistics for Prasuti Griha in 1995. 94 Table 15. Stillbirths analysed by birth weight, estimated gestation and clinical 94 condition from 14,771 deliveries at Prasuti Griha in 1995. Table 16. Birth details of 92 term infants with neonatal encephalopathy, 86 fresh 95 stillbirths of a weight of 2000g or more and 98 healthy term infants identified from 14,771 live deliveries at Prasuti Griha in 1995. Table 17. Review of the use of low one minute (less than or equal 96 to three) as a test for NE. Table 18. An analysis of the association between adverse outcome 97 and intrapartum adversity using two different measures. Table 19. Severity of neonatal encephalopathy with associated early neonatal 97 mortality risk. Table 20. Distribution of risk factors of interest amongst the four study groups. 100 Table 21. Univariate analysis of risk factors for neonatal encephalopathy 103 Table 22. Final model of risk factors for NE comparing 131 NE infants with 104 635 later controls. Table 23. Univariate analysis of risk factors for fresh stillbirth. 105 Table 24. Final model of risk factors for fresh stillbirth comparing 131 NE 106 infants with 635 later controls. Table 25. Rerun of model of risk factors for NE comparing 76 NE cases and 107 521 later controls in Kathmandu residents only. Table 26. Rerun of model of risk factors for NE comparing 105 NE cases and 107 594 later controls in planned deliverers only. Table 27. Maternal haemoglobin by study group. 108 Table 28. Odds Ratios for puerperal anaemia (HB<10) comparing 128 NE and 108 102 fresh stillbirths with 614 later controls. Table 29. Adjusted odds ratios for puerperal anaemia (HB<10) comparing 128 109 NE and 102 fresh stillbirths with 614 later controls. Table 30. Mean values (standard deviations) for biochemical measures of 110 maternal postpartum thyroid function by study group. Table 31. Comparative unpaired analysis of biochemical measures of thyroid 110 function comparing mothers of encephalopathic infants and fresh stillbirths with contemporary control population. Table 32. Mean values (standard deviations) for magnesium by study group. I ll Table 33. Comparative unpaired analysis of maternal serum magnesium levels 111 comparing mothers of encephalopathic infants and fresh stillbirths with contemporary control population. Table 34. Mean values (standard deviations) for maternal postpartum acute 112 phase proteins by study subgroup. Table 35. Comparative unpaired analysis of biochemical measures of acute 112 phase response comparing mothers of encephalopathic infants and fresh stillbirths with contemporary control population. Table 36. Acute phase proteins in contemporaneous controls and cases by 113 associated risk factors for sepsis. Table 37. Acute phase proteins by mode of delivery for all study mothers. 113 Table 38. Odds ratios for raised acute phase proteins (CRP>120 mg/1 and AGP 114 > 1.57 g/1) comparing NE cases with controls adjusting for time of sampling, mode of delivery and maternal risk factors for occult sepsis. 9

Table 39. Relationship between detected neonatal hypoglycaemia and grade of 115 neonatal encephalopathy. Table 40. Univariate analysis of perinatal infant characteristics comparing 131 119 NE cases with 208 contemporaneous controls. Table 41. Clinical details of co-twins of NE cases. 120 Table 42. Early clinical characteristics of the cases of NE. 120 Table 43. The distribution of neonatal deaths by grade. 123 Table 44. Clinical details of all hospital deaths following NE (n=40). 124 Table 45. Clinical progress and for all 9 postneonatal deaths. 126 Table 46. Developmental screening and initial neurological assessment 127 findings at 12 months of age for 102 control infants and 57 NE infants. Table 47 (a) Developmental screening findings and (b) Neurological screening 128 findings by grade of NE for 57 case infants at 12 months of age. Table 48. Neurodevelopmental findings for NE screening failures at one year. 130 Table 49. Outcome of cohort (102 NE cases and 106 controls) successfully 132 followed up to one year of age. Table 50. Relative risks of death (infant and postneonatal) or impairment (any, 132 major or minor) by grade of NE in comparison with control group. Table 51. Change in SD growth scores (ASD) over the first year of life 134 comparing growth in unimpaired (n=38) and impaired survivors (n=17) of NE with that of controls (n=99). Table 52. Pattern of echogenicity on cranial ultrasound by NE grade. 135 Table 53. Six week ultrasound findings (n= 72 cases and 114 controls). 136 Table 54. Comparative review data of the properties of various low cost 137 predictors of neurodevelopmental impairment in encephalopathic infants. Table 55. Clinical characteristics of the case and control populations comparing 138 those who completed 12 month follow up with those lost to follow up. Table 56. Proportion impaired at last point of clinical contact by NE grade and 139 age at last point of contact. Table 57. Criteria for the identification of high risk mothers. 156 Appended Tables: Appendix 6

Table 58. Review of the use of NE grade for the prediction of major 200 neurodevelopmental impairment in NE infants at one year. Table 59. Review of the use of abnormal neurological examination at discharge 200 in NE infants for the prediction of major neurodevelopmental impairment at one year. Table 60. Review of the use of echogenic appearance of the neonatal brain in 201 NE infants for the prediction of major neurodevelopmental impairment at one year. Table 61. Review of the use of cystic changes on cranial ultrasound in NE 201 infants at six weeks of age for the prediction of major neurodevelopmental impairment at one year. Table 62. Review of the use of acquired microcephaly (defined as a negative 202 change in the Occipito-frontal circumference SD score of two or more at six weeks) in NE infants for the prediction of major neurodevelopmental impairment at one year. 11

List of Figures Page Figure 1 Possible causal pathways relating prenatal risk factors, intrapartum 35 birth asphyxia, neonatal encephalopathy and cerebral palsy. Figure 2. Perinatal mortality (PNMR per 1000 livebirths) and proportion of 52 deliveries with trained attendant by subregion (WHO 1995 data). Figure 3. Time bar plot depicting the Kathmandu study process. 66 Figure 4. Interaction line plot for infant haemoglobin and NE grade split by 115 presence or absence of cephalhaematoma. Figure 5. Comparison of month of delivery of NE cases and later controls 116 Figure 6. Comparison of month of delivery of NE cases and later controls. 116 Figure 7. Temporal course of disease progression and resolution for newborns 121 with neonatal encephalopathy. Figure 8. Flow chart of cohort follow up. 122 Figure 9. Survival curve to one year of age (102 NE cases and 106 controls). 123 Figure 10. Cerebral palsy classification by pattern of involvment. 129 Figure 11 Growth trends for (a) infant weight (b) infant length and (c) infant 133 head circumference (SD score with 95% Cl bar) over the first year of life for cases and controls. Figure 12. Distribution of the average ventricular indices of case infants 136 examined between 30 and 60 days post-delivery. Figure 13. Age at last clinical contact of infants lost to follow up. 137 Appended Figures: Appendix 5 Figure 14. Distributions of maternal anthropometric data for all the study groups 195 combined. Figure 15. Distributions of infant anthropometric data for all the study groups 196 combined. Figure 16. Distribution of maternal haemoglobin for all groups combined. 197 Figure 17. Distribution of neonatal haemoglobin in NE cases only. 197 Figure 18. Distribution of maternal FT4 for all groups combined. 197 Figure 19. Distribution of maternal TSH for all groups combined. 198 Figure 20. Distribution of maternal magnesium for all groups combined. 198 Figure 21. Distribution of maternal CRP for all groups combined. 198 Figure 22. Distribution of maternal AGP for all groups combined. 199 Figure 23. Distribution of the average ventricular indices of control infants 199 examined between 30 and 60 days post-delivery. 12

Introduction Introductory comments This thesis describes the biology of birth asphyxia and explores its contribution to perinatal mortality and neurodevelopmental morbidity in children of a low income South Asian metropolitan area. It describes risk factors associated with fresh stillbirth and neonatal encephalopathy. Generalising to other low income countries in South Asia it suggests priority areas for future intervention programmes.

Chapter outlines In chapter one I briefly review current concepts of the effects of hypoxic-ischaemia on the central nervous system of the term infant. I proceed to critically review a range of approaches to the definition of birth asphyxia and describe an empirical approach for case definition based on fresh term stillbirths and the neurobehavioural signs of encephalopathy in term neonates. Risk factors for neonatal encephalopathy are discussed and set in the context of earlier work on birth asphyxia. In chapter two I review the existing body of knowledge describing birth asphyxia in low income countries. This is briefly compared with contemporary data from high income countries. I conclude with a series of hypotheses relating to birth asphyxia in low income countries. In chapter three I describe the study design and methodology used to test the study hypotheses. The setting is described both to contextualise the study and as a basis to assess its generalisability. The challenges of coordinating a three year project in a developing country setting are briefly discussed. Finally the analytical methods are described. In chapter four I present a 12 month study of the prevalence of fresh stillbirth and neonatal encephalopathy amongst a hospital delivering population in Kathmandu. In chapter five I present a study of risk factors comparing population samples resulting in either fresh term stillbirth or neonatal encephalopathy with a systematically selected population sample of term infants. In chapter six I present a cohort study comparing the neurodevelopmental outcome at one year of age of all term infants bom with neonatal encephalopathy over an eighteen month recruitment period with a control group of contemporary term infants who did not require neonatal special care. In chapter seven I summarise the principle findings of the three studies. I discuss the study limitations as well as its generalisability. The implications of the study findings are explored. I conclude with a discussion of future strategy for both research and intervention programmes to address the problem of birth asphyxia in low income countries. 13 The bibliography lists the more important sources I have consulted during the course of this research. The appendices contain those protocols, questionaires and data I consider of direct relevance to the interpretation of the study which could not be included in the body of the thesis for reasons of space. I conclude with a list of publications and presentations arising from this research project. 14

1. Birth asphyxia Biology of birth asphyxia The condition of a newborn infant is determined by a complex interaction of maternal, uterine, placental and fetal factors extending throughout pregnancy and the process of delivery. At the core of the feto-matemal unit is the process of placental gas exchange whereby oxygen from the maternal circulation and carbon dioxide from the fetal circulation passively diffuse across the placental membrane. The labour process imposes great strains on placental exchange. During normal labour uterine pressure frequently exceeds 30 mm Hg when placental exchange is abolished altogether. Studies using infra-red spectroscopic techniques during normal labour show that many infants undergo intermittent hypoxia during this process^1/ Despite this hypoxic stress most infants are bom in good condition. The human neonate is extraordinarily tolerant of hypoxia because of several protective mechanisms.

Nprmal fetal p^ygepation During intrauterine life the fetal blood oxygen tension gradually falls until by term it has a partial pressure of the order of 3.5 kPa (less than a third of that in the newborn infant/2). Fetal haemoglobin is well suited to oxygen transport at these low partial pressures which coincide with the steepest part of the fetal haemoglobin oxygen dissociation curve. This enables a high proportion of the oxygen bound to haemoglobin at the to be released to fetal tissues downstream. The relatively high fetal red cell mass and high cardiac output serves to maximize delivery of oxygenated haemoglobin to the circulation. Finally the fetal circulation maximizes oxygen supply to the coronary and cerebral circuits by the preferential shunting of oxygenated blood from the inferior vena cava via the foramen ovale to the aortic arch which supplies these vital regions. These mechanisms adequately compensate for the normally low fetal blood oxygen tension. Cardiovascular response to hvpoxia When oxygen levels fall below the compensatory capacity of these mechanisms the cardiovascular system, mediated by the sympathetic nervous system, actively redistributes the fetal circulation substantially increasing blood flow to the brain, myocardium and adrenals at the expense of the gut, and kidneys. Studies on monkey document the effectiveness of these safety mechanisms^3/ This experimental evidence suggests that not until oxygen tensions are reduced by 65% to 75% is there evidence of cardiac compromise with declining rate and . Even then does not necessarily occur unless these changes are maintained for several hours or oxygen tensions fall further to 85% to 90% of normal 15 values. The evidence from animal studies therefore suggests that cardiac impairment precedes cerebral damage^. Acute birth asphyxia The primary disturbance in birth asphyxia is a deficit of oxygen supply to the fetus. An important secondary effect is ischaemia resulting from reduced cardiac output. The limiting factor for the cardiovascular response to anoxic stress is most likely to be the extent of the myocardial store of glycogen available for anaerobic metabolism. The net result is hypoxic-ischaemic injury. Other organs tend to be adversely affected before cardiovascular decompensation affects the brain. In severely asphyxiated neonates evidence of cardiac, renal and dysfunction has been described^. Studies have described cardiac enlargement with evidence of myocardial ischaemia^, acute renal tubular dysfunction^ , hepatic dysfunction^ and pulmonary ^9^. For survivors the disturbances to these organ systems are generally reversible with little evidence of persisting cardiac, renal or pulmonary compromise The effects of birth asphyxia on the neonatal brain however may be irreversible and disabling. The mechanisms of brain injury centre on the cerebral circulation at an organ level and on biochemical energy pathways at a sub-cellular level. Effects of asphyxia on the cerebral blood supply Cerebral blood flow is normally kept constant despite variations in central blood pressure by auto regulatory mechanisms which are poorly understood. In the healthy human term neonate the limits of the auto regulatory range are not clearly defined. Volpe, recently summarising the research data, suggests the approximate auto regulatory range lies between 25 and 50 mm Hg mean arterial blood pressure^10). Within this range cerebral blood flow is kept relatively constant. There are however important regional differences in cerebral blood flow^n \ Animal studies have demonstrated that the grey matter (representing the neuronal cell bodies) of the cerebral cortex and brain stem receives a blood supply 5-10 fold that of the sub cortical white matter (representing the neuronal axons). Cortical and deep grey matter are undergoing rapid myelination in the term infant 12\ the associated high anabolic demand perhaps explaining this relative excess of blood supply. In contrast the parasagittal white matter due to its position at a relative watershed of supply between the anterior, middle and posterior cerebral artery circulations has been shown to receive an especially poor supply in the term neonate^13). If cardiac impairment causes blood pressure to fall below the lower limit of auto regulation cerebral ischaemia may result. Cerebral blood flow studies suggest auto regulation may be reversibly impaired following asphyxial insult, an impairment which may persist for several hours after an acute asphyxial insult. In the event of cardiac recovery from an acute asphyxial event the fetal/newborn brain, no longer protected by auto regulation, may experience a period of 16 hyperperfusion. This period of reperfusion coincides with the period when much of the irreversible cerebral injury discussed below occurs^ 10\ Effects of asphyxia on cellular energy metabolism Sub cellular events in the brain have been intensively studied both in animals and the human neonate. The human brain is dependent on a continuous supply of oxygen and substrate (principally glucose) from the cerebral circulation. Glycogen, the storage form of glucose, is found in relatively small amounts in brain tissue. The combination of hypoxia and ischaemia has major effects on cellular energy metabolism. Under aerobic conditions glucose enters the cell and after phosphorylation the major portion enters the glycolytic pathway in the cytosol culminating in the formation of pyruvate. This then enters the mitochondrion, is converted to acetyl-CoA which undergoes oxidation in the citric acid cycle to carbon dioxide, generating electrons in the process. The electrons enter the mitochondrial transport system which captures their energy in the bond formed when adenosine triphosphate (ATP) is generated from adenosine diphosphate (ADP) and inorganic phosphate. The final electron acceptor is oxygen which is ultimately converted to water. For each molecule of glucose oxidised 38 molecules of ATP are generated of which only two result from glycolysis within the cytosol. ATP is stored in the cytoplasm where it is in equilibrium, via the creatine kinase reaction, with phosphocreatine (PCr). The energy captured in ATP and PCr is utilised by the brain both for synthetic and ion transport purposes. Approximately 50- 60% of ATP is utilised for the maintenance of sodium, potassium and calcium ion gradients across cell boundaries to maintain homeostasis. In the absence of oxygen the citric acid cycle cannot function and the pyruvate formed by glycolysis instead goes to form lactate and hydrogen ions. For every molecule of glucose consumed anaerobically this pathway produces only two molecules of ATP and generates byproducts causing lactic acidosis. Therefore in the absence of oxygen much larger quantities of glucose must be metabolised if phosphate energy stores are to be maintained. This imposes high demands on glucose supply mechanisms. Animal studies have demonstrated that brain glucose levels fall almost immediately after the onset of asphyxia^14). Endogenous brain glucose reserves at the time of the insult determine to an important extent the eventual outcome^15). Brain glycogen levels seem far less important reflecting the relatively poor capacity of neonatal brain to rapidly mobilise glucose from glycogen stores. Brain tissue has been demonstrated to have the capacity to utilise other non-glucose substrates, including ketone bodies, pyruvate and lactate as alternative energy sources. Ketone bodies account for 12% of total cerebral oxygen consumption in term newborns subjected to six hour fasts in one study^10^ However, cerebral supply of ketone bodies is dependent on their plasma concentrations which in turn are determined by hepatic synthesis. A study amongst neonates in Kathmandu has shown that hydroxy-butyrate 17 levels (here an indicator of ketone body production) only become significant after the first twelve hours of life and that lactate may be more important in the first twelve hours of life (Costello A-personal communication). The animal evidence from dogs suggests ketone bodies are less important as a secondary source of energy under hypoglycaemic conditions^16). Amongst hypoglycaemic Kathmandu neonates there was no evidence of increased production of alternative fuels. This suggests these hypothetical alternative energy pathways may not be significantly protective of substrate deficiency during asphyxial insults in this population. Animal studies have documented that lactate levels rise rapidly (approximately 2 minutes) after the onset of asphyxia^14). The creatine kinase reaction initially buffers falls in ATP. Thus PCr levels fall rapidly whilst ATP levels withstand the first 5 minutes of asphyxia but decline thereafter. Events leading to This acute derangement of cellular energy metabolism does not necessarily precipitate cell death. Although for those cells at the heart of the insult early cell death may occur there appears to be a surrounding area (or ’penumbra') where many factors influence medium term neuronal viability^171. Studies of cerebral metabolism in neonates using magnetic resonance spectroscopy to measure the biochemical indicators of energy status demonstrate that some damaged cells return to normal energy function after the cessation of the asphyxial insult18). However a complex series of sub cellular events may be triggered by an episode of temporary energy depletion. Important components of this biochemical cascade recognised to date include the presynaptic extracellular release of the excitatory neurotransmitter glutamate, ingress into the cell cytosol of calcium, the release of free oxygen radicles, the generation of nitric oxide and the triggering of () (19\ There are complex interactions between these mechanisms resulting in a vicious amplificatory cycle of cellular injury. For example, of the four recognised glutamate receptors the one most potently triggered by N-methyl-D-aspartate, the so called NMDA receptor, is linked to an ion channel for calcium entry. Thus glutamate release triggers calcium ingress. Further glutamate release is in turn provoked by the influx of calcium into the presynaptic nerve ending. Glutamate has been shown to induce cell death by two different mechanisms^20). 'Rapid cell death' is initiated by glutamate ionotrophic receptor activation causing massive influx of sodium, passively accompanied by chloride and water resulting in cell lysis. 'Delayed cell death' appears to involve all the glutamate receptors and is in part mediated by raised calcium levels within the cell. The deleterious effects of elevated cytosolic calcium are multiple. It enters the mitochondrion and uncouples oxidative phosphorylation. This contributes to the so called secondary energy failure demonstrated some 72 hours after the primary and reversible energy failure which accompanies the asphyxial insult. Intracellular calcium also activates proteases and nucleases that directly injure the cytoskeleton and nucleus respectively. Calcium generates free radicals, highly reactive compounds with an uneven number of electrons in their outermost orbit. Free radicals are also generated directly by the electron transport system in anaerobic conditions. Free radicals cause molecular damage to DNA, proteins and membrane phospholipids. In contrast to this biochemical model of delayed cell death new research is now emphasizing the role of apoptosis, characterised by intact cell membranes and nuclear shrinkage as a pathway for delayed cell death^21) especially in the "penumbra" formed by the area of cerebral tissue surrounding an infarct. Apoptosis is a genetically regulated process of programmed cell death which, to use a computer software analogy, is the "default", prevented only by the presence of various growth factors. The gross biochemical derangement that accompanies acute cerebral hypoxic-ischaemia, by affecting the balance of neurotrophic factors, may remove that stay and trigger irreversible but delayed cell death. Thus an interlocking cascade of events may follow the original asphyxiating event resulting in secondary energy failure, irreversible neuronal injury and cell death many hours later. Several research groups are now intensively studying strategies for neuronal protection to interrupt this progression from the initial phase of reversible primary energy failure. Biochemical work is currently focusing on Insulin-like Growth Factor 1 (IGFi) which appears to be a potent anti-apoptotic agent (Gluckman P- personal communication). But the most promising strategy now entering clinical trials is the use of post injury moderate hypothermia which has been shown in animal models to produce a major reduction in impairment of cerebral energy metabolism^22). It is not clear by what mechanisms post injury cooling induces this neuroprotective effect. One hypothesis is that hypothermia delays commitment to apoptosis for long enough to enable endogenous protective mechanisms, including the production of neurotrophic factors, to be induced^23). Chronic antenatal asphyxia The majority of animal models used to understand the effects of hypoxic-ischaemia on the fetal brain are based on a single acute asphyxial event of limited duration. However early work with rhesus monkeys demonstrated that chronic partial in utero asphyxia may result in a different anatomical pattern of cerebral damage^24). Chronic human fetal hypoxia is well recognised and may result from: • reduced placental with compromised oxygen supply to the fetus (hypoxaemic hypoxia) • reduced fetal haemoglobin concentration (anaemic hypoxia) • reduced blood flow to the fetal tissues (ischaemic hypoxia) Studies using cordocentesis of the umbilical vein have demonstrated that some small for (SGA) fetuses have evidence of hypoxaemic hypoxia due to ^25). This is accompanied in some cases by hypoglycemia and hypoinsulinaemia^26). The resultant reduction in glycogen deposition, by reducing myocardial tolerance of acute hypoxia, may be one of the reasons for the observed excess of birth asphyxia in growth retarded newborns. Not all SGA infants result from utero-placental insufficiency. The development of Doppler ultrasound investigation of umbilical artery flow patterns has delineated a subgroup with evidence of impaired flow velocity. Absent or reversed end diastolic (ARED) flow is predictive of hypoxia and acidosis at birth^27). Observational studies of this group show a significantly increased risk of stillbirth and neonatal death^28). Pathological studies of the placentae following in which ARED flow patterns were detected have demonstrated occlusive lesions of the intraplacental vasculature and associated vascular remodeling^29). Following a multi-centre observational study cesarean section has been recommended in pregnancies complicated by ARED flow where neonatal intensive care is well developed^30). There are no large scale studies of the association with encephalopathy in term newborns since most of these pregnancies now culminate in preterm delivery. Pathological consequences of birth asphyxia The consequences of cerebral asphyxia vary considerably with the nature of the insult and most importantly with the gestation of the infant. The characteristic types of neuropathology described below in the human neonate may present separately or in combination^13). Selective neuronal Since the neurone is the cell type of the brain most vulnerable to hypoxic-ischaemic injury it is not surprising that this is the commonest variety of injury seen following birth asphyxia. It occurs in both premature and term infants. Neurones of the cerebral cortex, cerebellum, basal ganglia, brain stem and anterior horn of the spinal cord may be affected. Characteristic patterns of involvement may reflect in part the regional distribution of NMDA glutamate receptors. Parasagittal cerebral injury This is the main ischaemic lesion of the full term infant suffering . Necrosis is found of both the cortex and underlying white matter of the parasagittal cerebral convexities, usually bilaterally. Periventricular leukomalacia This is the characteristic lesion seen in sick, premature infants after several days of intensive life support including ventilation in the neonatal period. However it has also been increasingly recognised in term infants who manifest cerebral palsy having 20 experienced no perinatal adversity. The weight of evidence is shifting from a supply side explanation of the anatomical distribution (i.e. a supposed watershed between centripetal and centrifugal arterial supply) towards a demand side explanation based on the high requirements of this metabolically active area in the early third trimester^31). Although conventionally termed periventricular leukomalacia and presumed to be ischaemic in origin a recent review argues this term should be dropped in favour of the more general term white matter disease (WMD) which carries no assumptions as to aetiological mechanism^3 2\ Multifocal ischaemic necrosis Sometimes referred to as multicystic encephalomalacia this refers to multiple cavitated foci of cerebral necrosis. This reflects widespread cerebral white matter necrosis caused by generalised ischaemia. It may be seen in both premature and term asphyxiated infants. Status marmoratus of the basal ganglia and thalamus This is the least common lesion seen following asphyxia, usually in the term infant. It is characterised by neuronal loss and exuberant proliferation of myelin in the affected areas. It is now thought to be due to glutamate induced neuronal death. Cerebral oedema Finally there is the phenomenon of brain swelling observed in some asphyxiated infants. This is found in only a minority of affected infants, appears to always be accompanied by cerebral necrosis and reaches a maximum at least 24 hours after the initial insult. Therefore it is now considered to be a consequence of, and not a causal factor in, hypoxic-ischaemic brain injury^33). The timing of hvpoxic-ischaemic injury The relative weight of antenatal, intrapartum and early postnatal contributions to hypoxic-ischaemic insult is difficult to gauge. Volpe, reporting from well resourced intensive care environments in high income countries, tentatively suggests for term infants that 20% occurs exclusively in the antenatal period, 35% exclusively during labour and in 35% the timing of the insult is mixed<34). In 10% of cases affecting term infants he suggests the insult occurs postnatally . Outcomes of birth asphyxia

Stillbirth Fetal exposure to overwhelming asphyxia during delivery presents as fresh stillbirth. Wigglesworth has proposed the following scheme for the pathophysiological classification of perinatal deaths (including both stillbirths [SB] and neonatal deaths [NND])(35>. 21 • Group 1: normally formed macerated (SB) • Group 2: congenital malformations (SB or NND) • Group 3: conditions associated with immaturity (NND) • Group 4: asphyxial conditions developing in labour (fresh SB/NND) • Group 5: specific conditions other than the above In developing countries where stillbirth rates remain high the important contribution of fresh stillbirths to the overall problem of birth asphyxia needs to be taken into account. Neonatal death Severely asphyxiated term infants may be bom in poor condition and die in the early neonatal period (upto seven days after birth). The terminal event is usually ventilatory failure secondary to brain stem compromise ^36). Neonatal morbidity In addition to the direct multi-system involvement in the primary hypoxic-ischaemic process described above a wide range of clinical associations have been described. Thick meconium staining of the amniotic liquor is associated with poor condition at birth^37) and may directly contribute to fetal ischaemia by inducing vasospasm of the fetal chorionic vessel s(38). The well recognised association with meconium aspiration syndrome contributes to the neonatal morbidity seen in asphyxiated newborns @9\ There is an association between birth asphyxia and necrotising enterocolitis0*0) probably related to ischaemic injury sustained by the intestinal mucosa. Diminished splenic function^41), consumption of coagulation factors0*2) and thrombocytopenia^43) have all been reported following birth asphyxia. Clinical studies have reported an increased risk of both hypoglycaemia due to excess substrate consumption and hypothermia following hypothalamic damage in asphyxiated infants0*4). Asphyxia may also compromise immune function <45) and predispose the infant to early infection^46). It is not known whether this vulnerability to infection extends beyond the immediate neonatal period. There is also evidence that altered intestinal motility may predispose to feeding intolerance^47). Hypoglycaemia^48), hypothermia^49) , early infection^50) and feeding problems^51) are amongst the commonest problems faced by neonates in developing countries. These conditions may potentiate the contribution of hypoxic-ischaemic injury to both early and late morbidity and mortality. Disability Disability refers to a disorder of function resulting from one or more structural impairments^52). Following hypoxic-ischaemic cerebral injury there may be impairment of the motor system of the central nervous system (cerebral palsy[CP]) and the higher cortical centres (intellectual impairment)^53). There may be associated seizure disorders ^ ) and impairment of hearing^55) or vision ^56) due to cortical and/or cranial nuclear damage. For epidemiological purposes investigators generally classify children with multiple impairments dominated by motor involvement within the general category of cerebral palsy. The most commonly employed classification of cerebral palsy is based on the pattern of motor impairment^57). Birth asphyxia can result in several patterns of involvement depending on the predominant distribution of cerebral injury. In the most severely affected infants the widespread lesions result in spastic tetraplegic CP often accompanied by acquired microcephaly and cortical blindness^58). Less severely affected children may have spastic hemiplegia^59) or ^60). Dyskinetic CP may result from asphyxial damage to the extrapyramidal system, notably the basal ganglia^61) and cerebellar damage may result in ataxic CF*36). Children who develop CP following neonatal encephalopathy are more likely to have a spastic quadriparetic type of CP, to be non-walking and to have visual impairment than CP children who experienced an uneventful perinatal period^62). As Paneth has stated the impairment of asphyxiated infants with long term sequelae " is generally severe, often multiple, and virtually always involves the motor system"^53). His detailed review of the evidence suggests seizures or mental impairment in isolation do not result from birth asphyxia^63). A single long-term follow up study in Canada has shown that infants who experienced moderate neonatal encephalopathy without apparent neurological impairment had increased risk of educational difficulties at age 8 compared to an unaffected peer groups64). Importantly the largest prospective birth cohort study with primary neurological outcomes undertaken to date (the Collaborative Perinatal Project of the National Institute of Neurological and Communicative Disorders and Stroke [NCPP]) has shown that persisting neurological damage does not follow fetal distress in the absence of early neonatal encephalopathy^65). As these authors put it "the inevitable bridge between irreversible brain injury originating in the intrapartum period and eventual handicap is neonatal encephalopathy".

Clinical recognition of birth asphyxia The model of the pathophysiology of birth asphyxia I have sketched above is the result of a multitude of studies using animals, post mortem human specimens and evidence from metabolic and neuroimaging studies of the living neonate. Unfortunately we have no widely practical method for directly measuring fetal asphyxia during the birth process. As suggested above all infants undergo intermittent hypoxia during the process of labor. Therefore asphyxia cannot be categorised as present or absent, but rather it must be quantified and then ranges of normal described. These issues have been elegantly discussed by Blair in an important annotation^66). She argues that the term 'birth asphyxia' should be reserved for the conceptualisation of cellular pathogenesis and omitted from clinical practice altogether. 23 In the absence of a direct method of measurement clinical studies must rely on secondary markers of birth asphyxia. This has resulted in important differences of case definition which have muddled this field of research. This section critically reviews the historical development of clinical definitions of asphyxia concluding with an empirical definition adopted for the purposes of this study. Little In 1861 William John Little, then aged 51, collected together thirty years of experience to present a treatise on "The influence of abnormal parturition, premature birth, and asphyxia neonatorum, on the mental and physical condition of the child, especially in relation to deformities" to an early meeting of the Obstetrical Society of London^67). His own club-foot had led to a life-long interest in "deformities", the first descriptions of congenital spastic diplegia and the foundation of the concept of cerebral palsy. In the obstetric treatise he summarises the contemporary understanding of intrapartum fetal physiology, arguing "if pulmonary respiration be not immediately established, the state of suspended animation-asphyxia neonatorum-takes place". He discusses the earliest morbid anatomical studies of fresh stillbirths performed in Germany (where Little had studied) in the 1850's by Weber and Hecker. These described widespread hemorrhagic congestion and extravasation in the viscera including the brain. Little highlights in his discussion of this literature the absence of apparent mechanical injury in stillbirths which he argues were " caused by pressure on , premature separation of placenta, and uterine haemorrhage" and "in a case of descent of the umbilical cord no pelvic obstruction having existed' (Little’s emphases). Little took this evidence to expand his concept of the aetiological pathway of birth asphyxia to include interruption of the umbilical circulation prior to delivery, emphasising that mechanical injury is often not the cause of birth asphyxia. He went on to review 47 cases of cerebral palsy that presented to him in infancy and childhood. This highly selected group offer in many cases post-hoc accounts of obstetric difficulties. Little gives a good account of early post delivery neurological dysfunction in infants who survive birth asphyxia noting in the "first days after birth" following "the restoration of vegetative functions" the occurrence in some cases of "convulsions, opisthotonus, or laryngismus". This persuades him, in his own words, that "abnormal parturition, besides ending in death or recovery, not infrequently has another termination in other diseases". To use a phrase coined almost a century later he recognised a 'continuum of reproductive casual ty'(68\ Little goes on to discuss the patterns and progress of his patients' various neuro-muscular deformities. In so doing he describes a progression of clinical correlates to the morbid pathological findings "which correspond with or produce the symptoms of asphyxia, suspended animation, 24 apoplexy, torpidity, tetanic spasms, convulsions of new-born children, and the spastic rigidity, paralysis, and idiocy subsequently witnessed". This seminal paper, whilst representing a brilliant leap in knowledge at the time, created a paradigm that later came to mislead debate as to the origins of childhood disability. An assumed but unobservable aetiological process "birth asphyxia" is associated with an observable clinical condition "asphyxia neonatorum". The fact that neurologically abnormal infants sometimes do not breathe and therefore become asphyxiated compounds the confusion. Retrospective assignation of causation in a non-random sample on hearsay evidence led Little to overly emphasise perinatal factors in the causation of cerebral palsy and idiocy. Following his train of thought it became accepted faith that better obstetric care, to the extent that it prevented mechanical and fetal circulatory catastrophes, would reduce the incidence of stillbirths, asphyxia neonatorum, cerebral palsy and at least some cases of mental retardation. Little's paper was reprinted in its entirety in the first issue of the Cerebral Palsy Bulletin in 1958, an influential publication which matured into the Clinics in Developmental Medicine Series. Thus his ideas influenced the modem generation of neonatal neurologists. Apgar In 1953 Virginia Apgar described her now ubiquitous scoring system for the evaluation of the newborn infant^69). This permitted a quantitative expression of the early postnatal condition of infants. Given the continuing absence of any means of observing (never mind measuring) "birth asphyxia" it was inevitable that the Apgar score should be used as a proxy for the early condition of the infant in descriptive studies. Since an infant with "asphyxia neonatorum" would have a low Apgar score it subtly became a proxy for "birth asphyxia" itself. It continues to be misused as a marker of birth asphyxia right up to the present^70).

Clinical neonatal neurology Early attempts to link abnormal neurology in the neonatal period with later outcome suffered from a lack of inter observer agreement on the distinction between the normal and abnormal as highlighted by Donovan's 1962 study^71\ Gradually through the 1960's and 1970's standardised clinical neonatal neurological techniques were developed. This entailed virtually a reinvention of neurological examination techniques for the infant, the classical adult techniques being of limited application. For example Prechtl and Beintema's classic study of the term infant nervous system was published in 1964 (~72\ Various approaches to neonatal neurology were developed by different schools. In Europe behavioural items (i.e. what the infant did on simple observation) were to the fore. In North America formal examination procedures more closely allied to adult neurology were developed, Brazelton's detailed behavioural studies of newborn 25 infants being a notable exception^73). The subtleties of the examination of tone (the response of muscles to stretch) led especially to varied techniques for both its assessment and recording. In parallel with these developments investigators tested their developing clinical techniques on neurologically abnormal infants. In many cases these were "asphyxiated" term infants. Clinical markers of birth asphyxia The emphasis of clinical research has been to identify perinatal markers predictive of neurodevelopmental impairment so as to improve prognostication and the targeting of early therapeutic intervention. Such markers would ideally be specific to the presumed aetiology (birth asphyxia) and predictive of adverse outcome. Some investigators have studied intrapartum exposures (i.e. fetal distress or adverse obstetric events), whilst others have concentrated on early neonatal clinical, biochemical, neurophysiological or neuroimaging evidence. The majority of studies combine the two types of evidence. As we cannot measure birth asphyxia these putative markers can only be judged by their ability to predict later neurodevelopmental outcome. This has been investigated exhaustively and a brief summary of the more important studies only can be given here. Fetal distress The concept of fetal distress developed out of the classical observation of the association firstly of meconium staining of the amniotic liquor and later of abnormally fast or slow fetal heart rates with poor fetal outcome. The underlying mechanisms are poorly understood but are believed to reflect autonomic disturbances which may themselves be triggered by circulatory disturbances. Unfortunately evidence of fetal distress whether detected by intermittent fetal heart ausultation , more continuous cardiotocogram monitoring ^ or meconium staining of the amniotic fluid ^ is disappointingly non-specific. For example Nelson’s recent review of continuous cardiotocogram monitoring (75^ concludes that false positives occur in 99.8% of cases. Adverse obstetric events Using adverse obstetric events as a proxy for birth asphyxia ignores the well recognised phenomenon of acute fetal brain injury during apparently uncomplicated labor ^]1\ As Little suggested and Mann restated^78) undetected umbilical cord problems may significantly contribute to such cases. Obstetric complication markers have also been shown to have a very low sensitivity for neurodevelopmental sequelae^74). The Groningen School have tried to develop a different approach in order to explore the relationship between what have been conceived of as obstetric complications and neonatal neurological abnormalities^79^. In a long term research program Prechtl and colleagues have moved away from ideas of normality and abnormality to focus on what they term the optimality approach. This recognises a continuum of obstetric risk and generates an obstetric optimality score. This group related this to a quantitative measure 26 of neonatal neurological normality^80). The resulting association was found to be statistically significant but of little explanatory power. In terms of predicting the neurologically abnormal infant the obstetric optimality score has poor predictive power. The cut off must be set so low as to include 30% of the population if 50% of the actual neurologically abnormal infants are to be predicted (i.e. around 1000 pregnancies to detect 80 abnormal babies). Apgar score Early studies demonstrated a significant association between low Apgar scores and neonatal mortality^81). Therefore when the NCPP study was designed in the late 1950's the Apgar score was adjudged such an important variable to be measured that it was recorded by specially trained independent observers not involved in the care of the 50,000 plus infants enrolled^82). Subsequent analysis of this massive cohort study showed that the early Apgar score was a poor predictor of long term neurodevelopmental sequelae^83). However a very significant association was found between extended Apgar scores (15 and 20 minutes after birth) and long term neurodevelopmental impairment. An Apgar at 20 minutes of three or less carried a 59% mortality risk and a 57% CP risk in survivors. It was clear that it was not the condition of the infant in the first five minutes after birth that was informative but rather the condition of the infant in the succeeding minutes and hours. Early seizures In the 1980's there was much interest in the use of early seizures (occurring in the first 48 hours of life) as an indicator of birth asphyxia. Bergman writing in 1983 reviewed the early literature performing a meta-analysis of 1667 cases from special care hospital series and showed early seizures were associated with death or disability in 50% of cases^84). However this group's review of the aetiology of 131 cases of early seizures in their own neonatal intensive care unit highlights its poor specificity (60%) for birth asphyxia with a relatively high proportion of cases (40%) resulting from other causes. Nevertheless the principal investigators of the influential Dublin Collaborative Study of Fetal Monitoring in 1985 concluded that early seizures were the best available marker of asphyxia^85). Thus the neonatal asphyxia data from the Dublin study of cardiotocographic monitoring was reported in the form of a case-control study of neonatal seizures^86). In another influential study co-ordinated by the Oxford Perinatal Epidemiology Unit in 1988 Minchom and colleagues described a large series of infants affected by neonatal seizures in a geographically defined population^87). This group acknowledged the problems inherent in the clinical recognition of neonatal seizures. Their suspicions were confirmed by a Hammersmith study utilising continuous EEG monitoring which demonstrated that only 21% of electrical seizure activity was clinically recognised even by trained staff in this specialist unit with a high staff to patient ratio(88). In the context of under resourced developing country hospitals seizures 27 are even less likely to be recognised and this limits their usefulness as an indicator of birth asphyxia. An important re-analysis of the largest population based data base available (the NCPP) subsequently showed that isolated seizures were associated with death in 10% of cases and later handicap in only 1.5% of cases. However when interictal neurological signs were also present these predictive figures rose to 30% and 10% respectively^89^. It was apparent that the best clinical predictor of neurodevelopmental sequelae in infants presumed to have suffered cerebral injury due to perinatal asphyxia was likely to be some way of describing the early neurological condition of the infant which included seizures in its definition. Hvpoxic ischaemic encephalopathy (HIE) Clinical Neurological Studies Table 1 lists 20 studies performed since 1969 investigating birth asphyxia in which early neurological signs have been either entry criteria or the subject of the study. The key study was the first description of neonatal encephalopathy by Samat and Samat in 1976 but other influential studies preceded this description and others have subsequently modified it. Brown (1974) 2 years before Samat in Canada described neonatal encephalopathy Brown and colleagues used what they termed "behavioural symptoms" as enrolment criteria in their influential British study*90). They examined 760 neonates thought to have suffered birth asphyxia. They defined cases of birth asphyxia as those manifesting any of the following 'mixed bag' of presumed markers: • "antenatal"- maternal shock, antepartum haemorrhage, maternal , • "perinatal "-severe placental insufficiency, cord tightly around neck, cord prolapse, meconium stained liquor, cardiotocogram abnormalities • "postnatal"-Apgar <3 at 1 minute and <5 at 5 minutes, need for early ventilation, early serum acid base value<7.2, severe post-partum respiratory distress. They showed that only 11% displayed abnormal neurological behaviour in the neonatal period of whom 66% either died or developed neurodevelopmental disability. None of the neurologically normal neonates in this study subsequently appeared on the district handicap register. The "abnormal neurological behaviours" were non-specific abnormalities that drew nursing attention to infants (feeding difficulty requiring tube feeding, apnoeic/cyanotic attacks, apathy, convulsions, hypothermia, cerebral cry and persistent vomiting). Babies with a history of birth asphyxia who exhibited any of these symptoms were enrolled and underwent further detailed daily neurological examination. The investigators considered that "since abnormal muscle tone was so often present and prominent" in their series of 94 babies that they divided asphyxiated babies into four groups according to their predominant tone pattern. They distinguished between , hypotonia progressing to extensor hypertonus, extensor hypertonus and finally normal flexor tone. The latter clearly had the best outcome - only 15 babies had normal tone and all of these subsequently had normal neurodevelopmental follow-up apart from four with "minimal cerebral dysfunction".

Table X. Review of major studies of birth asphyxia in which early neurological signs have been entry criteria or the subject of the study.

Stuty Investigator (ref) Year Setting Study n case/ Purpose No Design control 1 Amiel-Tison^91) 1969 Paris, Descriptive 41/- to describe the prognosis France. case study of asphyxiated infants 2 Thom(92) 1969 Copenhagen Prospective 291/- to relate cerebral Sweden. follow up study symptoms to prognosis 3 Brown(90) 1974 Edinburgh, Prospective 94/- see text UK. follow up study 4 Samat(44) 1976 St Louis, Descriptive 21/- see text Canada. case study 5 Ziegle/93) 1976 Lausanne, Prospective 90/- to relate neurological Switzerland. follow up study signs to prognosis 6 Scott*-36) 1976 London, Descriptive 48/- to establish outcome of UK. case study, very severe asphyxia 7 de Souza^94) 1978 Manchester, Prospective 53/53 to assess outcome of UK follow up study abnormal infants 8 Fitzhardinge^9^ 1981 Toronto, Prospective 65/- to assess prognostic value Canada follow up study of CT scanning 9 Finer(96) 1981-7 Edmonton, Prospective a) 95/- to assess perinatal factors Canada. follow up study b)226/- to assess outcome of HIE 10 Lipp-Zwahlen^97) 1985 Lausanne, Prospective 25/- to predict outcome. Switzerland. follow up study 11 iW 98> 1985 Kingston, Descriptive 42/- to describe prevalence and Canada case study risk factors of NE 12 Levene^99) 1985-6 Leicester, Prospective 126/- to describe prevalence and UK. follow up study outcome of HIE 13 Upper*-100) 1986 New York, Prospective 45/- to derive prognostic USA. follow up study scoring system 14 Arche/101) 1986 Leicester, Prospective 49/49 to assess Doppler as early UK. follow up study predictor of outcome 15 Fernandez^102) 1987 Madrid, Prospective 33/20 to predict outcome Spain. follow up study 16 A1 Alfy(103) 1990 Kuwait, Case-control 43/43 to describe incidence and Kuwait. study risk factors 17 Shankarai/9) 1991 Detroit, Prospective 28/- to assess the extent of USA. follow up study multi system involvement 18 Airede^104) 1991 Maiduguri, Descriptive 166/- prevalence study study 19 Gray(105) 1993 Brisbane, Prospective 26/- to predict outcome. Australia. follow up study 20 Thompson^106) 1997 Cape Town, Prospective 45 to derive prognostic S. Africa. follow up study scoring system 29 Samat (1976) In 1976 Samat and Samat published a combined clinical and EEG study of 21 term infants who displayed evidence of fetal distress including fetal bradycardia, tachycardia, meconium stained amniotic fluid or an Apgar <5 (at either one or five minutes)^44). They described a syndrome of neurological and electroencephalogram (EEG) features that they labeled 'neonatal encephalopathy following fetal distress'. As described in their original study the syndrome was divided into three stages, with severely affected infants typically progressing from grade 1 to grade 3. The clinical characterization of the syndrome included abnormalities of level of consciousness, posture, tone, segmental myoclonus, stretch reflexes, Moro, tonic neck and oculovestibular reflexes, sucking behaviour, autonomic function (pupil size, heart rate, secretions, and gastrointestinal motility) and seizures. In their series all "grade 1" infants progressed to more severe grades. In other words grade 1 as they elaborated it was a transitory phase in the evolution of more severe encephalopathy rather than a mild and self-limiting grade of encephalopathy as used by later authors. Their interest in autonomic features was at least in part founded on their model of a progression of autonomic dysfunction with initial sympathetic over activity, followed in grade 2 by parasympathetic over activity, culminating in their stage 3 with generalised neuronal depression typified by depression of the entire autonomic nervous system. Later modifications to Samat staging De Souza in Manchester in 1978 applied Prechtl and Bentema's technique to a highly selected group of 53 "severely neurologically abnormal" infants with a history of fetal distress (94\ This group followed Brown in describing four categories of neonatal neurological status but added important observations on feeding, convulsions, apnoea and condition of the fontanelle. Lipp-Zwahlen in Switzerland in 1985 "followed" the Samat classification but with some modification they characterise grade 1 as "short lethargy after birth followed by hyper alertness, normal or weak primitive reflexes (suck, Moro, grasp), mildly abnormal muscular tone and no seizures" <97\ If compared to the Samat criteria this represents substantial modification. Likewise they subtly alter Samat grade 2 criteria of 'mild hypotonia' to substitute 'moderate tonus alteration'. It is apparent that as study of neonatal encephalopathy developed investigators came to realise that there was more variation in the constellations of neurological clinical signs than the Samat classification had acknowledged. The Levene (1985) modification Levene in Leicester in the same year implicitly recognised these difficulties by drawing up a grading system of his own . This was based on an influential review article by Fenichel (107) in which he built on Samat's work describing what he terms hypoxic- ischaemic encephalopathy (HIE). Levene's approach can be viewed as a pragmatic 30 clinical grading system with more of an emphasis on behavioural features. Thus, returning to Brown's earlier work, feeding is included and following de Sousa^94^ respiration is explicitly included. He also admits that not all features of each grade were necessary for diagnosis but gives no operational guidance in terms of "minimally required" features. The Leicester group focused on conscious level, tone, seizures, feeding, respiration and the time period to recovery to characterise the syndrome in three grades of severity. The mildest grade one by their definition showed complete resolution within 72 hours, the authors stating that "any infant who showed persistent neurological abnormalities were not included in grade 1". This introduces some methodological difficulty when they go on to relate the grading system to outcome. It is somewhat of a tautology to show that grade one, thus defined, carries an excellent prognosis. Leviton^108^ has argued that Levene's inclusion of a temporal dimension in the definition of grade one is inherently contradictory. Most researchers, including Levene, grade HIE according to the worst clinical grade observed during the entire duration of the clinical disturbance. Therefore if time is included in the definition of particular grades the scheme becomes inherently contradictory. Using this grading scheme Levene demonstrated that the grade of HIE manifest in the neonatal period was a more specific predictor of subsequent neurodevelopmental impairment than the Apgar score(109). This influential study ensured that HIE replaced the Apgar score as the most widely used marker for perinatal asphyxia in industrialised country studies. Other approaches Lipper and colleagues in New York (1986) departed from the qualitative approach previously employed to use a quantitative "post-asphyxial score" assigned during the first 24 hours of life <10°). This assigned scores in broad categories of neurological dysfunction as shown in Table 2.

Table 2. The New York post-asphyxial score.

Neurological category No of variables Total possible score

Conscious level 2 6 Cranial nerves 3 6 Tone 6 19 Sensory response 1 1 Reflexes 3 7 Intracranial pressure 1 1 Seizures 1 1 Totals 17 41 31 Unfortunately the development of this scoring system shows no attempt to assess discriminatory power of the various criteria included and thus the relative weighting of the categories appears arbitrary. Another scoring system reported in the literature is based on over seventeen neurobehavioural items^110) and as a result is rather unwieldy. Neither system has achieved widespread usage. More recently a Cape Town group have described a locally developed encephalopathy scoring system which assigns scores between 0 and 3 for each of 9 neurological features (conscious level, sucking behaviour, tone, posture, Moro and grasp responses, fits, respiratory pattern and fullness of the fontanelle)(106). This scheme has the merits that it broadly follows Fenichel's HIE grading system both in terms of the items selected and the criteria for grading each of those items. Total score is then grouped into three categories (0-10 mild, 11-14 moderate, 15-22 severe). The validity of this grading system was measured against Fenichel’s HIE grading system as a gold standard and disagreed in only 3 of 45 infants assessed. The inter-rater reliability was high with a kappa value for inter-tester reliability of 0.87 (on 10 subjects performed by 2 observers). The positive predictive power (PPP) of the scoring system for neurodevelopmental impairment at one year of age was assessed for a cohort of 45 encephalopathic infants. The authors here introduce the additional variable of the clinical condition of the infant on day seven. They demonstrate that when the maximum score exceeds 15 and the infant remains neurologically abnormal on day 7 then the PPP is high (92%) with a 100% negative predictive power (NPP) in their cohort. The authors argue that their scheme is quick, requires no special training and permits a non specialist to prognosticate with a high degree of confidence. Although detailed scoring of seizure frequency is open to criticism on practical grounds (score one if less than 3 fits per day, score two if more than 2 fits per day), it warrants further evaluation by other developing country groups. Unfortunately for the epidemiological investigator it is not useful for comparative purposes unless it achieves widespread acceptance. Low like Levene, adapts Fenichel's modification of the Samat criteria reducing the observed variables to four These are conscious level, tone, seizures and respiration. This differs from Levene mainly in assigning occasional seizures to the moderate NE category whilst reserving the severe category for multiple seizures. A1 Alfy and colleagues in Kuwait explicitly adopt "minimal criteria" for the diagnosis of what they describe as an "asphyxial syndrome" (103>. These are alterations in the infant's state of consciousness and abnormalities of muscle tone. They go on to describe a further modification of Fenichel's grading system, similar to that of Low, with the addition of the variable of feeding. Of the other studies referred to in Table 1 the investigators either do not employ a grading system ( studies 1,2,5, 6 ,8 ), or follow Samat (9,15,17), Fenichel's modification (19) or Levene's modification of the latter (14,18). 32 In the literature there are several review articles (as opposed to the actual studies listed above) of direct relevance. Amiel-Tison writing in 1986 ^ 1 ^ notes that inter study comparisons are difficult because the entry criteria vary with each investigator using neurological signs and symptoms of varying nature and severity. She advocates the conventional three stage model with the nuance of further binary subdivision of grades 2 and 3 according to the presence of seizures and brain stem signs respectively. Some authorities, including both the American Association of and College of and Gynaecology, advocate the inclusion of evidence of multisystem involvement in the diagnostic criteria for birth asphyxia ^112\ In a clinical series of 13 relatively severe cases of HIE three had no evidence of other organ involvement and in another 5 the only non-cerebral abnormality was a subtle renal disturbance (with transient oliguria and raised B2 microglobinuria) (5\ More recently a study of six cases of cerebral palsy satisfying what the authors term stringent criteria for intrapartum asphyxia found no evidence of multisystem involvement in two cases^113\ It would therefore seem likely to seriously reduce sensitivity for the key outcome of neurodevelopmental impairment if multi-system involvement was to be included in the case definition. HIE reviewed The vast majority of these studies included some marker of intrapartum compromise in the case definition of HIE, generally either fetal distress or a low early Apgar score or both. But in the absence of detailed obstetric monitoring data many clinicians continue to diagnose HIE in term infants. It is argued that HIE has a characteristic time course compared to other causes of encephalopathy. As Nelson has stated "the characteristic evolution of signs in HIE enables the diagnosis to be fairly safe retrospectively"^114^. This view has received experimental support from studies using magnetic resonance spectroscopy to describe patterns of derangement of cerebral energy metabolism at a subcellular level in human infants with clinical evidence of HIE I115). The temporal patterns of biochemical derangement reflect the clinical progress of the encephalopathy ( 116) Certain conclusions can be drawn from this review of the main clinical studies describing hypoxic-ischaemic encephalopath. Firstly, Samat and Samat’s original description was based on a small selective sample. Secondly, HIE as actually used in research studies is a surprisingly heterogeneous set of clinical symptoms and signs with significant variation in grading criteria G 17>. Thirdly, that the research definitions tend to broadly fall into three camps. The more formalist approach of Samat is favoured by many North American investigators. The more observational approach of Levene and Low has been used more widely (in geographical terms) for studies in the UK, Nigeria, Kuwait and Australia. The latter approach lends itself better to retrospective studies involving the post hoc review of case notes, such as the studies of Levene and Airede. 33 However it has also been successfully adapted to a prospective study design in which infants were graded as they were examined^103\ The third approach using scoring systems has not achieved widespread usage. Neonatal encephalopathy Leviton and Nelson have made a detailed critique of the definitions and classifications of neonatal encephalopathy^108). Some of their points have been made above. They conclude "the ideal grading system continues to be an unattained goal". These authors go on to criticise the whole concept of HIE. They argue that the inclusion of some marker of fetal distress in the case definition of HIE prejudges the aetiology of HIE since it assumes that all cases show evidence of intrapartum stress ^114\ They advocate the adoption of a wider definition of neurobehavioural abnormality in the newborn infant - neonatal encephalopathy (NE). NE is described in their review as a disturbance of neurological function in the earliest days of life in the term infant manifested by difficulty initiating and maintaining respiration, depression of tone and reflexes, subnormal level of consciousness and often by seizures. Clinical studies detailed above add that tone may be increased as well as decreased^97), that feeding difficulties are usually manifest and that brain stem signs (particularly the absence of comeal and gag reflexes) may be seen in severe encephalopathy ^11 It is axiomatic that not all cases of NE are the result of birth asphyxia and causes of neonatal encephalopathy recognised to date include: • hypoxic ischaemia • cerebral trauma • infection • inherited metabolic disorders • congenital neuromuscular disease • idiopathic cerebral infarction • structural brain malformations • congenital dysmorphic syndromes Clearly any study of birth asphyxia must seek to exclude these associated conditions if it is to avoid the erroneous ascription of encephalopathy to hypoxic-ischaemia. Even classical studies of HIE in centres of excellence such as Levene's have initially included cases that later proved to have another aetiology ( one case initially classified as grade 1 HIE which subsequently was diagnosed as congenital myopathy). A task force set up by the World Federation of Neurology Group for the Prevention of Cerebral Palsy and Related Disorders met in 1992 to agree the appropriate use of the term 'birth asphyxia'^118^.They recommended that "the terms perinatal or birth asphyxia, and hypoxic-ischaemic or post-asphyxial encephalopathy should not be used until or unless some evidence specific to an asphyxial origin for neurological illness in the neonate is available". They proposed "that the term 'neonatal encephalopathy of 34 early onset' be used instead, and that clinical information be recorded as to possible antecedents of the encephalopathy". Few studies have been published taking this approach and no consensus as to case definition has yet been reached^1191. The influential Western Australian Perth research groups 12°) have adopted a broad definition of NE including isolated seizures and any two of the following lasting for longer than 24 hours • abnormal consciousness • difficulty maintaining respiration (of presumed central origin) • difficulty feeding (of presumed central origin) • abnormal tone and reflexes A South West Thames group have adopted a similar definition but are yet more inclusive by including cases manifesting only one of the above e.g. tone abnormalities^119^. These studies are intentionally broad based exploratory analyses of factors associated with neonatal encephalopathy with the intention of generating causal hypotheses for later independent testing. They explicitly recognize the heterogeneity of the condition under study. They are not primarily designed to compare prevalence or outcome with previous studies of HIE. Birth asphyxia in preterm infants The syndrome of newborn encephalopathy describes early neurological abnormality in the term and post-term infant The immaturity of the central nervous system of the premature infant makes the clinical neurological definition of encephalopathy more problematic for this group. The lower tone and level of arousal found in healthy premature infants makes it difficult to apply the clinical criteria for encephalopathy described in the term infant to this group. There does not appear to be a striking acute clinical neurological syndrome associated with intrapartum cerebral injury in this group. In the absence of operational definitions of HIE in premature infants this group will remain outside the scope of syndromically defined birth asphyxia studies for the present. The neuropathology seen in premature infants (typically periventricular leukomalacia or white matter disease as it is increasingly termed) develops over the first days and weeks of life. There is a different anatomical distribution of involvement compared with the typical parasagittal pattern in term infants^121). There is some evidence that much of this pathology results from postnatal hypoxic-ischaemia134). The Wigglesworth pathophysiological classification of perinatal death likewise treats deaths associated with prematurity as separate from those in term infants resulting from birth asphyxia. In high income countries preterm infants make an increasingly important contribution to overall cerebral palsy rates^122). In the absence of neonatal intensive care they remain a much less important contributor to overall cerebral palsy rates in low income countries^123). 35 Birth asphyxia and CP The descriptive term cerebral palsy (CP) has proved clinically useful to define a population of disabled children with related problems due to non progressive central nervous system pathology predominantly affecting the motor system. However the term represents a heterogeneous group of disorders with little in common other than the phenotypic pattern of clinical expression. What proportion of cerebral palsy is caused by birth asphyxia? A review has explored the problematic causal chain involving birth asphyxia, NE and CF*124\ Possible causal pathways are graphically shown in Figure 1 adapted from this article. Even in cases sustaining significant intrapartum insult it has been hypothesised that CP may result only in infants who are vulnerable in some way (125\ In this model multiple 'hits', which individually may not result in long term sequelae, potentiate risk and in combination result in brain injury.

Figure 1. Possible causal pathways relating prenatal risk factors, intrapartum birth asphyxia, neonatal encephalopathy and cerebral palsy.

Prenatal factors Birth Neonatal Cerebral (early and late) Asphyxia Encephalopathy ------Palsy

Birth fe- Neonatal w Cerebral Asphyxia Encephalopathy Palsy

Prenatal factors Cerebral w Neonatal ...... P* (early and late) Encephalopathy Palsy

Prenatal factors (early and late)

Prenatal factors ------)». Birth Neonatal ------to. Normal (early and late) Asphyxia ...... 1^ Encephalopathy

Prenatal factors ib Birth |K- (early and late) Asphyxia Normal

Prenatal factors (early and late) w Normal

Epidemiological studies of CP generally exclude postnatally acquired brain pathology when examining the contribution made by birth asphyxia to CP. As more genetically determined disorders culminating in CP are recognised there is continuing debate over the inclusion of some syndromes, whether a molecular level of understanding has been achieved or nori 126,127\ These methodological issues notwithstanding studies in high income countries have consistently shown that birth asphyxia is the main cause of CP in only 10-20% of cases (63,113, l2S\ When considering a child with neuro- 36 developmental impairment it was suggested ten years ago that if the impairment is to be ascribed to birth asphyxia then the following four criteria should be met^76); • evidence of adverse intrapartum events • evidence of neonatal encephalopathy • disability of a nature consistent with presumed hypoxic ischaemic pathology • adequate neurological assessment to exclude other conditions The epidemiology of cerebral palsv Before considering risk factors for birth asphyxia it is relevant to consider the accumulated knowledge of the epidemiology of cerebral palsy which has been extensively investigated in high income countries. The earliest longitudinal data from an important Swedish disability register shows a modest decrease in overall CP rate from 2.2 per 1000 live births in 1956 to 1.5 per 1000 live births in 1970 followed by an apparent increase to 2.2 per 1000 live births in the 1970-1980 birth cohort (l29\ Although for the Mersey (UK) Regional registry this trend was not significant (13°) it is replicated for the period 1975-1985 in centres as widely spaced as Perth, Western Australia^124) and Tottori, Japan^131-. Most authorities believe this trend reflects a real increase in the prevalence of spastic diplegia secondary to the improved survival of very low birthweight infants with a high risk of CP (relative to term infants) following the introduction of neonatal intensive care in this period^ l32\ There is now evidence in the latest longitudinal data from Perth of a reversal of this trend in CP prevalence amongst children bom in the 1990's (F. Stanley-personal communication). But comparing all the available data from industrialised country disability registries the striking finding of the longitudinal record is that there has been remarkably little change since the 1950's in the prevalence of CP amongst term bom children which remains approximately 1 per 1000 live births <63) . This stands in stark contrast to the perinatal which for example in Sweden fell from over 30 per 1000 total births to below 10 per 1000 total births over the same period (133>. There are no equivalent estimates either from the historical past in Europe or from low income countries in the present day of cerebral palsy prevalence when perinatal mortality rates exceed 40 per 100 0 total births. Prematurity, pre-eclampsia and cerebral palsy Apart from birth asphyxia the other well recognised risk factor for CP is low birthweight. As has repeatedly been shown risk of CP rises sharply with decreasing maturity^134). However there remains an important birthweight effect after gestation has been controlled for <135). This has encouraged investigators to focus on antenatal risk factors which may mediate this association. The important factor of chronic placental insufficiency has been briefly discussed above. An important because common subgroup of this population have evidence of pre-eclampsia (PE) and a more 37 widespread vasculopathy. A current model of this incompletely understood phenomenon emphasises incomplete placentation, possibly as a result of immunologically based partial rejection of fetal material by the mother, resulting in a placental vasculature which remains responsive to vaso-active substances, impaired placental blood flow, placental ischaemia and the release of vasculopathic substances*136^. Studies of term bom infants have consistently shown a positive association between CP and PE; for instance an Oxford case control study of CP demonstrates an unadjusted odds ratio of 2.0 (95% Cl 1.2 -3.4) for all PE rising to 3.7 (95% Cl 1.4-9.9) for severe PE*137\ But when this association has been tested in premature infants the evidence for an independent effect over and above the propensity for pregnancy complicated by PE to result in premature delivery is less convincing. The Oxford group, for instance, have demonstrated that PE is protective against CP in very preterm singletons with an odds ratio of 0.4 (95% Cl 0.2-0.9). But if all CP children are considered there appeared to be no protective effect conferred by PE in Western Australia(138). It has been postulated that the stress of PE affecting preterm births may accelerate fetal maturational processes and thereby protect against the independent effect of prematurity 039) situation is further complicated by the observation of the California CP Project that magnesium sulphate, used for the treatment of pre-eclampsic toxaemia and to arrest premature labour, appeared to protect against cerebral palsy in very low birthweight infants ( odds ratio 0.08,95% Cl 0.02-0.67) O40). Other groups have failed to replicate these findings 041). the Perth group have pointed out since magnesium is only used in a group whose clinical characteristics and risk profile for CP differs from that of the no magnesium group it is impossible to demonstrate an independent effect of magnesium without performing a randomised control trial.

Infection, inflammation, cytokines and cerebral palsy Perinatal infection is a well recognised cause of cerebral palsy I124). Even in the absence of overt case-control studies of cerebral palsy have persistently demonstrated an association with prolonged rupture of membranes, and maternal pyrexia*74, l42\ More recently there has been growing interest in the effect of cytokine release during pregnancy*1431. Cytokines have been cited as a possible mechanism linking premature labor and cerebral white matter disease in premature infants. The theory supposes that cytokines in general (and Interleukin -6 and Tumour Necrosis Factor in particular) are released by the uteroplacental unit, circulate in the fetus and cross the blood-brain barrier where they are direcdy cytotoxic to the developing brain*1441. A recent case control study has demonstrated significantly higher levels of many members of the cytokine family in CP cases than normal controls in dried blood collected in the early neonatal period*1451. The short half life and biochemical instability of cytokines means that there remains considerable methodological problems in this field. Acute phase proteins, which are synthesised partially in response to cytokine signals, have longer half lives and are more stable and are therefore easier to assay. However both cytokines and acute phase proteins are influenced by a wide range of factors including infection, autoimmmune status, trauma, and ischaemia. Elevated levels are therefore not specific to infection and may simply mark a final common pathway linking a diverse range of mechanisms with neurodevelopmental sequelae. A growing body of animal evidence suggests they are perhaps best considered as yet another component of the complex cascade of events initiated by a range of mechanisms which may progress into irreversible neuronal injury*1461. Nevertheless there is growing interest in this line of research and its potential for linking observed epidemiological associations (e.g. maternal pyrexia) and observed brain pathology (e.g. white matter disease)*1471.

Neuroimaging and cerebral palsy Some progress has been made in elucidating the heterogeneity of cerebral pathology underlying cerebral palsy in recent years. The key factor here is improving access to neuroimaging with first computerised tomography (CT) and more recently magnetic resonance (MR) imaging. A single study describing MR findings in 56 children from a large German cohort of 200 children with bilateral spastic cerebral palsy at a wide range of ages (2-8) found periventricular cystic leukomalacia (PVL) to be the commonest pathology occurring in 37 ( 6 6 %) of this group*1211. Of the seven (12.5% of the cohort) term infants with evidence of perinatal hypoxic-ischaemia in this group only one had evidence of PVL alone. In contrast in the important subgroup of 19 term bom children with no evidence of intrapartum compromise 10 had evidence of PVL and three of neuronal migration disorder. Embryologically PVL seems to be a lesion seen after insult in the early third trimester in contrast with neuronal migration disorders which are related to insult in the late first and early second trimester. A Finnish MR study of 42 school aged preterm children (mean 31 weeks gestation) found that PVL was common (32%) and importantly found no instances of PVL in 42 matched term bom school aged control children*1481. In contrast a Japanese MR study on a non population based sample of 70 CP children found neuronal migration defects to be the single commonest radiological abnormality affecting 29%*1491. A recent report of MR in pregnancy has documented apparently acute white matter changes affecting a fetus at 29 weeks gestation associated with placental insufficiency and in utero death at 30 weeks*1501. Cerebro-vascular events and cerebral oalsv Vascular insults resulting in cerebral infarcts of the middle cerebral artery territory which are manifest at birth have been recognised for some time in a subgroup of infants who later manifest hemiplegic cerebral palsy*1511. Intriguing new data describing coagulation indicators in the dried blood of neonates who later manifested CP has 39 described an association with factor V Leiden mutation as well as protein C, an indicator of coagulation activation <145). Factor V Leiden mutation has been linked with familial thrombotic disease and this finding will trigger further investigation in this subgroup.

ThymidlUQgliQia ,an,d,.gerg.kal^.algy. There is renewed interest in the relationship between thyroid status and the developing fetal brain. Cretinism was clinically described in the European Alps as early as 1800. Neurological cretinism was later recognised to have spastic diplegic features in common with other types of cerebral palsy. Both neurological and myxoedematous cretinism are the result of maternal and fetal hypothyroxinaemia and are associated with iodine deficiency. It has been demonstrated in New Guinea that intramuscular iodised oil delivered to women of reproductive age prior to conception prevents endemic cretinism^152). More recently it has been suggested that the classic syndromes differ only in that the myxoedematous form suffer from continuing postnatal iodine deficiency^153). In the fetus both total and free thyroxine (T 4 ) and total and free triiodothyronine (T 3 ) increase significantly with increasing maturity^154). Although fetal thyroid development is partially independently of maternal thyroid status, limited transfers of thyroid hormone from mother to fetus may have protective effects if the fetus is hypothyroid^155). Currently physiologists are actively investigating the role of thyroid hormone as an epigenic signal agent intimately involved in the differentiation and organisation of the normal neuronal network of the fetal and infant brain. A recent review summarises its role "as to ensure the timed coordination of different developmental events through specific effects on the rate of cell differentiation and gene expression"^156). The NCPP study suggested a possible association between a history of exogenous maternal thyroxine supplementation and cerebral palsy^157). A case control study conducted as part of the Central New Jersey Neonatal Brain Haemorrhage project has demonstrated that severe hypothyroxinaemia in preterm infants is associated with subsequent disabling cerebral palsy (158) with an adjusted odds ratio of 4.4 (95%CI 1.0-18.6) on multivariate analysis. An association has also been detected in term infants between neonatal hypothyroxinaemia and the risk of cerebral palsy in a case control study in North America ^159). and cerebral palsy Although crude prevalence of cerebral palsy is higher in twins and triplets than in singletons, birthweight specific rates among low birthweight groups are the same. Only amongst those with a birthweight of 2500 g or more is cerebral palsy more prevalent 40 amongst twins (4.1 per 1000) than singletons (1.4 per 1000) in the Mersey Region registry^160). This increased risk has been reported elsewhere^161). An increased risk of CP in twin pregnancy complicated by the in utero death of a co­ twin has been shown in several studies ^160,161 * l62\ Whether this is due to differential outcome after a single insult affecting both twins or to a more direct causal link between the death of a co-twin and brain development in the surviving twin has not been determined. Risk factors associated with birth asphyxia This discussion has highlighted the heterogeneity of conditions amalgamated within the concept of cerebral palsy. What are the risk factors for birth asphyxia as it has been historically described? Table 3 includes all those exposures shown to be significantly associated on univariate analysis (at conventional statistical tolerance p<0.05) with markers of birth asphyxia in live term births in any one of four large studies conducted in a range of settings over a twenty year period prior to the design of the Kathmandu study. Case ascertainment was based on low Apgar scores^163, l64\ neonatal seizures^87) or neonatal encephalopathy^ 165\

Table 3. Risk factors shown to be significantly associated on univariate analysis (p<0.05) with birth asphyxia (variously defined) (see text for references).

Preconceptual Antepartum Intrapartum thyroxine exposure multiple birth induced delivery insulin dependent diabetes fetal growth retardation augmentation nulliparity pre-eclampsia breech/malpresentation previous perinatal death threatened prolonged 2 nd stage cord prolapse placental abruption operative delivery antepartum haemorrhage instrumental delivery polyhydramnios maternal pyrexia prolonged rupture of membranes

A recent study in a teaching hospital setting in Delhi defined birth asphyxia according to intrapartum and resuscitation criteria^ 166\ These criteria defined a high prevalence of asphyxiated infants, 8 6 of 2371 births (36 per 1000 total births), of whom 35 (41%) were stillborn. Multivariate analysis showed a prolonged second stage of labour (adjusted odds ratio [adj.OR] 9.4), vaginal breech delivery (adj.OR 4.6), elective 41 caesarian delivery (adj.OR 4.6), pregnancy induced hypertension (adj.OR 2.7) and fetal growth retardation (adj.OR 2.4) to be significant associated risk factors. Perth case control study of neonatal encephalopathy Most recently an important case control study from Perth, Western Australia of antepartum^120) and intrapartum^167) risk factors for neonatal encephalopathy has been published. This builds on an earlier pilot study <165) included in Table 1 and reports a multivariate comparative analysis of 164 term encephalopathic infants and 400 randomly selected controls. The case definition has been discussed above. Exclusions were preterm infants, infants with neural tube defects and Downs syndrome infants. Cases with were not excluded. 37 of the case infants (23%) had birth defects. Although the wide case definition adopted by this group leads to some of the same issues of heterogeneity that apply to the epidemiological study of cerebral palsy it remains the most important contribution in this field of research to date and is summarised below. Sociodemo graphic factors The Perth data shows increasing risk with increasing maternal age (adjusted odds ratio [adj.OR] 6 comparing over 35 year olds with under twenty year olds) and increasing risk with decreasing socioeceonomic status. Although there is a trend towards greater risk with nulliparity this fails to reach significance. Preconceptual medical conditions A family history of epilepsy or other neurological conditions is positively associated (adj.OR 2.7) whilst infertility treatment is strongly associated with encephalopathy in the Perth data (adj.OR 4.4). in pregnancy Severe pre-eclampsia (adj.OR 6.3), antepartum (adj.OR 3.6) and a documented medical attendance for viral illness (adj.OR 3.0) were all significant independent risk factors for encephalopathy. Fetal growth and gestational age Severe intrauterine growth restriction was the strongest association found in this study (adj.OR 38). The relationship with gestational age was J shaped with minimal risk at 39 weeks rising with increasing immaturity (adj.OR 2.3 at 37 weeks) and with postmaturity (adj.OR 13 at 42 weeks). Maternal anaemia Amongst those working in low income settings the suspicion remains that maternal anaemia may be an important risk factor for a subgroup of women whose infants suffer birth asphyxia (Lindblad B-personal communication). Given the role of maternal haemoglobin for oxygen carriage to the feto-placental unit this is certainly biologically plausible. A recent systematic review of the effect of the routine administration of iron 42 during pregnancy failed to detect any benefit for either maternal or infant outcome *168>. But, given the relative paucity of developing country data, it remains possible that for severely anaemic mothers there is additional risk of pregnancy culminating in birth asphyxia. The Perth study has no data on maternal anaemia which in that setting is closely monitored and corrected making their population relatively uninformative for this exposure variable. Thyroid status Maternal thyroid disease is strongly associated with neonatal encephalopathy in the small group affected in the Perth study (adjusted OR 9.7, n=7). Combined with the observed association of neonatal with CP discussed above this raises the intriguing possibility that the thyroid status of the developing fetus may contribute to the risk of perinatal brain injury. If this were to be the case it may be a factor of some importance in Nepal, an area of widespread iodine deficiency* 169\ However a recent report demonstrated that both thyroid stimulating hormone and thyroxine levels fall in response to birth asphyxia suggests that the observed association between neonatal hypothyroidism and later neurodevelopmental impairment may be accounted for by the confounding effect of acute brain injury* 170\

Magnesium status Glutamate is thought to play a pivotal role in the cytotoxic cascade culminating in neuronal loss in acute perinatal brain injury. Since magnesium is recognised to block glutamate receptors this provides biological plausibility to the apparent protective effect of magnesium sulphate for CP in preterm infants discussed above. It has been investigated as a rescue therapy for HIE, with disappointing results due to toxicity*19^. A case control study of mothers and encephalopathic infants in Cape Town has demonstrated lower red cell magnesium levels in both mothers and affected term infants compared with controls *171). Lower magnesium levels have also been described in the umbilical cord blood of moderate/severe encephalopathic term infants compared to normal controls in Estonia*172^. Maternal magnesium levels are related to diet and therefore offer a potential intervention point were they to be shown to be an important risk factor for perinatal brain injury. Infection and cvtokines Perinatal infection is a well recognised cause of neonatal encephalopathy*98^. Placental studies have shown that chorioamnionitis is more likely to be found on pathological examination in mothers experiencing unexpected adverse pregnancy outcome*38^. Maternal pyrexia was importantly associated with encephalopathy in the Perth data (adjusted OR 3.8) but the association with prolonged rupture of membranes failed to reach statistical significance. There has been no case control study reporting cytokine levels in encephalopathic infants reported to date. Mode of delivery Many studies of birth asphyxia have demonstrated an association with emergency operative delivery and instrumental delivery. The Perth data finds an approximate doubling of risk for encephalopathy in both groups compared with vaginal delivery. This association on univariate analysis presumes no causal link. For instance emergency operative delivery for obstruction following malpresentation is unlikely to be the cause of birth asphyxia; in a typical clinical scenario an obstructed breech delivery is delivered by emergency caesarian section - malpresentation is upstream of the operative delivery in the causal chain. Past studies have also shown an association with induced delivery although the recent Perth study did not find a significant association (adjusted OR 0.97). This issue is discussed more fully in chapter 7. Interestingly the Perth study has demonstrated an important protective effect of elective caesarian section (OR 0.17 95% Cl 0.05-0.56) when compared with spontaneous vaginal delivery. The authors explore this association in some detail and argue it is likely to be a robust finding. They conclude "Elective caesarian sections may exert their apparent beneficial effects by avoiding some of the intrapartum risk factors for encephalopathy". These include "post­ maturity, persistent occipito-posterior position, intrapartum maternal pyrexia and catastrophic events in pregnancy". Intrapartum hypoxia The question of how much neonatal encephalopathy is accounted for by intrapartum hypoxia is analogous to that of the contribution of birth asphyxia to cerebral palsy. Low investigated this question in the early 1980's on a case series of 92 encephalopathic infants of whom 42 were of term gestation*98^. The investigators classified 65 as mild/moderately encephalopathic and 27 as severely encephalopathic. They used a biochemical definition of fetal hypoxia as an umbilical artery buffer base concentration of <34 mmol/1. This criteria was met by only 12% of the mild/moderate group and 22% of the severe group. Microbiological evidence of pathogenic infection explained a further 3% and 22% respectively. The authors felt on clinical grounds that neonatal respiratory complications contributed in 14% and 44% respectively with some overlapping of the previous categories. By this group's account 72% of mild/moderate and 37% of severe remained unexplained. The merging of preterm with term infants and the restricted definition of fetal hypoxia limits the comparability of this study but it makes two important points. Firstly causes of early neonatal encephalopathy may extend into the neonatal period and importantly include respiratory complications. Secondly evidence of intrapartum compromise appears to be associated with more severe encephalopathy. Remarkably few studies have addressed this important question in the intervening 15 years. Only now with the growing research focus on neonatal encephalopathy as 44 defined by the World Federation of Neurology Group is it being revisited. The Perth Group have adapted a case definition for possible intrapartum hypoxia from a report of the American Academy of Paediatrics and College of Obstetricians^173*. The criteria are • abnormal intrapartum cardiotocogram or abnormal heart rate on auscultation or fresh meconium in labour accompanied by • a one minute Apgar score of less than three and a five minute Apgar score of less than seven 31 of 164 cases (19%) and 2 of 400 controls (0.5%) fulfilled these criteria. A further 16 cases had clinical evidence of a significant intrapartum event (9%). The authors conclude that 47 (29%) infants with encephalopathy as they broadly define it may have been exposed to intrapartum hypoxia. Of these only 7 (4% of their case series) had no antepartum risk factors found to be significantly associated with encephalopathy in their study. They conclude that intrapartum hypoxia alone accounts for only a small proportion of cases of newborn encephalopathy in this population. Neonatal factors There is a consistent excess of neonatal encephalopathy in male newborns which remains unexplained. It is apparent that hypoxic-ischaemic injury may continue into the neonatal period, exacerbating cerebral pathology originating in the intrapartum period or before. Important factors include: • early resuscitation efforts • hypoglycaemia • meconium aspiration syndrome • recurrent neonatal apnoea • cardiogenic shock As discussed above neonatal hypothermia, to the extent that it slows cerebral metabolism, may be cerebro-protective during acute hypoxic-ischaemic injury^174*. It has been suggested that the observed association of encephalopathy with maternal pyrexia may be due to this factor rather than occult infection. In practice it is impossible to distinguish the relative contribution made to hypoxic- ischaemic injury by intrapartum and early postnatal factors. Birth in the context of asphyxia is best considered as a broad perinatal period commencing with the onset of labor extending to 24 hours after birth. Early life experiences such as resuscitation, feeding practices and the management of meconium aspiration may all modify the effects of intrapartum asphyxia. 45 Investigations for birth asphyxia Biochemical Various investigators have sought a biochemical marker for asphyxia. Low has compared several presumed manifestations of birth asphyxia with a measure of acidosis (umbilical artery buffer base concentration of <34 mmol/L) as a gold standard^ 175\ He argues that the latter is likely to be the best available measure of cerebral tissue oxygenation. The derived sensitivity and false positive rates are given in Table 4.

Table 4. Comparison of several markers of birth asphyxia with a measure of umbilical artery base concentration as a gold standard.

Marker of birth asphyxia Sensitivity (%) False Positive (%) Late decelerations (fetal heart rate) 50 >50 moderate/severe meconium 32 95 Apgar 0-3 at one minute 46 84 Apgar 0-3 at five minutes 8 73 Grade 3 encephalopathy 23 79

This convincingly demonstrates the lack of congruence between acid-base measures and the other commonly adopted proxies for birth asphyxia. Others are sceptical of the use of acid base measures as a gold standard. Dennis both in a review of the literature and in a long term follow up study convincingly demonstrates that umbilical artery acid-base values themselves show no correlation with long term neurodevelopmental outcome (176) These authors even go on to speculate that the presence of acidosis is a good sign suggestive of healthy adaptation to intra-partum anoxic stress. It has been estimated that fully 80% of infants bom with an umbilical arterial blood pH below 7.00 are normal at follow up(177>. Although an important clinical tool it lacks specificity for epidemiological purposes. Other biochemical markers of cerebral damage are under investigation. Several groups have studied the behaviour of an isoenzyme of CreatinePhosphoKinase as a marker of perinatal asphyxia^102). More recently investigators have evaluated the prognostic power of a structural astroglial protein (glial fibrillary acidic protein) which is elevated in cerebrospinal fluid following asphyxia and a neuronal glycolytic enzyme (neuron specific enolase) which is also released by damaged cerebral tissue. These studies are summarised in a recent review (178\

Neurophvsiologv Neurophysiological assessment by EEG was initially used in the original description of the syndrome of HIE by Samat and Samat <44). Subsequently measures of abnormal 46 neurophysiology using single-channel amplitude-integrated EEG recordings as well as delayed or absent visually evoked potentials ^179) and somatosensory evoked potentials 080) jjave ajj been shown to have high positive predictive value for the detection of later neurodevelopmental adverse outcome.

Neuroimaging Space permits only the mention of some of the more important neuro-imaging studies which have assessed the power of ultrasound <181), computed cranial tomography (CCT) (97) and magnetic resonance imaging (MR) (182) to predict outcome in infants generally pre-selected using clinical criteria. These techniques are best considered as tools to study the pathophysiology of neonatal encephalopathy and to aid aetiogical classification in sophisticated research units rather than as epidemiological markers of birth asphyxia per se. I return to the only neuroimaging modality available in this Kathmandu study (cranial ultrasound) below in chapter 3.

Other investigative modalities Various techniques have been used to assess the cerebral haemodynamics associated with perinatal hypoxia-ischaemia. It has been shown that abnormal cerebral blood flow velocity (high or low) has a high positive predictive value for death or severe handicap. Other techniques, notably magnetic resonance spectroscopy and positron emission tomography have been used to relate measures of cerebral metabolism to neurodevelopemental outcome. This literature has recently been summarised^178). Conclusion It is clear that perinatal hypoxic-ischaemia can and does result in irreversible brain injury but explains only a small proportion of all cerebral palsy in high income countries. All infants with significant hypoxic-ischaemic brain injury present with encephalopathy in the early neonatal period. Perinatal hypoxic-ischaemia is only one of many causes of neonatal encephalopathy. In a single report from a high income country there was evidence of intrapartum hypoxia alone in only a small proporton of infants with neonatal encephalopathy. Four central points emerge from this review and influenced the case definition adopted for this study: 1. A prevalence study of birth asphyxia should include an estimate of all fresh stillbirths. 2. All cases of neonatal encephalopathy should be identified in the first instance. 3. The operational definition for encephalopathy should be comparable with previous studies and limited to the term infant. 4. Cases shown to have non-asphyxial causes should be excluded from a study focused on birth asphyxia. 47

2. International perinatal epidemiology The investigation of birth asphyxia and its outcomes in low income countries has been hampered by the problems of data collection endemic to these settings. In for example it is estimated that at least 80% of deliveries take place at home attended by untrained birth attendants 083) Most births are not registered at the time of delivery and stillbirths and neonatal deaths are usually never registered with the authorities. Descriptive studies of birth asphyxia in developing countries have generally used methods of ascertainment based on Apgar scores or the pathophysiological classification of perinatal mortality. More recently there have been three studies looking specifically at encephalopathy rates in the neonatal period. Best estimates of the prevalence of birth asphyxia in low income countries will be briefly reviewed and set in the wider context of perinatal mortality. Perinatal mortality and birth asphyxia The World Health Organisation (WHO) estimate that the overall perinatal mortality rate (PNMR) in less developed countries in 1995 was 57 per 1000 total births compared with 11 per 1000 in more developed countries*l84>. Globally it is estimated that seven million perinatal deaths occurred in 1995*51^. Although PNMR is estimated at 75 per 1000 in Africa compared with 53 per 1000 in Asia, the majority of perinatal deaths (4.5 million) occur in Asia reflecting the continent's high overall population, high proportion of women in the reproductive age group and relatively high birth rate. In another report WHO estimate that '840,000 newborns died of birth asphyxia in 1993 an equal number of survivors suffered brain damage’ *185). The source data for these estimates are indirect measures of birth asphyxia, principally Apgar scores, which have been shown to have poor predictive value for long term neurodevelopmental sequelae in comparison with hypoxic-ischaemic encephalopathy (109). This questions the validity of these estimates of the scale of birth asphyxia related morbidity and mortality in developing countries. A literature search identified 20 developing country studies published in the last 15 years in which the contribution of birth asphyxia to either perinatal mortality, stillbirth or early neonatal death rates was noted in the abstract. Table 5 presents summary data from these studies. Seven studies surveyed geographically defined populations, the remainder were hospital based. A variety of methods were employed. Several used pathophysiological criteria as advocated by Wigglesworth. The proportion of perinatal mortality attributed to birth asphyxia ranges from 24% to 61%. Where sufficient data was available a cause specific perinatal mortality rate for birth asphyxia has been calculated. This lies between 10-20 per thousand total births with the exception of rural Nepal where the overall perinatal mortality was exceptionally high. 48 Table 5. Review of 20 perinatal mortality studies published over the last 15 years which have assessed the relative importance of birth asphyxia in developing countries.

Location (re0 Setting f PNMR SBR Method of cause % PNMR birth asphyxia (Population) of death attributed to cause specific ______attribution ______birth asphyxia PNMR ______N. India C (2,243) 62 31 interview 15* (rural/183-*

Nepal(186) 4 sites Wigglesworth a-urban H (2,783) 23 15 35% 8 b-urban H(10,436) 48 28 44% 21 c-periurban C(470) 42 23 48% 20 d- rural C(1278) 96 34 45% 43

Jamaica^187) C ( 10,482) 41 clinical 61%#

Bangalore, H (5,000) 43 Wigglesworth 24%# India^188)

S Africa (189) C ( 1,795) 38 18 previous child 11

Trivandrum, H ( 13,964) 43 24 clinical 31% 13 S. India(190)

Bangladesh^191) C 75 37 clinical 26% 20

Zimbabwe^192) C 31 Wigglesworth 35% 11

India^193) H (10 57-106 Wigglesworth 26-41% (10 urban sites) studies)

Maharashstra, C (3,173) 28.4 clinical 36% India (rural/194)

Nairobi, H (5293) 36 post mortem 38% 14 Kenya^195) (71% sample) Notes: t setting: Community (C) or Hospital (H) * neonatal mortality only # term infants only PNMR=perinatal mortality rate per 1000 total births SBR=stillbirth rate per 1000 total births

These estimates give a useful indication of the scale of the perinatal mortality associated with birth asphyxia in the total neonatal population. However few distinguish between term and preterm infants. As discussed previously there are important differences between these two groups in terms of the pathophysiology of hypoxic-ischaemia. There may be significant differences in the effect of particular interventions on these two groups. In the perception of many developing country clinicians the prevention of birth asphyxia amongst term infants remains the top priority. There is a need for birth asphyxia specific perinatal mortality data amongst term infants in low income countries. 49

Neonatal morbidity Apgar studies Definitions of birth asphyxia based on Apgar scores vary widely. A review of birth records at the main hospitals in Central, East and Southern Africa based on a five minute Apgar score of less than 7 report a birth prevalence for birth asphyxia of 21.3% of all births^196). A prospective study of low Apgar scores (less than or equal to three at 1 minute) in India over the 1980's found an overall incidence of 7.6%(197). It is likely that these are a major overestimate of the scale of the problem.

Neonatal encephalopathy There are three published studies reporting encephalopathy rates in hospital delivering populations in developing countries (103,104,198). These are summarised in Table 6. Both the Nigerian and Kuwaiti studies were specifically designed to measure the prevalence of HIE. The Zimbabwe study in contrast was principally designed to investigate the prevalence of low Apgar infants but incidentally quotes a prevalence for moderate and severe NE. There is considerable variation between the study findings. This may be because of the varying degrees of selection bias introduced by differing referral patterns in the three areas or it may in part reflect methodological differences in case definition.

Table 6. Comparison of low income studies of neonatal encephalopathy.

Nigeria*404) Kuwait*103) Zimbabwe^198) Year of study. 1987-1989 1989 1986 Study population 6,261 4,591 5,632 Setting referral hospital main hospital main/referral annual deliveries 2 ,0 0 0 18,000 1 2 ,0 0 0 Study design retrospective prospective retrospective single centre single centre single centre

Grade 2/3 HIE/total HIE (%) 76/166 (46%) 22/43 (51%)

HIE Rate (/1000 LB) 26.5 (/1000 TLB) 9.4

Grade 1 HIE rate (/1000 LB) 14.4 (/1000 TLB) 4.6

Grade 2/3 HIE rate(/1000 LB) 12.1 7.8 (/1000 TLB) 4.8

Early Case fatality (%) 18.7 12 Early Neonatal Cause Specific Mortality Rate HIE (/1000 LB) 4.9 (/1000 TLB) 1.1 Notes: LB=livebirths TLB=term livebirths 50 Risk factors for birth asphyxia in low income countries In developing countries it is widely believed that there remains significant scope for the prevention of birth asphyxia by improved intrapartum and neonatal management. Kumari and colleagues describe the early neonatal condition and discharge outcome of 35,959 live births in four annual birth cohorts between 1981 and 1988 in a large maternity hospital in New Delhi^197^. The aim of the study was to determine whether or not new technology and equipment, resulting in improved obstetric and neonatal services, had affected the trends in neonatal outcome for those with very low Apgar scores (three or less at one minute). Their data shows no significant decrease in the proportions of infants with very low Apgar scores over this period (3% in 1981 compared with 4% in 1983 and 1986 and 2.6% in 1988). However there was a significant fall in mortality within the very low Apgar group between 1981 (46%) and 1988 (28%). They interpret their findings to show that 'improved obstetric and neonatal services increased neonatal survival'. They call for' earlier referral and regionalization of perinatal services' arguing this 'would further reduce asphyxia-related neonatal mortality'. This observational study cannot exclude changing population factors to explain the apparent improvement in outcome but it illustrates current hospital based views on strategies to reduce birth asphyxia in India. A more public health orientated approach is found in Daga and Daga's study of a low Apgar population in India. The authors attempt to assess population attributable fractions for risk factors associated with birth asphyxia^164). They found extremely variable attributable risk fractions over the three years of the study for many factors suggesting the limitations of their sample size of 541 infants. Moreover virtually none of the variables selected for study were amenable to intervention. Interestingly, the only factor with an attributable risk consistently above 10% was neonatal hypothermia. This is a timely warning of the dangers of uncontrolled hypothermia in the management of hypoxic ischaemic encephalopathy. A more clinically orientated approach was used by De Muylder to conduct an audit of perinatal mortality in a Zimbabwean district's health centres and central hospital over a 2 year period ^l92\ The investigators classified 111 deaths as Wigglesworth group 4 (fresh stillbirths) constituting 35% of the total perinatal mortality of 30 per 1000.33 fresh stillbirths (30%) followed abnormal labors complicated by inappropriately managed dystocia and a further 19 (17%) suffered either cord prolapse or placental abruption. In the remaining 59 cases (53%) the fetus was alive at the outset of labor and deteriorated during labor for no obvious reason. Of these 53 cases the author notes that the woman presented late in labor in 11 cases, there were problems with poor monitoring in 19 cases, interventions were inappropriate following the detection of abnormal heart rate in 19 cases and there was poor resuscitation in 5 cases. No avoidable factor could be found in 5 cases. The investigators conclude there were 51 avoidable factors involved in 8 6 % of the 111 cases and in more than half of these the avoidable factors were potentially correctable aspects of intrapartum care. For 'avoidable factors' to be translated into actual prevention several preconditions must be met. First and foremost amongst these the risk of impending birth asphyxia must be identified. This is not a straightforward process. Al-Alfy and colleagues report a case control study of clinically defined hypoxic ischaemic (HIE) encephalopathy!1031. Kuwait at the time of this study was a country in transition with high per capita income and is not representative of low income developing countries in general. For instance approximately 75% of both the cases and controls in the study were from either professional or semi-professional families. Virtually all mothers had received antenatal care. No data was collected on indicators of maternal nutritional or micro-nutrient status. Socioeconomic status of the family was not significantly associated with HIE. Consanguinity was significantly more common in HIE cases. There was no long term follow up of cases and therefore some recessive genetically determined conditions masquerading as birth asphyxia may have been missed. Of the obstetric factors nulliparity and maternal hypertension were both significantly more common in HIE cases than controls. The study also found that HIE cases were more likely to have meconium stained liquor and to have undergone longer labour than controls. The latter may simply reflect the excess of primigravid deliveries in this group. Surprisingly there was no excess of low birthweight infants or breech deliveries amongst the HIE group. In conclusion the Kuwait study identified only maternal hypertension as a risk factor amenable to intervention. Outcome Studies Virtually no data has been published describing the longer term outcome of birth asphyxia in low income settings. This is due to the problems of poor communication and population mobility common in developing countries. A recently published study of 45 infants treated for HIE in the intensive care unit of a centre of excellence in South Africa reports outcome at one year for 40 infants with HIE (89% follow up rate )!1061 # Of these four (10%) died within 6 months with evidence of neurological impairment, 13 (33%) survived with neurodevelopmental impairment, including cerebral palsy in all cases, and 23 (57%) infants were normal at one year. All 17 with adverse outcome had suffered either moderate or severe encephalopathy.

Maternity service delivery Comparison of the perinatal mortality data in Table 5 strongly suggests that the likelihood of perinatal death (including the component due to birth asphyxia) is higher in rural areas where obstetric services are less well developed. Figure 2 presents WHO sub regional data demonstrating an inverse relationship between the assistance of a trained birth attendant and perinatal mortality !1841. 52

Figure 2. Perinatal mortality (PNMR per 1000 totalbirths) and proportion of deliveries with trained attendant by sub region (WHO 1995 data).

100

O W. Africa 80 O E. Africa

O S.C. Asia 60 O' E W.Asia

* 40 S.E.Asia O s.America

20 Europe q

0 i — |— i— i— i— i— «— i— ■— 0 20 40 60 80 100

Deliveries with trained attendant (%)

In South Central Asia with a reported perinatal mortality in 1995 of 6 6 per 1000 total births only 30% of deliveries were supervised by a trained birth attendant. A recent study of perinatal mortality in Nepal describes a wide range of PNMR from 23 per 1000 total births in a Kathmandu mission hospital delivering population to 96 per 1000 in a rural population in the rugged North West of the country1186). The Nepal Family Health Survey has assessed the recent obstetric experience (based on three year recall) of over 8,000 women in 253 sampling areas selected to be nationally representative11"). In urban areas 46% of deliveries were attended by a health professional (30% doctor, 16% nurse) compared to 6 % in rural areas. Traditional birth attendants (TBA) were involved in 12% of urban deliveries and 23% of rural deliveries. In 36% of urban deliveries and 57% of rural deliveries a friend or relative was in attendance whilst over 10% of rural women delivered alone. Training for birth attendants Since the early 1970’s there has been considerable interest in the development of low cost training programmes for birth attendants in many developing countries1200). 15,000 TBAs have received training in Nepal alone since 19731201). In many parts of Africa and Asia traditional birth attendants are part of the culture but neonatal resuscitation is not generally part of their traditional practice1202). TBA training programmes are designed to incorporate the principles underlying modem obstetric care with traditional practices1203). Where there are no traditional practitioners mature 53 women are recruited and trained. The aims of TBA training are wide and include the reduction of perinatal mortality through the recognition and management of birth asphyxia^201). Potential strategies to achieve this aim are • antenatal advice/supplements to improve general maternal health in pregnancy • antepartum referral of high risk patients to birthing centres/obstetric units where more intensive monitoring and intervention is available • recognition of fetal distress and intrapartum transfer to an obstetric unit • early management of asphyxiated infants including basic resuscitation • appropriate referral of significantly asphyxiated infants

Antenatal risk asessment The identification and referral of high risk patients during pregnancy is an attractive concept. It underlies much of the rationale for antenatal check-ups. Many formal risk scoring systems have been developed and tested. The results are surprisingly equivocal. This is partly for methodological reasons. The introduction of a scoring system formalises obstetric management and usually brings with it a host of co-interventions which makes it difficult to identify the effect of risk scoring in isolation. In other words improvements in outcome usually result from a whole package of change of which formal risk assessment is only one part. This suggests the initiation of a system of risk assessment may be a strategic method for introducing a new package of service delivery. Different scoring systems have different objectives. Of interest in the context of birth asphyxia are those designed to predict either perinatal mortality or low Apgar scores. In a review of the published literature only 2 of 15 studies reviewed had a positive predictive value for perinatal death greater than 10%(204). All had high negative predictive values greater than 90%. In other words the risk assessment was good at identifying those at low risk of perinatal death but incorrectly assigned to the high risk group large numbers of women with a normal infant outcome. Since the proportion of the total delivering population considered at risk was high (10-50%) this has major service implications. Six studies are reported which have specifically predicted infants with low Apgar scores. The best of these achieved a positive predictive value of 41%, with 27% of the population considered at risk^205^. The setting was a remote rural community where the incidence of low Apgar scores was high (13%). Predictive values will vary markedly according to the population amongst which the scoring system is used. I could find no published data describing the predictive value of risk assessment schemes for the detection of NE in developing countries.

E.3-Q.e,Q£idi.Yery Although attractive in theory many TBA programmes suffer from poor referral pathways to centres where fetal monitoring, surgical obstetric intervention and 54 appropriate neonatal care is available. The author has personal experience of working in Humla in the rugged North West of Nepal where the nearest hospital facility is 5 days walk away. In my experience in such a remote rural setting antenatal referral by the TBA rarely, if ever, occurs. Interestingly I have documented a case of self referral in a woman with a poor obstetric history in Nepal. She moved from her rural home down to Kathmandu and wage laboured for the duration of her pregnancy in order to be near to a hospital facility at the time of delivery. Similar practices of informal temporary relocation by women from poor communities may be more common than is appreciated. Given the relatively short time lag (in some cases less than one hour) between the onset of labour difficulties and cerebral injury intrapartum referral is hazhardous where, as in many low income countries, communications are difficult. Maternity waiting homes (MWH), as described in a study from Zimbabwe, can be an effective approach to the problem of service delivery in rural areas(206). These are hostels neighbouring an obstetric unit where higher risk mothers are encouraged to come and spend the last few weeks of their pregnancy to be conveniently placed when labour commences to enjoy early monitoring and if necessary intervention. The investigators compared hospital outcomes of 1573 pregnancies who stayed at a maternity waiting home (MWH) with 2915 who did not and found the relative risk of perinatal death of infants bom to women coming from their own home to be 1.7 (95% Cl 1.1-2 .6 ) unadjusted and 1.5 (95% Cl 0.95-2.5) after adjustment for potential confounding factors. Next they compared outcomes for the subgroup with risk factors (950 women using MWH and 1488 presenting directly from home) defining risk to include all women with medical risk factors, a history of perinatal death or complicated delivery, height less than 150 cm, non-cephalic presentation and multiple pregnancy. In this subgroup the relative risk for perinatal death of infants bom to women coming from their own home compared with those staying at the MWH was 1.9 (1.1-3.4) when adjusted for potential confounding. It should be noted that the proportion of this population considered 'at risk ' was more than 50% suggesting considerable numbers would have to make use of the facility before significant reductions in perinatal mortality could be expected. It inevitably makes high demands on the user population who must therefore be convinced of its value if uptake is to be satisfactory. Even when health services are geographically accessible utilisation rates may differ widely from programme to programme. Perceptions of service quality and cost considerations are clearly relevant to user choice^207). Another approach is the provision of low cost midwife run birthing centres accessible to the target population. These allow for the prompt recognition of intrapartum complications and speed referral to more sophisticated obstetric units. Siting birthing centres in periurban areas can relieve the pressure on central specialist units which may translate into improved quality of care for a selected high risk population. This is 55 certainly the perception of Mission hospital staff in Kathmandu where a birthing centre has been located in the grounds of the Patan hospital (Adhikari N. - personal communication).

Domiciliary resuscitation In 1989 Indian investigators recommended that TBAs conducting more than 30 deliveries a year should be given a bag, mask and disposable mucus suction traps while those handling less cases should be given the disposable mucus suction traps and training in mouth-to-mouth resuscitation^208^. But what is the evidence that early domiciliary resuscitation modifies the effects of intrapartum asphyxia? A Chinese study has demonstrated that asphyxia case fatality fell from 7% to less than 1% in a Chinese province after grass root level health functionaries were trained in modem methods of neonatal resuscitation^209^. The low however suggests this group included many infants with transient neonatal depression, a self correcting phenomenon. A recent study from Chandighar, Northern India amongst a population with a perinatal mortality rate of 55 per 1000 total births claims a 20% reduction in perinatal mortality for the subgroup attended by more highly trained TBAs compared to those less well trained^210). The training emphasised the recognition of signs of asphyxia, the importance of clearing the airway and providing mouth to mouth resuscitation in severe cases. If this finding is confirmed by other studies it would be important validation of TBA programmes from the neonatal perspective. An important study has demonstrated that room air is equally efficacious as 100% oxygen for the resuscitation of asphyxiated infants of lOOOg birthweight or more ^21 ^ . A cheap, portable mask for use with mouth-to mask resuscitation has also been developed and demonstrated to be as effective as the more conventional bag and mask for neonatal resuscitation (2l2\ However the mouth-to-mask method was found to be tiring and uncomfortable by health service personnel in the two centres in and Bombay where the trial was performed. These developments strengthen the case for resuscitation training and equipment for TBAs - a measure yet to generate widespread support. But as the Chandighar group acknowledge an individual TBA deals with comparatively few births per year. Indeed the Nepal family health survey suggests TBAs deliver less than 25% of newborns even in rural areas. For example a TBA in rural Nepal will rarely deliver more than 5 infants per month in Nepal (Levitt M.-personal communication). Given a birth prevalence of birth asphyxia of 6 per 1000 term livebirths she will be called upon to resuscitate an asphyxiated term infant once every three years. Over 10,000 TBAs in Nepal alone would require equipping, training and periodic refresher training. The comparative 56 rarity of the problem and the resource requirements are likely to significantly limit the widespread utility of this approach.

Intrapartum management Fetal monitoring Several studies in industrialised countries have reported the disappointing impact of fetal monitoring on presumed outcomes of birth asphyxia. In the influential Dublin Study of fetal monitoring for instance, continuous electronic fetal monitoring (EFM) was found to have no advantage over intermittent auscultation when the outcome of cerebral palsy was considered^85). A recent review concludes that fetal monitoring generates a huge proportion of false positives that receive unnecessary and in some cases harmful interventions in order to prevent a handful of adverse outcomes^75). Generalising from this sort of study finding to low income countries can be misleading. The study methodology of the Dublin trial compared the best possible fetal auscultation practice with EFM in a setting of individual one to one care where birth asphyxia was relatively rare. In practice both the technique and frequency of fetal auscultation is often far from optimal in developing country maternity hospitals. A randomised trial of intrapartum fetal monitoring in Zimbabwe(213) compared the detection rate of abnormal fetal heart rates in 1255 labours randomly assigned to electronic fetal monitoring (EFM), hand-held Doppler ultrasound monitoring (Doppler), auscultation for the last 10 minutes of every half-hour by research midwives (Regular) and routine auscultation by the duty midwife (Routine). Compared with routine monitoring the relative risk (95% confidence intervals) of detecting abnormal fetal heart rate was EFM 6.1 (4.2-8.8), Doppler 3.6 (2.4-5.3) and Regular 1.7 (1.1-2.7) respectively. HIE rates were lowest in the EFM and Doppler groups and highest in the Regular and Routine auscultation groups although the power of the study was insufficient to assess the significance of this difference. Ironically perinatal deaths were equally common in the EFM group as in the routine monitoring group. The investigators ascribe this to delays in intervention either due to failure to follow protocols or to lack of available operating facilities. From this study it appears that the detection rate of fetal heart rate monitoring is dependent on both the motivation of staff and the technique of fetal monitoring used. It remains to be seen if in developing country hospital settings this can be translated into lower birth asphyxia rates. Portograms The concept of the active management of labor requires close monitoring of the progress of labour and a commitment to intervene when the labor process is found to be deviating from normal. The WHO, as part of the safe motherhood initiative, has developed a partograph to facilitate the monitoring of labor. The central feature is a 57 cervicograph which plots cervical dilatation against time. An "alert line" (recommending referral to a central hospital) is crossed when the rate of cervical dilatation falls below 1 cm/hour. An "action line" (recommending active intervention) lies parallel to and 4 hours beyond the alert line. An important multi-centre historically controlled trial of this partograph was performed in South-East Asia involving over 35,000 women in 1990(214^. No new interventions were introduced in any of the 8 participating hospitals as part of this study but the same formalised protocol for partography and the subsequent management of labor was introduced to all centres. The proportions of prolonged labor, labors requiring augmentation and emergency caesarian sections all fell significantly following the intervention. In this study intrapartum stillbirth refers to stillbirths with evidence of life at the time of hospital presentation. These fell from 93 (18%) of 516 total stillbirths in 18,254 total deliveries before the intervention to 55 (13%) of 413 total stillbirths in 17,230 total deliveries after the intervention. There was a significant reduction in the risk of stillbirth (risk ratio 0.85 95% C.I. 0.75-0.96) which was more marked for intrapartum stillbirth (risk ratio 0.63 95% C.I. 0.45-0.87). Survivors of birth asphyxia were assessed crudely by Apgar scores and admission rates to neonatal special care neither of which showed significant change. This study provides important evidence that partography, in countries where the stillbirth rate remains high (here 28 per 1000), can reduce stillbirth rates in general and intrapartum stillbirths in particular, provided it is accompanied by agreed intervention protocols which are adhered to. Amnioinfusion For women with meconium stained liquor noted during labour there is increasing evidence for the intervention of amnioinfusion whereby normal saline is infused via the patent . A recent study from Zimbabwe found an impressive protective effect with respect to hypoxic-ischaemic encephalopathy (OR 0.07 [95% Cl 0.01-0.56]) with no evidence of increased risk of puerperal pyrexia(215>. Although there remains uncertainty in its potential effect on vertical transmission of HIV infection a recent systematic review recommends its use where peripartum surveillance is limited as in many low income settings^216). 58 Industrialised country comparative data Prevalence data Several studies have assessed the prevalence of neonatal encephalopathy in industrial countries. The size of population, method of case definition and estimated prevalence (overall and by grade) for four published studies are given in Table 7. These were conducted in Sweden^217), Leicester^") and Derby^218^ UK and Perth, Australia^165).

Table 7. International comparison of the prevalence and early outcome of neonatal encephalopathy in high income countries.

Study Location (re0 Goteborg, Leicester, Derby, Derby, Perth, Sweden^217) u k (99) UK<218> u k (218) Australia^165) Study period 1985-1991 1980-1983 1976-80 1984-88 1992

Study population 42,203 20,975 24,824 24,265 11,200

Setting central central district district central hospital hospital population population hospitals

Study design retrospective retrospective retrospective retrospective prospective single centre single centre single centre single centre multicentre

Case definition fetal distress fetal distress HIE HIE NE (incl. + low Apgar plus HIE congenital plus HIE malformation)

NE rate per 1000 LB 7.7 (6.6-8.8) 4.6 (3.8-5.5) NE rate per 1000 term LB 1.8 (1.4-2.2) 6.0 (5.0-7.1) 7.9 (6.4-9.8)

Grade 1 NE rate per 1000 LB 5.1 (4.2-6.0) 2.8 (2.1-3.6) Grade 1 NE rate per lOOOtermLB 1.0 (0.8-1.5) 3.8 (3.0-4.7) 4.2 (3.1-5.6)

Grade 2/3 NE rate per 1000 LB 2.6 (2.0-3.2) 1.8 (1.3-2.4) Grade 2/3 NE rate per 1000 term LB 0.7 (0.5-1.0) 2.1 (1.6-2.9) 3.7 (2.7-5.1)

Grade 2/3 NE/total NE (%) 29/75 (39) 46/126 (37) 63/189 (33) 44/112(39) 42/89 (47)

Early NE Case fatality (%) 17 5 12 11 8 Early Neonatal Cause Specific NE Mortality Rate per 1(XX) LB 0.9 (0.6-1.3) 0.5 (0.3-0.9) per 1000 term LB 0.3 (0.2-0.5) 0.3 (0.1-0.6) 0.6 (0.3-1.3)

Grade 2/3 survivors per 1000LB 1.6(1.1-2.1) 1.3 (0.9-1.9) per 1000 term LB 0.4 (0.2-0.6) 1.7 (1.2-2.3) 3.1 (2.2-4.3) Notes: LB=livebirths term LB=term livebirths (95% Confidence Intervals)

There appears to be a trend of declining HIE rates in the industrialised world. The Perth data reports a significantly higher prevalence using an intentionally broad definition of encephalopathy. A preliminary report of a population based study in South West 59 Thames describes a prevalence of neonatal encephalopathy of 2/1000 despite using a similarly broad definition of encephalopathy^219^ . Outcome studies Many groups have reported cohort studies of HIE. A meta-analysis of the largest equivalent studies covers a total of 459 cases of HIE bom in North America or Europe between 1974 and 1984 with a follow up duration ranging from 1.5 to 3.5 years<117). The overall risk of death in the populations covered by these studies was 12.5%. The authors estimate the risk of death by grade of encephalopathy to be near zero for grade one, 5.6% for grade two and 61% for grade three. Regarding severe impairment the overall risk was 14.3% with grade specific risks of near zero for grade one, 20% for grade two and 72% for grade three. In a separate review of five studies (with follow up duration from 1 to 3.5 years) Levene reported the proportion dead or severely abnormal of 15-25% following moderate HIE and 50-100% following severe HIE. It is apparent that with optimal care the outcome of mild encephalopathy is excellent, severe encephalopathy has a poor prognosis and moderate encephalopathy has a variable outcome. lhswider,,cpirts,^y Perinatal mortality Wigglesworth, using a pathophysiological classification of stillbirths found 7 of the 10 deaths over 2,500 g birthweight were due to birth asphyxia in a major London referral centre in 1978/79^351. However, when all perinatal deaths were included this proportion fell to 7/74 (9%). A retrospective study comparing moderate and severe NE with intrapartum stillbirth rates among infants of 35 weeks or greater gestation in two districts of the UK between 1982 and 1984 provides comparative data(218). The investigators reported post-asphyxial stillbirth to be 0.4-0.8/1000 total births for the two districts and the prevalence of moderate/severe NE to be 0.6-1.9/1000 livebirths. Recent figures for the UK give a combined figure for perinatal death due to intrapartum asphyxia/trauma of 1.1/1000 total births, approximately 10% of overall perinatal mortality^220). Childhood disability What is the prevalence of cerebral palsy attributable to birth asphyxia in high income countries? The largest prospective cohort study (the NCPP) designed to assess the contribution of perinatal events to childhood impairment in an high income setting found just 21% of all cerebral palsy in their cohort of children by school age to be associated with adverse perinatal factors amongst a population of nearly 40,000 with a CP prevalence of 3.1 per lOOO^1281. This generates an estimate of birth asphyxia related cerebral palsy of 0.6/1000 total births. Hull and Dodd in Trent, UK found a CP rate attributable to birth asphyxia falling from 1.8/1000 in 1976-80 to 1/1000 in 1984- 60

88(2 18)> a recent study from Oxford estimated the contribution of birth asphyxia to cerebral palsy in a UK population by prospectively assessing the outcome at 5 years of age of 160 term newborns with evidence of intrapartum asphyxia drawn from a birth cohort of 11,100<113>. The authors found only three of this group developed major neurodevelopmental impairment by age five whilst from the same birth cohort a further 27 with no evidence of intrapartum asphyxia developed cerebral palsy. They estimated that the frequency of cerebral palsy associated with birth asphyxia was one in 3700 full- term births (0.3 per 1000). In this population birth asphyxia at term was associated with 10% [95% confidence interval (Cl) 2.1 - 26.5] of all cases of cerebral palsy and with 20% [95% Cl 4.3 - 48.1) of the 15 cases of cerebral palsy in children bom at term. In conclusion an upper estimate of the prevalence of major impairment following birth asphyxia lies between 0.3 and 1 per 1000 livebirths in high income countries.

There have been striking trends in the short historical record since regional handicap registers have permitted the sort of epidemiologically rigorous studies described above. Stillbirth rates have fallen steadily whilst CP rates have remained relatively unchanged ^l24\ The implication is that improving maternal health and perinatal services in industrialised countries have had a huge impact on perinatal mortality but relatively little impact on long-term disability. 61 Study hypotheses In the light of this literature review the study was designed to test the following hypotheses grouped into three broad areas:

The Prevalence of birth asphyxia

Neonatal encephalopathy(NE) is a clinically evident syndrome in term infants which can be reliably identified by junior doctors in a busy, low resource Kathmandu maternity hospital.

The birth prevalenceofNE is higher and its severity is worse in this low income setting than in high income settings.

The combined populations of fresh term stillbirths and term infants dying with NE in the early neonatal period make an important contribution to overall perinatal death rate.

A one minute Apgar score of three or less has a low positive predictive value for neonatal encephalopathy.

Risk factors for NE and fresh stillbirth

A higher proportion of cases of NE in a low income setting experience adverse intrapartum events in comparison with NE cases in high income countries. Fresh stillbirths experience greater intrapartum adversity than NE cases.

Maternal anaemia and hypothyroidism are independent risk factors for the outcomes of both fresh stillbirth and NE.

Antenatal care as currently practiced in Kathmandu protects against neonatal encephalopathy and fresh stillbirth.

Outcome of neonatal encephalopathy

The risk of neonatal (first month) mortality is higher in mild NE infants than normal controls in a low income setting. The risk of post neonatal first year mortality (between ages 1 and 12 months) amongst all survivorsofNE is higher than amongst normal controls.

The risk of neurodevelopmental disability at one year in infants who have experienced moderate NE is higher in low income countries than in comparably affected infants in high income countries.

During the course of the project as work elsewhere developed in the field of NE two further hypotheses were suggested which we hoped to test with the study data.

Maternal hypomagnesaemia is a risk factor for the outcomes of both NE and fresh stillbirth.

Markers of inflammation are raised postnatally in the mothers of infants who have suffered adverse pregnancy outcome (both NE and fresh stillbirth) in comparison with the mothers of normal controls. 3.Study design, setting and methodology Study design

EiML.study_de.siga:. For the purposes of presentation the study is best considered as composed of three interlocking but analytically distinct component studies. Some subjects appear in all three components of the study. The prevalence of birth asphyxia Design: Prospective descriptive cross-sectional survey Period: January 1995 - January 1996 Subjects: 14,771 births Outcomes: Fresh stillbirths over 2.0 kg Neonatal encephalopathy in term infants Perinatal mortality Risk factors for neonatal encephalopathy and fresh stillbirth Design: Retrospective case control study Recruitment: Cases: January 1995 - July 1996 Contemporaneous controls: January 1995 - July 1996 Later controls: July 1996 - July 1997 Subjects: 131 NE cases 107 fresh stillbirths 208 contemporaneous controls 635 later controls Outcomes following neonatal encephalopathy Design: Prospective cohort study Recruitment: January 1995 - July 1996 Follow up: 1 year Subjects: 131 NE cases and 208 normal controls Outcomes: Death by one year Neurodevelopmental impairment at one year

Initial aims, primary outcomes and sample size calculations The original aims and associated sample size calculations accepted by the Wellcome Trust for funding of this research project were as follows: Aim 1. To set up a screening and follow-up system for hypoxic-ischaemic encephalopathy in the government maternity hospital and the community it serves in Kathmandu, Nepal. Aim 2. To measure the incidence of hypoxic-ischaemic encephalopathy (HIE) among term newborn infants delivered in the maternity hospital. Sample size calculations estimated 64 that to measure birth prevalence with a relative precision of 2% and a confidence interval of 90% a birth sample size of 6766 would be required. Aim 3. To identify the most important preventable risk factors associated with the development of HIE so that cost-effective interventions may be designed. Assuming an incidence rate of HIE of 25 per 1000 (4 times the UK incidence) approximately 200 infants within a cohort of 8000 will develop HIE. All HIE affected infants will be enrolled for follow-up and control infants (1:1 ratio) will be systematically selected (the next eligible infant). The anticipated probability of exposure of normal controls will vary from perhaps 2% () to 60% (maternal anaemia). The odds ratios at which a significant difference (5% level, 80% power) will be detected for HIE infants (sample size n = 200) will range from 1.8 - 3.4. Aim 4. To prospectively compare a) the incidence and nature of early neurodevelopmental sequelae (up to 1 year of age) and b) the mortality rates in study and control infants. Assuming an rate of 10% in controls, the sample size required to detect a mortality relative risk of 2.5 in the HIE group is 168 (95% confidence interval, 90% power). Problems and modifications affecting study design It was apparent early in the study that the prediction of the prevalence of neonatal encephalopathy (NE) made for the purposes of sample size calculation significantly exceeded the observed prevalence. Various scenarios were explored in interim sample size calculations in the light of a revised prevalence estimate of 6.5/1000 after 8 months of recruitment (the point in the original study design when recruitment was to have been completed). Table 8 presents the revised case control sample size calculations for detectable odds ratios set for the same statistical tolerance as in the original calculations (5% level at 80% power) for exposures of interest at that time.

Table 8. Sample size calculations showing detectable odds ratios for exposures of potential interest based on different sample size and control:case scenarios.

Revised Odds Ratio Odds Ratio Odds Ratio Odds Ratio Exposure of interest exposure n=200 n=120 n=150 n=143 estimates 1 con./case 1 con./case 2 con./case 3 con./case

Maternal hypothyroidism 10% 2.3 2.9 2.3 2.0 Maternal anaemia 27% 1.8 2.2 1.8 1.5 No antenatal care 45% 1.8 2.1 1.8 1.3 Syntocinon in labour 15% 1.8 2.5 2.0 1.8

Note con./case = control:case ratio In order to retain similar power as the original design for the key exposures of interest it was decided to extend recruitment to eighteen months and double control recruitment to 2 controls per case. A second more detailed interim analysis after 12 months of recuitment of the control infants revealed that no controls had required special care, suggesting a recruitment bias. When I investigated this surprising finding I realised that due to a misunderstanding the research nurse assistants had excluded any infants admitted to the Special Care Baby Unit (SCBU) making the error of confusing the concepts of control and normality. This unfortunately had the net effect of excluding an interesting group of infants from the study. Since one of the admission criteria to SCBU is an Apgar at one minute of three or less it meant all those who were bom in poor condition but who did not go on to develop encephalopathy had been excluded from the control group. It seemed possible that exclusion of this group of infants would reduce the study power for the detection of the nature of any hypothetical protective factor (preventing acute distress from progressing to irreversible brain injury) leading to the possibility of type 2 ('false negative') error in a multivariate analysis. For this reason an intradepartmental assessor advised that whilst the group of normal controls recruited contemporaneously would be ideal as a comparison group for the follow up study, a new unbiased cohort should be recruited as controls for the case control study (McGregor S.- personal communication). Therefore the study was extended to permit recruitment of every 25th delivery who satisfied the case definition in all regards except for encephalopathy. This "later" control group was recruited over the 12 months immediately following the case recruitment period. Over the entire recruitment period there were no significant changes in hospital policy or resources. Given the often frustratingly slow pace of organisational change in Nepal short term secular change seemed unlikely to significantly affect the exposures of interest over a time period of only 12 months. We estimated this revised control recruitment protocol would generate an unmatched control: case ratio of between three and four to one. No other study design modifications were made. 66 Study Process Recruitment and follow up programmes were sustained for a 2.5 year period from January 1995 to July 1997. The various study components are presented as a time bar plot in Figure 3.

Figure 3. Time bar plot depicting the Kathmandu study process.

Wellcome S udy Funding: March 1994 - Vlarch 1997

Pilot recruitment -1 8 monthsCase study case follow up to age 12 i nonths (30 months total duratii in)

"conte nporary "control recruitment-18 months

"Later" contro recruitment - 12 months

Stillbirth recruitment - 18 month!;

perinat il data monitoring period lab analy: lis

v/riting up period

March 1994 March 1995 March 1996 March 1997 March 1998

Setting This study was performed in Nepal, a South-Central Asian state wedged between China and India.

GgQgrapfry pf Nqrel Nepal is a relatively small country, some 500 miles East/West and never more than 110 miles North/South. Geographically it divides into three distinct ecosystems with the low subtropical flat plains of the Terai contingent with India to the south and the high alpine mountain chain of the Himalayas forming the border with Tibet to the North. Between these climatic extremes the middle hills are densely populated undulating ridge systems between 1500 and 3500 metres in altitude crosscut by deep river valleys. The regional watershed lies to the North of the Himalayas on the high Tibetan plateau and six powerful river systems divide the relatively young and still growing mountain chain into a series of culturally distinct Himals, before draining into the Ganges away to the South. Until modem times the forrested Terai and its well deserved reputation for particularly virulent endemic malaria was an effective barrier to widespread population movement between the hills of Nepal and the urban centres of the Gangetic plains. To the North important trading routes wind up the river valleys through the Himalayan 67 chain to Tibet but the vast and inhospitable plateau remains an effective barrier to large scale population movements^221). Nepal and its health system in the twentieth century Nepal's geographical isolation, poor natural resources, reputation for bellicosity and location in the buffer zone between British and Russian imperial designs left the country largely free of external interference until the early 1950's. From 1846 to 1951 the ruling Rana dynasty of hereditary chief ministers maintained closed borders, manipulated the nominal royal family and in effect ran the country as their private feudal fiefdom. A department of health services was established in 1933 and some allopathic hospital facilities were developed but services were largely limited to the ruling elite and military personnel. Only with the support of the newly independant state of India was the constitutional monarchy restored in 1951 and an international airport opened in Kathmandu. At this time there were 12 allopathically trained Nepali doctors and 625 hospital beds in the whole country^222\ There was effectively no allopathic primary health care system available to the general population. Ayurvedha, based on the Hindhu belief system, Tibetan medicine derived from Buddhist concepts of health and disease and shamanistic healing were the mainstay of popular healthcare. The capital was finally linked by road with India only in 1953. Early international health assistance was led by Protestant missionary organisations but by 1956 the government had instigated the first five year development plan including a health infrastructural plan. There was some development of hospital services in the Kathmandu Valley at this time including the founding of Prasuti Griha Maternity Hospital in 1959 and the first locally based allopathic training schemes were instigated with foreign technical support. The first three five year plans (1956-1970) emphasised vertical programmes designed to control leprosy, smallpox, tuberculosis and malaria as well as provide maternal and child health including (usually irreversible) family planning. The Terai malarial eradication programme was especially important as it facilitated clearance of the southern forest belt and massive population resettlement programmes were officially endorsed in a bid to relieve the growing population pressure on the middle hills. During the 1970's and 1980's the government placed increasing emphasis on the development of primary health care in line with Nepal's status as a signatory of the Alma Ata declaration coordinated by WHO in 1978. The five year plans of this period emphasise the need for paramedical healthworkers, community health volunteers and an " Integrated Community Health Program". By the late 1970's this had resulted in the creation of 550 health posts staffed by paramedical workers and 70 hospitals staffed by 450 medical doctors and 350 trained nurses^223) . These impressive statistics are somewhat misleading: doctors and nurses remained a rarity outside urban areas and of 68 the rural healthposts less than a third were sited in their own premises - most being located in a rented village room. Judith Justice in an influential multidisciplinary study of culture and health development researched in the late 1970's describes some of the cultural factors why this well intentioned central planning process failed to actually deliver services at village level^224). In a fledgling bureaucracy ties of kinship and caste conflict with meritocratic notions of staff selection and advancement. Much time and effort is subverted from the ostensible objective of service provision into the parallel agenda of personal advancement. Since all decisions are made at the centre time spent in a peripheral clinic is counterproductive. The net effect is that clinical staff, whilst nominally registered in rural clinics, spent much of their time in Kathmandu activating kin networks (aaphno manchhe or literally 'own people') to influence the appointment system. Clinical services, Justice demonstrates, were actually delivered by the locally based paeon, an untrained janitor. Service development in urban areas, and especially in the Kathmandu valley, has been more effective principally because professionals are prepared to staff urban health facilities. Professional reluctance to work in rural areas is in part due to their desire to supplement their meagre government salaries with private fees which remain beyond the means of subsistence farmers. By 1996 and the completion of the eighth plan the integrated health care system conceptually had developed into a hub and spoke structure of over 3000 sub-health posts each providing basic primary health care to village development areas of some 4,000 people, referring into 200 primary health centres each equipped with a three bed unit backed up by 5 zonal referral hospitals. The current plan envisages urban health needs to be met by an urban government clinic for every 50,000 people backed up either by zonal referral hospitals or,in Kathmandu, by national hospital centres of excellence. In reality the majority of sub-health posts have patchy staffing at best and many rural primary health centres remain to be built. Hospital beds have expanded to over 5000 nationally^225). Kathmandu, however, remains a patchwork of service provision with every conceivable combinaton of private, non­ governmental, governmental and international agency supplying both primary and secondary services. There is little co-ordination of services and few effective referral pathways other than those following personal professional links.

Nepal: economic, demographic and health indicators Nepal remains a low income country ($200 per capita in 1995/226) with a high per annum population growth rate of 2.5% over the period 1991-1995 resulting in a population of 20.9 million as of 1995. Each year there are estimated to be over 700,000 births^227). The urban population, currently estimated at 13% of the population, grew at 69 a rate of 7.4% per annum between 1990-94 which was the highest among South Asean Association for Regional Cooperation (SAARC) member states^226). The Nepalese economy is based on subsistence agriculture with over 90% of the labour force working predominantly on the land and a manufacturing share of Gross Domestic Product of only 5%(228). Despite its dependence on subsistence agriculture it has poor natural resources with only 12% of its total land mass arable. This combined with unsustainable population growth has led to large scale deforestation and secondary problems of soil erosion. The vast potential of hydroelectric energy resources remain largely untapped due to the logistic difficulties of conducting major engineering schemes in isolated areas with poor physical and organisational infrastructure. Of every 1000 children bom in Nepal it is estimated that seven die on their first day of life; an additional 16 by the end of the first week; another 30 by the end of the first month and another 54 by the end of the first year. By the end of the fifth year 165 of every 1,000 livebom infants will have died<229). 65% of the adult population is illiterate with 82% female illiteracy. Total fertility rate (lifetime births) is estimated to be approximately 4 in urban areas and 6 in rural areas. Estimates of maternal mortality rates in Nepal vary between 515 and more than 1,000 maternal deaths per 100,000 births.

The Kathmandu Valiev The Kathmandu Valley consists of an alluvial plain some 30 kms by 20 kms ringed by hills reaching 2,000-2,700 metres. Despite an altitude of 1,200-1,500 metres it enjoys a subtropical climate owing to its southerly latitude of 27°. It is the site of a growing connurbation comprising of the three traditional municipalities of Kathmandu, Lalitpur and Bhaktapur with a combined population in the last national census in 1991 of nearly 600,000. Scattered agricultural villages towards the margins of the valley all lie within easy reach of the urban centre. When combined with the intervening urban sprawl the best estimate of the valley population exceeds one million^230). This has doubled in the last thirty years and reflects the centripetal forces typical of a developing country. The in-migration to the metropolitan centre can be viewed both in terms of pull factors (better urban facilities, a light industrial economy based on labor intensive carpet industry and a vast informal service economy financed both by national revenue and international aid dollars) and push factors (decreasing margins for subsistence farmers and poor educational facilities for rural youth). The net result is a restless, relatively young urban population. Nepal straddles a cultural, linguistic and ethnological fault line between Buddhist Tibeto-Burmese speaking peoples (such as the Sherpas) to the North and Hindu Indo- European speaking peoples to the South. Therefore the citizens of the newly democratic state of Nepal are physically and culturally heterogeneous, with more than 30 70 recognised ethnic groups. The traditional Kathmandu valley inhabitants, the Newars, still form 50% of the Valley population and occupy the centre ground with their varied physical appearance and uniquely syncretic Hindu/Buddhist religion. The Valley walls constitute a considerable barrier to communication traversed by only 3 roads. Travel on these is always slow and frequently precarious. For example the journey to the nearest neighbouring urban centre of Hetauda takes four hours despite being only 50 kms as the crow flies to the South.

Kathmandu maternity health institutions There are 3 main hospitals providing maternity care in the Kathmandu Valley. With their respective figures for total deliveries for a 12 month period during 1992/3 (Nepali calendar 2049) these are the Prasuti Griha Maternity Hospital (13367) - funded directly by government, Patan Hospital(4,002)-funded and run by expatriate protestant mission organisations and the Tribhuwan University Teaching Hospital (2,381) - largely Japanese Aid funded. In addition there are numerous small private nursing homes where the well off choose to deliver. A recent community based study performed by the MIRA research group showed that 81% of Valley residents living in or near Kathmandu choose an institutional delivery^231\ The hospitals levy charges at different rates and patients are asked to buy consumable medicines and equipment such as needles and cannulae as they are required. In a focus group discussion I held in an urban community clinic Kathmandu women estimated the total cost incurred by a routine delivery as 1000 Nepalese Rupees (100 NR=£1) at Pathan, 150NR at Prasuti Griha and 60NR (the price of a traditional 3.5 kg measure of rice) for the services of a sudeni (traditional birth attendant) at a home delivery. They were unsure of the cost at the Teaching Hospital but considered it to be to be the most expensive facility and beyond their means. Both Patan and Prasuti Griha operate an exemption scheme for the poorest mothers. A socioeconomic survey of women delivering at Prasuti Griha showed that total household monthly income lay between 1000 and 5000 NR (£10-50) for the vast majority (Shrestha P.- personal communication). This data supports my impression that Prasuti Griha serves the least well off families who choose to deliver in hospital. Place of birth data from this study sample (unpublished) show these include many first generation economic migrants to the Valley.

Prasuti Griha maternity hospital A recent study has demonstrated that just over 50% of all Kathmandu metropolitan deliveries occur at Prasuti Griha^231^. A recent audit showed less than 3% of women delivering at Prasuti Griha were resident outside the Kathmandu Valley (Manandhar D.- personal communication). Prasuti Griha has 220 beds and an annual total operating budget of £350,000. Between 30 and 40 babies a day are delivered either in labor 71 rooms, the admission room or the operating theatre. Fetal monitoring is by intermittent auscultation, partograms are not used and the caesarian section rate was 10% during the study period. When no fetal heart rate is detected prior to delivery and no pulsation of the umbilical cord is found at delivery resuscitation is not attempted. Otherwise resuscitation, including suction, oxygen by bag and mask or endotracheal tube, and cardiac massage follows standard procedures and is undertaken by the resident on-call paediatric staff. Minimum period of stay for maternity cases is 24 hours. The Special Care Baby Unit (SCBU) has 30 cots. There are no facilities for longterm ventilation. There is no routine referral facility available for level 3 care (ie long term ventilatory support). Between 15th January 1995 and the 14th January 1996,2,503 infants were admitted to SCBU, 17% of all deliveries. The admission policy includes the 24 hour observation of all newborn infants with an Apgar score less than or equal to three at 1 minute. A 1992 audit showed a perinatal mortality rate of 42/1000 total births at the hospital. In that year 1873 infants were admitted to the SCBU with an early neonatal death rate of 18/1000 live births. Severe birth asphyxia was considered to be the most common cause of death. (Manandhar D. -personal communication). Methodology Clinical methods Stillbirth definition Stillbirths were defined as infants bom in the hospital weighing 500 g or more with no evidence of life at birth (Apgar scores of zero at both 1 minute and 5 minutes). Following external examination by the attending clinicians the fetus was clinically classified as fresh or macerated and any evidence of congenital malformation was noted. In Nepal postmortems are only performed on medico-legal cases. Stillbirths were classified according to the pathophysiological criteria described by Wigglesworth for perinatal deaths*35). Stillbirth gestation It is impossible to accurately estimate the gestation of stillbirths by clinical examination. A study of term infants at the study hospital showed that over 30% were of low birthweight as conventionally defined with a birthweight of less than 2500 grams*232). Other studies have shown that intrauterine growth retardation is associated with an increased risk of stillbirth*233). Therefore as an estimate of fresh term stillbirths we included all those with a weight of 2000g or more. Stillbirth ascertainment Details of stillbirths occurring at Prasuti Griha are routinely recorded in both the birth register and a separate stillbirth register. The parents cannot take a stillborn infant from the hospital premises for without a carbon copy of the stillbirth register. Burial of a stillborn infant (in contrast to the usual Hindu practice for maturer children and adults 72 of ) is considered an important propitiating act which lessens the risk of haunting of the parents by the dead offspring's spirit. Both registers were manually checked by the principal investigator for the period January 1995 to January 1996 and details of all stillbirths entered into a computer database. During the case recruitment period paediatric staff were asked to make additional records for fresh stillbirths with a birthweight of 2000g or more and to interview their mothers. Apgar scoring All newborn infants were routinely assessed and awarded Apgar scores at one and five minutes of age. This was done by the attending clinical staff, either a paediatrician or midwife. No special training was given since it was our intention to compare the routine measurement of Apgar scores with subsequent encephalopathy grading. Neonatal encephalopathy - case definition A pilot study prior to final study design suggested detailed encephalopathy protocols used at University College, London based on Samat grading^were time consuming and required considerable experience to both execute and interpret satisfactorily. This conclusion has since also been reached independently by a South African group working in a centre of regional excellence (106). Therefore we decided to rework the examination protocol for the recognition of neonatal encephalopathy (NE) with the objectives of increasing user-friendliness whilst retaining inter observer agreement equivalence with other schemes. Following Fenichel neonatal encephalopathy was defined as an abnormal neurobehavioural state in a term infant commencing within the first 24 hours of life consisting of an altered conscious level with abnormalities of tone and/or sucking behaviour. There may in addition be seizures, abnormalities of respiratory control, primitive and brainstem reflexes. Formal operational criteria to grade the severity of encephalopathy in term infants were derived from the syndromic descriptions of Fenichel^107) modified in the light of more recent studies and observations during pilot testing. In particular these modifications incorporate the observations that mild NE cases may have signs of increased as well as decreased tone U11), that seizure activity may not be clinically detectable^88) and therefore cannot serve as a definitive feature in any grading system and that the inclusion of duration in the clinical definition of a grade ("Venders the scheme contradictory^108). In order to tighten the criteria for the grading of NE essential minimum criteria to reach each grade were described based on pilot study findings. In practise the distinction between minor neurological abnormality and mild (grade 1) encephalopathy was the hardest to define. The solution adopted here requires the observation of either poor suck or an abnormality of tone with the inherently more subjective recognition of irritabilty before the criteria of mild (grade 1) NE are met. During pilot testing we found the distinction between grade 1 and 2, and 2 73 and 3 was more clean cut. Table 9 presents the operational definitions for each of the three grades.

Table 9. Operational definitions for the clinical grading of neonatal encephalopathy.

Clinical variable Grade 1 (mild) Grade 2 (moderate) Grade 3 (severe) conscious level irritable/ lethargic comatose hyperalert

tone either a mildy moderately severely abnormal abnormal abnormal (hypo/hyper) (hypotonic or (hypotonia dissociated) suck or b abnormal poor absent primitive reflexes exaggerated depressed absent brain stem reflexes normal normal impaired seizures absent present present respiration tachypnoea occasional severe apnoeas apnoea Notes: The infant is examined daily and graded according to the clinical features described. The features in bold must be as described to meet the minimum requirements for each grade. a/b abnormality of tone or suck should accompany altered conscious level to assign grade 1. Features not in bold may be present but are not required to make the grade assignation. The overall description of the infant refers to the most severe grade observed during the neonatal period.

Gestational assessment Some gestational assessment techniques (e.g. the widely used system described by Dubowitz <234)) include neurological features which make them invalid for the assessment of the encephalopathic infant. This study uses the Parkin method which estimates gestational age during the first 24 hours of life from an assessment of skin colour and texture, ear firmness and breast size (235\ These four characteristics are scored on a scale of 0-3 (0-4 in the case of skin texture) giving a total possible score of 13. On white skinned infants the conversion of this score to gestational age has been shown to have mean 95% confidence limits of 15 days which is of the same order as other more complex scoring systems. It has been shown to be valid in infants with pigmented skin (236\ We followed Parkin's original validation study in using a cut off score of six or more for term gestation. 74 NE ascertainment strategy The original study design called for every term infant bom in the hospital during the study period to be clinically assessed for signs of NE. At the outset of the study period 1000 consecutive deliveries were examined on the routine post-natal wards for clinical signs of encephalopathy. No cases of NE who had not met the SCBU admission criteria described above were identified during this phase of the study .This is because the admission criteria for SCBU are wide including all newborns with a one minute Apgar score less than or equal to three, those with a birthweight less than 2.0 kg and all sick infants. Subsequently only SCBU infants were screened for NE. Daily visits were routinely made to all postnatal wards to ensure all sick neonates were detected and admitted to SCBU. Neurological examination All neurologically abnormal term infants were formally examined following a standard examination protocol developed locally based on the principles of neurological evaluation described by Amiel-Tison^237). The conscious level of the infant was assessed with reference to the normal cycles of the newborn term infant's state of arousal as described by Prechtl and BeintemaC2). Suck was assessed clinically in those having difficulty feeding by the examination of the infant's sucking behaviour when presented with the examiner's clean little finger. Poor or absent suction or a poorly co­ ordinated clamping of the gums on the finger were recognised to be abnormal^238). Tone assessment particularly emphasised the active neck tone assessed via the raise to sit and back to lying manoeuvre, passive tone of the upper limbs assessed by the Scarf sign and passive tone of the lower limbs assessed by measurement of the popliteal angle. Primitive reflexes were assessed using the Moro and grasp reactions. Brain stem reflexes were examined using the comeal and gag reflexes. Respiratory pattern was noted and any abnormal episodic movements suggestive of seizures were described. Detailed instmctions for performing the neonatal neurological examination are attached as Appendix 1. The findings were recorded daily in a standardised manner on the infant's clinical data sheet which is attached as Appendix 2. Case exclusions Predetermined conditions for case exclusion were • preterm (less than 37 completed weeks since the first day of the last menstrual period as assessed by a Parkin score of five or less) • dysmorphic infants with at least one major congenital abnormality including early cranial ultrasound evidence of major cerebral malformation or cyst • infants whose neurological condition normalised when hypoglycaemia was corrected • infants with hepatosplenomegaly, petechial rash and cataracts indicative of intrauterine infection 75 • infants subsequently found to have pathogenic microbacteria isolated from blood or cerebrospinal fluid samples taken within the first 24 hours of life Reliability and validity testing Inter-rater reliability of the grading system was tested by comparing the clinical findings of five junior paediatricians in training on 27 neurologically abnormal infants. The examinations were performed contemporaneously by the assessors blind to the infant's history and each others' findings. Not all assessors were available to examine each baby either because they were not blind to the infant's history or because they were not on duty. Table 10 presents the results of inter-rater reliability assessment. Of the 27 infants the majority verdicts were no encephalopathy (grade 0) in 5 cases, grade 1 in 10 cases, grade 2 in 8 cases and grade 3 in 4 cases. In no case did assessments of a subject differ by more than one grade from each other. An unweighted kappa analysis of interrater agreement was performed allowing for more than two observers and more than two outcomes (239). The kappa statistic is a measure of agreement where one indicates perfect agreement and zero no agreement better than chance alone. According to standard interpretation 0.6-0.8 represents 'substantial agreement' and 0.8-1.0 'almost perfect agreement' (24°). The kappa value is presented for each grade separately, with only that for grade 1 NE falling slightly below 0.8. The kappa value for the scheme overall was 0.87.

Table 10. Inter-rater agreement for neonatal encephalopathy grade.

Neurological kappa test results: Subjects n=27 examinations n=70 grade 0: 0.89 2 raters n=14 grade 0=14 grade 1: 0.79 3 raters n=10 grade 1=27 grade 2: 0.85 4 raters n=3 grade 2=21 grade 3: 1.0 grade 3=8 combined kappa: 0.87

The final assigned grade (one - mild, two - moderate or three - severe) as with Fenichel's scheme corresponded to the most severe encephalopathic state of the infant observed during the period of daily neurological examination. To test the validity of this grading system it was compared with Levene's modification of the Fenichel grading system developed for casenote review^"). The records of all 131 encephalopathic infants recruited to the study were retrospectively reviewed by a clinician blinded to the original grading. Four cases had the clinical features of Fenichel grade one encephalopathy which persisted beyond 48 hours of age. Using Levene's method these 76 infants would be classed as Grade 2 but his use of duration as a grading criteria has been criticised elsewhere^108) so the original grading was retained. All other cases were found to have congruent encephalopathy grading. Control recruitment The cohort study 'contemporary' controls were initially matched for season of birth with the NE cases. Since it took upto 24 hours after delivery to confirm a case of NE it proved necessary to allow a 24 hour lag between the birth of a case and the birth of a control. Therefore the control infant was the next term infant (as assessed by Parkin's method) bom alive at least 24 hours after a case who did not meet the case criteria described above and did not require special care. There are three separate birth registers at Prasuti Griha for labour ward deliveries (85% of all deliveries), operating theatre deliveries (10%) and admission room deliveries (5%). I decided to match the location of recruitment of the controls with that of the general hospital population. Therefore using block randomisation I predetermined which two of every twenty controls were to be recruited from the operating theatre and which one from the admission room. After some eight months when our revised sample size calculations suggested the need to extend the recruitment period it was decided to improve the power of the study by doubling the controls recruited to two per case. The same overall control recruitment strategy with a 24 hour time lag was followed but now two consecutive normal term infants were recruited for each case. For comparative descriptive purposes the prevalence study presents contemporary control data as well as NE case and fresh stillbirth data. For this 12 month dataset the second control was omitted from the analysis. For the 'later' set of controls a simpler procedure was adopted. On arrival at the hospital women are issued with the next available in-patient serial number. Using this serial number every 25th infant bom in the hospital was assessed and provided he/she met the case criteria in all respects with the exception of not being encephalopathic the infant/mother pair were recruited as controls. If the infant failed to meet these criteria (usually because of being preterm, occasionally because of congenital abnormalities) the next infant, according to in-patient serial number, was recruited. This procedure was followed for a twelve month period generating an unmatched control group of 635 subjects, nearly five controls per case. Perinatal clinical data collection A perinatal data form (Appendix 3) was completed by a junior paediatrician for all cases of NE, fresh stillbirths with a birthweight of 2000g or more and all controls. The data included demographic details, past maternal history and relevant obstetric history based on uncorroborated maternal account. Place of residence was noted and coded as either inside or outside of the Kathmandu Valley ringroad. This provides a convenient and readily identifiable boundary between the relatively homogeneous urban population 77 of Kathmandu and Patan and the more heterogeneous periurban and semirural areas of the outer Valley. Generally the interview took place 12-24 hours after delivery although it was occasionally some days later if the mother had suffered a stormy postpartum period. Whilst no mothers of living infants declined interview 35 (28%) of the mothers of stillborn infants were unavailable for interview, generally because of early discharge. A few mothers declined to be interviewed because of distress following the stillbirth. Relevant features of the perinatal history (e.g. time of arrival at the hospital, maternal complications, fetal heart rate and clinical management data) were abstracted from the maternity hospital records. All control infants were clinically examined at this stage to exclude neurobehavioural abnormality. All mothers and infants were measured for standard anthropometric variables. Maternal height was measured to the nearest 0.1 cm using a Raven Minimetre. Maternal weight was measured to the nearest 0.5 kg using Healthy Life electronic adult weighing scales. Infant weight was measured to the nearest 10 g using Soehnle electronic infant weighing scales. The occipito-frontal head circumference was measured to the nearest 0.1 cm using a Child Growth Foundation Lasso tape. The infant's length was measured to the nearest 0.1 cm using a Raven Rollametre. Maternal anthropometric data is presented using cut-offs derived from the longitudinal studies of the National Nutrition Monitoring Bureau of India^241^. The height and weight cut-offs are those conventionally used for the definition of obstetric high risk. The body mass index (BMI) cut-off used here is considered a useful indicator of chronic energy deficiency in the S. Asian context^242). The study mothers were measured in the puerperium when their weight was still above normal and therefore the measures of weight and BMI are likely to be significantly higher than pre-pregnancy measures.

Methods for investigation Haematology, biochemistry and microbiology Mothers of fresh stillbirths, NE cases and all controls underwent venepuncture if they consented to the procedure. Less than ten women of the more than one thousand approached refused the procedure. Approximately 2 mis of mother's venous blood was taken following completion of the perinatal data sheet. Maternal haemoglobin was estimated at the bedside using the Hemocue photometric method*243^. Of the remaining blood 0.5 ml was placed in a plain tube and 0.75 ml in a lithium heparin tube. Both tubes were spun immediately in a portable centrifuge at 10,000 rpm for three minutes and the resulting supernatant was stored as maternal plasma and maternal serum respectively at -10 deg centigrade in an onsite fridge. Biochemical assay was not available on site necessitating prolonged on site storage prior to transport back to the UK. Unfortunately occasional failure of the hospital backup generator during power 78 cuts (which are scheduled daily in the winter months) meant temperature control was less than ideal. Refrigerated transportation of the samples to the UK proved challenging to arrange. Dry ice is not manufactured in Nepal. Liquid nitrogen howevere is available locally to support the national livestock insemination programme. Given the number of samples a liquid nitrogen dry shipper large enough to accomodate a significant proportion of the samples was prohibitively expensive. After attempting to hand carry samples piecemeal in a small liquid nitrogen dry shipper back to the UK we abandoned this transport system. In a complex logistical operation the vast majority of samples were transported in one large batch at the end of the study on dry ice obtained in Dhaka, . On arrival in the UK the samples were stored at -50 deg C in the Institute of Child Health, London, sorted and then transferred to the Institute of Neurology biochemistry laboratory where they underwent assay. The assays measured free thyroxine (FT4), thyroid stimulating hormone (TSH), serum Magnesium (Mg) and two different acute phase proteins, namely c-reactive protein (CRP) and alpha -1- acid glycoprotein (AGP). The interval from sampling to measurement was between 18 months and three years for the cases and contemporaneous controls and between 6 and 18 months for the later controls. FT4 and TSH assays were performed using an enzyme linked fluorescent assay technique (marketed as Vidas by bioM&ieux SA). Magnesium assays measured the reflection density of a magnesium-dye complex fonned when the serum sample was mixed with a dye reagent component of Vitros Mg slides (Johnson and Johnson). Monoclonal antibodies linked with horseradish peroxidase were used in a similar technique to bind with CRP using Vitros CRP slides (Johnson and Johnson). Hydrogen peroxide in the final wash was used to trigger the enzyme mediated oxidation of dye which was also measured using reflection density. AGP was assayed using an immunoturbidimetric technique marketed as Unimate 3 AAGP by Roche Diagnostics. For each assay contemporary controls were compared with later controls to investigate the question of differential sample degradation due to the difference in age of the samples at the time of biochemical analysis. At the time of diagnosis of encephalopathy routine unit guidelines expected the junior clinical staff to take a neonatal blood sample for culture in the onsite microbiology laboratory aswell as to check for hypoglycaemia using cotside blood glucose strip monitoring. In addition the study protocol called for medical staff to take an extra 2 mis of venous blood whilst performing venepuncture on case infants. A haemoglobin estimation at the cotside using the Hemocue photometric method was performed. On ethical grounds advocated by local staff control infants were not subjected to venepuncture. Moderately or severely encephalopathic infants underwent lumbar puncture provided their fontanelle was not bulging and they were considered well enough to tolerate the procedure. Cerebrospinal fluid (CSF) was examined microscopically for red and white blood cells, organisms and protein and glucose estimation was performed. Surplus CSF was centrifuged and stored at -10° C. Cranial ultrasonography Early cranial ultrasonography was included in the investigation protocol for all case infants to contribute structural evidence of both major pre-existing intracranial pathology at the time of delivery and also of evolving structural pathology in the first six weeks of life. Resource limitations prevented routine early neonatal ultrasound of the control infants. Both cases and contemporary controls were rescanned at six weeks of age. This had the added value of permitting an estimate of the predictive power of this rapidly spreading technology for the early detection of perinatally aquired disabling neuroimpairment at one year. At the outset of the study the literature on the association between abnormal ultrasound findings and hypoxic-ischaemic encephalopathy was reviewed. A uniformly echogenic appearance on early cranial ultrasound, variously termed "bright brain "(244\ or more recently "fuzzy brain" (245>, was described in this study as generalised echogenicity. Several studies have described more localised patterns of echogenicity in the early neonatal period^245,246,247>. Localised echogenicity can be subdivided according to the area most affected into cortical, subcortical, parenchymal, periventricular and thalamic/basal ganglia. Most investigators (following Siegel's observation that the posterior periventricular area may exhibit very variable echogenicity^248^) consider periventricular echogenicity to be abnormal only if it extends anteriorly adjacent to the external angles of the lateral ventricles. For the purposes of analysis I describe one category of localised echogenicity. These localised but not well demarcated echogenic patterns are distinguished from the sharply defined focal echogenicity found in the arterial territory of major vessels in cases of (i.e. anterior, middle or posterior cerebral artery). Echogenicity on ultrasound images is not quantifiable and therefore remains somewhat subjective. A study of interobserver variation has described a 'good' kappa measure of interobserver agreement of 0.73 for the observation of abnormal parenchymal echodensity in optimal circumstances^249*. Secondary cerebral atrophy was first described by Babcock and Balf250*. A recent study has demonstrated the predictive value of ventricular enlargement in the absence of cyst formation as a marker of white matter disease^181*. I recognised at the outset that ventriculomegaly was a common sequel to NE. The most widely used measure of ventricular size is misleadingly termed the ventricular index (VI) (251*. This is the ventricular width as measured from the midline to the lateral-most point of the lateral ventricle in coronal section just posterior to the foramen of Monro. There is no standardised data describing ventricular proportions in an undernourished low income 80 Asian population. Therefore I decided to classify the six week ventricular dimensions of the case population according to the normal data we collected from controls. I defined ventriculomegaly as a VI exceeding the 2nd standard deviation of the control group. Localised cyst formation in the white matter (cystic leukomalacia) has been intensively studied for the structural clues it offers to the underlying mechanisms of hypoxic - ischaemic brain injury/ 13,252,253,254\ A recent authoritative work on neonatal brain ultrasonography suggests that cystic leukomalacia developing over the first weeks of life is characteristic of perinatal brain injury in the term infant and strongly correlates with residual motor dysfunction^2451. Recently this influential Dutch research group have emphasised the need for a 10 megahertz probe to detect superficial cortical and subcortical cysts. Unfortunately based on best current practice at the time of study design a 7.5 megahertz probe was ordered which is likely to have reduced the study sensitivity for the detection of cortical cysts. Early neonatal cranial ultrasound examination was performed on cases within 48 hours of birth whenever possible. The machine used was a Hitachi EUB - 405 which was specially purchased and imported for the first time introduction of systematic cranial neonatal ultrasound scanning to Nepal. The real time examination was generally conducted by a junior paediatrician trained by the principal investigator to interpret dynamic images generated by a 7.5 megahertz probe angled to beam ultrasonic waves through the anterior fontanelle. The gain control (which determines the power of the ultrasonic beam and hence the general brilliance of the image) was held constant for all examinations. A standard text was used for training purposes <2551. This illustrated for the junior investigators the common ultrasonic patterns of pathology aswell as the correct anatomical orientation to measure the ventricular index. The examiner was asked to sweep antero-posteriorly in the sagittal plane and from left to right in the coronal plane in a systematic manner. The contemporaneous opinion of the examiner was recorded as well as hard copy of three standard views (coronal in the plane of the foramina of Munro and left and right parasagittal to demonstrate the periventricular areas). Multiple hard copy recording of abnormal or puzzling features were encouraged. Finally the examiner used the ultrasound scanner linear measuring capability to measure VI bilaterally. The hard copy views were reviewed by the principal investigator blind to the original interpretation of the dynamic record. Each set of neonatal prints were classified as demonstrating localised, generalised, focal or no echogenicity aswell as the presence or absence of cysts. Neonatal ventricular index data was not coded for analysis since we found the VI was insensitive at detecting 'slit' ventricles, described as part of the appearances of "bright brain". This was because the lateral margins of the ventricle remain visible and of normal position in the coronal plane despite severe compression in 81 the rostro-caudal axis. Slit ventricles are in any event described in normal term infants^251). Each set of postneonatal prints taken at six weeks were classified as demonstrating subcortical and/or periventricular cysts or no cysts. In addition the VI was calculated by averaging the measured left and right VI. This cancelled out any assymetry which may be a normal finding^255). Contemporaneous descriptions of the examination were compared with review of hard copy performed by the principal investigator blind to the original assessment. In the minority of cases where the interpretation of either early or late images differed between the original examiner and the reviewer the casting opinion of a second reviewer was sought.

Case management Standard management guidelines were drawn up complying with local circumstances and are attached as Appendix 4. Excluded cases A priori exclusions In the study design I neglected to include a register of infants with encephalopathy excluded from the study according to the apriori criteria drawn up at the study outset. Circumstances did not permit me to retrospectively return to the clinical records of all those infants excluded on these grounds. Approximately ten infants were excluded for major congenital abnormalities (almost all neural tube defects) and five infants for severe symptomatic hypoglycaemia with complete resolution of the encephalopathy o nce the blood sugar normalised.Two infants were excluded with clinical evidence of intrauterine infection (thrombocytopenic purpura and hepatosplenomegaly). Exclusions following investigation Following the approach recommended by the World Federation of Neurology the remaining cases of 'neonatal encephalopathy of early onset' were fully investigated and where definite causes other than birth asphyxia found were separately analysed. Early cranial sonography detected no major structural brain malformations amongst NE cases in our cohort. NE is a well recognised complication of perinatal sepsis^98). There is also evidence of an increased risk of infection due to immunocompromise following birth asphyxia^46). Therefore an excess of late neonatal infection is likely and best considered secondary to the primary encephalopathy. I took the apriori view that perinatal infection severe enough to cause encephalopathy would have laboratory evidence of infection of blood and/or CSF on the first day of life. I therefore analysed as a separate group those cases shown to have positive blood or CSF cultures taken within the first 24 hours after birth. Of 142 NE cases the results of early blood culture are available for 95 (67%). Lumbar punctures were performed on 35 (42%) infants of a possible total of 83 with grade 2/3 82 NE. In some severely encephalopathic cases clinical deteriation was too rapid to permit collection of all samples and in other cases the protocol for routine investigations was not adhered too by hard pressed junior staff. On a few occasions samples went astray in the laboratory. As noted in the management guidelines the unit policy forbade lumbar puncture where severely raised intracranial pressure was clinically obvious with a bulging fontanelle.Some infants were judged to be too unwell to tolerate the procedure on the first day or two of life and some of these underwent lumbar puncture later in their first week of life. In eleven cases early (first day) blood cultures were positive; in no cases were early (first day) CSF cultures positive. On review of the early CSF microscopy findings only one case had a CSF white cell count elevated above the normal range. At post mortem case review infection was considered as a contributory or terminal factor in four cases. The clinical details are described below. In the absence of positive early neonatal cultures these cases were not excluded from the study.

Study no. 40 was a 24 year old primip with a history of three who presented early in labour with prolonged rupture of membranes. The infant's heart rate was normally auscultated once but thick meconium stained liqor was evident. She proceeded to a vaginal delivery assisted by vacuum 4 hours after admission. The infant was bom in reasonable condition (Apgar1 4 and Apgar5 7) with a low birthweight (2.2 kg) and was admitted to SCBU. He was normoglycaemic and mildly anaemic (Hb 11.2 g/dl) with an echobright neonatal ultrasound. He developed grade one encephalopathy within 24 hours and received antibiotics although cultures were not taken. He initially appeared to be stable but on day 4 deteriated and died on day 6.

Study no. 49 was a 21 year old primip who went into spontaneous labour at home at 19:00 arriving at hospital the following morning at 07:00 in the second stage of labour. Labour was augmented with syntocinon, despite which she took over an hour to deliver and the fetal heart was noted to have a bradycardia (<120) at 08:10, fifteen minutes prior to delivery. At delivery the baby girl of birthweight 2.1 kg was in poor condition (Apgar1 0 and Apgar5 1). She required full resuscitation including intubation, ventilation and cardiac massage not breathing spontaneously for 15 minutes. At 3 hours of age she was normoglycaemic and had a haemoglobin of 11 g/dl. Blood cultures were taken and intravenous antibiotics administered. The cultures were found to be negative. Clinically the infant had grade 3 NE with severe apnoeas with an echobright neonatal ultrasound. By the second day there was some improvement but the baby developed a fever, respiratory distress and died the same day. At death there were copious blood stained secretions in the larynx and LP showed uniformly blood stained CSF.

Study no 296 was a 20 year old primip who presented some 10 hours after the onset of labour with eclampsic fits. Despite this the fetal heart rate was within normal limits when listened to once. She underwent emergency caesarian section an hour and 35 minutes after arrival and had meconium stained liquor. The infant, of birthweight 2.8 kg, was in poor condition (Apgar1 4 and Apgar5 6) and failed to breathe in the first minute of life. During resuscitation thick meconium was seen in the trachea. He showed evidence of transient hypoglycaemia on the first day of life which resolved with intravenous dextrose infusion. His initial blood culture was negative. He had a grade 1 encephalopathy for three days before deteriating on day 4 to develop a grade 2 encephalopathy dying on day 6. Blood culture taken on day 4 grew Klebsiella. Post mortem laryngoscopy showed massive pulmonary haemorrhage.

Study no 330 was a 23 year old multip who presented 30 hours after the onset of labor, had a normally auscultated fetal heart on one occasion and proceeded to deliver vaginally 1 hour and ten minutes after arrival. The baby boy of birthweight 3.3kg was bom in reasonable condition (Apgar1 5 and Apgar5 6) requiring suction only, breathed by one minute and was admitted to SCBU with evidence of respiratory distress. He was normoglycaemic and initial blood culture was negative. An LP was not performed and antibiotics were not administered. He showed evidence of grade one encephalopathy but deteriated on day three developing status epilepticus dying the same day. Immediate postmortem lumbar puncture yielded turbid CSF, with "plenty" of white cells, from which E coli was cultured. 84 Clinical characteristics of the group excluded for sepsis Table 11 describes the clinical characteristics of the eleven infants excluded postnatally for early sepsis.

Table 11. Clinical characteristics of 11 infants excluded with evidence of early neonatal encephalopathy secondary to perinatal sepsis.

Variable n % Maternal resident outside city ringroad 2 18 demography never attended school 8 72 Maternal low weight (<40 kg) 2 18 anthropometry short stature (<1.45 m) 3 27 low BMI (<18 weight/height2) 1 9

Antenatal risk premature rupture of membranes 0 0 factors for offensive liquor 0 0 infection

Intrapartum syntocinon augmentation 5 45 obstetric uterine rupture 1 11 factors cord prolapse 1 11 (not mutually cephalic presentation 11 100 exclusive) emergency operative delivery 4 36 instrumental delivery 3 27 Fetal distress fetal tachycardia 0 0 fetal bradycardia 2 18 thick meconium 4 36 thin meconium 1 9 Infant male 5 45 (<2.5kg) 1 9

Apgar1 8-10 0 0 4-7 1 9 0-3 10 91

Apgar5 7-10 1 9 5-6 3 27 0-4 7 63 Investigation E. Coli 8 72 Klebsiella 1 9 Staphylococcus Aureus 2 18

Follow up strategy and outcome measurement Follow up process The key outcomes of the study were assessed at one year. Subsidiary data points were at hospital discharge, six weeks of age and six months of age. Follow up of both cases 85 and controls was extremely challenging in this deprived metropolitan setting characterised by a high degree of population mobility with no formal address system and hence no home delivered postal service. Private telephone lines are in limited supply and most people use public telephones sited in local shops. At the outset of the study an open access outpatient clinic offered primary infant health care including an immunisation service to all comers. The follow up of graduates of SCBU was integral to this arrangement. I decided to hold a dedicated twice weekly follow up clinic for study infants. All surviving cases and controls enrolled for the cohort study were invited to attend at 6 weeks, 6 months and one year. They were issued with an identification card which acted as a pass for the security staff of the hospital since access to clinical areas is restricted. They were given an appointment date which was entered on the study computer database. At the time of discharge the families of cases and controls were offered a lift home in the project vehicle. Some preferred not to wait and once they had left the hospital there was no effective means of contacting them. Those who accepted a lift in the project vehicle were taken home and the house location was mapped by project staff. A photograph of the proud parents with their newborn child was taken posed outside the house to facilitate later tracing. Parents were informed that a copy would be available when they attended for follow up at 6 weeks. Non-attenders were flagged up on the project database. Those whom we had taken home were visited by project staff repeatedly until direct contact with the mother was re­ established. It is customary for a nursing mother to spend prolonged periods at her mother's house (the maitighar, a secondary home of continuing importance long after marriage has brought the woman to her husband's family home) which often lay far from the Kathmandu Valley. Therefore follow up reminders were sometimes necessary over an extended period of time. When necessary a lift was offered to the infant's mother to facilitate follow up attendance. There was a small but continuing attrition rate over the following months mainly due to changes of address amongst the more mobile first generation urban migrants. In one case the family was uncontactable having been imprisoned for alleged involvement in girl trafficking, a serious problem in Nepal as elsewhere in South Asia. Only one family refused to attend despite these efforts and turned down the offer of a home assessment. Six weeks progress assessment At the first follow up appointment a combination of closed and open questions elicited any history of feeding problems, intercurrent illnesses, fits and mother's opinion of her baby's development. The neurological examination was unblinded and focussed on the quality of posture and spontaneous movement. Both passive and active tone aswell as tendon and primitive reflexes were assessed and anthropometry (weight, length and head circumference) was measured using the same equipment as for neonatal measurement. Cranial ultrasound examination followed the protocol described above. 86 If there was evidence of acute intercurrent illness appropriate action was taken. This arose in 13 cases of whom half (six) were suffering from acute respiratory infection (ARI) and were treated with oral co-trimoxazole with the exception of one infant who had significant respiratory distress and required urgent hospital admission. The remainder either required oral rehydration fluids for diarrhoeal disease or antibiotics for superficial skin sepsis. Where there was evidence of evolving cerebral palsy an early three month follow up appointment was offered. Six months progress assessment At six months the history and examination were again conducted by an unblinded clinician following similar guidelines to that of the six month check. Once again mother's opinion of her baby's development was sought, including the infant's ability to see and hear. On examination particular attention was paid to the of primitive reflexes especially the Moro and assymetric tonic neck reflex (ATNR) and the development of normal and symmetrical movement patterns including voluntary grasp and rolling. Again anthropometry was performed. Infants thought to be neurodevelopmentally impaired at six months (or three months if re-attenders) were referred to a Kathmandu cerebral palsy self-help group identified by the principal investigator to be the best available child focussed therapeutic facility for neurodisability in the city. One year outcome assessment The one year outcome assessments were performed by two Nepali paediatricians based at another hospital who assessed the children blind to their previous history. They were given further clinical training in neurodevelopmental assessment in general and the Denver Developmental Screening Test (DDST) in particular on secondment to a specialist child development centre in Dhaka, Bangladesh who have been using the tool on South Asian infants and children since 1981 (Khan N. - personal communication). The DDST is the most widely used developmental screening tool in the world and has been translated into seven languages^256). The DDST was selected because there is experience with this screening tool in neighbouring centres, it is cheap and in the public domain with no restrictive controls over training procedures and is an appropriate tool in the hands of general paediatricians in this setting. The DDST takes approximately 15 minutes to perform and when combined with a clinical assessment of vision, a distraction hearing test, a formal neurological assessment and anthropometric measurement the average 1 year old has generally reached the limits of their tolerance. The DDST is based on the comparison of a child's skills at a particular age with the age- spread of equivalent skill acquistion in a reference population. Until 18 months of age it is very largely based on direct observation of skills in each of the four major developmental areas and therefore avoids the errors inherent in parent reported developmental tools. At any particular age there are approximately six items per developmental area making a total of approximately 24 items. For each item the child is compared with the age range between which 25% and 90% respectively of the reference population acquired that particular skill. In practice the tester examines all those skills clustering around the subject's chronological age as marked by a vertical line on the standard DDST assessment chart. The chart includes boxes depicting centile data of the reference population. The tester is asked to interpret his findings following the guidelines below. The hard copy of the assessment chart allows checking of the interpretation of the screening test at a later date. The DDST distinguishes two different grades of screening failure according to the number of skill items that are delayed, the spread of those items across the different developmental areas and the extent of the delay.

Normal - the child fails none of the items which were passed by more than 90% of children of his age (i.e. all those items whose centile boxes fall entirely to the left of his chronological age line). Questionable - the child fails one of the test items (in any of the four sectors) passed by 90% of his age group (i.e. fails one item whose centile box falls entirely to the left of his age line) or fails to pass at least one item passed by 25% of his age group in all four sectors (i.e. fails one item transected by his age line in each sector) Delayed - the child fails at least two test items passed by 90% of his age group in any sector (i.e. two items whose boxes fall entirely to the left of his chronological age line).

At the outset of the assessment session the examiners once again asked about serious post neonatal illness, head injury and/or fits and sought the opinion of the parents regarding the child's neurodevelopment. On examination first the hearing was assessed clinically using a distraction hearing test ^257^. The conditions were not optimal (Asian cities are noisy and the poor building specifications of public buildings make sound proofing impossible) and therefore the assessors were trained to aim for a sound threshold of 40 decibels (dB) - sufficient to detect a disabling sensori-neural hearing impairment but not to detect the mild conductive hearing loss associated with secretory otitis media. The distractor (the examiner) and the tester (a trained project nurse) were asked to repeat the procedure for both ears until they both agreed there was a definite response or lack of response to both high and low frequency sound presented at 40 dB to either ear. An "uncertain" category was permitted if the assessors remained unsure of the infant's response. On examination of the eyes the assessors were first asked to record the presence or absence of either a squint or nystagmus. For vision testing a modification of the Stycar rolling ball was used with the subject sat on a parent's knees. Small (1 centimetre diameter) red and black balls were rolled silently across a white muslin sheet on a low surface 1 metre in front of the child. Visual tracking, pointing and reaching behaviour were clinically assessed. This form of test has been reported to have reasonable sensitivity (70%) for the detection of severe visual defects but is poor at detecting the mildly/moderately visually impaired^258). The DDST was then administered. Finally a formal neurological examination was performed following the examination procedures described by Amiel-Tison (237^ and the descriptive recording of central motor deficit advocated by Stewart (259\ The screening clinicians were asked to summate their neurological findings in three categories.

Normal no impairment detected Minor neurological impairment defined as reproducible neurological abnormalities of the motor system or sensory apparatus with no associated functional disability (e.g. abberrant tone or reflexes not interfering with movement orposture, a squint) Major neurological impairment with associated functional disability

All infants with abnormal neurological findings on examination or with questionable or serious delay on developmental screening were referred for a more detailed unblinded neurodevelopmental assessment involving administration of the Bayley 2 assessment tool and repeat neurological examination. This was held at a local cerebral palsy self help group in order to facilitate linkage with a local therapeutic community. Generally families were taken personally by the principal investigator to this voluntary sector centre. Ether the principle investigator or his principal clinical counterpart conducted the repeat assessment which for practical reasons did not permit more detailed evaluation of the hearing and vision than had already been performed. However it did permit a clinically important "second look". The Bayley 2 has undergone extensive validation on a large cohort of American children and is probably the most widely used developmental testing tool in the world today(26°). It generates two raw scores which when adjusted for chronological age are transformed into a mental development index (MDI) and psychomotor development index (PDI). Both indices are calibrated such that the 1st standard deviation (SD) of the scores of a normal American reference population score were 84 and 116 respectively and the 2nd standard deviation were 55 and 145 around a median of 100. Neither scale attempts to assign an index score below 50. Finally at reassessment the type of motor impairment was coded. For these purposes a widely accepted Swedish classification system for cerebral palsy was used^57^. The classification system, with four broad categories and a total of 8 possible subtypes is presented in Table 12. 89

Table 12. Classification of cerebral palsy (after Hagbergf57))

Category Subtype Spastic CP right hemiplegia left hemiplegia spastic diplegia spastic tetraplegia Ataxic CP ataxia

Dyskinetic CP mainly athetoid mainly dystonic

Mixed CP spastic/dyskinetic/ataxic

Interobserver reliability testing of this classification has been reported and was disappointing with a kappa value of only 03 in a study of 19 cases with 6 observers^261). To limit this interobserver error we limited the assignment of CP type to two people only (the author and principal clinical counterpart). Unfortunately service committments precluded us from measuring our inter rater agreement for this aspect of the study but at the outset we did review several cases together to standardise our clinical judgement criteria. Outcome definitions Where screening findings disagreed with fuller reassessment primacy was given to the latter. Following screening (and reassessment where applicable) the principal investigator made a final coding of one year outcomes following an a priori hierarchy of definitions.

Death Major neurodevelopmental impairment: neurodevelopmental or sensory impairments) resulting in functional disability and/or a mental developmental index less than 55 or a psychomotor developmental index less than 55 Minor neurodevelopmental impairment: neurodevelopmental or sensory impairments) not resulting in functional disability and /or a mental developmental index of 55-83 or a psychomotor developmental index of 55-83 Normal Lost to follow up

Verbal The families of infants found (on home visiting for non-attendance) to have died during the first year of life were visited and a verbal autopsy performed using the method described by Greenwood and colleagues^262) and the questionaire they have 90 developed^263\ This was always conducted by a medically trained paediatric specialist, either the principal investigator or my principal counterpart. Analysis

Data entry All clinical data was entered from paper hardcopy contemporaneously by project staff onto a Power Macintosh computer situated in the project office in the maternity hospital. The database was designed by the principal investigator in Filemaker Pro 2.0. Project staff were trained to use this patient database to administrate the follow up and recall system. The national calendar system of the Kingdom of Nepal is based on the lunar cycle, runs from Spring to Spring, and is published after final adjustments by court astrologers towards the end of the preceding year. All dates were transformed into the western calendar prior to data entry. Routine hospital data are recorded according to the national calendar and collated by lunar month. Since routinely collected demographic information describing total deliveries provided essential denominator data for the prevalence studies described here it was important that the study period coincided with national calendrical months. Therefore, although for ease of presentation I report the 12 month prevalence from January 1995 to January 1996, the actual period reported is from Magh 1 2051 (January 15 1995) to Push 30 2052 (January 14 1996). Data management All data was doublechecked by the principal investigator comparing hardcopy with electronic data fields prior to analysis. The Filemaker database containing all project data was then reviewed and project management data fields irrelevant to the analysis were cut. This allowed the reduction of 291 data fields per subject to 182 data fields for export to a data analysis software package. These data were transformed into a Statview 4.1 spreadsheet format on a Macintosh Performa computer in the UK. Data cleaning Data cleaning was performed mainly using range checking procedures. Outliers which were due to obvious data input error (eg date of birth out by a year, numerical data a factor of ten out of range) were corrected. The most troublesome area was the date data which had suffered from occasional miscalculation on conversion from the Nepali calendar into the Western calendar. Fortunately there were three dates (for commencement of labour, admission to hospital and delivery) for each individual which permitted cross checking and generally resolved the problem. All growth data was converted into standard deviation scores using the British 1990 growth reference data (264^ and conversion software available from the Child Growth Foundation. 91 At this stage five controls were excluded because of incomplete data (two for lack of maternal data and three for lack of infant data). Three stillbirths were excluded as not meeting the original case definition due to major malformations (two with anencephaly and one with hydrocephaly). Data coding Some open data fields and multiple subfields for certain variables had been permitted in the original data sheet for perinatal clinical data. These were recoded into conventional clinical categories which were generally self evident. Intrapartum complications were coded to be present where there was clinical evidence of obstructed labour, cord prolapse, eclampsia, placental abruption or uterine rupture. In the absence of any objective measure of birth asphyxia, following the Perth Groups 120\ I have defined intrapartum hypoxia to be evidence of fetal tachycardia (>160 bpm) or bradycardia (<120 bpm) on fetal auscultation or the presence of thick meconium along with an Apgar at one minute of less than 3 or an Apgar at five minutes of less than 6. Unlike the Perth group we did not have the benefit of cardiotocographs nor detailed clinical notes and therefore have also analysed the data in terms of definite or possible intrapartum suboptimality. These terms were defined so as to reflect obstetric clinical judgement and decision making as depicted in Table 13.

Table 13. Criteria used in the definition of definite and possible intrapartum suboptimality.

Definite intrapartum suboptimality Possible intrapartum suboptimaity emergency caesarian section instrumental delivery obstetric catastrophe (uterine rupture, fetal distress cord prolapse, abruption, eclampsia, (thick meconium/ hangi ng breech) fetal bradycardia < 120bpm/ fetal tachycardia >160bpm) clinically obstructed labor prolonged labor (prolonged 2nd stage [>1 hour] and /or (>24 hours primip, >12 hours multip) obvious problems with prolapsed limbs, or hanging breech) cord tightly round neck at delivery

The cleaned and coded dataset (50 variables per subject) was transformed into a Stata 5.0 data file to enable more complex epidemiological analysis including multivariate logistic regression of the case control data. 92 Analytical methodology Prevalence data Statistical analysis of the prevalence study was based on standard arithmetical calculation of proportions which are mainly quoted in terms of prevalence at birth of x per 1000 live or total births according to context. Confidence intervals for proportions were calculated using the exact binomial distribution. In standard texts birth prevalence data is often referred to as a rate. Although technically incorrect I have followed conventional usage in terms such as the 'perinatal mortality rate'. Risk factor analysis At the outset of the analysis of the case control study descriptive data of various exposures of interest were compared across all four population groups (NE cases, SB cases, contemporary controls and later controls). Clinical aspects of the two control groups were then compared using standard univariate tests of difference (two tailed Chi square) with conventional cutoffs for statistical significance (p < 0.05)(265). The suspected bias amongst contemporary controls was confirmed. Therefore risk factor analysis proceded to compare NE cases and later controls. Anthropometric variables were tested for normality of distribution (see Appendix 5). No important deviations from normality were detected. This and other continuous data were summarised using means with standard deviations and ranges according to context. Comparison of means was performed using Student's t test. Test for trend across ordered groups followed the method described by Cuzick^266^. Univariate analysis proceded to calculate odds ratios, confidence limits, Chi squared and p values for each exposure in a series of comparisons of study subgroups. For nominal stratified data a Mantel-Haenszel test procedure was used^267\ For continous data a score test for trend of the odds^268^was employed to derive odds ratios for an increase of a single unit and the strength of the association assessed using p values. A p value of 0.2 or less was adopted as an a priori cutoff for variables to be taken forward to multivariate analysis using logistic regression. Only those significant exposures which were considered to be primary risk factors in the causal chain were taken forward to the regression analysis. In a series of stepwise backward regressions exposures with non-significant odds ratios were sequentially removed from the model until only those exposures with odd ratios significantly departing from unity remained. The final logistic regression is presented using odds ratios as an estimate of the size of the effect and 95% confidence intervals to express the precision of that estimate as advocated in a recent editorial by Rothman^269^. The resulting model of risk factors for NE was rerun in two subgroup analyses to investigate the extent to which unplanned and out of town deliverers may have biased the population sample delivering at the maternity hospital. 93 The distribution of the laboratory data was first investigated for normality. All measures approximated to a normal distribution (see Appendix 5) with the exception of C reactive protein (CRP). CRP values above 120 mg/1 are not precisely quantified by the laboratory assay. These have been arbitrarily coded as 150 mg/1 for the purposes of analysis. Summary statistics of laboratory variables were compared in unpaired groups according to outcome using a two sample Student’s t test (265\ To test the null hypothesis that there was no association between seasonality and adverse pregnancy outcome a Kolmogorov-Smimov type statistic was employed. This is a non-parametric test which compares bivariate categorical distributions and tests the null hypothesis of no difference and has been previously described for the assessment of seasonal variation^270). Outcome data Various clinical variables associated with the outcome study groups were initially compared in a series of frequency tables using the Chi squared test The relative risk of adverse outcome at one year was then calculated comparing proportions in unmatched groups. Several variables are compared for their predictive power amongst the NE cases as tests for the primary outcome of major neurodevelopmental impairment at one year. The reviews of these tests follow the conventions described by Sackett (271^ and are presented more fully in Appendix 6. Ethical approval, informed consent and competing interests Ethical approval for this study was given by the Nepal Medical Research Council, the ethical committee of the Institute of Child Health, London and the Maternal and Infant Research Group (MIRA), Kathmandu governing body. The mothers of all cases were given a full explanation of the diagnosis and offered further follow up appointments to monitor neurodevelopmental outcome. It was explained that this was part of a study which we hoped would shed more light on why some babies were bom in poor condition. The mothers of both cases and controls were asked to undergo a single venepuncture. The investigations performed on the case infants constitute part of routine case management in high income countries and therefore explicit consent for each test was not sought. It was felt to be unethical to take blood from unaffected control infants. No competing interests are declared by the principal investigator. 94

4. Prevalence of fresh stillbirth and neonatal encephalopathy What is the prevalence of fresh stillbirth and NE (after overtly non asphyxial causes have been excluded) amongst term infants in this population? Perinatal vital statistics Table 14 presents the perinatal vital statistics for Prasuti Griha over a 12 month period from January 1995 to January 1996. Only infants actually bom within the hospital with a birthweight of 500g or more are included. Infants bom en route to the hospital ("taxi deliveries") are excluded from these data.

Table 14. Perinatal vital statistics for Prasuti Griha in 1995.

Total Births Livebirths Stillbirths Neonatal deaths! 14,771 14371 400 254 Note: t prior to hospital discharge. Stillbirth rates The total stillbirth rate was 27.1/1000 total births (95% Confidence Interval 24.5-29.8). Table 15 describes stillbirths of birthweight 500 g or more by birth weight, estimated gestation and clinical condition.

Table 15. Stillbirths analysed by birth weight, estimated gestation and clinical condition from 14,771 deliveries at Prasuti Griha hospital in 1995.

BirthWeight Stillbirths with Macerated Fresh Condition Total (%) (8) congenital Stillbirths Stillbirths unrecorded malformation "Preterm" 500-999 5 19 3 0 27(7) 1000-1499 9 23 15 4 51(13) 1500-1999 8 34 20 3 65(16) Subtotal (%) 22 (15) 76(53) 38 (27) 7(5) 143 "Term" 2000-2499 5 47 50 1 103 (26) >/= 2500 14 53 71 4 142 (36) Subtotal (%) 19(8) 100(41) 121 (49) 5(2) 245

Missing data 4 3 2 3 12(3) Total (%) 45(11) 179 (45) 161 (40) 15(4) 400(100%)

For 12 stillbirths the birthweight was not recorded. For 15 cases no details of the fetal clinical condition could be traced. For calculation of stillbirth rates the missing data was recoded assuming the same distribution by weight and condition as the full data set. As 95 an estimate of fresh term stillbirths I included all those with a weight of 2000 g or more. 121 stillbirths met these criteria; to these five were added as a proportionate correction factor for missing data. This gives a prevalence of fresh term stillbirths of 8.5 per 1000 total births (95% Cl 7.1-10.1). During the study period there were 14,086 total births with a birthweight of 2.0 kg or more, a fresh term stillbirth rate estimate of 8.9 per 1000 term births (95% Cl 7.5-10.6).

Neonatal encephalopathy prevalence During 1995,92 cases of NE met our case definition, a birth prevalence of 6.4 per 1000 live births (95% Cl 5.2-7.8). A study of low birthweight infants in the same hospital during the period of the study found 93% of all deliveries to be of term gestation on Parkin clinical assessment (unpublished data). Based on this I estimate a denominator value of live term births during the study period of 13,365 (93% of 14,371) giving a birth prevalence for NE of 6.9 per 1000 live term births (95% Cl 5.6- 8.4). 98 term control infants not requiring special care were recruited during this period for comparative purposes. Neonatal clinical characteristics Table 16 describes the condition of the infants at delivery.

Table 16. Birth details of 92 term infants with neonatal encephalopathy, 86 fresh stillbirths of a weight of 2000g or more and 98 healthy term infants identified from 14,771 live deliveries at Prasuti Griha in 1995.

Variable Category NE cases n (%) Stintmthsn(%) Controls n all n = 92 all n = 86 all n =98

Gender male 53(58) 56(65) 48(50) female 39 (42) 30 (35) 48(50)

Multiple birth one of twins 4(4) 3(3) 0 (0) singleton 88(96) 83(97) 98(100)

Birth weight 1500-1999 1 (1) 0 (0) 0 (0) (g) 2000-2499 26 (29) 33(37) 25(26) 2500-2999 39(43) 31(35) 55(56) 3000-3499 20 (22) 19(21) 15(15) 3500-3999 4(4) 3(3) 3(3)

Apgar score at 7 0 (0) 1 (1) Apgar score at <5 65 (34) 0 (0) 5 minutes 6-7 24(26) 1 (1) >7 3(3) 97(99) 96 Low Apgar scores as a predictor of neonatal encephalopathy During the study period a total of 735 infants with a birthweight of 500 g or more had an Apgar score at 1 minute (Apgar1) of three or less. Of these 548 (75%) had a birthweight of 2000 g or more. Table 17 presents a 2 by 2 table describing the qualities of the one minute Apgar score as a test for neonatal encephalopathy.

Table 17. Review of the use of low one minute Apgar score (less than or equal to three) as a test for neonatal encephalopathy.

NE No NE totals Apgar1 3 7 13629 13636 Total 92 14279 14371 Sensitivity 92.4% Specificity 95.4% LR+ 20.09 LR- 0.08 Positive predictive 11.6% value Negative predictive 99.9% value prevalence 0.64% Pre-test odds 0.0064 Post-test odds 0.129 Post-test probability 0.114 Note: LR = likelihood ratio Adverse intrapartum events and birth outcome Table 18 codes both case and control populations for clinical markers of intrapartum adversity (see chapter 3 for the definition of intrapartum sub optimality and intrapartum hypoxia). A Chi square analysis compares each case group with the control group. There is a highly significant excess of both obstetric and fetal markers of intrapartum compromise in the cases compared with the controls. In this setting many women present late in labour (40% of the women reported here delivered within four hours of arrival) and fetal monitoring is infrequent (in 25% of the deliveries reported here the fetal heart was auscultated only once). These measures may therefore underestimate the extent of intrapartum problems contributing both to encephalopathy but more especially to fresh stillbirth. Overall there was evidence of intrapartum compromise in 63 (68%) of the cases of NE and 65 (76%) of the stillbirths but only in 12 (12 %) of the controls. 97

Table 18. An analysis of the association between adverse pregnancy outcome and intrapartum adversity using two different measures.

Variable Category NE cases n (%) Stillbirths n (%) Controls (n = 92) (n = 86) (n =98) Chi2 v controls Chi2 v controls Intrapartum definite 27 (29) 50(58) 7(7) suboptimality possible 36 (39) 15(17) 5(5) (see text for none 29 (32) 21 (24) 8 6 (88) definition) 63.3 p<0.001 76.5 p<0.001

Intrapartum present 31(34) 61 (71) 0 (0) hypoxia absent 61(66) 25 (29) 98(100) (see text for 39.5 p<0.001 104 p<0.001 98(100) definition)

Grading and early neonatal outcome of NE Table 19 describes the clinical grading of encephalopathy and early outcome prior to hospital discharge. There were 27 survivors of moderate NE and 1 survivor of severe NE giving the prevalence of survivors of moderate or severe NE at discharge of 1.9 per 1000 livebirths (95% Cl 1.3-2.8).

Table 19. Severity of neonatal encephalopathy with associated early neonatal mortality risk.

NE Grade Cases Early deaths Early mortality risk n (column %) n (column %) (95% GI.)

Mild (1) 37(40) 3(10) 0.08(0.02-0.22) Moderate (2) 34(37) 7(23) 0.21 (0.09-0.38) Severe (3)______21 (23)______20 (67)______0.95 (0.76-1.0) Total 92(100) 30(100)

Perinatal mortality and birth asphyxia The overall perinatal mortality rate (prior to hospital discharge) was 44.3 per 1000 total births (95% Cl 41.0-47.7). This rate does not include deaths amongst well infants in the first week of life after hospital discharge. Three infants with a birthweight of 2000 g or more, with no evidence of major congenital abnormality, died during resuscitation efforts. Although neither fresh stillbirths nor NE cases as defined, they were included in a cause specific perinatal death estimate as, most likely, death resulted from intrapartum asphyxia. Combining the fresh stillbirths of 2000 g or more (126), the first week deaths from NE in term infants (30) and cases of failed resuscitation in normal infants of 2000 g or more (3) provides an estimate of birth asphyxia cause-specific perinatal mortality in term infants prior to hospital discharge of 10.8 per 1000 total births (95% Cl 9.2-12.6). This constitutes 24% (95% Cl 21-28%) of the overall perinatal mortality rate (PNMR). Over an almost contemporaneous twelve month period (2052 Nepali calendar, April 14 1995-April 13 1996) clinical audit of neonatal mortality recorded 41 early neonatal deaths prior to hospital discharge of infants with a birthweight of 2500 g or more. Seven were ascribed to congenital anomalies, two to pulmonary haemorrhage of unknown cause, one to severe hypothermia, four to perinatal sepsis and 25 to birth asphyxia (Manandhar D.- personal communication). In this normal birthweight group neonatal mortality ascribed to birth asphyxia constituted 61% of the early mortality affecting mature infants. Key findings The birth prevalence of fresh stillbirth and neonatal encephalopathy amongst term newborns in a low income country urban hospital (after the exclusion of all cases due to recognised non-asphyxial causes) was 8.5 per 1000 total births and 6.4 per 1000 livebirths respectively. There was evidence of intrapartum compromise in 63 (68%) of the cases of NE and 65 (76%) of the stillbirths but only in 12 (12 %) of the controls. 60% of the cases of NE were either moderate or severe and the early neonatal mortality resulting from NE was high with a 33% mortality risk in this setting. The overall estimate of birth asphyxia cause-specific perinatal mortality in term infants prior to hospital discharge in this hospital delivering population was 10.8 per 1000 total births (95% Cl 9.2-12.6). This constitutes 24% (95% Cl 21-28%) of the overall perinatal mortality rate. 99

5. Risk factors for neonatal encephalopathy and fresh stillbirth in term infants Chapter four shows that fresh stillbirth (SB) and neonatal encephalopathy (NE) in term infants together explain 25% of perinatal mortality in this population. What are the risk factors for these adverse outcomes? Are scarce resources best directed towards antenatal care, perinatal obstetric or neonatal interventions? This chapter presents a case control study comparing fresh term stillbirths and encephalopathic infants with non- encephalopathic term infants.

Comparison of control groups As explained in chapter three the unplanned bias in "contemporary" control recruitment led to the further recruitment of a "later" control group. The later control group of 635 term infants included 34 (5% of the sample) who required special care. In 19 infants (3% of the sample) the indication for admission was an Apgar1 of three or less at birth. On univariate comparative analysis of the two control groups using the Mantel- Haenszel procedure to compare odds ratios (OR) the later controls were less likely than contemporary controls to have attended for antenatal care (OR 0.8 [95% Cl 0.6-0.9]), were less likely to have planned to deliver at the maternity hospital (OR 0.5 [Cl 0.3- 0.9]) and were more likely to have a non-cephalic presentation (OR 4.2 [Cl 1.0-17.3]). These differences are significant and in the direction predicted and I interpret them as evidence of biased sampling towards a lower risk population amongst the contemporary control group. The alternative explanation of a secular downward trend in the uptake of antenatal and perinatal services as well as a changing pattern of fetal presentation is less likely. In respect of the other clinical risk factors presented below the differences between the groups did not meet conventional tests of statistical tolerance (p value < 0.05). There is no evidence of a significant secular upward trend in terms of improving maternal condition. For the purposes of this risk factor analysis the later control data has been used. Clinical risk factors The case control study groups comprised 131 NE cases, 107 SB cases and 635 later controls. Measures of maternal and infant anthropometry in the immediate postnatal period for all the study subgroups combined are normally distributed. Histograms describing the distribution for each variable are attached in Appendix 5. Summary descriptive indices of this normal data, coding categories for categorical risk factors of interest and coded data is presented in Table 20 by case and later control groupings. Equivalent data for the 208 contemporary controls are also included for comparative purposes. 100 Table 20. Distribution of risk factors of interest amongst the four study groups* Data with non Gaussian distributions are presented as categorical variables whilst normally distributed data are presented by mean values (with standard deviation). Exposure of Coding NE cases SB cases Contemp. Later interest categories if controls controls categorical n (%) or n (%) or n (%) or n (%) or data mean (SD) mean (SD) mean (SD) mean (SD) Maternal socioeconomic/demographic status maternal >5 years 29 (26) 12(13) 69 (36) 229 (41) education 1-5 years 20(18) 8(9) 29(15) 82(15) none 60(55) 71(78) 96(49) 254(45) total 109(100) 91(100) 194(100) 565(100) residence inner city 76(58) 53(50) 161(77) 521(82) periurban 55 (42) 53 (50) 47(23) 114(18) total 131 (100) 106(100) 208 (100) 635(100) Maternal anthropometry maternal 48.1 (7.1) 48.6 (6.6) 49.7 (7.3) 50.5 (7.2) weight (kg) total 131 102 203 627 maternal 148.0 (6.1) 149.6 (6.4) 150.4 (6.1) 150.9 (5.8; height (cm) total 131 102 205 625 maternal <18 11(8) 10(10) 24(12) 96(15) body mass 18-25 112(86) 85(83) 166(82) 484(78) index >25 8 (6) 7(7) 13(6) 43(7) (MW/MH2) total 131(100) 102(100) 203 (100) 623 (100) Obstetric/perinatal history parity multiparity 51 (40) 65 (61) 100 (49) 270 (43) nulliparity 77(60) 42 (39) 104(51) 364(57) total 128(100) 107 (100) 204(100) 634(100) inter-birth <24 months 7(15) 15(23) 18(19) 45(17) interval >24 months 40(85) 50(77) 78 (81) 223(83) (multips. only) total 47(100) 65(100) 96(100) 268(100) previous none 123 (97) 96 (91) 196 (98) 614 (97) stillbirth any 4(3) 10(9) 4(2) 19(3) total 127(100) 106(100) 200(100) 633 (100) previous none 114(93) 84(91) 187(96) 620 (98) neonatal any 8(9) 8(9) 7(4) 14(2) death total 122(100) 92(100) 194(100) 634(100) Features of this pregnancy place of PG Hospital 105(81) 67(66) 184(89) 594(94) intended elsewhere 25(19) 35 (34) 23(11) 37(6) delivery total 130(100) 102(100) 207(100) 631 (100)

Notes: MW = maternal weight, MH = maternal height 101 Table 20 (cont.) Distribution of risk factors of interest amongst the four study groups .

Exposure of Coding NE cases SB cases Contemp. Later interest categories if controls controls categorical n (%) or n (%) or n (%) or n (%) or ______mean (SD) mean (SD) mean (SD) mean (SD) antenatal 3 or more 75 (59) 45 (45) 126 (62) 444(71) care 1/2 visits 16(13) 20 (20) 39(15) 81(13) attendances never 35(28) 35 (35) 39(19) 100(16) total 126(100) 100(10) 204(100) 625 (100)

maternal reported 126 (98) 99(96) 204(99) 619 (98) anaemia not reported 3(2) 4(4) 2 (1) 13 (2) (history of) total 129(100) 103 (100) 206(100) 632 (100)

maternal PE reported 115(89) 92(89) 198 (96) 598 (95) (history of) not reported 14(11) 11(11) 8(4) 34(5) total 129(100) 103 (100) 206(100) 632(100)

twin singleton 125 (97) 102 (95) 206(99) 631(99) twin 5(4) 5(5) 1 (1) 2 (1) total 131 (100) 107(100) 207(100) 633 (100) risk factors absent 110(84) 87(81) 201 (97) 608(96) for infection present 21 (16) 20(19) 7(3) 27(4) (PROM/sm- total 131 (100) 107(100) 208 (100) 635(100) elly liquor) interval from <1 19(15) 16(15) 33 (16) 99(16) admission to 1-4 34 (26) 43(40) 55(27) 153 (25) delivery 5-12 40(31) 22 (21) 68(33) 208 (33) (hours) 12-24 22(17) 18(17) 28(14) 91 (15) >24 15(12) 8 (8) 2 2 (11) 71(11) total 130(100) 107 (100) 206(100) 622 (100) Intrapartum factors syntocinon none 68(52) 89 (85) 150(73) 452 (73) induction 12(9) 6 (6) 14(7) 29(5) augmentation 50(38) 10(9) 42 (20) 139 (22) total 130(100) 105(100) 206(100) 620 (100)

presentation cephalic 119(91) 74(70) 206(99) 611(96) other 12(9) 31 (30) 2 (1) 24(4) total 131(100) 105(100) 208(100) 635(100) mode of vaginal 85 (65) 58(55) 175(85) 553 (88) delivery instrumental 22(17) 5(5) 9(4) 20(3) operative 23(18) 43 (41) 2 2 (11) 58(9) total 130(100) 106(100) 206(100) 631 (100) intrapartum absent 92 (70) 46(43) 195 (94) 608 (96) complication present 39 (30) 61(57) 13 (6) 26(4) total 131(100) 107(100) 208(100) 634(100)

Notes: PE = pre- eclampsia, PROM = premature rupture of membranes 102 Table 20 (cont.) Distribution of risk factors of interest amongst the four study groups.

Exposure of Coding NE cases SB cases Contemp. Later interest categories if controls controls categorical n (%) or n (%) or n (%) or n (%) or mean (SD) mean (SD) mean (SD) mean (SD) obstetric team A 20(15) 18(17) 35(17) 93 (15) team teamB 20(15) 16(15) 26(13) 116(19) team C 23(18) 19(18) 31(15) 124(20) team D 22(17) 22 (21) 37(18) 87(14) team E 25(19) 18(17) 45 (22) 114(18) teamF 20(15) 13 (12) 34(16) 90(14) total 130(100) 106 (100) 208 (100) 624(100) fetal distress absent 65(50) 75 (70) 193 (93) 592 (93) present 66(50) 32 (30) 15(7) 43(7) total 131(100) 107(100) 208(100) 635(100) resuscitation none/0^ 47 (36) 204(98) 608(99) bag/tube/cpr 82(64) 4(2) 4(1) total 129(100) 208(100) 612(100) Fetal/infant characteristics infant male 80 (61) 66(62) 111(54) 336(53) gender female 51 (39) 41 (38) 93(46) 298 (47) total 131 (100) 107(100) 204(100) 634(100) absent 106 (82) 203 (99) 621(99) present 24(18) KD 4(1) total 130(100) 204(100) 625 (100) birthweight 2.75 (0.4) 2.64 (0.4) 2.75 (0.4) 2.80 (0.4) (kg) total 129 107 207 634 head 34.2(1.6) 33.4(1.3) 33.3 (1.3) drcumfercnoe total 131 208 633 (cm) Notes: (£ = oxygen via face mask, bag = oxygen via bag and mask, tube = oxygen via endotracheal tube, cpr = cardio-pulmonary resuscitation

Risk factors for neonatal encephalopathy The results of univariate analysis of the odds ratios for various exposures comparing 131 NE cases and 635 controls are presented in Table 21. In most cases these are binary categorical variables, but where the variable takes multiple values the odds ratios are calculated for a one unit/level change in the variable. A double asterisk (**) indicates those variables actually taken forward to the multivariate analysis. A single asterisk (*) denotes exposures reaching the a priori test of statistical significance (a p value of 0.2 or less) which were not taken forward to the multivariate analysis. 103 Table 21. Univariate analysis of risk factors for neonatal encephalopathy.

Risk Factor OR 95% Cl chi 2 Maternal socioeconomic status maternal education 5 years or less 1.9 1.1-3.0 7.5 0.006** Maternal anthropometry maternal weight < 40 kg 2.4 1.2-5.0 6.2 0.013* maternal height < 145 cm 3.0 1.9-4.8 26.1 <0.001** low maternal BMI 0.97 0.9-1.04 0.7 0.39 Obstetric/perinatal history nulliparity 1.1 0.8-1.7 0.3 0.567 previous stillbirth 1.1 0.4-3.1 0.0 0.93 previous neonatal death 3.1 1.3-7.6 6.8 0.009** Features of this pregnancy no antenatal care 2.0 1.3-3.2 9.9 0.002** maternal anaemia, history of 1.1 0.3-4.0 0.04 0.846 maternal pre eclampsia, history of 2.14 1.1-4.1 5.4 0.02** twin 12.5 2.4-66.5 14.7 <0.001* maternal risk of sepsis (PROM/smelly liquor) 4.3 2.3-8.0 25.6 <0.001** interval from admission to delivery 1.02 0.8-1.1 0.09 0.758 Intrapartum factors use of syntocinon 1.6 1.3-2.0 19 <0.001** non-cephalic presentation 2.56 1.2-5.3 7.0 0.008** mode of delivery 2.1 1.6-2.8 24.9 <0.001* long duration of labour 1.05 0.7-1.6 0.05 0.822 intrapartum complications 9.9 5.6-17.7 92 <0.001* obstetric team 1.0 0.9-1.2 0.39 0.53 fetal distress 14.0 8.3-23.6 169 <0.001* resuscitation 13.5 10.7-16.9 492 <0.001* Fetal/infant characteristics female infant gender 0.72 0.5-1.1 2.8 0.092** birth trauma 35.1 11-111.6 95.6 <0.001* high infant birthweight (z score) 0.87 0.7-1.1 1.63 0.20** high infant head circumference (z score) 1.81 1.5-2.1 46.8 <0.001**

Only those variables considered a priori to be primary risk factors in the causal chain were taken forward to the regression analysis. Where several variables describe related characteristics (e.g. maternal weight and BMI) the single variable with the strongest association was taken forward to the multivariate analysis. Where the variables were less clearly related (e.g. infant weight and head circumference z scores) both were taken forward provided the test of association met the cutoff. The variable describing multiple births was highly significant but due to the small overall numbers and lack of potential intervention to prevent twinning it was not carried forward to the multivariate analysis as an independent risk factor. The small numbers of twins (five SB, five NE and two later controls) precluded separate subgroup analysis. 104 After step-down (backward) logistic regression the final model of significant risk factors (i.e. those risk factors for which the 95% confidence interval for odds ratios did not include unity) had a value of 0.28 (a measure of the log-likelihood value of the model at predicting caseness where 0 corresponds to a "constant-only” model and 1 to perfect prediction). The risk factors and their associated odds ratios are given in Table 22.

Table 22. Final model of risk factors for NE comparing 131 NE infants with 635 later controls.

Risk Factor Odds Ratios (95% Cl) maternal education 5 years or less 2.37 1.29-4.34 maternal height (1 cm + increment) 0.92 0.87-0.96 previous neonatal death 5.56 1.71-18.08 no antenatal care 2.37 1.29-4.34 maternal risk of sepsis (PROM/smelly liquor) 4.90 2.00-12.00 syntocinon for induction 8.77 3.09-24.91 syntocinon for augmentation 4.41 2.48-7.82 non-cephalic presentation 2.86 1.04-7.92 infant weight (1 SD score + increment) 0.48 0.33-0.69 infant head circumference (1 SD score + inc.) 2.83 2.08-3.85

As with most perinatal studies low maternal educational attainment was importantly associated with adverse outcome. Previous neonatal death and short maternal stature are indicators for risk of NE. Maternal risk factors for occult infection are significantly associated with NE. Interestingly only a complete absence of any antenatal care remained a significant risk factor for NE and there appeared to be no incremental benefit from repeat attendances. The most striking preventable risk factor from this analysis was induction of delivery using syntocinon. A total of 12 NE cases (9%) compared to 29 controls (5%) were actually induced indicating that induction was significantly and independently associated with only a minority of cases of NE. The risk of NE was inversely related to birthweight SD score but this association did not hold when birthweight was coded conventionally using 2.5 kg as the cutoff for low birthweight (22% LBW rate for both NE cases and controls). Larger neonatal head circumference at birth also appeared to be a risk factor for NE. 105 Risk factors for fresh stillbirth The results of univariate analysis for fresh stillbirth are presented in Table 23. As for NE a double asterisk (**) indicates those variables actually carried forward to the multivariate analysis.

Table 23. Univariate analysis of risk factors for fresh stillbirth.

Risk Factor OR 95% Cl chi 2 p value Maternal socioeconomic status maternal education 5 years or less 4.49 2.36-8.5 25 <0.001** Maternal anthropometry maternal weight < 40 kg 0.96 0.93-0.99 5.7 0.016* maternal height < 145 cm 0.96 0.92-0.99 4.8 0.028** low maternal BMI 0.95 0.87-1.02 2.07 0.15* Obstetric/perinatal history nulliparity 0.5 0.3-0.7 12.2 <0.001** previous stillbirth 3.4 1.5-7.5 9.9 0.002** previous neonatal death 4.2 1.7-10.4 11.5 0.001** Features of this pregnancy no antenatal care 2.83 1.8-4.5 20.5 <0.001** maternal anaemia, history of 1.9 0.6-6.0 1.3 0.25 maternal pre eclampsia, history of 2.1 1.03-4.3 4.3 0.038** twin 15.4 2.88-82 18.4 <0.001* maternal sepsis (PROM or smelly liquor) 5.2 2.1-9.1 32 <0.001** interval from admission to delivery 0.87 0.74-1.03 2.38 0.123* Intrapartum factors use of syntocinon 0.69 0.53-0.89 8.3 0.004** non-cephalic presentation 10.7 5.7-19.9 87 <0.001** mode of delivery 3.7 2.8-5.0 77 <0.001* long duration of labour 2.37 1.5-3.8 14.25 <0.001* intrapartum complications 5.4 4.3-6.7 239 <0.001* obstetric team 0.99 0.87-1.1 0.03 0.87 fetal distress 5.9 3.4-10.0 54 <0.001* Fetal/infant characteristics infant gender 0.7 0.46-1.06 2.8 0.09** infant birthweight (z score) 0.64 0.51-0.8 14.6 <0.001**

After step-down (backward) logistic regression the final model of significant (p<0.05) risk factors had a value of 0.24 (a measure of the log-likelihood value of the model at predicting caseness where 0 corresponds to a "constant-only" model and 1 to perfect prediction). The variables and their associated odds ratios are given in Table 24. The analysis resembles that for NE with two important differences. For the outcome of fresh stillbirth no protective effect of antenatal care was demonstrated and this outcome was more common in multiparous women. 106 Table 24. Final model of risk factors for fresh stillbirth comparing 107 fresh stillbirths with 635 later controls.

Risk Factor Odds Ratios (95% Cl) maternal education less than 5 years 4.39 2.13-9.05 maternal nulliparity 0.48 0.28-0.85 previous stillbirth 3.81 1.41-10.23 maternal risk of sepsis (PROM/smelly liquor) 5.05 2.27-11.23 syntocinon for induction 4.51 1.42-14.30 non-cephalic presentation 8.21 4.05-16.7 infant weight (1 SD score + increment) 0.69 0.51-0.93

Subgroup analysis

Multiparous women and inter-birth interval To examine the association between inter-birth interval and adverse pregnancy outcome a univariate analysis was performed on the subgroup of multiparous women for the exposure variable inter-birth interval using the conventional cut-off of 24 months. 51 (40%) of the NE mothers were multiparous, 65 (61%) of the SB mothers and 270 (43%) of the later control mothers. There was no statistically significant difference comparing the proportions of mothers with short inter-birth intervals between either the NE group (OR 0.87 [Cl 0.37-2.06]) or the SB group (OR 1.49 [0.77-2.89]) with controls. NE amongst Kathmandu residents In view of the recruitment bias inherent in a hospital based study I decided to repeat the analysis for the primary outcome of interest (NE) amongst the subgroup of local residents delivering at the hospital. The best fit model derived above for NE with later controls was run excluding all those living outside the Kathmandu ring road leaving 76 cases and 521 controls (Table 25). There was little loss of explanatory power (R^ 0.25 for the urban subset compared with 0.28 for the overall population sample) and three previously significant explanatory variables (maternal education, non-cephalic presentation and previous neonatal death) lost statistical significance. Subgroup descriptive analysis showed the population of women from outside the Kathmandu ring road presenting to the maternity hospital were more likely to be uneducated (62% of periurban women in the whole study population had no education compared with 46% of inner city women), to have experienced previous neonatal death (6% compared with 3%), and to present with non cephalic presentation (12% compared with 4%). Table 25. Rerun of model of risk factors for NE comparing 76 NE cases and 521 later controls all of whom were Kathmandu residents.

Risk Factor ______Odds Ratios (95% Cl) maternal education 5 years or less 1.69 0.81-3.51 maternal height (1 cm + increment) 0.92 0.87-0.98 previous neonatal death 2.99 0.49-18.21 no antenatal care 4.74 2.2-10.23 maternal risk of sepsis (PROM/smelly liquor) 4.36 1.52-12.5 syntocinon for induction 8.66 2.50-29.91 syntocinon for augmentation 4.77 2.33-9.73 non-cephalic presentation 1.27 0.29-5.57 infant weight (1 SD score + increment) 0.44 0.28-0.70 infant head circumference (1 SD score + inc.) 3.08 2.08-4.59

NE amongst planned hospital deliveries The best fit model derived above for NE with later controls was run excluding all those who had planned to deliver elsewhere than Prasuti Griha hospital leaving 105 cases and 594 controls (Table 26). There was no change in explanatory power (R^ 0.29 for the planned deliverers compared with 0.28 for the overall population sample) and two previously significant explanatory variable (no antenatal care and non-cephalic presentation) lost statistical significance.

Table 26. Rerun of model of risk factors for NE comparing 105 NE cases and 594 later controls in planned deliverers only.

Risk Factor Odds Ratios (95% Cl) maternal education 5 years or less 2.05 1.07-3.92 maternal height (1 cm + increment) 0.91 0.86-0.96 previous neonatal death 5.39 1.52-19.1 no antenatal care 1.94 0.92-4.09 maternal risk of sepsis (PROM/smelly liquor) 4.87 1.81-13.1 syntocinon for induction 6.38 1.97-20.6 syntocinon for augmentation 5.21 2.77-9.79 non-cephalic presentation 1.82 0.54-6.20 infant weight (1 SD score + increment) 0.43 0.28-0.66 infant head circumference (1 SD score + inc.) 3.36 2.3-4.78 108 Laboratory data Maternal blood was taken postnatally from the mothers in all study groups. As described in more detail in chapter 3 haemoglobin was assayed contemporaneously whilst for operational reasons serum and plasma was stored on site and biochemically assayed at a later date.

Maternal haemoglobin Contemporaneous puerperal haemoglobin (HB) data is available for the majority of mothers studied (n=1050,97% of all subjects). The distribution of the maternal haemoglobin data for all groups combined was normal (see Appendix 5). Table 27 summarises the puerperal haemoglobin levels of the study mothers by study group. There appears to be a downward trend in HB with worsening pregnancy outcome.

Table 27. Maternal haemoglobin by study group.

Study Group mean HB SD n (% of g contemporary controls 12.12 2.25 206 (99) later controls 11.51 1.98 614 (97) NE cases 11.08 2.28 128 (98) fresh stillbirths 10.79 2.25 102 (95)

To investigate this further the haemoglobin data was coded following conventional criteria for anaemia in pregnancy (HB<10 g/dl) and a univariate analysis of the relationship between puerperal anaemia and both NE and SB performed as shown in Table 28.

Table 28. Odds ratios for puerperal anaemia (HB<10) comparing 128 NE and 102 fresh stillbirths with 614 later controls.

Outcome Odds Ratios 95% Cl NE 1.54 1.00-2.37 fresh stillbirth 2.35 1.50-3.68

This association suggests that maternal anaemia may be a primary risk factor for adverse pregnancy outcome. However it seemed likely that two peripartum factors may complicate the interpretation of postnatal haemoglobin levels. Maternal haemoglobin levels fall in the puerperium and therefore postnatal values are likely to vary with time of sampling. This relationship was investigated with all study subjects and a downward correlation of maternal haemoglobin levels with time of sampling post delivery 109 confirmed (correlation coefficient -0.20). Secondly instrumental and especially operative obstetric intervention is likely to lead to increased maternal blood loss. Investigating this relationship for all study subjects confirmed in this population that puerperal anaemia as defined was more common following operative delivery (61 of 140 mothers[43%]) than after instrumental delivery (11 of 54 mothers [20%]) and normal vaginal delivery (160 of 689 mothers[18%]). Therefore a multivariate analysis was performed to adjust for these two factors. Table 29 demonstrates that the adjusted odds ratios are no longer statistically significant.

Table 29. Adjusted* odds ratios for puerperal anaemia (HB<10) comparing 128 NE and 102 fresh stillbirths with 614 later controls.

Outcome Odds Ratios* 95% Cl NE 1.21 0.72-2.03 fresh stillbirth 1.43 0.86-2.40 Note: * adjusted for mode of delivery and time of blood sampling

Finally the adjusted odds ratios were recalculated using a lower cut-off for maternal haemoglobin of 8.0 g/dl. Again after adjustment as above there was no significant associated risk of NE (adjusted OR 1.61 95% Cl 0.67-3.86) but there was an apparently significant risk of fresh stillbirth (adjusted OR 2.46 95% Cl 1.09-5.54). However there were few mothers with a haemoglobin below 8 g/dl (NE 11/128 [9%], SB 11/102 [11%], controls 29/614 [5%]).

Maternal thvroid function Early postpartum maternal thyroid function was assessed using standard assay techniques to measure free thyroxine (FT4) and thyroid stimulating hormone (TSH). Data is available for 1050 of the 1081 study patients (97%). Of the whole population assayed 39 (3.7%) had a FT4 of less than 8.8 pmol/1 and 206 (19.6%) had a TSH of more than 5 mlu/l which are conventional cut-offs describing normal values in non pregnant high income populations^272^. Maternal serum concentrations of both FT4^155^ and TSH^154) do not change significantly during pregnancy and therefore normal ranges in pregnant and non-pregnant women are considered equivalent. The means and standard deviations of the assay results by study group are given in Table 30. 110 Table 30. Mean values (standard deviations) for biochemical measures of maternal postpartum thyroid function by study group.

Variable NE cases Stillbirths contemporary later controls (n=126) (n=99) controls (n=194) (n=631) mean (SD) mean (SD) mean (SD) mean (SD)

FT4 12.81 (2.43) 14.32(3.24) 12.46 (2.27) 11.95 (2.01) (pmol/1) TSH 3.58(2.61) 3.06(2.19) 3.41 (2.23) 3.32(2.33) (mIU/1)

There was no significant difference between the two control groups on Student's t test for TSH values (t=0.48 p=0.63) but the contemporary controls had significantly higher FT4 values than later controls (t=2.99 p=0.0014) which raised the question of sample degradation due to the difference in age of the samples at the time of biochemical analysis. Therefore both case groups were compared with the contemporary controls as shown in Table 31. The only significant difference was that FT4 was higher in the stillborn group than in the control group. The effect is in an opposite direction to that hypothesised. Therefore the hypothesis that maternal hypothyroidism is a risk factor for adverse pregnancy outcome is not supported by this data.

Table 31. Comparative unpaired analysis of biochemical measures of thyroid function comparing mothers of encephalopathic infants and fresh stillbirths with contemporary control population.

Measure Groups compared mean dif. (95% Cl) t test (df) p value FT4 NEv control 0.35 (-0.18_0.87) 1.3(318) 0.19 (pmol/1) SB v control ______1.86(1.21-2.50) 5.7(291) <0.0001* TSH NEv control 0.17 (-0.37_0.71) 0.6(318) 0.53 (mlu/l) SB v control -0.35 (-0.89_0.19) 1.2 (291) 0.20 Notes: dif.= difference, * = p value < 0.05 I ll Maternal magnesium Early postpartum maternal magnesium status was assessed using standard assay techniques. Data is available for 1046 of the 1081 study patients (97%) of whom 415 (39.6%) had levels below the conventional lower cut off for normal magnesium levels in high income populations (0.75-1.05 mmol/1)^272^. The means and standard deviations of the assay results by study group are given in Table 32.

Table 32. Mean values (standard deviations) for magnesium by study group.

Variable NE cases Stillbirths contemporary later controls (n=122) (n=99) controls (n=194) (n=631) mean (SD) mean (SD) mean (SD) mean (SD) Mg mmol/i 0.722 (0.09) 0.686 (0.12) 0.696 (0.08) 0.734 (0.09)

There was a significant difference between the two control groups on student t testing (t=5.30 p<0.0000) which again raised the question of sample degradation due to the difference in age of the samples at the time of biochemical analysis. Therefore comparison with control samples was limited to the contemporary group. The results (see T able 33) are conflicting with magnesium levels significantly higher in NE cases and non significantly lower in SB cases in comparison with contemporary controls. The hypothesis that maternal hypomagnesaemia is a risk factor for perinatal brain injury is not supported by this data.

Table 33. Comparative unpaired analysis of maternal serum magnesium levels comparing mothers of encephalopathic infants and fresh stillbirths with contemporary control population.

Measure Groups compared mean dif. (95% Cl) t test (df) p value Mg NEv control 0.02 (0.001_0.04) 2.1(314) 0.04* (mmol/1) SB v control -0.01(-0.03_0.013) 0.85(293) 0.39 Notes: dif. = difference, df = degrees of freedom, * = p value < 0.05 112 Acute Phase Proteins Early postpartum maternal acute phase proteins (C Reactive Protein [CRP] and Alpha 1 Acid Glycoprotein [AGP]) were assayed and data is available for 1046 of the 1081 study patients (97%). Using normal values derived from work with a mixed pregnant and lactating Zairian population^273) 243 (23%) had elevated CRP levels (above 60 mg/1) and 204 (19%) had elevated AGP levels (above 1.2 g/1). The means and standard deviations of the assay results by study group are given in Table 34.

Table 34. Mean values (standard deviations) for maternal postpartum acute phase proteins by study group.

Variable NE cases Stillbirths contemporary later controls (n=122) (n=99) controls (n= 194) (n=631) ______mean (SD) mean (SD) mean (SD)______mean (SD) CRP 67.9 73.96 44.7 38.0 (mg/1) (50.2) (52.1) (40.2) (36.3) AGP 1.21 1.14 0.87 0.79 (g/1) (0.39) (0.46) (0.35) (0.31)

There was a significant difference on Student's t test between the contemporary control values and the later control values for both CRP (t=2.19, p= 0.014) and AGP (t = 3.05, p=0.0012). This again raised the question of sample degradation and therefore comparison of case and control samples was limited to the contemporary control group (see Table 35).

Table 35. Comparative unpaired analysis of biochemical measures of acute phase response comparing mothers of encephalopathic infants and fresh stillbirths with contemporary control population.

Measure Groups compared mean dif. (95% Cl) t test (df) p value CRP NEv control 23.2(13.1-33.3) 4.5(314) <0.0001* (mg/1) SB v control ______29.3(18.4-40.1) 5.3(291) <0.0001* AGP NEv control 0.34(0.25-0.42) 8.0(314) <0.0001* (g/1) SB v control 0.27 (0.17-0.36) 5.6 (291) <0.0001* dif. = difference df = degrees of freedom * = p value <0.05

Both acute phase proteins were significantly higher in NE cases and SB cases in comparison with contemporary controls. To explore the association between sepsis and 113 acute phase response the contemporary controls and cases were analysed by the presence or absence of accompanying features associated with infection (prolonged rupture of membranes and/or smelly liquor) in Table 36. This demonstrates a modest increase in both CRP and AGP in control mothers with risk factors for sepsis. However case mothers appear to have raised acute phase proteins over and above that seen in association with septic risk factors in controls.

Table 36. Acute phase proteins in contemporaneous controls and cases by associated risk factors for sepsis. population sample CRP (mg/1) AGP (g/1) mean SD (n) mean SD (n) controls with no sepsis rf 44.3 40.1 (187) 0.86 0.34 (185) controls with sepsis rf 55.4 22.0 (7) 1.11 0.28 (7) case NE with no sepsis rf 63.0 47.9 (102) 1.22 0.40 (103) case NE with sepsis rf 93.2 55.7 (20) 1.22 0.36 (21) Note: rf = risk factors

To assess the effect of tissue injury on acute phase proteins CRP and AGP levels for the whole dataset (all cases and controls n=1046) were stratified by mode of delivery (Table 37). Both measures rise significantly with increasing intervention (Cuzick's test for trend in CRP z = 13.59, p<0.01, test for trend in AGP z = 12.97, p<0.01).

Table 37. Acute phase proteins by mode of delivery for all study mothers. population sample CRP (mg/1) AGP (g/L) mean SD (n) mean SD (n) Normal vaginal delivery 36.9 33.9 (850) 0.80 0.31 (849) Instrumental delivery 60.0 43.8 (51) 1.02 6.37 (52) Operative delivery 97.8 57.7 (137) 1.39 0.37 (137)

To further explore the acute phase protein response in mothers of NE cases CRP and AGP levels were recoded as a binary variable using 2 standard deviations of contemporary control values of CRP (120 mg/1) and AGP (1.57 g/1) as cut-off values. Because of the interactions demonstrated on univariate analysis with time of sampling, mode of delivery and associated risk factors for occult sepsis these variables were included in a multiple logistic regression for NE (n=122) compared with controls 114 (n=194). Table 38 demonstrates that acute phase proteins are raised in NE cases compared to controls and this effect appears to be independent of the confounding effects of premature ruptured membranes and/or smelly liquor, mode of delivery and time of sampling.

Table 38. Odds ratios for raised acute phase proteins (CRP>120 mg/1 and AGP > 1.57 g/1) comparing NE cases with controls adjusting for time of sampling, mode of delivery and maternal risk factors for occult sepsis.

Unadjusted OR (95% Cl) Adjusted OR (95% Cl) CRP (>120) 2.95 (1.53-5.70) 5.22 (2.21-12.28) AGP (>1.57) 4.63 (2.06-10.38) 3.99 (1.35-11.84)

Analysis of neonatal factors affecting grade of outcome

Infant haemoglobin The vast majority of NE infants underwent venepuncture as per the study protocol. Those bled more than 5 days after birth were excluded from the analysis since it is recognised neonatal haemoglobin falls sharply over time. The haemoglobin level read contemporaneously using the haemocue method for 116 samples (89% of the cohort) had a mean of 17.5 g/dl, a median of 17.7 g/dl, an inter quartile range of 3.9 g/dl and a standard deviation of 2.9 g/dl. 19 (16%) of this group had haemoglobin levels below normal values described for high income neonatal populations (14.5-22.5 g/dl )(274). The next stage of the analysis considered the possible univariate effect of gestational age at delivery, age at sampling, season of delivery, gender, maternal anaemia and mode of delivery on infant haemoglobin. None of these were significantly associated with neonatal haemoglobin. However the presence of neonatal cephalhaematoma was significantly associated with a lower neonatal haemoglobin. In order to explore the relationship between infant haemoglobin level and the severity of encephalopathy the cases were divided into two groups, mild (grade one) and moderate/severe (grades two / three). The haemoglobin level of the more severely affected NE infants (mean 16.95, SD 2.84) was significantly lower than that of those mildly affected (mean 18.54, SD 2.76) [Student's t test t=2.88, p=0.005]).The presence of cephalhaematoma confounds this association in seven affected cases but the trend remained when this was controlled for as depicted in Figure 4. 115 Figure 4. Interaction line plot for infant haemoglobin and NE grade split by presence or absence of cephalhaematoma.

^ 18.5 “ O' grade 1 ^ 17.5 - - O - grade 2 /3 S 16.5 = c Cephalhaematoma = 1 “ 15.5 “ No Cephalhaematoma = 0 14.5 “ 0 1

Blood sugar Blood sugar data was available for 125 encephalopathic infants. After the initial test repeat BM stix testing was performed only if hypoglycaemia was clinically suspected. In all 38 infants had blood sugar readings below 2.6 on at least one occasion. Of these 19 had grade one NE, 13 grade two NE and 6 grade three NE. This data is strongly biased by the shorter observation period and therefore reduced BM stix testing on the severe grade three infants many of whom died within hours of birth. Therefore grade three infants were omitted and risk of grade two NE compared with risk of grade one NE in association with hypoglycaemia (see Table 39). There was no significant difference in risk comparing these outcomes. This limited dataset does not support the hypothesis that neonatal hypoglycaemia importantly exacerbates the severity of neonatal encephalopathy.

Table 39. Relationship between detected neonatal hypoglycaemia and grade of neonatal encephalopathy.

Grade 2 NE Grade 1 NE Totals hypoglycaemia 13 19 32 normoglycaemia 40 27 67 Totals 53 46 n=99 Relative Risk =0.68, Chi square 3.17, P value 0.0751

Bilirubin 19 infants developed clinically significant jaundice requiring monitoring of their bilirubin. Of these 6 needed photo therapy. The peak bilirubin reading occurred most commonly on day 4 (range 3-6). No infant's bilirubin exceeded 350 mmol/1. Most authorities consider such levels of bilirubin to be harmless in the term infant<275). 116 Seasonality To assess the incidence of NE for evidence of seasonality I compared the distribution by month of birth of all NE cases bom in 1995 with the later control group which systematically recruited every 25th mother-child pair over a twelve month period. A Kolmogorov-Smimov analysis tested the null hypothesis that there was no difference in the monthly distribution of the NE cases and the later controls (Figure 5). The test found a Chi square of 2.0 with 2 degrees of freedom and an associated p value of 0.73 supporting the null hypothesis that there is no statistically significant difference between the distributions.

Figure 5. Comparison of month of delivery of NE cases and later controls.

later controls (n=632) NE cases (n=92) 70 12 60 10 50 8 § 40 o 6 o 30 o 20 4 10 2 0 0 “ i 6 8 10 12 14 6 8 10 12 14 Birth Month Birth Month

The comparison of the distributions of stillbirths with the later controls with regard to month of birth using the Kolmogorov-Smimov test showed even less difference with a Chi squared test value to 2 degrees of freedom of 0.7 and a p value of >0.99.

Figure 6. Comparison of month of delivery of NE cases and later controls.

later controls (n=632) SB cases (n=83) 16 7 0 1 60 - 50 " 12 c 4 ^ =1 40 " 3 o O 8 o 30 - O 20 " 4 10 - o -1 0 ~i 2 4 6 8 10 12 14 4 6 8 10 12 14 Birth Month Birth Month Key findings Both neonatal encephalopathy and fresh stillbirth are associated with lower maternal education, a broad marker for socioeconomic status in this population. There is no excess risk associated with nulliparity for either outcome in this population. Previous adverse pregnancy outcome is an independent risk factor. Non cephalic presentation and syntocinon for induction are importantly associated with both neonatal encephalopathy and fresh stillbirth. These outcomes are independent of measures of both maternal thyroid and magnesium status. The presence of risk factors for infection is importantly associated with both adverse outcomes. Acute phase proteins are raised in mothers of stillborn and encephalopathic infants, an association which is not explained by risk factors for infection or method of delivery. There is some evidence of increased risk of adverse outcome in mothers with severe anaemia and of increased severity of encephalopathy in infants with low neonatal haemoglobin levels. Both neonatal and fresh stillborn infants have lower birthweights than term controls, whilst encephalopathic infants appear to have larger head circumferences at birth. There appears to be no important seasonality associated with either outcome in this population. 118

6. Neurodevelopmental outcome of neonatal encephalopathy at one year

The dataset presented in this chapter compares 131 infants with neonatal encephalopathy of early onset for which no cause other than birth asphyxia could be detected and 208 contemporaneous normal controls who did not require special care.

Clinical characteristics of the infant population

Table 40 compares the perinatal characteristics of the cases and controls using Chi squared to test for significant differences betwen the two groups.There are more males amongst both cases and normal controls. All except one case infant had a birthweight between 2000 and 3999 g. Case infants were all bom with some evidence of depression at birth which persisted to five minutes in the vast majority of cases. For presentation of resuscitation data the infant appears in the group of the most aggressive intervention performed during the acute resuscitation process. The resuscitation was performed by a paediatric junior staff member in 103 of the 126 (82%) cases for which details are available. For the other 23 cases (18%) resuscitation was performed either by the midwife or obstetrician attending the delivery. The Parkin estimation of maturity is presented by raw score. The equivalent estimated weeks gestation is shown in parentheses. Nearly 20% of the cases had superficial evidence of birth trauma. These were mainly or moulding considered to be in excess of that commonly encountered following normal deliveries. There were five minor congenital abnormalities noted amongst the cases. These were a unilateral simian crease, a capillary haemangioma with an extensive distribution over the right upper arm and trunk, an infant with bilateral undescended testes but no evidence of intersex or a salt losing state and a fourth infant with mild micrognathia. A fifth infant had low set ears, flat nose, mild micrognathia and bilateral simian creases but did not fall into any named phenotype described in a standard work on dysmorphology^276^. 119 Table 40. Univariate analysis of perinatal infant characteristics comparing 131 NE cases with 208 contemporaneous controls.

Variable Category Cases(%) Control s(%) Chi 2 P value gender male 80 (61) 111(57) 1.44 0.23 female 51 (38) 93(42) total 131 204

Birth weight (g) 1500-1999 1 (1) 0 (0 ) 2.27 0 .6 8 2000-2499 27 (21) 49(24) 2500-2999 66(51) 98(47) 3000-3499 29 (23) 51(25) 3500-3999 6(5) 9(4) total 129 207

Apgar (one) 8 -1 0 0 (0 ) 3(2) 295 <0 .0 0 0 1 * 4-7 1 1 (8 ) 205 (99) 0-3 120 (92) 0 (0 ) total 131 208

Apgar (five) 7-10 3(2) 203 (98) 308 < 0 .0 0 0 1 * 5-6 40(31) 5(2) 0-4 88(67) 0 (0 ) total 131 208

Resuscitation none 2 (2 ) 206 (98) 313 < 0 .0 0 0 1 * facial oxygen 45 (35) 0 (0 ) handbagging 46(36) 2 (1) intubation 9(7) 0 (0 ) cardiac massage 27(21) 0 (0 ) total 129 208

Parkin 6 (37 weeks) 13 (10) 9(4) 8 .0 0 0.15 7 (38.5 weeks) 24(19) 34(16) 8 (39.5 weeks) 57 (45) 90(43) 9 (40 weeks) 32 (25) 69 (33) 10 (41 weeks) KD 5(2) total 127 207

Congenital none 126 207 3.28 0.07 abnormality minor 5 2 major 0 0 total 131 209

Birth Trauma none 106 (82) 203 37.3 <0 .0 0 0 1 * bruised arm 1 (1) 0 forceps marks 1 (1) 0 cephalhaematoma 9(7) 1 marked caput 12(9) 0 facial nerve palsy 1 (1) 0 total 131 204 Note: * = p value <0.05 120 Five of the cases (4%) and one of the controls (1%) were one of twins. The perinatal details and early outcome of the corresponding twin are presented in Table 41.

Table 41. Clinical details of co-twins of NE cases.

Birth order Birth Inter-twin Birth details Early outcome weight interval (kg) twin 1 3.0 lOmins cephalic pres. macerated normal vag. del. stillbirth twin 1 2 .0 30 mins footling breech, normal normal vag. del. twin 2 1.94 3 mins cephalic pres. normal normal vag. del. twin 2 1 .6 2 0 mins cephalic pres. anencephaly normal vag. del. twin 2 No data recorded

Notes: pres. = presentation, vag. del. = vaginal delivery

Table 42 presents relevant clinical features of the evolving encephalopathy in the 131 cases. Seizures were noted in 44% of cases, usually on the first day of life. Treatment included intravenous antibiotics in 64 (49%) of cases, antiepileptic medication in 100 (76%) of cases and phototherapy for in 10 ( 8 %) of cases. No infant's bilirubin exceeded 350 mmol/1.

Table 42. Early clinical characteristics of the cases of NE.

Variable Category n cases NE Grade mild (1) 48(37) moderate ( 2 ) 53(40) severe (3) 30 (23) total 131 seizures clinically observed no 74(56) yes 57(44) total 131 day seizures commenced 1 43 (75) 2 11(19) 3 or more 3(6) total 57 neurological condition at normal 40(44) discharge abnormal tone 26 (29) abn. tone and suck 1 0 ( 11) abn. tone/suck/conscious 15(16) total 91 Note: abn. = abnormal Figure 7. Temporal course of disease progression and resolution for for resolution and encephalopathy. progression neonatal with disease of newborns course Temporal 7. Figure

Count 20 u 20 o ■£ H 30 o 100 0 “ 40 40 20 60 80 10 0 I 50 |Tp|in n i ny |TTnpn| 0 0 1 3 5 7 9 0 1 2 3 4 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 1 3 5 7 9 0 1 2 3 4 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0 a nuooia anraiis ea t resolve to began abnormalities neurological day 3 5 7 9 0 1 2 3 4 15 14 13 12 11 10 9 8 7 6 5 4 3 2 day of occurrence of maximum ne grade ne maximum of occurrence of day |juj i 1 1 i | uUf 1 1 i f 1 1 11 1 1 day of death of day day of discharge of day riTT] , prTijTTTT] i 1 ------1 ------^ r 121 122 Figure 7 gives a visual impression of the disease progression with the modes for the day of death, maximum clinical expression, onset of recovery and hospital discharge at days 1,1,3 and 5 respectively. It should be noted that 51(44%) of the surviving 91 infants remained abnormal at discharge. Most commonly the persisting abnormality was axial hypotonia but for a significant minority of 25 infants (27%) there was persisting evidence of poor suck and associated feeding difficulties and in a small subgroup of 15 (16% of the total survivors) there was continuing evidence of abnormal consciousness either manifest as persistent irritability or lethargy.

Clinical outcome Follow up status of the original cohort to one year is graphically presented in Figure 8 . Full one year outcome data is available for 102 cases (78%), including 45 deaths and 106 controls (51%), including 4 deaths.

Figure 8. Flow chart of cohort follow up.

Ne cases (n =131) Controls (n=208)

131 208 40 early neonatal deaths 1 post discharge death 1 post discharge death ► 83 lost to follow up i 18 lost to follow up 6 weeks 124 follow up 72 2 deaths 3 deaths 8 lost to follow up 11 lost to follow up 3 returned to follow up

6 months 113 follow up 62

2 deaths 0 deaths 8 lost to follow up 17 lost to follow up 6 returned to follow up 5 returned to follow up

12 months follow up 57 102

Death In total 45 cases and 4 controls are known to have died before their first birthday. Figure 9 presents a survival curve for all 208 infants by caseness up to one year of age. It demonstrates that the vast majority of infant case deaths (39 of 45, 87%) occured in the first week of life. Two other minor clusters are seen for both cases and controls at 1-2 months of age and 4-6 months of age. Cause of death is considered in more detail below. 123

Figure 9. Survival curve to one year of age (102 NE cases and 106 controls).

1

.8

Cum. Survival (control) O Event Times (control) Cum. Survival (case) □ Event Times (case)

,2

0

0 30 60 90 120 150 180 210 240 270 300 330 360 Time (days)

Neonatal deaths All controls (who by the recruitment protocol were normal neonates who did not require special care) survived to hospital discharge. 40 case neonates died prior to hospital discharge and are presented by grade of NE in Table 43. Although severe NE accounts for the vast majority of these early deaths a small but significant proportion of both mild and moderate NE also died.

Table 43. The distribution of neonatal deaths by grade.

NE grade n dead/group total(%) mild NE - grade one 3/48 (6) moderate NE - grade two 9/53 (17) severe NE - grade three 28/30 (93) ______totals 40/131(31)

To explore the cause of death in more detail Table 44 presents relevant clinical details for each neonatal death in order of age at death. Clinically type 1 presented with worsening signs of respiratory distress as evidenced by tachypnoea and chest indrawing accompanied by hypoxaemia assessed using the single portable oxygen saturation monitor available to the unit. Type 2 respiratory (ventilatory) failure was characterised by worsening recurrent apnoeas. Meconium aspiration syndrome, associated with a variable degree of respiratory distress contributed to death in 17 of the 40 cases (42%) and was the principle cause of death in four cases (10%). 124

Table 44. Clinical details of all hospital deaths following NE (n=40).

Age at Grade Terminal event Contributory clinical factors death NE (hours) ______3 3 Type 2 thick meconium stained liquor, meconium in larynx at resus, 3 3 Type 2 4 3 Type 2 thick meconium stained liquor 6 3 Type 2 6 3 Type 2 thick meconium stained liquor cardiac failure 8 3 Type 2 8 2 Type 1 thick meconium stained liquor meconium in larynx at resus 8 3 Type 2 thick meconium stained liquor meconium not seen in larynx 9 3 Type 2 thick meconium stained liquor, meconium in larynx at resus, 9 3 Type 2 16 3 Type 2 17 3 Type 2 thick meconium stained liquor, meconium in larynx at resus, massive pulmonary haemorrhage, 19 3 Type 2 frothy secretions at post mortem laryngoscopy 2 0 3 Type 2 thick meconium stained liquor, meconium in larynx at resus, severe postnatal L to R shunting with 2 0 2 Type 1 thick meconium stained liquor, meconium in larynx at resus 21 3 Type 2 25 2 Type 1 thick meconium stained liquor, severe postnatal L to R shunting with cyanosis 29 1 Type 1 thick meconium stained liquor, severe postnatal L to R shunting with cyanosis 30 3 Type 2 30 2 Type 2 thick meconium stained liquor, 31 3 Type 2 39 3 Type 2 thick meconium stained liquor, meconium in larynx at resus 40 3 Type 1 pyrexia and respiratory distress day 2 , massive pulmonary haemorrhage,probable late sepsis 44 3 Type 2 48 2 Type 1 pyrexia and respiratory distress day 2 , probable late sepsis 48 3 Type 2 thick meconium stained liquor, 51 2 Type 1 52 2 Type 2 125 Table 44 (cont.) Clinical details of all hospital deaths following NE.

Age at Grade Terminal event Contributory clinical factors death NE (hours), ______53 2 Type 2 Respiratory Failure 56 3 Type 2 Respiratory Failure massive pulmonary haemorrhage 56 3 Type 2 Respiratory Failure 64 3 Type 2 Respiratory Failure 70 3 Type 2 Respiratory Failure cardiac failure 75 1 Type 2 Respiratory Failure thick meconium stained liquor, meconium in larynx at resus. 75 3 Type 2 Respiratory Failure cardiac failure 83 3 Type 2 Respiratory Failure 124 2 Massive pulmonary haemorrhage thick meconium stained liquor, meconium in laiynx at resus, Klebsiella isolated day 4 probable late onset sepsis 159 1 Type 2 Respiratory Failure thick meconium stained liquor, probable late onset sepsis 166 3 Type 2 Respiratory Failure 243 3 Type 2 Respiratory Failure Bleeding diathesis with gastrointestinal bleeding,

Three mildly encephalopathic infants died after the first 24 hours, two probably as a result of the complications of meconium aspiration syndrome and one as a result of late onset sepsis. Nine moderately encephalopathic infants died. Meconium aspiration syndrome contributed to death in five cases and in two cases late sepsis played a part. 28 severely encephalopathic died. Meconium aspiration syndrome contributed to death in nine of these cases. For the majority of grade three deaths the terminal event was ventilatory failure presumed due to brain stem involvement in the cerebral hypoxic- ischaemic pathological process. In three cases cardiac failure was clinically apparent, in one case late onset sepsis was a likely contributory factor and haemorrhagic disease of the newborn contributed to death in another case. It is noteworthy that ten deaths (ie 25% of all early mortality) had occurred by ten hours of age.

Post neonatal deaths There were nine deaths after hospital discharge to our certain knowledge, five cases and 4 controls. The clinical circumstances of the deaths are presented in Table 45 by order primarily of caseness and secondly by age at death. 126 Table 45. Clinical progress and cause of death for all 9 postneonatal deaths.

Ne Age at Discharge 6 week 6 month Verbal autopsy findings Grade death condition condition condition Primary cause of death ______(days)______(Contributory factors) 1 31 normal Septicaemia (umbilical infection) 1 34 abnormal FTT, ARI tone breastfed (diarrhoea and FTT)

2 7 42-180 abnormal spoon/bottle, death reported by neighbours tone hypertonic, family moved therefore no verbal ARI at 3 wks autopsy findings available

2 145 abnormal FTT, ARI tone and bulbar palsy, (protein energy malnutrition) suck, hypotonia

1 348 abnormal FTT, FTT, ARI tone breastfed, hypotonia hypotonia control 46 LBW nomal, ARI breastfed, growing well control 55 normal Diarrhoeal illness control 164 normal bottlefed Diarrhoeal illness control 180 normal mixed Aspiration breast/bottle Notes: FTT = failure to thrive, ARI = acute respiratory infection

Amongst the case deaths all but one left hospital with abnormal tone. Two infants had persisting serious feeding difficulties which contributed to their deaths as described below. Study No 206 Last assessed at three months of age had a complete bulbar palsy, severe acquired microcephaly, generalised indicative of an evolving spastic tetraplegia and probable cortical blindness. His impoverished rural mother's breastmilk supply failed lacking the stimulation of the infant's suck and as she could not afford artificial milk formula she was reduced to dipping a cloth in buffalo milk and wringing the milk into her infant's mouth. The infant had severe protein/energy malnutrition (weight 2.19 kg) when last seen at 3 months of age. Mother declined hospital admission on financial grounds and the infant died of a chest infection just before 5 months of age. Study No 237 Re-presented at three weeks of age with what seemed likely to be an aspiration secondary to an incomplete bulbar palsy for which he was treated with antibiotics. By six weeks he seemed to have recovered and was thriving but still requiring some spoon feeds. We had no 127 further direct contact with the family but neighbours informed our team that the baby had died, the father lost his job as a casual carpetworker and the family had returned to their hill village outside the Kathmandu valley. In two other cases the cause of death was acute respiratory infection (ARI) on a background of failure to thrive. In a fifth case there was convincing verbal autopsy evidence of periumbilical sepsis progressing onto septicaemia. Of the control deaths two infants died of diarrhoeal disease and one of ARI. The account of the fourth infant's death was poor as mother had left him with her own mother in the village and she was told the baby had acutely aspirated and died in the night. Although infanticide must be included in the differential diagnosis it seems unlikely given the infant's male gender and age, and I suspect this case would have met UK criteria for sudden infant death syndrome.

Impairment Neurodevelopment Table 46 presents the neurodevelopmental findings at initial blinded screening of all 159 members of the cohort examined at one year of age.

Table 46. Developmental screening and initial neurological assessment findings at 12 months of age for 102 control infants and 57 NE infants.

DDST screening Case Normal Questionable Abnormal Totals Status (col%) ______Neuro assessment ______Controls normal 100 0 0 100(98) min.neuro impairment 0 1 1 2(2) maj neuro impairment 0 0 0 0 (0) totals (row%) 100(98) 1(1) 1(1) 102(100)

Cases normal 37 0 0 37(64) min.neuro impairment 0 3 3 6(11) maj neuro impairment 0 0 14 14(25) totals (row%) 37(65) 3(5) 17(30) 57(100)

Of the 102 controls followed up to one year the first stage blinded DDST screening found one to have definite delay, one to have questionable delay and passed the remaining 100 controls as normal. The neurological assessment found evidence of minor motor impairment in these same two infants and detected no evidence of neurological impairment in the remaining 100. The relevant clinical progress of the two control infants with minor neuroimpairment at one year of age are presented below as abbreviated clinical notes. 128

Study no 7 Lowbirthweight infant bom by normal vaginal delivery with Apgar's of 5 at one and 8 at five. At six weeks thriving, breastfeeding and developing normally. Treated for two chest infections by six months of age when again found to be thriving and neurodevelopmentally normal. At one year assessment hearing was normal but the infant was markedly hypotonic in all four limbs although sitting independantly and a latent squint was queried. DDST classed as questionable. Referred for full assessment at which hypotonia of unknown cause confirmed but no evidence on cover testing of a squint. MDI 89, PDI 74.

Study no 25 Normal birthweight infant bom by normal vaginal delivery with Apgar's of 7 at one and 8 at five. Noted to have suffered an episode of chest infection prior to the six week check when the infant was thriving, breastfeeding and developing normally. Thought to be developing normally at six months. At one year screening he was thought to have normal visual tracking, reduced response to both high and low frequency sound presented on the left side, reduced tone in the left arm and leg and was seriously delayed according to the DDST. Full assessment found no evidence of assymetry of motor function or hearing but confirmed a significant mild motor delay with an MDI of 97 but a PDI of 65.

Of the 57 surviving NE cases who underwent neuro-developmental screening at one year of age 20 (35%) were found to have evidence of neuro-developmental impairment. Tables 47(a) and 47(b) present the developmental screening and initial neurological assessment findings at 12 months of age according to grade of N E .

Table 47(a) Developmental screening findings by grade of NE for 57 case infants at 12 months of age.

NE grade Normal Questionable Abnormal Totals l(row%) 26(90) 0 (0 ) 3(10) 29(100) 2 (row %) 10(37) 3(11) 14 (52) 27(100) 3 (row %) 1(100) 0 (0 ) 0 (0 ) 1 (1 0 0 ) Totals 37 (65) 3(5) 17 (30) n=57 (100)

Table 47(b) Neurological screening findings by grade of NE for 57 infants at 12 months of age.

NE grade normal Minor Major Totals impairment impairment l(row%) 26(90) 2(7) 1(3) 29(100) 2 (row %) 10(37) 4(15) 13(48) 27(100) 3(row%) 1(100) 0 0 1 (1 0 0 ) Totals 37(65) 6(11) 14(25) n=57(100) 129

The main clinical findings from the assessment process at 1 year of age on all 20 abnormal cases detected by screening are presented in Table 48 ranked in decreasing severity of neurodevelopmental impairment. 18/20 (90%) had major neuro- developmental impairment as defined at the outset of the study. The vast majority (16/18) had experienced a grade two (moderate) encephalopathy. Only two infants developed major sequelae following a grade one (mild) encephalopathy. One of these, study number 66 (see Table 48) displayed a pattern of neurodevelopmental impairment which appears atypical for hypoxic-ischaemic cerebral injury. There were some features of the history suggestive of Prader-Willi syndrome which remains a possibility but since this remains unconfirmed it was decided not to exclude this single case retrospectively. Only study numbers 82 and 176 satisfy our a priori criteria for a minor neurodevelopmental impairment with early signs of evolving mild right hemiplegia, no significant functional disability, no evidence of associated impairments of vision or hearing, no history of seizures and developmental indices within 2 standard deviations of the mean. Only one of the 20 affected families chose not to attend for reassessment (study number 212) and therefore did not undergo Bay ley testing. The verbatim clinical summary of his screening assessment reads "neurodevelopmental assessment marked motor developmental delay mainly affecting left upper arm ?defective hearing on left ear" and therefore this infant was classified as having functionally significant impairment and coded as having major impairment. The most striking aspect of the clinical findings is the severity and multiplicity of impairments suffered by the affected case infants. Of the 18 with major sequelae 11 (61%) have a spastic tetraplegic pattern of cerebral palsy (see Figure 10). 17 (94%) have evidence of cognitive impairment, 8 (44%) have major impairment of either vision or hearing and 7 (39%) have a seizure disorder. 10 (56%) have aquired microcephaly suggesting extensive cerebral damage.

Figure 10. Cerebral palsy by pattern of involvment (n=17).

Cerebral Palsy Subtypes

I 1 spastic tetraplegia H spastic hemiplegia | spastic diplegia | dyskinetic-dystonic 130 Table 48. Neurodevelopmental findings for NE case screening failures at one year (n=20).

Study NE MDI PDI neurological hearing viskn fits diagnosis No grade abnormalities 119 2 <55 <55 drooling, maj maj yes spastic tetraplegia opisthotonus, severe microcephaly, fits, hypertonia auditory inattention R=L / a=l cortical blindness 217 2 <55 <55 drooling, maj maj no spastic tetraplegia, opisthotonus cortical blindness, auditory hypertonia inattention, microcephaly R=L / a=l 220 2 <55 <55 opisthotonus maj maj yes spastic tetraplegia, cortical nystagmus blindness, microcephaly hypertonia R=L / a= 1 8 1 <55 <55 drooling min maj ? spastic tetraplegia, cortical hypertonia blindness, unilateral hearing L>R / a=l loss, microcephaly, 120 2 <55 <55 drooling, n maj yes spastic tetraplegia, fits, hypertonia microcephaly, cortical blindness, L=R / a=l 339 2 <55 <55 hypertonia maj min no spastic tetraplagia, hearing L>R/a>l impairment, microcephaly 271 2 <55 <55 hypertonia maj n yes spastic tetraplegia, microcephaly R=L / a=l hearing impairment 77 2 <55 <55 drooling n n yes spastic tetraplegia, fits, hypertonia microcephaly R=L / a=l 124 2 <55 <55 hypertonia n min no spastic tetraplegia, microcephaly L>R / a>l 266 2 <55 <55 dystonic posturing, n maj yes dyskinetic -mainly dystonic hypertonia microcephaly, R>L/ a > 1 287 2 <55 <55 hypertonia ? n no spastic tetrapegia R=L / a=l 244 2 <55 <55 hypertonia ? n no spastic tetraplegia R=L / a=l 190 2 <55 <55 mildl hypertonia n min no spastic tetraplegia R=L / a=l 88 2 <55 <55 mild hypertonia n n no spastic tetraplegia L=R / a=l 329 2 <55 <55 hypertonia ? n yes spastic diplegia L=R/ a < 1 212 2 fisting and ? n no left hemiplegia, hypertonia L a > 1 112 2 55- <55 mild assymetry of min n no left spastic hemiplegia 84 hand function 66 1 <55 55- mild generalised n n no obese (>97th %),hypotonia, 84 hypotonia behavioural difficulties, (female, no hypoglycaemia) ?Prader Willi 82 1 85- 55- mild assymetric n n no ? evolving right spastic 114 84 hypotonia hemiplegia R a > 1 176 2 55- 85- mild assymetry of n n no right spastic hemiplegia 84 115 tone increased R arm Key: to describe patterns of tone abnormality R=right L=left a=arm l=leg; n-normal min=minor impairment maj=major impairment =uncertain ? clinical findings 131 Hearing and vision On distraction hearing assessment 98 of the 102 control infants responded to high and low frequency sound bilaterally. In no control infants were the assessors clinically sure of absent responses. In three infants they remained uncertain as to a response unilaterally and one infant refused to co-operate with the test. One of these infants was clinically normal on repeat testing. The parents of the other three infants did not think there was any problem with their children's hearing and declined to reattend for repeat testing in the absence of any overt problems. In only one control infant was any eye or visual abnormality queried, a latent squint which was not confirmed on cover testing at re-assessment. Eight of the 57 cases (14%) were found to have clinical evidence of hearing impairment. The screening assessors were clinically convinced of absent responses to sound in four cases. Of these, three showed a complete lack of response to both high and low frequency bilaterally at the screening examination and these findings were replicated at re-assessment The screening assessors were sure of a bilateral lack of response to high frequency in the fourth case but remained uncertain as to the responses to low frequency. The reassessor remained uncertain as to the response to sound in this diplegic child. In a further sixteen cases the screening assessors were uncertain as to responses. On reassessment a hearing deficit was confirmed in two cases, suspected but considered of minor importance in a further three cases. Of the remaining 11 who failed screening nine were normal on retesting and two failed to attend for reassessment. In total there were five major and three minor hearing impairments detected at one year. On eye examination and vision testing nine of the 57 cases (16%) were found to be impaired. There were six infants with manifest squint, three with nystagmus and five appeared cortically blind (six infants had more than one abnormality). Since uncomplicated squint is not strictly functionally disabling at this age it was coded as minor visual impairment which therefore constitute three cases. The other 6 affected infants were coded as having major visual impairement.

One year cohort outcome analysis The overall outcome at one year for cases (by grade and overall) and controls is given in Table 49. The proportion dead or with major impairment following moderate HIE was 27/38 (71% [95% Cl 54 -84%]) and following severe NE to be 28/29 (97% [95% Cl 82-99]). 132 Table 49. Outcome of cohort (102 NE cases and 106 controls) successfully followed up to one year of age.

Normal Minor imp. Major imp. Dead Total Ne 1 (row%) 26 (74) 1(3) 2(6) 6(17) 35 (100) Ne 2 (row%) 10(26) 1(3) 16 (42) 11(29) 38(100) Ne 3 (row%) 1(3) 0(0) 0(0) 28(97) 29 (100) Total NE (row%) 37 (36) 2(2) 18(18) 45(44) 102 (100) Control (row%) 100 (94) 2(2) 0(0) 4(4) 106(100) Total cohort(row%) 137(66) 4(2) 18(9) 49(23) 208(100) Note: imp. = impairment

A series of two by two tables were constructed to study the relationship between the grade of encephalopathy and outcomes of interest (death or impairment) using controls as baseline data. For each outcome risk ratios (and 95% confidence intervals) were calculated by grade of NE and are presented in Table 50. Where the proportion affected of either cases or controls is zero the risk ratio (RR) could not be calculated. Therefore the relative risk for major impairment alone could not be estimated.

Table 50. Relative risks of death (infant and postneonatal) or impairment (any, major or minor) by grade of NE in comparison with control group.

Infant Postneonatal Major or Minor Minor Death Death Impairment Impairment ______RR (95% Cl) RR (95% Cl) RR (95% Cl) RR(95% Cl) grade 1 NE 4.5 (1.4-15.2) 2.5 (0.6-10.5) 5.3 (0.9-30) 1.9 (0.2-20.0) grade 2 NE 7.7(2.6-22.7) 1.8(0.4-9.5) 32.1 (7.9-131) 4.6(0.5-47.1) grade 3 NE 25.6(9.8-67.1) ne ______ne______ne______all NE 11.7(4.4-31.3) 2.1 (0.6-7.7) 17.9(4.3-73.8) 2.6(0.4-17.9) Note: ne=not estimable due to small numbers

Anthropometry To investigate the consequences of NE for subsequent growth the changes in SD scores for weight, length and head circumference (OFC) were analysed over the 12 month follow up period. Figure 11 presents the cohort (divided into cases and controls) anthropometric SD scores (and 95% Cl intervals) for weight, height, and head circumference at birth, 6 weeks, 6 months and 12 months. The number of subjects for whom data is available falls from 129 cases and 204 controls at birth to 55 cases and 99 controls at one year of age. Figure 11 (a). Growth trends for infant weight (SD score with 95% Cl bar) over the first year of life for cases and controls.

o in -.8 - ON « - 1.2 - V. o w - 1.6 O fucase (O fucontrol c -2

£ -2.4 1 -Co* a* - 2.8 > Birth 6 weeks 6 months 1 year Age of infant (not to scale) Figure 11 (b). Growth trends for infant length (SD score with 95% Cl bar) over the first year of life for cases and controls.

o .2 in ■■on_ ' 0 -.2 V. o -.4 w - e - Ct -.6 fucase (0 -.8 fucontrol c Of*0 -1 £ 1.2 a* 1.4 ac 1.6 Birth 6 weeks 6 months 1 year Age of infant (not to scale) Figure 11 (c). Growth trends for infant head circumference (SD score with 95% Cl bar) over the first year of life for cases and controls

o 0 in ■w-on -1 O f oV_ fucase wo 2 Ct —B— fucontrol CO c 3 Of £ o L u 4 O Birth 6 weeks 6 months 1 year Age of infant (not to scale) 134 At birth the cohort lay at between one and two standard deviations below an equivalent term UK population in all variables except for birth head circumference which showed some sparing of this effect. Interestingly although the cases and controls at birth were alike in other respects the cases had significantly larger heads than the controls (mean difference in SD 0.6 [95% Cl 0.3-0.8], student t test 4.5, p value <0.0001). As Figure 11 demonstrates there is significantly reduced linear growth and head growth in cases as compared with controls by one year but the difference in weight change does not reach significance. Aquired microcephaly is a well recognised manifestation of cerebral birth injury. The relative lack of brain growth was most marked in the case population over the first 6 weeks of life. Early aquired microcephaly was defined as a reduction in OFC SD score of 2 or more between birth and 6 weeks of life. 7 of the 17 (41%) impaired survivors and 2 of the 38 (5%) unimpaired survivors of NE met this definition at 6 weeks of age. Is this reduction in growth limited to overtly impaired survivors? Table 51 compares the change in SD scores (ASD) over the first year of life for each growth variable between firstly unimpaired survivors and normal controls and secondly impaired survivors and normal controls. Not surprisingly and in line with other cerebral palsy studies survivors with major impairment have severely affected growth in all variables. Head growth is worst affected. The unimpaired do not show a significant effect on weight gain but there is evidence of a smaller but still significant effect on length and head growth.

Table 51. Change in SD growth scores (ASD) over the first year of life comparing growth in unimpaired (n=38) and impaired survivors (n=17) of NE with that of controls (n=99). variable study group Mean ASD comp.v controls t test p-value weight control -0.14 unimpaired case -0.42 1.42 0.16 impaired case -1.19 3.37 0 .0 0 1 * length control 0.72 unimpaired case 0 .0 2 2.56 0 .0 1 2 * impaired case -0.93 4.07 0 .0 0 0 1 * OFC control -1.01 unimpaired case -1.76 3.41 0.0009* impaired case -4.38 9.64 <0 .0 0 0 1 * Note: * = p value <0.05 135 Cranial ultrasonography Earlv neonatal ultrasound Of the 131 cases a total of 107 (82%) underwent early ultrasound examination. Of those not examined either the machine was not available, a suitably trained investigator was not available or else the subject died before examination could be performed. 100 were imaged by 48 hours of age the remaining 7 were imaged within a week of life. Contemporaneous descriptions were compared with review of hard copy all performed by the principal investigator blind to the original assessment. There were differences of opinion over the presence of echogenicity in 29 cases, resolved by a second blinded review by the project supervisor. In the majority of cases (20/29) the original opinion was adjudged to be falsely positive, the degree of echogenicity being ruled to be within normal limits by both reviewers. The distribution of echogenicity on cranial ultrasound examination is presented below by NE grade. No cases of so called focal echogenicity were detected.

Table 52. Pattern of echogenicity on cranial ultrasound by NE grade.

NE grade Localised Generalised Normal Total echogenicity echogenicity l(row%) 3(7) 5(12) 34(80) 42(100) 2(row%) 1(2) 11(24) 34(74) 46(100) 3(row%) 0(0) ______8(44) ______10 (56)______18(100) total 4(4) 24(23) 78(74) 106(100) (row %)

Six week ultrasound data Cranial ultrasonography was performed on 114 (92%) of the 124 controls and all of the 72 cases attending for six week check. To establish the normal range of ventricuar index (VI) for this population the average values of the left and right ventricular indices were first plotted against age at examination. This comparison of normal individuals at ages varying from 3 to 17 weeks demonstrated a relatively low apparent growth velocity for VI (mean 0.25 cm/week). Therefore to estimate the normal range for VI at 6 weeks (42 days) all those control infants assessed between 30 and 60 days of age were included (n=103). The distribution of the averaged VI for this group approximates to normality (see Appendix 5 ) with a mean (range) of 11.4 mms (6.0-17.9) and standard deviation (SD) of 2.6. The upper 2nd SD of 16.6 mms was taken as the cutoff for the definition of ventriculomegaly in case infants examined between 30 and 60 days. Figure 12 shows the distribution of ventricular size in 60 case infants examined between 30 and 60 days of age. There appears to be a skewed distribution with an excess of infants with larger ventricles. Six cases had a VI exceeding 16.6 and therefore met the criteria for ventriculomegaly.

Figure 12. Distribution of the average ventricular indices of case infants examined between 30 and 60 days post-delivery (n=60).

14 I . » ■ 1 - I . I I . ■ .. . I ...... I. . I. . . I .

12 ------

10 “ —

8 - — ------

4 -

2 - — ------

0 ‘|tr it i'-t|-nr Itl-r )-i'■"!' rttm i'T ■ iHi- i rv '|"T■■■y i vt 1 t | t nH ■ ~ i i i4 ■ frrv t t |

4 6 8 10 12 14 16 18 20

Average Ventricular index (mm)

Cysts were reported contemporaneously in 8 cases. On hard copy review one of these was adjudged a false positive and a further two cases were detected making a total of 9 infants with cysts all of whom were NE cases. The overall cranial sonography results at six weeks are presented in Table 53 for 72 cases (by grade of NE) and 114 controls.

Table 53. Six week ultrasound findings for 72 cases and 114 controls.

Cysts V entriculomegaly Normal Total

NE grade 1 (row %) 0 (0 ) 1(3) 34 (97) 35(100) NE grade 2 (row %) 9(26) 5(15)* 21(62) 34(100) NE grade 3 (row %) 0 (0 ) 0 (0 ) 3(100) 3(100) All NE (row %) 9(13) 6*(7) 58 (80) 72 (100) Controls (row %) 0 (0 ) 5(4) 109(96) 114(100) Note: *includes one case also with cysts and therefore counted twice 137 Predictors of major neurodevelopmental impairment at one year In a series of analyses the value of various low cost predictors of major impairment in NE cases was reviewed. These predictors were based on the severity of encephalopathy (grade 2/3 v 1), abnormal neurology at neonatal hospital discharge, echogenicity on early ultrasound, cystic changes on late (six week) ultrasound and acquired microcephaly at six weeks. The full reviews, including the precise definitions and 2 by 2 tables of coded data are appended in Appendix 6. Table 54 compares various indices of their usefulness as predictive tests.

Table 54. Comparative review data of the properties of various low cost predictors of neurodevelopmental impairment in encephalopathic infants.

______Test NE severity abn d/c exam neo U/S 6/52 U/S 6/52 micro

Sensitivity 8 8 % 89% 50% 43% 41% Specificity 69% 83% 83% 1 0 0 % 95% (+) pre value 57% 52% 57% 100 % 78% (-) pre value 93% 92% 78% 80% 78% (+)LR 2 .8 5.2 2.9 > 1 0 0 0 8 .2 (-) LR 0.17 0.13 0 .6 0.57 0.62 Notes: (+) / (-) pre value = positive / negative predictive value, (+) / (-) LR = positive / negative likelihood ratio

Lost to follow up 29 cases (22%) and 102 controls (49%) were lost to follow up prior to their first birthdays. The points of last contact are graphically illustrated in Figure 13.

Figure 13. Age at last clinical contact of infants lost to follow up.

birth hospital discharge 6 weeks 6 months 1 year

115 106 131 102 CASES 16

130 117 208 ------106 CONTROLS 78 13 11

LOST to FOLLOW UP 94 22 15 131 138 Table 55 compares the clinical characteristics of the case and control populations completing follow up with those lost to follow up.

Table 55. Clinical characteristics of the case and control populations comparing those who completed 12 month follow up with those lost to follow up.

NE cases Controls characteristic followed lost to fu Chi 2 followed lost to fu Chi 2 (%nlQ2) (% n29) (p value) (%nl06) (%nl0 2 ) (p value) cold season 53 (52) 14(48) 0.12(0.7) 63(60) 45(44) 4.9(0.03)* inner city 63 (62) 13 (45) 2 .6 6 (0 .1) 95(90) 66(65) 18.(0.00)* residence no maternal 42 (41) 18(62) 3.9 (0.04)* 42(40) 54(53) 3.7(0.054) education short birth 4(4) 3(10) 1.84 (0.17) 9(8) 9(9) 0.01 (0.93) interval fetal distress 51(50) 15 (52) 0.03 (0.87) 6 (6 ) 9(9) 0.78 (0.38) non cephalic 9(9) 3(10) 0.06 (0.80) 1 (1) KD 0 (0.98) presentation instrumental/ 37 (36) 8(28) 0.76 (0.39) 17(16) 14(14) 0.22 (0.64) operative del. twin 4(4) 1(3) 0.01 (0.90) KD 0 (0 ) 0.97 (0.33) apgar^ <5 6 8 (67) 20 (69) 0.05 (0.82) 0 (0 ) 0 (0 ) low 21 (2 1 ) 7(24) 0.17 (0.68) 23(22) 26 (25) 0.42 (0.52) birthweight male 62(61) 18 (62) 0.02 (0.90) 53(50) 58 (57) 0.98 (0.32) grade 2/3 67(66) 16 (55) 1.0(0.29) Note: * = significant at p < 0.05

Amongst cases the infants of uneducated women were less likely to attend but there was no significant excess of moderate/severe NE amongst defaulters. Amongst controls defaulters were more likely to have been bom during the warm season and to reside out 139 of town. There was no gender bias in those attending for follow up. We know from the case control study that seasonality is not related to the risk of encephalopathy but that maternal lack of education is an important risk factor for encephalopathy. Generalising it is possible that this factor may adversely affect the longer term outcome of NE. Therefore those lost to follow up may be at risk of worse outcomes than the majority followed up and reported here. To investigate this further we examined the partial follow up data amongst our defaulters. The proportion impaired at the last point of contact of those lost to follow is shown in Table 56.

Table 56. Proportion impaired at last point of clinical contact by NE grade and age at last point of contact.

Hosp.discharge Six weeks Six months Total impaired/total impaired/total impaired/total (%) (%) (%) (%)______Ne 1 1/8(12) 0/4(0) 0/1(0) 1/13(8) Ne 2 6/8(76) 2/5(40) 1/2(50) 9/15(60) Ne 3______0/0 (0)______0/0 (0)______1/1 (100) 1/1 (100) Controls 0/78(0) 0/13 0/11 102

11 controls were last seen at 6 months when all were clinically normal whilst of four cases last seen at 6 months 2 were clinically normal (one grade 1, one grade 2 ) and two (SN 109 and 318) had evidence of major impairment. Their cases are described below. Study No 109 Female, NE grade 2, low birth weight (2.35 kg). No evidence of infection/subarachnoid haemorrhage on lumbar puncture day 2 (colourless CSF, 2-4 red cells per high power field), borderline enlarged ventricles (Ventricular Index 13.5/14.5 mms) on early (first week) ultrasound scan, discharged home feeding well but still hypotonic. At 6 weeks of age noted to have an an enlarging occipito-frontal circumference (OFC) of 38.9 cms (90th centile at 6 weeks having been 3rd-10th centile at birth). Clinically there were signs of evolving hydrocephalus with sutural diastasis, a full fontanelle and mild venous dilatation but the infant was breastfeeding well and developing normally. Cranial ultrasound demonstrated ventriculomegaly. At early (3 month) recall there was arrest of excessive head growth with an OFC of 41.0 cm (90th centile) and no clinical change. At six months his mother thought he could see and hear and had witnessed no fits.On examination there were signs of significant neurological impairment with persisting assymetric tonic neck reflex and assymetry of spontaneous movement with right hand fisting and a poverty of spontaneous movement. In summary there was evidence of spontaneously arrested hydrocephalus with an evolving right hemiplegia. 140

Study No 318 Male, NE grade 3, Low Birth Weight (2.2 kg), at discharge still not sucking well, hypotonic and lethargic. At six weeks not sucking (spoon fed), failing to thrive (3.5 kg), with marked axial hypotonia. At six months grossly malnourished (3.7 kg) with severe motor delay.

13 controls were last seen at 6 weeks and all were neurodevelopmentally normal. Nine cases were last seen at 6 weeks when seven were neurodevelopmentally normal and two were not. Their clinical details are described below.

Study No 14 Male, NE grade 2, birthweight 3.2 kg, at discharge feeding poorly and hypertonic. At six weeks bilateral fisting, opisthotonus, poor visual gaze and gross failure to thrive with a weight of 2.7 kg. Hospital admission declined (impoverished carpet worker, unaccompanied by husband) and early follow up at 3 months offered. When revisited had left her rented room, neighbours no knowledge of outcome.

Study No 185 Female, NE grade 2,birthweight 2.6 kg. At discharge lethargic with poor suck and hypotonia. At 6 weeks failing to thrive (2.9 kg) with severe neurodevelopmental abnormaities including persisting bulbar palsy, opisthotonus and absent visual tracking.

78 controls were last seen at hospital discharge when all were clinically normal. 16 cases were last seen at hospital dischaige.Seven of the eight infants with grade one NE had apparently made a full recovery and one only had persisting isolated hypotonia. Eight infants with grade two NE were last seen at discharge. Two had apparently made a full recovery , three had persisting signs of isolated hypotonia, one also had an abnormal suck and two had evidence of abnormal tone, suck and persisting lethargy. Therefore extrapolating from the partial follow up data there is no good evidence to suggest any of the 13 grade ones would have gone on to manifest major impairment at one year. In contrast the single grade three infant is likely to have died. Of the 7 grade two infants followed upto 6 weeks three were severely impaired of whom it seems reasonable to predict one surviving with serious impairment and two deaths. Of the eight not seen since hospital discharge it seems likely that at least three would have major impairment or death as outcomes given their markedly abnormal condition. Of the three with isolated hypotonia and the two with no abnormal neurological signs I would predict minor impairment at worst by one year of age. In summary based on the incomplete clinical data available of the 15 grade two NE infants lost to follow up I would predict 2 (13%) died, 4 (27%) were seriously impaired and 9 (60%) either had minor impairment or were normal. These outcome figures compare favourably with those described above for those successfully followed to one year. Therefore it is unlikely that the outcomes reported at one year significantly underestimate the neurodevelopmental sequelae of NE in this population. Estimated contribution of birth asphyxia to disability During the eighteen month period of surveillance and recruitment of NE cases there were 21,609 livebirths in the maternity hospital. Of these 131 developed neonatal encephalopathy as defined. Of the 102 successfully followed to one year 45 had died, 18 had major impairment and 39 were either normal or minimally impaired. Of the 29 lost to follow up on the available evidence I predict 3 deaths, 4 major impairments and 22 normal or minimally impaired. Therefore I estimate of 21,609 infants bom alive some 2 2 survived to one year with major impairment resulting in functional disability, an estimated prevalence of disability attributable to birth asphyxia at one year of age of 1 per 1000 birth cohort. 142

7. Implications for interventions in low income countries Principal findings

Prevalence Study Reliability of the grading ofNE This study shows neonatal encephalopathy to be a recognisable clinical syndrome which can be reliably graded by junior medical officers in an under resourced Kathmandu hospital with good inter-rater agreement (overall kappa value 0.87). Our operationally defined grading scheme demonstrated a high degree of congruence when compared (blind) with a recognised HIE grading system. New medical staff acquired the necessary expertise to reliably perform the neurological assessment within a month of starting their paediatric attachment and once trained could complete an assessment in approximately five minutes. Birth prevalence and severity ofNE In this population the birth prevalence of NE was 6.4 per 1000 live births (95% Cl 5.2- 7.8). Restricting the denominator only to term infants (estimated to be 93% of this population) the prevalence of NE was 6.9 per 1000 live term births (95% Cl 5.6-8.4). Of 92 NE infants 55 (60%) had moderate or severe encephalopathy. In comparison a recent UK study found the prevalence of NE to be 2 per 1000 total births^219\ Reported estimates of the proportion of encephalopathic neonates in high income countries who are moderately or severely affected vary from 33%-47%<277). Neonatal encephalopathy, fresh stillbirth and perinatal death This study shows the birth prevalence of fresh stillbirth and neonatal encephalopathy amongst term newborns in a low income country urban hospital (after the exclusion of all cases due to recognised non-asphyxial causes) to be 8.5 per 1000 total births and 6.4 per 1000 livebirths respectively. Early neonatal mortality resulting from NE was high in this setting with a 33% mortality risk overall. The cause specific perinatal mortality due to birth asphyxia in term infants was 10.8 per 1000 total births, 24% of the overall perinatal mortality. In contrast recent figures for the UK give a combined figure for perinatal death due to intrapartum asphyxia/trauma of 1.1 per 1000 total births, approximately 10% of overall perinatal mortality <22°). One minute Apgar score and neonatal encephalopathy A widely used definition of severe birth asphyxia based on an Apgar score at one minute of three or less identified a population eight times greater in magnitude than the population of term infants manifesting early neonatal encephalopathy not attributable to non-asphyxial causes. The vast majority (92%) of NE cases had an Apgar score at one minute of three or less and all had an Apgar score at one minute of 7 or less. 143 Perinatal events, NE and fresh stillbirth There was evidence of intrapartum compromise in 63 ( 6 8 %) of the NE cases and 65 (76%) of the stillbirths but only in 12 (12%) of the controls. In contrast only 31 of 164 (19%) NE cases (defined more broadly) and 2 of 400 (0.5%) controls in a recent Perth study had evidence of possible intrapartum compromise^167). Risk factors for NE and fresh stillbirth Antenatal care. Mothers who had not previously attended for antenatal care were at increased risk of delivering an encephalopathic infant ( adjusted OR 2.48 [Cl 1.30-4.76]) but no protective effect of antenatal care for the outcome of fresh stillbirth was detected. Syntocinon induction and augmentation Infants whose delivery was induced using syntocinon were at significantly greater risk ofNE than controls (adjusted OR 8 .8 [Cl 3.1-25.1]). Syntocinon induction also carried a significantly increased risk of fresh stillbirth in this setting (adjusted OR 5.0 [Cl 1.6- 16.1]). The use of syntocinon for augmentation was significantly associated with encephalopathy (adjusted OR 4.41 [Cl 2.48-7.82]) but since this is a common first line management response to intrapartum difficulties in this setting the nature of this association is difficult to interpret. Syntocinon for augmentation was not significantly associated with risk of stillbirth (adjusted OR 1.0 [Cl 0.4-2.5]) but the small numbers (10 [9%]) of mothers arriving at the hospital in time for this strategy to be implemented limit the significance of this observation. Maternal deficiency states This study found no significant association between NE or fresh stillbirth and maternal hypothyroidism or maternal hypomagnesaemia. There was an increased risk of fresh stillbirth ( adjusted OR 2.46 [Cl 1.09-5.54]) associated with severe postpartum anaemia (< 8 g/dl). Acute phase proteins Alpha-1 Acid Glycoprotein and C Reactive Protein levels were raised in mothers who had suffered adverse pregnancy outcomes compared with normal contemporaneous controls. Both remained significantly associated with NE when adjusted to allow for potential confounding by prolonged rupture of membranes, mode of delivery and time of sampling. Outcome study Death In Kathmandu neonatal encephalopathy of any severity carried an increased risk of death by one year. Compared to the control group the relative risk (95% Cl) of death in infancy was 5 (1.4-15) for grade one, 8 (3-23) for grade two and 26 (10-67) for grade three NE. There was no statistically significant difference in mortality risk for the post 144 neonatal period (1-12 months of age) between survivors of NE and normal controls. The terminal neonatal event was ventilatory failure presumed due to brain stem involvement in the hypoxic ischaemic process in the majority (80%) of cases. Meconium aspiration made an important contribution in 17 (42%) deaths. Late onset sepsis may have been an important factor in 4 (10%) of deaths. Neurodevelopmental impairment Moderate encephalopathy carried a high risk of major impairment for survivors with 16/27 affected (59% [95% Cl 39-78]). Mild encephalopathy was associated with major impairment in 2 of the 35 surviving infants ( 6 % [95% Cl 1-19]) but one of these cases had atypical impairment suggestive of a non asphyxial cause. There was no excess risk of minor impairment at one year following encephalopathy in this population. Overall one year outcome The proportion dead or severely impaired following moderate HIE was 27/38 (71% [95% Cl 54 -84]) and following severe NE was 28/29 (97% [95% Cl 82-99]). In comparison a review of five large studies in high income countries found the proportion dead or severely abnormal following moderate HIE to be between 15 and 25% and following severe HIE to be between 50 and 100% (277). The Kathmandu data suggests NE has a worse outcome at all grades of severity in this low income setting compared to high income countries. The contribution of birth asphyxia to childhood disability The prevalence of childhood disability at one year of age following birth asphyxia in this population is no more than 1 per 1000 birth cohort. In comparison a recent study from Oxford estimated the contribution of birth asphyxia to cerebral palsy in a UK population by prospectively assessing the outcome at 5 years of age of 160 term newborns with evidence of intrapartum asphyxia drawn from a birth cohort of 1 1 ,1 0 0 (113). They found three to have major neurodevelopmental impairment whilst from the same birth cohort a further 27 with no evidence of intrapartum asphyxia developed cerebral palsy. They estimated that the frequency of cerebral palsy associated with birth asphyxia was one in 3700 full-term births (0.27 per 1000). In Oxford birth asphyxia at term was associated with 10% [95% Cl 2.1-26.5] of all cases of cerebral palsy and with 2 0 % [95% Cl 4.3-48.1] of the 15 cases of cerebral palsy in children bom at term. Low cost predictors of major impairment following NE For infants with encephalopathy neurological abnormality at discharge was the most sensitive predictor of impairment at one year, closely followed by severity of NE. Cystic change seen on cranial ultrasound examination at 6 weeks was the most specific predictor, closely followed by acquired microcephaly at 6 weeks of age. 145 Limitations Recruitment bias associated with a hospital based study The study groups were drawn from the population of women delivering in a major public maternity hospital as opposed to a geographically defined population. A community enrolled cohort would have been preferable on methodological grounds but because NE is a relatively rare outcome (even in this setting) sample size considerations rendered this option impractical. A community based study of equivalent power would require the provision of community coverage for over half of the million residents of the Kathmandu Valley. The Maternal and Infant Research Activities (MIRA) Group have recruited a geographically defined population of pregnant Kathmandu Valley residents from two representative areas and compared the characteristics of women who deliver in hospital to those who deliver at home^278^. A baseline survey of over 6000 households identified 357 pregnant women. The delivery place of 334/357 (94%) of the pregnant women identified at the survey was determined. 272 (81%) had an institutional delivery and 62 (19%) delivered at home. On univariate analysis comparing home and institutional deliverers, maternal education, parity, and poverty indicators (income, size of house, ownership of house) were associated with place of delivery. After multivariate analysis, low maternal educational level (no education, OR 5.04 [95% Cl 1.61-15.8], class 1-10, OR 3.36 [95% Cl 1.04-10.8] compared to those with higher education) and multiparity (OR 3.1 [95% Cl 1.63-5.74] compared to primiparity) were significant risk factors for a home delivery. Of home deliverers, only 24% used a traditional birth attendant, and over half were actually unplanned deliveries due to precipitate labour or lack of transport. This population data and the residential subgroup analysis in chapter 6 suggest two distinct recruitment biases may affect this study's estimates of prevalence of adverse pregnancy outcome. Considering the population living close to the hospital (within the ring road) the omission of home deliverers excluded a group tending to be of lower educational attainment and higher parity whose labours are more likely to be precipitate. Their omission from the study population is likely to have led me to underestimate the true prevalence of both NE and fresh stillbirth in the Kathmandu Valley. Turning to the population living further from the hospital (outside the ring road) women whose labour is proceeding normally are less likely to present to the hospital for delivery. Those presenting in labour to the hospital are more likely to have complicated labours culminating in adverse outcome. Their inclusion in the study population is likely to have led me to overestimate the true prevalence of both NE and fresh stillbirth in the Kathmandu Valley. 146 These two sources of recruitment bias are therefore likely to have opposite effects on prevalence estimates. In an attempt to quantify their net effect I assume that adverse outcome is twice as common in home deliverers compared with hospital deliverers (based on recent comparative PNMR data from the Kathmandu valley^186)). Assuming one fifth of the local population deliver at home over the twelve month prevalence study period 35 infants with NE and 52 fresh stillbirths may have been omitted by the first recruitment bias. Conversely although approximately 20% of controls lived outside the ring road approximately 40% of NE and 50% SB cases were bom to families outside the ring road. This suggests 37 excess NE cases and 63 excess fresh stillbirths may have been included due to the second recruitment bias. I conclude that the two selection biases are likely to cancel each other out as regards prevalence estimates of both NE and SB. By repeating the risk factor analysis for the subgroups of within ring road residents and planned hospital deliverers I have attempted to allow for these sources of recruitment bias in the case control study. Regarding the follow up study it is likely that home deliverers have a worse early neonatal prognosis than that reported here. This issue is discussed more fully when considering the generalisability of the study findings. Ascertainment bias Our surveillance system was designed to detect all cases of encephalopathy occurring in a large maternity hospital where virtually all women stay with their infants for the first 24 hours after delivery. It is possible that a small number of mildly encephalopathic infants escaped our attention but moderate and severe encephalopathy is not a subtle syndrome and following birth asphyxia always presents within 24 hours of birth (107). Control recruitment The problems encountered due to the initially biased control recruitment and the modifications to study design this necessitated have been extensively discussed above. The solution of repeating control recruitment introduced a secular recruitment bias of 12 months during a period of notable political and economic stasis in Nepal. There were no important changes in health service delivery or obstetric practice in Kathmandu over this period. Background data It would have been useful to have birthweight data for the whole study baseline population but the resource implications of entering data on 15,000 infants on the study database proved prohibitive. A complete contemporaneous record of all encephalopathic infants excluded from the study would have improved its comparability with ongoing encephalopathy research. It would also have been helpful to have more detailed clinical data for all perinatal deaths occurring during the project period to place the proportion due to birth asphyxia in a more informative context. 147

Investigations The restricted laboratory and imaging services available in this resource-poor setting prevented documentation of the characteristic multi-organ hypoxic-ischaemic damage which is thought to accompany asphyxial NE^. However a recent study in a high income centre of excellence failed to identify multi-organ damage in a third of moderate or severe cases of NE suggesting this should not be viewed as a necessary criteria for case identification ^ 13^. It is difficult to estimate the false negative rate of microbiological culture techniques in this setting. The fact that raised white cells were not found in the vast majority of cerebrospinal fluid microscopic examinations supports the negative culture results. But it remains likely that a small proportion of infected infants were missed by the investigative techniques available. Post mortem pathology and placental pathology data would have added to the study but post mortems are not performed in this cultural setting and there is no local expertise in placental examination. These limitations reflect the challenges of investigating perinatal health in a low income setting.

Validity of fresh SB and NE as a measure of birth asphyxia. What proportion of fresh stillbirth and neonatal encephalopathy is due to birth asphyxia? As we have no method of measuring asphyxia the answer to these questions is currently uncertain even in the most sophisticated clinical settings. Due to late presentation in labour and poor fetal monitoring it is not feasible to separate antepartum and intrapartum fresh stillbirth in the challenging circumstances described here. The Wigglesworth classification of perinatal death was developed for situations like Kathmandu where sophisticated investigative techniques and autopsy findings are not available. Several studies have attempted to describe the causes of stillbirth in a more detailed manner. A recent descriptive study in Kenya of 212 stillbirths concluded that 123 (85%) of the 145 fresh non-dysmorphic stillbirths were attributable to birth asphyxia^279). A prospective UK study attempted to validate the Wigglesworth classification using pathological findings as a gold standard. Of 239 stillbirths 30 were classified by the Wigglesworth criteria as fresh term perinatal deaths due to birth asphyxia^280). In only 3 of these 30 cases (10%) was this classification revised in the light of full necropsy and histological findings. Therefore on the available evidence in a high income setting the Wigglesworth classification seems moderately robust. However it most be acknowledged that our current understanding of placental failure is incomplete and it maybe that in the future we will gain a better understanding of the mechanisms underlying fresh stillbirth. There are a number of potential causes of NE which have been suggested^114\ Neurological abnormality resulting from birth trauma is in practical terms difficult to distinguish from hypoxic-ischaemic encephalopathy and indeed often co-occurs^4). Infection remains an important cause of encephalopathy and as I have suggested above it is possible that a small proportion of affected cases in this study were the direct result of unrecognised perinatal sepsis. The more subtle hypothesised mechanism of cerebral injury resulting from cytokine toxicity is currently the focus of much research interest as summarised in chapter 1. It is possible that subclinical maternal infection leading to intrapartum compromise, fresh stillbirth and neonatal encephalopathy affected a proportion of the cases described. The relative importance of such an effect is currently speculative. On the available evidence I favour an alternative hypothesis that cytokines represent one aspect of the final common pathway triggered by a variety of mechanisms including acute hypoxic-ischaemia resulting in cerebral injury. Once hypoglycaemia and dysmorphic syndromes are excluded, as in this study, the remaining causes are rare. Structural brain malformations, amongst which neuronal migration defects are currently the best recognised, are often not detected by cranial ultrasound, especially if corpus callosum involvement is subtle. The emerging evidence from neuroimaging studies of children with cerebral palsy in high income settings suggest that the majority of children with cerebral malformations present later in infancy and childhood with no evidence of encephalopathy in the perinatal period^281\ Perinatal strokes are usually detectable with cranial ultrasound and are rare (best estimates are 0.1-0.2 per 1000). Finally metabolic disorders which may present with neonatal encephalopathy are no more than 0.2 per 1000 in the UK (J. Leonard - personal communication). Consanguineal marriage is culturally proscribed in all the ethnic groups of the Kathmandu Valley and metabolic encephalopathy is therefore unlikely to be any more prevalent in Nepal. I have shown in this population that even with the poor monitoring data at our disposal the proportion of NE cases satisfying the Perth criteria for intrapartum hypoxia was 42/92 (46%). When other evidence of intrapartum compromise is added this proportion rises to 63/96 ( 6 8 %). The prevalence estimate presented here is an upper limit for birth asphyxia in this setting. In my view it approximates to the real prevalence since as Nelson has said there is a 'characteristic evolution of signs that enable the diagnosis of encephalopathy due to perinatal asphyxia to be fairly safe retrospectively' I114). Maternal blood assays Maternal blood samples were taken in the puerperium as it was impractical to prospectively collect and store serum in the antepartum period from such a large baseline population of women. The unexpected degree of sample degradation during on site refrigerated storage further complicated the case control analysis. By using 149 contemporaneous control samples (thereby avoiding the bias introduced by different storage times) the principal causal hypotheses were tested. Maternal anthropometry Maternal anthropometry in the puerperium is complicated by the weight gain associated with pregnancy which can take several months to resolve. Maternal height is independent of pregnancy and and is therefore a robust measure of maternal stature.

Infant anthropometry Infant head circumference was measured on admission to SCBU in the first hours of life in cases whilst in controls it was measured typically at 24 hours of age. Measurement in the first hours may be complicated by molding and soft tissue trauma and this may have systematically biased the occipito-frontal circumference data. Raised intracranial pressure secondary to cerebral oedema is a secondary phenomenon in asphyxiated infants^33) and is unlikely to explain the increased head circumferences seen in the first hours of life in NE cases. Follow up period Follow up was demanding in the challenging circumstances described here and for logistical reasons was therefore limited to 12 months at the study design stage. Finer and Robertson report the longest follow up study of asphyxiated term infants in Canada with 77% follow up rate to age S^64*. They found the only neuroimpairment presenting de novo after the first year of life was an excess of educational difficulties associated with significantly lower IQ (10 points lower than controls) amongst otherwise unimpaired survivors of grade 2 NE. The Kathmandu study was designed only to detect functionally important major neurodisability. Subtle biological signals are obscured by environmental noise even in high income settings ^282^ and their detection in a low income setting (where to continue the analogy the 'noise is louder') is all the more problematic. Culture ^Dd.PeyiQdeyglQpmep.tal ,aggeggm and Japan (285* have shown local population differences in the rate of gross motor development in early childhood and in language acquisition in the preschool years. Both areas it can and has been argued are affected by marked cultural differences in child rearing practices (286). However Ueda's reliability testing of over 1000 Japanese children in an equivalent age range to that of Frankenberg showed no difference of more than two months in either the ages when 50% or 90% achieved any test item relevant to the assessment of the normal child at 12 months of age. The Bayley 2 has been used recently as part of a neurodevelopmental follow up study on a group of 43 infants at 10-20 months of age who had delivered at Prasuti Griha hospital, Kathmandu during 1995 (287^ and who therefore constitute a sample of this study's baseline population. All had been bom at term with birthweights in excess of 3.0 kg and had an uncomplicated perinatal course. They were drawn randomly from a tight geographically defined cohort from the traditional Newar satellite town of Kirtipur. This constitutes only one of the population groups who deliver at Prasuti Griha. At the time of assessment they were considered to be in good health and had no clinical evidence of neurological impairment. Their mean MDI was 90.7 (SD 11) and mean PDI was 96.8 (SD 11). The Nepali principal investigator of this study (a Newar mother herself) takes the view that these developmental quotients represent a real difference in skill maturity under the artificial conditions of the test in part related to the average Nepali child's unfamiliarity with the Western style consumer toys and bricks which are central to the Bayley test procedure (Manandhar N.- personal communication). However this small study assessed only one of the subgroups of our target population and therefore it is difficult to generalise to the wider study population. I hypothesise that the rather narrow standard deviations for MDI and PDI generated by this study reflect the culturally homogeneity of the sub-population studied. There are difficulties deriving new cutoff scores from this small study since the Bayley norms are based on raw scores adjusted for chronological age. Population based norms are derived from the standard deviations of raw scores across several age bands. The relatively small number of subjects in this Nepal based study precludes such analysis because of insufficient data. Therefore we elected to adopt the American cut-offs accepting that these may marginally overestimate abnormality.

Generalisability Place of delivery I have argued above that the study population is broadly representative of the surrounding urban population, which is in turn a representative poor South Asian conurbation. But only 8% of births occur in health facilities in Nepal^1"). Perinatal mortality rates have been shown to be approximately twice as high amongst home delivering populations when compared with neighbouring hospital delivering populations in Nepal although this relationship is confounded by significant socioeconomic differences between rural and urban populations^186^. Few studies have attempted to describe perinatal morbidity amongst low income home delivering Asian 151 populations. Kumar and colleagues based in Chandighar, North India (a setting with many cultural and economic features in common with rural Nepal) performed a prospective community-based study of birth asphyxia assessing early outcome for almost 2,000 livebirths in 54 villages^183). The perinatal mortality rate was 62 per 1000. The study attempted to quantify retrospectively the prevalence and early sequelae of birth asphyxia using home interviews. The rather inclusive folk definition of birth asphyxia adopted by the investigators added to the inclusion of premature infants is likely to have over estimated birth asphyxia compared with the more rigorous clinical definition of NE in term infants available to hospital investigators. This group report a prevalence for birth asphyxiated infants bom alive of 20 per 1000 livebirths with an early mortality of 74% . Their data suggests birth asphyxia related neonatal mortality rates after at home may be substantially higher (14 per 1000) than our data from a large maternity hospital (2.2 per 1000). Interestingly however the absolute number of infants surviving the early neonatal period following birth asphyxia (6 per 1000 in the North India home deliverers and 4 per 1000 in the Kathmandu hospital population) are similar. Therefore attempting to generalise from the Kathmandu data to rural South Asia birth asphyxia related perinatal mortality rates may be higher but the proportion of livebirths surviving neonatal encephalopathy is likely to be relatively constant. Obstetric care Prasuti Griha is the national training centre for obstetric care in the Kingdom of Nepal and obstetric practices found there are typical of the government sector in Nepal as a whole. The case control study findings are likely to be generalisable to equivalent institutions elsewhere in South Asia where more rigorous schemes for the monitoring of labour, such as partography, have not been instituted. As shown by the WHO study set in South East Asia discussed in chapter 2 <214) the introduction of partography may have an important effect on fresh stillbirth rates although an effect on NE as an outcome remains to be demonstrated. Infant care Major neurodevelopmental sequelae associated with birth asphyxia are the result of severe cerebral injury and have proven refractory to therapeutic intervention in the best child development centres. These major impairments are biologically caused and little modulated by cultural or social factors. Therefore the outcome findings of this study relating to major neurodevelopmental impairment are probably generalisable to populations who have equivalent perinatal experiences elsewhere in South Asia. Africa Comparing regional estimates of perinatal mortality rates Asia, with 53 per 1000, and Africa, with 75 per 1000, are in a league of their own O 8 4) £)0es the Kathmandu data 152 shed any light on birth asphyxia in Africa? Interestingly both the Nigerian and Zimbabwean studies described in chapter 2 found a significantly higher birth prevalence of moderate and severe NE than in Kathmandu. Both studies were set in referral hospitals which may have importantly biased their prevalence estimates. Also more intensive obstetric monitoring may have improved outcome shifting some otherwise stillborn infants into the encephalopathic live birth group. In the absence of contemporaneous stillbirth data this possible confounding factor cannot be assessed. Could the higher reported prevalence estimates in Africa reflect a real difference with Asian populations? Indirect support for this comes from a study of the aetiology of severe learning difficulties (developmental quotient < 50) in children attending a Cape Town child development centre (288>. Although not population based the proportions of children with learning difficulties from different ethnic groups were representative of the surrounding population with 150 coloured, 75 black and 7 white children. Following retrospective case analysis the investigators concluded that 26 full term infants had acquired brain injury as a result of birth asphyxia. Of these 7 were coloured and 19 were black. Clearly this data is likely to be importantly confounded by differential access to obstetric care but it again points to a higher prevalence of birth asphyxia amongst black Africans than amongst Asians. A possible biological factor of potential relevance is the tendency to larger size (in all dimensions) of African neonates compared with Asian neonates. This observation along with the huge cultural and economic differences between the two continents suggests caution should be exercised when generalising from this Asian dataset to Africa . Implications

Prevalence study The relative contribution of asphyxia to stillbirth and NE The only comparable study to our own combining cause-specific estimates of neonatal morbidity and mortality due to birth asphyxia was a retrospective study comparing moderate and severe NE with intrapartum stillbirth rates among infants of 35 weeks or greater gestation in two districts of the UK between 1982 and 1984 (218>. The investigators reported post-asphyxial stillbirth to be 0.4-0.8 per 1000 total births for the two districts and the prevalence of moderate/severe NE to be 0.6-1.9 per 1000 livebirths. The equivalent data from our study is 8.5 per 1000 and 3.8 per 1000 respectively. In the UK an infant is more likely to be bom with serious neonatal neurological morbidity than to be stillborn following asphyxia. Conversely in Kathmandu fresh stillbirth is a more likely outcome than moderate or severe encephalopathy. In rural areas, in the absence of interventional obstetric care, this trend is likely to even more marked. 153 Use of early Apgar scores as selection criteria for special care This study shows that early Apgar scores overestimate the scale of the problem of significant birth asphyxia as indicated by neonatal encephalopathy but remain a useful screening test in busy units. The Kathmandu data suggests that an admission policy for special care selecting all infants bom alive with a one minute Apgar of three or less will select approximately 5% of all deliveries and will include more than 90% of all cases of NE. Indicators of birth asphyxia 79% (126/159) of the perinatal deaths resulting from birth asphyxia in this study were fresh stillbirths with a birthweight of 2.0 kg or more. This group of births is relatively easy to measure requiring no more from a birth attendant than simple observation and access to a set of weighing scales. Conversely to identify infants with NE and to exclude other causes even in the basic manner attempted in this study requires at the very least an inpatient facility with basic laboratory services and medical staff trained to identify and grade NE. Although the grading scheme adopted for this study is relatively straightforward it will only achieve widespread usage at the level of the district hospital. Health planners' more urgent requirements include practical outcome indicators with which to monitor intervention programmes. Ideally the Wigglesworth perinatal audit scheme provides important outcome information reflecting aspects of both antenatal and perinatal care. However where resources aTe limited this study shows that the proportion of all births with a birthweight of 2.0 kg or more which culminate in a fresh stillbirth is a practical indicator of the problem of birth asphyxia amongst term births where perinatal mortality rates remain high. Risk factor study Induction The most striking potentially preventable risk factor for adverse outcome found during this analysis was induction of delivery with an adjusted odds ratio for the outcome of NE of 8.8 (95% Cl 3.1-25.1) and for SB of 5.0 [Cl 1.6-16.1]. A total of 12 NE cases (9%) compared to 29 controls (5%) were actually induced. The calculated population attributable fraction of NE following induced delivery in this population is 5%. This assumes there is a causal relationship underlying the observed association. Does induction in Kathmandu cause birth asphyxia or is the obstetrician more likely to induce delivery in a subgroup of pregnancies who are more prone to adverse outcome for other reasons? Certainly according to the American College of Obstetricians and Gynecologists "induction of labour is undertaken when, in the opinion of the physician, the risks of delivery to the mother, to the fetus or both are less than the risks of continuing pregnancy "(289). In the classical model of labour, 'the passenger' and 'the passage' are fixed variables leaving 'the powers' or uterine force as the only variable 154 open to manipulation. This can be increased using uterine stimulants such as oxytocin or by artificial rupture of membranes by amniotomy. The few well designed clinical trials so far conducted show little tangible benefit from either intervention^290^. Common indications for induction in Kathmandu are postmaturity and pre eclampsia when amniotomy is not routinely performed. Intravenous syntocinon infusion is commenced at 2 mU per minute titrated upto 15 mU per minute (K. Malla - personal communication). The fact that induction in Kathmandu remains a significant risk factor for adverse outcome after adjustment for important co-variables suggests there may be an independent causal effect. Hypertonic uterine contraction, a dose dependent side effect of syntocinon reduces placental perfusion and hence oxygen supply to the fetus. Amongst the study control population induced labour was likely to be more protracted (mean 19 hours) than spontaneous labour (mean 14 hours). Although fetal monitoring was recorded more frequently during induced labours (mean 10 times) than during spontaneous labours (mean three times) there was no systematic monitoring of uterine contractions. It is therefore biologically plausible that the apparent association between induction and adverse pregnancy outcome is in part causal. The Jamaican Perinatal Mortality Survey described a significant association (Odds Ratio 2.3) between induction and intrapartum death <290. The Perth group^165^ have previously reported an odds ratio of 2.3 (1.2-4.6) for NE associated with induction on univariate analysis. On multivariate analysis of a larger cohort this group however found an adjusted odds ratio to be non significant at 0.97 (0.57-1.68). Similarly no association between induction and early neonatal seizures was found during the Cardiff Births Survey (OR 1.1 [0.5-2.1])<87>. A study of elective induction for postmaturity in comparison with conservative management utilising serial antenatal monitoring has demonstrated no difference in perinatal mortality or neonatal morbidity between the two groups in the late 1980's in Canada^292). A retrospective study of induced deliveries in a low risk term population demonstrated no difference in neonatal morbidity but an increased proportion of operative and instrumental deliveries in the induced group^293^. A recent study from Jerusalem assessed the intrapartum complication rate for three different methods of induction amongst a high risk population (n=210) in a teaching hospital setting. This group show an increased intra-partum complication rate (fetal heart rate changes, meconium, cord prolapse, amnionitis and ) in the oxytocin group (adjusted OR 5.1 [95% Cl 1.5-17.5]) compared with the amniotomy group. This difference was not seen when the prostaglandin group was compared with the amniotomy group (adjusted OR 0.7 [95% Cl 0.3-1.6]). Elective induction of labor although commonly practiced in high income countries remains poorly studied and generally discouraged^294^. Even in high income countries labour induction is performed more often in the least capable facilities with no consistent management protocols^295). A recent review comments that induction of labour should only be countenanced "when personnel are comfortable with the emergency measures to treat uterine hypertonus and resulting fetal compromise because medical induction of labor carries with it more potential for uterine hyper stimulation than spontaneous labor"^296^. I hypothesise that labour induction is safe where obstetric care is optimal but poses a significant added risk to the fetus where, as in Kathmandu, fetal and uterine monitoring is rudimentary. Antenatal care Complete absence of any antenatal care was a significant risk factor for N E . Although one or two attendances were protective for NE (but not for SB) there appeared to be no incremental benefit from repeat attendances beyond this minimal level. The safe motherhood policy of the government of Nepal calls for a minimum of three antenatal visits during pregnancy focusing on risk screening, tetanus toxoid immunisation, anaemia prophylaxis and treatment and health and nutrition education. The protective effect of one or two antenatal visits seen in this study suggests some positive benefit but the mechanism remains unclear. Maternal anaemia A recent systematic review of routine iron supplementation failed to demonstrate appreciable benefit to either the mother or the fetus in the populations studied to date(i68). However the treatment of severe anaemia in women presenting in late pregnancy is widely considered to be worthwhile and current management guidelines at the Kathmandu maternity hospital recommend blood transfusion for mothers with a haemoglobin level less than 8 g/dl. This cut-off is supported indirectly by the postpartum haemoglobin data reported in this study. This study provides no support for routine antenatal supplementation designed to optimise thyroid or magnesium status. Risk assessment The use of antenatal risk assessment schemes to guide referral policy is commonplace in developing countries yet the evidence base to support their use is sparse. Table 57 shows a set of simple community-based criteria for the identification of high risk mothers currently advocated in India^297^. This study provides support for the inclusion of low maternal height, previous perinatal death, non-cephalic presentation and severe anaemia as important independent risk factors for adverse pregnancy outcome from the infant’s perspective. In this study maternal age and primiparity were not independently associated with adverse fetal/neonatal outcome. The remaining risk factors were not systematically investigated due to inadequate antepartum clinical data. 156 Table 57. Criteria for the identification of high risk mothers. primi gravida young (<18 years) or old (>35 years) short (<145 cm) or light (<40 kg) mother history of two or more previous fetal/neonatal deaths history of previous low birthweight baby chronic systemic disease severe anaemia pregnancy induced hypertension

Occult maternal infection This study in common with previous studies of adverse intrapartum outcome has described an association with maternal risk factors for infection. Postpartum assay of acute phase proteins in maternal serum were raised in both NE and SB cases compared with controls. There is much current interest in the possibility of perinatal cytokine mediated cerebral injury. Does this data support the theory? Unfortunately the maternal perinatal data in this study is limited and acute phase proteins may be triggered by a wide range of inflammatory processes. This study was not designed to test this hypothesis and although the data is in line with current thinking it would constitute over-interpretation to suggest it either supported or contradicted the cytokine hypothesis. Newborn anthropometry and maternal nutrition A recent cross cultural survey of beliefs surrounding childbirth added weight to widespread anecdotal reports that the restriction of staple foods in the third trimester is common amongst women "to avoid a big baby and a difficult delivery" (203). An anthropologist working in Nepal was explicitly told by hill women that they smoked cigarettes and worked hard during pregnancy to keep birthweights down and facilitate delivery ^298^. This study of adverse perinatal outcome like many of its predecessors demonstrates lower risk as birthweight increases. No excess risk of either NE or stillbirth was seen with high birthweight presumably because the heaviest infant bom enrolled to the study weighed only 4 kg. Intriguingly this study suggests there may be a higher risk of NE with increasing head circumference at birth. Head circumference at birth is associated with maternal nutritional status <299) but dietary intervention in the third trimester has been shown recently in The Gambia to cause only a 3 mm increase in head circumference and resulted in a significant reduction in perinatal mortality(30°). In conclusion there appears to be little objective evidence to support this widely held folk belief. Dietary restriction in pregnancy should be consigned to the relatively small group of harmful traditional practices. Outcome study Excess early neonatal mortality Our data strongly suggests NE has a worse outcome at all grades of severity in this low income setting compared to industrialised countries. Early neonatal death was importantly associated with the meconium aspiration syndrome. This lends indirect support to the developing case for the routine use of intrapartum amnioinfusion where labour is complicated by the passage of fresh meconium in utero. In the current absence of effective interventions for hypoxic-ischaemic encephalopathy resuscitation and early neonatal care should focus on the prevention of this complication. The relatively new technique of intrapartum amnioinfusion represents an appropriate intervention of great potential benefit which deserves further investigation. Neurodevelopmental sequelae Generalising from this study it seems likely that the contribution of birth asphyxia to neurodevelopmental impairment in low income settings has been overestimated 085) WHO estimate 840,000 infants suffer brain damage globally every year following birth asphyxia. The United Nations Population Division estimates there were 132 million births in 1996*301). This denominator generates a crude WHO estimate of over 6 brain damaged children per 1000 birth cohort in contrast to the 1 per 1000 estimated by this study. Historical experience in industrialised countries has shown that the transformation in maternal and perinatal death rates effected by modem obstetric care practices has had no equivalent effect on childhood disability. This is because the vast majority of neurodevelopmental impairment in childhood is not associated with perinatal events. The largest prospective cohort study (the NCPP) designed to assess the contribution of perinatal events to childhood impairment in an industrialised setting found just 21% of all cerebral palsy in their cohort of children by school age to be associated with adverse perinatal factors O28). This was in a population of nearly 40,000 with a CP prevalence of 3.1/1000. Equivalent population based cohort studies have not been attempted in developing countries for a variety of reasons. The best CP data in developing countries has been collected from carefully designed cross sectional surveys using clinical validation of self reported disability^302). In Bangladesh, a South Asian low income country, this methodology has described a prevalence for serious physical impairment in school aged children of 3.8/1000 (ranging between 3.6/1000 in urban areas and 4/1000 in rural areas). Assuming this overall prevalence of physical disability is generalisable to South Asia as a whole and applying it as a denominator to our Kathmandu cohort would predict that neonatal encephalopathy is associated with no more than 25% of childhood neurodevelopmental impairment in this setting. Future research priorities in low income settings Bi.rthAsphma apd childhood disability Further epidemiological research is ideally needed to compare these findings with the outcome of a large home delivering population in order to test the hypothesis generated by this study that birth asphyxia has been overstated as a cause of childhood disability in low income countries. However there are significant sample size difficulties inherent in such a project. Perhaps the question may be most practically explored as a retrospective population based descriptive study of childhood disability. I have performed a non population based pilot study of children presenting to a cerebral palsy centre in Kathmandu^303). I found mothers could recall their children's condition in the first day of life sufficiently well to give clear answers to questions seeking the presence or absence of feeding difficulties, floppiness or fits. Of 97 consecutively presenting children with cerebral palsy mothers recalled first day evidence of NE (indicated by two of these three cardinal signs) in 27 cases (28%). This suggests a similar population based study would be possible, perhaps as part of a wider investigation of preventable causes of childhood disability in South Asia.

Antepartum care and birth asphyxia This study has identified several simple risk indicators for birth asphyxia. Most are already incorporated into risk assessment schemes but the addition of more inclusive criteria (previous low birth weight infant and primiparity for example) limit the predictive power of such schemes. In rural areas where antenatal care consists of only one or two visits to a primary health centre for the vast majority of women and only a small minority will reach a delivery centre an important function of the antenatal visit is to identify those mothers at maximal risk. This requires rigorous testing of the predictive power of simple risk assessment measures in low income settings. Intrapartum care The use of syntocinon for the induction of labour in the absence of partography and cardiotocographic monitoring needs urgent reassessment. This would be a suitable topic for audit through the newly established South Asian network for perinatal audit. Partography has been shown to be effective in reducing perinatal mortality in tertiary centres and now requires testing at the district hospital level. Fetal monitoring remains a major challenge; the use of hand held Doppler to monitor fetal heart beat discussed in chapter two deserves wider dissemination and assessment. The most promising intrapartum intervention judging from the results of recent Southern Africa trials is intrapartum amnioinfusion in women with fresh meconium staining of the amniotic 159 fluid. It will be important to test the generalisability of these findings in the context of South Asia. Neonatal interventions The search for effective neuroprotective strategies to limit the extent of cerebral hypoxic-ischaemic injury continues to drive basic scientific research in this field in high income settings. Paradoxically hypoxic-ischaemic injury in term infants in these settings is becoming increasingly rare (Wyatt J.- personal communication). Attempts to identify neuroprotective agents capable of interrupting the cascade of subcellular events culminating in cell death have so far proved disappointing^19^. In general terms neuroprotective effects are not evident at low dosage levels and high dosage levels of the various agents employed to date have proved toxic to other physiological systems. Cost effective agents of relevance to low income settings remain a distant proposition. Animal studies have shown that moderate cooling of the brain is a potentially useful strategy for neural rescue treatment23\ Clinical trials are now being planned employing a simple water flow system to induce selective head cooling. In the medium term such a strategy may prove feasible in developing country centres. Designers of multicentre trials would be well advised to liaise with colleagues in low income settings at an early stage both to maximise recruitment and to ensure widespread applicability of this developing technology.

Broader intervention programmes Safe Motherhood initiatives are predicated on ready access to facilities capable of delivering effective interventions, including as the bottom line operative delivery. Low cost midwife operated birthing centres with ready access to more intensive obstetric units are being developed by some programmes. Their evaluation might usefully include the proportion of pregnancies resulting in fresh stillbirths with a birthweight of 2000 g or more as an outcome indicator for birth asphyxia. I hypothesise this measure will fall with better antenatal coverage and targeting of women for more intensive intrapartum monitoring with earlier operative intervention in selected cases. These strategies compliment the primary aim of Safe Motherhood to reduce .

Conclusions Fresh stillbirth and neonatal encephalopathy are relatively common amongst term births in a Kathmandu hospital resulting in perinatal death in over 1% of the population delivered there. The majority of these perinatal deaths occurred before birth. There was evidence of intrapartum compromise in more than half of the deliveries resulting in encephalopathy and three quarters of those ending in fresh stillbirth. Induction of delivery with syntocinon was a significant independent risk factor for both outcomes but explained only 5% of the cases of neonatal encephalopathy and a smaller proportion of fresh stillbirths. Antenatal care as currently practiced was marginally protective only for encephalopathy. Indirect evidence supported vigorous treatment of severe anaemia but neither maternal thyroid status nor magnesium status were found to be significant risk factors. This study failed to identify a simple and affordable public health measure likely to have a major impact on the prevalence of birth asphyxia. Neonatal encephalopathy in this setting tends to be more severe with a worse outcome than in high income countries. The observed clinical association between meconium aspiration syndrome, encephalopathy and neonatal death suggests a possible intervention point. Closer attention to the management of intrapartum meconium both in the uterus prior to delivery and in the newborn airway immediately after delivery may help to improve outcome. The expected excess of neurodevelopmental impairment accruing from low resource perinatal care proved to be surprisingly modest. This study has shown that for the vast majority of infants in poor countries the outcome of a difficult delivery is perinatal death or normal survival. Improvements in obstetric care in low income countries are unlikely to have a major impact on the prevalence of childhood disability. This is in line with epidemiological data from the recent past in high income countries. In high income countries attention has shifted from primary to secondary prevention of asphyxial brain injury with the evaluation of neuroprotective strategies^174). This study serves as a timely reminder that rescue strategies designed to interrupt the ampliflcatory cascade of pathological events which are set in train by acute hypoxic-ischaemia must aim sufficiently upstream to avoid the risk of simply transforming death into impaired survival. Researchers would be well advised to involve South Asian neonatologists at this early stage of research and development of brain cooling equipment. The historical record of falling maternal and perinatal mortality in high income countries reflects in part social change and improvements in women's health and in part improvements in obstetric and perinatal health services^304). In the long term enlightened education programmes designed to maximise the schooling of girls remain the single most important intervention available to government. In the medium term a complete maternal and child health package is required to significantly improve pregnancy outcomes in developing countries. In the short term inappropriate resourcing of neonatal intensive care in the absence of commensurate improvements to obstetric care may result in an increase in absolute numbers of impaired infant survivors. Most deaths due to birth asphyxia in South Asia occur during the hours leading up to delivery. Available resources are best aimed at the extension of obstetric services to reach more precisely targeted risk groups at the outset of labour. This will require better linkage between the primary sector and maternity institutions, imaginative organisation of those institutions and the development of appropriate intrapartum protocols. Safe Motherhood initiatives, as advocated by the World Health Organisation, provide an ideal vehicle for future intervention studies. 161

Bibliography

1. Peebles DM, Edwards AD, Wyatt JS, Cope M, Delpy DT, Reynold EO. Changes in human fetal cerebral haemoglobin concentration and oxygenation during labour measured by near-infrared spectroscopy. Am J Obstet Gynecol 1992;166:1369- 73.

2. Dear PRF. Hypoxia and the neonatal brain. In: Meadow R, ed. Recent Advances in Paediatrics 8. Edinburgh: Churchill Livingstone, 1986. 140-155.

3. Myers RE. The clinical and pathological effects of asphyxiation in the fetal rhesus monkey. In: Adamsons K, ed. Diagnosis and treatment of fetal disorders. New York: Springer-Verlag, 1968. 226-249.

4. Fenichel JM. Neonatal Neurology. New York: Churchill Livingstone, 1990.

5. Perlman JM, Tack ED, Martin T, Shackelford MD, Amon E. Acute systemic organ injuiy in term infants after asphyxia. Am J Dis Child 1989;143:617-20.

6. Primhak RA, Jedeikin R, Elli G, Mahela SK, Rowe RD. Myocardial ischaemia in asphyxia neonatorum. Acta Paediatr Scand 1985;7 4:595-600.

7. Jayashree G, Dutta AK, Sama MS, Saili A. Acute renal failure in asphyxiated newborns. Indian Pediatr 1991;28:19-23.

8. Saili A, Sama MS, Gathwala G, Kumari S, Dutta AK. dysfunction in severe birth asphyxia. Indian Pediatr 1990;27:1291-1294.

9. Shankaran S, Woldt E, Koepke T, Bedard MP, Nandyal R. Acute neonatal morbidity and longterm CNS sequelae in term infants. Early Hum Dev 1991 ;2 5:135- 48.

10. Volpe JJ. Hypoxic-ischaemic encephalopathy: biochemical and physiological aspects. In: Volpe J, ed. Neurology of the Newborn. Boston: WB Saunders, 1995. 211-259.

11. Pryds O, Edwards AD. Cerebral blood flow in the newborn infant. Arch Dis Child 1996;7 4:F63-F69. 162 12. Hill A. Current concepts of hypoxic-ischaemic injury in the term newborn. Paediatr Neurol 1991 ;7:317-25.

13. Volpe JJ. Hypoxic-ischaemic encephalopathyrneuropathology and pathogenesis. In: Volpe J, ed. Neurology of the Newborn. Boston: WB Saunders, 1995. 279-313.

14. Vannucci RC, Duffy TE. Cerebral metabolism in newborn dogs during acute hypoxia and recovery. Ann Neurol 1977;1:528-534.

15. Vannucci RC. Cerebral carbohydrate and energy metabolism in perinatal hypoxic-ischaemic brain damage. Brain Pathol 1992;2:229-234.

16. Vannucci RC, Mardis EE, Vannucci SJ. Cerebral carbohydrate and energy metabolism during hypoglycaemia in newborn dogs. Am J Physiol 1981 ;240:192.

17. Hossman KA. Viability thresholds and the penumbra of focal ischaemia. Ann Neurol 1994;36:557-563.

18. Wyatt JS, Edwards AD, Azzopardi D, Reynolds EOR. Magnetic resonance and near infra red spectroscopy for investigation of perinatal hypoxic-ischaemic brain injury. Arch Dis Child 1989;64:953-963.

19. Vannucci RC, Perlman JM. Interventions for perinatal hypoxic-ischaemic encephalopathy. Pediatrics 1997;100:1004-1014.

20. Clark GD. Role of excitatory amino acids in brain injury. Clin Perinatol 1989;16:458-467.

21. Mehmet H, Edwards AD. Hypoxia, ischaemia, and apoptosis. Arch Dis Child 1996;7 5:F73-F75.

22. Thorensen M, Penrice J, Lorek A, Cady EB, Wylezinska M, Kirkbride V, et al. Mild hypothermia after severe hypoxia-ischemia ameliorates delayed energy failure in the newborn piglet. Pediatr Res 1995;37:667-670.

23. Edwards AD, Wyatt JS, Thoresen M. Treatment of hypoxic-ischaemic brain damage by moderate hypothermia. Arch Dis Child 1998;7 8:F85-F91. 163 24. Brann AW, Myers RE. Central nervous system findings in the newborn monkey following severe in utero partial asphyxia. Neurology 1975;25:327-338.

25. Nicolaides KH, Economides DL, Soothill PW. Blood gases and pH and lactate in appropriate and small for gestational age fetuses. Am J Obstet Gynecol 1989;161:996-1001.

26. Economides DL, Proudler A, Nicolaides KH. Plasma insulin in appropriate and small for gestational age fetuses. Am J Obstet Gynecol 1989;160:1091-1094.

27. Tyrrell S, Obaid AH, Lilford RJ. Umbilical artery Doppler velocimetry as a predictor of fetal hypoxia and acidosis at birth. Obstet Gynecol 1989;74:332-7.

28. Montenegro N, Santos F, Tavares E, Matias A, Barros H, Leite LP. Outcome of 88 pregnancies with absent or reversed end-diastolic blood flow (ARED flow) in the umbilical arteries. Eur J Obstet Gynecol Reprod Biol 1998;7 9:43-6.

29. Salafia CM, Pezzullo JC, Minior VK, Divon MY. Placental pathology of absent and reversed end-diastolic flow in growth-restricted fetuses. Obstet Gynecol 1997;90:830-6.

30. Karsdorp VH, van Vugt JM, van Geijn HP, Kostense PJ, Arduini D, Montenegro N, et al. Clinical significance of absent or reversed end diastolic velocity waveforms in umbilical artery. Lancet 1994;17:1664-8.

31. Mayer PL, Kier EL. The controversy of the periventricular white matter circulation: a review of the anatomic literature. Am J Neuroradiol 1991;12:223-228.

32. Kuban CK. White-matter disease of prematurity, periventricular leukomalacia, and ischemic lesions. Dev Med Child Neurol 1998;40:571-573.

33. Levene MI, Evans DH, Forde A, Archer LNJ. The value of intracranial pressure monitoring in asphyxiated newborn infants. Dev Med Child Neurol 1987;29:311-319.

34. Volpe JJ. Hypoxic-ischaemic encephalopathy: clinical aspects. In: Volpe J, ed. Neurology of the Newborn. Boston: WB Saunders, 1995. 314-369. 164 35. Wigglesworth JS. Monitoring perinatal mortality - a pathophysiological approach. Lancet 1980;2:684-686.

36. Scott H. Outcome of very severe birth asphyxia. Arch Dis Child 1976;51:712- 716.

37. Meis PJ, Hall M, Marshall JR. Meconium Passage: A new classification for risk assessment during labor. Am J Obstet Gynecol 1978;131:509-513.

38. Redline RW. Placental pathology: a neglected link between basic disease mechanisms and untoward pregnancy outcome. Curr Opin Obstet Gynecol 1995;7:10- 15.

39. Wiswell TE, Bent RC. Meconium staining and the meconium aspiration syndrome. Pediatr Clin North Am 1993;40:955-981.

40. Boo NY, Goon HK. Epidemiology of necrotising enterocolitis in Malaysian neonates. Singapore Med J1989;30:444-448.

41. McKay JG, Hermansen MC, Maley BE. Diminished splenic function in asphyxiated term infants. JPerinatol 1990;10:12-15.

42. Andrew M, Brodovich H, Mitchell L. Fetal lamb coagulation system during birth asphyxia. AmJHematol 1988;28:201-203.

43. Castle V, Andrew M, Kelton J, Giron D, Johnston M, Carter C. Frequency and mechanism of neonatal thrombocytopenia. J Pediatr 1986;108:749-755.

44. Samat HB, Samat MS. Neonatal encephalopathy following fetal distress. Arch Neurol 1976;3 3:696-705.

45. Gaur U, Saili A. The immunological profile in asphyxiated newborns. National Neonatal Forum 1996; 19:3-4.

46. Antia-Obong OE, Utsalo SJ, Udo JJ, Udo KT. Neonatal septicaemia in Calabar, Nigeria. Cent Afr J Med 1992;38:161-165.

47. Berseth CL, McCoy HH. Birth asphyxia alters neonatal intestinal motility in term neonates. Pediatrics. 1992;90:669-673. 165 48. Anderson S, Shakya KN, Shrestha LN, Costello AM de L. Hypoglycaemia: a common problem among uncomplicated newborn infants in Nepal. J Trop Paeds 1993;39:273-277.

49. Ellis M, Manandar N, Shakya U, Manandar DS, Fawdry A, Costello AM de L. Postnatal hypothermia and cold stress among newborn infants in Nepal monitored by continuous ambulatory recording. Arch Dis Child 1996;75:F42-F45.

50. Jalil F, Lindblad BS, Hanson LA, Khan SR, Yaqoob M, Karlberg J. Early child health in Lahore, Chapter 9. Perinatal events. ActaPaediatr 1993;3 90 suppl 1:95-107.

51. Costello AM de L. Perinatal health in developing countries. Trans R Soc Trop Med Hyg 1993;87:1-2.

52. Davies P. Speaking the same language. Dev Med Child Neurol 1994;36:189- 90.

53. Paneth N, Stark RI. Cerebral palsy and mental retardation in relation to indicators of perinatal asphyxia. Am J Obstet Gynecol 1983;147(8):960-966.

54. Robertson CMT, Finer NN. Term infants with hypoxic-ischaemic encephalopathy: Outcome at 3.5 years. Dev Med Child Neurol 1985;27:473-484.

55. D’Souza SW, McCartney E, Nolan M, Taylor IG. Hearing, speech and language in survivors of severe perinatal asphyxia. Arch Dis Child 1981;56:245-52.

56. Eken P, de Vries LS, van der Graaf Y, Meiners LC, van Nieuwenhuizun O. Haemorrhagic-ischaemic lesions of the neonatal brain: correlation between cerebral visual impairment, neurodevelopmental outcome and MRI in infancy. Dev Med Child Neurol 1995;37:41-55.

57. Hagberg B, Hagberg G, Olow I. The changing panorama of cerebral palsy in Sweden 1954-1970. Acta Paediatr Scand 1975;6 4:193-200.

58. Brown JK. Disorders of the central nervous system. In: Campbell AGM, McIntosh N, ed. Forfar andArneil's Textbook of Paediatrics. 4th ed. Edinburgh: Churchill Livingstone, 1992. 802-820. 166 59. Powell TG, Pharoah POD, Cooke RWI, Rosenbloom L. Cerebral palsy in low- birthweight infants. 1. Spastic hemiplegia: associations with intra-partum stress. Dev Med Child Neurol 1988;30:11-18.

60. Mulligan JC, Painter MJ, O'Donoghue PA, Macdonald HM, Allen AC, Taylor PM. Neonatal Asphyxia 2: Neonatal mortality and long term sequelae. J Pediatr 1980;96:903-907.

61. Rosenbloom L. Dyskinetic cerebral palsy and birth asphyxia. Dev Med Child Neurol 1994;36:285-89.

62. Gaffney G, Flavell V, Johnson A, Squier M, Sellers S. Cerebral palsy and neonatal encephalopathy. Arch Dis Child 1994;7 0:FI95-200.

63. Paneth N, Kiely J. The Frequency of Cerebral Palsy: A review of population studies in industrialised nations since 1950. In: Stanley FJ, Alberman E, ed. The epidemiology of the cerebral palsies. Clinics in Developmental Medicine Oxford: 87. Blackwell Scientific, 1984.46-56.

64. Robertson CMT, Finer NN, Grace MGA. School performance of survivors of neonatal encephalopathy associated with birth asphyxia at term. J Pediatr 1989;114:753-760.

65. Nelson KB, Ellenberg JH. Asymptomatic newborn and risk of cerebral palsy. Am J Dis Child 1987;141:1333-1335.

66. Blair E. A research definition for birth asphyxia? Dev Med Child Neurol 1993;35:449-455.

67. Little WJ. On the influence of abnormal parturition, difficult labours, premature birth, and asphyxia neonatorum, on the mental and physical condition of the child, especially in relation to deformities. Transactions of the Obstetric Society of London 1862;3:293-344.

68. Lilienfield AM, Parkhurst E. A study of the association of factors of pregnancy and parturition with the development of cerebral palsy: a preliminary report. Am J Hyg 1951;19 51(53):262-8. 69. Apgar V. A proposal for a new method of evaluation of the newborn infant. AnaesthAnalg 1953;3 2:260.

70. Mane AV, Daga SR. Changing profile of birth asphyxia. Indian Pediatr 1997;33:326-328.

71. Donovan DE, Coues P, Paine RS. Prognostic implications of neurogical abnormalities in the newborn period. Neurology 1962;12:910.

72. Prechtl HER, Beintema DJ. The neurological examination of the full term newborn infant. Clinics in developmental medicine London: 12. Heinemann, 1964.

73. Brazelton TB. Neonatal behavioural assessment scale. Clinics in developmental medicine 88. London: Heinemann, 1984.

74. Nelson KB, Ellenberg J. Obstetrical complications as risk factors for cerebral palsy or seizure disorders. JAMA 1984;251:1843-1848.

75. Nelson KB, Dambrosia JM, Ting TY, Grether JK. Uncertain value of electronic fetal monitoring in predicting cerebral palsy. N Eng J Med 1996;334:613-618.

76. Freeman JM, Nelson KB. Intrapartum asphyxia and cerebral palsy. Pediatrics 1988;82:240-249.

77. Menticoglou SM, Manning FA, Harman CR, Morrison I. Severe fetal brain injury without evident intrapartum asphyxia or trauma. Obstet Gynecol 1989;7 4:457- 61.

78. Mann LI. Pregnancy events and brain damage. Am J Obstet Gynecol 1986;155:6-9.

79. Prechtl HFR. The optimality concept. Early Hum Dev 1980;4:201-5.

80. Touwen BCL, Huisjes HJ, Jurgens van der Zee AD, Bierman-van Eendenburg ME, Smrkovsky M, Olinga AA. Obstetrical condition and neonatal neurological morbidity. An analysis with the help of the optimality concept. Early Hum Dev 1980;4:207-228. 81. Drage JS, Kennedy C, Schwarz BK. The Apgar score as an index of neonatal mortality. Obstet Gynecol 1964;2 4:222.

82. Niswander KR, Gordon M. The women and their pregnancies. In: Collaborative Perinatal Study of the National Institute of Neurological Diseases and Stroke. Philadelphia: WB Saunders, 1972.73-379.

83. Nelson KB, Ellenberg JK. Apgar scores as predictors of chronic neurological disability. Pediatrics 1981;68:36-44.

84. Bergman I, Painter MJ, Hirsch RP, Crumrine PK, David R. Outcome in neonates with convulsions treated in an intensive care unit. Ann Neurol 1983;14:642- 647.

85. Macdonald D, Grant A, Sheridan-Pereira M, Boylan P, Chalmers I. The Dublin randomised controlled trial of intrapartum fetal heart rate monitoring. Am J Obstet Gynecol 1985;152:524-39.

86. Curtis PD, Matthews TG, Clarke TA, Darling M, Crowley P, Griffin E, et al. Neonatal seizures: the Dublin Collaborative Study. Arch Dis Child 1988;63:1065-68.

87. Minchom P, Niswander K, Chalmers I, Dauncey M, Newcombe R, Elboume D, et al. Antecedents and outcome of very early neonatal seizures in infants bom at or after term. Br J Obstet Gynaecol 1987;9 4:431-39.

88. Connel J, Oozer R, DeVries L, Dubowitz LMS, Dubowitz V. Continuous EEG monitoring of neonatal seizures: diagnostic and prognostic considerations. Arch Dis Child 1989;64:452-458.

89. Ellenberg JH, Nelson KB. Cluster of perinatal events identifying infants at high risk for death or disability. J Pediatr 1988;113:546-552.

90. Brown JK, Purvis RJ, Forfar JO, Cockbum F. Neurological aspects of perinatal asphyxia. Dev Med Child Neurol 1974;16:567-580.

91. Amiel-Tison C. Cerebral damage in full-term newborn. Aetiological factors, neonatal status and longterm follow-up. Biol Neonate 1969;14:234-250. 169 92. Thom I. Cerebral symptoms in the newborn. Diagnostic and prognostic significance of symptoms of presumed cerebral origin. Acta Paediatr Scand 1969;195 suppl 1:1-113.

93. Ziegler AL, Calame A, Marchand C, Passera M, Reymond-Gonil, Prod'hom LS. Cerebral distress in full-term newborns and its prognostic value. Helv Paediat Acta 1976;31:299-317.

94. De Souza SW, Richards B. Neurological sequelae in newborn babies after perinatal asphyxia. Arch Dis Child 1978;53:564-69.

95. Fitzhardinge PM, Flodmark O, Fitz CR, Ashby S. The prognostic value of CT as an adjunct to assessment of the term infant with postasphyxial encephalopathy. J Pediatr 1981;99:777-781.

96. Finer NN, Robertson CM, Richards RT, Pinnell LE, Peters KL. Hypoxic- ischaemic encephalopathy in term neonates: Perinatal factors and outcome. J Pediatr 1981;98:112-117.

97. Lipp-Zwahlen A, Deonna T, Micheli J, Calame A. Prognostic value of neonatal CT scan in asphyxiated term babies: low density score compared with neonatal neurological signs. Neuropediatrics 1985;16:209-17.

98. Low JA, Galbraith RS, Muir DW, Killen HL, Pater EA, Karchmar EJ. The relationship between perinatal hypoxia and newborn encephalopathy. Am J Obstet Gynecol 1985;152:256-60.

99. Levene MI, Komberg J, Williams THC. The incidence and severity of post­ asphyxial encephalopathy in full term infants. Early Hum Dev 1985;11:21-26.

100. Lipper EG, Voorhies TM, Ross G, Vannucci RC, Auld PAM. Early predictors of one-year outcome for infants asphyxiated at birth. Dev Med Child Neurol 1986;28:303-309.

101. Archer LNJ, Levene MI, Evans DH. Cerebral artery Doppler ultrasonography for prediction of outcome after perinatal asphyxia. Lancet 1986;1:67-69. 102. Fernandez F, Verdu A, Quero J, Perez-Higueras A. Serum CPK-BB isoenzyme in the assessment of brain damage in asphyctic term infants. Acta Paediatr Scand 1987;76:914-8.

103. Al-Alfy, Carroll JE, Devarajan LV, Moussa MAA. Term infant asphyxia in Kuwait. Ann Trop Paediatr 1990;10:355-361.

104. Airede Al. Birth asphyxia and hypoxic-ischaemic encephalopathy:incidence and severity. Ann Trop Paediatr 1991;11:331-335.

105. Gray PH, Tudehope DI, Masel JP, Bums YR, Mohay HA, O'Callaghan MJ, et al. Perinatal hypoxic-ischaemic brain injury: prediction of outcome. Dev Med Child Neurol 1993;35:965-973.

106. Thompson CM, Puterman AS, Linley LL, Hann FM, van der Elst CW, Molteno CD, et al. The value of a scoring system for hypoxic ischaemic encephalopathy in predicting neurodevelopmental outcome. Acta Paediatr 1997;86:757-61.

107. Fenichel JM. Hypoxic-ischaemic encephalopathy in the newborn. Arch Neurol 1983;40:261-6.

108. Leviton A, Nelson KB. Problems with definitions and classifications of newborn encephalopathy. Pediatr Neurol 1992;8:85-90.

109. Levene MI, Sands C, Grindulis H, Moore JR. Comparison of two methods of predicting outcome in perinatal Asphyxia. Lancet 1986;1:67-9.

110. Portman RJ, Carter BS, Gaylord MS, Murphy MG, Thieme RE, Merenstein GB. Predicting neonatal morbidity after perinatal asphyxia: a scoring system. Am J Obstet Gynecol 1990;162:174-181.

111. Amiel-Tison C, Ellison P. Birth Asphyxia in the fullterm newborn: early assessment and outcome. Dev Med Child Neurol 1986;28:671-682.

112. Carter BS, Haverkamp AD, Merenstein GB. The definition of acute perinatal asphyxia. Clinics in Perinatology 1993;20:287-304. 113. Yudkin PL, Johnson A, Clover LM, Murphy KW. Assessing the contribution of birth asphyxia to cerebral palsy in term singletons. Paediatr Perinat Epidem 1995;9:156-70.

114. Nelson KB, Leviton A. How much of neonatal encephalopathy is due to birth asphyxia? Am J Dis Child 1991;145:1325-1331.

115. Wyatt JS. Noninvasive assessment of cerebral oxidative metabolism in the human newborn. J R Coll Physicians Lond 1994;28:126-131.

116. Roth SC, Edwards AD, Cady EB, Delpy DT, Wyatt JS, Azzopardi D, et al. Relation between cerebral oxidative metabolism following birth asphyxia, neurodevelopmental outcome and brain growth at one year. Dev Med Child Neurol 1992;34:285-295.

117. Peliowski A, Finer NN. Birth asphyxia in the term infant. In: Sinclair JC, Bracken MB, ed. Effective care of the newborn infant. Oxford: Oxford University Press, 1992. 249-279.

118. Editorial. Birth Asphyxia: a statement. Dev Med Child Neurol 1993;3 5:1022- 24.

119. Badawi N, Kurinczuk JJ, Hall D, Field D, Pemberton PJ, Stanley FJ. Newborn encephlopathy in term infants: three approaches to population-based investigation. Semin Neonatol 1997;2:181-188.

120. Badawi N, Kurinczuk JJ, Keogh JM, Alessandri LM, O’Sullivan F, Burton PR, et al. Antepartum risk factors for newborn encephalopathy: the Western Australian case-control study. BMJ 1998;317:1549-53.

121. Krageloh-Mann I, Petersen D, Hagberg G, Vollmer B, Hagberg B, Michaelis R. Bilateral spastic cerebral palsy-MRI pathology and origin. Analysis from a representative series of 56 cases. Dev Med Child Neurol 1995;37:379-97.

122. Pharoah POD, Platt MJ, Cooke T. The changing epidemiology of cerebral palsy. Arch Dis Child 1996;75:F169-F173.

123. Laisram N, Srivastava VK, Srivastava RK. Cerebral palsy - an etiological study. Indian J Pediatr 1992;59:723-728. 172 124. Stanley FJ, Blair E. Cerebral Palsy. In: Pless IB, ed. The Epidemiology of Childhood Disorders. New York: Oxford University Press, 1994. 473-497.

125. Hall DMB. Intrapartum events and cerebral palsy. Br J Obstet Gynaecol 1994;101:745-7.

126. Williams K, Albeiman E. The impact of diagnostic labelling in population-based research into cerebral palsy. Developmental Medicine and Child Neurology 1998;40:182-185.

127. Badawi N, Watson L, Petterson B, Blair E, Slee J, Haan E, et al. What constitute cerebral palsy? Dev Med Child Neurol 1998;40:520-527.

128. Nelson KB. What proportion of cerebral palsy is related to birth asphyxia? [Editorial]. J Pediatr 1988;112:572-4.

129. Hagberg B, Hagberg G. The changing panorama of infantile hydrocephalus and cerebral palsy over forty years: a Swedish survey. Brain Dev 1989;11:368-73.

130. Pharoah POD, Cooke T, Rosenbloom L, Cooke RW1. Trends in the birth prevalence of cerebral palsy. Arch Dis Child 1987;62:379-82.

131. Takeshita K, Ando Y, Takashima S. Cerebral palsy in Tottori, Japan. Benefits and risks of progress in perinatal medicine. Neuroepidemiology 1989;19 89(8):184- 192.

132. Paneth N. Neonatal care and patterns of handicap in the community. In: Chard J, Richards MPM, ed. Obstetrics in the Nineties. Clinics in Developmental Medicine 123/124. Mac Keith Press, 1992. 232-241.

133. Hagberg B, Hagberg G, Olow I, Wendt LV. The changing panorama of cerebral palsy in Sweden 7. Prevalence and origin in the birth year period 1987-90. Acta Paediatr 1996;8 5:954-60.

134. Pharoah POD, Cooke T, Cooke RWI, Rosenbloom L. Birthweight specific trends in cerebral palsy. Arch Dis Child 1990;65:602-606. 173 135. Blair E, Stanley FJ. Intrauterine growth and spastic cerebral palsy. 1: Association with birth weight for gestational age. Am J Obstet Gynecol 1990;162:229- 237.

136. Brown MA. Pre-eclampsia: a case of nerves? Lancet 1997;349:297-298.

137. Gaffney G, Sellers S, Flavell V, Squier M, Johnson A. Case-control study of intrapartum care, cerebral palsy, and perinatal death. BMJ 1994;308:743-750.

138. Blair E, Palmer L, Stanley F. Does preeclampsia really protect against CP? Pediatrics 1997;97:780-2.

139. Collins M, Paneth N. Preeclampsia and cerebral palsy: are they related? Dev Med Child Neurol 1998;40:207-211.

140. Grether JK, Nelson KB, Emery ES, Cummins SK. Prenatal and perinatal factors and cerebral palsy in very low birth weight infants. J Pediatr 1996;128:407- 414.

141. Leviton A, Paneth N, Susser M, Reuss M, Allred E, Kuban K, et al. Maternal receipt of magnesium sulfate does not appear to reduce the risk of neonatal white matter damage. Pediatrics 1997;9 9:URL:http://www.pediatrics.org/cgi/content/full/99/4/e2.

142. Grether JK, Nelson KB. Maternal infection and cerebral palsy in infants of normal birth weight. JAMA 1997;27 8:207-211.

143. Adinolfl M. Infectious disease in pregnancy, cytokines and neurological impairment: an hypothesis. Dev Med Child Neurol 1993;35:549-553.

144. Dammann O, Leviton A. Maternal intrauterine infection, cytokines, and brain damage in the preterm newborn. Pediatr Res 1997;42:1-8.

145. Nelson KB, Dambrosia JM, Grether JK, Phillips TM. Neonatal cytokines and coagulation factors in children with cerebral palsy. Ann Neurol 1998;44:665-675.

146. Silverstein FS, Barks JD, Hagan P, Liu XH, Ivacko J, Szaflarski J. Cytokines and perinatal brain injury. Neurochem Int 1997;30:375-83, 174 147. Nelson KB, Dambrosia JM, Phillips TM, Grether JK. Cytokines and cerebral palsy. Neurology 1998;50:1 (abstract).

148. Olsen P, Paakko E, Vainionpaa L , Pyhtinen J, Jarvelin MR. Magnetic resonance imaging of periventricular leukomalacia and its clinical correlation in children. Annals of Neurology 1997;41:754-61.

149. Sugimoto T, Woo M, Nishida N, Araki A, Hara T, Yasuhara A, et al. When do brain abnormalities in cerebral palsy occur? an MRI study. Dev Med Child Neurol 1995;37:285-292.

150. Sibony O, Stempfle N, Luton D, Oury JF, Blot PH. In utero fetal cerebral intraparenchymal ischemia diagnosed by nuclear magnetic resonance. Dev Med Child Neurol 1998;40:122-123.

151. Scher MS, Belfar H, Martin J, Painter MJ. Destructive brain lesions of presumed fetal onset: antepartum causes of cerebral palsy. Pediatrics 1991;88:898-906.

152. Pharoah POD, Buttfield IH, Hetzel BS. Neurological damage to the fetus resulting from severe iodine deficiency during pregnancy. Lancet 1971;1:308-310.

153. Halpem JP, Boyages SC, Maberly GF, Collins JK, Eastman CJ, Morris JGL. The neurology of endemic cretinism; a study of two endemias. Brain 1991 ;114:825- 41.

154. Thorpe-Beeston JG, Nicolaides KH, Felton CV, Butler J, McGregor AM. Maturation of the secretion of thyroid hormone and thyroid-stimulating hormone in the fetus. N Engl J Med 1991;324:532-536.

155. Burrow GN, Fisher DA, Larsen PR. Maternal and fetal thyroid function. New Engl J Med 1994;331:1072-1077.

156. Bernal J, Nunez J. Thyroid hormones and brain development. Eur J Endocrinol 1995;133:390-398.

157. Nelson KB, Ellenberg JH. Antecedents of cerebral palsy . Univariate analysis of risks. Am J Dis Child 1985;139:1031-8. 175 158. Reus ML, Paneth N, Pinto-Martin JA, Lorenz JM, Susser M. Transient hypothyroxinaemia in preterm infants and neurodevelopment at age two years. N Engl J Med 1996;334:82-7.

159. Kuroda M, Grether JK, Nelson KB, Hoogstrate J. Neonatal hypothyroxinaemia and risk of cerebral palsy in term infants in a population-based study. Paediatr Perinat Epidem 1996; 10:A17.

160. Pharoah POD, Cooke T. Cerebral palsy and multiple births. Arch Dis Child 1996;7 5:F174-F177.

161. Petterson B, Nelson KB, Watson L, Stanley F. Twins, triplets, and cerebral palsy births in Western Australia in the 1980's. BMJ 1993;307:1239-43.

162. Grether JK, Nelson KB, Cummins SK. Twinning and cerebral palsy in four northern California counties: births 1983 through 1985. Pediatrics 1993;92:854-858.

163. Macdonald HM, Mulligan JC, Allen AC, Taylor PM. Neonatal Asphyxia. 1. Relationship of obstetric and neonatal complications to neonatal mortality in 38,405 consecutive deliveries. J Paediatr 1980;96:898-907.

164. Daga AS, Daga SR, Patole SK. Risk assessment in birth asphyxia. Journal of Tropical Pediatrics 1990;36:34-39.

165. Adamson SJ, Alessandri LM, Badawi N, Burton PR, Pemberton PJ, Stanley F. Predictors of neonatal encephalopathy in full term infants. BMJ 1995;311:598-602.

166. Chandra S, Ramji S, Thirupuram S. Perinatal asphyxia: multivariate analysis of risk factors in hospital births. Indian Pediatr 1997;34:206-212.

167. Badawi N, Kurinczuk JJ, Keogh JM, Alessandri LM, O'Sullivan F, Burton PR, et al. Intrapartum risk factors for newborn encephalopathy: the Western Australian case-control study. BMJ 1998;317:1554-8.

168. Mahomed K. Routine iron supplementation during pregnancy. In: Neilson JP, Crowther CA, Hodnett ED, Hofmeyr GJ, Keirse MJNC, ed. Pregnancy and childbirth module of the Cochrane database of systematic reviews. Oxford: Update Software, 1997. The Cochrane Library; Issue 3. 176 169. National Planning Commission. Nepal multiple indicator surveillance. Kathmandu: His Majesty's Government of Nepal, 1995.

170. Procianoy RS, Pereira DN. Effect of birth asphyxia on thyroid hormones in full term infants. Proceedings of the Pediatric Academic Societies Annual Meeting. New Orleans: 1998.

171. Harrison V, Peat G. Red blood cell magnesium and hypoxic- ischaemic encephalopathy. Early Hum Dev 1997;47:287-296.

172. lives P, Blennow M, Kutt E, Magi ML, Kudijavtseva G, Lagercrantz H, et al. Concentrations of magnesium and ionized calcium in umbilical cord blood in distressed term newborn infants with hypoxic ischaemic encephalopathy. Acta Paediatr. 1996;85:1348-50.

173. Committee on Fetus and Newborn of the American Academy of Pediatrics and Committee on Obstetric Practice of the American College of Obstetrics and Gynecology. Use and abuse of the Apgar score. Pediatrics 1996;98:141-2.

174. Wyatt JS, Thoresen M. Hypothermia treatment and the newborn. Pediatrics 1997;100:1028-1030.

175. Low JA. The significance of fetal asphyxia in regard to motor and cognitive deficits in infancy and childhood. In: Tejani J, ed. Obstetrical Events and Development Sequelae. Boca-Raton: CRC Press, 1990. 43-58.

176. Dennis J, Johnson A, Mutch L, Yudkin P, Johnson P. Acid-base status at birth and neurodevelopmental outcome at 4.5 years. Am J Obstet Gynecol 1989;161:213- 20.

177. Goodwin TM, Belai I, Hernandez P, Durand M, Paul RH. Asphyxial complications in the term newborn with severe umbilical acidaemia. Am J Obstet Gynecol 1992;167:1506-1512.

178. Patel J, Edwards AD. Prediction of outcome after perinatal asphyxia. Curr Opin Pediatr 1997;9:128-132. 177 179. Eken P, Toet MC, Groenendaal F, de Vries LS. Predictive value of early neuroimaging, pulsed doppler and neurophysiology in full term infants with hypoxic- ischaemic encephalopathy. Arch Dis Child 1995;7 3:F75-F80.

180. Harbord MG, Weston P. Somatosensory evoked potentials predict neurologic outcome in full-term neonates with asphyxia. J Paediatr Child Health 1995;31:148-51.

181. Pinto-Martin JA, Riolo S, Cnaan A, Holzman C, Susser MW, Paneth N. Cranial ultrasound prediction of disabling and nondisabling cerebral palsy at age two in a low birth weight population. Pediatrics 1995;95:249-254.

182. Kuenzle CH, Baenziger O, Martin E, Thun-Hohenstein L, Steinlin M, Good M, et al. Prognostic value of early MR imaging in term infants with severe perinatal asphyxia. Neuropediatrics 1994;25:191-200.

183. Kumar R. Birth asphyxia in a rural community of north India. J Trop Pediatr 1995;41:5-7.

184. World Health Organisation. Perinatal mortality: a listing of available information. WHO Geneva. 1996 WHO/FRH/MSM/96.7

185. World Health Organisation. Essential Newborn Care. WHO Geneva. 1996 WHO/FRH/MSM/96.13

186. Geetha T, Chenoy R, Stevens D, Johanson RB. A multicentre study of perinatal mortality in Nepal. Paediatr Perinat Epidem 1995;9:74-89.

187. Keeling JW. Perinatal mortality of mothers and children: the Jamaica studies. In: Baum D, ed. Nestle Nutrition Workshop Series New31. York: Raven Press, 1993. 25-33.

188. Raghuveer G. Perinatal deaths: relevance of Wigglesworth’s classification. Paediatr Perinat Epidem 1992;6:45-50.

189. Buchmann EJ, Crofton Briggs IG, McIntyre JA. Previous birth outcome of antenatal attenders in northern Kwazulu - perinatal and infant mortality rates. S Afr Med J 1992;81:419-421. 178 190. Bai NS, Mathews E, Nair PM, Sabarinathan K, Harikumar C. Perinatal mortality rate in a south Indian population. J Indian Med Assoc 1991;89:97-98.

191. Fauveau V, Wojtyniak B, Mostafa G, Sarder AM, Chakraborty J. Perinatal mortality in Matlab, Bangladesh: a community-based study. Int J Epidemiol 1990;19:606-612.

192. De Muylder X. Perinatal mortality audit in a Zimbabwean district. Paediatr Perinat Epidem 1989;3:284-293.

193. Singh M. Hospital based data on perinatal and neonatal mortality in India: special review article. Indian Paediatrics 1986;23:579-584.

194. Shah U, Pratinidhi AK, Bhatlawande PV. Perinatal mortality in rural India: a strategy for reduction through primary care 1: Stillbirths. J Epidemiol Community Health 1984;38:134-7.

195. Mati JK, Aggarwal VP, Lucas S, Sanghvi HC, Corkhill R. The Nairobi Birth Survey IV. Early perinatal mortality rate. J of Obstet Gynecol E Cent Afr 1983;2:129- 33.

196. Kinoti SN. Asphyxia of the newborn in East, Central and Southern Africa. East Afr MedJ 1993;70:422-33.

197. Kumari S, Sharma M, Yadav M, Saraf A, Kabra M, Mehra R. Trends in neonatal outcome with low Apgar scores. Indian J Paediatr 1993;60:415-22.

198. Nathoo KJ, Chimbra TH, Mtiavalye LA. Mortality and immediate morbidity in term babies with low Apgar scores. Ann Trop Paediatr 1990;10:239-244.

199. Pradhan A, Aryal RH, Regmi G, Ban B, Govindasamy P. Nepal family health survey 1996. Kathmandu, Nepal and Calverton, Maryland: Ministry of Health (Nepal), New Era and Macro International, 1997.

200. Verderese M de L, Turbull LM. The traditional birth attendant in maternal and child health and family planning: a guide to her training and utilisation. WHO Geneva. 1975 offset publication No. 18. 179 201. Levitt M. Traditional birth attendant training. In: Costello AM de L, Manandhar DS, ed. Improving health of the newborn infant in developing countries. London: Imperial Press, In press.

202. Levitt MJ. A systematic study of birth and traditional birth attendants in Nepal. Kathmandu: John Snow International. 1993 ASB-0135-C-00-1042-00.

203. Lefeber Y, Voorhoever H. Practices and beliefs of traditional birth attendants: lessons for obstetrics in the north? Trop Med Int Health 1997;2:1175-1179.

204. Alexander S, Keirse MJNC. Formal risk scoring during pregnancy. In: Chalmers I, Enkin M, Keirse MJNC, ed. Effective care in pregnancy and childbirth. Vol 1. Oxford: Oxford University Press, 1989. 345-365.

205. Smith M, Stratton WC, Roi L. Labour risk assessment in a rural community hospital. Am J Obstet Gynecol 1985;151:569-574.

206. Chandramohan D, Cutts F, Millard P. The effect of stay in a maternity home on perinatal mortality in rural Zimbabwe. J Trop Med Hyg 1995;98:261-7.

207. Timyan J, Griffey-Brechin SJ, Measham DM, Ogunleye B. Access to care: more than a problem of distance. In: Koblinsky M, Timyan J, Gay J, ed. The health of women. Oxford: Westview press, 1993. 217-235.

208. Raina N, Kumar V, Management of birth asphyxia by traditional birth attendants. World Health Forum 1989;20:243-246.

209. Xiaoyu Z. Neonatal resuscitation. World Health Forum 1993;1993(14):289- 290.

210. Kumar R. Effect of training on the reuscitation practices of traditional birth attendants. Trans R Soc Trop Med Hyg 1994;88:159-60.

211. Ramji S, Ahuja S, Thirupurum S, Rootwelt T, Rooth G, Saugstad OD. Resuscitation of asphyxic newborn infants with room air or 100% oxygen. Pediatr Res 1993;34:809-12. 180 212. Massawe A, Kilewo C, Irani S, Verma RJ, Chakrapam AB, Ribbe T, et al. Assessment of mouth-to-mask ventilation in resuscitation of asphyxic newborn babies. A pilot study. Trop Med Int Health 1996;1:865-73.

213. Mahomed K, Nyoni R, Mulambo T, Kasule J, Jacobus E. Randomised controlled trial of intrapartum fetal heart rate monitoring. BMJ 1994; 308:497-500.

214. Kwast BE, Lennox CE, Farley TMM. World Health organisation partograph in management of labour. Lancet 1994;343:1399-404.

215. Mahomed K, Mulambo T, Woelk G, Hofmeyr GJ, Gulmezoglu AM. The collaborative randomised amnioinfusion for meconium project (CRAMP): 2. Zimbabwe. Br J Obstet Gynaecol 1998;105:309-313.

216. Hofmeyer GJ. Amnioinfusion for meconium-stained liquor in labour (Cochrane Review). The Cochrane Library, Oxford: Update Software 1998;,

217. Thomberg E, Thiringer K, Odeback A, Milsom I. Birth asphyxia: incidence, clinical course and outcome in a Swedish population. Acta Paediatr 1995;84:927-932.

218. Hull J, Dodd KL. Falling incidence of hypoxic-ischaemic encephalopathy in term infants. Br J Obstet Gynaecol 1992;99:386-391.

219. Hall DMB, Evans K. Epidemiology of neonatal encephalopathy. Perinat Paediatr Epidemiol 1996; 10:A14.

220. Infancy. CEiSaDi. 3rd Annual Report. London: Department of Health, 1996.

221. Von Furer-Haimendorf C. Himalayan traders. London: John Murray, 1975.

222. Dixit H. The quest for health. Kathmandu: Educational Enterprise, 1995.

223. Ministry of Health. Mid term health review 2035. Kathmandu: HMG Nepal, 1979.

224. Justice J. Policies, plans and people. Berkeley: University of California Press, 1986. 181 225. Ministry of Health. Annual report of the department of health services. Kathmandu: Planning and foreign aid division, MOH, HMG, 1996.

226. World Bank. World Development Report. Oxford: Oxford University Press, 1996.

227. Unicef. The State of the World's Children. Oxford: Oxford University Press, 1997.

228. Economist Intelligence Unit. India and Nepal Country Profiles. London: The Economist, 1992.

229. Unicef and HMG Nepal. Children and Women of Nepal: a situational analysis. Kathmandu: Unicef and HMG, 1992.

230. Central Bureau of Statistics. The Analysis of the 1991 Population Census. Kathmandu: His Majesty’s Government of Nepal, 1993.

231. Bolam AJ, Manandhar DS, Shrestha P, Manandhar B, Ellis M, Costello AM de L. Maternity care utilisation in the Kathmandu Valley: a community based study. J Nep Med Assoc 1997;35:122-129.

232. Manandhar DS, Rajbhandari S, Pal D, Costello AM de L. Anthropometry of the term newborn and postnatal mother in Nepal. J Nep Med Assoc 1997;3 5:150-157.

233. Hellier JL, Goldstein H. The use of birthweight and gestation to assess perinatal mortality risk .J Epidemiol Community Health 1979;33:183-185.

234. Dubowitz LMS, Dubowitz V, Goldberg C. Clinical assessment of gestational age in the newborn infant. JPediatr 1970;7 7:1.

235. Parkin JM, Hey EN, Clowes JS. Rapid assessment of gestational age at birth. Arch Dis Child 1976;51:259-263.

236. Brueton MJ, Palit A, Prosser R. Gestational age assessment in Nigerian newborn infants. Arch Dis Child 1973;48:318. 182 237. Amiel-Tison C. Clinical assessment of the infant nervous system. In: Levene MI, Lilford RJ, Bennett MJ, Punt J, ed. Fetal and neonatal neurology and neurosurgery. Edinburgh: Churchill Livingstone, 1995. 83-104.

238. Volpe JJ. The neurological examination: normal and abnormal features. In: Volpe J, ed. Neurology of the Newborn. Boston: WB Saunders, 1995. 314-369.

239. Landis JR, Koch GG. A one-way components of variance model for categorical data. Biometrics 1977;33:671-679.

240. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159-74.

241. Reddy V, Shekar M, Rao P, Gillespie S. Nutrition in India. Hyderabad: National Institute of Nutrition, 1992.

242. Sachdev HP. Epidemiological trends in maternal and neonatal nutritional status. In: Costello AM de L, Manandhar DS, ed. Improving health of the newborn in developing countries. London: Imperial Press, In press.

243. Makarem A. Principles and Technics. In: Henry RJ, Cannon DC, Winkelman JW, ed. Clinical Chemistry. Hagerstown: Harper and Row, 1974.

244. Skeffington FS, Pearse RG. The 'bright brain'. Arch Dis Child 1983;5 8:509- 511.

245. Govaert P, de Vries LS. An atlas of neonatal brain sonography. Clinics in developmental medicine 141/2. Cambridge: Cambridge University Press, 1997.

246. Kreusser KL, Schmidt RE, Shackleford RD, Volpe JJ. Value of ultrasound for identification of acute haemorrhagic necrosis of thalamus and basal ganglia in an asphyxiated term infant. Ann Neurol 1984;16:361-363.

247. Eken P, Jansen GH, Groenendaal F, Rademaker KJ, de Vries LS. Intracranial lesions in the full term infant with hypoxic-ischaemic encephalopathy: ultrasound and autopsy correlation. Neuropediatrics 1994;25:301-307. 248. Siegel MJ, Shackelford GD, Perlman JM, Fulling KH. Hypoxic-ischaemic encephalopathy in term infants: diagnosis and prognosis evaluated by ultrasound. Radiology 1984;152:395-399.

249. Pinto J, Paneth N, Kazam E, Kairam R, Wallenstein S, Rose W, et al. Interobserver variability in neonatal cranial ultrasonography. Paediatr arid Perinat Epidemiol 1988;2:43-58.

250. Babcock D, Ball W. Postasphyxial encephalopathy in full-term infants: ultrasound diagnosis. Radiology 1983;148:417-423.

251. Levene MI. Measurement of the growth of the lateral ventricles in preterm infants with real-time ultrasound. Arch Dis Child 1981;56:900-904.

252. Martin DJ, Hill A, Fltz CR, Daneman A, Havill DA, Becker LE. Hypoxic- ischaemic cerebral injury in the neonatal brain: a report of sonographic features with computed tomographic correlation. Pediatr Radiol 1983;13:307-331.

253. Trounce JQ, Levene MI. Diagnosis and outcome of subcortical cystic leucomalacia. Arch Dis Child 1985;60:1041-1044.

254. Gaffney G, Squier M, Johnson A, Flavell V, Sellers S. Clinical associations of prenatal ischaemic white matter injury. Arch Dis Child 1994;7 0:F101-6.

255. Levene MI, Williams JL, Fawer C. Ultrasound of the infant brain. Clinics in developmental medicine 92. London: Heinneman, 1985.

256. Chaudhari S. Developmental assessment tests:scope and limitations. Indian Pediatr 1996;33:541-545.

257. T aylor IG. The neurological mechanisms of hearing and speech in children. Manchester: Lanchester University Press, 1964.

258. Hall SM, Pugh AG, Hall DMB. Vision screening in the under fives. BMJ 1982;285:1096-1098.

259. Stewart A. Standard recording of central motor deficit. Dev Med Child Neurol 1989;31:117-129. 184 260. Bayley N. Bayley scales of infant development manual. (2nd ed.) New York: The Psychological Corporation, 1993.

261. Blair E, Stanley FJ. Interobserver agreement in the classification of cerebral palsy. Dev Med Child Neurol 1985;27:615-622.

262. Greenwood BM, Greenwood AM, Bradley AK, Tulloch S, Hayes R, Oldfield FSJ. Deaths in infancy and early childhood in a well vaccinated, rural, West African population. Ann Trop Paediatr 1987;7:91-9.

263. Smith PG, Morrow RH. Methods for field trials of interventions against tropical diseases: a 'toolbox'. Oxford: Oxford University Press, 1991.

264. Freeman JV, Cole TJ, Chinn S, Jones PRM, White EM, Preece MA. Cross- sectional stature and weight reference curves for the UK, 1990. Arch Dis Child 1995;73:17-24.

265. Altman DG. Practical statistics for medical research. London: Chapman and Hall, 1991.

266. Cuzick J. A Wilcoxon-type test for trend. Stat Med 1985;4:87-90.

267. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;22:153-157.

268. Clayton D, Hills M. Statistical Models in Epidemiology. Oxford: Oxford University Press, 1993.

269. Lang JM, Rothman KJ, Cann Cl. That confounded p-value. Epidemiology 1998;9:7-8.

270. Freedman LS. The use of a Kolmogorov-Smimov type statistic in testing hypotheses about seasonal variation. J Epidemiol Community Health 1979;3 3:223- 228.

271. Sackett DL, Richardson WS, Rosenberg W, Haynes RB. How to practice and teach evidence-based medicine. Edinburgh: Churchill Livingstone, 1997. 185 272. Giles AM, Holloway P. Reference intervals for biochemical data. In: Weatherall DJ, Ledingham JGG, Warrell DA, ed. Oxford textbook of medicine. Oxford: Oxford University Press, 1996. 4363-4376.

273. Kuvibidila S, Warrier RP, Yu L, Ode D, Mbele V. Reference levels of acute phase reactant proteins in healthy Zairean woman in the reproductive age group. J Trop Med Hyg 1994;97:239-243.

274. Behrman RE, Vaughan VC, Nelson WE. Nelson Textbook of Paediatrics. Philadelphia: WB Saunders, 1987.

275. Roberton NRC. A manual of neonatal intensive London:care. Edward Arnold, 1993.

276. Smith DW. Recognisable patterns of human malformation. Philadelphia: WB Saunders, 1982.

277. Levene MI. Management and outcome of birth asphyxia. In: Levene MI, Lilford RJ, Bennett MJ, Punt J, ed. Fetal and Neonatal Neurology and Neurosurgery. London: Churchill Livingstone, 1995. 427-442.

278. Bolam AJ, Manandhar DS, Shrestha P, Ellis M, Malla K, Costello AM de L. Factors affecting home delivery in the Kathmandu Valley, Nepal. Health Pol Plan 1998;13:152-158.

279. Were EO. Stillbirths at Eldoret district hospital: a retrospective study. E Afr MedJ 1994;71:607-610.

280. Keeling JW, MacGillivray I, Golding J, Wigglesworth J, Berry J, Dunn PM. Classification of perinatal death. Arch Dis Child 1989;64:1345-1351.

281. Hayakawa K, Kanda T, Hashimoto K, Okuno Y, Yamori Y. MR of spastic tetraplegia. Am J Neuroradiol 1997; 18:247-253.

282. Aylward GP, Verhulst SJ, Bell S. Correlation of asphyxia and other risk factors with outcome: a contemporary view. Dev Med Child Neurol 1989;31:329-340.

283. Frankenburg WK, Dodds JB. The Denver Developmental Screening Test. J Pediatr 1967;71:181-191. 186 284. Bryant GM, Davies KJ, Richards FM, Voorhees S. A preliminary study of the use of the Denver Developmental Screening Test in a health department. Dev Med Child Neurol 1973;15:33-40.

285. Ueda R. Standardization of the Denver Developmental Screening Test on Tokyo Children. Dev Med Child Neurol 1978;20:647-656.

286. LeVine RA, Dixon S, LeVine S, Richman A, Leiderman PH, Keefer CH, et al. Child Care and Culture: Lessons from Africa. Cambridge: Cambridge University Press, 1996.

287. Manandhar N. Neurodevelopmental outcome of term low birth weight and appropriate birth weight infants aged 10-24 months in Nepal[MSc dissertation]. London: University College, 1996.

288. Molteno C, Lachman P. The aetiology of learning disability in preschool children with special reference to preventability. Ann Trop Paediatr 1996;16:141-8.

289. American College of Obstetricians and Gynecologists. Induction and augmentation of labour. Washington DC: 1987.ACOG Tech Bull; vol 110.

290. Olah KS, Gee H. The active management of labour. Br J Obstet Gynaecol 1996;103:729-731.

291. Escoffery C, Greenwood R, Ashley D, Coard K, Keeling J, Golding J. Deaths associated with intrapartum asphyxia in Jamaica. Paed Perinat Epidem 1994;8 suppl 1:119-142.

292. Hannah ME, Hannah WJ, Heilman J, Hewson S, Milner R, Willan A. Induction of labor as compared with serial antenatal monitoring in post-term pregnancy. N Engl J Med 1992;326:1587-92.

293. Yudkin P, Frumar AM, Anderson ABM. A retrospective study of induction of labour. Br J Obstet Gynaecol 1979;86:257-265.

294. Jackson M, Regan C. Elective induction of labor. Clin Obstet Gynecol 1997;40:496-509. 187 295. Jarvelin MR, Hartikainen-Sorri AL, Rantakallio P. Labour induction policy in hospitals of different levels of specialisation. Br J Obstet Gynaecol 1993;100:310- 315.

296. Laube DW. Induction of labor. Clin Obstet Gynecol 1997;40:485-495.

297. Singh M. Current recommendations for the care of high risk babies in developing countries. In: Costello AM de L, Manandhar DS, ed. Improving health of the newborn infant in developing countries. London: Imperial Press, 1998.

298. Panter-Brick. Women's working behaviour and matemal-child health in rural Nepal. In: Norgan N, ed. Physical activity and health. Cambridge: Cambridge University Press, 1992. 190-206.

299. Thame M, Wilks RJ, Mcfarlane-Anderson N, Bennett H, Forrester TE. Relationship between maternal nutritional status and infant's weight and body proportions at birth. EurJ Clin Nutr 1997;51:134-8.

300. Ceesay MS, Prentice AM, Cole TJ, Foord F, Weaver LT, Poskitt EM, et al. Effects on birth weight and perinatal mortality of maternal dietary supplements in rural Gambia: 5 year randomised controlled trial. BMJ 1997;315:786-790.

301. Unicef. The State of the World's Children. Oxford: Oxford University Press, 1998.

302. Khan N, Durkin M. Prevalence. In: Zinkin P, McConachie H, ed. Disabled children and developing countries. Oxford: Mackeith Press, 1995.

303. Ellis M, Bhattarai S, Pandey J. Cerebral palsy in Nepal:a descriptive study of 136 disabled children presenting to a cerebral palsy centre in Kathmandu. J Instit Med Nep 1998;20:12-19.

304. Alberman E. Social and biological effects on perinatal mortality. J R Soc Med 1979;72:295-297. Appendices

Appendix. INeonatal neurological examination protocol

1. Conscious level Prechtl and Beintema C2) have drawn attention to the normal cycles of the newborn term infant's state of arousal through the following states: State 1: eyes closed, regular respiration, no movements. State 2: eyes closed, irregular respiration, no gross movements. State 3: eyes open, no gross movement. State 4: eyes open, gross movements, no crying State 5: eyes open or closed, crying. • If the infant cycles normally through these states allowing examination during state 3 the infant's conscious level is classified as normal. • If state 1/2 predominates over repeated examinations classify the infant as lethargic/comatose. • If state 4/5 predominates over repeated examinations classify the infant as irritable/hyper-alert. 2. Neuromuscular tone This is assessed separately for (a)axial tone, (b)upper limb tone and (c) lower limb tone by 3 simple manoeuvres. sl) Axial Tone The supine infant is held by the shoulders and slowly pulled to the sitting position. In this position he is gently rocked backwards and forwards. Then he is gently lent backwards to the supine position. The reaction of the opposing flexor and extensor neck muscles are assessed by noting the resulting movements of the head relative to the trunk. • In the normal infant there is minimal headlag either forwards or backwards. • In the hypertonic infant there is a predominance of neck extensor tone such that the head is abnormally extended during the manoeuvre. • In the hypotonic infant there is marked head lag in both flexor and extensor dimensions during both phases of the manoeuvre. b.) The upper limb tone is assessed by the Scarf sign. The examiner pulls the infant's hand across the chest towards the opposite shoulder with the infant supported by the examiner's other hand in the semi-reclining supine position. • In the normal infant the elbow does not quite reach the midline. • In the hypertonic infant the elbow does not approach the midline. • In the hypotonic infant the elbow crosses the midline 189 (c.)The lower limb tone is assessed by measuring the popliteal angle. The examiner fflexes the infant's thighs laterally beside the abdomen and then extends the knee to its llimit. • In the normal infant the popliteal angle is 80-100°. •• In the hypertonic infant the popliteal angle is <80°. • In the hypotonic infant the popliteal angle is >100°. After completing these manoeuvres the following conclusions may be reached. • Neuromuscular tone is mildly abnormal if there is abnormality of tone according tto the preceding description in only one system or if there is generalised hypertonicity in sail three systems. «• Neuromuscular tone is moderately abnormal if there is marked dissociation of ttone between the systems e.g. lower limb tone » upper limb tone or marked axial Hiypotonicity with relatively normal limb tone. • Neuromuscular tone is severely abnormal if there is profound generalised Ihypotonia in all three systems.

3. Sucking behaviour After careful attention to hand-washing the little finger is inserted into the infant's imouth and his subsequent sucking efforts are evaluated. •• The normal infant sucks strongly in bursts of at least 8 movements with sufficient suction to exert the sensation of attempted stripping of the examiner's skin off tthe finger. • The suck is poor if either the burst is less than 8 movements, the suction effect iis poor or there is incoordination of suck such that the infant tends to grip the examiner's finger with his gums. • The suck is absent if no sucking efforts are stimulated over 30 seconds. 4. Reflexes a.) The Moro reflex may be • Exaggerated i.e. elicited with minimal stimulation by merely tapping the cot • Normal with abduction and adduction phases • Reduced with a slow and delayed abduction phase only •• Absent with no response • Asymmetrical lb.) The Grasp reflex may be • Exaggerated i.e. baby can be suspended in mid air by grasp reflex only. • Normal i.e. can pull to sitting with grasp reflex only. • Reduced i.e. will attempt to grasp your finger but unable to take any weight •• Absent •» Asymmetrical. 190

In the hypotonic infant with poor suck the following reflexes should also be assessed: c.) The Comeal reflex (wisp of cotton applied to the cornea) which may be present or absent d.) The Gag reflex (washed little finger applied to the posterior pharyngeal wall) which may be present or absent.

5. Seizures Given the many clinical manifestations of seizures in the newborn period and the high patient to nurse ratios in many developing countries these should be noted to be present and described if witnessed or reported. Attempts to quantify them precisely are not necessary.

6 . Respiratory Pattern Three encephalopathic respiratory patterns can be recognised. • Tachypnoea with a respiratory rate >60/min. • Occasional apnoeas responding to stimulation only • Severe recurrent apnoeas requiring ventilatory support 191

Appendix 2.Neonatal neurological examination data form MIRA HM ini©yir@l®gJ©iil n§@(i@§inni®iniflexion extHI exionext>fiexion ext>flexion □ ext>f lexion ext>f lexion equa l hypotoni equal hypotonequal hypotoni equal hypotoniequal hypotoni equal hypotonequal equal hypotonit P equal hypotoni equal hypotoni normal norma norma normal normal □ normal normal Ventral Suspension hypotonic hypotonic hypotonic hypotonic hypotonic □ hypotonic hypotonic hypertonic hypertonic hypertonic hypertonic hypertonic □ hypertonic hypertonic 1 elbow does not) I elbow does not! I elbow does not □ elbow does notl 1 elbow just react I elbow just react I elbow just reac □ elbow just reac| ] elbow crosses I I elbow crosses [ I elbow crosses □ elbow crosses i

I no resistance I no resistance I I no resistance I I no resistance □ no resistance I I no resistance I mild resistance | I mild resistanci I I mild resistance! I mild resistance □ mild resistance I I mild resistance I ++ resistance • resistance I I ++ resistance I I ++ resistance □ ++resistance I I ++ resistance present t!j present □ present b present □ □ present present Se.lsg cortical thumb I absent P absent P absent P absent P □ absent 1 absent popliteal angle [gogjiteal^ngle^ heel ear angle [heehaaj^ngji^^J^ adductor angle ankle angle □ Normal ...... □ Normal □ Normal (□Normal □ Normal 1 TONE ASSESSMENT 0 Normal □ Normal □ Mild. Abnorm* □ Mild. Abnomti □ Mild. Abnorm □ Mild. Abnorm □ Mild. Abnorm □ Mild. Abnorm □ Mild. AbnormJ □ Mod. Abnorma □ Mod. Abnorma □ Mod. Abnorme □ Mod. Abnorme □ Mod. Abnorm □ Mod. Abnorma □ Mod. Abnorma □ Sev. Abnormal □ Sev. Abnorma □ Sev. Abnorma □ Sev. Abnorma □ Sev. Abnorme □ Sev, Abnormal □ Sev. Abnormal day of life Copy Sbv of I3» Cgw 1 □ exaggerated □ exaggerated □ exaggerated □ exaggerated □ exaggerated □ exaggerated □ exaggerated □ normal □ normal □ normal □ normal □ normal □ normal □ normal □ reduced □ reduced □ reduced □ reduced □ reduced □ reduced □ reduced □ absent □ absent □ absent □ absent □ absent □ absent □ absent □ assymetrical □ assymetrical □ assymetrical □ assymetrical □ assymetrical □ assymetrical □ assymetrical □ exaggerated □ exaggerated □ exaggerated □ exaggerated □ exaggerated □ exaggerated exaggerated graap □ normal □ normal □ normal □ normal □ normal □ normal □ normal □ reduced □ reduced □ reduced □ reduced □ reduced □ reduced □ reduced □ absent □ absent □ absent □ absent □ absent □ absent □ absent □ assymetrical □ assymetrical □ assymetrical □ assymetrical □ assymetrical □ assymetrical □ assymetrical □ present □ present □ present □ present □ present □ present □ present □ absent □ absent □ absent □ absent □ absent □ absent □ absent

Describe Seizures

gag reflex □ present □ present □ present □ present □ present □ present □ present □ absent □ absent □ absent □ absent □ absent □ absent □ absent □ present □ present □ present □ present □ present □ present □ present comeal reflex □ absent □ absent □ absent □ absent □ absent □ absent □ absent □ Normal □ Normal □ Norma □ Norma □ Normal □ Normal □ Normal 7 . Respiratory Pattern □ Tachypnoea □ Tachypnoea □ Tachypnoea □ Tachypnoea □ Tachypnoea □ Tachypnoea □ Tachypnoea □ Apnoea-mild □ Apnoea-mild □ Apnoea-mild □ Apnoea-mild □ Apnoea-mild □ Apnoea-mild □ Apnoea-mild □ Apnoea-severe □ Apnoea-severe □ Apnoea-severe □ Apnoea-severe □ Apnoea-severe □ Apnoea-severe □ Apnoea-severe □ normal □ normal □ normal □ normal □ normal □ normal □ normal @. Visual Movements □ abnorm. specif) □ abnorm. specif) □ abnorm.specif) □ abnorm.specif) □ abnorm.specif) □ abnorm. specify □ abnorm.spec if)

Head circumference Head circumfer □ depressed □ depressed □ depressed □ depressed □ depressed □ depressed □ depressed D @ , Fontsnelle □ normal □ normal □ normal □ normal □ normal □ normal □ normal □ full □ full □ full □ full □ full □ full □ full □ tense □ tense □ tense □ tense □ tense □ ten se □ tense HIE GRADE J------□ phenobarb. □ phenobarb. □ phenobarb. □ phenobarb. □ phenobarb. □ phenobarb. □ phenobarb. treatment □ antibiotics □ antibiotics □ antibiotics □ antibiotics □ antibiotics □ antibiotics □ antibiotics Initials of □ phototherapy □ phototherapy □ phototherapy □ phototherapy □ phototherapy □ phototherapy □ phototherapy completing doctor ______1______1...... I— ...... i______192 Appendix 3. Perinatal clinical data form MfllRA HOE ©TODY - MATE^INAL BHE Q(5fl®iraligffllffiatll]©ora D o t t s S l S & l i S & S S L f l E t S S S ^ L I t l & S l f f i L S S S l Siam y otua^jv^^l IP ® IT II JJ SB ft (II0 HOsHtSIII Collect data from maternal notes supplemented by discussion with the obstetric staff. Dote of Hospital admission LtfiifiUilLliO&Bllai Date of delivery Time of Hospital admission IT.manfHnan.tal Time of delivery

Syntoclnon 9 Y es------9 Induction ot Ikbdur ------Infusion O No If yes, O Augmentation of labout Administered O don't know Syntoclnon for O d o n 't know

O yee Fetal heart rate Total number of times Tntal niimhnr of monitoring performed O no fetal heart monitored Itimaa fatal haart O don't know d» 2 hourly O d o n 't know

Fetal tachycardia >180 bpm noted |Q yos Q n o Q don't know-]

Fetal bradycardia <120 bpm noted iq yes & no Q don't know |

Time abnormal fetal haartrate first detected

Meconium Stained Uquor KJinicK o inin oNione p uont KrioW I

Time from recognition of fetal dlstreee to delivery, minute □ h1 hr) Mode of delivery □ O th e r... □ spontaneous vaginal □ vaginal with vacuum □ vaginal with high forceps □ emergency c/s I induced vaginal □ vaginal with low forceps Q elective caesarian section □ Other.

Indication for Intervention Indication for intervention ( If mor* than on* Hat m descending order of importance ) PDsnsa Turn Twin delivery IP Yes Q IVO | ( If yes complete separate perinatal data sheets for each twin

Obstetric Group p a : s . b : g g MIIbI* of oomploUnq doctor

s i d e a o r j Husband’s name Maternal Age Husband's occupation IHuspand's occupatioTr LMP EDD t o Mother's years completed schooling LdfiUlStifiJISal Place of Birth (if outside KTM valley) Currently staying within Kathmandu Ringroad IS y es' f l n o I (including Kirtipur, Patan and Boudha) Planning to remain within Kathmandu Ringroad for 1 year |Q y e s Q n o Q d o n 't kriov7"|

y o a o ib Bed No Mother’s O N ew ar 0 L a m a Ward O New Paying Ward ed'Nb ** Ethnic O B rahm in O G urung O N ew W ard Group O C hhetrl O O th e r... E O T a m an g INFORMED CONSENT IQ YfeS ONOI O R ai

O to®t® 11 F 0® Do not include the current delivery No of , lees than 28 weeks |Nn at J No of completed singleton pregnane Is . areeter than 28 weeks pTnT-| jWS?..... ,9,1. No of completed twin pregnancies, greater than 28 weeks Total number of live blrthe | < PTaj... ] Total number of stillbirths p n td l greater than 28 weeks Li .-. L. Months since last live born delivery IMohths fclrtca Do the parents have any handicapped children? tO'nS* "OVed. . dbdfcity ~ Is there a history of previous neonatal death*? to no Q yes .specify No of attendances at Antenatal car* Antenatal Complications I Pffbtmhals Place of intended delivery jOPrasuti (jrina q other nospitai q nome History of decreas ed fetal movement ves o n6 o aon't Know i Membrane* ruptured more than 24 hour* before delivery IQ ves & no O don't know I Date of onset of labour* Tim* of onset of labour (“when regular PAINFUL contractions began)

Haemoglobin QgyQflflEhO Tim* from delivery in hours £ a I (at least 12 hours) Maternal Weight paternal M„t*rnal Height E 22 S H 2 E 1 Cheeked by j^ h eg k e^j 193 Appendix. 4NE management protocol

IhrnnzlsQMioi Environmental temperature of SCBU was maintained above 25 degrees. The infants were nursed mainly on locally manufactured trolleys with an integral 100W electric bulb overhead heat source. Fluid management During the early stages of encephalopathy a restricted fluid intake of 50 ml/kg/day was given to allow for the risk of fluid overload associated with inappropriate anti diuretic hormone syndrome. When the child was sucking normally with an intact gag reflex this was given orally and breast feeding was encouraged. In the absence of a normal suck provided the infant had a normal gag reflex the local variant of spoon feeding was permitted using expressed breast milk via a traditional ritual oil lamp with small spout. If the gag reflex was absent or the baby was deeply lethargic breast milk was given by nasogastric tube. In a comatose or fitting infant 10% dextrose was administered via intravenous infusion. Hvpoglvcaemia Blood sugar was measured after diagnosis by glucose strip testing (Boehringer Mannheim strips [BM] and Reflolux-S meter) to eliminate hypoglycaemia (blood glucose less than 2.6 mmol/L). If hypoglycaemia was detected and clinical condition allowed it a feed was offered immediately. If this failed to raise the BM above 2.6 after 1 hour on repeat testing an iv infusion of 10% dextrose was commenced and the blood BM checked again after 1 hour. If the BM remained below 2.6 mmol/1, 15% dextrose was administered. If despite this the infant remained hypoglycaemic then the infant was excluded from the study and hydrocortisone was given with a diagnosis of encephalopathy due to symptomatic refractory hypoglycaemia.

Hvpoxaemia Where the infant was clinically cyanosed oxygen was administered generously by nasal prongs. A single portable oxygen saturation monitor was used if available but otherwise oxygen therapy continued to be administered since these were term babies with little risk of oxygen toxicity. During early resuscitation efforts artificial pressure limited ventilation was performed using the single available ventilator for a period of upto one hour. Only if there was evidence of independent respiratory effort at this time was ventilation continued up to a maximum of 6 hours on occasion.

Infection If infection was clinically suspected or investigations suggested it was a possibility first line antibiotics, ampicillin and gentamycin, were administered intravenously. In the 194 event of deterioration and a decision to change antibiotics to a second line regime (usually cefotaxime) investigations were repeated.

Seizures When seizures were noted they were carefully described, blood sugar was checked and the baby commenced on intravenous phenobarbitone at a loading dose of 2 0 mg/kg . Raised intracranial pressure In the presence of marked clinical signs of raised intracranial pressure at consultant discretion a single 20% mannitol infusion was given in a dose of 7 ml/kg. 195 Appendix 5.Parametric data distributions

Figure 14. Distributions of maternal anthropometric data for all the study groups combined (n=1063).

350 - 300 - 250 - § 200 - o 150 “

100 " 50 “ 0 - 30 40 50 60 70 80 90 maternal weight (kg)

350 300 ^ 250 § 200 o 150 100 50 0 120 130 140 150 160 170 180 190 maternal height (cms) rus obnd (n=1077). combined groups iue 5 itiuin fifn nhooercdt o l h study the all for data anthropometric infant of Distributions 15. Figure

Count Count Count 250 200 300 350 200 - 250 “ 300 200 “ 250 ~ 300 350 100 150 100 ~ 150 100 “ 150 50 0 - 50 0 “ 50 0 0 0

- - ~ - - “ 1 8 0 2 4 6 8 40 38 36 34 32 30 28 5 0 5 50 45 40 35

5 25 3.5 3 2.5 2 .5 nat edcrufrne (cms) circumference head Infant nat itlnt (cms) birthlength infant T nat itwih (kgs) birthweight infant T 55

197

Figure 16. Distribution of maternal haemoglobin for all groups combined (n=1050).

350 - 300 " 250 " § 200 " o 150 " 100 - 50 " 0 - Lrr ^ '^ T '"T-T- 4 6 8 10 12 14 16 18 20 22

maternal haemoglobin (g/dl)

Figure 17. Distribution of neonatal haemoglobin in NE cases only (n=122). i i i

3c 1 oO

2 -

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 inf h b ( g / d l )

Figure 18. Distribution of maternal FT4 for all groups combined (n=1050).

207 ~

0 “

FT4 pmol/l Figure 19. Distribution of maternal TSH for all groups combined (n=1050).

259 ~

£ U.

0 “

TSH rnKJ/1

Figure 20. Distribution of maternal magnesium for all groups combined (n=1046).

213 “

0 - _ .75 Magnesium mmols/I

Figure 21. Distribution of maternal CRP for all groups combined (n=1046).

260 “

0 - 100 125 150 CRP (mg/1) Note: CRP >120 coded as 150 mg/1 Figure 22. Distribution of maternal AGP for all groups combined (n=1046).

159 ~

uZ

o - 2.5 AGP g /l

Figure 23. Distribution of the average ventricular indices of control infants examined between 30 and 60 days post-delivery. (n=103). 200

Appendix. 6Review tables of low cost predictors of neuroimpairment in NE infants

Table 58. Review of the use of NE grade for the prediction of major neurodevelopmental impairment in NE infants at one year.

Major Impairment No/minor impairment Total NE2/3 16 12 28 NE 1 2 27 29 total 18 39 57 Sensitivity 88% Specificity 69% Pre-test probability 31% Predictive value (+) 57% Predictive value (-) 93% Likelihood Ratio (+) 2.8 Likelihood Ratio (-) 0.17 Post-test probability0.56

Table 59. Review of the use of abnormal neurological examination at discharge in NE infants for the prediction of major neurodevelopmental impairment at one year.

Major Impairment No/minor impairment Total abnormal neurology 16 15 31 normal neurology 2 24 26 total 18 39 57 Sensitivity 89% Specificity 83% Pre-test probability 31% Predictive value (+) 52% Predictive value (-) 92% Likelihood Ratio (+)5.2 Likelihood Ratio (-) 0.13 Post-test probability 0.70 201 Table 60. Review of the use of echogenic appearance of the neonatal brain in NE infants for the prediction of major neurodevelopmental impairment at one year.

Major Impairment No/minor impairment Total echogenic 8 6 14 normal 8 29 37 total 16 35 51 Sensitivity 50% Specificity 83% Pre-test probability31% Predictive value (+) 57% Predictive value (-) 78% Likelihood Ratio (+)2.9 Likelihood Ratio (-) 0.6 Post-test probability0.57

Table 61. Review of the use of cystic changes on cranial ultrasound at six weeks of age in NE infants for the prediction of major neurodevelopmental impairment at one year.

Major Impairment No/minor impairment Total cysts at 6 weeks 7 0 7 no cysts 9 36 45 total 16 36 52 Sensitivity 43% Specificity 100% Pre-test probability 31% Predictive value (+) 100% Predictive value (-) 80% Likelihood Ratio (+) >1000 Likelihood Ratio (-) 0.57 Post-test probabilityapprox. 100% 202 Table 62. Review of the use of acquired microcephaly (defined as a negative change in the occipito-frontal circumference SD score of two or more at six weeks) in NE infants for the prediction of major neurodevelopmental impairment at one year.

Major Impairment No/minor impairment Total acquired microcephaly 7 2 9 no microcephaly 10 36 46 total 17 38 55 Sensitivity 41% Specificity 95% Pre-test probability 31% Predictive value (+) 78% Predictive value (-) 78% Likelihood Ratio (+) 8.2 Likelihood Ratio (-) 0.62 Post-test probability79% Appendix 7: Publications and Presentations

Principal publications Stillbirths and neonatal encephalopathy in Kathmandu, Nepal: an estimate of the contribution of birth asphyxia to perinatal mortality in a low income urban population. Ellis M, Manandhar DS, Manandhar N, Wyatt J, Bolam AJ, Costello AM de L .Paediatr Perinat Epidem In Press

One year outcome of neonatal encephalopathy in Kathmandu, Nepal. Ellis M, Manandhar N, Shrestha L, Shrestha PS, Manandhar DS, Costello AM de L. Dev Med Child Neurol In Press

Birth asphyxia in developing countries: biology, sequelae, epidemiology and prevention. Ellis M. Department for International Development Working Paper (in conjunction with Centre for International Child Health, London) Available on the internet at http://www.ich.ucl.ac.uk

Progress in perinatal asphyxia Ellis M.Semin Neonat In press.

Birth asphyxia Ellis M in Improving health of the newborn infant in developing countries. Costello AM de L, Manandhar DS eds. London: Imperial Press, In press.

Birth asphyxia, Apgar score and neonatal encephalopathy. Ellis M, Costello AM de L. Indian Pediatr 1997; 34: 975-8.

An Apgar score of three or less at one minute is not diagnostic of birth asphyxia but is a useful screening test for neonatal encephalopathy. Ellis M, Manandhar N, Manandhar DS, Costello AM de L. Indian Pediatr 1998; 35:415-421.

Cerebral palsy in Nepal: a descriptive study of 136 disabled children presenting to a cerebral palsy centre in Kathmandu. Ellis M, Bhattarai S, Pandey J. J Inst Med Nep 1998;20:12-19.

Cerebral palsy in developing countries. Ellis M. Postgrad Doc 1997; 13:117-20.

Incidence and severity of neonatal encephalopathy in Nepal, a least developed country (abstract). Ellis M, Manandhar N, Manandhar DS, Wyatt JS, Costello AM de L. Paediatr Perinat Epidem 1996; 10: A 17. 204 Principal presentations

Seventeenth Conference of the German Tropical Paediatric Society, Witten 1999 Neonatal encephalopathy: the Kathmandu study.

Royal College of Paediatrics and Child Health Annual Conference, York 1998 (plenary presentation in the presence of the Royal Patron) Prevalence and one year outcome of neonatal encephalopathy in ,Kathmandu Nepal.

Second European Congress on Tropical Medicine, Liverpool 1998 One year outcome of neonatal encephalopathy in Nepal

Nepal Paediatric Society Biennial Conference, Kathmandu 1998 Neonatal encephalopathy and cerebral palsy: lessons from a cohort study in Nepal.

ODA International Workshop on Perinatal Health in South Asia, Kathmandu 1997 Birth asphyxia in South Asia and Informatics in South Asia

International Conference on the Epidemiology of Cerebral Palsy, Beme 1996 Incidence and severity of neonatal encephalopathy in, a Nepalleast developed country.

Neonatal Forum of all India, Patna 1995 A practical system for grading neonatal encephalopathy

Related publications The effects of postnatal health education for mothers on infant care and family planning practices in Nepal: a randomised controlled trial. Bolam AJ, Shrestha P, Manadhar DS, Ellis M, Costello AM de L. BMJ 1998; 316: 805-11.

Liquid crystal thermometry for the detection of neonatal hypothermia in Nepal. Manandhar N, Ellis M, Manandhar DS, Morley D, Costello AM de L.J Trop Pediatr 1998; 44: 15-17.

Factors affecting home delivery in the Kathmandu Valley, Nepal. Bolam AJ, Manandhar DS, Shrestha P, Ellis M, Malla K, Costello AM de L. Health Pol Plan 1998 ; 13: 152-8. Maternity care utilisation in the Kathmandu Valley: a community based study. Bolam AJ, Manandhar DS, Shrestha P, Manandhar B, Ellis M, Costello AM de L. JNep Med Assoc 1997;35:122-129.

Postnatal hypothermia and cold stress among newborn infants in Nepal monitored by continuous ambulatory recording. Ellis M, Manandhar N, Shakya U, Manandhar DS, Fawdry A, Costello A.M de L. Arch Dis Child 1996; 7 5: F42-F45.

Comparison of two cotside methods for the detection of hypoglycaemia among newborns in Nepal. Ellis M, Manandhar DS, Land J, Patel N, Costello AM de L. Arch Dis Child 1996; 75: F122-F125.

Newborns who need special care. Ellis M, Manandhar DS. Child Health Dialogue 1996; 2: 7.

Related presentations

Nepal Paediatric Society Biennial Conference, Kathmandu 1996 Neonatal hypothermia in Nepal.

British Paediatric Association Annual Conference, York 1995 (Plenary session on international child health) Postnatal hypothermia and cold stress among newborn infantsNepal. in

The Asia and Oceania Perinatal Society, International workshop, Lahore 1995 The effect of ambient temperature on neonatal core hypothermia and cold stress.