Risk factors for neonatal encephalopathy in Kathmandu, Nepal, a developing country: unmatched case-control study

Setting

Nepal is a low income country (£120 per capita in 1995) with a population of 21 million and an infant mortality rate estimated to be 95 per 1000 live births in 1994.16 The capital, Kathmandu, is a rapidly growing conurbation situated in a large valley surrounded by hills. The estimated population of the valley is 1 100 000.17 A recent community based study showed that 81% of mothers living in or near Kathmandu chose to give birth in hospital.18

Prasuti Griha is the major maternity hospital in Kathmandu, delivering over 14 000 babies annually, around 50% of all the deliveries in Kathmandu. Poor communications exist with the hilly regions, and a recent audit showed that less than 3% of women delivering at Prasuti Griha were resident outside the valley. In Nepal there is no functioning referral system for pregnant women. The hospital budget of £350 000 is supplemented by a modest user charge and offers the most competitively priced delivery service in the valley. We therefore consider the mothers delivering at the hospital to be broadly representative of the surrounding population.

Fetal monitoring is by intermittent auscultation; partograms are not used. The caesarean section rate was 10% during the study period. For labour induction or augmentation, oxytocin infusion is given in dosages of 2 mU per minute titrated up to 15 mU per minute. Amniotomy is not routinely performed. The minimum period of stay for birth is 24 hours. Neonatal resuscitation including suction, oxygen by bag and mask or endotracheal tube, and cardiac massage follows standard procedures and is undertaken by the resident paediatric staff. These are junior doctors serving a training attachment in neonatology. The special care baby unit has 30 cots. There are no facilities for long term ventilation.

Case definition and recruitment

Neonatal encephalopathy is defined as an abnormal neurobehavioural state commencing within the first 24 hours of life, which consists of an altered conscious level with abnormalities of neuromuscular tone or sucking behaviour.19 In addition there may be seizures or abnormalities of respiratory control, primitive reflexes, and brainstem reflexes.

During the 18 month recruitment period a research oriented surveillance system ran throughout the hospital.20 All term infants with evidence of neurobehavioural disturbance were examined daily by the junior doctors who had been trained by the principal investigator (ME) using a protocol based on that of Amiel-Tison.21 The findings were recorded daily on a standard proforma. Formal operational criteria to grade the severity of encephalopathy in term infants were derived from the syndromic descriptions of Fenichel modified according to more recent studies.19 In particular these modifications incorporated the observations that infants with mild neonatal encephalopathy may have signs of not only decreased but increased tone,22 that seizure activity may not be clinically detectable and therefore cannot serve as a definitive feature in any grading system,23 and that the inclusion of duration in the clinical definition of a grade24 renders the scheme contradictory.1 Infants with mild (grade 1) encephalopathy were irritable or hyperalert, with either poor suck or an abnormality of tone. Infants with moderate (grade 2) encephalopathy were lethargic, with moderately abnormal tone, poor suck, and depressed Moro and grasp reflexes (seizures were often clinically evident). Infants with severe (grade 3) encephalopathy were comatose, with severely abnormal tone, absent suck, and brainstem malfunction including impaired respiratory drive. Interobserver reliability of the grading system was tested among five paediatric doctors on 27 neurologically abnormal infants. The unweighted kappa measure of interrater agreement was 0.87, which according to standard interpretations is almost perfect.25 26 The final assigned grade corresponded to the worst condition of the infant observed during the period of neurological abnormality.

We included all infants with a gestational age greater than 37 weeks27 with evidence of neonatal encephalopathy at 6-24 hours after birth. Exclusion criteria were: preterm infants (less than 37 weeks); severely dysmorphic infants or infants with at least one major congenital abnormality; infants with hepatosplenomegaly, cataracts, and thrombocytopenia indicative of intrauterine infection; infants with microbiological evidence of early neonatal infection, with positive bacteriological culture of blood or cerebrospinal fluid collected within 24 hours of birth; and infants whose neurological condition normalised when hypoglycaemia was corrected.

We excluded 20 of 151 encephalopathic infants recruited;13 had evidence of infection, two with clinical evidence of intrauterine infection and 11 with early bacterial septicaemia.

Recruitment of controls

On arrival at the hospital women were issued with an in-patient serial number, given out in chronological order. According to this number every 25th infant born in the hospital was clinically examined, and provided that the infant met the inclusion criteria—with the exception of not being encephalopathic—the infant and mother were recruited as controls. If the infant failed to meet these criteria (usually because of being preterm, occasionally because of congenital abnormalities) we recruited the next infant according to the number system. We followed this procedure for 12 months, generating an unmatched control group of 635 subjects.

Data collection

Maternal characteristics and antepartum history were obtained during an interview of the mothers 12-24 hours after delivery, using structured proformas completed by the resident paediatricians. Relevant details of the perinatal history were abstracted systematically from the maternity hospital records at this time. The records form an account of clinical progress and intervention during the intrapartum period and do not include data from the antenatal clinic. All mothers and infants were measured for standard anthropometric variables. Maternal height was measured to the nearest 0.1 cm with a minimetre (Raven, Dunmow, United Kingdom). Infant weight was measured to the nearest 10 g with Soehnle (Raven) electronic infant weighing scales. Puerperal blood samples were collected from the mothers by standard venepuncture. The concentration of maternal haemoglobin was estimated at the bedside using the Hemocue photometric method.28 Plasma and serum samples were prepared immediately and stored at −10°C. All samples were later transferred for assay. Thyroid stimulating hormone assays were performed using an enzyme linked fluorescent assay technique (Vidas bioMérieux, Lyons, France). Assays for free magnesium measured the reflection density of a magnesium-dye complex (Johnson and Johnson, Rochester, NY).

Data analysis

All data were entered into a computer database, doublechecked, and analysed with Statview (version 4.1) and Stata (version 5.0). The prevalence of neonatal encephalopathy was estimated from routinely collected hospital vital statistics for the denominator data. Confidence intervals for proportions were calculated with the exact binomial distribution.

A core model containing preconceptual and antepartum variables was constructed using those variables found to be significantly associated with neonatal encephalopathy in previous studies. Bivariate associations between neonatal encephalopathy (yes or no) and variables of antepartum exposure were estimated by unconditional logistic regression. We tested potentially confounding variables using a likelihood ratio test procedure (twice the change in the log likelihood), and we included only those additional terms that represented a significant (P<0.05) component of the model. For the final model we estimated adjusted odds ratios using logistic regression, combining all the significant variables.5

For intrapartum and puerperal assay variables of interest, the association with the binary variable neonatal encephalopathy (yes or no) was first estimated using unconditional logistic regression and subsequently adjusted for all the variables in the final antepartum model.

Preconceptual and antepartum risk factors

Table 3 lists the preconceptual and antepartum risk factors. High maternal age (more than 35 years, odds ratio 4.35, 1.04 to 18.2) and primiparity (2.0, 1.10 to 3.61) were significant independent risk factors. Maternal education in this population is a proxy for socioeconomic status and did not seem to be a significant independent risk factor. This population sample, however, showed evidence of widespread deprivation (nearly half of the women had never attended school) and this may have diluted any effect. Short stature, defined conventionally in South Asia as a standing height of less than 145 cm, was significantly associated with neonatal encephalopathy (3.16, 1.50 to 6.66).29 Although over 5% of the mothers of encephalopathic infants reported previous neonatal death, the numbers were low and were not significant when adjusted for other variables (1.89, 0.53 to 6.81). A significantly increased risk was found among those mothers who never accessed antenatal care (2.05, 1.16 to 3.66). A history of pre-eclampsia was associated with a non-significant increased risk (1.86, 0.82 to 4.22). Over 20% of both encephalopathic infants and controls were of low birth weight (less than 2500 g) and no study infants had a birth weight of more than 4000 g. No association between encephalopathy and birth weight was found in this population. An excess of boys was observed among encephalopathic infants but this was not independently significant (1.38, 0.84 to 2.26). Plurality (twins in all cases) carried a highly significant increased risk despite the small numbers in this study (22.11, 3.45 to 141.47). Of the five cotwins of encephalopathic infants, two were neurologically normal, one was anencephalic, one was a macerated stillbirth, and no data were available for one. Of the exposures of a priori interest a history of pre-eclampsia and sex of the infant did not contribute significantly to the model (P values on likelihood ratio testing of 0.066 and 0.215 respectively). These variables have been retained for comparative purposes. Potentially confounding variables that did not contribute significantly to the model were month of birth and a history of maternal anaemia.

Intrapartum risk factors

Table 4 lists the intrapartum risk factors. Around 5% of term infants had a non-cephalic presentation among whom there was a significant excess risk of encephalopathy (3.35,1.38 to 8.11). A significant risk was associated with prolonged rupture of membranes (3.84, 1.56 to 9.47). Forty one of 766 deliveries (5%)were induced with oxytocin, of which 12 (29%) resulted in encephalopathic infants (5.28, 2.03 to 13.76). No encephalopathic infants were delivered by elective caesarean section. Instrumental delivery was associated with greater risk to the infant than emergency caesarean section. Serious intrapartum complications were more common among encephalopathic infants (28; 21%) than controls (17; 3%). In encephalopathic infants these included obstructed labour resulting in a prolonged second stage, instrumental or operative delivery (19 infants), cord prolapse (four), maternal eclampsia (two), intrapartum haemorrhage (two), and uterine rupture (one). Meconium stained liquor and especially thick (particulate) meconium was importantly associated with encephalopathy (18.21, 8.05 to 41.18). When oxytocin administration was coded according to augmentation or induction, induction was more significantly associated with adverse neonatal outcome (9.09, 3.32 to 24.83) even after adjustment for antenatal confounders including a history of pre-eclampsia. Among the control population induced labour was likely to be more protracted (mean 19 hours) than spontaneous labour (14). 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. Despite the liberal use of oxytocin for augmentation, prolonged labour as conventionally defined was common but not significantly associated with encephalopathy (1.04, 0.60 to 1.80). Nor did there seem to be an important association with time from admission to delivery. No association was found with day of delivery, a proxy for the obstetric team providing care.

Like the Perth group,7 we used the following criteria to estimate the proportion of infants who had been exposed to possible intrapartum hypoxia: the presence of an abnormal fetal heart rate on auscultation (less than 120 beats per minute or more than 160) or thick meconium in labour, or both, together with a one minute Apgar score of less than three and a five minute Apgar score of less than seven. Overall, 56 (43%) encephalopathic infants and 19 (3%) controls fulfilled these criteria. A further 22 encephalopathic infants (17%) did not meet these criteria but had evidence of a significant intrapartum event; clinical evidence of obstructed labour resulting in either a prolonged second stage or instrumental or operative delivery (18 infants), cord prolapse (3), and major intrapartum haemaorrhage (1). Overall, 78 encephalopathic infants (60%) compared with 36 controls (6%) either had evidence of intrapartum compromise or were born after an intrapartum difficulty likely to result in fetal compromise.

Puerperal maternal assays

Routine antenatal blood assay is not performed in low income settings. Given the large baseline population for this study, antenatal blood sampling was not feasible. We therefore had to rely upon maternal blood assays performed in the puerperium. table 5 presents the assay data for haemoglobin, thyroid stimulating hormone, and magnesium. Given the differing cut off points used in the definition of maternal anaemia, haemoglobin concentration was stratified in 2 g/dl bands. Low haemoglobin concentrations (less than 8 g/dl) were significantly associated with encephalopathy (odds ratio 2.64, 1.00 to 7.01). Overall, 16% of control mothers had increased concentrations of thyroid stimulating hormone (greater than 5 mIU/l) suggestive of thyroid hormone deficiency. This proportion increased to 24% in mothers of encephalopathic infants, with a significantly increased risk of encephalopathy after adjustment for antepartum variables (2.14, 1.19 to 3.82). Almost half of the control mothers had magnesium concentrations below a conventional cut off point. This was not associated with a significantly increased risk of encephalopathy in their offspring.

Neonatal encephalopathy

In the low income setting of Kathmandu, as in the high income setting of Perth, increasing maternal age and multiple birth are associated with an increased risk of neonatal encephalopathy. Unlike Perth, in the less well nourished Kathmandu population both short maternal stature and primiparity are significantly associated with encephalopathy. Indirect data suggested an association with maternal hypothyroidism and severe anaemia. Almost 1 in 5 of this urban population had never received antenatal care and their infants were at significantly greater risk of encephalopathy than those who had attended for antenatal care. In contrast to Perth, there was no apparent association with maternal socioeconomic status, low infant birth weight, or sex of the infant.

Intrapartum monitoring detected evidence of intrapartum hypoxia affecting 43% of the encephalopathic infants. As in Perth there was a particularly strong association with particulate meconium. Infants presenting in a non-cephalic position were at significantly greater risk. Intrapartum complications were importantly associated with encephalopathy. In Kathmandu 60% of encephalopathic infants had evidence of possible intrapartum hypoxia compared with 29% in Perth. The adjusted odds ratios for intrapartum management were higher in Kathmandu than in Perth for all modes of delivery intervention but, as in Perth, emergency caesarean section seemed to be associated with a lower risk of encephalopathy than instrumental vaginal deliveries. Twelve encephalopathic infants (9%)were born after induction with oxytocin.

Study limitations

There is, as yet, no widely agreed definition for neonatal encephalopathy.30 We chose to define neonatal encephalopathy so as to enable comparison with prevalence studies of hypoxic-ischaemic encephalopathy in other settings.4 We have shown elsewhere that grade 1 (mild) encephalopathy is associated with significant neonatal mortality in Kathmandu unlike in high income settings.3 We therefore report this risk factor study for all grades of encephalopathy combined. Our surveillance system was designed to detect all term infants with early neonatal encephalopathy. 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 presents within 24 hours of birth.19 Unlike the Perth group we did not have the benefit of detailed maternity records combining antenatal findings with continuously monitored intrapartum data. Our antepartum data relied on retrospective maternal interview, which may have been open to recall bias. Neither the interviewers nor the participants were familiar with the precise study aims and had no explicit gain from biased replies, and no women refused interview. The intrapartum data were taken from contemporaneous clinical notes and are therefore of higher reliability within the constraints of this low income setting. Logistics dictated a reduced period of 12 months for recruitment of controls—the same recruitment period as for the cases would have been preferable. The limitations of locally available microbiological facilities may have had some effect on our detection rate for early neonatal infection. Assays performed in the puerperium are only indirect measures of prior status during pregnancy. These assay data may also be confounded by several variables including type of delivery and interval beween birth and venepuncture.

Implications for perinatal care in low income settings

We have previously shown that neonatal encephalopathy in low income settings carries a high risk of death in infancy, with 441 deaths per 1000 live births by one year of age (95% confidence interval 343 to 543) compared with 37 deaths per 1000 live births among term newborn infants not requiring special care (10 to 94). Most encephalopathic infants who died (89%) did so in the early neonatal period. Of the 27 survivors of moderate encephalopathy, 16 developed major neurodevelopmental impairment (59%, 39% to 78%). Neonatal encephalopathy continues to importantly contribute to neonatal mortality and long term morbidity in this population.

Absent antenatal care was associated with an increased risk of encephalopathy in our study. The safe motherhood policy of Nepal's government recommends a minimum of three antenatal visits during pregnancy, focusing on risk screening, immunisation against tetanus, anaemia prophylaxis and treatment, and health and nutrition education. The apparent protective effect of antenatal care in Kathmandu needs to be interpreted with care. It is likely that both exposure and outcome are confounded by socioeconomic factors. The only proxy for socioeconomic status we were able to measure was maternal educational status. The effect remains after adjustment for this. It is biologically plausible that risk screening (when combined with appropriate advice and management) and anaemia prophylaxis may be protective to the fetus.

Particulate meconium was strongly associated with encephalopathy in Kathmandu as elsewhere.31–34 This lends indirect support to the developing case for the routine use of intrapartum amnioinfusion—the infusion of normal saline through the patent cervix when labour is complicated by the passage of fresh meconium in utero. A recent study from Zimbabwe found a protective effect of amnioinfusion in hypoxic-ischaemic encephalopathy (odds ratio 0.07, 0.01 to 0.56), with no evidence of increased risk of puerperal pyrexia.35 Although there remains uncertainty about the potential effect of amniofusion on vertical transmission of HIV, a recent systematic review recommended its use where peripartum surveillance is limited, as in many low income settings.36

The most potentially preventable risk factor for encephalopathy was induction of delivery with oxytocin. Overall, 12 encephalopathic infants (9%) compared with 29 controls (5%) were induced. The calculated population attributable fraction of encephalopathy (the proportion that may be reversible) after induced delivery in this population is 5%. This assumes a causal relation. Common indications for induction in Kathmandu are postmaturity and pre-eclampsia. That induction in Kathmandu remains a significant independent risk factor for encephalopathy after adjustment for other significant exposures argues for an independent causal effect. Hypertonic uterine contraction, a dose dependent side effect of oxytocin, reduces placental perfusion and hence oxygen supply to the fetus. It is therefore plausible that the association between induction and adverse pregnancy outcome is in part causal.

The Perth group have previously reported an odds ratio of 2.3 (1.2 to 4.6) on univariate analysis for encephalopathy associated with induction.37 On multivariate analysis of a larger cohort, however, they found the adjusted odds ratio to be non-significant (0.97, 0.57 to 1.68). Similarly no association between induction and early neonatal seizures was found during the Cardiff births' survey (1.1, 0.5 to 2.1).38

In the late 1980s in Canada a study of elective induction for postmaturity in comparison with conservative management utilising serial antenatal monitoring found no difference in perinatal mortality or neonatal morbidity between the two groups.39 A retrospective study of induced deliveries in a UK population at low risk at term found no difference in neonatal morbidity but an increased proportion of operative and instrumental deliveries in the induced group.40

The Jamaican perinatal mortality survey by contrast described a significant association (odds ratio 2.3) between induction and intrapartum death.41 A recent study from Jerusalem assessed the rate of intrapartum complications for three different methods of induction among 210 women at high risk in a teaching hospital setting. Compared with the amniotomy group, the oxytocin group showed an increased rate of intrapartum complications (adjusted odds ratio 5.1, 1.5 to 17.5). This difference was not seen when the prostaglandin group was compared with the amniotomy group (0.7, 0.3 to 1.6).

Elective induction of labour, although commonly practised in high income countries, remains poorly studied and generally discouraged.42 Even in high income countries labour induction is performed more often in the least capable facilities with no consistent management protocols.43 A recent review suggests 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 hyperstimulation than spontaneous labor.”44 We hypothesise that labour induction is safe when obstetric care is optimal but poses a significant added risk to the fetus when fetal monitoring is suboptimal and uterine monitoring is non-existent, as in Kathmandu.

Conclusion

Both antepartum and intrapartum factors are important in the causation of neonatal encephalopathy in developing countries. In the short term there are likely to be continuing gains from improvements to intrapartum care. Research and development priorities are audit of outcome after oxytocin induction and further investigation of the use of amnioinfusion. Improvements in the public health of women with associated gains in female growth and nutrition must remain a longer term goal.