Study design
The aim of this study was to describe the clinical outcomes of neonates who were potentially exposed to ketamine, diazepam and lidocaine in maternal breast milk, and identify any relationships between anesthetic doses and commonly observed adverse neonatal outcomes. Clinical data extracted from the records of women undergoing postpartum tubal ligation and their infants was explored and analyzed using logistic regression to compare infant outcomes against maternal anesthetic drug doses.
Setting
Records of tubal ligations performed at clinics of the Shoklo Malaria Research Unit (SMRU) between 2013 and 2018 were reviewed. The clinics are located in Tak Province, Thailand, and provide free antenatal, delivery, family planning, and general medical care to rural migrant workers from Myanmar. Contraceptive options include combined oral contraceptives, medroxyprogesterone acetate injection, contraceptive implant, copper intrauterine device and postpartum tubal ligation.
Tubal ligations are performed by qualified medical doctors with appropriate surgical experience. Locally trained skilled birth attendants (training reported elsewhere [27]) support as surgical assistants, take vital signs, and administer medications at the doctor’s request.
There have not been any specific guidelines about breastfeeding after anesthesia at SMRU clinics. Women were encouraged to breastfeed when they were awake enough to do so, usually within 2–4 h following the procedure. During the operation, small amounts of milk were provided (e.g. 10–30 ml) by syringe if babies were crying. Syringe feeding of larger amounts was used based on clinical indications (weight loss or severe jaundice), and was usually maternal breast milk but formula was allowed when maternal breast milk was insufficient.
Available anesthetics were lidocaine, diazepam and ketamine, and most operations were performed with a combination of these three medications. A loading dose of 1 mg/kg ketamine was used for induction, followed by 0.5 mg/kg as needed (usually to 1.5–2.5 mg/kg total dose). Diazepam was available as 5 mg tablets and 10 mg (2 ml) vials, and was usually given as 5 mg per dose. For crude analysis doses were divided into low and high dose categories. Low dose ketamine was defined as < 1.16 mg/kg and high dose was ≥1.16 mg/kg maternal body weight, based on a median dose of 1.16 mg/kg in this cohort. A threshold of 1 mg/10 kg, approximately a 5 mg IV dose, was used for diazepam: low dose IV diazepam was defined as < 1 mg/10 kg, and high dose was ≥1 mg/10 kg maternal body weight.
CSWL was defined as reaching or exceeding the 95th percentile of weight loss for the respective day of life compared to a historical cohort of infants of multiparous women in a similar population [28], the day after the operation. Infants were weighed each morning on Seca scales with 5 g precision.
NH in this report is defined as hyperbilirubinemia requiring phototherapy. From 2008 onwards phototherapy units [29], total serum bilirubin (SBR) measurement and G6PD assessment were available for jaundiced newborns. G6PD deficiency is a major cause of NH in this population and was assessed using the fluorescent spot test. An on-site bilirubinometer measured capillary blood SBR and phototherapy was started when SBR results exceeded thresholds based on the UK NICE guidelines [30]. A prospective study of jaundice at SMRU [31] from 2015 to 16 resulted in screening of SBR and G6PD for all neonates before discharge.
Staff assessed low risk infants daily and high-risk infants 6 hourly; the soonest evaluation after breastfeeding resumed following anesthesia was taken. Parity was defined as number of infants born at an estimated gestational age (EGA) of 28 weeks or more. EGA at delivery was estimated by ultrasound before 24 weeks of gestation, and by Dubowitz exam if an early ultrasound was not available. Apgar scores were calculated by skilled birth attendants at delivery. Blood loss was measured by weight and postpartum hemorrhage was defined as blood loss of more than 1 l. SGA was calculated using Intergrowth-21 standards [32]. Body mass index (BMI) categories followed those recommended for use in Asian populations [33].
Neurodevelopmental data from prospective infant cohort studies [31, 34] was included when available. Inclusion in these studies was not related to contraceptive choice or any of the variables in this study. Neurodevelopment at 1 year was assessed using the locally developed and validated Shoklo Developmental Test [35, 36].
Data extraction
Operation logbooks and notes were both searched to identify eligible women. Women were included if they had term, live, singleton infants who were breastfeeding on the day of the operation and for whom data on the variables of interest was available. Extracted data was entered into a STATA 15 (StataCorp, College Station, TX, USA) data file.
Statistical analysis
Data was analyzed using STATA 15. Chi-squared test was used for comparison of binary variables with binary outcomes. Non-parametric tests of association were used for crude comparisons of non-normally distributed variables: Mann-Whitney U for binary outcomes and Spearman’s correlation coefficient for continuous outcomes.
A causal model was drawn and crude associations with weight loss and NH were investigated before creating the logistic regression models. Logistic regression was used for relationships between drug doses and the binary outcomes (weight loss and NH) and linear regression was used for continuous neurodevelopmental scores. Theoretical and statistical factors were considered and any variable whose inclusion in the model changed the effect estimate by 10% or more was automatically included.
Systemic anesthetic agents (ketamine and both intravenous and oral diazepam) were included in both the weight loss and NH regression models a priori. Lidocaine was excluded from the model due to lack of impact on effect estimates, and missing values (lidocaine dose was not always recorded). For the weight loss model, postpartum hemorrhage, Apgar scores, and EGA were included as a priori confounders. For the NH model maternal factors of Karen ethnicity, pre-eclampsia and obesity, and infant factors of sex and EGA were considered a priori confounders [31, 37].