Neonatal and infant mortality associated with spina bifida: A systematic review and meta-analysis

Objectives A systematic review was conducted in high-income country settings to analyse: (i) spina bifida neonatal and IMRs over time, and (ii) clinical and socio-demographic factors associated with mortality in the first year after birth in infants affected by spina bifida. Data sources PubMed, Embase, Ovid, Web of Science, CINAHL, Scopus and the Cochrane Library were searched from 1st January, 1990 to 31st August, 2020 to review evidence. Study selection Population-based studies that provided data for spina bifida infant mortality and case fatality according to clinical and socio-demographical characteristics were included. Studies were excluded if they were conducted solely in tertiary centres. Spina bifida occulta or syndromal spina bifida were excluded where possible. Data extraction and synthesis Independent reviewers extracted data and assessed their quality using MOOSE guideline. Pooled mortality estimates were calculated using random-effects (+/- fixed effects) models meta-analyses. Heterogeneity between studies was assessed using the Cochrane Q test and I2 statistics. Meta-regression was performed to examine the impact of year of birth cohort on spina bifida infant mortality. Results Twenty studies met the full inclusion criteria with a total study population of over 30 million liveborn infants and approximately 12,000 spina bifida-affected infants. Significant declines in spina bifida associated infant and neonatal mortality rates (e.g. 4.76% decrease in IMR per 100, 000 live births per year) and case fatality (e.g. 2.70% decrease in infant case fatality per year) were consistently observed over time. Preterm birth (RR 4.45; 2.30–8.60) and low birthweight (RR 4.77; 2.67–8.55) are the strongest risk factors associated with increased spina bifida infant case fatality. Significance Significant declines in spina bifida associated infant/neonatal mortality and case fatality were consistently observed, advances in treatment and mandatory folic acid food fortification both likely play an important role. Particular attention is warranted from clinicians caring for preterm and low birthweight babies affected by spina bifida.


Introduction
Neural tube defects (NTDs) constitute the largest group of congenital anomalies of the central nervous system [1]; the aetiology of spina bifida is multifactorial [2,3]. Mortality among infants with spina bifida has been previously investigated in several studies, mainly restricted to selected geographical regions especially in high-income countries where data are more available [4].
Infant mortality associated with spina bifida has been changing over time depending on various factors including folic acid supplementation and food fortification programme [5], prenatal screening [6], treatment and termination of pregnancy [7], and the health care system to tract and link all cases with death registers [4]. We conducted a systematic review and metaanalysis of population-based studies focusing on liveborn infants with spina bifida in highincome countries.
The aims of this study were to assess: (1) spina bifida-specific neonatal and infant mortality rates over time; and (2) the socio-demographic and clinical factors associated with mortality in the first year for infants affected by spina bifida.

Methods
The methods for the overall systematic review have been registered as a review protocol in the International Prospective Register of Systematic Reviews (PROSPERO) [8], registration number CRD42018081353.

Search strategy
Comprehensive literature searches of PubMed, Embase, Ovid, Web of Science, CINAHL, Scopus and the Cochrane Library was performed from 1 st January, 1990 to 31 st August, 2020. MeSH terms and keywords which included neural tube defect or congenital brain malformation or abnormalities or spina bifida and infant or neonatal or perinatal and mortality or death or survival etc. were entered systematically into the databases. The detailed search terms are shown in S1 Table. There were no language restrictions applied. Manual searches of reference lists were performed on all the included publications. Attempts were also made to contact the study authors for any relevant unpublished data where appropriate.
Titles and abstracts of all identified studies were screened by PH according to the inclusion criteria. For articles which satisfied the first screen, full articles were then screened independently by two reviewers (PH and DT/ CB) for their eligibility according to the full inclusion and exclusion criteria. Any discrepancies were reviewed by the author review team (PH, MQ and JK) to reach a final decision.

Eligibility criteria
Population-based studies were included if they (1) ascertained all individuals born with spina bifida in a predefined population; (2) provided data for spina bifida infant mortality (defined as the number of spina bifida associated deaths under one year of age occurring among all livebirths in a given population) or case fatality (defined as the proportion of spina bifida associated deaths among all spina bifida cases under one year of age in a given population), or case fatality estimates according to clinical or socio-demographical characteristics; and (3) were conducted in high-income countries as defined by the World Bank [9].
Studies were excluded if: (1) they were conducted solely in tertiary or referral centres; and (2) spina bifida-affected individuals were not followed up from birth. Spina bifida occulta (i.e. a mild form of spina bifida with no clinical consequence) or spina bifida related to a syndrome were excluded where possible.

Data extraction
PH conducted the literature searches with the assistance of a specialist information librarian, screening of abstracts and review of 368 eligible full articles. DT and CB each reviewed 50% of the eligible full articles, validated decisions about the final included articles and extracted data.
The following data were extracted: study population, study design, sample size, country of study, and descriptive/ mortality data about maternal age, maternal ethnicity, smoking, maternal education, previous live births, prenatal care, induction of labour, mode of delivery, infant sex, gestational age, birthweight, year of birth, plurality, lesion level, the presence of multiple defects, hydrocephalus and major cardiac defects and period of folic acid fortification.
Spina bifida mortality rates, case fatality and the corresponding 95% CI were extracted at the age of one month and one year. If the 95% CI of the mortality rates were not reported, they were estimated using 'binomial exact', assuming no cases were censored. In studies where 95% CI of relative risks (RR) of factors for mortality were not reported, they were estimated using log RR ± 1.96 × Standard Error, where standard error was derived from the counts and proportions.
Statistical analysis. As all the pooled estimates for mortality risk factors rate ratios came from two studies [10][11][12]23], both random-effects and fixed effects models were used to calculate and compare the pooled estimates. These studies were conducted during pre-mandatory folic acid food fortification period, which justified the decision for the effects pooling. Heterogeneity between studies was assessed using the Cochrane Q test and I 2 statistics, p < 0.10 was considered as statistically significant with I 2 > 50% indicating a substantial level of heterogeneity. Random-effects meta-regression was used to assess the effect of year of birth. Results of meta-regressions are presented in bubble plots, where the weights used to determine the bubble size are the inverses of the effect-size variances and hence the size of the bubble is proportional to the precision of each study. The slopes of the meta-regression lines were calculated, with p < 0.05 considered as statistically significant.
All data analyses were performed using Stata version 13 (StataCorp). The reporting of results was in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram and Meta-analysis of Observation Studies in Epidemiology (MOOSE) checklist [24,25].
Quality appraisal. Risk of bias (RoB) assessment was conducted using a refined quality tool based on Quality in Prognostic Studies (QUIP) [26] for the following domains: study participation, measurement of outcome(s), study attrition, measurement of exposure or prognostic factor(s) and statistical analysis and reporting. Each domain was rated either as 'high', 'moderate' or 'low' RoB by two independent reviewers (PH and DT/CB). The overall RoB for each included study was assessed using the method described by Hayden et al. [27] in line with the Cochrane Risk of Bias tool [28].

Results
The flow of literature in this review is summarised in the PRISMA diagram (Fig 1). Of 5,023 articles identified, 20 met the inclusion criteria.

Quality appraisal
Results of the quality appraisal of evidence are summarised in Fig 12. The majority of the included studies satisfied the study population domain as they were national or register-based. The exposure/ prognostic factors measurement domain was satisfied in about half of the included studies, this was mainly due to the differences in treatment and preventative Table 2. Neonatal and infant mortality rates from spina bifida.

Mid-year of birth cohort
Population size (no. of livebirths)

Spina bifida
Wen (     interventions available over time and between countries of the studies, as well as variations between studies in reporting syndromal and multiple defects which is a major prognostic factor. The statistical analysis and reporting domain was satisfied only in one-fifth of the included studies due to a combination of concurrent cases [11,38], missing data, unreported 95% CI [12,29], and the potential for incomplete ascertainment of deaths particularly in early surveillance programmes [11,30,35,41]. In addition, some studies did not perform mortality analysis but only reported the number of infants alive or dead, in which cases censoring was not accounted for. Infant mortality and case fatality estimates may have been underestimated in these studies.

Discussion
This systematic review included 20 population-based cohort studies with a total study population of over 30 million liveborn infants, in which approximately 12,000 liveborn infants were affected by spina bifida; outcomes were reported based on generally high-quality evidence from studies mainly with a low to moderate risk of bias. The review spanned 49 years with births from 1966 to 2015. Variation in study period is perhaps the main source of heterogeneity for the mortality estimates. This is largely indicative of different clinical and interventional exposures to which the patient populations were subject to over time. Spina bifida infant mortality encompasses both to the chance of being born with spina bifida as well as surviving with the condition during infancy (i.e. case fatality), both of which have significantly decreased over time. There are several possible mechanisms including: First, primary prevention: Five large population-based US cohort studies in this review have consistently shown a significant decline over time in spina bifida infant mortality post mandatory folic acid fortification of U.S. grain supply introduced in September 1998 [3,13,30,[33][34][35][36][37][38][39][40][41][42][43][44][45]. This finding is consistent with long term surveillance data from the U.S., Canada, Chile and Costa Rica where folic acid food fortification had been successfully implemented, showing that the NTD birth prevalence can be reduced over time to as low as 5-6 per 100,000 pregnancies [9,46]. Indeed, the original large (n = 1,817 women at high risk of having a pregnancy with a NTD) international randomised double-blind prevention trial conducted by the MRC Vitamin Study Research Group showed a 72% protective effect against the occurrence of an NTD (i.e. spina bifida, encephalocele and anencephaly) pregnancy for women taking 4 mg folic acid supplementation daily prior to conception [5]. Similarly, in a large population-based cohort study, Berry et al. (1999) have shown that periconceptional intake of folic acid reduce the risk of NTDs in areas with variable prevalence rates in China [47]. These findings established the specific role of folic acid supplementation in the prevention of NTDs.
Second, the studies by Kalucy et al. (1994) [23], Riley et al. (1998) [35], Wen et al. (1999) [32] and Davidoff et al. (2002) [3] have shown that spina bifida infant mortality and case fatality declined over time prior to national implementation of folic acid food fortification in Australia and the U.S‥ This finding suggests that improved technology in and access to prenatal screening and diagnosis over time in high-income countries have led to significant declines in both late fetal death and birth of spina bifida affected infants as a result of early prenatal diagnosis and termination of pregnancy [23,48]. According to EUROCAT data, over 50-75% of cases of spina bifida are terminated in Europe [49].
Possible explanations for the decline in spina bifida infant case fatality over time include: First, folic acid food fortification: This review has shown a significant decline over time in spina bifida infant case fatality post mandatory folic acid fortification of the U.S. grain supply [3,13,30,33,34]. Infants born with a spina bifida during the period of mandatory fortification were almost one-third less likely to die in infancy compared to infants born with a spina bifida during pre-fortification period [11]. Studies suggest that folic acid food fortification, may reduce the risk of severe types of spina bifida by moving the lesion caudally along the developing spine [11,[50][51][52], as upper-level spina bifida is associated with more severe anomalous

PLOS ONE
brain development and more system wide complications compared with lower-level spina bifida [42,53]. Second, advances in medical and surgical interventions: Studies have suggested that improved ventilator support in neonatal intensive care and the use of antibiotics to treat CNS infections have provided better treatment of spina bifida and associated co-morbidities such as prematurity and low birthweight which have been shown as the strongest predictors for infant deaths in this review [54]. In addition, advances in surgical treatment for spina bifida and associated hydrocephalus and cardiac defects may also have contributed to the reductions in case fatality [31,50,[55][56][57][58][59][60].
Regarding other risk factors, mortality among Black and Hispanic infants with spina bifida has declined with time but remains consistently higher when compared to White infants [12]. Possible explanations include barriers to access to health care particularly among women of lower socioeconomic classes [12,61,62]. Hispanic populations also have a high incidence of C677T methylenetetrahydrofolate reductase (MTHFR) homozygous genotype mutations [62,63] which have been shown to be associated with upper level spina bifida defects [64]. Fetuses with an anomaly often have other developmental effects during gestation which may also lead to preterm birth [65,66]. Therefore, preterm birth and congenital anomalies are likely to have a multiplicative effect on infant mortality. Multiple anomalies are an indicator of disease complications and severity, and hence tend to be associated with poorer prognosis when major systems are involved, compared to isolated anomalies [43,67].
The main strengths of this review include its comprehensive search strategy with no language restriction, robust inclusion criteria and complete coverage of liveborn spina bifida population-based studies. Unpublished relevant data were obtained from study authors and included in the analysis where possible. A further strength is the generally high quality of the included studies with the majority of studies rated as low to moderate risk of bias according to the refined QUIP and Cochrane RoB tool [26,28].
There are several limitations. First, mortality in infants with spina bifida depends on various factors including health care systems, treatment, termination of pregnancy and ability to track and link all cases with death registers. Without the availability of these data, it is not possible to determine the true underlying cause of the reduction in infant case fatality and mortality rates. In addition, we could not verify whether multiple defects and syndromal spina bifida were included or excluded in the majority of the included studies. Consequently, these potentially caused high variability and heterogeneity in the estimates. Second, study data about risk factors for spina bifida case fatality were limited; in particular, data about women intake of folic acid supplement was not available in most of the included studies, which is an important factor to account for. Finally, cases of multiple defects and syndrome associated with spina bifida cannot be completely excluded from the analysis as they cannot be reliably identified in most of the included studies.

Conclusions
This study has reviewed the evidence about spina bifida infant mortality in high-income countries. This study has shown that a decline in spina bifida associated infant/neonatal mortality and case fatality were consistently observed, in which advances in medical and surgical treatment, and mandatory folic acid food fortification likely to play an important role. Preterm birth and low birthweight are strongest risk factors associated with increased case fatality of infants with spina bifida, which warrant particular attention from clinicians caring for these vulnerable babies.