ELBW and ELGAN outcomes in developing nations-Systematic review and meta-analysis.

CONTEXT
Morbidity and mortality amongst extremely low birth weight (ELBW) and extremely low gestational age neonates (ELGANs) in developing nations has not been well studied.


OBJECTIVES
Evaluate survival until discharge, short- and long-term morbidities of ELBW and ELGANs in LMICs.


DATA SOURCES
CENTRAL, EMBASE, MEDLINE and Web of Science.


STUDY SELECTION
Prospective and retrospective observational studies were included.


DATA EXTRACTION AND SYNTHESIS
Four authors extracted data independently. Random-effects meta-analysis of proportions was used to synthesize data, modified QUIPS scale to evaluate quality of studies and GRADE approach to ascertain the certainty of evidence (CoE).


RESULTS
192 studies enrolling 22,278 ELBW and 18,338 ELGANs were included. Survival was 34% (95% CI: 31% - 37%) (CoE-low) for ELBW and 39% (34% - 44%) (CoE-moderate) for ELGANs. For ELBW neonates, the survival for low-income (LI), lower middle-income (LMI) and upper middle income (UMI) countries was 18% (11% - 28%), 28% (21% - 35%) and 39% (36% - 42%), respectively. For ELGANs, it was 13% (8% - 20%) for LI, 28% (21% - 36%) for LMI and 48% (42% - 53%) for UMI countries. There was no difference in survival between two epochs: 2000-2009 and 2010-2020. Except for necrotising enterocolitis [ELBW and ELGANs-8% (7% - 10%)] and periventricular leukomalacia [ELBW-7% (4% - 11%); ELGANs-6% (5%-7%)], rates of all other morbidities were higher compared to developed nations. Rates of neurodevelopmental impairment was 17% (7% - 34%) in ELBW neonates and 29% (23% - 37%) in ELGANs.


LIMITATIONS
CoE was very low to low for all secondary outcomes.


CONCLUSIONS
Mortality and morbidity amongst ELBW and ELGANs is still a significant burden in LMICs. CoE was very low to low for all the secondary outcomes, emphasizing the need for high quality prospective cohort studies.


TRIAL REGISTRATION
PROSPERO (CRD42020222873).

Introduction Prematurity is one of the leading causes of childhood mortality, rates of which are on the rise across the globe [1][2][3][4]. Available data indicate that about 80% of the preterm births happen in the geographical regions of Africa and South Asia [2,3]. When countries are grouped by their World Bank income categories, it is found that approximately 90% of all preterm births occurs in low-and middle-income countries (LMICs). The average preterm birth rate for LMICs is close to 9-12%, compared to approximately 9% for most of high-income countries (HICs) [4].
The quantum of impact on decreasing the neonatal mortality rate (NMR) is more pronounced when community based interventions targeting the late preterm and term neonates are implemented successfully across LMICs [1]. It is estimated that about 80% of neonatal deaths can be thwarted in LMICs by achieving a 95% coverage rate of simple neonatal interventions such as neonatal resuscitation, avoiding hypothermia, clean cord practices, kangaroo mother care as well as supporting breast feeding, clubbed with antenatal practices [1]. Henceforth, higher level NICU care targeting extremely low birth weight (ELBW) neonates with a birth weight of less than 1000 grams and extremely low gestational age neonates (ELGANs) born at less than 28 weeks' gestation might not be a priority as of now in many LMICs.
Every Newborn action plan by WHO envisages reducing national NMR to 12 per 1000 live births by the year 2030 [5]. Historical trends of NMR from developed nations reveal that once a NMR of 15 per 1000 live births is attained, augmenting community based health programmes with facility based neonatal care through establishment of higher level NICUs would facilitate achieving single digit NMR in LMICs [6]. Improved survival rates amongst ELBW and ELGANs in LMICs with such an approach might also be associated with increased rates of typical morbidities of prematurity, including bronchopulmonary dysplasia (BPD), retinopathy of prematurity (ROP) and long term neurodevelopmental impairment (NDI) [7][8][9][10]. While previous systematic reviews have tried to quantify the burden of mortality and morbidity in high risk neonates including preterm infants from LMICs, none have been published till date with exclusive focus on the particularly vulnerable group of ELBW neonates and ELGANs [11][12][13]. Assessing the burden of mortality and morbidity in these infants is vital in developing future newborn targeted health interventions and policies, especially for those LMICs that are transitioning from community based care to a facility based one. Hence, this systematic review and meta-analysis was conducted to study the survival, short-and long-term outcomes of ELGANs and ELBW neonates from LMICs.

Methods
The protocol for the systematic review was registered with PROSPERO (CRD42020222873) [14]. The reporting of the review is in accordance with PRISMA guidelines [15].

Literature search
Electronic databases: MEDLINE, EMBASE, Web of Science and Cochrane CENTRAL were searched from 1 st January 2000 till 21 st November 2020. There were no language restrictions. Google translate (California, USA) was used for translating non-English literature. Studies published in Urdu and Persian were translated with the help of a translator as there were some limitations in Google translate for these languages. Studies published as abstracts were also eligible for inclusion. Only published data was used in this systematic review. Four authors (TA, NBS and PB, DN) conducted the literature search independently in pairs of two using Rayyan-QCRI software [16]. The search strategy for the different databases is provided in S1 Table in S1 File.

Inclusion criteria
Retrospective as well as prospective observational studies that had reported on outcomes of ELGANs (born at less than 28 weeks' of gestation) and / or ELBW neonates (birth weight of less than 1000 grams) from a LMIC as per the World Bank country classifications by income levels (2020) were eligible for inclusion [17]. Randomized controlled trials were excluded as certain ELGANs and ELBW neonates who would otherwise have been eligible for inclusion might have been excluded due to a variety of reasons such as stringent inclusion criteria, refusal of consent and presence of co-morbid conditions among others.

Outcomes
The primary outcome of the review was proportion of neonates who had survived until discharge. Secondary outcomes included prevalence of intraventricular hemorrhage (IVH) (> grade II) [18], periventricular leukomalacia (PVL) (any severity), necrotizing enterocolitis (NEC) (stage II or more) [19], PDA requiring surgical or medical treatment, BPD [oxygen requirement at 36 weeks' postmenstrual age (PMA)], sepsis (early onset and late onset sepsis) diagnosed based on blood culture as well as on other blood markers, ROP (any ROP, ROP stage � 2 as per ICROP classification [20], ROP requiring laser therapy or intravitreal bevacizumab), extrauterine growth retardation (EUGR) (defined as weight less than 10 th centile at 36 weeks' PMA), any NDI (as defined by authors) and cerebral palsy of any severity, both assessed at 18-24 months' corrected age.

Risk of bias (ROB) assessment
The risk of bias assessment was done using a modified QUIPS scale [21]. Four parameters namely, representativeness of the sample, definition and assessment of the outcomes, evaluation of baseline characteristics of the enrolled subjects known to affect the outcomes and method of data collection were evaluated. Studies that had satisfied all the four criteria were classified as having a low risk of bias. Those fulfilling two or three criteria were adjudged as having an intermediate risk of bias and those satisfying one or none were classified as high risk of bias studies. Two authors (TA, NBS) evaluated the risk of bias independently. Disagreements were resolved by consulting a third author (VVR).

Data collection and synthesis
The data for relevant outcomes were extracted using a pre-specified proforma by two authors independently (PB, DN). Statistical analysis was done using the R-Software. Meta-analysis of proportions was used for synthesis of data. Raw data was used if the proportions were between 0.20 to 0.80. Otherwise, data was logit transformed. Freeman-Tukey Double arcsine transformation of data was preferred over logit transformation when there were many proportions with 0 or 1 values. 'Metaprop' package in R-software was used for meta-analysis of proportions. 'Metaprop' can analysis meta-analysis of raw data based on proportions as well as convert raw data to logit transformation or Freeman-Tukey Double arcsine transformation and re-transform to proportions by default for meaningful interpretation. A random effects model was chosen as significant heterogeneity was anticipated as reported in prior studies on newborn survival [22]. An inverse variance method was chosen with DerSimonian Liard estimator being used for assessing between study variance [23]. The final estimates were expressed as proportions with 95% confidence interval.

Certainty of evidence (CoE) assessment
CoE was assessed using a modified GRADE approach [24]. Outcomes from prospective studies started as high quality evidence and other type of studies as low. The parameters of risk of bias, inconsistency, imprecision, indirectness and publication bias were evaluated for downgrading the evidence further. I 2 value was used to adjudge significant heterogeneity. Imprecision was evaluated based on the point estimate and the 95% confidence interval. If these were to cross a decision threshold to intervene, the CoE was downgraded by one level for imprecision. Though publication bias assessment in meta-analysis of prognosis studies is a contentious topic, we assessed publication bias using funnel plots and Begg's rank test based on GRADE working group guidelines [24,25].

Sensitivity analyses
The following sensitivity analyses were performed for the primary outcome • Categorizing studies from countries based on three income levels-low income (LI), lower middle-income (LMI) and upper middle-income (UMI).
• Categorizing studies based on geographical regions where they are situated.
• Excluding studies with low sample size (less than 50 subjects). • Analyzing studies which had evaluated neonates with varying baseline sickness.

Results
A total of 21,535 studies were identified from the literature search, of which 2157 full texts were assessed for eligibility after removal of duplicates as well as title, abstract screening and 192 studies (ELBW neonates-22,278; ELGANs-18,338) were included in the final synthesis (S1 File references

Primary outcome
1a. Survival until discharge for ELBW neonates. The overall survival was 34% (95% CI: 31% -37%). There was significant heterogeneity in the survival rates between studies. Subgroup analysis to address the heterogeneity was done by analyzing studies based on the income status, region and country of origin. While there were widely varying survival rates between countries of similar economic status as well as region of origin, it was much lesser when studies were analyzed according to the country of origin.
For ELBW neonates, the survival for LI, LMI and UMI countries was 18% (11% -28%), 28% (21% -35%) and 39% (36% -42%), respectively (S1 and S2 Figs in S1 File). While most of the countries from the African subcontinent had a survival rate of less than 20%, single center studies from Benin, Eritrea, South Africa and Tanzania reported better survival rates of more than 40% for ELBW neonates. While single center studies from China and India had reported survival rates of more than 60%, it was much lesser from the other Asian countries. While the countries from South America had a reported survival of 39% -48%, the only middle-income country from North America, Mexico had a survival rate of 43% (37% -49%). Amongst the countries from the Middle East, Turkey had the highest survival rate of 43% (31% -57%). Country-wise survival outcomes for ELBW neonates is given in Fig 2. Publication bias was detected for this outcome (S3 Fig in S1 File) 1b. Survival until discharge for ELGANs. Survival until discharge was 39% (34% -44%). It was 13% (8% -20%) for LI, 28% (21% -36%) for LMI and 48% (42% -53%) for UMI countries. Similar to ELBW survival, there was considerable variation between countries of similar economic classification and geographic region (S4-S6 Figs in S1 File) Single-center outlier studies from Benin, China, South Africa, Thailand and Turkey had reported relatively higher survival of more than 60% when compared to other countries (Fig 3). The survival as assessed by synthesizing data from multiple studies was highest in China [61% (53% -68%)].
NEC stage II or more. The rates of NEC were uniform across countries with its prevalence being 8% (7% -10%) for ELBW neonates as well as ELGANs. Publication bias was detected for studies reporting on NEC in ELBW neonates (S32-S35 Figs in S1 File).
Data related to ROP is given in S37-S48 Figs in S1 File. Sensitivity analyses. 4a. Excluding studies with low sample size. Analysis of survival outcome after excluding studies with low sample size did not result in any significant changes in the effect estimate when compared to the primary analysis for both ELBW neonates and ELGANs (S49 and S50 Figs in S1 File). There was no significant difference in the survival rates between the two epochs for ELBW neonates (Test of moderators p value-0.83) and ELGANs. (Test of moderators p value-0.78) (S51 and S52 Figs in S1 File).
4c. Analysing studies which had evaluated neonates with varying baseline sickness. While of most of the studies were single centre studies, survival rates were reported by five studies for

Certainty of evidence
CoE for the primary outcome measure of survival until discharge for ELGANs was moderate and for ELBW neonates was low. While most of the included studies reporting on the primary outcome measure were prospective in design and started with high CoE, they were downgraded for serious inconsistency by one level. The outcome of survival until discharge for ELBW neonates was further downgraded by one level for publication bias. The CoE for other secondary outcomes were very low to low, with retrospective study design and heterogeneity being reasons for downgrading the CoE. The CoE is given in Table 2.

Discussion
This systematic review and meta-analysis included 192 studies (ELBW neonates-22,278; ELGANs-18,338) from LMICs situated in different geographical regions across the globe. To the best of our knowledge, this is the only systematic review evaluating survival and morbidities in ELGANs or ELBW neonates from LMICs. The primary outcome of the review was the proportion of neonates who had survived until discharge. The results of this study indicate that the overall survival until discharge of ELBW neonates was 34% and ELGANs was 39%, with significant heterogeneity between studies based on the income status, region as well as country of origin. These survival rates are much lower than that reported from HICs [22]. Myrhaug et al. in their meta-analysis (2019) had reported a survival rate of 74% at 25 weeks' and 90% at 27 weeks' amongst neonates born alive in HICs [22]. Such stark differences in survival between LMICs and HICs can be explained by many reasons. Attitudes towards providing life-saving intervention to these immature infants are quite different between LMICs and HICs. In LMICs, the focus of healthcare programmes to reduce NMR comprise predominantly of cost-effective high impact interventions comprising of early essential newborn care services which are predominantly targeted to address mortality in relatively bigger neonates [6,26]. Decreasing mortality rates in ELGANs in LMICs through establishment of higher level NICUs might impose further financial burden as well as risk of inequity, diverting resources from the relatively more mature neonates. Further, absence of healthcare insurance schemes in LMICs result in financial constraints for parents who eventually have to bear most of the costs incurred [27]. Finally, 'denominator bias' due to lack of surveillance data and consequent inconsistent reporting from LMICs regarding live births versus NICU admissions might also result in significant variability in survival rates in LMICs when compared to HICs [28,29].
There was significant variation in survival between LMICs as well. Similar findings are also seen in the studies published from HICs [30]. While some studies from HICs have pointed out factors such as differing gestational age cut-offs for instituting active care, selective versus comprehensive care and variations in clinical care practices for babies born at less than 25 weeks' gestation being some of the reasons for varying survival rates between different countries, such comparative studies are lacking for LMICs [31][32][33]. Differing survival rates between studies from the same LMIC might be explained by the survival gap between rural and urban areas [34]. Whether there are any differences in survival between private versus public sector systems in LMICs still remains a contested topic, it was beyond the scope of our review to look into the same [35]. Differences in health policy, financial resources, access to and use of health  [22]. However, some neonatal network studies from other HICs have shown both an improving as well as static trend for survival in the past few decades [39][40][41][42][43]. The rates of some secondary outcomes such as NEC and PVL were comparable with data from HICs [40][41][42]. This might be due to the fact that many of the HICs provide active care for neonates born from 22 weeks' gestation who are at the highest risk of major brain injury, whilst most of the ELBW neonates and ELGANs enrolled in studies from LMICs were of higher gestational ages ranging from 25-27 weeks [31,33,40,44]. However, rates of all other morbidities were higher in LMICs when compared to HICs. Information from limited studies in the meta-analysis have shown that the rates of NDI was 17% (7% -34%) and 29% (23% -37%) in ELBW neonates and ELGANs, respectively. Milner et al. in their systematic review and meta-analysis of long-term neurodevelopmental outcomes of preterm VLBW neonates in resource limited settings had reported a lower NDI rate of 21.4% (11.6% -30.8%) [12]. This could be because of the fact that Milner et al.'s study consisted of neonates who were gestationally more mature. Also, many studies from HICs such as Croatia, Korea, Poland and Taiwan were included in their study.
This systematic review had some limitations. There was considerable heterogeneity in the various outcome measures despite performing multiple sensitivity and sub-group analyses to address the same. The analysis might be limited by denominator bias as survival rates could not be assessed based on different denominators such as live births versus NICU admissions, due to inconsistent reporting in studies. Accurate gestational age assessment is still a major issue in LMICs and might have influenced our final estimates. Further, analysis by stratification on the basis of gestational age or birth weight could not be performed for ELGANs and ELBW neonates. Important confounders determining survival such as antenatal corticosteroid use, multiple gestation, SGA status, gender, chorioamnionitis and level of NICU care were not reported by majority of the included studies, thus precluding any sensitivity analysis based on these parameters. The overall CoE was very low to low for all the secondary outcomes. Finally, though the literature search included the standard databases as recommended by the Cochrane group, studies from LMICs might be more likely to be published in alternative databases as well.

Conclusion
Mortality and morbidity of ELBW neonates and ELGANs is still a huge burden in LMICs, with significant differences in their occurrence between countries of similar economic status and geographical region of location. For ELBW neonates, the survival for LI, LMI and UMI countries was 18% (11% -28%), 28% (21% -35%) and 39% (36% -42%), respectively. For ELGANs, it was 13% (8% -20%) for LI, 28% (21% -36%) for LMI and 48% (42% -53%) for UMI countries. These are significantly lower than the survival rates reported from HICs. The CoE for most of the outcomes was very low to low, emphasizing the need for surveillance and high quality prospective cohort studies from LMICs on this sub-group of vulnerable neonates. Such studies should provide data related to still births, live births, delivery room deaths, important baseline characteristics such as antenatal corticosteroid use, multiple gestation status, SGA, gender, chorioamnionitis and level of NICU care, and sub-group data on different gestational ages or birth weights for ELGANs or ELBW neonates. Such data would not only quantify the burden of mortality and morbidity in these preterm infants, but also enable evaluating the post-implementation impact of important public health interventions.