Birth prevalence of neural tube defects in eastern Africa: a systematic review and meta-analysis

This study is being reported per the reporting guidance provided in the 2020 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement and Meta-analysis of Observational Studies in Epidemiology (MOOSE) [17, 18]. See completed PRISMA checklist in Supplementary Table 1. We searched PubMed (Medline), Embase, and Cochrane Library databases from inception to December 17, 2021. We also searched the grey or difficult-to-locate literature. We performed hand-searching of the reference lists of included studies, relevant reviews, or other relevant documents using Google Scholar. No limitations were imposed based on study design and language. The search terms of interest included Medical Search Headings (MeSH) and keywords—“congenital abnormalities” OR “neural tube defects” OR “anencephaly” OR” encephalocele” OR “spina bifida” OR “meningocele” OR “myelomeningocele” AND [“Uganda” OR “Kenya” OR “Tanzania” OR “Ethiopia” OR “Malawi” OR “Eritrea” OR “Burundi” OR “Comoros” OR “Djibouti” OR “Madagascar” OR “Mauritius” OR “Mayotte” OR “Mozambique” OR “Reunion” OR “Rwanda” OR “Seychelles” OR “Somalia” OR “South Sudan” OR “Zambia” OR “Zimbabwe”]. For a full list of search terms, see Supplementary Table 2. Duplicate studies were initially extracted via Endnote software. Two reviewers (ESH and AES) independently screened titles and abstracts of the studies for inclusion eligibility. The comprehensive list of studies from our initial search was transferred into Endnote, which further removed duplicate studies. The inclusion criteria for this meta-analysis and systematic review was defined as 1) observational studies (cohort, cross-sectional, nested case-control) and randomized controlled trials reporting the incidence or prevalence estimates with and without confidence bounds of neural tube defects (encephalocele, spina bifida, or anencephaly), 2) conducted in human subjects, 3) conducted in Eastern African countries listed above, 4) population-based (all cases in a defined geographical area, or hospital or community-based surveillance). We excluded studies 1) not conducted in human subjects, 2) did not report the rates of NTDs or congenital anomalies, 3) meta-analyses, 4) literature reviews, and 5) commentaries.

After the reviewers initially screened titles and abstracts of potential articles, full-text articles were independently screened by two reviewers (ESH and AE) for eligibility. If duplicate articles were identified, we included only mutually exclusive data. All included articles were scored for methodological quality using the quality assessment tool for NTDs published by Atta et al. (2016) [4]. The GRADE (Grading of Recommendations, Assessment, Development and Evaluation) approach was used to assess the certainty (quality) of the evidence. Discrepancies between the reviewers were resolved by discussion until consensus was reached. However, if an agreement was not reached, a third independent reviewer (PS) was involved in the discussion until consensus was reached. Articles that met inclusion criteria had appropriate data extracted using a standard data collection form. We extracted the following information: the year of publication, country, year of study, study design, sample size, proportion male, the proportion with HIV, number of births with anencephaly, spina bifida, and encephalocele. Incidence and prevalence estimates were all classified as prevalence because, as discussed by Mason et al., prevalence is the preferred measure of the frequency of birth defects epidemiology [19]. Due to high rates of pregnancy loss induced by congenital defects, the observed rates suffer from selection bias and do not fully represent the true incidence.

All statistical analyses were performed with R software version 3.6.2 (R Project for Statistical Computing). The primary outcome was the birth prevalence of NTDs reported per 10,000 total informative births (live births only or live births + stillbirths only or live births + stillbirths + spontaneous abortions). The metaprop function from the R package meta was used to calculate the pooled effect estimates [20]. Meta-analysis was performed with the DerSimonian-Laird random-effects model with Hartung-Knapp-Sidik-Jonkman variance correction [21]. Individual and pooled estimates were graphically displayed using forest plots. A random-effects model assumes the observed estimates of NTDs can vary across studies because of real differences in the effect in each study and sampling variability (chance). Between-study heterogeneity was assessed using I² statistics, expressed as % (low (25%), moderate (50%), and high (75%) and Cochrane’s Q statistic (significance level < 0.05) [22]. To investigate the sources of heterogeneity, we conducted a subgroup analysis estimating NTDs by country in which the study was conducted. Potential ascertainment bias (as might be caused by publication bias) was assessed with funnel plots by plotting the study effect size against standard errors of the effect size and Egger’s test [23].

Our initial searches yielded 465 studies, of which 123 underwent full-text screening (Fig. 1). Of these, 20 studies were eligible for inclusion. Of the included studies, nine were conducted in Ethiopia, [24‐32] three from Tanzania, [33‐35] three from Uganda, [36‐38] three from Kenya, [39‐41] and one from Malawi and Eritrea each [42, 43]. The total sample size of this meta-analysis and systematic review was 752,936 individuals. No studies were identified from the rest of the United Nations region of Eastern Africa—Burundi, Comoros, Djibouti, Madagascar, Mauritius, Mayotte, Mozambique, Reunion, Rwanda, Seychelles, Somalia, South Sudan, Zambia, and Zimbabwe. Wu et al. (2013) and Mosha et al. (2014) were community-based surveillance studies, but the rest were hospital-based surveillance. The median study quality score for studies reporting on the incidence or prevalence of NTDs was 5 out of 6 (range = 4–6). Study-specific details and references are given in Table 1.

The pooled period prevalence of all neural tube defect events per 10,000 observations was 33.30 (95% CI: 21.58 to 51.34). Between-study heterogeneity was high (I² = 97%, p < 0.0001), Fig. 2). To assess country-specific differences in the pooled rates of NTDs, a sub-group meta-analysis stratified by country was conducted (Fig. 3). Ethiopia demonstrated the highest birth prevalence of NTDs, 59.74 (95% CI: 42.10 to 84.71, I² = 96%, p < 0.0001), followed by Eriteria. The rate was lowest in Malawi, 4.69 (95% CI: 2.67 to 8.26, I² = NA). Next, we assess specific NTDs—Spina bifida, anencephaly, and encephalocele. Among the NTDs, spina bifida rate was the highest, 20.47 per 10,000 (95% CI: 12.21 to 34.29, I² = 95%, p < 0.0001, Fig. 4A) followed by anencephaly; 8.66 (95% CI: 4.94 to 15.17, I² = 97%, p < 0.0001, Fig. 4B) and encephaloceles were reported less frequently, 2.33 per 10,000 (95% CI: 1.16 to 4.66, I² = 75%, p < 0.0001, Fig. 4C). Using the median year of study (not publication year), and starting from 1983, the rate of NTDs increased by 4% per year (p = 0.07) (Fig. 5).

To assess the potential for outlier and influential studies affecting the robustness of the pooled estimates, we conducted influence sensitivity analyses for the birth prevalence of NTDs (Supplementary Fig. 1), spina bifida (Supplementary Fig. 2), encephalocele (Supplementary Fig. 3) and anencephaly (Supplementary Fig. 4) separately [44]. In this meta-analysis, a named study was omitted and replaced one study at a time (leave-one-out method) from the meta-analysis and recalculated the pooled estimates for the remaining studies. The estimates remained close to the overall pooled estimate, indicating that no individual study significantly influenced the pooled estimate. The funnel plot and the value of the Egger’s test (p = 0.04) revealed the presence of publication bias in the reporting of the overall NTDs (Supplementary Fig. 5). Trim and fill analysis was conducted to adjust for the potential publication bias. Analyses suggest that the adjusted effect estimates would fall in the range of 33 to 92 per 10,000 births, and 7 additional studies were added to the funnel plots (Supplementary Fig. 6).

Establishing national and regional registries and harmonizing and improving diagnostic criteria would result in a better estimation of the disease burden. Some populations at risk will need special attention, including childbearing age women with epilepsy. They will need to be systematically identified and followed up, preferably by providers with epilepsy-care experience through a well-structured referral system. The prevalence of the use of antiepileptic drugs with high teratogenic potential will need to be established. Additionally, besides benefiting from fortified nutrients, those women will need to take folic supplements peri-conceptually. Finally, for children born with NTDs, prompt referral to specialized centers will need to be made as soon as the disease is detected.