Lead (Pb) contamination in drinking water in low- and middle-income countries: a systematic review and meta-analysis

Michael B. Fisher , Timothy Purvis , Zakariah Seidu , Siddhartha Roy , Michelle Cawley, Carrie Baldwin-SoRelle , Ryan D. Cronk , Aaron B. Salzberg , Amy Z. Guo, J. B. Tracy , Emily Browning, Jamie K. Bartram

Lead (Pb) in drinking water causes organ damage, cardiovascular disease, cancers, and lifelong neurological and developmental impairment, disproportionately harming infants and developing fetuses. However, evidence on lead in drinking water in low- and middle-income countries (LMICs) is limited and lacks robust synthesis, impeding action. We conducted a systematic review and meta-analysis of peer-reviewed studies reporting lead in drinking water in LMICs, according to PRISMA guidelines.
We searched PubMed, Scopus, and Web of Science for studies published in English since 1969 reporting lead concentrations in drinking water. Study relevance was ranked using supervised clustering and machine learning. Relevant studies were manually screened for inclusion; data were manually extracted from included studies. Within-study risk of bias was scored using quality items defined in this work. Between-study bias was assessed based on continuity and symmetry of the (roughly lognormal) distribution of data included in the review. 
Of approximately 40,000 search results (for both lead and other TMs), 16,868 scored relevant using a trained machine learning algorithm; of these, 3,187 met inclusion criteria. Approximately one third (n=1,088) reported on lead. Central and Southern Asia accounted for 45% of included datasets, while 31% of LMICs were unrepresented. Many studies (45%) had sites purposively selected for known or suspected contamination; after excluding such “targeted” studies, metaregression indicated that >20% (95% CI: 20-27%) of samples exceeded World Health Organization (WHO) guideline values for lead (n=523).
Potential sources of bias within studies include sample collection and analysis limitations (lower quality studies) and purposive site selection (“targeted” studies); potential sources of bias among studies include language limitations, exclusion of grey literature, geographic heterogeneity of datasets, and potential publication bias. This work suggests that lead contaminates drinking water at levels of health concern in LMICs worldwide, and that increased collective efforts to prevent, manage, and monitor such contamination are merited.