Ethiopia draws drinking water from boreholes, wells, springs, rivers and lakes, yet the metal concentrations reported across this literature have never been pooled. We searched PubMed/MEDLINE, Europe PMC, OpenAlex and Crossref from inception to 20 September 2026 and screened records in duplicate. Study-level means were pooled by random-effects meta-analysis of log concentrations, guideline exceedance by logit-transformed proportions, and certainty rated with GRADE adapted for environmental evidence. Of 44 studies with extractable concentration data, 31 reported the information needed for random-effects pooling; pooled concentrations exceeded drinking-water benchmarks for 7 of 9 metals, lead reaching 221.8 µg/L (95% CI 70.3–699.5) and cadmium 120.0 µg/L (95% CI 24.8–581.6). More than half of sampled sites exceeded the benchmark for 6 of 8 metals. Heterogeneity was extreme (I² > 98%) and certainty very low throughout. Translated into burden, lead-attributable cognitive loss was 1.4 IQ points per exposed child even where water merely meets the 10 µg/L guideline, and 7.3 points at the pooled concentration; the carcinogens implied 1,712 disability-adjusted life years (DALYs) per 100,000 exposed as a lower bound, and the annual economic loss spanned 0.21–3.65% of GDP across exposure scenarios. This literature is not a probability sample of Ethiopian water: it is concentrated on sources near known pollution, reports dispersion inconsistently, and at the pooled lead concentration implies a population blood lead that is not credible. The pooled values therefore characterise the sources investigators selected, not national exposure. A probability-based national monitoring programme, with mandatory reporting of sample size, dispersion and detection limits, is the prerequisite for turning this evidence into policy.
