1. Introduction
Countries with widespread blood lead level (BLL) testing have documented that average BLLs declined precipitously following the elimination of leaded gasoline, with ranges reported from 30–48% within 5 years of phaseouts (Fewtrell et al., Reference Fewtrell, Kaufmann and Prüss-Üstün2003). Nonetheless lead remains common in the environment and many children are exposed at levels known to cause adverse effects.
Lead is a potent neurotoxicant that adversely impacts most of the body’s enzyme systems (Flora et al., Reference Flora, Gupta and Tiwari2012). There is no known safe BLL for children, who are particularly vulnerable. Children seldom exhibit symptoms at the time of exposure and the many long-term impacts, including decreased Intelligence Quotient (IQ) and behavioural challenges limiting academic and professional success, cannot be recognised until later in life (Breit et al., Reference Breit, Brunner and Preckel2021). The World Health Organization (WHO) has estimated that lead exposure accounts for 30% of the global burden of idiopathic developmental intellectual disability (WHO, 2021).
The economic cost of paediatric lead exposure can be conservatively quantified as the effect of reduced cognitive capacity on lifetime economic productivity (Grosse et al., Reference Grosse, Matte, Schwartz and Jackson2002; Landrigan et al., Reference Landrigan, Fuller, Acosta, Adeyi, Arnold, Basu, Baldé, Bertollini, Bose-O’Reilly, Boufford, Breysse, Chiles, Mahidol, Coll-Seck, Cropper, Fobil, Fuster, Greenstone, Haines, Hanrahan, Hunter, Khare, Krupnick, Lanphear, Lohani, Martin, Mathiasen, McTeer, Murray, Ndahimananjara, Perera, Potočnik, Preker, Ramesh, Rockström, Salinas, Samson, Sandilya, Sly, Smith, Steiner, Stewart, Suk, van Schayck, Yadama, Yumkella and Zhong2017). Lead exposure impairs the development of grey matter in the brain and reduces IQ. Effect size estimates vary widely; loss of a single IQ point has been quantified as a 0.5–2.5% reduction in lifetime earning capacity (Grosse and Zhou, Reference Grosse and Zhou2021). Lead’s impact is proportionately greater at lower levels of exposure (Crump et al., Reference Crump, Van Landingham, Bowers, Cahoy and Chandalia2013; Lanphear et al., Reference Lanphear, Hornung, Khoury, Yolton, Baghurst, Bellinger, Canfield, Dietrich, Bornschein, Greene, Rothenberg, Needleman, Schnaas, Wasserman, Graziano and Roberts2005; Budtz-Jørgensen et al., Reference Budtz-Jørgensen, Bellinger, Lanphear, Grandjean, Lanphear, Hornung, Khoury, Yolton, Baghurst, Bellinger, Canfield, Dietrich, Bornschein, Greene, Rothenberg, Needleman, Schnaas, Wasserman, Graziano and Roberts2013). For example, BLLs of 5 and 15 µg/dL have been associated with decrements of 2.9 and 5.5 IQ points, respectively (Budtz-Jørgensen et al., Reference Budtz-Jørgensen, Bellinger, Lanphear, Grandjean, Lanphear, Hornung, Khoury, Yolton, Baghurst, Bellinger, Canfield, Dietrich, Bornschein, Greene, Rothenberg, Needleman, Schnaas, Wasserman, Graziano and Roberts2013). However, deficits continue at higher levels of exposure, with BLLs ≥ 65 µg/dL being associated with major intellectual disability, seizure disorders and death (ATSDR, 2020).
Two studies since 2010 have calculated productivity losses attributable to paediatric lead exposure in low- and middle-income countries (LMICs). First, Attina and Trasande (Reference Attina and Trasande2013) evaluated a one-year cohort of < 5-year-old children, finding total lifetime losses of international $ 977 billion (2011 Purchasing Power Parity [PPP]) (Attina and Trasande, Reference Attina and Trasande2013). Ten years later, relying on a larger BLL dataset, Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023) found substantially higher losses, totalling approximately international $ 1.61 trillion (2019 PPP) (Larsen and Sánchez-Triana, Reference Larsen and Sánchez-Triana2023). Here we develop a third valuation of lead-attributable productivity losses in LMICs. We utilise BLL data from a substantially larger systematic review than previous studies, comprised of 388 unique publications. For context, the datasets used by preceding efforts comprised fewer than 100 studies. Further, we interpolate BLL data for countries where blood lead information was unavailable, providing a larger dataset of country-specific BLLs. We then provide a comparison of our results with preceding studies to demonstrate the significance of major assumptions and data gaps.
2. Methods
The published literature on BLLs in LMICs was reviewed and summarised. IQ decrement and consequent productivity losses were then calculated for a one-year cohort of 5-year-old children in each evaluated country.
2.1 Estimation of national level BLLs
We evaluated all 136 LMICs of the World Bank’s June 2020 list of economies (World Bank, 2020). BLL estimates for each country were taken from a recent systematic review that presents background BLL levels for 34 countries and hotspot BLL levels for 23 countries (Ericson et al., Reference Ericson, Hu, Nash, Ferraro, Sinitsky and Taylor2021). Values for the remaining countries were identified through a literature review or imputed if no studies were identified.
A 2022 PubMed search was conducted using the search terms ‘[country name]’, ‘blood’ and ‘lead’ (all fields, MeSH terms) for studies published after the country’s leaded petrol phaseout date. Results were further filtered by species (Human) and age (Child: birth–18 years). Only those countries missing estimates in a recent systematic review were included in the search (Ericson et al., Reference Ericson, Hu, Nash, Ferraro, Sinitsky and Taylor2021).
Results returned by PubMed were screened for relevance through a title and abstract review. The full text of relevant studies was then assessed against five inclusion criteria: contained BLLs from children < 18 years residing in the target country; comprised of ≥ 30 participants; presented BLL data (ie. central tendency) derived from venous, capillary, or umbilical cord samples of whole blood; data collected at least 12 months after the leaded petrol phaseout; and contained data that had not been reported elsewhere (ie. excluding systematic reviews or subsequent studies of the same cohort). Studies meeting these criteria were assessed for bias and quality using a bespoke tool developed for this study and available in the supplementary materials.
BLL data were extracted from studies meeting the inclusion criteria and passing the bias assessment. Available measures of central tendency (eg. mean, median) and dispersion (eg. SD, 95% CI) were extracted for children < 18 years of age. Other descriptive statistics and the method of laboratory analysis (eg. ICP-MS, AAS) were also extracted if available. Subgroups were coded as either ‘background’ or ‘elevated exposure’ the latter comprised of populations identified as having lead exposure uncharacteristic of the general population (eg. proximity to a hotspot or use of leaded cosmetics). The subgroup ‘background’ was intended to capture general population exposure and included population assessments, control groups in case-control studies, and similar efforts. Results were taken at face value and were not reinterpreted.
In countries with more than one study, the mean of the log-transformed sample-weighted values was taken for each, following previous work (Fewtrell et al., Reference Fewtrell, Kaufmann and Prüss-Üstün2003; Attina and Trasande, Reference Attina and Trasande2013; Ericson et al., Reference Ericson, Hu, Nash, Ferraro, Sinitsky and Taylor2021). Background and elevated exposure subgroups were analysed separately. Missing mean and standard deviation values were imputed using indicators of central tendency and dispersion as described elsewhere (Ericson et al., Reference Ericson, Hu, Nash, Ferraro, Sinitsky and Taylor2021). Two countries, China and the Republic of Georgia, had recent population-based studies of BLLs that met our inclusion criteria (UNICEF, 2019; Li et al., Reference Li, Gao, Dong, Wu, Yan, Cao, Ma, Wang, Gong, Xu, Cai, Chen, Xu, Tong, Tang, Zhang and Yan2020). The results from these studies only were used and were not pooled with those from other efforts. Finally, the most recent available citation list from the Institute for Health Metrics and Evaluation’s (IHME) 2017 Global Burden of Disease was reviewed to identify any studies that may have not been identified in the preceding steps (IHME, 2018). All pooled BLLs, regardless of the year of data collection or the age of the children, were taken as reflecting the cumulative lifetime exposure to a cohort of 5-year-old children 2019.
If a country’s background BLLs were not identified, we spatially interpolated them based on known values using Triangulated Irregular Network (TIN) interpolation.
2.2 Estimation of hotspot exposure
BLLs for elevated exposure subgroups for each country were also identified based on the five inclusion criteria. The number of children with elevated exposures was then conservatively approximated as 0.1% of a country’s population based on a 2016 study of informal lead recycling in LMICs (Ericson et al., Reference Ericson, Landrigan, Taylor, Frostad, Caravanos, Keith and Fuller2016).
2.3 Economic cost methods
The effect of lead exposure on IQ was calculated using pooled BLLs indexed against associated piecewise linear IQ decrement previously estimated (Budtz-Jørgensen et al., Reference Budtz-Jørgensen, Bellinger, Lanphear, Grandjean, Lanphear, Hornung, Khoury, Yolton, Baghurst, Bellinger, Canfield, Dietrich, Bornschein, Greene, Rothenberg, Needleman, Schnaas, Wasserman, Graziano and Roberts2013). Here, IQ decrement is proportionally greater at lower BLLs, below a breakpoint of 7.5 ug/dL (see Figure 1 in the supplementary materials). Thus, in the present study IQ decrement was calculated for different BLL intervals, beginning with 0.2–1 ug/dL. Intervals then increased by units of 1 ug/dL until the 25 ug/dL, with the final interval capturing BLLs greater than 25 ug/dL. These values were then summed to arrive at total IQ decrement and associated costs for a given country.
Figure 1. Literature review flowchart.
The loss of a single IQ point has been quantified as a 0.5–2.5% reduction in lifetime earnings. We used the recently conservatively estimated IQ value as 1.4% of the present value of lifetime earnings for both market and nonmarket labour (Grosse and Zhou, Reference Grosse and Zhou2021). The present value of lifetime earnings is the discounted average earnings for workers in a specific economy over the course of their lives. It is calculated as labour’s share of GDP divided by the number of workers and adjusted for both the rate of growth in output per worker and a social discount rate (Landrigan et al., Reference Landrigan, Fuller, Acosta, Adeyi, Arnold, Basu, Baldé, Bertollini, Bose-O’Reilly, Boufford, Breysse, Chiles, Mahidol, Coll-Seck, Cropper, Fobil, Fuster, Greenstone, Haines, Hanrahan, Hunter, Khare, Krupnick, Lanphear, Lohani, Martin, Mathiasen, McTeer, Murray, Ndahimananjara, Perera, Potočnik, Preker, Ramesh, Rockström, Salinas, Samson, Sandilya, Sly, Smith, Steiner, Stewart, Suk, van Schayck, Yadama, Yumkella and Zhong2017; Krupnick and Cropper, Reference Krupnick and Cropper2019). We assume that the social discount rate will exceed growth per worker by 1.5% (low) or 3% (high). Thus, it is calculated as follows: [1/(1+d)], where d = 0.015 or 0.03 (Landrigan et al., Reference Landrigan, Fuller, Acosta, Adeyi, Arnold, Basu, Baldé, Bertollini, Bose-O’Reilly, Boufford, Breysse, Chiles, Mahidol, Coll-Seck, Cropper, Fobil, Fuster, Greenstone, Haines, Hanrahan, Hunter, Khare, Krupnick, Lanphear, Lohani, Martin, Mathiasen, McTeer, Murray, Ndahimananjara, Perera, Potočnik, Preker, Ramesh, Rockström, Salinas, Samson, Sandilya, Sly, Smith, Steiner, Stewart, Suk, van Schayck, Yadama, Yumkella and Zhong2017). We further include the value of nonmarket labour at 25% of market labour following previous work and based on US Bureau of Economic Analysis estimates (US Bureau of Economic Analysis, 2022; Krupnick and Cropper, Reference Krupnick and Cropper2019). Lifetime earnings are then weighted by the probability that a worker will survive from one year to the next using the Global Health Observatory survival rates (WHO, 2022). Lifetime earnings are calculated for a one-year cohort of 5-year-old children. To control for the impact of the COVID-19 pandemic, all economic data are from the year 2019. All results are presented in 2019 USD unless stated otherwise. Further detail, including underlying equations and sources, is provided in the supplementary materials.
2.4 Data analysis
Data were aggregated and analysed in Microsoft Excel and Stata. IQ decrement and associated productivity losses were estimated in Microsoft Excel. Spatial analysis and interpolation were conducted in QGIS (QGIS Development Team, 2024; Microsoft Corporation, 2022; StataCorp. LP, 2017).
2.5 Sensitivity analysis
Our sensitivity analysis evaluated the influence of the percent difference in market productivity attributable to a single IQ point, using the estimates described above and varying the values to 1 or 2%. As part of our sensitivity analysis, interpolation was conducted with Inverse Distance Weighting (IDW) which more heavily weights known proximate points than distant points. The IDW interpolation was conducted with 6 different weights (p coefficients).
3. Results
3.1 Literature Review
Our PubMed search was conducted for 101 countries and returned 676 titles. Of these 54 titles were considered relevant and reviewed against the inclusion criteria and are reported in the supplementary materials. Thirty-nine studies were excluded at this stage, leaving 15 studies available for inclusion. A review of the IHME citation list returned one additional study which passed our bias assessment. Thus the literature review resulted in a total of 16 additional titles (See Figure 1). Data on background groups were available for six countries (n = 12 studies), while data on elevated exposures were available for one (n = 4 studies). These 16 studies were then joined with previous work as well as with recent studies from China and Georgia for the analysis, resulting in 388 unique publications (Ericson et al., Reference Ericson, Hu, Nash, Ferraro, Sinitsky and Taylor2021; UNICEF, 2019; Li et al., Reference Li, Gao, Dong, Wu, Yan, Cao, Ma, Wang, Gong, Xu, Cai, Chen, Xu, Tong, Tang, Zhang and Yan2020).
3.2 Estimation of national level BLLs
Background BLL data were pooled for a total of 39 countries, combining previous data for 33 countries with data from 6 countries extracted as part of the literature review. Romania, included in previous work, was no longer classified by the World Bank as an LMIC in 2019 and was therefore excluded. Argentina, excluded from previous work, became classified as an LMIC in 2019 and was therefore included.
Data on sub-populations with elevated exposures were available in 25 countries. BLLs in the subgroups were highly variable ranging from 1.76 µg/dL to 58.5 µg/dL for Bolivia and Nigeria, respectively. Of these 19 (76%) had mean BLLs for the elevated exposure subgroup that exceeded 5 ug/dL; 40% had means exceeding 10 µg/dL.
3.3 National IQ and productivity losses
National BLL estimates could be calculated using existing studies for 39 countries and were interpolated for an additional 82 countries using the TIN approach, with the remaining countries being too distally located from known values to be approximated. Total lead-attributable IQ losses were calculated as 403 million points. Consequent productivity losses in LMICs ranged from USD 305 billion in our high discount scenario to 499 billion in our low discount scenario for each one-year cohort of 5-year-old children (2019 USD). Productivity losses as a share of GDP were lowest in East Asia and the Pacific (range: 0.7–1.1%) and highest in Sub-Saharan Africa (range: 2.6–4.2%). Conversely, productivity losses in East Asia and the Pacific were far larger in absolute terms than other regions, accounting for nearly 38% of the LMIC total.
Table 1 presents productivity loss estimates for the 121 countries for which BLL data were reported or could be interpolated. The results are organised by World Bank region and include ranges in country-level estimates of BLL, IQ decrement, and associated costs. Full country-level results are provided in the supplementary materials.
Table 1. Productivity loss estimates for the 121 countries, by World Bank Region, for which BLL data were reported or could be interpolated. All values in 2019 USD
3.4 Hotspot IQ and productivity losses
National productivity losses attributable to hotspots ranged from USD 117,000 for Armenia to nearly USD 300 million for China. Table 2 presents productivity loss estimates for the 25 countries for which BLL data were reported. Our method for quantifying the IQ decrement associated with a given BLL was capped at 25 µg/dL (IQ decrement of 7.03 points). Pooled BLLs in three countries exceeded this level: Ecuador, Egypt and Nigeria.
Table 2. Productivity loss estimates attributable to hotspot exposures in the 25 countries for which BLL data were reported (low estimate assuming 0.1% of the children exposed). All values in 2019 USD
3.5 Results of sensitivity analysis
Adjusting the value of an IQ point from 1.4 to 2% of lifetime earnings increased global productivity losses from USD 305–499 billion to USD 436–712 billion. Lowering this value to 1% reduced total losses to USD 218–356 billion. Using IDW interpolation resulted in comparable though less well-fitting BLLs than TIN when compared to the original dataset (see supplementary materials).
4. Discussion
This study calculated productivity losses of USD 305 billion to USD 499 billion in LMICs for each one-year cohort of 5-year-old children (2019 USD), depending on the discount scenario used. In terms of their share of GDP, these losses range from 0.7 to 1.1% in East Asia and the Pacific to 2.6–4.2% in Sub-Saharan Africa. Results in East Asia and the Pacific were driven largely by the values for one country, China, which comprises roughly 50% of IQ points lost and 76% of absolute productivity losses in the region. China is only one of two countries in the study with BLLs based on representative data, thus eliminating a major limitation of the analysis and underlining the importance of improved data collection.
These estimates are conservative. The spectrum of health effects from lead exposure ranges from subtle to severe, and exposure is associated with toxicity to every organ system. Societal economic costs are typically attributed to adverse outcomes that result in decreased productivity or increased need for healthcare. Adult lead exposure has well-documented effects on the cardiovascular system, increasing both systolic and diastolic blood pressure (ATSDR, 2020). Associated increases in morbidity and mortality costs are substantial and have been calculated elsewhere (Lamas et al., Reference Lamas, Ujueta and Navas-Acien2021; Larsen and Sánchez-Triana, Reference Larsen and Sánchez-Triana2023). The global cost of heart failure in a single year (2012) has been estimated as USD 108 billion, with 60% being borne by healthcare systems and 40% associated with indirect costs, such as lost productivity and unpaid medical bills (Cook et al., Reference Cook, Cole, Asaria, Jabbour and Francis2014). The attributable fractions of low-level lead exposure for cardiovascular disease mortality and ischemic heart disease have been calculated as high as 28.7% and 38.4%, respectively (Lanphear et al., Reference Lanphear, Rauch, Auinger, Allen and Hornung2018). Thus by estimating only those productivity costs associated with paediatric lead exposure, the present study vastly underestimates lead’s overall impact.
For hotspots, children’s BLLs can reach levels exceeding 100 µg/dL (Bose-O’Reilly et al., Reference Bose-O’Reilly, John, Joseph, Paul, Bret and Jack2017; Dooyema et al., Reference Dooyema, Neri, Lo, Durant, Dargan, Swarthout, Biya, Gidado, Haladu, Sani-Gwarzo, Nguku, Akpan, Idris, Bashir and Brown2012). Some impacts of very high blood lead levels on children are well understood including severe intellectual disability, seizure disorders, blindness and death. However, most studies base their IQ decrement estimates on cohorts of children in affluent countries with much lower exposure levels. Thus in the present study, IQ decrement could be estimated for BLLs up to 25 µg/dL only; all BLLs exceeding this level were assumed to have an equivalent IQ loss of 7.08 points. The neurological impacts of lead exposure do not stop at 25 µg/dL and the literature is replete with case studies and reports of the drastic impact of frank childhood lead poisoning. Further, there is evidence that attributable IQ decrement continues beyond childhood into adolescence and adulthood, implicating costs not captured here (Reuben et al., Reference Reuben, Caspi, Belsky, Broadbent, Harrington, Sugden, Houts, Ramrakha, Poulton and Moffitt2017).
A key reason that childhood lead exposure does not receive attention commensurate with its impact is lack of comprehensive and publicly available information. The existing studies provide a compelling argument that supports taking a closer look at blood and environmental lead levels in most countries. Biomonitoring and population-based surveys generalisable to an entire country can provide estimates of national lead exposure. Biomonitoring also provides measures against which population-wide BLL reduction can be assessed. Systematic small-area studies in communities or subpopulations known or suspected to be at higher risk for lead exposure provide estimates that inform specific remediation strategies to these communities and can also be compared to the general population (Egan et al., Reference Egan, Dignam, Brown, Bayleyegn and Blanton2022).
There is a great need to strengthen methods for conducting and reporting data on lead in children’s blood and environment, as reflected through the small number of studies that could be included here. It is critical that a key set of demographic, geographic, environmental, and BLL reporting standards be developed to improve data quality.
The routine reporting of data, including that from the peer-reviewed literature, to a central repository could support efforts to understand the nature and extent of global lead exposure. This evidence base could further be used in the identification and remediation of lead sources and monitoring and evaluating efforts to eliminate lead exposure worldwide.
4.1 Comparison with existing studies
We benchmarked our main results against Attina and Trasande (Reference Attina and Trasande2013) and Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023). Only Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023) report on the results of their BLL calculations, which they do by World Bank region. Here the results of the two studies are similar, with all regional population-weighted BLL estimates being in the range of 0–0.63 µg/dL of each other. An exception was Europe and Central Asia where the present study found a population-weighted BLL 2.63 µg/dL greater than Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023). We attribute the variation to the small number of studies in the region and thus increased relative influence of any one study’s results on the regional estimate.
With regard to IQ decrement, Attina and Trasande (Reference Attina and Trasande2013) found a total loss of 407 million IQ points in LMICs. This study identified a nearly equivalent loss of 403 million IQ points. By contrast, Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023) find a substantially larger loss of 729 million IQ points. More detailed comparisons between the three studies were not possible given differences in reporting. Most significantly Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023) report the majority of their findings at the global or regional level only, without disaggregating by country income level.
Both Attina and Trasande (Reference Attina and Trasande2013) and Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023) assessed the value of an IQ point as 2% of lifetime earnings. Attina and Trasande (Reference Attina and Trasande2013) also report results in international dollars (ie. at PPP). This is distinct from the present study, which utilises a value of 1.4% of lifetime earnings and presents results in 2019 USD. International dollars are adjusted to take into account the relative purchasing power of a US dollar in a given economy. Thus in international dollar terms, the value of one dollar in the United States is reflected as $ 1, while the same dollar would be valued at $ 2.34 in Kenya or $ 0.97 in Japan (2019 international dollars). For the purpose of comparison, we adjusted our assessment of lifetime earnings upward to 2% and converted our results to 2019 international dollars. We also converted the Attina and Trasande (2011) results to 2019 international dollars. Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023) report their findings in both 2019 international dollars and USD, however, they do not fully disaggregate LMICs from their global results. We thus approximate total losses based on information provided.
Following these conversions, the main results of the three studies are similar with the present study finding international $ 1–1.7 trillion in losses compared to international $ 1.3 trillion (0.9–1.58) found by Attina and Trasande (Reference Attina and Trasande2013) and international $ 1.61 trillion (736–2,415) found by Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023) (2019 International dollars). We attribute the higher values of the Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023) to their use of a more severe dose–response model to calculate IQ decrement than those employed in the other two studies. This observation was also made by Larsen and Sánchez-Triana (Reference Larsen and Sánchez-Triana2023).
4.2 Limitations
Our results are limited by multiple factors. The most substantial among these is the dearth of data available on BLLs in LMICs. Nationally representative studies were available in only two of the 136 LMICs (China and Georgia). Background BLL data on children could be identified in 39 countries only. With the exception of the two countries noted above, these data were not initially collected for the purpose of inferring a nationally representative value. BLLs for all other countries (n = 82) were interpolated based on data reported in proximal countries. This necessarily introduces error into our estimates as in reality a large number of factors influence actual BLL concentrations. IHME (2017) attempted to improve on simple geographic interpolation with the use of predictive covariates, relating primarily to the effects of leaded petrol and urbanicity. It is unclear how and if the inclusion of these covariates improves on the methods used here. Lead exposure in LMICs is attributable to multiple and disparate sources, including adulterated spices and ceramic glazes that are unrelated to urbanicity or leaded petrol. Attina and Trasande (Reference Attina and Trasande2013) similarly employ a simple geographic interpolation. Given the level of uncertainly in the interpolated country-specific estimates, we provide only regional summaries in our main results. To facilitate comparison and transparency, the full list of country-specific estimates is provided in the supplementary materials.
Subgroups were characterised as ‘background’ or ‘elevated exposure’. Some studies pooled these subgroups together without calculating separate subgroup means. In these cases, we conservatively coded the sample as ‘elevated exposure’. However, this presents an artificially low BLL for the ‘elevated exposure’ subgroup. Samples from certain mining communities or smelter towns for example were pooled across areas of high and low ambient contamination when only a subsample lived in places that could be characterised as a hotspot. One example is Oruro, Bolivia where the studies report BLLs from a wide geographic area as representative of hotspot exposure. However in reality, the environmental contamination is much more confined affecting only a smaller subset of the population (Pure Earth, 2018).
Assessment of childhood BLLs cannot depend alone on national or regional estimates which fail to identify areas where lead exposure is disproportionately high. Pooling results over multiple sites increases statistical stability but also results in conservative estimates of BLLs. For example, a mass lead poisoning event in Nigeria resulted in BLLs in excess of 130 µg/dL (Dooyema et al., Reference Dooyema, Neri, Lo, Durant, Dargan, Swarthout, Biya, Gidado, Haladu, Sani-Gwarzo, Nguku, Akpan, Idris, Bashir and Brown2012). When these estimates were pooled with BLLs from other elevated exposure groups in Nigeria the result was 58.5 µg/dL. A related limitation is our pooling of BLLs for all children < 18 years of age to calculate a representative BLL for each cohort of 5-year-old children. Because BLLs tend to decrease with age, this likely resulted in an underestimate of exposure.
There were certain limitations related to the availability of economic data. These are reported in detail in the supplementary materials and were unlikely to discernibly influence the results; in most cases, reasonable imputations or proxies could be made.
There are implicit limitations associated with making these sorts of projections. The models that underly our study are based on observations in high-income countries. We have tried to adapt these to LMICs by using country-specific economic and demographic data. We have also endeavoured to be clear about the uncertainty in our analysis and have presented ranges where possible. Finally, we have made an effort to present conservative estimates by valuing an IQ point at 1.4% of lifetime earnings and using more conservative estimates of attributable IQ point loss. The results may in fact be an underestimate of losses. We have made our full dataset available in the supplementary materials and encourage further analysis and interpretation by other researchers.
5. Conclusions
This study found USD 305–499 billion (2019 USD) in lost economic productivity attributable to paediatric lead exposure in the 121 LMICs for which values could be calculated. This finding is consistent with previous estimates and further validates those efforts with a substantially expanded dataset. Differences in the findings resulted primarily from the use of a more conservative dose–response model in the present study. Improved reporting of BLLs is essential to refine future estimates and could be facilitated through a centralised registry of study results.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1744133125100121
Acknowledgements
We are grateful to Scott Grosse for reviewing an earlier draft of this study. We are grateful to Bren Fitzmaurice, Hailey Mainieri and Lily Stauble who contributed to background research.
Financial support
We do not declare any funding.
Competing interests
Bret Ericson declares that he is an independent technical expert that consults for UN organisations and NGOs on chemicals exposures, including those relating to lead. Mary Jean Brown declares that she provides technical assistance and advice to Meridian Biosciences makers of the blood lead point of care instrument LeadCare II. Meridian Biosciences is not referenced in this document. Dr. Brown has also provided technical assistance for plaintiffs in legal cases involving lead exposure.
Open access
