Study population
The study population has been described before [8]: 14–15-year old adolescents were sampled from the Flemish population in 2003–2004 (FLEHSI), 2008–2009 (FLEHSII), and 2013–2014 (FLEHSIII). In FLEHS I, 1679 adolescents were recruited by a randomized two stage sampling design in nine areas in Flanders with a different pollution pressure (two industrial sites, two harbors, two cities, a rural area, a zone around waste incinerators and a fruit cultivating area). In total, the area in which the participants were recruited covered 3035 km2 and 1.2 million inhabitants which is equal to 22% of the Flemish region and 20% of the Flemish population, respectively. In FLEHS II and III, a representative sample of the general Flemish adolescent population of respectively 210 and 208 participants was recruited by a two stage sampling design with the five districts of Flanders as primary sampling units and schools as secondary sampling units. The number of participants per province was proportional to the number of inhabitants of that province. To account for seasonal variation, recruitment was spread over one year with no recruitment of adolescents during examination periods and summer holidays (June, July, August). The studies were approved by the medical–ethical committee of the University of Antwerp (reference number A03 053, UA A08 09, and B300201316515). For all three surveys, inclusion criteria were (1) residing at least 10 years in Flanders (FLEHSII and III) or at least 5 years in one of the selected areas (FLEHSI), (2) giving written informed consent,(3) being able to fill out an extensive Dutch questionnaire, and (4) being in the 3rd year of secondary school. More detailed information on the selection and recruitment of the participants in the studies was described earlier [16,17,18,19].
Exposure assessment
Peripheral blood samples were collected during a clinical examination in the schools. Blood lead was measured by high resolution inductively coupled plasma-mass spectrometry (HR-ICP-MS,Thermo Element II) after micro-wave acid digestion using HNO3 and H2O2 [8, 20].
Blood lead level distribution
B-Pb levels reported by Schoeters et al. [8] enabled to derive the log-normal B-Pb distribution for each sampling period (FLEHSI, FLEHSII, and FLEHSIII). The data of FLEHS I were weighed to correct for unequal sample sizes in the eight geographical areas spread over Flanders. The geometric mean (GM) adolescent B-Pb in FLEHSI, FLEHSII, and FLEHSIII were respectively 22.5 μg/L (95% Confidence Interval (CI): 21.8–23.3), 14.6 μg/L (95% CI: 13.8–15.5), and 9.5 μg/L (95% CI: 9.0–9.9). The results were adjusted for age, sex, and smoking to ensure comparability between the studies. The number of participants during the first, second, and third cycle were respectively 1679, 210, and 208. The participants age ranged from 13.8 to 17.2 years of age with participants between 14.5 and 15.5 years the strongest represented (about 60%). Sex was equally distributed throughout the different studies. Smoking was reported by about 13–14% of the participants.
Estimation of IQ loss attributable to blood lead levels above 20 μg Pb/L
To estimate the IQ loss attributable to elevated blood lead levels in the study populations we used a dose response function from the available literature that describes IQ decrement as a function of B-Pb. We focused on the study of Lanphear et al. (2005) [7] as it covered the low exposure range, it was based on international pooled analysis of 7 cohorts (total N = 1333), and the data had been obtained from children followed from birth or infancy until 5–10 years of age.
They established a linear-log inverse relationship between IQ and B-Pb, in which a doubling in B-Pb was associated with a decrease in IQ of 1.88 points (95% CI: 1.16–2.59). The linear-log relationship implies that, for a given absolute increase in blood lead, the associated IQ loss is higher in the low level range. A statistical re-evaluation of the data used confirmed that the effect was higher in the low level range and that the conclusions were robust [21]. Based on the linear-log relationship proposed by Lanphear, Gould et al. suggested in 2009 that a uniform decrease (i.e. linear relationship) may be assumed over three ranges, i.e. for B-Pb between 20 and 100 μg/L; between 100 and 200 μg/L; and between 200 and 300 μg/L [22]. The estimated decrease in IQ points for an increase in B-Pb with 1 μg/L was equal to 0.054 (95% CI: 0.034–0.075); 0.019 (95% CI: 0.012–0.026); 0.011 (95% CI: 0.007–0.015) for blood levels between respectively 20 and 100 μg/L; 100 and 200 μg/L; and between 200 and 300 μg/L. There is little difference in the linear-log and the linear-interval dose response relationship at higher exposure levels. However, the linear-interval dose response relationship is more conservative in the lower dose range (blood lead level below 100 μg/L), as visualised in Fig. 1.
We estimated IQ loss above the threshold of 20 μg/L as dose response estimates are lacking in the lower dose region. Furthermore, 20 μg/L is considered as a relevant action level for B-Pb, as was already argued by Gilbert and Weiss in 2006 to lower the CDC action level to 20 μg/L [23]. We applied the linear-log dose response function to derive our main conclusions, the linear-interval dose response function was used as a sensitivity analysis. Based on the dose response relationships, the average IQ loss (and 95% CI) per individual within the population exceeding 20 μg/L for each sampling period was calculated by simulating 1,000,000 samples from the log-normal B-Pb distributions and taking the average. This was subsequently multiplied by the fraction of the population with B-Pb exceeding 20 μg/L and by 100,000 to estimate the IQ loss - attributable to B-Pb above 20 μg/L - per 100,000 individuals in Flanders. Although reliable dose response functions are lacking below 20 μg/L, it has been argued that there is no safe threshold for B-Pb [24]. Hence, we calculated potential additional IQ loss below 20 μg/L, by extrapolating the linear dose response function reported by Gould et al. [22] for the range between 20 μg/L and 100 μg/L.
Next, we compared the estimated IQ loss - attributable to B-Pb above 20 μg/L - between 2000 and 2014 with the estimated IQ loss that is expected between 2015 and 2029 for the Flemish population. The FLEHSI exposure distribution was fixed for the period between 2000 and 2004, the FLEHSII exposure distribution for the period between 2005 and 2009, and the FLEHSIII exposure distribution for the period between 2010 and 2014 and also for the period between 2015 and 2029. Although a further decrease in B-Pb between 2015 and 2029 may be expected, it is uncertain to what extent. Hence, the calculations for 2015–2029 may be considered as an upper limit. For each sub-period, the IQ loss and economic loss attributable to B-Pb above 20 μg/L in Flanders were calculated by taking into account the number of Flemish 15 year-olds: The average IQ loss (and 95% CI) per individual within the population exceeding 20 μg/L – calculated as described in the previous paragraph – was multiplied by the fraction of the population with B-Pb exceeding 20 μg/L and with the number of Flemish 15 year-olds within a given period. According to demographic data obtained from http://statbel.fgov.be, the number of Flemish 15 year-olds was 340,355 in 2000–2004, 364,704 in 2005–2009, and 347,761 in 2010–2014, which sums up to 1,052,820 for the period between 2000 and 2014. For the period from 2014 to 2029 we assumed that the number of 15 year-olds remains constant as the differences between the different sub-periods from 2000 to 2014 were relatively small.
Estimation of social costs attributable to blood lead levels above 20 μg Pb/L
Social costs of IQ decrement were valued by calculating lifetime earning loss per person. We used the estimated lifetime value of 1 IQ point reported by Bellanger et al. [25] as a basis. In this publication, the authors used the life time value of 1 IQ point that was calculated for France in 2008 based on data from the US by Pichery et al. [26] (€ 17,363) and adjusted it for differences in purchasing power to derive an estimate for other countries. As such, the Belgian lifetime value of 1 IQ point was estimated at € 16,458. Based on the harmonised index of consumer prices (https://ec.europa.eu/eurostat/web/hicp/data/database) this value was adjusted for inflation, which results in an estimated lifetime value of 1 IQ point of € 19,464 for Belgium in 2018.
It shall be noted that this estimate is mainly based on studies carried out in the United States [22, 27],so we assumed that differences in lifetime incomes are the same in Europe which is not necessarily true. Adjustment for differences in purchasing power parity has been included to take this issue partially into account. As we focused on lifetime earnings only, our estimate is probably an underestimate of the total benefits of Pb control. We did not consider direct medical costs linked to treatment or interventions for children with neurodevelopmental disorders, costs related to special education or additional years of schooling for children as a consequence of these disorders [25]. Furthermore, our estimate is an underestimate of total costs attributable to Pb exposure, as we did not consider e.g. cardiovascular effects.