We identified all papers published on or before December 2010 that reported air or blood lead measurements that had been collected from US work sites that contained information on job or industry by searching the web-based bibliographic databases MEDLINE, Web of Science, Scopus, SciFinder, and NIOSHTIC2 using the search terms ‘lead exposure’, ‘worker’, ‘occupation’, and ‘occupational exposure’ and by reviewing the citations of the identified papers.
From each paper, we extracted lead measurements, which were primarily reported as summary statistics (including results presented in figures), and corresponding ancillary data. We excluded blood zinc protoporphyrin and free erythrocyte protoporphyrin measurements because these tests are not specific to lead [Baxter and Igisu 2010 ], pre-employment baseline biologic measurements, post-medical removal biologic measurements, and subsequent reports of data reported in multiple papers.
Air and blood lead concentrations were entered using units of µg/m3 and µg/dL, respectively, using conversion factor 1.0 µmol/L=20.7 µg/dL. If the number of measurements was provided as a range, the mean number was used in descriptive statistics and analyses. The three air summary statistics reported as below the limit of detection (LOD) were replaced with the LOD/2. Individual measurements in the same job/facility were aggregated. Missing arithmetic means (AM) were calculated from the geometric mean (GM) and geometric standard deviation (GSD) usingequation 1 . Missing GMs were assigned the median, if reported. Missing GMs and GSDs were calculated using equations 2 and 3 , respectively, if the range was reported, then the AM was calculated using equation 1 . If missing the GSD but the GM was available, we assumed a GSD =2.56 [Kromhout, et al. 1993 (link)].
Extracted ancillary data included exposure category, industry, job, task or area description, sample year (if a range, the midpoint was assigned; if missing, assigned publication year minus 2), exposure source, sample type, sampling method, analytic method, type of ventilation used, type of respiratory protection used, whether measurements represented worse case exposure scenarios (e.g., elevated blood lead levels, employee concerns, regulatory violations), whether the work being performed was a lead-based paint removal activity, and whether workplace containment structures were erected. Industry was coded to two-digit 1987 Standard Industrial Classification (SIC) codes [OMB 1987 ]. For air measurements, we extracted sampler location, whether the sample was task-based or full-shift, sampling duration, and particle size. For blood measurements, we extracted the time of sampling.
Hereafter, we focus on personal air and blood lead measurements because they are considered the preferred measures of personal lead exposure [ACGIH 2001 ]. To facilitate broad comparisons, we calculated industry-specific weighted arithmetic means (WAM≥1970, weighted by the number of measurements) for all personal air and blood lead summary statistics collected from 1970 onwards. Personal air WAM≥1970 calculations were restricted to total suspended particle and inhalable particle samples with sample durations >1 hour and that reported the number of measurements collected. WAMs≥1970 are reported here only for industries with ≥10 measurements. Statistical modeling of these data is reported separately.
From each paper, we extracted lead measurements, which were primarily reported as summary statistics (including results presented in figures), and corresponding ancillary data. We excluded blood zinc protoporphyrin and free erythrocyte protoporphyrin measurements because these tests are not specific to lead [Baxter and Igisu 2010 ], pre-employment baseline biologic measurements, post-medical removal biologic measurements, and subsequent reports of data reported in multiple papers.
Air and blood lead concentrations were entered using units of µg/m3 and µg/dL, respectively, using conversion factor 1.0 µmol/L=20.7 µg/dL. If the number of measurements was provided as a range, the mean number was used in descriptive statistics and analyses. The three air summary statistics reported as below the limit of detection (LOD) were replaced with the LOD/2. Individual measurements in the same job/facility were aggregated. Missing arithmetic means (AM) were calculated from the geometric mean (GM) and geometric standard deviation (GSD) using
Extracted ancillary data included exposure category, industry, job, task or area description, sample year (if a range, the midpoint was assigned; if missing, assigned publication year minus 2), exposure source, sample type, sampling method, analytic method, type of ventilation used, type of respiratory protection used, whether measurements represented worse case exposure scenarios (e.g., elevated blood lead levels, employee concerns, regulatory violations), whether the work being performed was a lead-based paint removal activity, and whether workplace containment structures were erected. Industry was coded to two-digit 1987 Standard Industrial Classification (SIC) codes [OMB 1987 ]. For air measurements, we extracted sampler location, whether the sample was task-based or full-shift, sampling duration, and particle size. For blood measurements, we extracted the time of sampling.
Hereafter, we focus on personal air and blood lead measurements because they are considered the preferred measures of personal lead exposure [ACGIH 2001 ]. To facilitate broad comparisons, we calculated industry-specific weighted arithmetic means (WAM≥1970, weighted by the number of measurements) for all personal air and blood lead summary statistics collected from 1970 onwards. Personal air WAM≥1970 calculations were restricted to total suspended particle and inhalable particle samples with sample durations >1 hour and that reported the number of measurements collected. WAMs≥1970 are reported here only for industries with ≥10 measurements. Statistical modeling of these data is reported separately.
