Estimating distributions of long-term particulate matter and manganese exposures for residents of Toronto, Canada

Methylcyclopentadienyl manganese tricarbonyl (MMT), a manganese-based gasol ine additive, has been used in Canadian gasoline for about 20 yr. Because M MT potentially increases manganese levels in particulate matter resulting f rom automotive exhausts, a population-based study conducted in Toronto, Can ada assessed the levels of personal manganese exposures. Integrated 3-day p articulate matter (PM2.5) exposure measurements, obtained for 922 participa nt periods over the course of a year (September 1995-August 1996), were ana lyzed for several constituent elements, including Mn. The 922 measurements included 542 participants who provided a single 3-day observation plus 190 participants who provided two observations (in two different months). In ad dition to characterizing the distributions of 3-day average exposures, whic h can be estimated directly from the data, including the second observation for some participants enabled us to use a model-based approach to estimate the long-term (i.e. annual) exposure distributions for PM2.5 mass and Mn. The model assumes that individuals' 3-day average exposure measurements wit hin a given month are lognormally distributed and that the correlation betw een 3-day log-scale measurements k months apart(after seasonal adjustment) depends only on the lag time, k, and not on the time of year. The approach produces a set of simulated annual exposures from which an annual distribut ion can be inferred using estimated correlations and monthly means and vari ances (log scale) as model inputs. The model appeared to perform reasonably well for the overall population distribution of PM2.5 exposures (mean = 28 mu g m(-3)). For example, the model predicted the 95th percentile of the a nnual distribution to be 62.9 mu g m(-3) while the corresponding percentile estimated for the 3-day data was 86.6 mu g m(-3). The assumptions of the m odel did not appear to hold for the overall population of Mn exposures (mea n = 13.1 ng m(-3)). Since the population included persons who were potentia lly occupationally exposed to Mn (in non-vehicle-related jobs), we used res ponses to questionnaire items to form a subgroup consisting of non-occupati onally exposed participants (671 participant periods), for which the model assumptions did appear to hold. For that subpopulation (mean = 9.2 ng m(-3) ), the model-predicted 95th percentile of the annual Mn distribution was 16 .3-ng m(-3), compared with 21.1 ng m(-3) estimated for the 3-day data. (C) 1999 Elsevier Science Ltd. All rights reserved.