Summary Prolonged exposure by inhalation to respirable sized dusts containing crystalline silica is known to cause respiratory diseases, including lung cancer and silicosis. Many studies have identified the overexposure of construction workers to crystalline silica since this compound is present in many materials used on construction sites. Evaluation of crystalline silica exposure in this industry is a challenge due to the many working conditions and the temporary nature of the sites.In order to improve knowledge about working conditions that can expose construction workers to silica and to respond to an initial request from the CSST, a joint team from the Université de Montréal and the IRSST developed an occupational exposure database for this contaminant based on the literature. This initial project was funded by the IRSST and was the subject of a report published in 2011 by this organization. The database contains more than 10,000 exposure measurements, originating from scientific periodicals, external databases and reports from research organizations, collected following an exhaustive literature review covering the last 25 years. Descriptive analyses produced from quantitative measurements from the database identified the professions and tasks associated with the highest exposures. The researchers nevertheless emphasized the important potential for additional information. This potential is associated with the use of multivariate analytical techniques, allowing the evaluation of the simultaneous influence of several variables on the exposure levels. The present project had the primary objective of furthering the use of the database in order to refine the portrait of crystalline silica exposures and to evaluate the contribution of the different exposure determinants.The crystalline silica exposure data were first selected on the basis of the sampling strategy, resulting in two complementary analyses whose objective was to estimate the exposure levels in relation to the threshold limit values (TLVs) over eight hours according to the occupation title, and the task in relation to its performance duration. The Monte Carlo method was used to recreate the samples originating from data reported in the form of synthesis parameters (e.g., geometric mean and geometric standard deviation, arithmetic mean, range), allowing their combined analysis with the results reported in the form of single measurements. Statistical models including variables such as occupation title, the task performed, the measurement duration, the year and the sampling strategy, the type of project (demolition, new construction and renovation), the activity sector (e.g., civil engineering, residential), the environment (outdoor, indoor) and the control methods were developed and interpreted using multi-model inference. These analyses were also performed for the respirable dust samples in order to evaluate the presence of effects potentially different from the exposure determinants. Furthermore, cross-tabulation of data on crystalline silica and respirable dusts made it possible to evaluate the differences between the percentages of crystalline silica in the air samples by occupation title, task, tool, material and source control method.The analysis of crystalline silica data for comparison with a TLV was based on 1346 measurements covering 11 categories of occupation titles. The model containing all the variables explained 22% of the variability in the measurements, whereas the year and sampling strategy (regulatory compliance vs. surveillance) were the variables with the most impact on the exposure. The increase in the sampling duration was associated with lower levels, whereas the time trends by strategy went in opposite directions with a reduction of 17% per year (regulatory compliance) compared to an increase of 9% per year (surveillance). The use of means of control (without considering the specific type) reduced the concentrations by 18% outdoors and 24% indoors. The highest predicted geometric means for the year 1999 over eight hours were found in drillers (0.24 mg/m3), underground workers (0.22 mg/m3), roofers working on concrete slab roofing (0.15 mg/m3) and cement finishers (0.13 mg/m3). The effects of the determinants for respirable dusts, estimated from 1137 measurements, were comparable to those for crystalline silica; however, the agreement between the predicted geometric means for the two types of contaminants was moderate (Spearman correlation coefficient of 0.45).For the analysis of the task-related levels, 1466 crystalline silica measurements in 27 task categories were selected. The model containing all the variables explained 60% of the variations in exposure levels, and all of the contextual variables were highly predictive. The predicted geometric means for the year 1998, based on the median duration by task, were higher during concrete bush hammering (0,73 mg/m3), breaking masonry with multiple tools (including jackhammers/percussion drills) (0.59 mg/m3), tunnel boring (0.27 mg/m3), abrasive blasting (0.19 mg/m3) and brick joint grinding (0.19 mg/m3). A major reduction in concentrations was observed with the spraying systems (71%) and dust vacuum systems (69%) integrated into the tools. For respirable dusts (1566 measurements), the efficiencies of the means of control were generally higher, with exposure reductions of 88% and 81% for these same categories. The agreement between the predictions for the two contaminants was also higher, with a Spearman correlation coefficient of 0.70.The median percentage of respirable crystalline silica, calculated from 924 samples, was 11%. The majority of the percentages by category of occupation title, task, tool, material and source control method were between 6% and 16%, with the highest value (19%) being found for the "sand" material category.The data analysis showed a generalized overexposure in relation to the TLV of the American Conference of Governmental Industrial Hygienists (ACGIH) and to the Québec standard, indicating a long-term risk of occupational diseases for all the occupation titles studied. The results obtained for the task-related evaluation show that this strategy provides a better characterization of the exposure-related factors and a better identification of the intervention priorities for controlling the crystalline silica exposure levels on construction sites during a work shift.