Thiese et al. 1 which raised questions about our recently published article in AJIM 2 and two other publications 3,4 from our study of working conditions and musculoskeletal symptoms among railroad maintenance-of-way (MOW) workers. We conducted the study jointly with two other research teams who were studying mortality and the social and economic impact of diseases and injuries among MOW workers. We followed guidance from our respective Institutional Review Boards (IRBs) (see publications) and an outside independent scientific advisory panel (see below).We reported the data in detail and acknowledged in our publications the potential limitations of our study, including use of workers' self-report of symptoms and workplace exposures, and a low survey response rate. However, Thiese et al. failed to mention that we assessed the possibility and direction of selection bias in two ways. First, we compared survey respondents to the national membership of the Brotherhood of Maintenance-of-Way Employes Division (BMWED), the union representing MOW workers, on available demographic information. Second, we compared survey respondents to a random sample of nonrespondents, who agreed to complete a 10-question version of the survey by telephone. 3 Compared with nonrespondents, active members who completed the survey were younger, had slightly better working conditions, and tended to be healthier (except for back pain). These results suggest that our analyses may underestimate associations between working conditions and musculoskeletal symptoms among MOW workers. 2 Thiese et al. also fail to mention that we explained how missing data was handled. We stated that "The prevalence of specific working conditions and measures of ill health from survey respondents were computed by dividing the number of respondents reporting that working condition, symptom, or diagnosis, by the number of respondents who completed at least one question on that page of the survey, as the response rate tended to decline in later pages of the survey." 3 p. 587 There was no imputation of missing data.