We agree with Dr Dorinson and do not want to "dismiss attempts at prevention." However, our research shows that employers and employees should not depend on back belts for prevention. Instead, prevention efforts should be based on a comprehensive workplace evaluation of work tasks and the work environment to reduce the hazards of lifting. Whatever mechanical or psychological effects that back belts may have on workers, our study shows that these effects do not result in decreased incidence of back pain or back injury claims. We are now completing a study to assess biomechanical and physiological effects of back belts. Dorinson has observed that many workers wear belts improperly. In our study, employees had training on proper lifting techniques and proper belt wearing when first hired. Although not reported in our article, belt wearing was validated through direct observation of 227 employees in 8 stores. Back belt use was observed in 77% of participants who reported wearing belts " usually every day" based on a short observation period on a single day. However, belt tightness was not measured. The observations were made as soon as could be arranged after the interview with a median of 16 days between interview and work observation. Dorinson suggests that back belts might have different effects for people with "some preexisting weakness or mechanical deficiency in the spine." We adjusted for a "previous back injury" in our main analysis and also evaluated 2 subgroups of employees separately (both with and without a history of previous back injury) to determine any effects of belt wearing. Back belts provided no statistically significant protection from "biomechanical challenge" for either subgroup based on our study of back pain incidence and injury claim rates. Dr Marshall points out that the unadjusted and adjusted odds ratios and rate ratios were "almost identical" for our evaluation of belt-wearing effects on back pain and injury. Our results do not imply that future studies can dispense with individual-level risk factor data, especially if conducted in different settings or among different populations. Prior to estimating regression results, we recognized many possibilities for effect modification and confounding. Women are less likely to report wearing belts "usually every day" and are more likely to report back pain; current smokers are more likely to report wearing belts "usually every day" and also have a higher rate of back injury claims. The complicated interrelationships among these variables requires a careful regression analysis for an unbiased evaluation of the effects of belts. We did indeed conduct "a full sequence of stratified analyses" and also diagnostic evaluation of the regression models. We chose to emphasize results for the main objectives of our study and the consistency of these results in subgroup, interaction, and other ancillary statistical analyses of data, all of which demonstrate the lack of effects of back belts on back pain and injury.