The aim of this study was to review the literature regarding human noise-induced permanent threshold shift and to determine whether the observed data agreed with the predictions of two different exchange rates (ERs).
An initial list of possibly relevant studies included those cited by authors who endorsed the 3 dB ER, as well as studies in personal files, studies retrieved by a MEDLINE search, and the reference lists of all of these. Criteria for relevance were designed to ensure that exposures were sufficiently intermittent or fluctuating that effective exposure levels based on the 3 dB (LAeq8h) and 5 dB (time-weighted average [TWA]) ERs would differ by at least 1 dB, that at least one of these metrics could be estimated, and that audiometric data were available for groups of defined age, sex, and exposure. Relevant studies were reviewed in detail, and their audiometric data were compared with the predictions of the ISO-1999/ANSI S3.44 model.
Nine relevant studies were identified. For six articles, the reported hearing levels were substantially less than would have been predicted from LAeq8h. In each of these cases, TWA would have predicted lower hearing levels than LAeq8h and would have better fit the observed data. In three cases it was not possible to say which ER would have better fit the observed data.
The 3 dB ER systematically overestimates the risk of noise-induced hearing loss for intermittent or fluctuating noise. The 5 dB ER appears to be more accurate, but also overestimates risk, particularly for exposures above 100 dBA.