Relative risk of elevated hearing threshold compared to ISO1999 normative populations for Royal Australian Air Force male personnel

Abstract

Objective: This paper introduces a new method to calculate relative risks of elevated hearing thresholds, at various ages and frequencies, between a study population and ISO1999:2003: Annex A Screened, Annex B Unscreened and ISO1999 Section 5.3 adjustment for noise exposure using Annex A Screened data. We demonstrate this method on a study population of male Royal Australian Air Force personnel.

Study design: Using a retrospective cohort design, hearing thresholds were assessed in 583 F-111 aircraft maintenance personnel, 377 technical-trade comparisons and 492 non-technical comparisons using pure-tone audiometry. A quantile regression model was used determine whether an association exists between median hearing thresholds and F-111 maintenance, adjusting for possible confounders. The new method involves using quantile regression models with bootstrapped standard errors to estimate percentiles for the study population and thus determine the probability of a greater than 25 dB hearing threshold. This was done for the three ISO datasets as follows; for the ISO1999 Annex A screened population data the formula provided allows the calculation of these probabilities. ISO1999 Annex B unscreened population data only provides the values for the 10th, 50th and 90th percentiles at ages 30, 40, 50 and 60 only, therefore it was necessary to fit a curve to these values in order to estimate the probabilities. For ISO1999 Section 5.3 adjustment for noise exposure population we used the Annex A screened population data plus the formula. The probabilities were then divided to give the relative risks of a greater than 25 dB hearing threshold, at various ages and frequencies.

Results: While no difference was observed between the three groups, the model identified a number of significant confounders, namely tinnitus, smoking, diabetes and the use of anti-depressant medications. Relative risks were high at frequencies 2 kHz and less for the study population of all ages compared to ISO A screened data. The increased relative risks at 4 and 6 kHz give the appearance of a "noise notch" for ages 30 and 40 years. The comparison with the ISO B unscreened data are significantly less than one for frequencies above 2 kHz, particularly for young men and greater than one less than 2 kHz. The relative risks for the comparison to the ISO A screened data with ISO 5.3 adjustments, are highest for young men decreasing with age, with the highest relative risk are at frequencies less than 2 kHz.

Conclusions: This paper demonstrates a new method for quantifying the probability of a clinically relevant hearing loss and the relative risk of the loss due to a risk factor. Prior to this, researchers were reduced to simplistic methods such as visual comparison of deciles which did not enable the estimation of risk. The new method can use all observed hearing thresholds per study participant, adjust for known confounding factors such age and gender, and calculate the relative risk of a clinically relevant increase in hearing threshold due to a risk factor of interest.