Concerns with amplitude variation in calibrated audiometer systems in clinical simulations

Christopher A Barlow¹, Lee Davison¹, Mark Ashmore²,  
¹ Maritime and Technology Faculty, Southampton Solent University, Southampton, England, United Kingdom
² Strategic Audiology Services, Somerset, United Kingdom

Correspondence Address:
Christopher A Barlow
Maritime and Technology Faculty, Southampton Solent University, Southampton, England
United Kingdom

Full Text

Sir,

In their commentary in the May edition, [1] Byrne and colleagues discuss the points of concern regarding our study of "Amplitude variation in calibrated audiometer systems in clinical simulations." [2]

Unfortunately, they appear to have misinterpreted the study. This study aimed to examine the accuracy of standard pure tone testing in normal practice and to what degree there was variation in the sound pressure level of tones presented at the ear for patients undergoing audiometric testing, for instance, in a clinic or in an occupational health test. This is particularly important given the degree of litigation relating to occupational hearing loss, which often relies on pure tone tests and where a misdiagnosis could prove very costly.

The study simulated clinical practice using a range of different audiometers, purchased from different companies, and calibrated by their appropriate laboratories. This replicates the situation that would be found in different clinics in normal practice. The tests were undertaken on a Head and Torso Simulator (HATS) (Brüel & Kjær) as this allows simulation of headphone placement on a patient, and the study aimed to replicate the clinical situation of testing different patients with a qualified audiometrist removing and replacing the headphones in between each test.

Byrne et al. state that "nothing is known about the calibration lab(s) that were used and an independent verification was not performed" and use this to claim that this makes the results of the study unreliable. A key tenet of this study was to undertake the simulated tests in as close to a "real world" situation as possible and this requires using a range of audiometers, calibrated by a range of providers. Any variation in calibration between these devices would be replicated between real clinics. If the suggestion of Byrne et al. to verify that all devices were producing identical outputs was followed, this would result in unacceptable sample bias as this would not replicate normal practice.

Byrne et al. also state that 'an audiometer set to generate a 50-dB(HL) tone at 6,000 Hz could produce anywhere from 43.8 dB(HL) to 56.2 dB(HL) and be considered "in calibration."' While this is the case, if a patient goes for a hearing test at a particular clinic, they will assume (as will many audiometrists) that the system is accurate and reliable and that they would get the same results at another clinic. In an ideal world, different calibrated meters should "theoretically" provide identical sound pressure level for each tone presentation. In reality they will not, due to calibration tolerances, and this is a contributing factor to the level of variation seen in this study. It does raise the question of whether the calibration tolerances are too high, given the potential for compound errors in clinical practice.

The results of simulating a number of patients attending different clinics show significant variations of sound pressure levels at the eardrum. These fall into the following two categories: Variation within the same audiometer (test-retest variation) and variation between audiometers (between test variation).

This is a compound effect that includes issues with the lack of control of laboratory calibration, variation in headphone placement, problems with the way in which transducers are calibrated, "allowable" calibration tolerances, dated transducer design, and variations in headband tension.

Byrne et al. state that "the measured differences cannot be assumed to be solely due to earphone placement." Yet, assuming that there is no significant component drift in an audiometer between tests, any test-retest variation of results from the same audiometer is not likely to be caused by transducer or calibration variation. Differences in laboratory calibration between meters will not affect a test-retest result of a single meter and the only parameter that varies in the test-retest simulation is that of headphone placement.

The worst case test-retest variation by a single meter was 11.4 dB. We have, therefore, suggested that variations in tone presentation shown in the results (particularly the often contentious 6-kHz tone) are likely to be caused by small changes in headphone placement. Further studies are needed to identify to what degree this is the case and in particular what the specific effects of varying these parameters might be.

There is also a significant degree of "between test" variation. As Byrne et al. point out, 'some of these variations are only just outside the ±3.7 dB calibration tolerance.' However, these variations go up to ±10.5 dB and give a worst case variation in excess of 20 dB for the same tone in tests by different audiometers. This is far in excess of the calibration tolerances. Given the test-retest variations discussed earlier, the level of variation in results between audiometers is highly likely to be contributed to by headphone placement and further exacerbated by variations in calibration procedure and quality of the laboratories.

Byrne et al. also state that, "Presumably, the only variation between the four audiometers in this study would be caused by headphone positioning, because the same HATS measurement system was used to evaluate all four of the audiometers."

However, we have quite clearly stated that the key issues relate to the calibration process and standards. A significant problem here is the use of audiocups that appears to add considerably to variation in results, both between and within the audiometers. As we have stated in the paper, "the calibration methods specified by the International Electrotechnical Commission (IEC) 60318-1 dictate the use of a standardized coupler or artificial ear and use of these systems requires removal of the supra-aural Telephonics Dynamic Headphone (TDH)-39 from the attenuating cups. When they are replaced, the attenuating cups themselves are likely to introduce a degree of resonance that will change the frequency response of headphones to the ear." We also state that the "current standards for audiometry calibration do not take into account the issues faced in clinical practice."

We also suggest that a potentially significant contributor to test variations could be acoustic coupling issues caused by differences in headband tension, given that there appeared to be variations in the headband tension between the audiometers. This forms part of the discussion of the measured results and is a suggestion that this is a likely contributor. We were clear that headband tension was not measured as part of this study and that "further research needs to be done on the exact impact of headband tension on results." A key point to note is that while headband tension should provide a nominal static force of 4.5 N ± 0.5 N, it is not commonly tested during the calibration process where a jig or a weight is used to provide appropriate coupling to the artificial ear used for calibration. Any set of headphones placed on different patients will be subject to variations in tension due to physiological differences in the size and shape of the head and this is something that could have a significant impact on acoustic coupling.

The results of this study clearly show that a patient could expect up to a variation of ±11 dB in the results at some frequencies if tested in different clinics. They could also expect a high degree of variation (greater than the calibration tolerance) for different tests within the same clinic using the same audiometer. Given that all tests were done on a uniform "head," testing on real patients could further exacerbate these variations.

The margin of error at some frequencies in a test could be sufficient to change the diagnosis of a patient on the borderline of a hearing loss categorization, depending on the clinic(s) that they attend. This could have a significant impact on the patient in terms of their access to treatment and will also have a serious impact on any litigation involved.

In conclusion, despite being regularly referred to as the "gold standard," pure tone audiometry, as it currently stands, has a very high degree of potential error, particularly in a clinical environment. We reject the assertion that "accurate interpretation of the study measurements is not possible" as stated by Byrne et al. The study used repeated measurements of the output of different audiometers bought from and calibrated by different companies. This is representative of clinical practice and the tests included all the contributing factors that can cause variation aside from the differences in the physiology of the patient. The variations in sound pressure levels at the ear in an audiometric test measured here can, therefore, be realistically assumed to occur between different clinics and different patients.

Further studies on a larger scale are still required to identify the relative contribution to this variation of each parameter of the audiometric test. This will help to design improvements in the technology as well as bring forth improvements in the calibration process that are needed to resolve these issues.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

References