DOI: 10.1097/MAO.0b013e3181be6884
,
Issn Print: 1531-7129
Publication Date: 2010/02/01
Facebook
Twitter
LinkedIn
Email
 |
Checking for direct PDF access through Ovid | |
Excerpt
Saliba et al. (1) addressed an important issue: in cases of asymmetric hearing loss, what is the best audiometric criterion for referral for magnetic resonance imaging (MRI) to rule out acoustic tumor? This question is usually couched in sensitivity and specificity, with MRI as the reference standard. If one criterion (e.g., asymmetry of ≥15 dB at 3 kHz) has both better sensitivity and better specificity than another, it is considered preferable.
This type of analysis is most valid when a single group of patients all receive both the diagnostic test (audiometry) and the reference standard test (MRI). Jaeschke et al. (2) point out that the performance of a diagnostic test will be considerably overestimated when investigators enroll separate disease and normal populations. They also state that, "the properties of a diagnostic test will be distorted if its results influence whether patients undergo confirmation by the reference standard." Both of these biases seem to apply to the study by Saliba et al. The tumor patients had all been referred to them, while a separate control group consisted of their own patients who underwent MRI with no tumor found. Moreover, the tumor patients had been referred to their tertiary care clinic by clinicians - presumably otolaryngologists - who almost certainly had used audiometric data to decide whether an MRI was necessary.
These biases make it impossible to estimate either sensitivity or specificity applicable to patients presenting with asymmetrical hearing loss. For a given audiometric criterion, the appropriate estimates would be:
We do not know the denominator for sensitivity. If the referring doctors had offered MRI only to patients with large asymmetries, they would have missed many tumors; to estimate sensitivity without knowing how many tumors had been missed in the referring practices would lead to an overestimate of sensitivity.
We know neither numerator nor denominator needed to calculate specificity for patients in the referring practice.
A more subtle point is that this study cannot even appropriately rank different decision variables. The authors claim that a 3-kHz asymmetry is better than asymmetry at other frequencies, but if the referring doctors had considered asymmetry at this frequency essential to deciding who needed an MRI, the tumors they found would necessarily be in people with the 3-kHz asymmetry.
