HSC Section 8_April 2017

I. WEGNER ET AL.

those studies with a low prevalence of otosclerosis. Prevalence of disease in these studies may have been influenced by the choice of inclusion and exclusion criteria (suspected otosclerosis based on clinical history and physical examination versus intraoperatively con- firmed otosclerosis) and the choice of reference test (middle-ear surgery versus histopathology). Previous studies and reviews highlight the importance of CT in the diagnostic evaluation of otosclerosis (2,3,28). However, the prevalence of otosclerosis in patients with a clinical suspicion of otosclerosis and/or an indication for surgery is generally high. As a result, preoperative CT has little to add in establishing otoscle- rosis and may not be necessary to confirm the diagnosis. Although its role in diagnosing otosclerosis is limited, CT imaging may still be useful in establishing the extent of disease and cochlear involvement, and in detecting concomitant abnormalities. CT scans were qualitatively analyzed in the included studies. Otosclerosis is usually confirmed on CT on the basis of visual confirmation of double ring signs, hypo- densities around the otic capsule, and/or thickening of the footplate. Several authors adopted a more quantitative approach and measured bone densities in the area immediately anterior to the oval window: the fissula ante fenestram (20,21,29,30). These studies did find statisti- cally significant differences in Hounsfield units measured over the fissula ante fenestram region between patients with otosclerosis and control patients, but not in several other regions surrounding the otic capsule. Tringali et al. (20) performed analyses in a subgroup of patients with otosclerosis and normal-appearing CT scan and found no significant differences for densitometric measurements in this subgroup compared with control subjects without otologic disease and control patients with cholesteatoma. Unfortunately, none of these studies defined a cut-off value that can be used to create two-by-two contingency tables and calculate diagnostic test measures. Quality of Evidence The majority of the included studies were character- ized by a moderate relevance and moderate risk of bias. Most studies only included patients with surgically con- firmed otosclerosis. As a result, prevalence of otoscle- rosis is 100% in these studies and only sensitivities can be calculated. Second, only studies using surgery or histology as a reference test were included, since this is the only adequate reference test for confirming oto- sclerosis. This implies that all included patients had an indication for surgery. This causes a substantial risk of selection bias, because the study populations probably will not include patients with very mild disease or patients with a severe sensorineural component. There- fore, these results will not translate into the general group of patients with otosclerosis. Potential Biases in Review To our knowledge, this is the first review to not only systematically evaluate, but also critically appraise the

100%, but 74% (3). In one study, which included various patients with an indication for middle ear surgery, the prevalence was very low (19). Diagnostic Test Measures The sensitivity ranged between 60 and 95% in seven studies (2,3,6,16–20). The two studies (2,3) that com- pared CT to histopathology found lower sensitivities of 60 and 66% than the studies comparing CT to middle ear surgery (6,16–20). Specificity was 100% in two studies (3,18). In another two studies specificity was lower: 83% and 75% (16,17,19). Specificity could be calculated for one of the studies that compared CT to histopathology and was not different from the specificities found in the studies that compared CT to middle-ear surgery (100% (3) compared with 75 to 100% [16–19]). Post-test Probability of Otosclerosis In patients with a clinical suspicion of otosclerosis, the positive post-test probability of otosclerosis with an abnormal CT was 99% (16,17). The positive post-test probability was much lower in a study population of patients with an indication for middle-ear surgery (23%) (19). The negative post-test probabilities were reported in three studies: 51%, 53%, and 67% (3,16–18). In patients with an indication for middle-ear surgery, the negative post-test probability of otosclerosis with a normal CT was only 3% (19). Summary of Main Results The pretest probability or prevalence of otosclerosis was very high in the majority of the included studies (2,3,6,16–18,20). In those studies with a high prevalence of disease (74–97%), both positive and negative post-test probabilities were (relatively) high (3,16–18). Positive post-test probability was 99% in one of these studies (16,17) and negative post-test probabilities ranged between 51% and 67% (3,16–18). In one study (19) with a low prevalence of disease (9%), both positive and negative post-test probabilities were much lower com- pared with studies with a high prevalence of disease (23% and 3%, respectively). Overall, reported sensitivities ranged between 60% and 95% (2,3,6,16–20). The sen- sitivity and specificity for operatively confirmed otoscle- rosis were 61 to 95% (6,16–20) and 75 to 100% (16–19), respectively. The sensitivity and specificity for histopa- thologically confirmed otosclerosis were 60 to 66% (2,3) and 100% (3), respectively. The diagnostic measures and post-test probabilities were generally higher in newer studies. The advance- ments in scanners and techniques may have attributed to higher detection rates in newer studies. Prevalence influ- ences post-test probabilities; post-test probabilities are generally higher in studies with a high prevalence of disease and lower in studies with a low prevalence of disease. Indeed, the studies with a high prevalence of otosclerosis reported higher post-test probabilities than DISCUSSION

Otology & Neurotology, Vol. 37, No. 1, 2016

146