Discussion

Based on the results of this cohort, pre-operative imaging of the mastoid has good positive predictive values for cholesteatomas and associated complications (90 per cent and 88 per cent respectively). These high values are comparable to those of other studies.^{Reference Prata, Antunes, de Abreu, Frazatto and Lima1,Reference Yildirim-Baylan, Ozmen, Gun, Yorgancilar, Akkus and Topcu3,Reference Khavasi, Malashetti and Chandrashekarayya4,Reference Mohamed, Yousif and Mehnab7,Reference Jackler, Dillon and Schindler10,Reference Alzoubi, Odat, Al-Balas and Saeed11,Reference Payal, Pranjal, Gul, Mittal and Rai15,Reference Park, Park, Kwon, Kim, Park and Sung16,Reference O'Donoghue, Bates, Anslow and Rothera18} Imaging was not as accurate when predicting a non-diseased ear (65 per cent negative predictive value for cholesteatomas and 76 per cent negative predictive value for complications). There were, however, a large proportion of ‘undecided’ reports (49 per cent) that were not included, as either the radiology report or the operation notes failed to comment on one of the independent variables, thereby precluding valid comparison. This is a limitation of retrospective analysis. Similar results from the literature show that sensitivity values are not as reliable, suggesting that the role of CT in predicting complications of CSOM is not absolute.^{Reference Faramarzi and Roosta21}

Our cohort was representative of the population, as several studies had mean patient ages ranging from 24.02 years to 38.2 years.^{Reference Prata, Antunes, de Abreu, Frazatto and Lima1,Reference Yildirim-Baylan, Ozmen, Gun, Yorgancilar, Akkus and Topcu3,Reference Karki, Pokharel, Suwal and Poudel6,Reference Mohamed, Yousif and Mehnab7,Reference Alzoubi, Odat, Al-Balas and Saeed11,Reference Gerami, Naghavi, Wahabi-Moghadam, Forghanparast and Akbar13,Reference Rogha, Hashemi, Mokhtarinejad, Eshaghian and Dadgostar14} Das et al.^{Reference Das, Ranjan Rout, Shilpa Sri, Mohanty and Vittal Prasad5} and Chintale et al.^{Reference Chintale, Kirdak, Jatale and Shaikh8} had age ranges that reflected our cohort. Khavasi et al.^{Reference Khavasi, Malashetti and Chandrashekarayya4} had a younger cohort, with 45 per cent of their patients with cholesteatomas being between 11 and 20 years.

We had a large number of patients in our cohort. The largest number of patients found in comparative literature was 80.

Invariably, disease progresses chronologically. It can therefore be theorised that a longer time delay between a CT scan and an operation can influence the discrepancy between the two findings. Our median time lag was over seven months. This duration could quite easily allow changes to occur between what is reported on a CT scan and what is found at operation. This is what was found in our cohort, as the prediction of a non-diseased middle ear was lower than that for predicting a cholesteatoma (65 per cent vs 90 per cent respectively). What is not taken into account in our calculations, but is appreciated as a limitation, is that a proportion of the CT scans were used for surveillance or diagnosis, and not as a pre-operative planning adjunct.

It is also worth noting that CSOM and cholesteatomas rarely affect the anatomical locations described in our study in isolation.^{Reference Sethi, Rafferty and Kayarkar22} Variations in the reported frequencies of individual anatomical locations of cholesteatomas must therefore be considered, as well as the calculations of sensitivity and positive predictive value.

A further limitation of this study is the assumptions made associated with the nomenclature used in the reporting of the findings in order to categorise the data effectively. If this was not done, direct comparisons between verbatim reports would be unfeasible and lead to multiple, substantially smaller cohorts.

The sensitivity for predicting cholesteatoma based on pre-operative CT imaging varies in the literature. Our results are most comparable with those of Prata et al.,^{Reference Prata, Antunes, de Abreu, Frazatto and Lima1} Jackler et al.^{Reference Jackler, Dillon and Schindler10} and Alzoubi et al.^{Reference Alzoubi, Odat, Al-Balas and Saeed11} (who reported sensitivity values of 72.73–80 per cent). Other sensitivity values reported ranged from 61.8 per cent to 100 per cent.^{Reference Walshe, McConn Walsh, Brennan and Walsh2–Reference Khavasi, Malashetti and Chandrashekarayya4,Reference Mohamed, Yousif and Mehnab7,Reference Tatlipinar, Tuncel, Öğredik, Gökçeer and Uslu12,Reference Gerami, Naghavi, Wahabi-Moghadam, Forghanparast and Akbar13,Reference Payal, Pranjal, Gul, Mittal and Rai15,Reference Park, Park, Kwon, Kim, Park and Sung16,Reference O'Donoghue, Bates, Anslow and Rothera18}

Much like our sensitivity findings, our specificity was similar to Prata et al.^{Reference Prata, Antunes, de Abreu, Frazatto and Lima1} (81 per cent vs 82.5 per cent respectively). However, specificity ranged from 48 per cent to 94.7 per cent in studies by Yildirim-Baylan et al.^{Reference Yildirim-Baylan, Ozmen, Gun, Yorgancilar, Akkus and Topcu3} and Alzoubi et al.^{Reference Alzoubi, Odat, Al-Balas and Saeed11}

There was variability in both the positive predictive value and negative predictive value within the literature, ranging from 60.6 per cent to 97.1 per cent and from 66.67 per cent to 85.7 per cent, respectively. Multiple studies have demonstrated positive predictive values comparable to our results. The positive predictive values were 77.27 per cent in the study by Prata et al.,^{Reference Prata, Antunes, de Abreu, Frazatto and Lima1} 85.2 per cent in that by Yildirim-Baylan et al.,^{Reference Yildirim-Baylan, Ozmen, Gun, Yorgancilar, Akkus and Topcu3} and 94.3 per cent in the study by Kumaresan and Nirmala.^{Reference Kumaresan and Nirmala9} Payal et al.^{Reference Payal, Pranjal, Gul, Mittal and Rai15} had a considerably lower positive predictive value of 68.95 per cent.

There was a big discrepancy in the sensitivity values for erosion or dehiscence of the facial nerve in the literature. A low sensitivity was also found by Khavasi et al.,^{Reference Khavasi, Malashetti and Chandrashekarayya4} Das et al.^{Reference Das, Ranjan Rout, Shilpa Sri, Mohanty and Vittal Prasad5} and Payal et al.^{Reference Payal, Pranjal, Gul, Mittal and Rai15} (42 per cent, 42.9 per cent and 40 per cent, respectively). These values were considerably lower than in some other studies, such as those by Prata et al.,^{Reference Prata, Antunes, de Abreu, Frazatto and Lima1} Yildirim-Baylan et al.^{Reference Yildirim-Baylan, Ozmen, Gun, Yorgancilar, Akkus and Topcu3} and Karki et al.,^{Reference Karki, Pokharel, Suwal and Poudel6} which reported sensitivity values for facial nerve compromise of over 98 per cent.

Again, there was variability in each of the parameters measured (complications of the ossicles, facial nerve, lateral semi-circular canals and tegmen). Sensitivity values ranged from 33 per cent to 100 per cent.^{Reference Prata, Antunes, de Abreu, Frazatto and Lima1–Reference Khavasi, Malashetti and Chandrashekarayya4,Reference Karki, Pokharel, Suwal and Poudel6,Reference Chintale, Kirdak, Jatale and Shaikh8,Reference Kumaresan and Nirmala9,Reference Tatlipinar, Tuncel, Öğredik, Gökçeer and Uslu12,Reference Payal, Pranjal, Gul, Mittal and Rai15,Reference Park, Park, Kwon, Kim, Park and Sung16,Reference O'Donoghue, Bates, Anslow and Rothera18} Specificities were also widely spread, ranging from 48 per cent to 100 per cent. The incidence of facial nerve dehiscence is relatively low, and therefore the sample size of patients needed to generate accurate statistical measures would generally be larger than the cohorts we have compared against.^{Reference Faramarzi and Roosta21}

Several studies reported high sensitivity values for correctly identifying tegmen erosion on the pre-operative CT scans. Tatlipinar et al.,^{Reference Tatlipinar, Tuncel, Öğredik, Gökçeer and Uslu12} Yildirim-Baylan et al.^{Reference Yildirim-Baylan, Ozmen, Gun, Yorgancilar, Akkus and Topcu3} and Karki et al.^{Reference Karki, Pokharel, Suwal and Poudel6} reported sensitivity values of 90 per cent, 98 per cent and 100 per cent, respectively. Our cohort had a sensitivity of 75 per cent, which is more in keeping with the studies of Khavasi et al.,^{Reference Khavasi, Malashetti and Chandrashekarayya4} Chintale et al.,^{Reference Chintale, Kirdak, Jatale and Shaikh8} Kumaresan and Nirmala,^{Reference Kumaresan and Nirmala9} and Rogha et al.^{Reference Rogha, Hashemi, Mokhtarinejad, Eshaghian and Dadgostar14} (which reported sensitivity values of 66–75 per cent). Our specificity and positive predictive values were high, at 100 per cent. Such values were also found in studies by Prata et al.,^{Reference Prata, Antunes, de Abreu, Frazatto and Lima1} Karki et al.,^{Reference Karki, Pokharel, Suwal and Poudel6} Kumaresan and Nirmala,^{Reference Kumaresan and Nirmala9} and Rogha et al.,^{Reference Rogha, Hashemi, Mokhtarinejad, Eshaghian and Dadgostar14} with a range of 91.93–100 per cent.

This study highlights the scarcity of reported key negative findings in mastoid surgery. It has been suggested by the author that the operation notes more often include positive findings, and that if there is no comment on complications it can be assumed that there are none. This could potentially raise issues during revision surgery, or even from a medicolegal perspective in cases of an unexpected, unreported finding.^{Reference Bateman, Carney and Gibbin23} Locally, we have plans to introduce a standardised operation note to include all relevant pertinent findings, including the absence of such findings.

As suggested by other studies in the literature,^{Reference Walshe, McConn Walsh, Brennan and Walsh2,Reference Yildirim-Baylan, Ozmen, Gun, Yorgancilar, Akkus and Topcu3,Reference Alzoubi, Odat, Al-Balas and Saeed11,Reference Tatlipinar, Tuncel, Öğredik, Gökçeer and Uslu12} pre-operative CT imaging performed prior to mastoid surgery can be a useful adjunct when planning surgery, in terms of anticipating risk factors present and determining the extent of localised complications. However, given the wide variation of correlational findings between studies, further discussion is warranted regarding whether imaging, particularly ionising radiation, should be conducted for all patients undergoing surgery.