HSC Section 8_April 2017

is higher in younger ( < 9 years) than in older CWU patients. 12 Secondly, the ratio between residual and recurrent disease might depend on follow-up length. Therefore, a 5-year follow-up is essential to evaluate both types of disease recurrence. 7,13,14 Thirdly, distinction within the study between residual and recurrent disease was important because of different key aspects in etiology. 8 The aforementioned items for relevance were appointed priority items (see Table I, marked with *). Pri- ority items for validity were: “loss to follow-up”, “missing data”, and “confounding by indication”. Both a high loss to follow-up percentage and missing data could lead to biased outcomes. In the third place, confounding by indication is important because severe cholesteatoma cases tend to be treated by CWD rather than CWU. 15 Subsequent to scoring each individual CAT item, studies received an overall score for both relevance and validity: low (L) (0–2.5 points), moderate (M) (3–5.5 points) or high (H) ( 6 points) (Table I). Articles received 1 point per item when a plus ( 1 ) was scored and 0.5 point when a plus/minus ( 6 ) was scored; priority items received double points. We selected stud- ies with a moderate to high relevance for inclusion in the cur- rent review. Calculations and Statistics We performed statistical pooling of data when there was similarity between studies in: patients’ age groups, applied out- come measures, type of applied statistical analysis and elected follow-up moments. We used Review Manager (RevMan) 5.3 software to establish analysis (version 5.3. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). We preserved meta-analysis results when the I 2 was between 0% and 40%. 16 We extracted original data from included studies to calculate the absolute risk reduction (ARR) and relative risk (RR). We used the Fisher’s exact test in SPSS 22 (IBM Corp., Armonk, NY) to calculate P values. We considered a value of P < .05 significant. We retrieved a total of 2,060 articles, which were all screened by evaluating the title and abstract. From this process, 125 articles were selected for full-text screening (Fig. 1). We considered seven articles eligible to answer our research question (Fig. 1). Snowballing revealed an eighth article 4 , a master’s thesis, which was not indexed in selected databases (Fig. 1). Reverse snow- balling revealed no additional articles. Assessing Studies The overall relevance was moderate in six studies (Table I). One study 4 scored high and one study 17 scored low on relevance. The overall validity of the retrieved studies was graded low to moderate (Table I). Publica- tion years of the articles ranged from 1979 to 2010. Two studies included only adults in their studied cohort. 4,18 In addition, two studies only included acquired choles- teatomas. 4,7 Seven studies did not report whether reop- erations were included. Only one study 19 provided well- described baseline characteristics. Three studies defined the difference between residual and recurrent dis- ease 4,15,18 (Table III). The follow-up duration was at least 5 years in four studies. 6,17,19,20 In only three stud- RESULTS Retrieving Studies

ies, all patients were operated on by the same sur- geon. 7,15,18 In four studies 7,15,17,20 the overall validity was moderate; whereas it was low in the remaining studies. 4,6,18,19 Loss to follow-up was 20% or less in three studies 15,17,20 and the method of handling of missing data was reported in only one study. 15 In two studies, 6,19 the operation technique was described in a protocol. In the study of Brown 6 , the elected surgical technique for cholesteatoma removal was the intact canal wall tympa- noplasty and mastoidectomy, as advocated by House and Sheehy. 21 Only one of our included studies (Ajal- loueyan 19 ) applied an obliteration technique, published previously by Quaranta et al. 22 In the remaining four studies 7,15,17,20 , the operation protocols were not clearly defined. The method of determining whether cholestea- toma recidivism had developed, occurred according to a well-defined protocol in two studies. 6,19 Seven articles were biased by confounding by indication. 4,6,15,17–20 We excluded one of the eight studies 17 with a low relevance: report of adult residual disease only. Conclusions are based on the remaining seven included studies with a moderate to high relevance (Table I). The study reported by Declerck 4 seemed to be the most relevant to answer our research query because adult patients with an acquired cholesteatoma were included. However, the studies of Nyrop 7 and Stankovic 15 scored the highest values on validity in our CAT (Table I). Therefore, results from the latter two studies were suggested to provide the most unbiased insight. There was a high amount of statistical heterogeneity in the reported risk differences (I 2 5 96%). We decided not to include Palmg- ren’s study 20 in the heterogeneity analysis because patient numbers could not be derived from recurrence percentages. In addition, we defined whether heteroge- neity was different for residual or recurrence numbers of the six studies: 89% and 92%, respectively. We consid- ered an I 2 below 40% to be acceptable; therefore, we decided not to pool results. Data Extraction The extracted data are presented in Tables II and III. The selected studies included 1,268 operated ears for the CWU group and 1,038 ears for the CWD group (Table II). The follow-up ranged between 6 months and 10 years. Two studies 4,18 reported results for patients above 18 years and four studies reported results for patients over 15 years. 6,7,15,19 The remaining study 20 mentioned the mean ages of the youngest 50 patients and the oldest 50 patients (Table II). The data of the study of Declerck 4 are displayed in separate rows because one group was followed until the second-look operation and the second had a longer follow-up (mean follow-up: 2.5 years) (Table II). Six articles 4,6,7,18–20 reported a higher percentage of disease recidivism after the CWU (range 15% to 61%) than after the CWD proce- dure (0 to 13%) (Table II). Four of these differences were statistically significant (Table II). Contrarily, Stankovic 15 reported disease recidivism of 8% after the CWU and 22% after the CWD technique ( P < .001). Both the abso- lute risk reduction (ARR range 2 14% to 61%) and the

Laryngoscope 126: April 2016

Kerckhoffs et al.: A Review on Cholesteatoma Recidivism After CWU and CWD

108