Evaluation of Residual Disease Following Transcanal Totally Endoscopic vs Postauricular Surgery Among Children With Middle Ear and Attic Cholesteatoma

This propensity score–matched cohort study of pediatric tympanic cholesteatoma resections investigates the association between the surgical method—transcanal endoscopic or postauricular approach—and rates of residual disease at follow-up.

Key Points

Question

For cholesteatoma limited to the middle ear and/or attic in children, is a 2-handed, postauricular approach associated with reduced risk of residual disease compared with use of transcanal totally endoscopic resection?

Findings

In this propensity score–matched analysis of 209 cases of primary cholesteatomas limited to the middle ear and/or attic, the incidence of residual cholesteatoma after transcanal totally endoscopic ear surgery was not statistically significantly higher than that after open surgery with a postauricular approach using an operating microscope with or without an endoscope.

Meaning

This study’s findings suggest that the ability to use 2 hands offered by use of an operating microscope is not associated with reduced incidence of residual cholesteatoma; therefore, transcanal totally endoscopic ear surgery may be an effective alternative for some children.

Abstract

Importance

Transcanal totally endoscopic ear surgery (TEES) improves tympanic cavity visibility during cholesteatoma resection but does not readily permit 2-handed surgery, which is possible with a postauricular approach and use of an operating microscope. Improved visibility and 2-handed access may reduce rates of incomplete surgical clearance and residual cholesteatoma.

Objective

To compare rates of residual cholesteatoma after TEES vs those after open surgery with a postauricular approach for initial disease limited to the middle ear and/or attic.

Design, Setting, and Participants

This propensity score–matched cohort study considered a consecutive sample of 209 pediatric cases of cholesteatomas resected at a tertiary referral center between January 1, 2005, and December 31, 2017, by either TEES or postauricular approach. The 177 cases (84.7%) in 172 patients that were eligible for inclusion in the study had undergone at least 1 of the following assessments: second-look tympanoplasty, diffusion-weighted magnetic resonance imaging, or in-clinic follow-up examination at least 2 years after the operation. Ears were matched on the basis of the propensity score, a function of age, sex, comorbid diagnoses, etiology, disease extent, ossicular chain condition, presence or absence of tympanic membrane perforation, and otorrhea status at the time of surgery.

Exposures

Transcanal totally endoscopic ear surgery was defined as a transcanal approach with surgeon-reported use of an endoscope for resection of cholesteatoma. Use of an operating microscope was permitted in the TEES group for graft placement during tympanoplasty. The postauricular approach was characterized by a postauricular incision and use of an operative microscope with or without adjunct use of an endoscope.

Main Outcomes and Measures

Residual cholesteatoma.

Results

Of 177 cases included in the study, the unmatched cohort consisted of 65 ears that had undergone TEES (mean [SD] age, 9.9 [3.6] years; 33 [50.8%] female) and 112 ears of patients who had undergone surgery with a postauricular approach (mean [SD] age, 9.9 [3.5] years; 72 [64.3%] male). Propensity score matching was accomplished for a total of 128 cases, with 64 in each group. After matching, standardized differences between TEES and postauricular approaches were less than 0.20 for all measured baseline characteristics, indicating balanced groups. The TEES group had a residual disease incidence of 4 of 64 (6.3%) compared with 7 of 64 (10.9%) in the propensity score–matched postauricular approach group (absolute risk difference, –4.7%; 95% CI, –14.8% to 5.4%).

Conclusions and Relevance

This study’s findings suggest that there was no apparent advantage of the 2-handed postauricular microscopic approach with respect to surgical clearance of cholesteatoma limited to the middle ear and/or attic. Transcanal totally endoscopic ear surgery may be an effective alternative to a postauricular approach for removal of limited cholesteatoma in children.

Introduction

Complete resection of cholesteatoma is a requisite step toward establishing a safe, dry, waterproof, and hearing ear—the ultimate goal of cholesteatoma surgery. Residual cholesteatoma may be present after primary resection without recurrence of the process that generated the disease initially; it results from incomplete resection. On occasion, this is intentional, such as when complete resection of the cholesteatoma matrix risks greater harm than its partial removal; however, typically, incomplete resection results from inadvertent failure to achieve adequate dissection of cholesteatoma matrix from underlying bone. Inability to visualize or access the limits of disease contributes to this risk.

Endoscopes offer improved visualization beyond that achieved with operating microscopes for both transcanal¹ and posterior tympanotomy² approaches to cholesteatoma and have therefore been lauded as a useful adjunct to identify residual disease following microscopic resection.^3,4,5 The improved visualization offered by the endoscope may also obviate a postauricular incision and mastoidectomy, which is a more invasive and more costly approach.⁶ However, because endoscopic surgery is completed using only 1 hand, the other being required to position the endoscope, it may be considered that the postauricular, microscope-guided approach provides the potential advantage of using 2 hands for cholesteatoma dissection. The application of transcanal endoscopic ear surgery (TEES) to avoid a postauricular incision and mastoidectomy would be more difficult to justify if the 2-handed postauricular approach improves the probability of complete disease clearance.

One of the challenges in comparing observed outcomes after alternative surgical approaches is that allocation to surgical approach is not random. Allocation bias can in part be accounted for in the analysis by modeling the likelihood of using each approach for each patient, generating what is known as a propensity score. Patients can then be matched according to their propensity scores, and outcomes can be compared across matched pairs of individuals who are similarly likely to have received either surgical approach. We sought to compare the incidence of residual disease among pediatric patients with cholesteatoma that was amenable to resection with either TEES or a postauricular approach by using a propensity score–matched analysis.

Methods

Study Design, Setting, and Participants

Pediatric patients (aged <18 years) undergoing primary resection of a cholesteatoma between January 1, 2005, and December 31, 2017, by a single surgeon were enrolled for prospective collection of demographic and clinical characteristics, details of surgical intervention, and recidivism outcomes. Consecutive cases of TEES or surgery with a postauricular approach with an operating microscope for cholesteatoma limited to the middle ear and/or attic were eligible for study inclusion. Those who underwent second-look tympanoplasty or postoperative diffusion-weighted magnetic resonance imaging for the purpose of detecting recidivism or who had at least 2 years of follow-up without clinical evidence of recidivism were included. Cases in which cholesteatoma extended beyond the attic into the antrum or other mastoid air cells were excluded as being potentially beyond the limits of the TEES approach. The study was approved by the Research Ethics Board at the Hospital for Sick Children, Toronto, Ontario, Canada. Written informed consent was obtained from decision-makers of included participants.

Surgical Technique

Study participants were assigned to either the TEES group or the postauricular approach with operating microscope (PA) group according to the surgical approach and technique used for primary resection of cholesteatoma. TEES was defined as a transcanal approach with surgeon-reported use of a 2.7-, 3.0-, or 4.0-mm 0°, 30°, or 45° rigid endoscope for resection of cholesteatoma. Use of an operating microscope was permitted in the TEES group for facilitating graft placement during tympanoplasty only. With the use of International Otology Outcome Group descriptors, the TEES approach can be summarized as S1, A1, MX or M2a, Ex or E2, Ox.⁷ Postauricular access was characterized by a postauricular incision with or without cortical mastoidectomy and use of an operating microscope with or without adjunct use of an endoscope for cholesteatoma resection (S1, A4, Mx or M1 or M2a, Ex or E2, Ox).

Residual Cholesteatoma

Residual cholesteatoma was identified within 2 years after the primary surgery by second-look tympanoplasty or by diffusion-weighted magnetic resonance imaging. Residual disease was differentiated from recurrent disease by the absence of continuity with tympanic membrane retraction or a potential implantation event between the date of primary surgery and the identification of recidivism.

Covariates

Demographic and clinical characteristics were recorded at the time of surgery and included those plausibly associated with allocation to the surgical technique. Comorbid diagnoses recorded included cleft palate and a diagnosed syndrome or syndromic features. Etiology was operationalized as congenital, primary acquired (from retraction of the pars tensa, the pars flaccida, or both), or secondary acquired (from previous myringotomy or tympanoplasty). Disease extent was staged prospectively according to the system proposed by Saleh and Mills⁸ separately for site (S1-S5) and ossicular chain involvement (O0-O3). The maximum possible S stage in our sample was 3 given the exclusion of cases with disease extending beyond the attic and middle ear. Owing to limited numbers of S3 and O3 stages in the cohort, the S stage was operationalized as S1 or greater than S1, and the O stage was operationalized as O0, O1, or greater than O1.

Statistical Methods

Propensity score matching was used to mitigate potential bias introduced by nonrandom allocation of patients to surgical techniques. A logistic regression was used to model the log odds of the probability that a patient would receive TEES (propensity score) as a function of age, sex, comorbid diagnoses, etiology, disease extent, ossicular chain condition, and presence or absence of perforation and otorrhea at the time of surgery. Each TEES case was matched 1:1 with a PA case on the basis of this modeled propensity score for each patient. An optimal matching algorithm was used that minimized the pairwise differences in propensity scores between TEES and PA cases. The maximum allowable distance between propensity scores (ie, the caliper) was set to 0.25 times the SD of the logit of the propensity score. Differences in baseline characteristics between PA and TEES groups were reported as standardized differences, which represent the difference between means or proportions in units of pooled SD.⁹ Values less than 0.25 are consistent with adequately balanced groups on measured variables.¹⁰

For the unmatched and propensity-matched cohorts, incidences of residual disease in TEES and PA groups were reported as the absolute residual disease risk difference with the 95% CI. All analyses were performed using SAS Studio, version 3.8 (SAS Institute Inc).

Results

Cohort Characteristics

Between 2005 and 2017, a total of 222 primary cholesteatomas limited to the middle ear and/or attic that were resected by one of us (A.L.J.) were considered for inclusion in this cohort study. TEES or PA was used for resection of 209 lesions (94.1%). A total of 177 cases in 172 patients met inclusion criteria for potential detection of residual disease with at least 1 of the following examinations: second-look tympanoplasty (76 [42.9%]), diffusion-weighted magnetic resonance imaging (29 [16.4%]), or 2-year or postoperative follow-up assessment without clinical evidence of recidivism on history or physical examination results (78 [44.1%]).

The unmatched cohort included 65 cases in which TEES was performed (mean [SD] age, 9.9 [3.6] years; 33 [50.8%] female) and 112 PA cases (mean [SD] age, 9.9 [3.5] years; 72 [64.3%] male) (Table 1). There was imbalance between the groups in regard to sex (standardized difference, 0.30), presence of a diagnosed syndrome or syndromic features (2 of 65 [3.1%] in the TEES group vs 14 of 112 [12.5%] in the PA group; standardized difference, 0.34), and presence of otorrhea at surgery (3 of 65 [4.6%] in the TEES group vs 18 of 112 [15.2%] in the PA group; standardized difference, 0.35). The distributions of Saleh and Mills⁸ stages were similar between groups, with standardized differences all below 0.20 in this cohort restricted to cholesteatoma limited to the middle ear and attic. The proportion of TEES cases had a linear association with year, with 42% of cases (5 of 12) in 2011 and 93% (13 of 14) in 2016 (R² = 0.84).

Table 1. Baseline Characteristics of Cohorts Before and After Propensity Score Matching^a.

Characteristic	Before Matching			After Matching
	TEES (n = 65)	PA (n = 112)	Standardized Difference	TEES (n = 64)	PA (n = 64)	Standardized Difference
Age, mean (SD), y	9.9 (3.6)	9.9 (3.5)	0.01	9.9 (3.6)	9.3 (3.7)	0.16
Female	33 (50.8)	40 (35.7)	0.30	32 (50.0)	26 (40.6)	0.19
Cleft palate	6 (9.2)	16 (14.3)	0.15	6 (9.4)	6 (9.4)	0.00
Missing	0	2 (1.8)
Syndromic	2 (3.1)	14 (12.5)	0.34	2 (3.1)	1 (1.6)	−0.06
Missing	0	2 (1.8)
Etiology
Congenital	17 (26.2)	23 (20.5)	−0.12	17 (26.6)	19 (29.7)	0.07
Canal	0	4 (3.6)		0	0
Flaccida	6 (9.2)	7 (6.3)	−0.10	6 (9.4)	6 (9.4)	0.00
Tensa	33 (50.8)	70 (62.5)	0.21	33 (51.6)	33 (51.6)	0.00
Flaccida and tensa	4 (6.2)	4 (3.6)	−0.05	3 (4.7)	2 (3.1)	−0.08
Secondary acquired	5 (7.7)	4 (3.6)	−0.17	5 (7.5)	4 (6.3)	−0.07
Saleh and Mills S stageb
S1	35 (53.9)	61 (54.5)	−0.01	35 (54.7)	34 (53.1)	−0.03
S2 or S3	30 (46.2)	45 (40.2)	−0.08	29 (45.3)	26 (40.6)	−0.09
Missing	0	6 (5.4)
Saleh and Mills O stagec
O0	28 (43.1)	40 (35.7)	−0.15	28 (43.8)	27 (42.2)	−0.03
O1	17 (26.2)	31 (27.7)	0.07	16 (25.0)	18 (28.1)	0.07
O2 or O3	20 (30.8)	39 (34.8)	0.09	20 (31.3)	19 (29.7)	−0.03
Missing	0	2 (1.8)
Otorrhea at surgery	3 (4.6)	18 (15.2)	0.35	3 (4.7)	2 (3.1)	−0.05
Missing	0	1 (0.9)
Perforation at surgery	26 (40.0)	49 (43.8)	0.09	26 (40.6)	27 (42.2)	0.03
Missing	1 (1.5)	1 (0.9)

Abbreviations: Missing, cases with data missing; PA, postauricular approach with operative microscope; TEES, transcanal endoscopic ear surgery.

^aData are presented as number (percentage) unless otherwise indicated.

^bSaleh and Mills⁸ S stage indicates the number of sites affected by cholesteatoma: S1 indicates a single disease site; S2, 2 sites; and S3, 3 sites.

^cSaleh and Mills⁸ O stage indicates the presence and extent of ossicular involvement: O0 indicates intact ossicles; O1, incus erosion and ossicular discontinuity; O2, incus and stapes arch erosion; and O3, absent malleus hand and incus and stapes arch erosion.

Propensity Score Matching

Successful propensity score matching was accomplished for 128 cases, with 64 in each group. After matching, standardized differences between TEES and PA cases for all baseline characteristics were less than 0.20, indicating balanced groups (Table 1).

Residual Disease Incidence

Among all 177 included cases, 16 (9.0%) had residual disease. All cases of detected residual disease were confirmed with second-look tympanoplasty. The incidence of residual disease in the TEES group was 6.2% (4 of 65) compared with 10.7% (12 of 112) in the PA group (absolute risk difference, −4.6%; 95% CI, −12.7% to 3.6%; Table 2).

Table 2. Incidence of Residual Disease According to Operative Technique.

Cohort	No./Total No. (%)		Absolute Risk Difference, % (95% CI)
	Transcanal Totally Endoscopic Ear Surgery	Postauricular Approach With Operating Microscope
Unmatched	4/65 (6.2)	12/112 (10.7)	−4.6 (−12.7 to 3.6)
Propensity score–matched	4/64 (6.3)	7/64 (10.9)	−4.7 (−14.8 to 5.4)

In the propensity score–matched analysis, the TEES group had a residual disease incidence of 6.3% (4 of 64) compared with 10.9% (7 of 64) in the PA group (absolute risk difference, −4.7%; 95% CI, −14.8% to 5.4%).

Discussion

In this propensity score–matched cohort study, TEES was not associated with an increased risk of residual disease compared with PA (with or without use of an endoscope) among children with cholesteatoma limited to the middle ear and/or attic.

The adjunct use of the endoscope with the microscope in a hybrid approach has been shown to improve visualization and can reduce the risk of residual disease.^1,5,11 Previous analysis of an earlier subset of our data also found a lower residual disease rate for middle ear cholesteatoma when adjunct endoscopy was used for dissection rather than simple inspection as part of a postauricular approach.¹² The current study is the largest published cohort study comparing TEES with a postauricular approach to cholesteatoma, to our knowledge. Our findings suggest that the use of both hands to dissect cholesteatoma, which is possible while using an operating microscope but not with an endoscope, does not provide an advantage with respect to completeness of resection of cholesteatomas limited to the middle ear and/or attic in children. This result has important implications for the management of such cases given the demonstrated incentives to avoid a postauricular incision, including reduced morbidity and greater cost savings.⁶

The observed residual rates in our cohort were lower than comparable published estimates, which range from 23% to 34%.^3,13 This is explained by the restriction of our cohort to limited disease, our inclusion of cases of congenital cholesteatoma, and our approach to defining eligible participants. Other studies have defined residual cholesteatoma by its detection on second-look inspection, but patient selection for a second look is frequently motivated by difficult dissection or another reason for high preinspection probability of residual disease. Inclusion of patients who had no clinical evidence of residual disease after 2 or more years of follow-up and who did not undergo a second-look procedure would therefore have a relatively reduced observed residual rate.

A challenge of comparing different operative techniques is the potential for allocation bias to explain a portion of observed associations. Restricting the analyzed cohort to patients with cholesteatoma that could be feasibly approached by either TEES or a postauricular microscopic dissection is an important step in facilitating a clinically meaningful analysis. However, the decision to proceed with TEES or with a postauricular approach for any particular case is not random. We observed a linear increase in the proportion of TEES cases over time, which is consistent with the hypothesis that learned surgical proficiency and comfort are important factors in approach allocation. This observation is intuitive: earlier in any transition to a new surgical technique, a surgeon is likely to have a lower threshold for reverting to an alternative technique he or she is more comfortable with. Similarly, cases that are completed using the new technique early in its adoption are likely to be less technically challenging and therefore less likely to have poor outcomes, a conclusion supported by the observed differences in baseline characteristics between groups in our cohort. For these reasons, a strength of our study is the use of propensity score matching to mitigate the risk of selection bias. We were able to balance groups on measured characteristics with this approach, providing for a relevant comparison of residual disease rates.

Limitations

Our study was limited by a low event rate that was due in part to the relative rarity of residual cholesteatoma after resection of limited disease. The resulting CIs did not exclude the possibility of equal residual disease risk despite an observed risk difference of 4.7% favoring totally endoscopic ear surgery. Nonetheless, our findings suggest that TEES is unlikely to be clinically significantly worse than a postauricular approach with respect to risk of residual cholesteatoma in our cohort. Like any observational study, our conclusions are also limited by the availability and completeness of recorded data. There may be unmeasured characteristics associated with allocation to treatment groups that also may affect residual disease rates. For example, intraoperative bleeding has been shown to be associated with risk of residual disease, but these data were inconsistently recorded for our cohort.¹⁴ The risk of underspecifying the propensity score model by omission of this factor is mitigated by inclusion of variables that are likely to be colinear with bleeding, including the presence of otorrhea and extent of cholesteatoma.

Another limitation is that second-look tympanoplasty, the criterion standard for detection of residual disease, was not performed for all patients, and residual cholesteatomas may therefore have been missed. However, the absence of clinical evidence of residual disease after 2 years of follow-up has been used as a marker of disease clearance and is more rigorous than the 1-year limit we and others have used previously.^4,12

A desire to continue to improve the quality of available evidence invites consideration of alternative study designs. A randomized clinical trial eliminates the potential treatment allocation bias, but this is not a feasible design to approach the study question. The need for surgeons to develop technical proficiency with a new technique is the principal barrier to randomization, and several other issues related to the application of randomized trials to evaluate the effectiveness of procedural interventions have been described.¹⁵ A continued reliance on observational evidence is more likely. The relative rarity of residual cholesteatoma and the limited number of endoscopic cases currently performed early in the adoption cycle make it challenging to collect the data necessary to draw strong conclusions about its effectiveness from any single institution. Multicenter collaboration and data sharing are therefore particularly important initiatives, and they demand standardized data collection and reporting.^7,16 Even with standardized data, interpretation is challenged by inconsistent indications for endoscopic surgery and varying thresholds for a more invasive approach. Propensity methods may therefore be an important analytic technique in future collaborative efforts.

This study was restricted to analysis of surgery for cholesteatoma confined to the middle ear and attic, sites readily within reach of TEES. We did not include cholesteatoma extending into the mastoid antrum, even though this can often be removed effectively with TEES. The lateral semicircular canal is often used as a guide for an achievable limit to TEES dissection, but use of this structure as a technical boundary has not yet been validated. We note that the larger atticotomy or atticoantrostomy (M2b in International Otology Outcome Group nomenclature⁷) that is required for access can be difficult to reconstruct effectively, at least in children, and question whether this larger bone defect may increase the risk of subsequent recurrent cholesteatoma.¹⁷ Further study of the comparative effectiveness of TEES and PA with respect to rates of recidivism in children with more extensive disease is needed.

Conclusions

The findings in this propensity-matched cohort study suggest that TEES may be an effective alternative to a postauricular approach for removal of cholesteatoma limited to the middle ear and/or attic in children.