Endoscopic type I tympanoplasty is as effective as microscopic type I tympanoplasty but less invasive—A meta‐analysis

Endoscopic type I tympanoplasty was originally introduced in the 1990s, and the extensive spread of this practice can be easily observed. The conventional technique performed involves the repair of a tympanic membrane perforation, and is defined as microscopic type I tympanoplasty.


| INTRODUC TI ON
The increased proportion amongst subjects suffering from chronic tympanic membrane perforations and, interestingly, do not want to be operated on, suggests there is a considerable need for novel therapeutic procedures. 1 An additional systematic review analysing the burden of disease caused by otitis media, in which chronic suppurative otitis media incidence rate was 4.76‰ or at or about 31 million cases, concluded that 22.6% of cases occurring annually were patients under five years old. Otitis media-related hearing impairment has a prevalence of 30.82 per ten-thousand. Annually, 21 000 individuals succumb due to complications of otitis media. 2 A study found the prevalence of chronic tympanic membrane perforation amongst adults 0.9%, eight out of nine subjects refused tympanoplasty for various reasons. 1 The microscopic tympanoplasty has been the standard procedure regarding the effective reconstruction of a perforated tympanic membrane, dating back to the middle of the 20th century. This approach has many disadvantages including poor cosmetic results, caused by a retroauricular incision and the necessity in performing canaloplasty primarily in cases of anterior perforation. 3 Since endoscopic ear surgery was first performed in the 1990s, 4 it represents the ever-increasing, minimally invasive branch of otologic surgery. An endoscope is an ideal tool, enabling the surgeon to perform transcanal endoscopic ear surgery in its entirety while the classical microscopic approach will most likely require an additional external incision, or a mastoidectomy. 3 In regards to the endoscope, the external incision, soft tissue dissection and bone removal can effectively be avoided. In addition, the consumption of medical resources can also be diminished and partly reflecting a shorter rate of hospitalisation. 5 This approach also provides a wider perspective in reference to enabling access into 'hidden spaces' deep within the middle ear cavity. 6

| Objectives
In 2016, a superbly written meta-analysis was published which collected evidence regarding both endoscopic and microscopic type I tympanoplasty. 7 Since this topic is of particularly high interest, several new studies, including high-quality randomised controlled trials (RCTs), have been recently published. The large body of evidence accumulated proved inspirational, resulting in our intensive review of the plethora of recently published literature. All relevant papers comparing the safety and efficacy of endoscopic type I tympanoplasty to that of microscopic type I tympanoplasty were collated. Our paper aims to determine which method proves more effective in and applicable to the treatment of tympanic membrane perforation.

| MATERIAL S AND ME THODS
The Preferred Reporting Items for Systematic Reviews and Metaanalysis (PRISMA) guidelines were used to report our results. 8 The protocol of this study was registered with PROSPERO (registration number: CRD42018095616; www.crd.york.ac.uk/PROSPERO).

| Ethical considerations
There were no ethical considerations.

| Eligibility criteria
Following the PICO framework, eligible studies examined patients (P) at various age, suffering from dry, central tympanic membrane perforation and, who underwent type I tympanoplasty, carried out with either the endoscope (I) or the microscope (C), and discussed strategic, postoperative outcomes (O). Patients with sensorineural or mixed hearing loss, discharging ear, cholesteatoma, ossicular chain abnormality or combating an active ENT infection were excluded.
Randomised and non-randomised studies (nRCT), including conference abstracts, were all included. Case reports, case series, review articles, letters, editorials and comments were excluded.
Conclusions: Based on our meta-analysis, the surgical outcomes of endoscopic type I tympanoplasty in terms of graft uptake rate, postoperative hearing results and operation time were comparable to the microscopic type I tympanoplasty. In regards to cosmetics, an increase in desirable results was achieved in the endoscopic group, particularly the incidence of canaloplasty which proved to be significantly lower.

Keypoints
• Endoscopic type I tympanoplasty is as effective as microscopic type I tympanoplasty but less invasive.
• The surgical outcomes of endoscopic type I tympanoplasty in terms of graft uptake rate, postoperative hearing results, and operation time were comparable to the microscopic type I tympanoplasty.
• In regards to cosmetics, an increase in desirable results was achieved in the endoscopic group, particularly the incidence of canaloplasty which proved to be significantly lower.

| Outcome measures
The graft uptake rate served as our primary outcome. Regarding inclusion, at least a six-month follow-up was required. Our secondary outcomes included postoperative audiological outcomes, the need for canaloplasty, average operation time and cosmetic results. Seven studies 6,[9][10][11][12][13][14] compared the postoperative hearing outcomes based on the average air-bone gap (ABG); hence, we used the ABG means in support of our analysis. However, four of the included studies [15][16][17][18] reported the postoperative ABG, in accordance with the Committee on Hearing and Equilibrium of the American Academy of Otolaryngology, Head and Neck Surgery. 19 In this event, we determined the postoperative hearing results amongst three groups. Group 1 shows the ABG ranging from 0 to 10 dB, Group 2 contains outcomes with ABG ranging from 11 to 20 dB, while in Group 3, the ABG range was 21-30 dB. The necessity of canaloplasty and the duration of the surgery were determined in the review of the surgical reports. The preferred method for reporting cosmetic results was discrepant across the studies, which may influence the precision of the evaluation of cosmetic results.
To evaluate the postoperative cosmetic outcome, we collated data creating two groups: good cosmetic results and poor cosmetic results, respectively (Appendix S1.).

| Search and selection
We performed a systematic search in the databases of PubMed, Embase,

| Data extraction and management
The two authors (IP and IT) independently imputed the extracted data onto a previously edited Excel table. We collected general information including the name of the first author and the year of publication, study design, initial population, the total number of in-

| Risk of bias (ROB) assessment
RoB in the individual studies was independently assessed by two authors (IP and IT) (Appendix S1). Randomised controlled trials were assessed using the Cochrane Risk of Bias Tool, in compliance with the following domains: random sequence generation, allocation concealment, blinding of participants, personnel, and outcome assessment, incomplete outcome data, selective reporting and other bias. 20 In the case of nRCTs, we used the topic-tailored version of Newcastle-Ottawa Scale (NOS) regarding three domains: selection, comparability and outcome assessment. 21 If an agreement could not be reached, a third-party arbitration was adopted to settle the dispute (I.Sz.).

| Statistical analysis
Statistical analysis was performed in the use of the Stata 11 SE (StataCorp). For dichotomous outcomes (graft uptake rate, canaloplasty rate, cosmetics and hearing outcomes, in accordance with the Committee on Hearing and Equilibrium of the American Academy of Otolaryngology, Head and Neck Surgery), we calculated pooled odds ratios (ORs) with 95% confidence intervals (CIs).
In some cases, we applied the Peto method, 20 due to the potential, occasional rare event. Weighted mean difference (WMD) with 95% CIs was calculated for continuous outcomes (operation time and ABG means).
Subgroup analyses were applied in accordance with the study design and included (RCT vs nRCT), if, at least three RCTs were available (graft uptake rate, canaloplasty rate, cosmetics and operation time). We applied the random effect model with DerSimonian-Laird estimation. I 2 and chi 2 tests were used to quantify statistical heterogeneity and gain probability-values, respectively; P < 0.1 indicated a significant heterogeneity. To check for publication bias, a visual inspection of funnel plots and Egger's test were performed. 20

Sensitivity analysis was performed by omitting studies (individually)
from the analyses and recalculating to investigate the impact of the individual studies upon the summary estimate.
The trial sequential analysis (TSA) was conducted regarding the graft uptake rate to observe the futility threshold. Notably, TSA is a methodology which combines an information size calculation (cumulated sample sizes of all included trials) for a meta-analysis with the threshold of statistical significance. In the operational use of this tool, we can quantify the statistical reliability of data in the cumulative meta-analysis adjusting significance levels for sparse data and repetitive testing on accumulating data. 22

| The GRADE approach
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) was used for assessing the quality of evidence of critical and important outcomes assessed. 23 The grade of evidence was assessed in support of subgroups of RCTs, if possible. Figure 1 represents the study flow chart. Our primary search yielded a total of 8700 records. In the screening of the records by title, abstract and full-text, we found 18 articles which met the eligibility criteria. 6,[9][10][11][12][13][14][15][16][17][18][24][25][26][27][28][29][30] We found two publications with identical cohorts of patients; therefore, we excluded the recently published one. 25 Another study was excluded due to insufficient data. 28 Inevitably, 16 articles were included in the meta-analysis. Characteristics of the included articles are shown in Table 1. The quality of evidence for each outcome is represented in Table 2.
Based on the Egger's test, the small study effect is unlikely to occur (P = 0.727) (Appendix S1). We performed a sensitivity analysis which demonstrated if any study is excluded from the analysis, the overall results are not affected nor altered. To overcome the potential risk of poor peer review caused by the inclusion of conference abstracts, we performed sensitivity analysis by omitting abstracts from the relating analysis which confirmed that the inclusion of these publications did not change the direction of the main association (Appendix S1).

| Canaloplasty rate: endoscopic type I tympanoplasty performs better compared with microscopic type I tympanoplasty (high grade of evidence)
Six studies analysed the canaloplasty rate, including a total of 594 interventions. 6,12,13,15,16,27 The rate of canaloplasty proved eight times higher in the use of the microscope when compared to that of the endoscope (15% vs 0%; OR = 7.96; 95% CI: 4.30-14.76; I 2 = 0.0%, P = 1.000) (Figure 3). In the performance of a sensitivity analysis, if and when any study was excluded from the analysis, the overall results are not affected nor altered.

| Cosmetic results: endoscopic type I tympanoplasty performs better than microscopic type I tympanoplasty (moderate grade of evidence)
Cosmetic results were reported in four articles including 279 interventions. 12,15-17 Microscopic type I tympanoplasty was reportedly 19 times more likely to result in a poor cosmetic outcome when compared to the endoscopic approach (58.3% vs 0%; OR = 19.29; 95% CI: 11.37-32.73; I 2 = 0.0%, P = 0.839) (Figure 4). We performed a sensitivity analysis, based on, if and when any study is excluded from the analysis, the overall results are not affected nor altered.
Undeniably, a serious inconsistency was evaluated in the event of high heterogeneity at the following two outcomes: Postoperative mean ABG (I 2 = 78.1%) and, the duration of the operation (I 2 = 98.0%).
The evaluation of indirectness yielded irrelevant results, due to the entirety of the included studies featured a clear PICO, which corresponded to our PICO.
Regarding the cosmetic results, imprecision was determined as relevant, since the assessment of cosmetic results was not performed using a standardised questionnaire. The optimal duration of operation was not precisely defined. In consideration of the balance of the outcomes, imprecision was not as serious amongst the outcomes based on RR.
Publication bias was not detected amongst critical outcomes. Regarding the remaining three important outcomes, publication bias could not be evaluated, due to the decreased number of included studies

| Postoperative pain, quality of life and perioperative complications
Three studies reported data on postoperative pain, 6,11,27 all used subjective methods for pain assessment. The studies did not detect a significant difference in the level of postoperative pain between the endoscopic and microscopic techniques.
Only one study investigated quality of life. 11 According to their findings, after endoscopic type I tympanoplasty, the patient`s quality of life improved significantly.
Since the discrepant reporting of the findings, these outcomes proved to be ineligible for meta-analysis.

| D ISCUSS I ON
Endoscopic ear surgery represents the ever-increasing, minimally invasive branch of otologic surgery. New results culminating in the past recent years warranted an updated meta-analysis to resolve debates concerning the two approaches. Our results indicate that the endoscopic approach competitively achieved graft uptake and hearing restoration equally and the microscopic method, yet it resulted in an increase in the preferred cosmetic results and a lower canaloplasty rate (ie less invasive) ( Table 2). These findings support the fact that the endoscope is an ideal tool in the management of chronic suppurative otitis media.
With regards to the primary outcome in association with the graft uptake, both approaches performed well. During endoscopic ear surgery, one-handed dissection is performed. Thus, one might assume that the thickness and manoeuverability of the materials used for the reconstruction of the tympanic membrane can F I G U R E 2 Represents the forest plot of our primary outcome, referred to as the graft uptake rate influence the primary outcome. Although technical differences across studies (eg reconstruction of the membrane with fascia, perichondrium or cartilage) elevated reasonable concern in reference to clinical heterogeneity (Table 3), it did not manifest itself in statistical heterogeneity (I 2 = 0.0%, see Figure 2). The potential selection bias is a persistent worry surrounding surgical studies, however, results on efficacy were also consistent in the subgroup of RCTs (Figure 2). Similarly, the number of interventions proved to be generally satisfactory towards excluding beta-type errors (Appendix S1).
Audiological outcomes reflect graft uptake. 31 Not surprisingly, the postoperative ABGs showed no difference amongst both groups. However, using ABG as a categorical variable, yielded results favouring endoscopy. Here, we believe the difference observed may be biased due to the potential differences in hearing at the baseline. Factually, this hypothesis in which we failed to detect any difference regarding the graft uptake rate, determines the recovery in hearing.
One of the primary advantages regarding endoscopic type I tympanoplasty is that there is no necessity for performing a canaloplasty. Non-invasiveness was proven both in the subgroup

| Strengths and limitations
Our meta-analysis possesses several strengths. A thorough systematic search and RoB assessment was performed using the Admittedly, amongst our strengths, our meta-analysis has several limitations. Two papers were conference abstracts, which potentially carry the risk, due to a less than ideal peer review process. us towards applying a high RoB (Appendix S1), if and when, a minimum period of six months was not achieved.

| Implications for research
Our meta-analysis provides a high level of evidence towards justifying the introduction of endoscopic type I tympanoplasty. Distinctly, based upon the TSA performed on graft uptake rates amongst the two methods (endoscopic vs microscopic), it can be stated, in which there is no need to perform additional, randomised controlled trials debating this outcome. However, the potential modifying effect of the location of tympanic membrane perforation and the learning curve in surgical practice should be further investigated. Due to the discrepant reporting of the cosmetic results, the importance of developing standardised questionnaires for quality of life evaluation should be emphasised.

| Implications for clinical practice
Our results imply how endoscopy, if and when accessible, is preferred when compared with microscopy throughout the routine practice. Hence, the implementation of endoscopic type I tympanoplasty regarding the treatment guidelines of chronic suppurative otitis media is strongly recommended.

ACK N OWLED G EM ENTS
We would like to thank Jon Marquette for English proofreading of our manuscript.

CO N FLI C T O F I NTE R E S T
The authors have no involvement in any financial or non-financial interest in the subject discussed in this manuscript.