xRead Articles - November 2022

Risk of Bias and Study Quality Assessment Methodological quality of included studies was assessed inde pendently by two separate authors (E.R. and A.D.V.). The quality of each study was assessed according to the Methodological Index for Non-randomized Studies (MINORS) quality assessment tool 30 or the Risk of Bias Assessment Tool for randomized controlled trials (RCTs) recommended by the Cochrane Collaboration, 31 as appropri ate. A funnel plot was created using the effect size for each of the fi ve intervention in order to examine a potential publication bias. Data Synthesis and Statistical Analysis The data from each study were transcribed in tabular form. Categorical variables were summarized by counts and percentage, while continuous variables were reported as median ± interquartile range (IQR: 25th and 75th). Given that only a few RCTs have been published on this topic, an arm-based network analysis was conducted to compare different surgical interventions. In particular, a meta-analysis of proportions was performed for both subjective and objective FS incidence. Net work meta-analysis is a statistical approach able to synthetize direct and indirect evidences about multiple treatment comparisons from independent studies. In particular, an arm-based approach describes population-averaged absolute effect sizes for each treatment arm, and it differs from the contrast-based method that focuses on relative effect size within each study. As a consequence, an arm-based analy sis allows for the inclusion of single-arm studies other than compara tive prospective or retrospective studies, enhancing the robustness of the meta-analysis in the setting of a lack of studies that directly com pare different treatment strategies. 32 A Bayesian hierarchical model based in Markov Chain Monte Carlo (MCMC) convergence simula tion was used. 33 In particular, the number of iterations for the adap tation process was set at 5.000, while the number of iterations in each chain was 100.000. A Bayesian framework incorporates a prior probability distribution based on external evidences, and a likelihood estimate based on the data reported by the included studies. The results of Bayesian inference are indeed probabilistic. The MCMC algorithm generate posterior samples that are used to estimate the parameters of interest (e.g., relative treatment effects and the corresponding credible intervals). Moreover, posterior samples can empirically rank treatments in terms of their probabilities to be the best treatment on the basis of the estimated effect size. Network meta-analysis introduce several criteria to deter mine if one intervention is better than another. The absolute effect size is calculated for each treatment arm in the arm-based approach, but a measure of relative effect size is also derived. In particular, we chose the odds ratio (OR) given that also retro spective studies have been included in the analysis. Moreover, the Bayesian approach allows for treatment ranking based on these data, using the abovementioned MCMC algorithm. 34 Pairwise associations between each treatment modality were depicted by a graphical representation of the network. Network esti mates from all outcome variables were presented as OR with 95% credible intervals (CIs). Rank probability plots (rankogram) were constructed to display the calculated treatment ranks. Posterior density plots were also used to determine the absolute risks of the various treatment alternatives and to identify potential overlapping. Cochrane ’ s Q method was used to assess between studies heterogeneity in each treatment group. I 2 was calculated as a measure of heterogeneity. An I 2 value represents the percentage of total variation across studies caused by heterogeneity rather than by chance. According to the Cochrane criteria, values from 0% to 40% may signify low heterogeneity, 30% to 60% may repre sent moderate heterogeneity, 50% to 90% may represent sub stantial heterogeneity, and 75% to 100% represent considerable heterogeneity. Scatter plots (Baujat plot) were created for each

The aim of this systematic review and meta-analysis was indeed to compare different treatments in the pre vention of FS after parotidectomy. In particular, we tried to de fi ne the best surgical technique able to reduce the subjective and objective FS incidence. MATERIALS AND METHODS The study was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 29 Institutional review board approval and informed consent were not required for this review of previously published studies. No review protocol was registered for this study. Eligibility Criteria This systematic review was conducted according to PICOS acronym: Patients (P), adults underwent partial or total parotidectomy; Intervention (I), temporoparietal fascia (TPFF), free fat graft (FFG), acellular dermal matrix (ADM), sternocleidomastoid muscle (SCM) fl ap, and super fi cial musculoaponeurotic system (SMAS) fl ap; Comparator (C), observation; Outcomes (O), FS inci dence (subjective and objective); Study design (S), both retrospective and prospective studies. Exclusion criteria were as follows: non-English language; insuf fi cient reported data; not extractable data; multiple tech niques; subgroup analysis of patients from a larger study; article type was either review, case report, conference abstract, letter to the editor, or book chapter. In addition, treatments, including less than 50 patients in both outcomes (subjective and objective FS incidence) were excluded to reduce the distortion of the pooled estimates. Data Source and Study Searching A comprehensive electronic search was performed on PubMed/MEDLINE, Cochrane Library, and Google Scholar data bases. Relevant keywords, phrases, and MeSH terms were adjusted to fi t the speci fi c requirements for each of the individual databases. The Medical Subject Heading keywords and phrases searched included: “ parotidectomy, ” “ parotid, ” “ Frey syndrome, ” “ reconstruction, ” and “ prevention. ” To minimize the risk of miss ing relevant data, a cross-reference search of the selected articles was performed, and the “ cited by ” function on Google Scholar was also used to obtain other relevant articles for the study. The last search was conducted on August 15, 2020. Data Collection Process Two independent reviewers (E.R. and A.D.V.) separately conducted the electronic search. The title and the abstract of all studies were reviewed to identify potentially relevant studies. The full-text of studies that appeared to be relevant to our search were downloaded and analyzed to determine eligibility. A third author (G.S.) resolved the con fl ict between reviewers. Data extraction from the included studies was systematically made using a structured form. In particular, one author (E.R.) indepen dently compiled a standardized form to extract the following characteristics of included studies: study design, number of patients, patient demographics, outcome de fi nitions, treatment protocols, and duration of follow-up. Then, another author (A.C.) checked for accuracy. Studies ’ general characteristics are shown in Supporting Table 1. The FS subjective and objective incidence rate was calculated for each study (Supporting Table 1) to be included in the network meta-analysis.

Laryngoscope 131: August 2021

De Virgilio et al.: Surgical prevention of Frey syndrome

1762