2018 Section 6 - Laryngology, Voice Disorders, and Bronchoesophalogy

invasive cancers, but may also be caused by scarring, vocal overuse, or profound inflammation. Due to this, diagnostic difficulties of VS have been decribed. 5–7 How- ever, a normal mucosal wave seems to exclude extensive vocal ligament invasion. 7 In the case of obvious cancer (e.g., large tumor or vocal fold immobility), VS may not provide additional information. However, for smaller lesions, impaired amplitude of vibration and mucosal wave will raise the suspicion of malignant invasion into the deeper layers of the mucosa. The aim of this study was to estimate the diagnostic accuracy of VS in differentiating glottic cancer from non- invasive lesions by conducting a systematic review and meta-analysis of published studies. An explicit state- ment of the question being addressed is provided in the Methods section. METHODS A systematic review was performed according to the Pre- ferred Reporting Items for Systematic Reviews and Meta- Analysis statement. 8 The review protocol has been registered in the PROS- PERO International Prospective Register of Systematic Reviews, registration number CRD42015017597. Systematic Literature Search A PICOS (population, intervention, comparison, outcome, and study design) strategy 9 was adopted for the search. The target population consisted of patients with neoplastic lesions of the true vocal cords. VS was the intervention, whereas biopsy with histopathological evaluation was the comparator. The out- come measures were sensitivity, specificity, and positive and negative predictive values of the stroboscopic evaluation com- pared to biopsy. Finally, the study design was without restric- tions, as long as original data were provided. We searched the electronic bibliographic databases PubMed and Embase without restrictions on publication date. Languages were limited to English, Swedish, Norwegian, and Danish. The search was conducted on February 24, 2015. The search strategy was based on a combination of all available free text and Medical Subject Heading terms for VS, vocal cords, and malignancy or premalignancy. (See Supporting Information, Appendix, in the online version of this article for the complete search strategy). Study Selection The search strategy was developed with help from a scien- tific librarian. The electronic search was performed by one author ( C . S . M .), and studies identified were entered into Endnote version X7 (Thompson Reuters, Carlsbad, CA). Duplicates were removed manually. The eligibility process was conducted in two steps. Two authors ( C . S . M . and T . R .) screened titles and abstracts independently using predefined criteria for inclusion and exclu- sion. The same two authors independently evaluated the full- text version of publications that passed through the first step. Publications meeting all of the following inclusion criteria were included: patients with neoplasia on the true vocal cords evaluated with VS, original data provided, and histopathological evaluation performed and results provided Publications were excluded in cases of missing data for VS and/or histopathology; missing data for true positives (TPs),

false positives (FPs), true negatives (TNs), and false negatives (FNs); case reports; conference abstracts; or unacceptable qual- ity according to the Scottish Intercollegiate Guidelines Network (SIGN) recommendations. 10 At each step, discrepancies were solved by discussion, and if needed, a third evaluator ( A . M . G . or C . G .) acted as arbitrator. In case of any doubt, the article proceeded to the next level. In case of missing data, an attempt was made to obtain these by contact to the corresponding author of the study in question. Data Items and Collection Process Two authors ( C . S . M . and T . R .) performed data extraction, and any disagreements were solved by discussion. Data were extracted on patient selection criteria, number of patients, histopathological data, TP, FP, TN, and FN of stroboscopic evaluation. In case of inclusion of patient subgroups irrelevant to our study, those were extracted from the study population before statistical analysis. In one case, the corresponding author was contacted because of a misleading table, and data were confirmed. 11 Another attempt to clarify reported outcome of a study by contact to the author 12 failed, and we chose to perform calculations of accuracy solely based on the raw data provided in the published article. Risk of Bias Assessment Risk of bias in individual studies was assessed using the SIGN Checklist for Studies of Diagnostic Accuracy, 10 and qual- ity of the study was assessed according to these criteria. Publi- cations assessed to be of unacceptable quality according to SIGN recommendations were excluded. Synthesis of Results A diagnostic test accuracy meta-analysis was conducted for VS. For each study included, the TP, FP, TN, and FN were extracted and used for calculation of the derived sensitivity, specificity, positive predictive value (PPV), and negative predic- tive value (NPV). Those were the principal summary measures. Sensitivity was calculated as the proportion of individuals cor- rectly identified as having cancer and specificity as the propor- tion of patients correctly identified as not having vocal cord cancer. PPV was calculated as the proportion of positive test results determined as TPs and NPV calculated as the propor- tion of negative test results determined as TNs. Wilson scores were calculated for point estimates for mean and 95% confidence interval (CI) for sensitivities and specific- ities. 13 Egger’s test was used to assess the risk of publication bias. All statistical analyses were carried out in Stata 13.1 (Sta- taCorp, College Station, TX). Our systematic literature search resulted in 705 publications. The study selection process is schemati- cally illustrated in Figure 1. Fifty-five publications were analyzed as full-text, and the SIGN checklist was used for quality assessment. Five of these publications met the eligibility criteria and were included in the analysis. All five studies were also eligible for subsequent meta-analysis. Study characteris- tics and data from each of these studies are presented in Table I and Figures 2, 3, and 4. RESULTS Study Selection and Characteristics

Laryngoscope 126: September 2016

Mehlum et al.: VS and Prediction of Early Glottic Cancer

22