2018 Section 6 - Laryngology, Voice Disorders, and Bronchoesophalogy

TABLE II. Risk of Bias Assessment Within Studies.

Flow and Timing

Overall Assess- ment of the Study

Patient Selection

Index Test

Reference Standard

Peretti et al. 18

Unclear

Low

Unclear

Unclear

Acceptable

Gugatschka et al. 11 El-Demerdash et al. 17

Unclear

Low Low

Unclear Unclear

Unclear Unclear

Acceptable

Low

High quality

Colden et al. 7 Caffier et al. 12

Unclear

Low

Low

Unclear

High quality

Low

Low

Unclear

Unclear

High quality

dure for assessing glottic lesions, but it should be supple- mented by further diagnostic procedures due to the inability to correctly identify patients without cancer. If normal mucosal waves are present, the lesion is almost certainly not malignant. On the contrary, a reduced or absent wave is not a specific finding and will lead to fur- ther diagnostic procedures and often surgery eventually. 7 The SIGN Checklist was used for quality assess- ment of all the included studies to avoid bias in the study result implied by diversity in the methodological aspects of selected studies. The overall quality of included studies was acceptable or high, even though the associated risk of bias was unclear in a significant proportion of the assessed domains. We aimed at identifying studies that tested the diagnostic accuracy of VS in differentiating cancer from noninvasive lesions. Some of the included studies sought to demonstrate the superiority of another diagnostic

This systematic review and meta-analysis quantita- tively evaluated the diagnostic accuracy of VS in differenti- ating cancer from noninvasive lesions on the vocal cords compared to biopsy. To address this exact question, we left out patients without stroboscopic rating 7 and added patients with carcinoma in situ to the patient group with noninvasive lesions 12 according to World Health Organiza- tion classification. For all of the included studies, the term noninvasive covered the entire histopathological range of lesions, not classified as malignant (invasive), from inflam- mation, hyperplasia, and keratosis to premalignant lesions (i.e., severe dysplasia and carcinoma in situ or LIN III). The raw data provided were used to calculate Wilson scores. This explains why our calculations differ somewhat from the reported numbers. All of the included studies reported a high sensitiv- ity of VS, whereas the specificity was reported with great variation. The studies by El-Demerdash et al., 17 Peretti et al., 18 and Caffier et al. 12 all reported a high specificity of VS. In contrast to this, the data provided from Gugatschka et al. 11 and Colden et al. 7 suggested very low specificity of VS. We did not discover any obvious differences among the studies (i.e., study popula- tion or study design that would explain this issue). The results of the meta-analysis showed that VS is generally able to identify 96% of patients with cancer, but only 65% were correctly identified as not having cancer. This has huge practical consequences for the clinical departments in terms of need for further diagnostic proce- dures and/or surgery, as well as psychological and practical implications for the patients. It highlights the need for fur- ther research concerning diagnostic accuracy of pre- and intraoperative assessment of patients with a glottic lesion. Superior diagnostic equipment will help optimize preopera- tive assessment of future patients to provide sufficient treatment in case of premalignant lesions or cancer and at the same time help avoid unnecessary operations and voice impairment. This is also relevant for healthcare pro- viders and policymakers to ensure a cost-effective treat- ment strategy for this group of patients. Several studies suggest an additional positive effect of combining other diagnostic modalities (i.e., autofluores- cence, 19 saline infusion, 18 brush cytology, 11 contact endos- copy, and other imaging techniques enhancing the vascularity 20–22 (i.e., narrow-band imaging) with VS to increase diagnostic accuracy. Results are promising but not consistent, and further research is needed. 12,23 VS probably remains the gold standard as a screening proce-

Fig. 5. Bubble plot of the meta-analysis of the sensitivity and specificity of videostroboscopy compared to histology. HSROC 5 hierarchical summary receiver operating characteristic. [Color figure can be viewed in the online issue, which is available at www.laryngoscope.com.]

Laryngoscope 126: September 2016

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

25