2017 Section 7 Green Book

Otolaryngology–Head and Neck Surgery 154(3)

treatment have consistently found that the accuracy of the scans varies with timing, with a longer time interval associated with a greater accuracy. 48-51 Our study confirms these findings, with scans for local or regional disease performed after 12 weeks having a greater specificity ( P = .009 and P = .004, respec- tively), but no difference in sensitivity, when compared with those performed before 12 weeks. These results are slightly dif- ferent from 2 previous meta-analyses that demonstrated an improvement in sensitivity, rather than specificity, with a delay of 10 to 12 weeks after the completion of treatment before scan- ning. However, both of these meta-analyses included retrospec- tive as well as prospective studies in their analysis. The lower specificity in the immediate posttreatment period found in our review is likely related to the increased vascularity, edema, and inflammatory changes at the pri- mary site and in the neck after (chemo)radiation, which results in an increased physiological uptake of FDG 52 and hence more false-positive readings. However, a deliberate delay before the first posttherapy scan is not without conse- quences; a prolonged period between the completion of che- motherapy and salvage surgery allows time for extensive postradiation fibrosis to develop, leading to an increased fre- quency and severity of surgical complications. 53 In deter- mining the optimal time for the initial scan, we must therefore balance the need for prompt diagnosis and man- agement of disease against the risk of misleading results if scans are performed too early. However, the timing of the first posttreatment scan remains somewhat controversial despite numerous diagnostic accuracy studies. Based on the results of our review, we would support a delay of 12 weeks after (chemo)radiotherapy before imaging because of the improvement in diagnostic accuracy seen with a later scan. Strengths and Limitations There are several strengths of this meta-analysis. We included only prospective studies in our review, thus reducing the number of articles included in our study compared with previ- ous meta-analyses. However, this inclusion may help reduce the risk of bias that may be found with retrospective studies. Because studies with positive results are more likely to be pub- lished, there is always the risk of publication bias with sys- tematic reviews. We attempted to minimize the potential for such bias by using a comprehensive search strategy with no language restrictions. Our exclusion of conference abstracts, letters, editorials, and gray literature may affect the results; however, we believe that this would have minimal impact overall. Publication bias was detected for the nodal sites only, and based on the large fail-safe number ( . 1000), we believe it is highly unlikely that these studies would have not been found using our comprehensive search strategy. The studies identified in our review had some limitations. Most notably, the reference standard was not consistent across all studies; histopathology was performed in every patient in only 4 of the 27 included studies. In most cases, histopathological confirmation was used only in patients with a positive PET or PET/CT because of the invasive nature of biopsies and neck dissections. Clinical follow-up,

with and without conventional imaging, formed the basis of the reference standard in those with negative PET scans. This may potentially result in the overestimation of test sen- sitivity and underestimation of test specificity. 54 There was also substantial variability in the sensitivity and specificity estimates among studies. Although the difference in imaging modality and timing explained this to some extent, some heterogeneity remained despite subgroup analysis. Other variables such as the stage and location of the tumor at initial diagnosis, the reference standard used, and the clinical presen- tation at recurrence may have contributed to the heterogeneity among studies. We could not assess the impact of these factors on test accuracy because of inconsistent reporting of data. Conclusion This is a meta-analysis focused on the diagnostic accuracy of PET and PET/CT for the detection of residual and/or recurrent local and regional disease and distant metastases in patients with HNSCCs using only prospective data. We found that both modalities had a good overall diagnostic accuracy for detection of residual and/or recurrent disease at local, nodal, and distant sites, with PET/CT being more specific than PET alone for the detection of disease at the primary site. The accuracy of visual assessment and semi- quantitative analysis of images were comparable at local, nodal, and distant sites. The timing of the scan had an impact on accuracy, with later scans being more specific than earlier scans. This study has determined that the most ideal strategy for follow-up scans is after 12 weeks post- treatment with the use of combined PET and CT. Author Contributions Phylannie K. F. Cheung , study concept and design, acquisition of data, analysis and interpretation of data, drafting of the manuscript, critical revision of the manuscript for important intellectual con- tent, and statistical analysis; Ronald Y. Chin , study concept and design, analysis and interpretation of data, drafting of the manu- script, critical revision of the manuscript for important intellectual content, and study supervision; Guy D. Eslick , study concept and design, analysis and interpretation of data, drafting of the manu- script, critical revision of the manuscript for important intellectual content, statistical analysis, and study supervision.

Disclosures Competing interests: None. Sponsorships: None. Funding source: None.

Supplemental Material Additional supporting information may be found at http://otojournal .org/supplemental. References 1. Curado MP, Hashibe M. Recent changes in the epidemiology of head and neck cancer. Curr Opin Oncol . 2009;21:194-200. 2. Marur S, Forastiere AA. Head and neck cancer: changing epi- demiology, diagnosis, and treatment. Mayo Clinic Proc . 2008; 83:489-501.

31