2017 Section 7 Green Book

Cheung et al

that reported exclusively on residual/recurrent disease at the primary site and 5 studies on residual/recurrent disease at the neck only. There were no prospective studies that evalu- ated the detection of distant metastases exclusively; all 3 studies included in our analysis for distant disease also reported on residual/recurrent disease at the primary site and in the neck. We found that FDG-PET and FDG-PET/CT had a high overall accuracy in detecting local, nodal, and distant resi- dual/recurrent disease after (chemo)radiotherapy in patients initially diagnosed with HNSCCs. The pooled sensitivity, specificity, PPV, and NPV for local and regional residual/ recurrent disease were similar to the findings of previous meta-analyses. Gupta et al 38 reported a sensitivity of 79.9% and 72.7%, specificity of 87.5% and 87.6%, PPV of 58.6% and 52.1%, and NPV of 95.1% and 94.5% for primate site and nodal recurrences, respectively. Similarly, a systematic review by Isles et al 39 that included data from dual-head gamma cameras reported a sensitivity of 94% and 74%, spe- cificity of 82% and 88%, PPV of 75% and 49%, and NPV of 95% and 96% for local and regional recurrent disease, respectively. For the detection of distant metastases in recurrent head and neck cancers, not specifically SCCs, Gao et al 40 reported a sensitivity of 92% and a specificity of 95%. Our results indicated that the overall diagnostic accuracy was slightly lower for the detection of residual and recurrent disease in the neck, when compared with detection at pri- mary, distant, or multiple sites considered together. The lower sensitivity of PET for nodal disease compared with other sites may be related to the spectrum of disease in the studies that reported on regional recurrence; nearly a third of these studies included only patients in whom nodal dis- ease was present at initial diagnosis. Micrometastases in the lymph nodes may not be detected by imaging, leading to a higher false-negative rate and lower sensitivity. The pooled NPVs for residual and recurrent disease at local, nodal, distant, and all sites combined were nevertheless quite high, suggesting that PET and PET/CT scans can reliably exclude residual/recurrent locoregional disease and distant metastases. A negative posttreatment scan can therefore guide the ongoing management of patients with HNSCCs and poten- tially reduce the need for more invasive diagnostic procedures. PET/CT has largely superseded the use of PET alone in clinical practice, and we wanted to explore whether this newer technology would have an impact on the test diagnos- tic accuracy. We found that there was a small benefit of PET/CT over PET alone for the detection of residual/recur- rent disease but only at the primary site ( P \ .001). The results of our subgroup analyses suggest that PET/CT has greater specificity, but no difference in sensitivity, when compared with PET alone for the detection of local recur- rences. We found no significant difference between the ima- ging modalities in terms of sensitivity or specificity in the detection of residual/recurrent nodal disease. The increased specificity with PET/CT for the identifica- tion of recurrent disease at the primary site may be related

to the improvement in anatomical localization possible with the co-registration of anatomical and functional information. Our results differ from the findings of a previous meta- analysis by Gupta et al, 38 which showed no difference between PET and PET/CT in terms of diagnostic perfor- mance. While there were no prospective studies directly comparing the use of PET/CT and PET in head and neck cancers, the few retrospective studies directly comparing the use of the 2 modalities generally reflect the findings of our study. Fakhry et al 41 compared the use of PET and PET/CT in 32 patients who presented with a suspicion of recurrent HNSCC. They found no difference in sensitivity (94% for both modalities) and a nonsignificantly higher specificity for PET/CT (57% vs 36%-50%), and they concluded that PET/CT was more accurate than PET alone. Likewise, a study by Ishitaka et al 42 involving 129 patients with sus- pected head and neck (including thyroid) cancer recurrence demonstrated no significant sensitivity benefit of PET/CT over PET (sensitivity 93.9% vs 91.4%, respectively) but a significant improvement in specificity when integrated PET/ CT is used (specificity 97.2% vs 74.4%). Similarly, when Chan et al 43 compared the use of the 2 modalities in 67 patients with papillomavirus-associated oropharyngeal SCCs, the findings showed that PET/CT had a better NPV compared with PET alone (98.2%-95% vs 95.7%-100%) for the detection of nodal recurrence. On the other hand, a study by Halpern et al 44 in patients with suspected local recurrence found that integrated PET/CT did not signifi- cantly improve the detection of recurrence compared with PET alone. FDG uptake by tissues can be assessed qualitatively using visual comparison of the abnormal and normal tissue or semiquantitatively through the calculation of standardized uptake values (SUVs). While results indicated that there was a trend toward a greater sensitivity or specificity with the use of semiquantitative methods for image assessment, the difference was not statistically significant at the primary site or in the neck, suggesting that either method can be used to interpret PET scans with a reliable degree of accuracy. This is consistent with a previous study 45 that suggested that the accuracy of visual interpretation by an experienced nuclear physician is comparable to SUV-based assessments. While the calculation of SUV may be viewed as a more objective index in assessing the uptake of FDG, it is never- theless affected by technical aspects such as the uptake time and the selection of the region of interest. 46 Moreover, despite 3 of the included studies 21,27,29 nominating a spe- cific cutoff value for diagnostic purposes, it has been argued that such thresholds are somewhat arbitrary because of the considerable overlap in SUVs between patients with benign and malignant lesions. 47 While many institutions report SUVs as part of their protocol, based on the results of our study, visual assessment alone is sufficient for characteriz- ing residual/recurrent HNSCCs. Retrospective studies directly comparing the diagnostic accuracy of PET and PET/CT at different time points after

30