This retrospective study was approved under a waiver of informed consent by the institutional review board. All patients with biopsy-proven HNSCC between 2000 and 2013 who underwent surgical resection as primary treatment and had a PET/CT scan in our center were identified. Unknown primary cancers were excluded. A total of 98 patients who underwent ¹⁸F-FDG PET/CT within 6 months after completion of treatment were identified and were included in this study. The median time interval for performing PET/CT after completion of postsurgical chemoradiation was 3.4 months and after surgery was 2.9 months (for those who underwent surgical resection only).

PET/CT studies were performed according to institutional clinical head and neck protocol. All patients fasted for at least 6 hours before the PET/CT examination and had a blood glucose level lower than 200 mg/dL at the time of injection of FDG. An injection of an average of 5.55 MBq/kg (0.068 mCi/lb) of FDG was administered with an uptake time of approximately 60 minutes. Two PET/CT scanners (Discovery LS 2D and Discovery VCT 3D, GE Healthcare) were used. For body images, patients were scanned with arms up from midthigh to chin. The head and neck images were then acquired with arms down from carina to skull vertex. The acquisition time per bed position was 3 minutes, and the images were acquired in 128 × 128 matrix. Ordered-subsets expectation maximization (2D and 3D) was used to reconstruct all PET images. All PET data were reconstructed with and without CT-based attenuation correction. Helical CT images (120 kV; 20–200 mA; 8.0 noise index) were obtained with a 512 × 512 matrix. Beam collimation was 128 × 0.6 mm with a pitch of 0.8. Slice thickness was 3 mm, and the FOVs were 50 cm² (body) and 30 cm² (neck).

All FDG PET/CT images were interpreted by board-certified nuclear medicine physicians at the time of imaging in accordance with the routine clinical reporting. A nuclear medicine postdoctoral research fellow retrospectively read the scan reports and categorized them into three groups. When a report contained clear evidence of recurrence or metastasis, the scans were recorded as positive. When there was no evidence of recurrence or metastasis in a report, the scans were recorded as negative. Some scan reports had terminology such as “indeterminate” or “cannot exclude recurrence” in the impression, and these scans were categorized as indeterminate. For the purpose of analysis, indeterminate reports were grouped with the negative scan results because we found in previous studies [4, 16] that most indeterminate reports are true-negative for tumor.

The accuracy of the posttreatment PET/CT studies was established with either histopathologic confirmation of a suspected lesion (when biopsy was performed) or 6-month clinical follow-up findings from the date of the PET/CT study. Twenty-seven patients had biopsy confirmation (four image guided, 21 surgical excisional or endoscopic biopsies, two fine-needle aspiration of the neck mass without image guidance). Hospital electronic medical records and imaging records were carefully reviewed to establish the 6-month clinical follow-up findings. Change in management was recorded after each posttreatment PET/CT study. We also investigated the added value of PET/CT to clinical assessment. For each patient, the therapy-assessment PET/CT study was verified as to whether the requesting physician had previous clinical suspicion of residual disease at the time of the study. This was identified from the indication for the study as stated in the PET/CT reports and from careful review of the office and hospital visit and electronic medical records before the date of the PET/CT study. We further established the impact of PET/CT on the treatment strategy. The primary patient outcome measure evaluated in this study was overall survival. The survival status of patients was obtained from a public death registry (http://www.ancestry.com) and review of electronic medical records at our institution. The survival data for patients who were alive were censored at the last date of follow-up at our institution.

Mean ± SD was used to present the descriptive features of variables. If data were not in normal distribution, the median with 25th and 75th centiles (interquartile range [IQR]) was reported. We evaluated the accuracy of therapy-assessment PET/CT scans obtained within 6 months of completion of primary treatment. The classic 2 × 2 table was used to calculate the sensitivity, specificity, positive and negative predictive values, and accuracy of PET/CT. Chi-square analysis was performed to evaluate the impact of PET/CT findings on treatment. The added value for clinical assessment and impact on management of PET/CT were calculated as fractions. Overall survival was estimated with Kaplan-Meier curves. Hazard ratios were calculated with the Cox proportional hazards model. Multivariate regression analysis was performed with variables that were significant in univariate analyses. All statistical analyses were conducted with two-sided tests with statistical significance considered p < 0.05. All analyses were performed with SPSS software (version 15.0, IBM-SPSS).

The mean age ± SD of the 98 study patients was 58 ± 12 years. The demographic and tumor characteristics of patients included in the study are summarized in Table 1. Of the 98 patients, 31 (31.6%) were treated with surgery alone (group A), and 67 (68.4%) received adjuvant radiotherapy with or without chemotherapy after primary surgical resection (group B). The median time point of PET/CT after surgery (group A) was 2.9 months, and the median time of PET/CT after completion of postsurgical chemoradiation therapy (group B) was 3.4 months. The median follow-up period from the posttreatment study was 35 months (IQR, 14–53 months). Of the total 98 patients, 24 (24.5%) died, and 74 (75.5%) survived the follow-up period.

Of a total of 98 posttreatment scans, the reference standard was available for 96 scans, including biopsy confirmation (27 scans) and 6-month clinical follow-up (69 scans). Two scans with no information to confirm the results of PET/CT (neither clinical nor pathologic report) were excluded from the accuracy analysis. Most indeterminate FDG PET/CT reports are truly negative for tumor [4] in HNSCC, so we grouped the indeterminate reports along with the negative scan results for statistical analysis. Of the total 96 posttreatment scans with available confirmation data, 24 (25%) were positive and 72 (75%) were negative (including 30 indeterminate scans). Among the 24 positive scans, 16 were confirmed as true-positive by biopsy, or clinical follow-up, and eight were confirmed as false-positive by biopsy or clinical follow-up. Among the 72 negative scans, 68 were confirmed as true-negative by biopsy or clinical follow-up, and four scans (all four were originally interpreted as indeterminate scans) were confirmed as false-negative by biopsy. The sensitivity of posttreatment FDG PET/CT in assessment of therapy of patients with HNSCC treated primarily with surgical resection was 80.0%; specificity, 89.5%; positive predictive value, 66.7%; negative predictive value, 94.4%; and accuracy, 87.5%.

According to the reason for obtaining the scan cited by the treating physician, all scans were categorized into two types: with and without prior clinical suspicion of residual disease. A total of 22 (22.4%) scans were requested because of clinical suspicion of residual tumor. The other 76 (77.6%) scans were obtained without prior clinical suspicion of residual tumor. Fifteen of the 22 (68.2%) scans obtained with prior clinical suspicion were positive (13 true-positive, two false-positive), and seven (31.8%) were negative (five true-negative, two false-negative). The two false-positive scans were confirmed to be false-positive by biopsy. The two false-negative scans were read as indeterminate, and because of high clinical suspicion, biopsy was performed and confirmed residual disease. Among the scans obtained without prior clinical suspicion of residual disease, 13.2% (10/76) were positive (three true-positive, six false-positive, one with no confirmation data) and 86.8% (66/76) were negative (63 true-negative, two false-negative, one with no confirmation data) for residual disease. Among the six false-positive scans, biopsy confirmed no disease. The treating clinician believed FDG avidity was due to inflammation, and the FDG avidity resolved in follow-up imaging. The two false-negative scans were read as indeterminate, and because of high clinical suspicion, biopsy was performed and showed residual disease. Excluding false-positive and false-negative results, posttreatment PET/CT results were helpful in excluding tumor in 22.7% (5/22) of patients with clinical suspicion of residual disease and helped identify suspected disease in 3.9% (3/76) of patients with no prior clinical suspicion (Fig. 1A).

Among patients treated with surgical resection alone (group A), a total of 10/31 (32.2%) scans were requested with prior clinical suspicion of disease. The other 21 (67.8%) scans were obtained without prior clinical suspicion of residual tumor. Of the scans obtained with prior clinical suspicion, 7/10 (70%) were positive (all true-positive) and 3/10 (30%) scans were negative (two true-negative, one false-negative). Among the scans obtained without prior clinical suspicion, 19.0% (4/21) were positive (three true-positive, one false-positive) and 81.0% (17/21) were negative (16 true-negative, one false-negative) for residual disease. Therefore, excluding false-positive and false-negative results, posttreatment PET/CT results were helpful in excluding tumor in 20% (2/10) of patients with clinical suspicion of residual disease and helped identify suspected disease in 14.3% (3/21) of patients with no prior clinical suspicion (Fig. 1B).

Among patients treated with surgical resection plus adjuvant radiotherapy with or without concurrent chemotherapy (group B), a total of 12/67 (17.9%) scans were requested with prior clinical suspicion of disease. The other 55/67 (82.1%) scans were obtained without prior clinical suspicion of residual tumor. Of the scans that were obtained with prior clinical suspicion, 8/12 (66.7%) scans were positive (six true-positive, two false-positive), and 4/12 (33.3%) were negative (three true-negative, one false-negative). Among the scans obtained without prior clinical suspicion, 10.9% (6/55) were positive (five false-positive, one with no confirmation) and 89.1% (49/55) were negative (47 true-negative, one false-negative, one with no confirmation) for residual disease. Therefore, excluding false-positive and false-negative results, posttreatment PET/CT results were helpful in excluding tumor in 25% (3/12) of patients with clinical suspicion of residual disease (Fig. 1C). The median interval between completion of treatment and imaging was 2.5 months for patients with false-positive results.

After the PET/CT, no new treatment was initiated in 76 patients (77.6%). Of these 76 scans, 43 were read as negative, 26 as indeterminate, and seven as positive. Of the seven positive scans, five were confirmed as false-positive after histopathologic examination, and the clinicians believed that the FDG avidity on the other two scans was due to inflammation. Among 26 indeterminate studies, 22 were followed clinically with no treatment, and histopathologic examinations were performed after four studies. The results of 20 (20.4%) scans (17 positive, three indeterminate) prompted a change in treatment. On all three indeterminate scans, PET showed uptake, biopsies of these lesions showed recurrence, and new treatment was started. Finally, in two (2%) scans, the impact of PET/CT result on treatment was not known because there was no follow-up information. In summary, the start of new treatment was more likely after a positive PET/CT result. Observing a patient without treatment was more likely after the negative follow-up PET/CT results (chi-square p < 0.001) (Table 2).

Subgroup analysis showed that among those treated with surgical resection alone (group A, n = 31) no new treatment was started after follow-up PET/CT in 19 patients (13 negative, six indeterminate). The results of the other 12 scans (11 positive, one indeterminate) in this group led to the start of new treatment. The biopsy result was positive for disease recurrence after the indeterminate scan (Table 2).

Among those treated with surgical resection plus adjuvant radiotherapy with or without concurrent chemotherapy (group B, n = 67), no new treatment was started after follow-up PET/CT in 57 patients (30 negative, 20 indeterminate, seven positive). In the seven positive scans, biopsies ruled out disease recurrence after five scans, and in the cases of the other two scans, the treating physician suspected that the FDG avidity was due to inflammation. The results of eight scans (six positive, two indeterminate) in this group led to the start of new treatments (the two indeterminate scans were confirmed to be positive by biopsy). Finally, in the case of two scans, the impact of PET/CT result on treatment was not known because there was no follow-up information (Table 2).

Clinical factors such as age, sex, race, primary site (oral cavity, pharynx, larynx, and other sites), TNM category, overall stage, primary treatment (surgery, surgery plus radiotherapy, surgery plus chemotherapy, surgery plus chemoradiation therapy), clinical suspicion (scans with or without previous clinical suspicion of residual disease), and PET/CT result (positive versus negative) were included in the Cox regression analysis. In the multivariate regression analysis, T category (p = 0.002), grade (p = 0.041), clinical suspicion scan type (p = 0.002), and scan result (p = 0.005) were statistically significant and independent predictive factors of overall survival (Table 3).

Among the 98 patients with HNSCC, 25 (25.5%) had positive posttreatment scans (14 patients dead, 11 alive at last follow-up). The posttreatment scan results of the other 73 patients (74.5%) were negative (10 patients dead, 63 alive at last follow-up). The median survival time in the positive scan group was 10 months (IQR, 3–57 months), and the median survival in the negative scan group was 39 months (IQR, 20–52 months). Kaplan-Meier analysis based on the PET/CT scan results was performed, and the differences were statistically significant (log rank p < 0.001) (Fig. 2A).

Subgroup analyses were performed on data from patients treated with surgery alone (31 patients) or surgery followed by adjuvant treatment (67 patients). Among the 31 patients treated with surgery alone, 20 had negative postoperative PET/CT scans (18 patients who survived, two who died; median survival, 40.6 months; IQR, 23.6–64 months), and 11 had positive postoperative scans (five patients who survived, six who died; median survival, 11.6 months; IQR, 3–41 months). The difference was statistically significant (log rank p = 0.004) (Fig. 2B). Among the other 67 patients, who underwent postsurgical adjuvant radiotherapy with or without concurrent chemotherapy, 53 had negative posttreatment PET/CT scans (45 patients who survived, eight who died; median survival, 38.8 months; IQR, 18–52 months), and 14 had positive posttreatment scans (six patients who survived, eight who died; median survival, 6.9 months; IQR, 3.6–66 months). The difference was also statistically significant (log rank p = 0.001) (Fig. 2C).

Among the patients with early stage (stage I or II) HNSCC (20/98, 20.4%), two patients died (both had positive posttreatment scans; median follow-up, 12.6 months), and 18 patients were alive at last follow-up (14 patients had negative scans, four patients had positive scans; median follow-up, 42.7 months). There was a significant difference in survival between patients with positive and those with negative posttreatment PET/CT scans (log rank p = 0.03). Seventy-three (74.5%) of the patients presented with advanced-stage disease (stage III or IV). Among these patients, 21 died (12 patients had positive scans, nine patients had negative scans; median follow-up, 6.8 months) and 52 patients were alive at last follow-up (45 patients had negative scans, seven patients had positive scans; median follow-up, 40.6 months). The Kaplan-Meier analysis based on the PET/CT scan results showed a significant difference in survival (log rank p < 0.001) (Fig. 3).