In head and neck cancer (HNC), 3-month post-treatment positron emission tomography (PET)/computed tomography (CT) reliably identifies persistent/recurrent disease. However, further PET/CT surveillance has unclear benefit. The impact of post-treatment PET/CT surveillance on outcomes is assessed at 12 and 24 months.
A 10-year retrospective analysis of HNC patients was carried out with long-term serial imaging. Imaging at 3 months included either PET/CT or magnetic resonance imaging, with all subsequent imaging comprised of PET/CT. PET/CT scans at 12 and 24 months were evaluated only if preceding interval scans were negative. Of 1114 identified patients, 284 had 3-month scans, 175 had 3- and 12-month scans, and 77 had 3-, 12-, and 24-month scans.
PET/CT detection rates in clinically occult patients were 9% (15 of 175) at 12 months, and 4% (3 of 77) at 24 months. No difference in outcomes was identified between PET/CT-detected and clinically detected recurrences, with similar 3-year disease-free survival (41% vs 46%, P = .91) and 3-year overall survival (60% vs 54%, P = .70) rates. Compared with 3-month PET/CT, 12-month PET/CT demonstrated fewer equivocal reads (26% vs 10%, P < .001). Of scans deemed equivocal, 6% (5 of 89) were ultimately found to be positive.
The treatment of head and neck cancer (HNC) remains challenging due to high mortality, as well as considerable difficulties in restoring cosmesis and function. After primary therapy it is important to accurately identify early recurrence, as salvage treatment is less successful once disease reaches advanced stages.1, 2 Distinguishing true disease recurrence from false positives would also spare patients the morbidity and expense of unnecessary further intervention.
The use of (18F)fluoro-deoxy-D-glucose (FDG) positron emission tomography (PET) has improved post-treatment surveillance by conveying the functional residual activity of treated sites, as well as identifying new sites of disease involvement. Multiple studies have shown PET to be a generally more sensitive diagnostic tool than computed tomography (CT), magnetic resonance imaging (MRI), ultrasound, or physical examination,3, 4 providing more refined TNM staging.5 Single-unit combined PET/CT scanners have made hybrid PET/CTs even more effective than PET alone; such fused scans reduce the number of equivocal interpretations through anatomic correlation, and have further improved diagnostic accuracy.6
With an excellent negative predictive rate of approximately 95%, a 3-month PET/CT appears to successfully distinguish persistent disease from inflammation, nonviable tumor, or treatment sequelae.7-9 In the context of a negative 3-month scan, however, it is unclear whether further PET/CT surveillance is useful or cost-effective. This question is especially relevant given the escalating use and high cost of PET/CT.10 Given that approximately 65% to 79% of recurrences occur within 1 year, and 89% to 95% of recurrences occur within 2 years,11, 12 a 3-month PET/CT would theoretically miss at least some later recurrences. With less tissue distortion and inflammation, subsequent PET/CTs may also be more diagnostically definitive, generating fewer equivocal reads and less ambiguity. Previous research has investigated PET/CT surveillance, basing recommendations on the detection of recurrence as the primary indicator.12-16 This study examines the ability of serial PET/CT to identify disease and improve survival relative to clinical examination, in a cohort of patients with negative 3-month imaging.
MATERIALS AND METHODS
All HNC patients who underwent PET/CT imaging over a 10-year period (between June 2002 and June 2012) at a single academic medical center were identified using the STRIDE (Stanford Translational Research Integrated Database Environment) tool, a standards-based informatics platform supporting clinical and translational research.17 Study protocols were approved by the Stanford University Institutional Review Board. TNM staging was per the American Joint Committee on Cancer, 7th edition staging guidelines.18 Patients were seen by a multidisciplinary team once every 1 to 3 months for at least the first 2 years.
Inclusion criteria were previously untreated head and neck squamous cell carcinoma, therapy intended for cure, and 3-month post-treatment imaging (PET/CT or MRI; Fig. 1). If the 3-month PET/CT or MRI was negative, the 12-month PET/CT result was recorded if performed. Likewise, if the 12-month PET/CT was negative, the 24-month PET/CT result was recorded if performed. PET/CT-detected recurrences were defined as clinically occult disease identified at the 12- or 24-month imaging time point. Clinically detected recurrences were defined as disease identified at routine clinic follow-up (between 3 and 12 months, or 12-24 months). To model realistic surveillance conditions, PET/CTs subsequent to any recurrence were excluded from analysis.
Table 1. Cohort Demographics
Abbreviations: CRT, chemoradiation therapy; RT, radiation therapy; SD, standard deviation.
Median ± SD
57 ± 13.1
Median ± SD
3 ± 1.6
Surgery + RT
Surgery + CRT
PET/CT interpretations were confirmed against the following gold standards: histopathologic verification, progressive disease on subsequent imaging, or at least 6 months of follow-up in the case of negative findings. Patients with stage IVC and previous HNC diagnoses were excluded. Other cancers excluded were esophageal malignancy, thyroid malignancy, and metastatic disease to the head and neck from other sites.
PET/CT scans were acquired on 1 of 3 scanners: GE Discovery D690, GE Discovery D600, or GE Discovery LS (GE Medical Systems, Milwaukee, Wis). The following protocol was used for the GE Discovery D690 and D600 scanners: 10-17 mCi of FDG was injected (based on weight), with a tracer uptake time of 60 minutes. Patients were scanned with 2 separate acquisitions: 1) first acquisition was of the base of skull to midthigh, with the patient lying with arms above their head, scanning for 7 to 8 bed positions, 2- to 5-minute acquisition per bed position; and 2) second acquisition was of the head and neck with the arms down, scanning for 2 to 3 bed positions, 2- to 5-minute acquisition per bed position. For the Discovery D690 and D600 scanners, slice thickness was 3.75 mm, pitch was 39.37, current was 85 mA, and energy was 140 kVp. For the Discovery LS scanner, slick thickness was 4.25 mm, pitch was 42.45, current was 100 mA, and energy was 120 kVp.
Results were based on a review of PET/CT reports. Imaging was originally interpreted by Division of Nuclear Medicine and Molecular Imaging faculty, each with at least 10 years of experience reading PET and PET/CT. Reports were dictated with access to patient history information and previous films. Image interpretation was based on visual and semiquantitative analysis, with no strict SUV cutoff used.19 A positive read was recorded when persistent, progressive, or new hypermetabolic areas were deemed suspicious and not otherwise explained. An equivocal read was recorded when malignancy could not be excluded. A negative read was recorded when no evidence of metabolically active disease was seen. Local, regional, and distant sites were assessed independently, with the interpretation as a composite of all sites. Any false-positive or false-negative finding at a site led to this classification for the entire read. Conversely, a true-negative finding required all 3 sites to be negative.
True positives, true negatives, false positives, and false negatives were determined for each PET/CT scan, and statistical measures (sensitivity, specificity, positive predictive value [PPV], negative predictive value [NPV], and accuracy) were calculated for all 3 intervals. Imaging intervals, median disease-free interval (DFI), median survival, disease-free survival (DFS), and overall survival (OS) were calculated from the end of treatment. To assess the survival benefits of PET/CT, DFI and DFS after recurrence were measured from the end of treatment of first recurrence until identification of the second. Patients treated palliatively were not included in DFI and DFS metrics, but were included in median survival and OS calculations.
Associations between variables and endpoints were assessed with Fisher exact test or Welch t test with unequal variance. Three-year DFS and OS were estimated using the Kaplan-Meier method and analyzed using the log-rank test. Events were defined as diagnosed second recurrence. All P values were 2-sided, with significance set at P < .05. Statistical analyses were performed using Prism 5 (GraphPad Software, La Jolla, Calif).
Of 1114 HNC patients identified with PET/CT imaging, 284 had 3-month post-treatment scans (257 PET/CTs and 27 MRIs). Of those with negative 3-month imaging, 175 had 12-month PET/CT scans. Of those with negative 3-month and negative 12-month imaging, 77 had 24-month PET/CT scans (Fig. 1). The majority of excluded patients had only staging PET/CTs, with no post-treatment scans.
Patient demographics and treatment details are listed in Table 1. Median age was 57 ± 13 years, with a median follow-up period of 3 ± 1.6 years. The oropharynx was the most common primary site (49%), followed by oral cavity (19%) and nasopharynx (18%). Advanced stage disease encompassed 85% of patients, and accordingly multimodality therapy was used in 88%. Recurrences were identified in 14% (40 of 284) of the study population. The 3-year DFS and 3-year OS were 86% and 89%, respectively.
PET/CT Diagnostic Performance
PET/CT characteristics by time interval are summarized in Table 2, whereas the efficacy of PET/CT reads as measured against actual outcome is summarized in Table 3. The respective rates of negative PET/CT reads at 3, 12, and 24 months were 65%, 78%, and 90%. Similarly, the respective rates of equivocal PET/CT reads at 3, 12, and 24 months were 26%, 10%, and 5% (Fig. 2). The proportion of equivocal PET/CT reads decreased over time significantly between the 3-month and 12-month intervals (P < .001), but not between the 12- and 24-month intervals (P = .33). Only 7% (5 of 67) of 3-month equivocal scans were later confirmed as positive, whereas none of the 12-month scans (0 of 18) or 24-month scans (0 of 4) was positive.
Performance values (sensitivity, specificity, NPV, PPV, accuracy) for 3-month, 12-month, and 24-month PET/CT are listed in Table 4. Upon comparison between 3-month and 12-month results, only PPV significantly improved (P = .02). No significant differences in diagnostic performance were noted between 12-month and 24-month results.
Table 4. Positron Emission Tomography/Computed Tomography Performance by Time Interval
There were 2 false negatives in which disease was missed by PET/CT, and 15 false positives (Table 3). Overall, 4% (9 of 257) of patients had clinically occult disease detected by 3-month PET/CT (Table 5). Locoregional recurrence occurred in 2% (5 of 257) of patients (neck ×2, oropharynx ×3), whereas distant metastases occurred in 2% (4 of 257; lung ×1, bone ×2, liver ×1).
All true recurrences were correctly read as positive by PET/CT, but there were 6 false positives (Table 3). Overall, 9% (15 of 175) of patients had clinically occult disease detected by 12-month PET/CT (Table 5). Locoregional recurrence occurred in 3% (5 of 175) of patients (neck ×3, oral cavity ×1, nasopharynx ×1), whereas distant metastases occurred in 5% (10 of 175; lung ×8, kidney ×1, retroperitoneum ×1).
Twenty-Four-Month PET/CT Findings
Similarly, all true recurrences were read as positive by PET/CT, but there was 1 false positive (Table 3). Overall, 4% (3 of 77) of patients had clinically occult disease identified by 24-month PET/CT (Table 5). All detected findings were distant metastases (lung ×1, kidney ×1, bone ×1).
Salvage Outcomes for PET/CT-Detected Recurrences (True Positives, No Clinical Suspicion)
All 5 patients with locoregional recurrences found by PET/CT underwent salvage with curative intent. After salvage, median DFI was 14 months, and median survival was 21 months. The 3-year estimated DFS and OS after recurrence were 40% and 60%, respectively (Table 6).
Table 6. Recurrences Detected by Long-Term (12- and 24-Month) PET/CT or Clinical Examination
No clinically detected distant metastasis recurrences were identified.
Total patients, No.
Initial advanced stage, %
Recurrent advanced stage, %
Salvage tx with curative intent, %
Median DFI after recurrence, mo
Median survival after recurrence, mo
3-year DFS after recurrence, %
3-year OS after recurrence, %
Of the patients with distant metastases found by PET/CT, 69% (9 of 13) underwent salvage with curative intent. After salvage, median DFI was 21 months, and median survival was 19 months. The 3-year estimated DFS and OS after recurrence were 39% and 69%, respectively.
Salvage Outcomes for Recurrences Detected by Clinical Examination
To better evaluate the impact of PET/CT-detected recurrences, outcomes of clinically detected recurrences that occurred during the same period were determined for comparison. Of those patients with negative 3-month imaging, 13 recurrences were clinically identified independent of PET/CT (Fig. 1). All were identified by physical examination, and none by laboratory studies or conventional imaging (CT, MRI). All were locoregional recurrences, with 6 detected in the 3- to 12-month post-treatment interval, and 7 detected in the 12- to 24-month interval. All had negative routine imaging before their recurrence, and 85% (11 of 13) underwent salvage with curative intent. Median DFI after recurrence was 27 months, and median survival after recurrence for the entire group was 27 months. The 3-year estimated DFS and OS after recurrence were 46% and 54%, respectively (Table 6).
Comparing PET/CT-detected recurrences (n = 18) to clinically detected recurrences (n = 13), no significant differences were identified in median DFI (P = .93) or median survival (P = .61). Similarly, no significant differences were found in 3-year DFS (P = .91) or 3-year OS (P = .47; Fig. 3). All outcomes were based on the time after recurrence.
Salvage of HNC recurrence detected at earlier stages has been reported to confer favorable survival benefit.2, 20, 21 Previous studies have established that a 3-month PET/CT after therapy is highly effective in detecting persistent disease,22, 23 and it is now a standard part of tumor surveillance at many institutions. The case for additional PET/CT surveillance in uncomplicated patients, however, remains uncertain. The purported advantages of subsequent imaging include improved accuracy from less inflammation, and improved survival outcomes from earlier intervention.
This study examined the impact of routine PET/CT scans at 12- and 24-month intervals in HNC patients treated for cure. In concordance with the longer time course, these scans had substantially fewer equivocal reads than at the 3-month interval, and significantly better PPV. Given the clinical dilemmas that equivocal scans often generate, such greater diagnostic certainty may be useful in reducing the need for unnecessary intervention. However, measures such as sensitivity, specificity, NPV, and accuracy were no better than 3-month PET/CT results, likely due to the marginal room for improvement, given already superb rates at 3 months. In addition, 94% of such reads were ultimately found to be negative in this cohort. Together, this argues against repeating PET/CT imaging simply to reduce the confounding effects of inflammation.
From a locoregional perspective, long-term PET/CT imaging in this 3-month–negatve PET/CT population did not identify many recurrences; the 12-month PET/CT detection rate was only 2%, and the 24-month PET/CT did not detect any locoregional recurrences at all (Table 5). Such small percentages likely reflect the effectiveness of negative baseline 3-month imaging, a known favorable prognostic factor.24
In terms of distant metastasis, routine PET/CT showed only a somewhat higher rate of recurrence. The 6% and 4% detection rates by respective 12-month and 24-month PET/CT imaging are in line with the 2% to 18% overall incidence reported in other non-PET/CT screening studies,25, 26 and lower than the 11% to 21% incidence seen in patients with high-risk factors such as advanced stage.27, 28 It might be expected that earlier detection by PET/CT would catch metastatic disease before it becomes widespread, and lead to better salvage results.1, 2 Yet outcomes of PET/CT-detected metastases in this study were not necessarily superior. Previous non-PET/CT reports examining surgical salvage of metastatic HNC demonstrate comparable numbers; 5-year DFS has been reported at 25%, 5-year OS at 59%, median DFI at 27 months, and median survival at 19 to 26 months.29-33 Spector et al found similar results in 170 HNC patients with distant metastasis, concluding that PET/CT relative to conventional imaging modalities did not improve life expectancy.34
In this study, PET/CT-detected recurrences appeared to fare no better than those detected clinically (Fig. 3). Such results do not necessarily confirm parity between PET/CT-detected and clinically detected outcomes, but only that significance could not be demonstrated. This may reflect an insufficient number of recurrences for detecting a difference, or a true lack of difference. It is important to note that lead-time bias, whereby earlier detection misleadingly inflates reported survival data, should have benefited the PET/CT-detected cohort. The finding that this bias did not facilitate superior PET/CT survival outcomes suggests that long-term surveillance may be of limited yield. Overall, not only were PET/CT detection rates in this study low, but identified recurrences appeared no more actionable than clinically detected recurrence in terms of survival. The prerequisite of negative 3-month imaging may play a role in this finding. It is also possible that the 12- and 24-month intervals are too lengthy for PET/CT to impact survival, but the low detection rates do not suggest this to be the case.
Few studies have examined PET/CT efficacy beyond 3 months, and none has considered the effect of systematic long-term surveillance on outcomes. Abgral et al and Lee et al recommended 12-month PET/CT based on detection rates of 33% and 20%, respectively.13, 14 However, patients did not have baseline negative 3-month imaging in either study, biasing the data toward more detected recurrences and not reflecting current clinical practice. Beswick et al recently reported a 28% overall recurrence rate using surveillance PET/CT at 2, 5, 8, and 14 months.12 PET/CT surveillance was recommended for 24 months, as 95% of detected recurrences occurred within this period. Results are difficult to compare given the varied imaging frequency; imaging was performed at the physician's discretion until late in the study, and the ensuing protocol did not include a 24-month scan. It is also unclear if the short-interval PET/CT-detected recurrences would have been otherwise identified via clinical means. The present study's detection rates of 9% at 12 months and 4% at 24 months suggest that routine PET/CT imaging at such brief intervals is less compelling. Moreover, the lack of survival data in all of these previously published reports makes it challenging to verify whether the imaging frequency and duration used were of clinical consequence.
Limitations of this study include its retrospective design and inclusion of 3-month MRI in the study cohort. Given the focus on long-term PET/CT imaging, and the high NPV of MRI together with clinical examination, the authors felt that inclusion of such patients for the purposes of investigating surveillance PET/CT was justified. The 3-month PPV of 35% was also somewhat low, although in line with the 19% to 58% rates reported by others.19, 35-37 In addition, the eligibility requirement of negative prior imaging with 12- or 24-month surveillance PET/CT was stringent, excluding >70% of patients receiving PET/CT scans. Under such conditions, this study's results apply to patients with straightforward treatment and history; higher-risk patients with previous recurrences or clinical suspicion may deserve confirmatory imaging or closer surveillance.38 Finally, it is clear that small recurrence numbers limit the reported survival data. A larger, appropriately powered trial following more recurrences would better delineate and extend this study's preliminary findings. Nonetheless, this cohort is a sizeable, accurate representation of patients with benign baseline imaging and long-term follow-up.
In HNC patients with negative 3-month imaging and low clinical suspicion, routine 12- and 24-month PET/CT scans appear to confer only minor benefit. In addition to low detection rates, earlier discovery by 12- and 24-month surveillance PET/CT was not found to improve salvage outcomes when compared with clinically identified recurrence. Higher-risk groups may still gain from serial PET/CT. Larger prospective trials are warranted to further investigate the survival benefits of long-term PET/CT surveillance.