2257

Detection of Recurrent Oral Squamous Cell Carcinoma

by [18F]-2-Fluorodeoxyglucose-Positron Emission
Tomography
Implications for Prognosis and Patient Management

Martin Kunkel, M.D., Ph.D.1 BACKGROUND. Patients with recurrent oral squamous cell carcinoma (OSCC) have
Gregor J. Förster, M.D.2 a dismal prognosis and represent a therapeutic challenge. A positron emission
Torsten E. Reichert, M.D., Ph.D.1 tomography (PET) scan with [18F]-2-fluorodeoxyglucose ([18F]FDG) can improve
Jong-Hyeon Jeong, Ph.D.3 early cancer detection. The current study evaluates the prognostic value of
Peter Benz, M.D.4 [18F]FDG-PET scan in patients with recurrent OSCC.
Peter Bartenstein, M.D., Ph.D.2 METHODS. The authors studied 97 patients with previously resected OSCC who
Wilfried Wagner, M.D., Ph.D.1 were restaged by PET scanning. Of the 97 patients, 64 had no evidence of clinical
Theresa L. Whiteside, Ph.D.5,6,7 disease and 33 were suspected of having disease by imaging, clinical findings, or
pathologic evaluation. The median follow-up period was 35.4 months after a PET
1
Department of Oral and Maxillofacial Surgery, scan. The end points included disease recurrence, a disease recurrence-free period
University Hospital Mainz, Mainz, Germany. 6 months after a PET scan, or death.
2
Department of Nuclear Medicine, University Hos- RESULTS. The overall sensitivity of a PET scan did not exceed 90% and its specificity
pital Mainz, Mainz, Germany. varied from 67% for local disease recurrence/second primaries to 99% for lymph
3 node metastasis. Increased [18F]FDG uptake predicted increased hazard of death
Department of Biostatistics, University of Pitts-
burgh, Pittsburgh, Pennsylvania. (hazard ratio: 6.83; P ⫽ 0.00034) and proved to be a highly predictive marker of
4
disease status. A significant association was established for incremental standard-
PET-Unit of the University Hospital Mainz, Mainz,
ized uptake values and 3-year patient survival (P⫽0.0089), indicating that intense
Germany.
glucose metabolism in the tumor is a negative marker of survival in recurrent
5
Department of Pathology, University of Pittsburgh OSCC. Overall, survival was longer in patients with a negative rather than a positive
School of Medicine, Pittsburgh, Pennsylvania.
PET scan (P ⬍ 0.00001).
6
Department of Immunology, University of Pitts- CONCLUSIONS. PET scanning was found to be highly valuable for diagnosing OSCC
burgh School of Medicine, Pittsburgh, Pennsylva- recurrence in a postoperative setting. It provided prognostic information and
nia.
played an important role in patient counseling and management. Cancer 2003;98:
7
Department of Otolaryngology, University of 2257–65. © 2003 American Cancer Society.
Pittsburgh School of Medicine, Pittsburgh, Penn-
sylvania.
KEYWORDS: oral cancer, disease recurrence, prognosis, positron emission tomog-
raphy.
Supported in part by a grant of the “Mainzer
Forschungsförderungsprogramm des Fachbe-
reichs Medizin” to M. Kunkel and T.E. Reichert
(MAIFOR: Glukosetransporter).
R ecurrence of oral carcinoma is always a life-threatening event,
leading to death in the majority of patients. Results of therapy in
patients presenting with recurrent oral carcinoma are modest. The
Address for reprints: Theresa L. Whiteside, Ph.D., 2-year survival rates seldom exceed 30%.1–3 Early detection of disease
University of Pittsburgh Cancer Institute, Research recurrence is critically important and there is a persistent clinical
Pavilion at the Hillman Cancer Center, Suite 1.27, need to improve restaging procedures to exclude or verify tumor
5117 Centre Avenue, Pittsburgh, PA 15213-1863; recurrence at stages that allow for a successful therapeutic interven-
Fax: (412) 624-0264; E-mail: whitesidetl@
tion.
msx.upmc.edu
Because increased glycolytic metabolism has been established as
Received May 13, 2003; revision received August a consistent characteristic of malignant cells, positron emission to-
4, 2003; accepted August 7, 2003. mography (PET) with [18F]-2-fluorodeoxyglucose ([18F]FDG) became

© 2003 American Cancer Society

DOI 10.1002/cncr.11763
2258 CANCER November 15, 2003 / Volume 98 / Number 10

an important component of the diagnostic regimens included in the study were considered to have an
and surveillance of various cancers.4 –10 [18F]FDG-PET increased risk of disease recurrence. Postoperative ra-
evaluates the intensity of glucose metabolism to dis- diotherapy was administered to patients with positive
criminate between malignant and nonmalignant tis- resection margins confirmed by pathology, extracap-
sues. [18F]FDG-PET can be used successfully to detect sular spread of involved lymph nodes, carcinomatous
recurrent oral squamous cell carcinomas (OSCC), be- lymphangiosis, as well as the majority of patients with
cause FDG uptake by the tumor exceeds that by the disease recurrence. From 1997 to 2002, all 97 patients
nonirradiated and irradiated soft tissues of the oral received an [18F]FDG-PET scan for restaging of sus-
cavity area.11,12 In the last few years, the superior pected recurrent disease. At the time of the [18F]FDG-
sensitivity and specificity of [18F]FDG-PET compared PET scan, 64 of 97 patients had no evidence of disease,
with other imaging modalities such as computed to- whereas recurrent disease was suspected in 33 of 97
mography (CT) or magnetic resonance imaging scans patients based on clinical findings, pathology results,
(MRI) have been reported.13–16 Recently, Wong et al.17 or morphologic imaging. The patients were scheduled
confirmed the successful diagnostic performance of to undergo [18F]FDG-PET beginning with the fourth
[18F]FDG-PET in a large series of patients with various month after radiation, typically 6 –9 months after tu-
malignancies of the head and neck. They demon- mor resection, before planned major secondary re-
strated the principal prognostic value of overall PET construction procedures, or when disease recurrence
interpretation and glucose uptake measured via the was suspected on the basis of clinical findings or mor-
standardized uptake value (SUV). However, it still re- phologic imaging. Informed consent was obtained
mains unconfirmed whether a comparable improve- from all patients for the diagnostic procedure and for
ment in the diagnostic accuracy can be associated therapy. Inclusion criteria were OSCC, potentially cur-
with a better prognosis in a largely homogeneous
ative radical resection of the primary tumor or the last
group of surgically treated patients with oral squa-
disease recurrence, high probability of disease recur-
mons cell carcinoma.
rence (advanced stages of primary tumors, patients
In the current study, the hypothesis was tested
already treated for disease recurrence, positive/doubt-
that [18F]FDG-PET provides clinically relevant prog-
ful margins by pathology) or recurrent disease already
nostic information for patients with oral carcinoma in
detected by another diagnostic method, an [18F]FDG-
the postoperative setting. Our results verify that in-
PET scan, and follow-up data.
creased posttreatment glucose metabolism is highly
All patients had follow-up to at least one of the
predictive of local and/or systemic disease recurrence
following three end points: disease recurrence either
in a largely homogeneous population of patients with
proven pathologically or by obvious clinical progres-
oral carcinoma.
sion, a minimum disease recurrence-free period of 6
MATERIALS AND METHODS months after the PET scan, or death. For survival
Treatment analysis, death was an end point. The median follow-
Ninety-seven patients comprosed the study popula- up period for surviving patients was 35.4 months after
tion. Their mean age was 55 years (range, 29 – 81 the [18F]FDG-PET scan and 46.8 months after tumor
years). The patients underwent potentially curative resection. At the time of survival analysis, 36 of 97
radical surgery or combined surgical and radiotherapy patients had died. Two patients died of causes unre-
for primary or recurrent OSCC. The tumors were re- lated to OSCC (i.e., intestinal bleeding, pulmonary
sected with a clinical safety margin of at least 10 mm. edema). Tumor localizations of the primary tumors
Surgical management of the lymph nodes followed the were the maxilla including the palate (n ⫽ 14), floor of
recommendations of the DÖSAK cooperative group the mouth (n ⫽ 31), tongue (n ⫽ 21), gingiva of the
(Deutsch Österreichisch Schweizerischer Arbeitskreis mandible (n ⫽ 10), retromolar trigone (n ⫽ 9), and all
für Tumoren im Kiefer- und Gesichtsbereich) based other locations (n ⫽ 12). The TNM staging categories
on the criteria of tumor localization and histologic were determined according to the criteria established
evaluation of frozen tissue sections obtained from in- by the American Joint Committee on Cancer (AJCC)
traoperative lymph node biopsies.18 According to and the International Union Against Cancer (UICC ).19
these recommendations, the N0 neck was treated by a Twenty patients had Stage I disease, 17 patients had
selective anterolateral neck dissection. When intraop- Stage II disease, 15 patients had Stage III disease, and
erative biopsy of suspicious lymph nodes revealed 45 patients had Stage IV disease. On histopathologic
metastases, a complete neck dissection was per- evaluation of lymph node involvement, 62 primary
formed. All patients in this series completed primary tumors were classified N0, 15 were N1, and 20 were
treatment as recommended. The patients who were N2.
 [18F]FDG-PET and Survival in Oral Ca/Kunkel et al. 2259

PET Scanning rection was not applied. The (local) sites of the pri-
PET scans were performed on a Siemens ECAT EXACT mary tumors, the cervical lymph node area, and dis-
922 (Siemens/CTI, Knoxville, TN) whole-body camera. tant sites were assessed independently. Any single
This device acquires 47 planes within an axial field of positive site was sufficient to consider the overall PET
view of 16.2 cm with a patient port of 56.2 cm in scan as positive.
dimension. The scanner achieves a transaxial spatial
resolution of 5.2 mm full width at half maximum Statistical Analysis
(FWHM) in the center of the field. The average axial The sensitivity and specificity of [18F]FDG-PET scans
resolution changes from 5.2 mm at the center to 8.1 were calculated based on pathologic or obvious clinical
mm FWHM at R ⫽ 20 cm. The patients fasted a min- evidence of any tumor progression identified in the en-
imum of 6 hours before a PET scan. During and at tire follow-up period for the local tumor site (tumor
least 30 minutes after administration of FDG, patients recurrence or second primary tumors), the cervical
reclined in a horizontal position. They were advised to lymph node area, and distant sites. Both neck sides were
keep at rest, to avoid speaking, and to minimize swal- considered separately, as therapeutic decisions followed
lowing for reduction of local, unspecific FDG uptake the identification of tumor involvement per neck side.
due to muscular activation. Survival probabilities were estimated by the
Approximately 45 minutes after intravenous injec- Kaplan–Meier method21 and statistically significant
tion of 370 MBq (⫾ 10%) of [18F]FDG, two-dimen- prognostic factors were identified using the propor-
sional emission measurements were obtained for tional hazards model.22 The end point of interest was
three to four bed positions (depending on the pa- time from PET scan to death (in months). The two
tients’ height) for imaging the viscerocranium, the patients who died from of other causes than tumor
neck, the thorax, and the epigastric region. For atten- were coded as censored observations. The log-rank
uation correction, transmission scanning of the same test23 was used to obtain P values for the test of sta-
region was performed thereafter using external 68[Ge] tistical significance for the comparison among groups
ring sources. The duration of emission and transmis- in case of univariate analysis. For multivariate analy-
sion scans was 10 minutes each. Emission data were sis, P values were obtained from the Wald tests using
reconstructed by filtered back projection using a Han- the Cox model to test for any statistical significance of
ning filter transmission correction (cut-off frequency regression coefficients. Confidence intervals (95% CIs)
of 0.5 bin⫺1). For (semi)quantitative evaluation, mean have a coverage probability of 95%.
SUVs were calculated for regions of interest (ROI) of 1 In the survival analysis, patients were categorized
cm in diameter drawn at the site of maximum according to PET scan results, indicating vital tumor
[18F]FDG uptake according to Zasadny and Wahl as (positive PET scan) or tumor control (negative PET
shown in Equation 1.20 scan). Multivariate survival analysis was performed
using the proportional hazards model with the follow-
SUV ⫽ [A(ROI)]/[V(ROI)] ⫻ [BM]/[A(total)] (1) ing dichotomized covariates as independent variables:
PET scan (negative vs. positive), UICC stage of the
where [A(ROI)] is the decay-corrected activity of the ROI primary tumor (Stage I and II vs. Stage III and IV), age
(MBq), [V(ROI)] is the volume of the ROI (mL), [BM] is (younger than median age vs. older than median age),
the body mass (kg), and [A(total)] is the applied activity irradiation (irradiated vs. nonirradiated), and type of
(MBq). disease (primary tumor vs. tumor recurrence).
[18F]FDG-PET images were interpreted primarily Quantitative data on glucose metabolism (SUV)
on the basis of a visual analysis by two experienced were available for 45 of 49 patients with progressive
senior nuclear medicine physicians, who were aware disease. For univariate survival analysis, these patients
of all clinical data, including the results of a physical were categorized according to the intensity of glucose
examination and morphologic imaging. Intense focal metabolism (SUV ⬍ 2; 2 聿 SUV ⬍ 4; 4 聿 SUV). The
or asymmetric FDG uptake not related to anatomic cut-off values were based on previous studies,10,24 in-
structures indicated tumor recurrence or metastasis. cluding our own, which showed that an SUV after
Visual interpretation was used to dichotomize PET radiotherapy (i.e., SUV 聿 4) almost always allowed for
scan results as either negative (no pathologic uptake) successful tumor resection, whereas an SUV greater
or positive (suspicious or distinct pathologic uptake). than 4 predicted local tumor recurrence in spite of
In lesions apparent by visual interpretation, the mean radical surgery in almost 80% of the patients.25 The
SUV was calculated based on an ROI placed within the SUV was also considered as a continuous variable in
borders of the lesion. Because no lesion was detected the proportional hazards model, including the covari-
with a dimension less than 1 cm, partial volume cor- ates UICC stage (Stage I and II vs. Stage III and IV), age
2260 CANCER November 15, 2003 / Volume 98 / Number 10

FIGURE 1. (A–C) An [18F]-2-fluorode-

oxyglucose-positron emission tomogra-
phy ([18F]FDG-PET) scan was used to
restage a second primary tumor in the
hypopharynx 10 years after surgery for a
patient with a retromolar carcinoma. The
PET scan revealed a synchronous
esophageal carcinoma and a third pul-
monary lesion.

TABLE 1
Performance Characteristics of [18F]-2-Fluorodeoxyglucose-Positron Emission Tomography

Positive Negative
Characteristics Sensitivity Specificity Prevalence predictive value predictive value

Local disease recurrence/second

primary tumors 0.87 0.67 0.32 0.55 0.92
Lymph node metastasesa 0.87 0.99 0.15 0.96 0.98
Distant metastases 0.71 0.93 0.18 0.67 0.94

a
Both neck sides were considered separately.

(younger than median age vs. older than median age), Ten lesions became positive by morphologic imaging
irradiation (irradiated vs. nonirradiated), type of dis- (large pulmonary masses, n ⫽ 5; cervical masses, n
ease (primary tumor vs. tumor recurrence), and sal- ⫽ 4; tumor at the base of the scull, n ⫽ 1). Except for
vage treatment approach (operative/curative vs. pal- 1 patient, all the patients (n ⫽ 12) with the above-
liative). described 18 lesions died. One patients was still alive
at the time of last follow-up but had developed pro-
RESULTS gressive disease at the tumor site. Overall, the pres-
Diagnostic Accuracy of [18F]-2-FDG PET Scans ence of these lesions, which were small and unappar-
In the follow-up period, 78 sites of tumor progression ent at the time of a PET scan, was confirmed by the
were recognized in 49 of 97 patients. [18F]FDG-PET unequivocal progress of disease observed either clin-
scans identified 65 of 78 tumor sites and 60 of 78 were
ically or by morphologic imaging in the follow-up
confirmed by pathologic evaluation. PET scans iden-
period.
tified local tumor recurrence or second primary tu-
The negative predictive values of [18F]FDG-PET
mors (27 of 31 sites), secondary lymph node metasta-
scans were 0.92 for the region of the primary tumor,
ses (26 of 30 sites), and distant metastasis or distant
0.98 for the neck area, and 0.94 for the sites of distant
metachronous cancers (12 of 17 sites). Twenty-six pa-
metastases. Table 1 summarizes the diagnostic results
tients had 1, 17 patients had 2, and 6 patients had 3
distinct sites of tumor progression. Figure 1 shows for the different regions and Table 2 shows the clinical
[18F]FDG-PET scan findings for a patient who simul- data for the 13 false-negative PET scan results for 11
taneously developed a second primary tumor in the patients. Three metachronous secondary primary tu-
hypopharynx, a third primary tumor in the esophagus, mors were diagnosed 7 months, 9 months, and 22
and a pulmonary metastasis (or a fourth primary tu- months after a [18F]FDG-PET scan. It is, thus, ques-
mor in the lung). Of 18 lesions not confirmed by tionable whether these tumors and their metastases
pathology, 8 became evident after clinical follow-up as already existed at the time of the [18F]FDG-PET scan.
exulceration of cervical metastases (n ⫽ 6), exulcer- In five patients, false-positive [18F]FDG uptake was
ation of a paratracheal mass (n ⫽ 1), and exulceration traced retrospectively to inflammatory lesions (e.g.,
of a large retromaxillary/infratemporal mass (n ⫽ 1). due to dentoalveolar surgery or osteoradionecrosis).
 [18F]FDG-PET and Survival in Oral Ca/Kunkel et al. 2261

TABLE 2
Clinical Variables of Patients Considered to be False Negative in the Study

Patient Method of tumor

no. Type of progress site detection Comment

1 Second primary (trachea) at the time of [18F]FDG-PET Endoscopy False-negative PET

2 Superficial second primary (⬍ 1 cm) and ipsilateral cervical lymph node Clinical investigation Existence of tumor at the time of [18F]FDG-PET
metastasis, detected 7 mos after [18F]FDG-PET is questionable
3 Superficial second primary (⬍ 1 cm) detected 22 mos after [18F]FDG-PET Clinical investigation Existence of tumor at the time of [18F]FDG-PET
is highly unlikely
4 Second primary at the time of [18F]FDG-PET and ipsilateral cervical lymph node Clinical investigation False-negative PET
metastasis detected 6 mos after [18F]FDG-PET
5 Secondary cervical lymph node metastasis detected 16 mos after [18F]FDG-PET Ultrasonography False-negative PET
6 Secondary lymph node metastasis at the time of [18F]FDG-PET Pathology (resection False-negative PET
specimen)
7 Second primary (lung) detected 9 mos after [18F]FDG-PET [18F]FDG-PET Existence of tumor at the time of [18F]FDG-PET
is questionable
8 Distant metastases detected 2 mos after [18F]FDG-PET and after palliative [18F]FDG-PET Time of tumor generalization is questionable
radiation of a large cervical tumor mass
9 Distant metastasis detected 7 mos after [18F]FDG-PET [18F]FDG-PET False-negative PET
10 Distant metastasis detected 4 mos after [18F]FDG-PET [18F]FDG-PET False-negative PET
11 Distant metastasis at the time of [18F]FDG-PET CT False-negative PET

[18F]FDG-PET: [18F]fluorodeoxyglucose-positron emission tomography scan; CT: computed tomography scan.

Association between Overall [18F]FDG PET Results and

Patient Survival
In the follow-up period, 4 of 39 patients in the PET
scan-negative group and 30 of 58 patients in the PET
scan-positive group died of their disease. The 3-year
survival rates according to Kaplan–Meier were 86% in
the PET scan-negative group and 44% in the PET
scan-positive group. The difference in survival was
highly significant (P ⬍ 0.00005, log-rank test). The
survival curves in Figure 2 demonstrate that increased
[18F]FDG uptake predicts an increased hazard of death
in patients investigated for recurrent disease. How-
ever, if pathologic glucose metabolism was once ruled
out via an [18F]FDG-PET scan, death due to any type of
tumor progression was highly unlikely in the near FIGURE 2. Kaplan–Meier survival estimation (disease-specific survival) ac-
future. Only two patients died of disease within 2 years cording to overall [18F]-2-fluorodeoxyglucose-positron emission tomograpy
after the [18F]FDG-PET suggested a complete tumor (PET) scan results. Glucose metabolism has a significant predictive value for
eradication. When controlled for tumor stage, age, patient’s survival. A physiologic pattern of [18F]FDG uptake strongly indicates
radiotherapy, and former events of tumor recurrence, tumor control (P ⬍ 0.00005, log-rank test). Hazard ratio ⫽ 7.4 (P ⫽ 0.00001).
an [18F]FDG-PET scan status was confirmed to be a Black line, PET overall negative (n ⫽ 39); dashed line, PET overall positive (n
significant independent prognostic marker in a pro- ⫽ 58).
portional hazards regression analysis. Pathologic glu-
cose uptake was assigned a hazard ratio of 6.47 (P
⫽ 0.00052, Wald test). Hazard ratios, 95% CIs, and P highly predictive marker of the disease status in pa-
values of the Wald test for all variables are given in tients with OSCC. Normalized glucose uptake strongly
Table 3. When covariates with P greater than 0.1 were indicated tumor control and vice versa.
eliminated, 2 factors of significant predictive value
remained: pathologic glucose uptake (hazard ratio, Association between Standardized Uptake Value and
6.83; P ⫽ 0.00034) and former events of tumor recur- Patient Survival
rence (hazard ratio, 2.09; P ⫽ 0.0385). The data indi- When patients with progressive disease were catego-
cated that after therapy, glucose metabolism was a rized according to the intensity of glucose metabolism
2262 CANCER November 15, 2003 / Volume 98 / Number 10

TABLE 3
Cox Regression Analysis of 97 Patients Investigated by [18F]Fluorodeoxyglucose-Positron Emission Tomography: Glucose
Metabolism Controlled for Other Important Variables in Oral Squamous Cell Carcinomas

Hazard ratio (group 2 95% confidence P

Variable Definition of groups (Group 1 vs. Group 2) relative to group 1) interval value (Wald)

Glucose metabolism Normalized vs. increased 6.47 2.25–18.56 0.00052

Tumor stage (primary tumor) Stages I ⫹ II vs. III ⫹ IV 0.98 0.46–2.1 0.96
Age Younger than median vs. older than median 0.65 0.31–1.33 0.24
Radiotherapy No previous irradiation vs. previous irradiation 0.56 0.26–1.24 0.15
Former disease recurrence No previous disease recurrence vs. previous disease recurrence 2.67 1.22–5.85 0.0144

DISCUSSION
Primary treatment of advanced stages of oral squa-
mous cell carcinoma (OSCC) most commonly utilizes
combined protocols of radical tumor ablation, radia-
tion, and/or chemotherapy. The local effects of these
treatment protocols often interfere with accurate clin-
ical or radiological detection of local recurrent and
secondary disease.26 With regard to oral carcinoma,
posttreatment distortions of the tissues are further
aggravated by the required reconstruction procedures
involving bone plates, bone grafts, and a large variety
of complex soft tissue flaps. In consideration of these
region-specific implications, our study was confined
to a homogeneous population of patients with OSCC.
FIGURE 3. Kaplan–Meier survival estimation (disease-specific survival) ac- A high degree of accuracy has been reported for
cording to quantitative glucose uptake. The curves demonstrate that in the [18F]FDG-PET in the diagnosis of SCCHN at various
postoperative setting, a high standardized uptake value (SUV) was significantly sites, including the larynx, pharynx, and the sinus-
associated with poor survival. Black line, SUV ⫽ 0.00 –2.00 (n ⫽ 8); line with
es.17,27,28 However, these studies only occasionally in-
small dashes, SUV ⫽ 2.01– 4.00 (n ⫽ 16); line with small/large dashes, SUV
cluded patients with oral carcinoma. Moreover, the
⬎ 4.01 (n ⫽ 21).
predominant therapeutic modality in all of these se-
ries was irradiation. Therefore, our study was con-
(Fig. 3), the outcome was 3 of 8 deaths in the SUV ⱕ 2 ducted to determine whether diagnostic results of
group, 11 of 16 deaths in the 2 less than SUV ⱕ 4 similar quality could be obtained postsurgery for pa-
group, and 17 of 21 deaths in the SUV greater than 4 tients with OSCC. Although 52 of 97 patients had un-
group. The 3-year survival rates according to Kaplan– dergone at least one major flap reconstruction, which
Meier were 60%, 24%, and 12%, respectively (P typically may camouflage regrowth of the tumor,
⫽ 0.0089, log-rank test). A univariate Cox analysis [18F]FDG-PET was highly sensitive in revealing tumor
further confirmed the impact of SUV on survival (P recurrence. Negative PET was a strong prognostic de-
⫽ 0.00085; hazard ratio ⫽ 1.316 with a 95% CI of terminant of a favorable clinical outcome. However,
1.12–1.55). When the data were adjusted for tumor compared with other series,17,29 the local and overall
stage, age, radiotherapy, former events of tumor re- sensitivity of our PET scans were slightly lower and did
currence and operability, SUV (considered as a con- not exceed 90%. It is likely that this difference resulted
tinuous variable in the Cox regression model) was from our definition of the true disease status that
assigned a hazard ratio of 1.38 per increase in 1 unit of included any tumor site detected throughout the en-
SUV (P ⫽ 0.00073, Wald test), demonstrating an incre- tire follow-up period. Metachronous second primary
mental prognostic value of glucose uptake. The hazard tumors (Patients 2, 3, and 7 in Table 2) and their
ratios, 95% CIs, and P values of the Wald test for all metastases added significantly to the reduced level of
variables are summarized in Table 4. Overall, these sensitivity observed in the current study. Applying the
data indicate that intense glucose metabolism is a criteria of Wong et al.17 (disease status defined by
negative marker of prognosis in patients with recur- lesions detected within 6 months after PET), the sen-
rent OSCC. sitivity of our series would have been 93% for local
 [18F]FDG-PET and Survival in Oral Ca/Kunkel et al. 2263

TABLE 4
Cox Regression Analysis in 45 patients with Progressive Disease: Standardized Uptake Value Controlled
for Other Important Variables in Oral Squamous Cell Carcinomas

Hazard ratio (Group 2 95% confidence P

Variable Definition of groups (Group 1 vs. Group 2) relative to Group 1) interval value (Wald)

SUV Continuous variable 1.38a 1.15–1.67 0.00073

Tumor stage (primary tumor) Stages I ⫹ II vs. III ⫹ IV 0.88 0.39–1.96 0.75
Age Younger than median vs. older than median 0.98 0.47–2.0 0.95
Radiotherapy No previous irradiation vs. previous irradiation 0.81 0.3–2.2 0.67
Former disease recurrence No previous disease recurrence vs. previous disease recurrence 1.57 0.63–3.87 0.33
Operability of recurrence Surgery vs. no surgery 3.57 1.62–7.89 0.00167

SUV: standardized uptake value.

a
Hazard ratio per increase in 1 SUV.

tumor recurrence, 93% for secondary lymph node in- OSCC, the definition of a low-risk status by
volvement, and 80% for distant metastases. Con- [18F]FDG-PET has a substantial impact on patient
versely, late singular cervical and mediastinal lymph management. Rehabilitation of patients with oral
node metastases in patients without secondary pri- carcinoma frequently requires multiple surgical in-
mary tumors (Patients 5 and 9) clearly demonstrate terventions for secondary reconstruction of the
that vital tumor tissue was obviously present at the jaws, improvement of mastication, speech and swal-
time when the [18F]FDG-PET scan was performed, but lowing, and, finally, prosthodontics, which typically
was not recognized until 16 months later. As the pres- involves dental implants for anchoring. The costs
ence of such “dormant” tumor cells cannot be de- and effort of these reconstructive procedures often
tected by any diagnostic method, it appeared justified exceed the costs of the primary treatment.30 Conse-
to us to define the disease status by every occurrence quently, these procedures are often postponed for
of tumor progression in the entire follow-up period. up to 2 years to ensure lasting tumor control. The
Currently, the major clinical limitation of results of the current study suggest that a negative
[18F]FDG-PET is its lack of specificity. Only a few false- PET scan justifies an early onset of reconstruction
positive FDG accumulations (i.e., 5 of 29) were related measures. If, however, glucose uptake is high, major
to distinct pathologic lesions such as local soft tissue reconstruction steps should be delayed until further
inflammation, residuals of dentoalveolar surgery, or evaluation.
osteoradionecrosis. In all other cases, the true causes How do we manage the patient who is undergoing
of increased glucose uptake remained obscure. For screening and whose PET scan becomes positive? Our
example, Figure 4 demonstrates intense focal FDG course of action is to follow a series of steps, which
uptake (SUV ⫽ 3.2), which is highly suspicious for might include complementary morphologic imaging
local tumor recurrence. Nevertheless, in this patient, (MRI/CT scans), direct biopsy of all suspicious lesions
the follow-up period was uneventful for 3 years up to when a morphologic correlative is identified, repeat
the present time. It is, of course, possible that the imaging after 2–3 months when no morphologic cor-
tumor will recur at a later date, as [18F]FDG-PET is relative is found, a second PET scan if still no mor-
capable of identifying small lesions invisible to mor- phologic correlative is found, biopsies by PET scan
phologic imaging.29 Still, a diagnostic dilemma is to guidance (“blind” with respect to morphological im-
discriminate between false-positive glucose metabo- aging) if FDG uptake increases, and “watchful waiting”
lism and a true pathologic FDG uptake due to lesions if FDG uptake decreases. Two recently published stud-
not detected by other imaging modalities. We wit- ies on primary tumors suggested that high FDG up-
nessed nine PET-positive sites of true tumor recur- take may be a useful parameter for identifying partic-
rence initially missed by a CT scan. In 2 patients, ularly aggressive tumors.31,32 Our results are
morphologic evidence of disease was not obtained consistent with these findings and further extend the
until 6 months later. Thus, a positive PET scan without prognostic relevance of glucose metabolism to predict
any morphologic correlate requires a follow up and prognosis in patients with recurrent OSCC. These
repeated morphologic and/or functional imaging. findings strongly support the hypothesis that the gly-
Despite the above-mentioned false-negative colytic tumor phenotype is of critical importance in
and false-positive results, [18F]FDG-PET was highly oral carcinoma progression. Because a 3-year survival
predictive of the clinical outcome. In patients with rate of only 12% was achieved when the SUV of the
2264 CANCER November 15, 2003 / Volume 98 / Number 10

FIGURE 4. A false-positive [18F]-2-

fluorodeoxyglucose-positron emission
tomography ([18F]FDG-PET) scan. The
reconstructed images show intense
(standardized uptake value [SUV] ⫽ 3.2)
focal FDG uptake, which is highly sus-
picious for local tumor recurrence. Be-
cause a computed tomography scan and
biopsy did not reveal the presence of a
tumor, a second [18F]FDG-PET scan was
performed, and it confirmed the site and
intensity (SUV ⫽ 3.6) of pathologic glu-
cose metabolism. Neither a second bi-
opsy nor repeated imaging ascertained
tumor regrowth. Ten months later, the
pattern of FDG uptake had completely
resolved (SUV ⫽ 0.68) without any ther-
apeutic intervention.

tumor recurrence exceeded 4, radical salvage surgery REFERENCES

should be considered with caution for this subset of
1. Lacy PD, Spitznagel EL, Piccirillo JF. Development of a new
patients. staging system for recurrent oral cavity and oropharyngel
The data from the current study, based on squamous cell carcinoma. Cancer. 1999;86:1387–1395.
[18F]FDG-PET results obtained in a homogeneous co- 2. Schwartz GJ, Metha RH, Wenig BL, et al. Salvage treatment
hort of 97 patients with oral carcinoma, provided for recurrent squamous cell carcinoma of the oral cavity.
overwhelming evidence that [18F]FDG-PET is highly Head Neck. 2000;22:34 – 41.
3. Sun L, Leung SW, Su C, et al. The relapse pattern and
valuable for the postoperative diagnosis of these pa-
outcome of postoperative recurrent tongue cancer. J Oral
tients. In addition to providing crucial prognostic in-
Maxillofac Surg. 1997;55:827– 831.
formation, [18F]FDG-PET plays a key role in patient 4. Beggs AD, Hain SF, Curran KM, et al. FDG-PET as a “met-
counseling and management, especially for the timing abolic biopsy” tool in non-lung lesions with indeterminate
of reconstruction procedures. biopsy. Eur J Nucl Med. 2002;29:542–546.
 [18F]FDG-PET and Survival in Oral Ca/Kunkel et al. 2265

5. De Potter T, Flamen P, van Cutsem E, et al. Whole-body PET 18. Bier J. Definitionen zum radikalchirurgischenVorgehen bei
with FDG for the diagnosis of recurrent gastric cancer. Eur Plattenepithelkarzinomen der Mundhöhle [Definitions of
J Nucl Med. 2002;29:525–529. radical surgical resection of oral squamous cell carcinoma].
6. Jerusalem G, Beguin Y, Fassotte MF, et al. Whole-body Dtsch Z Mund Kiefer Gesichts Chir. 1981;6:369 –372.
positron emission tomography using 18F-fluorodeoxyglu- 19. Wittekind C, Wagner G. TNM-Klassifikation maligner Tu-
cose for posttreatment evaluation in Hodgkin’s disease and moren [TNM classification of malignant tumors], 5th edi-
non-Hodgkin’s lymphoma has higher diagnostic and prog- tion. Berlin: Springer, 1997.
nostic value than classical computed tomography scan im- 20. Zasadny KR, Wahl RL. Standardized uptake values of normal
aging. Blood. 1999;94:429 – 433. tissues at PET with 2-[fluorine-18]-fluoro-2-deoxy-D-glu-
7. Mac Manus MP, Hicks RJ, Ball DL, et al. [18F] fluorodeoxy- cose: variations with body weight and a method for correc-
glucose positron emission tomography staging in radical tion. Radiology. 1993;189:847– 850.
radiotherapy candidates with nonsmall cell lung carcinoma. 21. Kaplan EL, Meier P. Nonparametric estimation from incom-
Cancer. 2002;92:886 – 895. plete observations. J Am Stat Assoc. 1958;53:457– 481.
8. Rose C, Dose J, Avril N. Positron emission tomography for 22. Cox DR. Regression models and life-tables. J R Stat Soc [B].
the diagnosis of breast cancer. Nucl Med Commun. 2002;23: 1972;30:248 –275.
613– 618. 23. Peto R, Peto J. Asymptotically efficient rank invariant test
9. Spaepen K, Stroobants S, Dupont P, et al. Can positron procedures. J R Stat Soc [A]. 1972;135:185–206.
emission tomography with [18F]-fluorodeoxyglucose after 24. Kitagawa Y, Sadato N, Azuma H, et al. FDG PET to evaluate
first line treatment distinguish Hodgkin⬘s disease patients combined intra-arterial chemotherapy and radiotherapy of
who need additional therapy from others in whom addi- head and neck neoplasms. J Nucl Med. 1999;40:1132–1137.
tional therapy would mean avoidable toxicity? Br J Haema- 25. Kunkel M., Förster GJ, Reichert TE, et al. Radiation response
tol. 2001;115:272–278. non-invasively imaged by [18F]FDG-PET predicts local tu-
10. Strauss LG, Conti PS. The applications of PET in clinical mor control and survival in advanced oral squamous cell
oncology. J Nucl Med. 1991;32:623– 648. carcinoma. Oral Oncol. 2003;39:170 –177.
11. Rege SD, Chaiken L, Hoh CK, et al. Change induced by 26. Lell M, Baum U, Greess H, et al. Head and neck tumors:
radiation therapy in FDG uptake in normal and malignant imaging recurrent tumor and post-therapeutic changes with
structures of the head and neck: quantitation with PET. CT and MRI. Eur J Radiol. 2000;33:239 –247.
Radiology. 1993;189:807– 812. 27. Fischbein NJ, Aassar, OS, Caputo GR, et al. Clinical utility of
12. Kunkel M, Reichert TE, Benz P, et al. Overexpression of positron emission tomography with 18F-fluorodeoxyglu-
Glut-1 and increased glucose metabolism in the tumor are cose in detecting residual/recurrent squamous cell carci-
associated with poor prognosis in oral squamous cell carci- noma of the head and neck [see comments]. AJNR Am J
noma. Cancer. 2003;97:1015–1024. Neuroradiol. 1998;19:1189 –1196.
13. Keyes JW Jr., Watson NE Jr., Williams DW 3rd, et al. FDG PET 28. McGuirt WF, Greven K, Williams D 3rd, et al. PET scanning
in head and neck cancer. AJR Am J Roentgenol. 1997;169: in head and neck oncology: a review. Head Neck. 1998;20:
1663–1669. 208 –215.
14. Kresnik E, Mikosch P, Gallowitsch HJ, et al. Evaluation of 29. Terhaard CH, Bongers V, Rijk PP, et al. F-18-fluoro-deoxy-
head and neck cancer with 18F-FDG PET: a comparison glucose positron-emission tomography scanning in detec-
with conventional methods. Eur J Nucl Med 2001;28:816 – tion of local recurrence after radiotherapy for laryngeal/
821. pharyngeal cancer. Head Neck. 2001;23:933–941.
15. Li P, Zhuang H, Mozley PD, et al. Evaluation of recurrent 30. Angthoven M, v. Ineveld BM, de Boer MF, et al. The cost of
squamous cell cacinoma of the head and neck with FDG head and neck oncology: primary tumors, recurrent tumors
positron emission tomography. Clin Nucl Med. 2001;26:131– and long-term follow-up. Eur J Cancer. 2001;37:2204 –2211.
135. 31. Allal AS, Dulguerov P, Allaoua M, et al. Standardized uptake
16. Lowe VJ, Boyd JH, Dunphy FR, et al. Surveillance for recur- value of 2-[18F]fluoro-2-deoxy-D-glucose in predicting out-
rent head and neck cancer using positron emission tomog- come in head and neck carcinomas treated by radiotherapy
raphy. J Clin Oncol. 2000;18:651– 658. with or without chemotherapy. J Clin Oncol. 2002;20:1398 –
17. Wong RJ, Lin DT, Schöder H, et al. Diagnostic and prognos- 1404.
tic value of [18F]fluorodeoxyglucose positron emission to- 32. Halfpenny W, Hain SF, Biassoni L, et al. FDG-PET. A possi-
mography for recurrent head and neck squamous cell car- ble prognostic factor in head and neck cancer. Br J Cancer.
cinoma. J Clin Oncol. 2002;20:4199 – 4208. 2002;86:512–516.