|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2002 American Society for Clinical Oncology Diagnostic and Prognostic Value of [18F]Fluorodeoxyglucose Positron Emission Tomography for Recurrent Head and Neck Squamous Cell CarcinomaByFrom the Departments of Surgery, Nuclear Medicine, Medicine, Radiation Oncology, and Epidemiology and Biostatistics, Memorial Sloan-Kettering Cancer Center, New York, NY. Address reprint requests to Dennis H. Kraus, MD, Head and Neck Service, Department of Surgery, Memorial Sloan-Kettering Cancer Center, 1275 York Ave, Box 285, New York, NY 10021; email: krausd{at}mskcc.org
PURPOSE: Patients with recurrent head and neck squamous cell carcinoma (HNSCC) present a diagnostic and therapeutic challenge. We evaluated the diagnostic accuracy and prognostic value of [18F]fluorodeoxyglucose positron emission tomography (PET) in this patient population. PATIENTS AND METHODS: We performed a retrospective review of 143 patients with previously treated HNSCC who underwent 181 PET scans at our institution from May 1996 through April 2001 to detect recurrent disease. Disease recurrence within 6 months was used as the gold standard for assessing true disease status at PET. RESULTS: With equivocal sites considered positive, the sensitivity and specificity of PET for detecting recurrence overall were 96% and 72%, respectively. PET was highly sensitive and specific at regional and distant sites. At local sites, sensitivity was high, but specificity was lower because of false-positive findings. One fifth of all false-positive PET scans occurred at sites of known inflammation or infection. The area under the curve for a receiver operator characteristic curve on the basis of standardized uptake value (SUV) was 0.882 ± 0.025. PET interpretation, SUV, and physical examination were independent predictors of relapse-free and overall survival in a time-dependent, multivariate proportional hazards model. An increase in SUV by one unit increased the relative risk (RR) of relapse by 11% and the RR of death by 14%. A positive PET interpretation increased the RR of relapse by four-fold and the RR of death by seven-fold. CONCLUSION: PET was a highly sensitive method of detecting recurrent HNSCC and provided important prognostic information for relapse-free and overall survival.
HEAD AND NECK squamous cell carcinoma (HNSCC) is the eighth most common malignancy in the United States1 and refers to a group of malignancies involving the upper aerodigestive tract including the oral cavity, oropharynx, nasopharynx, hypopharynx, and larynx. Unfortunately, approximately two thirds of patients with head and neck cancer present with locally or regionally advanced disease, and less than 30% of these patients are cured with currently available treatment modalities.2 Treatment often requires a multimodality approach that includes combinations of surgery, radiation, and chemotherapy. Such treatment frequently causes significant distortions of normal tissues such as edema or fibrosis, which may obscure the early detection of recurrent disease and hamper subsequent salvage treatment. [18F]Fluorodeoxyglucose (FDG) is a radiolabeled glucose analog that is taken up by cells through the glucose transporter and is subsequently phosphorylated by hexokinase. FDG distribution within the body is a measure of glucose metabolism and may be detected by positron emission tomography (PET). Because cancer cells generally exhibit increased glucose metabolism,3,4 PET imaging with FDG may be used for the detection and localization of a wide variety of malignancies.5 Several small retrospective reports have suggested that FDG-PET may be accurate in detecting recurrent HNSCC, with sensitivities ranging from 84% to 100% and specificities ranging from 61% to 93%.6-10 Two prospective studies have also suggested that PET has value in the routine surveillance for recurrent HNSCC after therapy.11,12 None of these studies, however, examined the utility of PET as a predictor of outcome. In the present study, we retrospectively reviewed a larger series of patients with previously treated HNSCC who underwent FDG-PET imaging for the assessment of possible tumor recurrence. The aims of this study were (1) to determine the diagnostic accuracy of FDG-PET for the detection of recurrent HNSCC, and (2) to evaluate the prognostic value of PET in this patient population.
Patient Review The records of 219 patients with pathologically proven HNSCC who underwent 259 PET scans for any indication at our institution from May 1996 through April 2001 were reviewed. Twenty patients who underwent 20 PET scans were excluded from analysis because of inadequate clinical follow-up. Also excluded were 56 patients with histologically proven, recurrent HNSCC who underwent 58 PET scans to assess for the presence of metastatic disease. The remaining 143 patients all had previously treated HNSCC and were without pathologic evidence of recurrence at the time of PET. These patients underwent 181 PET scans to assess for possible recurrent disease, and form the cohort analyzed in this study. All of the patients included on this study were required to have clinical follow-up to at least one of the following end points after PET: (1) disease recurrence proven pathologically; (2) obvious disease recurrence determined on the basis of clinical examination or radiographic (computed tomographic ([CT] scan and/or magnetic resonance imaging [MRI]) imaging; (3) death; and (4) a minimum 6-month relapse-free interval after PET. The medical records of these patients were reviewed and the following information was retrieved: age, sex, primary tumor site, date of initial diagnosis, initial histologic tumor grade, initial tumor-node-metastasis stage, number and type of prior treatments (surgery, radiation, and chemotherapy), patient symptoms at the time of PET, clinical examination at the time of PET, date and results of all biopsies performed, presence of concurrent infection or inflammation at the time of PET, PET interpretations (positive, equivocal, or negative) on the basis of official PET reports, highest standardized uptake values (SUVs) reported by PET at each site (local, regional, or distant), evidence of recurrent disease after PET (histologic, radiographic, or clinical), site of recurrent disease (local, regional, or distant), date of recurrence, subsequent treatment after recurrence, and status (alive or dead, relapsed or relapse-free) at date of last follow-up.
FDG-PET Imaging
PET Image Interpretation
To minimize partial volume effects, the maximum SUV within a ROI was used. The official PET report was used for this analysis. The PET was primarily interpreted on the basis of visual analysis, using the SUV as an additional criterion. If a finding was initially characterized as clearly normal or suspicious, the reading was not altered by the SUV, although the SUV may have altered the final interpretation of equivocal visual findings. Sites of FDG uptake that were interpreted as suspicious for disease were considered positive, those in which no determination could be rendered were considered equivocal, and those without any evidence to suggest malignancy were considered negative. PET data were acquired from the report of the original PET interpretation.
Correlation of PET Findings Local, regional, and distant sites were assessed independently. An assessment of the overall PET interpretation was made as a composite of the local, regional, and distant sites. For the overall PET interpretation, any single true-positive site was deemed sufficient to consider the overall PET true-positive. All three sites (local, regional, and distant) were required to be true-negative in order to consider the overall PET true-negative. Studies with a false-positive site in the absence of a true-positive focus in any other area were considered false-positive for the overall analysis. Similarly, studies with a false-negative site in the absence of a true-positive finding in other areas were considered false-negative for the overall interpretation of the study.
Statistical Analysis The prognostic value of PET was determined through both univariate and multivariate analyses of relapse-free survival and overall survival. A Cox proportional hazards regression model was used to assess the value of PET and other factors as independent predictors of outcome.14 Both SUV (as a continuous variable) and PET interpretation served as time-dependent covariates in this analysis. Time-dependent covariates were used to make full use of information in patients with repeated PET scans. Model selection was performed in a stepwise manner using likelihood-based measures to assess the incremental prognostic value of PET. The lower and upper quartiles of the SUVs (2 and 6) were used to form three groups (less than 2, 2 to 6, and more than 6). Survival probabilities were calculated on the basis of PET interpretation and SUV grouping using the Kaplan-Meier method and the first PET scan for each patient. A separate analysis of PET sensitivity, specificity, and a multivariate proportional hazards survival model were performed independently for patients who had been treated with chemoradiation therapy before PET scan.
All 143 patients were previously treated for biopsy-proven HNSCC, and underwent 181 PET scans for the detection of recurrent disease. The median follow-up period was 19.5 months from the initial PET scan (range, 1 to 55 months). There were 38 repeat PET studies included that were performed in patients who had undergone a previous PET scan. The mean interval between serial PET studies within the same patient was 7.6 months. The patient and tumor characteristics at initial presentation are listed in Table 1. Treatment before PET consisted of surgery in 115 (64%) of 181 cases, radiation therapy in 176 (97%) of 181 cases, and chemotherapy with or without radiation therapy in 122 (67%) of 181 cases. Chemoradiation therapy preceded PET in 117 of 181 cases (65% of all PET studies), with concurrent treatment in 87 (74%) of 117 cases and sequential treatment in 30 (26%) of 117 cases. The average period between completion of chemotherapy and PET scanning was 6.9 months.
Specific factors that prompted PET included the following: (1) new or suspicious patient symptoms (74 [41%] of 181), (2) equivocal or suspicious findings on clinical examination (69 [38%] of 181), and (3) equivocal or abnormal CT and/or MRI scan before PET (66 [36%] of 181). In 55 cases (30%), PET scans were obtained in the absence of any apparent indication and for posttreatment surveillance. PET was interpreted as locally positive in 59 (33%) of 181 cases and equivocal in seven (4%) of 181; regionally positive in 28 (15%) of 181 cases and equivocal in four (25) of 181; and distantly positive in 37 (20%) of 181 cases and equivocal in two (1%) of 181. A 6-month clinical follow-up interval was used as the gold standard to determine whether disease was present at the time of PET. The majority of recurrences occurred within this interval; of 85 recurrences that occurred at any time after PET, 72 (85%) occurred within this 6-month period. In seven patients, a secondary recurrence was detected after an initial local or regional recurrence was already documented. These secondary sites were not considered to be indicative of disease at the time of PET because they were likely seeded by the initial recurrence. The calculated sensitivity and specificity of the PET interpretation overall, and divided locally, regionally, and distantly, are listed in Table 2. The positive and negative predictive values of the overall PET interpretation were 69% and 96%, respectively.
Overall PET interpretation resulted in 69 true-positive, 78 true-negative, 31 false-positive, and three false-negative studies. Examples of a true-positive and a false-positive PET scan are shown in Fig 1. In nine true-positive studies, there was also a coexisting false-positive focus of FDG uptake. There were no PET scans in which a false-positive and false-negative focus coexisted. Of 31 false-positive PET scans, six (19%) were in patients with known inflammation or infection at the sites of FDG uptake. Of these, three were in patients who underwent biopsies just before the scan, two had concurrent osteoradionecrosis, and two had concurrent infections including herpes shingles and pneumonia. In 36 PET scans (20%), a true-positive focus of disease was detected that was not previously suspected on the basis of prior clinical examination, patient symptoms, or conventional radiologic imaging.
Of the 181 PET scans reviewed, 180 had a physical examination performed at the time of the PET. Physical examination (PE) at the time of PET was positive in 69 cases and negative in 111 cases. The sensitivity and specificity of PE in detecting recurrence was 60% and 76%, respectively (Table 2). The positive and negative predictive values were 62% and 74%, respectively. The mean SUV for patients who suffered any subsequent recurrence (n = 69) was 5.8 ± 3.7, compared with 2.0 ± 2.3 for those who did not recur (n = 74). The mean SUV was 5.8 ± 3.8 for patients who died (n = 46), compared with 2.9 ± 3.1 for those who were alive at last follow-up (n = 97). An ROC curve was plotted on the basis of an analysis of the highest SUV for each PET overall (Fig 2). The AUC of the ROC curve for SUV overall was 0.882 ± 0.025. When examined independently at different sites, these AUC values were 0.918 ± 0.024 for local, 0.963 ± 0.022 for regional, and 0.94 ± 0.025 for distant sites, respectively.
On the basis of the ROC, selecting an SUV of 4 as a determinant of malignancy would yield a sensitivity of 79% and a specificity of 78% (Fig 2). Selecting an SUV of 3.2 as a cutoff value would yield a sensitivity of 92% and a specificity of 70%, which more closely approach the performance of the physicians’ PET interpretation. There was no SUV operating point on the ROC that equalled or surpassed the sensitivity (96%) and specificity (72%) of the physicians’ PET interpretation. PET interpretation (Fig 3) and highest SUV (Fig 4) demonstrated a significant impact on both relapse-free survival (RFS) and overall survival (OS). In patients with a positive PET, the 2-year RFS and OS were 23% and 48%, respectively, compared with 82% and 97%, respectively, in patients with a negative PET.
PET interpretation, highest SUV (as a continuous variable), and PE were all significant and independent predictors of RFS and OS in both univariate and multivariate proportional hazards models (Tables 3 and 4). Because symptoms was highly correlated with PE (Kendall’s tau = 0.47, P < .0001), it was not included in the final multivariate model, although it was highly significant in the univariate analysis. Highly significant changes were observed for the likelihood ratio test when PET interpretation (P = .0011 for OS and P = .0003 for RFS) and SUV (P = .0045 and P = .003) were added to the model after adjustments for clinical examination, demonstrating the incremental prognostic value of the PET scan. When analyzed as a continuous variable, an increase in one unit of SUV increased the patient’s relative risk (RR) of relapse by 11% and the RR of death by 14% (Tables 3 and 4). A positive PET interpretation by the nuclear medicine physician resulted in the greatest relative risk increases for both relapse (four-fold) and death (seven-fold) of all the parameters analyzed.
A separate subgroup analysis was performed of patients who had undergone chemoradiation therapy before PET scan (n = 117). For this subgroup, the local, regional, and distant PET sensitivities and specificities for detecting recurrence were 95% and 81%, 88% and 94%, and 92% and 93%, respectively. The overall PET sensitivity and specificity were 92% and 68%, respectively. A multivariate proportional hazards analysis was performed for the chemoradiation subgroup, and demonstrated that both PET interpretation and SUV were significant predictors of OS, with RRs of 5.7 and 1.18 (per unit increase in SUV), respectively. PET therefore exhibited a diagnostic and prognostic value for the chemoradiation subgroup similar to that for the overall group.
This study shows that FDG-PET is highly sensitive in detecting recurrent HNSCC. PET was most accurate for detecting recurrent disease at regional and distant sites, whereas the specificity at local sites was diminished because of false-positive findings. In addition, positive PET interpretation by a nuclear medicine physician and higher SUV were both independent and strong prognostic determinants of adverse clinical outcome. The initial treatment of advanced-stage HNSCC often requires a multimodality approach using combinations of surgery, radiation, and chemotherapy. Such treatment may cause significant distortions of the treated tissues in the head and neck region. These anatomic alterations often impair an accurate clinical and radiographic examination of the primary tumor site, hampering the early detection of recurrent disease. The late detection of recurrent HNSCC has been shown to be an adverse factor in clinical outcome after salvage surgery. Patients with recurrent, early-stage HNSCC who undergo salvage surgery have a 70% 2-year RFS, whereas those with recurrent, advanced-stage HNSCC undergoing surgical salvage have just a 22% 2-year RFS.15 The early detection of recurrent HNSCC is therefore critically important for achieving successful surgical salvage. The sensitivity and specificity for the overall PET interpretation in detecting recurrence were 96% and 72%, respectively. For this analysis, the identification of any one true-positive site of FDG uptake was deemed sufficient to designate the overall PET true-positive. This designation reflects the fact that the successful detection of any true recurrence in a given patient is likely the most clinically relevant factor in determining that patient’s subsequent therapy and prognosis. When analyzed by site of recurrence, the specificity of PET was diminished at local sites because of a higher rate of false-positive findings. In contrast, PET had excellent sensitivity and specificity at regional and distant sites. An analysis of SUV as a determinant of recurrence demonstrated a similar pattern. The ROC curves for SUV had higher AUC values for regional and distant sites than local sites, indicating a greater diagnostic accuracy at these areas. The higher false-positive findings at local sites were likely related to inflammation and other posttreatment effects, which may enhance FDG uptake in the absence of malignancy. Any anatomic site with increased glucose metabolism will also concentrate FDG, and may be falsely interpreted as a site of malignancy. In this study, one fifth of all false-positive PET studies had known inflammation, infection, or recent biopsy at the sites of FDG uptake. Closer communication between clinicians and nuclear medicine physicians, and the avoidance of biopsies immediately before PET scanning, may potentially reduce the number of false-positive interpretations in the future. The optimal timing of PET after completion of therapy to limit false-positive interpretations requires further investigation. We assessed the utility of using SUV as a sole determinant of malignancy. Our ROC analysis demonstrated that an SUV of 3.2 yielded sensitivity and specificity most closely approaching that of PET interpretation by the nuclear medicine physician. There was no SUV that surpassed the physician’s performance. This finding is likely related to several factors, including the physician’s understanding of the patient’s clinical history and prior radiologic studies, and the physician’s expertise in interpreting the PET study. SUV itself was dependent on the physician. Experienced nuclear medicine physicians are familiar with typical and atypical patterns of FDG uptake, and calculate SUV only in areas that are considered suspicious or equivocal by visual analysis. In addition, SUV may be affected by many different factors, including body fat, time of measurement after injection, blood supply to a given organ, and the size of the ROI.16,17 Therefore, an optimal SUV cutoff identified for diagnostic accuracy should not necessarily be interpreted as a guideline for use in clinical practice. It must be emphasized that SUV is not a replacement for, but merely an aid to, the physician in his or her interpretation of PET findings. This study was not designed or intended to compare different modalities for detecting recurrence. Many patients on this study were selected for PET because of equivocal or ambiguous findings on CT scanning, MRI (data not shown), or PE. This therefore poses a selection bias that precludes a fair comparison between these modalities. Furthermore, because the assessment of locoregional recurrence was the most common focus in these patients, there was a decreased intensity of surveillance for distant metastases by anatomic imaging. However, this algorithm does closely reflect the clinical practice at our institution and at many others. PE, CT scanning, and MRI are often part of the routine evaluation for patients with HNSCC after therapy, and may guide the clinician’s decision to obtain a PET scan. Our results demonstrate that PET was remarkably sensitive and fairly specific in those problematic patients who had ambiguous findings by conventional methods and in whom the diagnosis remained elusive. For these cases, a positive PET scan may encourage a clinician to pursue more aggressive biopsies in an attempt to establish a definitive pathologic diagnosis. Survival analyses were performed using a time-dependent covariate proportional hazards model. Initial tumor factors such as stage and grade were not significant factors, an understandable finding in this setting where patients have already had their initial disease treated. PET interpretation by the nuclear medicine physician, SUV, and PE were all independent, significant predictors for RFS and OS in a multivariate model. A positive PET interpretation led to significantly increased risks of both relapse and death. Because modeling was performed in a stepwise manner, we may also conclude that PET interpretation and SUV have significant incremental information beyond clinical examination. The finding that PET carries significant prognostic value may have clinical implications regarding patient counseling, subsequent treatment, and surveillance. An important limitation of this retrospective study is recognizing that these results are specific to the characteristics of the patient population seen at this tertiary level institution. It is possible that the false-positive rate of PET might be lower within a population of patients that did not have any suspicion of recurrence who underwent PET solely as a means of surveillance. Similarly, the sensitivity of PET might be lower within such a population without suspicion of recurrence, as the disease burden may be lower. Therefore, these results may not necessarily apply to any population of patients undergoing PET to detect recurrence. PET interpretation might also have had effects on subsequent clinical decisions that may have affected subsequent clinical outcome. Our results also reflect the expertise of this institution’s nuclear medicine physicians in interpreting PET scans. Other limitations of this study include possible variability of interpretation between different nuclear medicine physicians, bias introduced from multiple PET studies being performed in certain patients, and potential underreporting of false-positive foci in PET studies with multiple foci of FDG uptake. In conclusion, these findings demonstrate that PET was an accurate method of detecting recurrent HNSCC, with a high sensitivity at all sites, and a high specificity at regional and distant sites for our patient population. The specificity of PET at local sites was somewhat reduced because of false-positive findings, which were likely related to treatment effects. The diagnostic performance of PET in this population is remarkable when considering that many patients had inconclusive results from prior imaging studies and remained diagnostic dilemmas at the time of PET. These results also demonstrate that PET interpretation and SUV were both highly significant and independent predictors for RFS and OS. PET represents a novel prognostic parameter with possible implications for patient counseling, treatment, and surveillance. Although prospective clinical trials are needed to fully evaluate the role of PET in patients with suspected recurrent HNSCC, this study suggests that PET has an important diagnostic and prognostic utility in these problematic patients.
Supported in part by training grant no. T32CA09685 from the National Institutes of Health.
Presented at the Annual Meeting of the American Head and Neck Society, Boca Raton, FL, May 11, 2002, and received the Robert M. Byers Award.
1. Jemal A, Thomas A, Murray T, Thun M: Cancer statistics, 2002. CA Cancer J Clin 52: 23-47, 2002
2. Vokes EE, Weichselbaum RR, Lippman SM, Hong WK: Head and neck cancer. N Engl J Med 328: 184-194, 1993 3. Warburg O: Metabolism of Tumors. New York NY, Richard R Smith, 1931 4. Weber G: Enzymology of cancer cells: Part 2. N Engl J Med 296: 541-551, 1977[Medline]
5. Phelps ME: Positron emission tomography provides molecular imaging of biological processes. Proc Natl Acad Sci U S A 97: 9226-9233, 2000 6. Lonneaux M, Lawson G, Ide C, et al: Positron emission tomography with fluorodeoxyglucose for suspected head and neck tumor recurrence in the symptomatic patient. Laryngoscope 110: 1493-1497, 2000[CrossRef][Medline] 7. Stokkel MP, Terhaard CH, Hordijk GJ, et al: The detection of local recurrent head and neck cancer with fluorine-18 fluorodeoxyglucose dual-head positron emission tomography. Eur J Nucl Med 26: 767-773, 1999[CrossRef][Medline] 8. Lapela M, Eigtved A, Jyrkkio S, et al: Experience in qualitative and quantitative FDG PET in follow-up of patients with suspected recurrence from head and neck cancer. Eur J Cancer 36: 858-867, 2000[CrossRef][Medline] 9. Farber LA, Benard F, Machtay M, et al: Detection of recurrent head and neck squamous cell carcinomas after radiation therapy with 2-18F-fluoro-2-deoxy-D-glucose positron emission tomography. Laryngoscope 109: 970-975, 1999[CrossRef][Medline] 10. Greven KM, Williams DW, McGuirt WF, et al: Serial positron emission tomography scans following radiation therapy of patients with head and neck cancer. Head Neck 23: 942-946, 2001[CrossRef][Medline]
11. Lowe VJ, Boyd JH, Dunphy FR, et al: Surveillance for recurrent head and neck cancer using positron emission tomography. J Clin Oncol 18: 651-658, 2000 12. Terhaard CH, Bongers V, van Rijk PP, et al: F-18-Fluoro-deoxy-glucose positron-emission tomography scanning in detection of local recurrence after radiotherapy for laryngeal/pharyngeal cancer. Head Neck 23: 933-941, 2001[CrossRef][Medline] 13. Bamber D: The area above the ordinal dominance graph and the area below the receiver operating characteristic graph. J Math Psychol 12: 387-415, 1975[CrossRef] 14. Kalbfleisch JD, Prenctice RL: The Statistical Analysis of Failure Time Data. New York NY, Wiley, 1980 15. Goodwin WJ: Salvage surgery for patients with recurrent squamous cell carcinoma of the upper aerodigestive tract: When do the ends justify the means? Laryngoscope 110: 1-18, 2000 (suppl 93)[Medline]
16. Sugawara Y, Zasadny KR, Neuhoff AW, Wahl RL: Reevaluation of the standardized uptake value for FDG: Variations with body weight and methods for correction. Radiology 213: 521-525, 1999
17. Keyes JW Jr: SUV: Standard uptake or silly useless value? J Nucl Med 36: 1836-1839, 1995 Submitted February 19, 2002; accepted July 8, 2002.
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2002 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
