Detection of Hematogenous Bone Metastasis in Cervical Cancer

Detection of Hematogenous Bone Metastasis in Cervical Cancer 18 F-Fluorodeoxyglucose Positron Emission Tomography Versus Computed Tomography and Magnetic Resonance Imaging Feng-Yuan Liu, MD 1 ; Tzu-Chen Yen, MD, PhD 1 ; Min-Yu Chen, MD 2 ; Chyong-Huey Lai, MD 2 ; Ting-Chang Chang, MD 2 ; Hung-Hsueh Chou, MD 2 ; Ji-Hong Hong, MD 3 ; Yu-Ruei Chen, MD 4 ; and Koon-Kwan Ng, MD 4 BACKGROUND: In this large-scale, retrospective study, the authors evaluated the diagnostic performances of computed tomography (CT), magnetic resonance (MR) imaging, and 18 F-fluorodeoxyglucose-positron emission tomography ( 18 F-FDG PET) in detecting hematogenous bone metastasis in patients with cervical cancer. The associated risk factors also were analyzed. METHODS: Patients with invasive cervical cancer who had both 18 F-FDG PET studies and CT or MR imaging studies were selected. Patients who had either International Federation of Gynecology and Obstetrics (FIGO) stage III/IV disease or positive lymph node metastasis at the time of primary staging and patients who had suspected recurrent disease were included in the analyses. The diagnostic performances of PET was compared with the performance of CT and MR imaging by using the area under the receiver-operating-characteristic curve (AUC). Both univariate and multivariate analyses were applied to assess the risk factors for hematogenous bone metastasis at primary staging. RESULTS: PET was more sensitive than CT (P ¼.004) and was more specific than MR imaging (P ¼.04). The diagnostic performance of PET was significantly superior to the performance CT (AUC, 0.964 vs 0.662; P <.001) and MR (AUC, 0.966 vs 0.833; P ¼.033). Both FIGO stage and the extent of lymph node metastases were associated with hematogenous bone metastasis in univariate analysis. However, the extent of lymph node metastases was the only significant risk factor in multivariate analysis (P ¼.025). CONCLUSIONS: The current study demonstrated the superiority of 18 F-FDG PET over CT and MR imaging for detecting hematogenous bone metastasis in patients with advanced cervical cancer. Hematogenous bone metastasis in cervical cancer was associated with the extent of lymph node metastases rather than with FIGO stage. Cancer 2009;115:5470 80. VC 2009 American Cancer Society. KEY WORDS: cervical cancer, hematogenous bone metastasis, 18 F-fluorodeoxyglucose positron emission tomography, computed tomography, magnetic resonance imaging. Corresponding author: Koon-Kwan Ng, MD, Department of Diagnostic Radiology, Chang Gung Memorial Hospital, No. 5, Fusing Street, Gueishan Township, Taoyuan County 33305, Taiwan; Fax: (011) 886-3-2110052; ngkk0412@adm.cgmh.org.tw 1 Department of Nuclear Medicine and Molecular Imaging Center, Chang Gung Memorial Hospital, Chang Gung University College of Medicine, Taoyuan, Taiwan; 2 Department of Obstetrics and Gynecology, Chang Gung Memorial Hospital, Chang Gung University College of Medicine, Taoyuan, Taiwan; 3 Department of Radiation Oncology, Chang Gung Memorial Hospital, Chang Gung University College of Medicine, Taoyuan, Taiwan; 4 Department of Diagnostic Radiology, Chang Gung Memorial Hospital, Chang Gung University College of Medicine, Taoyuan, Taiwan Koon-Kwen Ng and Tzu-Chen Yen contributed equally to this work. We thank all members of the cross-disciplinary team for gynecologic oncology in our institution for their devotion and help. Received: January 13, 2009; Revised: March 2, 2009; Accepted: March 5, 2009 Published online September 8, 2009 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/cncr.24599, www.interscience.wiley.com 5470 Cancer December 1, 2009

Detecting Bone Metastasis in CC/Liu et al Cervical cancer remains the second most common malignancy among women and 1 of the leading causes of cancer-related death worldwide. The most common sites for hematogenous spread are the lung, bone, and liver. 1-3 Bone metastasis in patients with cervical cancer occurs occasionally and is associated with advanced disease and a poor prognosis. 3,4 Accurate detection of bone metastasis is important for staging and proper management, such as the use of bisphosphonates. 5 In the past, patients who had suspected metastasis usually were worked up with x-ray, computed tomography (CT), ultrasonography, and whole-body bone scintigraphy. In the modern era, magnetic resonance (MR) imaging and positron emission tomography (PET) or integrated PET/CT using 18 F-fluorodeoxyglucose ( 18 F- FDG) increasingly are used. Both MR and PET are considered to have high sensitivity for detecting bone marrow or osteolytic bone metastasis. 6-11 Because hematogenous bone metastasis is considered to start in the bone marrow, and the majority of metastatic bone lesions in cervical cancer seem to be of an osteolytic nature, 12 both MR imaging and PET may facilitate the detection of bone metastasis. To the best of our knowledge, no study had been conducted to evaluate the diagnostic performances of CT, MR imaging, and PET in detecting hematogenous bone metastasis in patients with invasive cervical cancer. In this largescale, retrospective analysis, we tried to address this question. We also analyzed the factors associated with hematogenous bone metastasis in cervical cancer as a second objective. MATERIALS AND METHODS Patients The current study was based on patients with invasive cervical cancer who had 18 F-FDG PET and either CT or MR studies performed within 30 days and had signed informed consent for research. Patients who had a history of other malignancy or a follow-up duration <180 days after 18 F-FDG PET (except those who died of disease within 180 days) were excluded. Because bone lesions resulting from direct extension of contiguous soft tissue masses, such as locoregional tumors, lymphadenopathies, or distant soft tissue metastases, were not considered hematogenous bone metastases, patients who had only those types of bone lesions also were excluded. The PET machine had been installed in 2001 at our institution and was replaced in May 2006 by a new PET/CT machine. Studies performed by the new PET/CT machine were not included in the current analysis. The imaging data were grouped as either data obtained for primary staging or data obtained to evaluate suspected recurrent disease. For primary staging, patients with either International Federation of Gynecology and Obstetrics (FIGO) stage III/IV disease or positive lymph node metastasis were included. The criteria for positive lymph nodes were detailed in a previous study. 13 In the setting of suspicious disease recurrence, no further selection criteria were applied. In our institution, patients with pathologically confirmed, invasive cervical cancer undergo either CT or, preferably, MR imaging of the pelvis and the abdomen, and sometimes both. Several prospective studies concerning 18 F-FDG PET for primary staging in cervical cancer have been conducted. After curative-intent treatment, the surveillance program consists of visits at 3-month intervals for 2 years, 4-month intervals for the third year, 6-month intervals for the fourth and fifth years, and yearly intervals thereafter if no recurrence is suspected. Clinical history, physical and pelvic examinations, and Papanicolaou (Pap) smears along with serum tumor markers, including squamous cell carcinoma antigen and/or carcinoembryonic antigen, are checked at every visit. Chest x-rays are obtained yearly, and CT or MR imaging studies are obtained when disease recurrence is suspected (suspicious symptoms/signs, positive Pap smear, or tumor marker elevation). 18 F-FDG PET images also were obtained in several prospective studies of patients after they received curative-intent treatment for restaging. Computed Tomography CT images were obtained with a Hi-Speed Advantage CT Scanner (GE Medical Systems, Milwaukee, Wis) or with a Somatom Plus 4 multislice CT scanner (Volume Zone, version A40; Siemens AG Medical, Forscheim, Germany). Patients fasted for at least 4 hours before the study and received oral diatrizoate meglumine (Gastrografin; Schering Health Care, Burgess Hill, United Kingdom) as a gastrointestinal contrast. Contiguous 5-mm-thick slices Cancer December 1, 2009 5471

were obtained in a craniocaudal direction from the lower chest to the pelvis. Contrast-enhanced imaging was performed approximately 30 seconds after intravenous injection of 100 ml iohexol (Omnipaque; GE Healthcare, Little Chalfont, United Kingdom). Magnetic Resonance Imaging MR images were obtained with a 1.5-T Magnetom Vision MR scanner (Siemens Medical Systems, Erlangen, Germany) using a phased-array body coil. For the abdomen, a T2-weighted, transverse spin-echo sequence (repetition time/echo time, 6000 msec/120 msec; number of signals acquired, 5; flip angle, 180 degrees; section thickness, 8 mm; gap, 2 mm; matrix, 256 256; field of view, 40 cm) was applied. For the pelvis, a T1- weighted, transverse spin-echo sequence (repetition time/echo time, 500 msec/15 msec; number of signals acquired, 2; flip angle, 90 degrees; section thickness, 5 mm; gap, 2 mm; matrix, 256 256; field of view, 20 cm); a T2-weighted, transverse spin-echo sequence (repetition time/echo time, 4000 msec/99 msec; number of signals acquired, 2; flip angle, 180 degrees; section thickness, 5 mm; gap, 2 mm; matrix, 256 256; field of view, 20 cm); and a T2-weighted, sagittal spin-echo sequence (repetition time/echo time, 4000 msec/99 msec; number of signals acquired, 2; flip angle, 180 degrees; section thickness, 5 mm; gap, 2 mm; matrix, 256 256; field of view, 20 cm) were applied. Another T2-weighted, coronal spin-echo sequence (repetition time/echo time, 5500 msec/130 msec; number of signals acquired, 2; flip angle 180 degrees; section thickness, 5 mm; gap, 2 mm; matrix, 256 256; field of view, 40 cm) was applied for the pelvic girdle and lower lumbar spine. Two or 3 sets of T1-weighted, spin-echo images with fat saturation were obtained immediately after the administration of intravenous gadopentetate dimeglumine (Magnevist; Bertex Laboratories, Wayne, NJ) at a dose of 0.1 mmol/kg. These included a transverse sequence (repetition time/echo time, 1100 msec/ 20 msec; number of signals acquired, 2; flip angle, 90 degrees; section thickness, 5 mm; gap, 2 mm; matrix, 256 256; field of view, 20 cm) and a sagittal sequence (repetition time/echo time, 1100 msec/20 msec; number of signals acquired, 2; flip angle, 90 degrees; section thickness, 5 mm; gap, 2 mm; matrix, 256 256; field of view, 20 cm) regularly. Additional postcontrast coronal sequence (repetition time/echo time, 1100 msec/20 msec; number of signals acquired 2; flip angle, 90 degrees; section thickness, 5 mm; gap, 2 mm; matrix, 256 256; field of view, 40 cm) would be acquired if the precontrast coronal images for the pelvic girdle and lower lumbar spine revealed suspicious abnormalities. 18 F-Fluorodeoxyglucose Positron Emission Tomography All patients fasted for at least 6 hours before the examination and received intravenous injections of 370 megabecquerels 10% 18 F-FDG after initial preparation. For optimal tumor identification, all patients were catheterized to reduce bladder activity. An ECAT EXACT HRþ PET scanner (CTI/Siemens, Knoxville, Tenn) with a resolution of 4.5 mm (full width at half-maximum) and a 15-cm axial field of view was used to acquire images. Early-phase PET acquisition from the skull base to the upper thighs was started 40 minutes after injection. Delayed-phase PET imaging for 1 or 2 beds was acquired at about 3 hours after injection under the attending physician s discretion. Transmission scans using the germanium-68 rod source were obtained for attenuation correction. The accelerated maximum-likelihood algorithm with ordered subset expectation maximization was applied for image reconstruction. The obtained maximal standardized uptake values (SUVmax) of the lesions were available to interpreters. The maximal SUVmax of the osseous lesion(s) interpreted as metastases in the same patient was recorded. Image Interpretation The following 5-point scoring system for imaging findings was used: 0 indicated negative; 1, probably benign; 2, equivocal; 3, possibly malignant; and 4, malignant. For CT and MR studies, the imaging findings were interpreted by 2 radiologists with 21 years (K-K.N.) and 3 years (Y-R.C.) of experience in consensus. For PET studies, the imaging findings were interpreted by 2 nuclear physicians with 10 years (T-C.Y.) and 4 years (F-Y.L.) of 5472 Cancer December 1, 2009

Detecting Bone Metastasis in CC/Liu et al experience in consensus. The SUVmax for each lesion was provided to interpreters, but no threshold was used for scoring. The interpreters were blinded to the patient s history and clinical data. For calculating sensitivity, specificity, and other statistical analyses, scores of 3 and 4 were considered as positive, and other scores were considered negative. Then, the results of imaging studies were presented in the weekly combined conference of gynecologic oncology. Further imaging studies would be arranged if the positive lesion on the PET study was outside the imaging field of the original CT or MR image. These additional studies would not alter the scoring in the current study but would be referenced for the determination of bone metastasis, as detailed below. Determining the Presence of Hematogenous Bone Metastasis Patients were considered to have hematogenous bone metastasis if the detected bone lesion was positive on both 18 F-FDG PET studies and either CT or MR studies along with a concordant clinical course of progression with or without a transient response to palliative treatment. Because bone biopsy is an invasive procedure, histologic proof of bone metastasis was considered necessary only if it was critical for decisions regarding further management when the imaging results were not consistent (positive on 1 modality and negative on the other modality). For patients who had concurrent distant visceral metastasis, bone biopsy would not be arranged, and progressive findings on imaging follow-up studies were used as the standard of proof. New evidence of hematogenous bone metastasis that appeared within 180 days from the date of 18 F-FDG PET imaging also was considered positive if the above criteria were met. Analysis of Risk Factors for Hematogenous Bone Metastasis An analysis of risk factors for hematogenous bone metastasis, including age, tumor cell type, tumor differentiation, SUVmax of the primary tumor, FIGO stage, and extent of lymph node metastases, was performed in patients for primary staging. Patients who had either FIGO stage III/IV disease or positive lymph node metastasis were included. Age was classified as <45 years, 45 years to 65 years, and >65 years. Tumor cell type was classified as squamous cell carcinoma, adenocarcinoma, adenosquamous carcinoma, and neuroendocrine tumor. Tumor differentiation was classified as well differentiated, moderately differentiated, and poorly differentiated. For the SUVmax of the primary tumor, the median of the values obtained from whole-body PET images was determined, and patients were classified accordingly into 2 groups. FIGO stage was determined by the patient s attending gynecologist. The extent of lymph node metastases was classified as negative, pelvic, para-aortic or inguinal, supraclavicular, and mediastinal according to the most distant positive site along the route of lymphatic spread as determined by criteria detailed in a previous study. 13 Statistical Analysis Sensitivities, specificities, and accuracies of the CT, MR imaging, and PET modalities for detecting hematogenous bone metastasis were calculated using the standard method and were compared by using the McNemar paired-sample test. Positive and negative predictive values were compared using the method proposed by Leisenring et al. 14 The diagnostic performances of PET were compared with those of CT and MR imaging using receiver operating characteristic (ROC) curves. Areas under the empirical ROC curves (AUCs) were estimated nonparametrically and were compared using the method of Hanley and McNeil. 15 Both univariate exact chi-square analysis and a multivariate logistic regression model were applied to analyze the risk factors for patients with hematogenous bone metastasis. All statistical analyses were 2- sided, and the significance level was fixed at P ¼.05. Statistical analyses were performed using the MedCalc (version 7.6.0; MedCalc Software, Mariakerke, Belgium) or SPSS (version 13.0; SPSS Inc., Chicago, Ill) statistical software packages. RESULTS Patient Demographics For the comparison of CT and PET, 233 imaging pairs in 190 patients were eligible (Group CT/PET). For the Cancer December 1, 2009 5473

Table 1. Characteristics of the Study Groups No. of Patients Characteristic Group CT/PET Group MR/PET Total No. of imaging pairs 233 245 Indication for imaging Primary staging 40 146 165 Suspected disease recurrence 193 99 277 No. of patients 190 228 356 Age at initial diagnosis: MeanSD, y 5211 5312 5212 Tumor cell type Squamous cell carcinoma 144 177 273 Adenocarcinoma 20 25 37 Adenosquamous carcinoma 17 17 29 Neuroendocrine tumor 9 9 17 Initial FIGO stage I 83 65 131 II 63 97 140 III 37 53 69 IV 7 13 16 CT indicates computed tomography; PET, positron emission tomography; SD, standard deviation; FIGO, International Federation of Gynecology and Obstetrics. comparison of MR imaging and PET, 245 imaging pairs in 228 patients were eligible (Group MR/PET). All CT and MR studies included precontrast and postcontrast images. Baseline characteristics of the patient groups are presented in Table 1. Some patients had more than 1 pair of imaging studies throughout the disease course, and some patients had both CT and MR studies, which could be compared with the same PET study. In total, there were 356 patients, and their mean age was 52 years at the initial diagnosis of cervical cancer. The percentage of patients with squamous cell carcinoma, adenocarcinoma, adenosquamous carcinoma, and neuroendocrine tumor was 77%, 10%, 8%, and 5%, respectively. Initial FIGO staging was stage I in 37% of patients, stage II in 39% of patients, stage III in 19% of patients, and stage IV in 4% of patients. MR images had been obtained more frequently for primary staging, and CT scans had been more frequently for suspected recurrent disease. MR imaging was preferred over CT for primary staging, because MR imaging can provide more precise local tumor status. In contrast, CT scans were performed more frequently in patients who had suspected recurrent disease because of the problem with scheduling delays for MR imaging. Patients With Hematogenous Bone Metastasis In Group CT/PET, 15 patients were positive for hematogenous bone metastasis, including 4 patients at the time of primary staging and 11 patients at the time of recurrence. In Group MR/PET, 17 patients were positive for hematogenous bone metastasis, including 10 patients at the time of primary staging and 7 patients at the time of recurrence. In total, there were 29 patients who were positive for hematogenous bone metastasis (3 patients were in both Group CT/PET and Group MR/PET). Excluding 1 patient who had a false-negative PET study, the maximal SUVmax of osseous metastases in 28 patients ranged from 2.3 to 15.7 (mean standard deviation, 6.4 3.5). Fourteen patients were positive for hematogenous bone metastasis by findings from both imaging modalities with a concordant clinical course, 6 patients had histologic proof of bone metastasis by biopsy, and the remaining 9 patients were considered positive for hematogenous bone metastasis on follow-up imaging studies. In 165 patients at the time of primary staging (FIGO stage III/IV disease or positive lymph node metastasis), 12 patients (7.3%) were considered 5474 Cancer December 1, 2009

Detecting Bone Metastasis in CC/Liu et al Table 2. Comparison of the Diagnostic Performance, Sensitivity, Specificity, Positive Predictive Value, Negative Predictive Value, and Accuracy of Computed Tomography, Magnetic Resonance Imaging, and 18 F-Fluorodeoxyglucose Positron Emission Tomography for Hematogenous Bone Metastasis Performance Parameter AUC Sensitivity Specificity PPV NPV Accuracy Group CT/PET Primary staging CT 0.625 0.250 1.000 1.000 0.923 0.925 PET 1.000 1.000 1.000 1.000 1.000 1.000 P 0.017 0.25 0.07 0.25 Suspected disease recurrence CT 0.676 0.364 0.989 0.667 0.963 0.953 PET 0.952 0.909 0.995 0.909 0.995 0.990 P 0.005 0.03 1.0 0.26 0.01 0.04 Total CT 0.662 0.333 0.991 0.714 0.956 0.948 PET 0.964 0.933 0.995 0.933 0.995 0.991 P <0.001 0.004 1.0 0.24 0.002 0.006 Group MR/PET Primary staging MR 0.893 0.800 0.985 0.800 0.985 0.973 PET 0.996 1.000 0.993 0.909 1.000 0.993 P 0.12 0.5 1.0 0.22 0.15 0.25 Suspected disease recurrence MR 0.748 0.571 0.924 0.364 0.966 0.899 PET 0.923 0.857 0.989 0.857 0.989 0.980 P 0.14 0.5 0.07 0.008 0.14 0.02 Total MR 0.833 0.706 0.961 0.571 0.978 0.943 PET 0.966 0.941 0.991 0.889 0.996 0.988 P 0.033 0.13 0.04 0.004 0.04 0.003 AUC indicates area under the receiver operating characteristic curve; PPV, positive predictive value; NPV, negative predictive value; CT, computed tomography; MR, magnetic resonance; PET, positron emission tomography. positive for hematogenous bone metastasis. In 277 imaging pairs from 226 patients who had suspected recurrent disease, 17 images (6.1%) from 17 patients (7.5%) were considered positive for hematogenous bone metastasis. Sensitivity, Specificity, Predictive Values, Accuracy, and Diagnostic Performance The sensitivity, specificity, accuracy, positive predictive value, negative predictive value, and diagnostic performance of CT, MR imaging, and PET are presented in Table 2. In Group CT/PET, hematogenous bone metastases were detected by CT and PET in 5 of 15 instances and in 14 of 15 instances, respectively; and CT and PET results were false-positive in 2 instances and in 1 instance, respectively, all in the subgroup of patients who had suspected recurrent disease. In Group MR/PET, hematogenous bone metastases were detected by MR imaging and PET in 12 of 17 instances and in 16 of 17 instances, respectively; and MR imaging and PET results were falsepositive in 9 instances and in 2 instances, respectively. Seven false-positive MR imaging results were observed the subgroup of patients who had suspected recurrent disease. Overall, PET was more sensitive than CT (P ¼.004) but was not more sensitive than MR imaging (P ¼.13). PET was more specific than MR imaging (P ¼.04) but was not more specific than CT (P ¼ 1.0). PET had a higher positive predictive value than MR imaging (P ¼.004) and a higher negative predictive value than both CT (P ¼.002) and MR imaging (P ¼.04). PET was more accurate than both CT (P ¼.006) and MR imaging (P ¼.003). The diagnostic performance of PET, as expressed Cancer December 1, 2009 5475

FIGURE 1. These charts compare the area under the receiver operating characteristic curve (AUC) for (A) positron emission tomography (PET) versus computed tomography (CT) and for (B) PET versus magnetic resonance (MR) imaging in detecting hematogenous bone metastasis in patients with invasive cervical cancer. by AUC values, was significantly superior to the performance of CT (P <.001) and MR imaging (P ¼.033) (Fig. 1). A patient who had a true-positive PET scan and a false-negative CT scan (Fig. 2) and another patient who had a true-negative PET scan and a false-positive MR image (Fig. 3) are provided. In the subgroup analysis, CT was less sensitive than PET for evaluating suspected recurrent disease (P ¼.03) but not for primary staging (P ¼.25), possibly because of the smaller sample size of Group CT/PET at the time of primary staging. MR imaging had a lower positive predictive value than PET for suspected recurrent disease (P ¼.008) but not for primary staging (P ¼.22). FIGURE 2. These images were obtained from a woman aged 75 years who had cervical squamous cell carcinoma that was classified as International Federation of Gynecology and Obstetrics stage IIB disease with both pelvic and inguinal lymph node metastasis. (A) This transverse, sectional 18 F-fluorodeoxyglucose positron emission tomography image revealed 2 foci of metastases in the right ischial ramus. (B) This transverse, sectional, contrast-enhanced computed tomography image revealed slightly increased bone marrow density in the right ischial ramus that was not interpreted as positive for bone metastasis. Analysis of Risk Factors for Hematogenous Bone Metastasis An analysis of the risk factors for hematogenous bone metastasis is presented in Table 3. In total, 165 patients with either FIGO stage III/IV disease or positive lymph node metastasis at primary staging were included for analysis. Hematogenous bone metastasis were identified in 4% of patients aged <45 years, in 6% of patients ages 45 years to 65 years, and in 16% of patients aged >65 years. The percentage of patients with hematogenous bone metastasis was 7% for squamous cell carcinoma, 7% for adenocarcinoma, 6% for adenosquamous carcinoma, and 11% for neuroendocrine tumors. The percentage of patients with hematogenous bone metastasis was 0% for well differentiated tumors, 9% for moderately differentiated tumors, and 7% for poorly differentiated tumors. The median 5476 Cancer December 1, 2009

Detecting Bone Metastasis in CC/Liu et al FIGURE 3. These images were obtained from a woman aged 60 years who had elevated levels of serum squamous cell carcinoma antigen about 5 years after she received chemoradiation for cervical squamous cell carcinoma that was categorized as International Federation of Gynecology and Obstetrics stage IIB. (A) This transverse, sectional, T1- weighted magnetic resonance image with contrast enhancement and fat saturation revealed a left iliac bone abnormality (arrow), which was favored to be a skeletal metastasis. (B) This transverse, sectional 18 F-fluorodeoxyglucose (FDG) positron emission tomography image revealed that the lesion was non-fdg avid. It was verified later that this patient had paraaortic lymph node metastasis. Clinical and imaging follow-up revealed no evidence of bone metastasis. SUVmax value of the primary tumor was 11.5. Hematogenous bone metastases were identified in 9% of patients who had an SUVmax of the primary tumor <11.5 and in 6% of the remaining patients. The factors described above were not associated significantly with hematogenous bone metastasis in either the univariate or multivariate analyses. The percentage of patients with hematogenous bone metastasis was 0% for FIGO stage I, 6% for stage II, 8% for stage III, and 33% for stage IV. It was 0%, 2%, 6%, 11%, and 55% for patients who had negative, pelvic, para-aortic or inguinal, supraclavicular, and mediastinal lymph node spread, respectively. Both FIGO stage and the extent of lymph node metastases were associated with hematogenous bone metastasis in univariate analysis. However, only the extent of lymph node metastases was associated significantly with hematogenous bone metastasis in the multivariate model (P ¼.025). DISCUSSION 18 F-FDG PET is a useful diagnostic tool with which to detect distant lymph node metastases in patients with cervical cancer. 16,17 18 F-FDG PET also occasionally detects unsuspected distant metastatic sites other than lymph nodes. Thus, 18 F-FDG PET is considered a promising method for the primary staging of advanced cervical cancer. For recurrent cervical cancer, 18 F-FDG PET is useful for restaging and, thus, allows the determination of optimal treatment. 18 18 F-FDG PET has been considered useful for detecting active bone metastasis in breast, lung, and head and neck cancers. 6-8,19,20 In cervical cancer, 18 F-FDG PET or PET/CT has detected unsuspected bone metastasis. 21,22 However, to our knowledge, no large-scale study has been performed to compare the performances of 18 F- FDG PET and CT or MR imaging in patients with cervical cancer. The objective of the current study was to clarify this issue and included a large number of patients with cervical cancer who received both PET and CT or MR imaging in a single institution. In previous studies, bone metastasis in patients with cervical cancer was associated with advanced disease. 3,4 FIGO staging still is the most frequently used clinical parameter for prognostic stratification. However, it was demonstrated in a pooled study of the Gynecologic Oncology Group 23 that lymph node status is a more significant prognosticator of survival for patients with invasive cervical cancer. In another study of 44 patients, the distant metastasis-free rate differed significantly among lymph node-negative, possibly lymph node-positive, and probably lymph node-positive groups. 24 In a previous study from our group, both positive pelvic lymph nodes and FIGO stage at primary staging were prognostic for distant lymph node or systemic failure. 2 In our experience, distant bone metastasis at primary staging has not been observed in patients who have FIGO stage I/II cervical cancer and are negative for lymph node metastasis. This is why we included only patients with FIGO stage III/IV disease or positive lymph node metastasis in the primary setting. It is worth noting that we wondered whether the presence of bone metastasis would be associated with the extent of lymph node metastases rather than with FIGO stage. If this is true, then the importance of lymph node status will be corroborated. Two mechanisms for osseous metastasis in cervical cancer have been described. 4 Direct invasion into bone from locoregional or distant soft tissue tumor masses is one mechanism, and hematogenous spread through the systemic circulation or the Batson venous plexus is another. The periosteum and bony cortex are involved first through direct invasion. The bone marrow is involved first in patients with hematogenous metastasis. Cancer December 1, 2009 5477

Table 3. Analysis of Risk Factors for Hematogenous Bone Metastasis at Primary Staging No. of Patients P Characteristic BM Positive BM Negative Univariate: Exact Chi-Square Multivariate: Logistic Regression Total no. of patients 12 153 Age, y.177.187 <45 2 43 45-65 6 89 >65 4 21 Tumor cell type 1.0.968 Squamous cell carcinoma 9 117 Adenocarcinoma 1 13 Adenosquamous carcinoma 1 15 Neuroendocrine tumor 1 8 Tumor differentiation.706.135 Well differentiated 0 7 Moderately differentiated 6 62 Poorly differentiated 6 84 SUVmax of primary tumor.565.347 <11.5 7 75 11.5 5 78 FIGO stage.003.383 I 0 38 II 4 61 III 4 46 IV 4 8 Extent of lymph node metastases <.001.025 Negative 0 7 Pelvic 2 88 Para-aortic or inguinal 3 45 Supraclavicular 1 8 Mediastinal 6 5 BM indicates hematogenous bone metastasis; SUVmax, maximal standardized uptake value; FIGO, International Federation of Gynecology and Obstetrics. Because these 2 types of osseous metastases are distinct in mechanism and may originate from tumors with different characteristics, the current analysis focused on hematogenous bone metastases. Combining the imaging studies from the time of primary staging and from the time patients were evaluated for suspected disease recurrence, the diagnostic performance of 18 F-FDG PET was superior to the performance of both CT and MR imaging. In our experience, it is difficult for CT to detect early bone or bone marrow metastasis. Mild bone or bone marrow abnormalities on CT are not specific for metastasis and usually are neglected by interpreters. The sensitivity of MR imaging for detecting early bone marrow metastasis is better than that of CT. However, the limited imaging field and the large amount of soft tissue information may impair the interpretative accuracy of MR imaging. In addition, MR imaging may yield false-positive findings in patients who have bone changes after treatment, such as radiotherapy. For 18 F- FDG PET, the higher metabolic contrast between metastatic lesions and background and the large imaging field from the skull base to the upper thigh may result in its higher diagnostic performance. Benign bone changes, such as osteonecrosis or avascular necrosis, may occur after radiotherapy and can account for the lower specificity of MR imaging. Osteonecrosis, particularly in the acute-subacute setting, may be misdiagnosed as an osseous metastasis on MR imaging. In the current study, many false-positive bone lesions detected by MR had no obvious FDG avidity on PET 5478 Cancer December 1, 2009

Detecting Bone Metastasis in CC/Liu et al imaging. This may explain the lower positive predictive value of MR for suspected recurrent disease. According to our statistical analyses, the current study indicated the hematogenous bone metastasis have a stronger association with the extent of lymph node metastases than with FIGO stage. The FIGO staging system has been adopted globally because it does not require expensive examinations. However, the current results corroborate the importance of lymph node status in the management of invasive cervical cancer. Greater than 50% of the patients who had mediastinal lymph node metastasis had concomitant hematogenous bone metastasis. For other risk factors, we observed a mild tendency for older patients to have a higher incidence of bone metastasis at presentation; this may be related to their reluctance and delay in seeking medical help. It also is noteworthy that none of the 7 patients who had well differentiated tumors had hematogenous bone metastasis. With regard to the SUVmax of the primary tumor, it was not associated with hematogenous bone metastasis. Whether it is associated with local recurrence is an interesting topic and deserves further study. Although recurrent cervical cancer was traditionally considered to have a poor prognosis, recent studies and experiences suggest that a significant portion of patients with confined para-aortic or supraclavicular lymph node metastasis can achieve long-term survival after combined-modality treatment, such as concurrent chemoradiotherapy. 2,25-27 Because hematogenous bone metastasis is associated strongly with the extent of lymph node metastases, early identification and proper management of lymph node metastasis at the time of primary staging or disease recurrence can be an important step toward reducing distant hematogenous failure and, thus, can be beneficial for patients with invasive cervical cancer. The current study has the inherent limitations associated with retrospective studies. Some limitations can be remedied in part by the large number of eligible patients and the standard management protocol for patients with cervical cancer in our institution. However, some selection biases still cannot be avoided. Another limitation of the current study is that not all patients who were determined to have bone metastasis had histologic proof. Even considered as a gold standard, bone biopsy still has a high rate of sampling error and is inappropriate ethically for patients who have distant visceral metastasis. We have tried our best to establish the presence of bone metastasis by correlating different imaging modalities with adequate follow-up. Ever more sophisticated whole-body imaging technologies are being used for oncologic staging, including combined PET/CT and whole-body MR imaging. Prospective studies concerning the roles of these newer imaging modalities in the management of cervical cancer patients are warranted. In conclusion, the current study demonstrated the relative superiority of 18 F-FDG PET over CT and MR imaging for detecting hematogenous bone metastasis in patients with advanced cervical cancer. Hematogenous bone metastasis in cervical cancer is associated with the extent of lymph node metastases rather than the FIGO stage. Early identification and proper management of lymph node metastasis may be important for decreasing the rate of distant hematogenous recurrence in patients with invasive cervical cancer. Conflict of Interest Disclosures The authors made no disclosures. References 1. Wang CJ, Lai CH, Huang HJ, et al. Recurrent cervical carcinoma after primary radical surgery. Am J Obstet Gynecol. 1999;181:518-524. 2. Hong JH, Tsai CS, Lai CH, et al. Recurrent squamous cell carcinoma of cervix after definitive radiotherapy. Int J Radiat Oncol Biol Phys. 2004;60:249-257. 3. Fagundes H, Perez CA, Grigsby PW, Lockett MA. Distant metastases after irradiation alone in carcinoma of the uterine cervix. Int J Radiat Oncol Biol Phys. 1992;24:197-204. 4. Ratanatharathorn V, Powers WE, Steverson N, Han I, Ahmad K, Grimm J. Bone metastasis from cervical cancer. Cancer. 1994;73:2372-2379. 5. Aapro M, Abrahamsson PA, Body JJ, et al. Guidance on the use of bisphosphonates in solid tumours: recommendations of an international expert panel. Ann Oncol. 2008; 19:420-432. 6. Hamaoka T, Madewell JE, Podoloff DA, et al. Bone imaging in metastatic breast cancer. J Clin Oncol. 2004;22: 2942-2953. 7. Cheran SK, Herndon JE II, Patz EF Jr. Comparison of whole-body FDG-PET to bone scan for detection of bone metastases in patients with a new diagnosis of lung cancer. Lung Cancer. 2004;44:317-325. Cancer December 1, 2009 5479

8. Liu FY, Chang JT, Wang HM, et al. [ 18 F]fluorodeoxyglucose positron emission tomography is more sensitive than skeletal scintigraphy for detecting bone metastasis in endemic nasopharyngeal carcinoma at initial staging. J Clin Oncol. 2006;24:599-604. 9. Du Y, Cullum I, Illidge TM, et al. Fusion of metabolic function and morphology: sequential [ 18 F]fluorodeoxyglucose positron-emission tomography/computed tomography studies yield new insights into the natural history of bone metastases in breast cancer. J Clin Oncol. 2007;25:3440-3447. 10. Even-Sapir E. Imaging of malignant bone involvement by morphologic, scintigraphic, and hybrid modalities. J Nucl Med. 2005;46:1356-1367. 11. Ghanem N, Uhl M, Brink I, et al. Diagnostic value of MRI in comparison to scintigraphy, PET, MS-CT and PET/CT for the detection of metastases of bone. Eur J Radiol. 2005;55:41-55. 12. Matsuyama T, Tsukamoto N, Imachi M, Nakano H. Bone metastasis from cervix cancer. Gynecol Oncol. 1989;32:72-75. 13. Yen TC, Ng KK, Ma SY, et al. Value of dual-phase 2-fluoro-2-deoxy-D-glucose positron emission tomography in cervical cancer. J Clin Oncol. 2003;21:3651-3658. 14. Leisenring W, Alonzo T, Pepe MS. Comparisons of predictive values of binary medical diagnostic tests for paired designs. Biometrics. 2000;56:345-351. 15. Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology. 1983;148:839-843. 16. Grigsby PW, Siegel BA, Dehdashti F. Lymph node staging by positron emission tomography in patients with carcinoma of the cervix. J Clin Oncol. 2001;19:3745-3749. 17. Lai CH, Yen TC, Chang TC. Positron emission tomography imaging for gynecologic malignancy. Curr Opin Obstet Gynecol. 2007;19:37-41. 18. Lai CH, Huang KG, See LC, et al. Restaging of recurrent cervical carcinoma with dual-phase [ 18 F]fluoro-2-deoxy-Dglucose positron emission tomography. Cancer. 2004;100: 544-552. 19. Basu D, Siegel BA, McDonald DJ, et al. Detection of occult bone metastases from head and neck squamous cell carcinoma: impact of positron emission tomography computed tomography with fluorodeoxyglucose F18. Arch Otolaryngol Head Neck Surg. 2007;133:801-805. 20. Schirrmeister H. Detection of bone metastases in breast cancer by positron emission tomography. Radiol Clin North Am. 2007;45:669-676. 21. Belhocine T, Thille A, Fridman V, et al. Contribution of whole-body 18 FDG PET imaging in the management of cervical cancer. Gynecol Oncol. 2002;87:90-97. 22. Loft A, Berthelsen AK, Roed H, et al. The diagnostic value of PET/CT scanning in patients with cervical cancer: a prospective study. Gynecol Oncol. 2007;106:29-34. 23. Stehman FB, Bundy BN, DiSaia PJ, Keys HM, Larson JE, Fowler WC. Carcinoma of the cervix treated with radiation therapy. I. A multi-variate analysis of prognostic variables in the Gynecologic Oncology Group. Cancer. 1991; 67:2776-2785. 24. Ohara K, Tanaka YO, Tsunoda H, et al. Nonoperative assessment of nodal status for locally advanced cervical squamous cell carcinoma treated by radiotherapy with regard to patterns of treatment failure. Int J Radiat Oncol Biol Phys. 2003;55:354-361. 25. Sakurai H, Mitsuhashi N, Takahashi M, et al. Analysis of recurrence of squamous cell carcinoma of the uterine cervix after definitive radiation therapy alone: patterns of recurrence, latent periods, and prognosis. Int J Radiat Oncol Biol Phys. 2001;50:1136-1144. 26. Chou HH, Wang CC, Lai CH, et al. Isolated paraaortic lymph node recurrence after definitive irradiation for cervical carcinoma. Int J Radiat Oncol Biol Phys. 2001;51:442-448. 27. Qiu JT, Ho KC, Lai CH, et al. Supraclavicular lymph node metastases in cervical cancer. Eur J Gynaecol Oncol. 2007;28:33-38. 5480 Cancer December 1, 2009