Defining the Priority of Using ¹⁸F-FDG PET for Recurrent Cervical Cancer

Patients

Patients enrolled were from 2 independent prospective studies investigating the role of ¹⁸F-FDG PET in cervical cancer patients with documented failure after definitive treatment (CTRP-018) or unexplained elevated tumor marker squamous cell carcinoma antigen (SCC-Ag) or carcinoembryonic antigen (CEA) serum levels (CTRP-016) and proven relapse after PET. These 2 trials were approved by our institutional review board. Informed written consent was obtained from each patient before entry in the CTRP-016 or CTRP-018 study. The common eligibility criteria were (a) completion of definitive radiotherapy (RT) or surgery; (b) without contraindications to and willing to undergo contrast-enhanced CT/MRI and PET scans; (c) willing to receive an image-guided biopsy or surgical exploration if indicated; and (d) potentially curable and willing to receive curative salvage therapy if restaging confirmed the feasibility. Biopsy-documented recurrent or persistent cervical carcinoma was required to enter the CTRP-018 study. For the CTRP-016 study, an additional criterion was either a serum level of SCC-Ag > 2 ng/mL or CEA > 10 ng/mL on 2 successive tests 1 mo apart, but no definite recurrence was identified by conventional imaging studies (including CT/MRI). Patients in the CTRP-016 study were enrolled after proven relapse. Though elevated serum CEA was removed from the inclusion criteria in the amendment of the CTRP-016 project to homogenize the patient characteristics of that trial, we did not exclude those subjects in this analysis.

The common ineligibility criteria were if (a) the patient had received salvage therapy for previous recurrence, (b) the patient was medically or psychologically unfit to receive curative salvage therapy, and (c) the patient had a history of another malignancy excluding basal cell carcinoma of the skin. For CTRP-018, additional exclusion criteria were (a) only a superficial lesion on the cervix or vaginal cuff, (b) disseminated abdominal or pleural lesions with positive fluid cytology, (c) >2 regions involved or a previously irradiated recurrent or persistent pelvic tumor with extrapelvic metastasis.

“Potentially curable” were defined as (a) the proven recurrent disease was confined to the pelvis without previous primary or adjuvant RT; (b) if RT had been given before relapse, the relapse should be limited to the central pelvis without pelvic side wall or extrapelvic involvement; (c) distant recurrences at a sole site that could be completely resected or encompassed by a curative RT procedure, such as paraaortic, supraclavicular, or inguinal lymph nodes (LNs); or pulmonary metastasis without spread to mediastinal LNs, solitary or multiple (≤4 nodules) but confined to 1 lobe (7,23). Every effort was made to differentiate recurrent cervical cancer from new primary malignancy.

CT and MRI

All CT and MR images were obtained with a spiral CT scanner (Hi-Speed; General Electric Medical Systems) or a multislice CT scanner (Somatom Plus 4, Volume Zoom, Version A40; Siemens AG Medical) and a 1.5-T scanner (either a Magnetom Vision or Magnetom Expert; Siemens Medical Systems) with the method described in our previous study (21).

PET

The FDG used in the PET scans was produced by the Institute of Nuclear Energy Research of Taiwan and the images were obtained with an ECAT EXACT HR+ PET camera (CTI), using a full width at half maximum of 4.5 mm and 15-cm transaxial field of view. The preparation of each patient and data acquisition were according to our previous study (21). Transmission scans were obtained with ⁶⁸Ge rod sources. Reconstruction of both transmission and emission scans used accelerated maximum-likelihood reconstruction and ordered-subset expectation maximization, which reduce image noise and avoid reconstruction artifacts, resulting from filtered backprojection reconstruction of data with low count densities.

Image Analysis

Three experienced nuclear physicians interpreted the PET data with consensus between at least 2. Any foci of abnormally increased ¹⁸F-FDG uptake were evaluated using transaxial, sagittal, and coronal displays and compared between early and delayed scans. ¹⁸F-FDG accumulation was classified on the basis of visual scores using 5 grades: 0 = normal, 1 = probably normal, 2 = equivocal, 3 = probably abnormal, and 4 = definitely abnormal. The official PET report was primarily interpreted on the basis of visual analysis. The standardized uptake value was used as an additional criterion (21). Images of CT/MRI scans were analyzed by an experienced radiologist 3 times with consensus at least twice. The 5-grade scoring criteria were also used for interpretation of the CT/MR image results: grade 0, a completely normal finding; grade 1, visible nodes with size < 0.5 cm and considered as reactive; grade 2, any single visible node was between 0.5 and 1 cm in length and considered as equivocal for malignancy; grade 3, any LNs > 1 cm in its short axis or clustered nodes (n ≥ 3) with sizes between 0.5 and 1 cm; and grade 4, confluent nodes with central necrosis or an irregular contour (21). Image readings were designated positive with grades 3 and 4, and negative with grades 0, 1, and 2. Disease status was determined from pathologic findings or clinical follow-up.

Study Procedures and Determination of Lesion Status

Abdominal and pelvic CT/MRI were performed on all patients before enrollment. PET and CT/MRI were performed within a period of 2 wk. For CTRP-016, a neck and chest CT scan was done in those patients with suspected lung metastases or supraclavicular node metastases clinically or after PET. For CTRP-018, all patients had neck and chest CT scans. Image fusion of CT/MRI and PET with a commercially available software program (Hermes; Nuclear Diagnostics AB) was applied in an abnormally elevated region of ¹⁸F-FDG uptake or discrepant results (24). We examined the suspicious (grade ≥ 3) lesions by a CT- or ultrasound-guided biopsy to confirm lesions demonstrated by coregistered images, if feasible. If a distal site LN metastasis was confirmed histologically, we considered all proximal adenopathy with abnormal uptake (grade ≥ 3) as positive based on the fact that LN metastasis usually follows in a sequential fashion. A biopsy was attempted whenever discrepant findings existed between the CT/MRI and PET scans. If biopsy of the lesion of interest was not feasible, or yielded a negative result, follow-up with CT/MRI or PET was performed 3–6 mo later to assess the interval status of the lesion and to avoid a false-negative (FN) biopsy result.

Statistical Analysis

We prospectively recorded the following data: birth date, date of initial diagnosis and recurrence, initial FIGO staging, initial tumor grade, initial tumor-node-metastasis status, number and type of prior treatment(s) (surgery, radiation, and chemotherapy), serum tumor marker levels (SCC-Ag and CEA), patient’s symptoms and signs at the time of enrollment, results of all biopsies performed, the presence of concurrent infection or inflammation at the time of the PET scan, the official CT/MRI and PET reports, the highest standardized uptake values (SUVs) reported from PET scans for each positive site (local, regional, or distant), evidence (histologic, radiographic, or clinical) of recurrent disease after salvage therapy, site and date of recurrent disease (local, regional, or distant), date of recurrence, subsequent treatment (palliative or curative, modality and field of RT) after recurrence, and status (alive or dead, relapsed or relapse-free) at last follow-up.

Sensitivity, specificity, and accuracy were calculated for PET and CT/MRI results. The area under the receiver-operating characteristic (ROC) curve was calculated for evaluating diagnostic efficacy. The area under the curve (AUC) comparison between PET and CT/MRI was made using the method of Metz (25). Kaplan–Meier estimates and log-rank tests were used to examine how overall survival (OS) varied with prognostic variables at initial diagnosis and recurrence by univariate analysis (26). A Cox proportional hazards model with forward selection was made to identify independent risk factors and to estimate the adjusted hazard ratio (HR) of death (26). A scoring system using these significant covariates was formulated to define distinct prognostic groups. All P values were 2-sided and the significance level was 0.05.

Initial and Recurrent Characteristics

Between February 1, 2001, and January 31, 2003, 55 patients were eligible. Forty patients from CTRP-018 were deemed eligible. Twenty-seven patients from CTRP-016 were assessed, in which 15 had documented relapse after PET and were enrolled for this study. Of these 15 enrolled CTRP-016 patients, 12 had unexplained serum elevation of SCC-Ag and 3 had elevated CEA. The other 12 CTRP-016 patients were excluded due to lack of evidence of lesion presence after tissue studies, image examination, and clinical follow-up. The median age at first recurrence in these 55 patients was 51 y (range, 25–86 y). They were monitored to July 31, 2003. The median follow-up from recurrence was 16 mo (range, 8–28 mo). The interval between the initial diagnosis of cervical cancer and the first documented recurrence was 4–276 mo (mean ± SD, 33 ± 50 mo). Their initial FIGO stages were IB or IIA in 26 patients and IIB to IVA in 29 patients. Twenty had primary radical surgery and 35 received primary RT. Histologic examination revealed squamous cell carcinoma in 44 patients and adenocarcinoma or adenosquamous carcinoma in 11 patients.

Diagnostic Efficacy of PET

Ten sites, including peritoneum, bone, liver/spleen, lung, mediastinal LN, supraclavicular LN, paraaortic LN, pelvic LN, inguinal LN, and central or parametrial persistent or recurrent tumors were evaluated for each patient. Therefore, a total of 550 regions of interest (ROIs) from PET and CT/MRI were studied in the 55 patients (Figs. 1 and 2). Either PET or CT/MRI recognized 104 ROIs as positive (grade ≥ 3), and their final diagnoses were 89 malignant and 15 benign. Of the 550 ROIs, 441 were considered as true-negative (TN). Of these 441 TN ROIs, 9 were proven by biopsy, whereas the remaining 432 ROIs were considered to be TN without biopsy because both PET and CT/MRI scans were negative and the patients remained free of disease for at least 6 mo after clinical and imaging follow-up. Five FN results (proven by positive CT-guided biopsy or clinical imaging follow-up) were found in both CT/MRI and PET. Table 1 shows comparisons of the 550 ROIs of ¹⁸F-FDG PET scans and MRI/CT in all patients studied. In all, sensitivity (TP = true-positive) was significantly higher with PET than with CT/MRI for metastatic lesion detection (89.2% vs. 39.2%; P < 0.0001), but there was no significance for local lesion detection (90.0% vs. 80.0%; P = 0.472). The ROC curve indicated that PET was superior to CT/MRI for all lesions (AUC: 0.968 vs. 0.702; P < 0.0001) and metastatic lesions (other than local recurrence or persistent tumor) (AUC: 0.971 vs. 0.684; P < 0.0001), though a marginal effect was observed in local recurrence or persistent tumor (AUC: 0.956 vs. 0.922; P = 0.053).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1.

A 25-y-old woman with poorly differentiated squamous cell carcinoma of uterine cervix, FIGO stage IIIb, underwent concurrent chemoradiation therapy. Three months later, a left neck mass was palpated. Abdominopelvic MRI and chest CT showed no definite abnormal findings except an enlarged supraclavicular LN. Balancing between salvage RT and palliation treatment, PET was performed and suggested nodal metastases at the left supraclavicular, the bilateral upper and lower paraaortic, and the bilateral pelvic regions. After the left supraclavicular and paraaortic nodal metastases were confirmed histopathologically, she received palliation treatment.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2.

A 52-y-old woman with poorly differentiated squamous cell carcinoma of uterine cervix, FIGO stage IIa, underwent concurrent chemoradiation therapy. Four months after complement of treatment, an elevated serum SCC-Ag of 2.23 ng/mL was noted. Abdominopelvic MRI and chest RT showed negative findings. Two months later, her serum SCC-Ag level was 7.36 ng/mL. Conventional images still showed negative findings. A PET scan was then obtained and disclosed a metastatic lesion in the right lower lung. She subsequently received pneumonectomy and was well for 1 y.

Modification of Treatment Due to PET, OS, and Prognostic Factors

Of the 55 study patients who were considered potentially curable for salvage therapy at enrollment, 36 (65.5%) patients had treatment plan modification, and the remaining 19 patients were treated according to the initial plan. Among the 36 patients, 25.0% (n = 9) had treatment remaining curative in intent, yet the modality or field of radiation changed and 75.0% (n = 27) received palliative therapy. Three of the 9 patients (33%) had their treatment modified but still with a curative intent were down staged (CT/MRI grade 3 or 4, with biopsy negative or unfeasible and follow-up disease free).

Up to July 31, 2003, 18 patients had died with disease, 22 were alive with disease, and 15 patients were alive and free of disease. The median survival of the study patients (n = 55) was 19.4 mo. The 6-mo, 1-y, and 2-y OS rates were 92%, 70%, and 29%, respectively. SCC-Ag levels at recurrence were categorized into 2 groups: ≤4 and >4 ng/mL. The serum SCC-Ag cutoff of 4 ng/mL was determined according to the log-rank test with the lowest P value. Univariate analyses showed covariates such as SCC-Ag ≥ 4 ng/mL at recurrence (P = 0.0041) and primary treatment (P = 0.0081) were significantly associated with survival. The presence of symptoms at recurrence (P = 0.0536), pelvic node metastasis before primary treatment (P = 0.0585), and the type of salvage treatment (P = 0.0642) was marginally significant. There has been no death (0/9) among those who remained with a curative intent yet had modified treatment compared with those with a curative intent and no change (6/19); however, the difference was not significant. Other covariates such as initial FIGO stage, histology type, grade of differentiation, or pattern of recurrence did not significantly influence OS by the log-rank test (Table 2). The Cox proportional hazard model using multivariate analysis for forward selection identified SCC-Ag serum levels at recurrence (HR = 5.82; 95% confidence interval [CI] = 1.53–22.04; P = 0.010), primary treatment (HR = 14.62; 95% CI = 2.74–77.92; P = 0.002), and the presence of symptoms (HR = 6.24; 95% CI = 1.99–19.61; P = 0.002) as independent prognostic factors (Table 3).

Prognostic Scoring System

A prognostic scoring system was formulated by summing up the 3 significant covariates: 0 for SCC-Ag of ≤4 ng/mL, 1 for >4 ng/mL; 0 for the absence of symptoms and 1 for the presence of symptoms; 0 for primary treatment with radical surgery and 1 for primary treatment with RT (Table 4). The probability of dying in 2 y was the highest in the group with a score of 3 (high risk) (HR = 60.46; 95% CI = 9.68–378.09); this was followed by the group with a score of 2 (intermediate risk) (HR = 6.91; 95% CI = 1.49–32.14) compared with the group with a score of ≤1 (low risk) (P < 0.0001) (Fig. 3).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 3.

Kaplan–Meier curves for 2-y OS rates in patients with recurrent cervical cancer. Patients are categorized by risk score of ≤1 (bold solid line), 2 (thin solid line), and 3 (dashed line) (P < 0.0001).

In the low-risk group (n = 27), 10 patients were shifted to palliation and the other 17 remained with curative salvage therapy (7 with plan changed, 10 unchanged). Furthermore, in the intermediate-risk group (n = 19), 12 patients were switched to palliative treatment, whereas the other 7 patients stayed on curative treatment (2 with plan changed, 5 unchanged). Each of the 2 patients, who had changed treatment planning but still remained with a curative intent after PET, was alive at the time of analysis. Among the 6 patients in the high-risk group, 5 (83.3%) have changed treatment planning to palliation after PET, and 4 of these 5 (80%) were dead. The remaining 1 patient on curative therapy still died of disease (Table 4).