ABSTRACT

Objective:

This study aimed to evaluate the diagnostic accuracy of preoperative ¹⁸F-fluoro-deoxy- glucose (FDG) positron emission tomography/computed tomography (PET/CT) metabolic parameters for the prediction of risk factors and detection of lymph node metastasis (LNM) in patients with endometrial cancer.

Material and Methods:

This study included 26 patients with endometrioid carcinoma who underwent preoperative PET/CT and treated with adjuvant local radiotherapy. The maximum standard uptake value of the tumor (SUV_max-T), SUV_max of the pelvic and/or para-aortic LNs, metabolic tumor volume (MTV), and tumor lesion glycolysis (TLG) with cut-off values of 30-40% were calculated. International Federation of Gynecology and Obstetrics stages 3 and 4, high-grade disease, lymphovascular invasion (LVI), cervical involvement (CI), and myometrial invasion (MI) ≥50% were established as high-risk features. Disease-free survival and overall survival were analyzed in comparison with ¹⁸F-FDG PET/CT parameters.

Results:

SUV_max-T was only associated with tumor diameter (p=0.01). It was not correlated with MI, high-grade disease, CI, or LNM. With SUV_maxP ≥2.81 as a cut-off value, the sensitivity, specificity, and accuracy in the detection of LNM were high (90%, 83.3%, 71.4%, respectively). For LNM, the mean MTV-30 (p=0.021), TLG-30 (p=0.030), and SUV_max-P (p=0.009) were significant predictors. According to the regression analysis, MTV-40 (p=0.043) was an independent predictor of LNM, and LVI (p=0.037) was the only significant predictor of MI. MTV-30 was a significant predictor of CI (p=0.04).

Conclusion:

SUV_max-P, MTV, and TLG cut-off values, to predict LN metastases, increase diagnostic accuracy for EC.

What is known on this subject?

The known is that pre-operative ¹⁸F-fluoro-deoxy-glucose (¹⁸F-FDG) positron emission tomography/computed tomography (PET/CT) can be used for staging in endometrial cancer patients.

What this study adds?

¹⁸F-FDG PET/CT is not standart diagnostic image in endometrial cancer (EC). According to our results, we found out that the metabolic parameters on ¹⁸F-FDG PET/CT, for prediction of lymph node metastases increase diagnostic accuracy for EC.

Introduction

Endometrial cancer (EC) is the most common gynecologic malignancy in developed countries, with adenocarcinoma as the most common histologic type (1). The majority of patients with EC are diagnosed at an early stage with the disease confined to the primary site (67%). However, the spread to regional organs and lymph nodes (LNs) (21%) and distant metastases (8%) are less frequent (2). Although EC staging is performed with surgery, the identification of disease extent before surgery is very important for treatment planning.

Imaging modalities play an important role for staging and treatment planning of patients with EC. ¹⁸F-fluoro-deoxy-glucose (FDG) positron emission tomography/computed tomography (PET/CT) combines morphology with physiology and is the preferred imaging modality, especially in clinical oncology. Its accuracy of staging and determination of the aggressiveness of EC have also been investigated (3,4). The maximum standardized uptake value (SUV_max) of the tumor, the most widely used PET parameter, was considered an important indicator that reflects tumor aggressiveness, such as myometrial invasion (MI), cervical involvement (CI), LN metastases (LNM), and high-risk disease in EC (4,5). The metabolic tumor volume (MTV) and total lesion glycolysis (TLG) were reported to have prognostic effect on several cancers, including cervical, ovarian, and lung cancer; however, data regarding EC are limited (6,7,8).

In this study, we aimed to evaluate the prognostic importance and diagnostic accuracy of preoperative ¹⁸F-FDG PET/CT metabolic parameters of the local tumor and pelvic and/or para-aortic LN SUV_max of patients with EC treated with intracavitary brachytherapy (ICRT) and/or pelvic external beam radiation therapy (EBRT) postoperatively and to examine the correlation of results with histopathology.

Material and Methods

Patients

The study was approved as a retrospective study by the İstanbul Training and Research Hospital Clinical Research Ethics Committee (decision no: 1447, date: 28.09.2018), and the requirement to obtain informed written consent was abandoned. A total of 90 patients with histopathologically verified EC treated with three-dimensional high dynamic range (3D HDR) ICRT and/or pelvic EBRT at a single center between August 2016 and October 2019 were analyzed. Of those, 26 patients who underwent preoperative ¹⁸F-FDG PET/CT were included in this study. Patients with previous or concurrent diagnosis of any other primary malignancy, patients with follow-up duration <6 months, patients without pretreatment ¹⁸F-FDG PET/CT, and patients without adequate surgical staging (total abdominal hysterectomy and bilateral salpingoopherectomy, pelvic- para-aortic LN dissection) were excluded from the study.

Patients with EC were surgically staged according to the 2009 International Federation of Gynecology and Obstetrics (FIGO) (9). Tumor histopathologic types were classified as endometrioid endometrial carcinoma (grades 1, 2 and 3), serous carcinoma, mixed-type endometrial carcinoma, and carcinosarcoma. The estimated 3-year NFS, DFS, and overall survival rates were 84.2%, 86.1%, and 87.5%, respectively.

¹⁸F-FDG PET/CT Image Acquisition and Analysis

All ¹⁸F-FDG PET/CT records were retrospectively analyzed by the investigators without knowledge of patients’ clinical and histopathological information. Imaging of patients who fasted for at least 6 h before intravenous administration of 5-6 MBq/kg ¹⁸F-FDG and whose blood glucose concentrations were <180 mg/dL was perfromed using an integrated PET/CT system. Combined image acquisition began approximately 60 min after ¹⁸F-FDG injection from the vertex to the mid-thigh. Sagittal, coronal, and transaxial images and fused images were analyzed on workstation (Syngo.via Siemens Molecular Imaging).

Qualitative and quantitative (or semi-quantitative) image analyses were performed by an experienced nuclear medicine physician (B.Y.) with significant experience in reading ¹⁸F-FDG PET/CT scans (average 140 reads/month individually). Pretreatment FDG uptake in both local tumor and LNs was quantitatively assessed using SUV_max. For each FDG PET/CT study, SUV_max values of the most FDG-avid pelvic and/or para-aortic and inter-aortocaval LNs were measured.

The volume of interest (VOI) was defined over the primary tumoral lesion. The tumor contours were semi-automatically delineated by using thresholds of 30% and 40% of the SUV_max within the lesion to calculate MTV (7,8). MTV values were used to calculate TLG by multiplying the mean SUV within the VOI both for 30% and 40% thresholds. In the pretreatment PET/CT for the primary tumor area, SUV_max-of the tumor (T), MTV-30, MTV-40, TLG-30, and TLG-40; for pelvic LNs SUV_max-P, for para-aortic LNs SUV_max-PA, for interaortocaval LNs, SUV_max-invasive adenocarcinoma (IAC) were recorded (Figure 1). In addition, any suspicious distant metastatic site was noted and verified by other imaging modalities.

Treatment and Follow-up

3D HDR ICRT was delivered once a week in three or five fractions, and D90 ≥5.5 Gy or 7 Gy was prescribed for the planning target volume (PTV) in all patients. The PTV was defined as the upper 1/3 and 5 mm deep of the vagina using a cylinder applicator on the same-day CT scan and a new plan in each brachytherapy fraction. Twelve (46.2%) patients received pelvic intensity modulated radiation therapy (IMRT) technique with Rapid Arc in 1.8 Gy daily fractions, five times a week, for a median total dose of 45 Gy (range, 45-50.4). Para-aortic radiation was delivered to cases with para-aortic LN histopathological involvement (n=3), with a dose up to 45 Gy with one isocenter field in field IMRT. Adjuvant chemotherapy (cisplatin and paxlitaxel, 4-6 cycles) was only administered to patients with high-risk EC (30.8%; n=8).

During follow-up patients underwent bimanual pelvic examination and speculum examination. Serum CA-125 levels were measured and imaging studies were performed every 3 months for 2 years, every 6 months from 2 to 5 years, and annually thereafter.

Statistical Analysis

 All statistical analyses were performed using SPSS software (version 15.0; SPSS Inc.), with p<0.05 considered significant. Descriptive data are expressed as mean ± standard deviation and percentages. Student’s t-test was used to compare the mean values between two independent groups, and the chi-square test was used to compare nominal values between two groups. The metabolic parameters among the groups were compared using the Mann-Whitney U test. Correlations among the PET parameters were analyzed using Spearman rank correlation analysis. With respect to SUV_max, MTV, and TLG, receiver-operating characteristic (ROC) curve analysis was performed to determine the cut-off values for predicting LNM and clinicopathologic characteristics. The optimal cut-off values of SUV_max, MTV, and TLG were those giving the highest sensitivity and specificity.

The sensitivity, specificity, and area under the curve (AUC) values of the ¹⁸F-FDG PET/CT were also calculated. Multivariate logistic regression analysis was performed to determine the independent variables associated with LNM and clinicopathologic characteristics by including all significant factors (p<0.25) from the univariate analysis.

Results

Patient Characteristics

Patient’s clinicopathological findings were summarized according to risk stratification (Table 1). The median age was 63 (range, 45-84) years, and the median follow-up time was 22 (range, 9-36) months. Sixteen patients had FIGO stage I disease, and 22 patients were found to have endometrioid histology. While 14 patients had intermediate-risk EC, 12 patients had high-risk features. Moreover, 26 patients had pelvic LN dissection, and 15 patients underwent further para-aortic LN dissection, of which 888 LNs were retrieved. Six (23.1%) patients had LNM on pathologic examination (Figure 2).

Correlation of Preoperative ¹⁸F-FDG PET/CT Metabolic Parameters with Clinicopathological Factors

According to the presence of LNM, metabolic parameters of PET and clinicopathological findings are shown in Table 2. The mean TLG-30, TLG-40, and SUV_max-T were significantly higher in patients with tumor diameter ≥2.5 cm (p<0.01), and only TLG-40 was significantly related with high-grade EC (p=0.045). The mean TLG-30, TLG-40, MTV-30, and MTV-40 of the local tumor were significantly higher in patients with locally advanced disease (p<0.03) (Table 3). Meanwhile, the mean SUV_max-P for LNM was significantly higher in the node-positive group than in the node-negative group (p=0.009). By contrast, SUV_max-T could not predict pelvic and/or para-aortic LNM. Moreover, no significant difference was found for SUV_max-T between endometrioid and non-endometrioid subtypes, with mean SUV_max of 13.02 and 13.68, respectively (p>0.05). Besides, the mean SUV_max-PA and SUV_max-IAC were not higher in patients with para-aortic LN metastases (n=3; p<0.05).

Cut-off Values of PET Parameters for Predicting Risk Factors

The ROC curve for SUV_max-P for discriminating LNM is shown in Figure 3 (AUC 0.900; 95% confidence interval (CI) 0.729-1.000; p=0.003). Using 2.81 as a cut-off value of SUV_max-P, the specificity, accuracy, sensitivity, positive predictive value, and negative predictive value of ¹⁸F-FDG PET/CT in the detection of LNM in all patients (n=26) were 95%, 83.3%, 88.5%, 94.7%, and 71.4%, respectively (Table 3). The relationship between SUV_max-P and DFS and OS was not significant (p=0.3; p=0.5, respectively). The cut-off values of SUV_max-T, SUV_max-PA, and SUV_max-IAC were not significant to discriminate LNM, high-grade tumor, MI, or CI. For the prediction of LNM, MTV-30 and MTV-40 with cut-off values of 11.9 cm³ and 24.8 cm³ yielded sensitivity and specificity of 83.3-60% (p=0.021; AUC 0.817) and 66.7-100% (p=0.051; AUC 0.767), respectively.

For the prediction of CI and MI, cut-off values of metabolic PET parameters were also evaluated with ROC curve analysis (Figure 4, 5). For CI prediction, MTV-30 and MTV-40 with cut-off values of 20.7 cm³ and 14.3 cm³ yielded sensitivity and specificity of 75-88.9% (p=0.006; AUC 0.844) and 75-88.9%, respectively (p<0.006; AUC 0.819).

The area under the ROC plot for detecting LNM using TLG-30 with a cut-off value of 99.09 g/mL cm³ was 0.792 (sensitivity 83.3%; specificity 75%; p=0.033). Additionally, the AUC of TLG-30 with a cut-off value of 82.06 g/mL cm³ was 0.788 (sensitivity 68.8%; specificity 80%; p=0.015) and the AUC of TLG-40 with a cut-off value of 49.01 g/mL cm³ was 0.781 (sensitivity 75%, specificity 70%, p=0.018), and they were significant predictors of MI. Furthermore, mean tumor diameter and lymphovascular invasion had significant relation with MI.

Multiple Logistic Regression Analysis

According to the regression analysis, MTV-40 [p=0.043; odds ratio (OR) 1.123; 95% CI 1.004-1.258] was an independent predictor of LNM, and the lymphovascular invasion (p=0.037; OR 64.006; 95% CI 1.291-3172.4) was the only significant predictive factor of MI. In addition, MTV-30 was a significant predictor of CI (p=0.04; OR 1.108; 95% CI 1.005-1.223; Table 4).

Discussion

In accordance with the literature, we found that SUV_max-P, MTV-30, and TLG-30 were significantly correlated with LNM in patients with EC. Moreover, we found significant correlation between these metabolic parameters and MI and CI, which are well-known prognostic factors to predict LNM in EC.

In patients with inoperaple EC, primary radiotherapy is the preferred treatment (10). By contrast, adjuvant radiotherapy is widely used, depending on the individual risk factors, including histological subtype, grading, lymphovascular-stromal invasion, MI and CI, tumor size, and LNM (9). While these prognostic factors were previously determined through surgery and pathological examination, nowadays, imaging modalities allow evaluation of tumor size, MI and CI, and LNM to some extent.

Patients with EC with early stage, grade 1, and grade 2 endometrioid histology, preoperative imaging usually does not significantly change the baseline management or prognosis. However, staging of patients with high-risk EC with ¹⁸F-FDG PET/CT has become more common gradually. This is an important issue because LNM is a major factor in treatment planning and prediction of prognosis. However, few studies have examined the diagnostic accuracy of PET/CT for the detection of LNM in EC, and available results show variable accuracy (11). Studies have reported that ¹⁸F-FDG PET/CT have high specificity in detecting metastatic nodes; however, its sensitivity was only modest and affected by the size of the metastatic deposit (12,13). In our study, we found high sensitivity, specificity, and accuracy of SUV_max-P with a specific cut-off value in the detection of LNM, and this finding was diffferent from those of previous studies.

Previous studies have demonstrated that high SUV_max-T can be associated with the aggres­siveness of EC (3,4), although its effect on overall survival or locoregional relapse remains controversial (5,11). Yahata et al. (13) reported that a high SUV_max-T was predictive of risk factors, such as deep MI, locally advanced stage, and node metastasis in EC. In contrast to these studies, the present study showed that SUV_max-T had significant relation only with tumor diameter in patients with EC.

In recent years, several metabolic parameters of PET/CT, besides the SUV_max, were reported to be useful in EC. Kitajima et al. (14) demonstrated that the MTV and TLG of local tumors were correlated with pathological features and were suggested useful for differentiating high‑ from low‑risk EC. Consistently, Chung et al. (8) reported that MTV was an independent prognostic factor for disease recurrence in EC, and Husby et al. (15) reported that MTV was useful to classify patients with high‑risk EC. Lee et al. (16) showed that preoperative TLG was related with disease recurrence in 28 patients with carcinosarcoma. Additionally, Shim et al. (17) stated that preoperative MTV and TLG could be independent prognostic factors to predict EC recurrence. In the present study, we found that MTV and/or TLG may be a new tool to assess well-established surrogate markers for poor outcome: High-grade disease, advanced FIGO stage, CI, and LNM. Therefore, we evaluated potential cut-offs to help identify patients at a higher risk of having these markers. For CI and LNM prediction, we found specific cut-off values for MTV-30 and MTV-40. Additionally, TLG-40 was a significant predictor of high-grade tumors, and TLG-30 and TLG-40 were higher in patients with EC with high FIGO stages. Our results also suggest the potential importance ofMTV and TLG for the preoperative classification of patients with high-risk status and improve the ability to tailor surgical and systemic therapies accordingly. Our results are similar with those of previous studies (15,17) that emphasize MTV and TLG as significant predictors of several clinicopathologic characteristics and superior to SUV_max-T in differentiating patients with high-risk status from those with low-risk status.

However, we could not achieve significant cut-off values for SUV_max-T, SUV_max-PA, and SUV_max-IAC to predict LNM, deep MI, CI, and high-grade EC. As SUV_max-T only represents the single greatest point of metabolic activity within the tumor, it cannot evaluate the entire metabolic tumor burden (18). Meanwhile, MTV and TLG can evaluate metabolic activity throughout the tumor volume. Therefore, these parameters could reflect tumor histology, prognosis, and treatment response more precisely than SUV_max-T.

Study Limitations

This study has some limitations. First, it was a retrospective study. Second, the study was conducted with a relatively small number of patients. Third, the study cohort was composed of patients with intermediate- or high-risk status and these findings do not represent those with low-risk status. Prospective studies with a larger number of patients and longer follow-up periods are required to confirm our findings. The potential added value of ¹⁸F-FDG PET/CT as a predictive biomarker is promising but requires further evaluation.

Conclusion

SUV_max-P, MTV, and TLG cut-off values on ¹⁸F-FDG PET/CT for the prediction of LNM increase the diagnostic accuracy and aid pretreatment identification of patients with high-risk status. Especially, SUV_max-P can be useful in deciding the extent of LN dissection and radiation therapy field for patients with medically inoperable intermediate-high-risk EC or for patients with inadequate surgical staging.