The Role of the Metabolic Parameters of 18F-FDG PET/CT in Patients With Locally Advanced Cervical Cancer

Purpose To evaluate the role of the pre-treatment cervical and lymph node (LN) metabolic parameters of 18F-fluorodeoxyglucose positron emission tomography-computed tomography (18F-FDG PET/CT) for locally advanced cervical cancer (LACC) patients receiving concurrent chemoradiotherapy or radiotherapy. Methods we reviewed 125 consecutive patients with LACC who underwent pre-treatment 18F-FDG PET/CT examination and concurrent chemoradiotherapy or radiotherapy from February 2010 to December 2015 at our institute. The mean standardized uptake value (SUVmean), maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG) of cervical lesion and lymph node (LN) were recorded. Receiver operator characteristic curve, C-index, Kaplan-Meier method, and Cox proportional hazards models were performed. Results The median follow-up was 62 months (range, 4-114 months). For 125 included patients with cervical cancer, the 5-year overall survival (OS), disease-free survival (DFS), local control (LC) and distant metastasis-free survival (DMFS) rates were 83.6%, 75.1%, 92.3% and 79.9%, respectively. Cervical MTV (c-index 0.59-0.61) and cervical TLG (c-index 0.60-0.62) values calculated with a threshold of 40% SUVmax presented stronger prediction capability than cervical SUVmean (c-index 0.51-0.58) and cervical SUVmax (c-index 0.53-0.57) for OS, DFS, LC, and DMFS. In univariate analysis, cervical TLG ≥ 113.4 had worse DFS and DMFS. Cervical MTV ≥ 18.3 cm3 had worse OS and DMFS. In multivariate analysis, cervical TLG ≥ 113.4 implied worse OS, DFS, and DMFS. In either univariate or multivariate analyses, cervical SUVmean and cervical SUVmax had no statistically significant correlation with OS, DFS, LC and DMFS. For 55 cervical cancer patients with positive LN, LN SUVmax presented strongest prediction capability for OS (c-index = 0.79), DFS (c-index = 0.72), LC (c-index = 0.62), and DMFS (c-index = 0.79). In multivariate analysis, LN SUVmax remained significant biomarker linked to OS, DFS, and DMFS. Conclusion Pre-treatment cervical and LN metabolic parameters were associated with survival outcomes in patients with LACC. In our study, we found that pre-treatment cervical TLG and LN SUVmax may be important prognostic biomarkers for OS, DFS, and DMFS. However, further prospective studies with a large number of patients are required to evaluate the value of the metabolic parameters in survival outcomes prediction.


INTRODUCTION
Cervical cancer is a global health problem and the leading cause of cancer death for women in developing countries (1). Cervical cancer ranks eighth in incidence and mortality in China (2). Almost half of the patients present with locally advanced disease at the time of diagnosis. Currently, the primary therapeutic method for patients with locally advanced cervical cancer (LACC) is concurrent chemoradiotherapy. In approximately 80% of patients with disease recurrence, disease failure of cervical cancer occurs within 2 years after initial treatment. Some prognostic factors have been associated with clinical outcomes, including age, stage, tumor pathology, primary tumor size, lymph node status, squamous cell carcinoma antigen, and human papillomavirus (3)(4)(5)(6). 18 F-fluorodeoxyglucose positron emission tomographycomputed tomography ( 18 F-FDG PET/CT) has become an essential imaging tool in oncology in addition to conventional radiologic methods such as computed tomography (CT) and magnetic resonance imaging (MRI) (7). It is widely used in the diagnosis, clinical staging, response evaluation, curative effect observation, failure mode and prognostic analysis of cervical cancer and other tumors (8)(9)(10)(11)(12). In recent years, the association between the metabolic parameters of pre-and post-treatment 18 F-FDG PET/CT and treatment failure or survival in cervical cancer has become a research hotspot. Metabolic parameters, such as the mean standardized uptake value (SUVmean), maximum standardized uptake value (SUVmax), metabolic tumor volume (MTV), and total lesion glycolysis (TLG), are the focus of attention (13). On the one hand, some studies have reported the correlations between metabolic parameters and the clinical outcomes of cervical cancer. Patients with cervical cancer with a high SUVmax primary lesion show a worse prognosis (14,15). The baseline SUVmean can effectively predict the histopathological partial response of the primary tumor in LACC patients treated with chemoradiotherapy followed by surgery, suggesting the potential role of 18 F-FDG PET/CT in personalized treatment (16). Pre-treatment MTV and TLG are predictors of response to therapy and are correlated with overall survival in cervical cancer patients treated with chemoradiotherapy (17,18). On the other hand, there are some dissenting views. The role of SUVmax and SUVmean as prognostic factors for cervical cancer is still controversial (19). Whether MTV and TLG are important prognostic indicators of cervical cancer remains to be further studied (20).
In this study, we reviewed cervical cancer patients with pretreatment 18 F-FDG PET/CT and analyzed the associations between metabolic parameters and treatment failure or survival.

Patients
We reviewed patients with LACC who received a pre-treatment 18 F-FDG PET/CT scan and were treated with concurrent chemoradiotherapy or radiotherapy between February 2010 and December 2015 at our institute. The inclusion criteria were as follows: (1) pathologically proven cervical cancer; (2) 2009 FIGO stage IB2, IIA2 and IIB-IVA; (3) underwent 18 F-FDG PET/CT scan before primary treatment; (4) no evidence of distant metastases; and (5) treated with concurrent chemoradiotherapy or radiotherapy. The exclusion criteria were as follows: (1) underwent conization of the cervix; (2) previous or concurrent diagnosis of secondary primary tumor; (3) Karnofsky performance score <70; and (4) diagnosis of diabetes mellitus.
Pre-treatment evaluations included history, physical, and gynecological examinations, complete blood count, liver function test, renal function studies, chest and abdomen CT or whole-body PET/CT, and pelvic MRI.

PET/CT Technique and Image Analysis
The imaging agent 18 F-FDG, which has both a radiochemical purity and chemical purity greater than 98% and negative 24 h bacterial culture and bacterial endotoxin test by the gel method, was synthesized by the PET center of Peking Union Medical College Hospital. All patients fasted for at least 4 hours and rested for 90 minutes before an intravenous administration of 3.7-7.4 (0.1-0.2 mCi) MBq/kg body weight 18 F-FDG. The blood glucose level was less than 8 mmol/l before the administration of the radiotracer. The PET images were acquired in 3-dimensional mode with a Siemens Biograph 64 PET/CT system from the skull base to the symphysis pubis 1 hour after injection.
The acquired data were reconstructed using the ordered-subset expectation maximization method (two iterations, eight subsets, Gaussian filter, image matrix size 168 × 168). The attenuationcorrected volumetric images were collected in axial, coronal, and sagittal views, and they were independently performed by 2 senior PET physicians. The readers came to a consensus over controversial viewpoints. The SUVmax of the primary lesion of the cervix or positive lymph node (LN) was measured. A contouring around the primary cervical lesions or positive LN inside the boundaries was automatically determined, and the region of interest (ROI) with 40% SUVmax of the primary lesion of the cervix or positive LN within the contouring margin was delineated to define the cervical or LN MTV (20). The TLG of the primary lesion of the cervix or positive LN was calculated by multiplying the cervical or LN MTV by its SUVmean. The LN MTV and LN TLG analyzed in this study were calculated from the most FDG-avid lymph node (21).

Treatment
All patients were treated with external beam radiation therapy (EBRT) and high-dose-rate brachytherapy. EBRT was delivered with intensity-modulated radiation therapy (IMRT), volumetricmodulated arc therapy (VMAT), or helical tomotherapy (HT). A total of 50.4 Gy external radiation (1.8 Gy per fraction daily) was delivered to the elective regional lymphatics, and 59.36-61.6 Gy (2.12-2.2 Gy per fraction daily) was prescribed for the positive lymph nodes with simultaneous integrated boost (SIB) targets. For patients with para-aortic nodal involvement, the superior border extended to the level of renal vessels or to the upper margin of T12. High-dose-rate brachytherapy was delivered with an Ir-192 source, with 24-36 Gy (biologically effective dose 38.4-57.6 Gy) in four to six fractions to point A. The first-line recommendation of concurrent chemotherapy was weekly cisplatin (40 mg/m 2 ). A small number of patients received radical radiotherapy alone.

Follow-Up
Patients underwent follow-up examinations every 3 months in the first 2 years, every 6 months from 3 to 5 years, and once per year thereafter. Disease failure was confirmed by pathology or evidence of disease recurrence based on a series of imaging data. Overall survival (OS) was defined as the time from the start of treatment to death from any cause or the last follow-up. Diseasefree survival (DFS) was defined as the time from the end of treatment to recurrence or the last follow-up. Local control (LC) was defined as the time from the end of treatment to local failure or the last follow-up. Distant metastasis-free survival (DMFS) was defined as the time from the end of treatment to distant metastasis or the last follow-up.

Statistical Analysis
R and SPSS software (version 26.0; SPSS Inc., Chicago, Illinois, USA) were used for statistical analyses. ROC curve analysis was performed to determine the cut-off values of SUVmax, SUVmean, MTV and TLG of the primary lesion of the cervix or positive lymph node that indicated the optimal trade-off by maximizing the sum of sensitivity and specificity for survival outcomes. The c-index values of SUVmean, SUVmax, MTV, and TLG of the primary lesion of the cervix or lymph node were calculated to present the prediction capability of the metabolic parameters for survival outcomes. The Kaplan-Meier method was used to estimate OS, DFS, LC, and DMFS. Univariate and multivariate analyses of the patient characteristics were performed using the log-rank test and Cox proportional hazards model. To avoid many of the previously significant relationships falling out of the 0.05 significance level, we decided to include variables with p < 0.1 values in the multivariate analysis, and p < 0.05 values were statistically significant.

RESULTS
A summary of the detailed characteristics of all patients is shown in Table 1. In accordance with the inclusion and exclusion criteria of the study, 125 of the 1560 patients were finally included in this study. A total of 112 patients (89.6%) presented with stage IIB or above cervical cancer. A total of 114 patients (91.2%) had squamous cell carcinoma, 9 patients (7.2%) had adenocarcinoma, 1 patient had clear cell carcinoma, and the remaining patient had Mullerian carcinosarcoma. Fifty-two patients (41.6%) had a tumor size greater than 4 cm by gynecological examination. Forty-three patients (34.4%) had positive pelvic metastatic lymph nodes (MLNs) and 2 patients (1.6%) had positive paraaortic MLNs confirmed by 18 F-FDG PET/CT; 10 patients (8%) with positive para-aortic MLNs had concomitant pelvic lymph nodes metastasis.
All 125 patients completed concurrent chemoradiotherapy or radiotherapy with a median time of 51 days (range, 42-98 days). Twelve patients (9.6%) received neoadjuvant chemotherapy followed by chemoradiotherapy or radiotherapy alone, 102 patients (81.6%) received concurrent chemoradiotherapy as the primary therapy, and the remaining 11 patients (8.8%) received radiotherapy alone. Ninety-seven patients (77.6%) completed more than or equal to four cycles of chemotherapy. A total of 111 patients (88.8%) underwent a total point A equivalent dose at 2 Gy (EQD 2Gy ) greater than or equal to 85 Gy.
The median follow-up period for all patients was 62 months (range, 4-114 months). Of the 125 patients, 24% (n=30) experienced disease failure, including 6 patients with pelvic failure, 21 patients with distant metastases, and 3 patients with concurrent local and distant progression. The total local recurrence and distant failure rates were 7.2% and 19.2%, respectively. The cervix uterus was the most common site of pelvic recurrence, and the lung was the most common site of distant metastases. The 5-year overall survival, disease-free survival, local control and distant metastasis-free survival rates were 83.6%, 75.1%, 92.3% and 79.9%, respectively ( Figure 1).
ROC curve analysis was carried out to determine the best cutoff values of SUVmean, SUVmax, MTV, and TLG of the primary lesion of the cervix or positive lymph node in predicting the prognosis of cervical cancer, considering the sensitivity and specificity for survival outcomes ( The univariate analysis showed that para-aortic MLNs, total point A EQD 2Gy , and cervical TLG were significantly associated with DFS ( Table 3). In multivariate analysis, para-aortic MLNs, total point A EQD 2Gy < 85 Gy, and cervical TLG ≥ 113.4 remained significant in predicting DFS (Table 4). Furthermore, in multivariate analysis, 2009 FIGO stage, para-aortic MLNs, total point A EQD 2Gy , and cervical TLG were significant prognostic factors for OS. Para-aortic MLNs was a poor prognostic factor for LC. Para-aortic MLNs, total point A EQD 2Gy , and cervical TLG had important impacts on DMFS in multivariate analysis. cervical MTV was an important prognostic factor for OS and DMFS in univariate analysis; however, no significant differences were identified for OS and DMFS in multivariate analysis. Moreover, in either univariate or multivariate analyses, cervical SUVmean and cervical SUVmax had no statistically significant correlation with OS, LC, DFS and DMFS.
The 5-year overall survival, disease-free survival, local control and distant metastasis-free survival rates for patients with cervical TLG levels <113. 4

DISCUSSION
At present, various metabolic parameters of PET, such as MTV and TLG, have particularly become a research hotspot for predicting the prognosis of cervical cancer. However, there are different opinions on the role of PET metabolic parameters in the prognosis of cervical cancer. Some studies have shown that metabolic parameters of PET, such as SUVmax, SUVmean, MTV and TLG, play an important role in predicting the prognosis of cervical cancer ( Table 5). However, other studies have found no significant correlation between these parameters and survival. Therefore, in our study, we investigated the relationships between clinical characteristics  and PET metabolic parameters and the recurrence and long-term survival of cervical cancer. Our study shows that pre-treatment cervical TLG, LN SUVmax, Para-aortic MLNs, and Total point A EQD 2Gy are important independent prognostic factors for recurrence and survival. SUV can reflect metabolic activity as a semiquantitative marker of tumor uptake and has been demonstrated to play an important role in predicting the prognosis of cervical cancer in previous studies. A meta-analysis demonstrated that a significantly worse prognosis was associated with a higher SUVmax of the primary lesion in cervical cancer. However, SUVmax was not a significant independent prognostic factor in most of the enrolled studies in that meta-analysis (14). There are various reasons for this contradictory result, especially publication bias, which cannot be ignored. In addition, there are several limitations, such as missing data, the small sample size of each enrolled study and inconsistent treatment methods in different medical centers, which may cause differences in results. Voglimacci et al. (15) also suggested that cervical SUVmax as a continuous variable was a critical predictive index for survival outcomes, but the difference was not statistically significant when using the cut-off value. Herrera et al. (19) reported that pretreatment cervical SUVmean ≥ 5 was a significantly poor prognostic factor of OS (57% vs. 86%, p =0.03), DFS (36% vs. 88%, p = 0.004) and LC (65% vs. 88%, p = 0.04) in univariate analysis. However, statistically significant associations were not found between cervical SUVmean and survival outcomes in multivariate analysis. Meanwhile, the demonstration of an association between cervical SUVmax and prognosis would have been more challenging to interpret. In our study, cervical SUVmean and cervical SUVmax had no statistically significant correlation with OS, DFS, LC or DMFS. There are several articles with similar results to our study (13,22). However, LN SUVmax was significant associated with survival outcomes. Similarly, Martinez et al. (21) indicated that LN SUVmax was significantly linked to para-aortic nodal involvement only in univariate analysis. SUV may be affected by many factors, including blood glucose level, body mass index, scan duration, and reconstruction algorithm (23)(24)(25). Therefore, the role of SUVmean and SUVmax in predicting the prognosis of cervical cancer is still controversial and remains to be further studied.      MTV represents the volume of metabolically active malignant lesions, which is similar but more accurate than the measurement of tumor size on physical examination and may be significantly correlated to the prognosis of the disease. Leseur et al. (13) demonstrated that cervical MTV calculated with a segmentation of 55% SUVmax from pre-treatment PET/CT was applicable for predicting patient survival outcomes after concurrent chemoradiotherapy for patients with locally advanced cervical cancer. Similarly, Sun et al. (26) also considered that cervical MTV accumulation with a threshold of 40% SUVmax was a critical prognostic factor for cervical cancer patients and should be used to guide oncologists in selecting individualized therapies. Martinez et al. (21) proposed that cervical MTV calculated with a threshold of 40% SUVmax was an independent prognostic biomarker on para-aortic nodal involvement prediction. Guler et al. (27) took the opposite view that the role of using cervical MTV, calculated with the primary cervical tumor equal to or greater than an SUV of 2.5, to predict the prognosis of patients with cervical cancer and to develop patient treatment strategies required further confirmation. In our study, cervical MTV, calculated with a threshold of 40% SUVmax, presented an obvious association with OS but failed to reach the 0.05 significance level for DFS in univariate analysis; however, there was no significant association between cervical MTV and OS in multivariate analysis. For cervical cancer patients with positive LN, LN MTV remained significant correlation with DMFS in multivariate. We considered that the reasons for these different results may be related to the inconsistency in the definition of MTV in different studies. Therefore, we believe that MTV alone is not rigorous enough to predict the prognosis of patients with cervical cancer in the absence of a consistent definition of MTV.
The combination of MTV and TLG is a more resultful prognostic factor that takes into consideration both tumor volume and metabolic activity as crucial parameters of tumor response to treatment. Yoo et al. (28) highlighted that cervical TLG (cut-off, 7600), a volume-based metabolic parameter for primary cervical tumors, was a significant predictor of recurrence in cervical cancer in both univariate analysis and multivariate analysis. Likewise, Liang et al. (29) reported that total TLG was obviously correlated with survival outcomes in patients with locally advanced cervical cancer. Similarly, Carpenter et al. (18) indicated that total TLG measured by 18 F-FDG PET/CT was correlated with OS in high-risk cervical cancer patients treated with chemoradiotherapy and brachytherapy. Lima et al. (17) also preliminary suggested that although its p value seems to be below the critical value, pre-treatment total TLG was a significant independent predictor of response to therapy. However, the sample sizes of these studies were less than 100 cases. In our group, we obtained similar results and had a larger sample size. Although TLG confronts the same challenges as MTV, we still believe that the combination of multiple parameters makes predictions more effective.
In our study, we found that distant metastasis in patients with LACC treated with chemoradiotherapy and brachytherapy was a major pattern of treatment failure. This finding was consistent with that of previous research. Importantly, we found that cervical TLG and LN SUVmax were important prognostic factors for OS, DFS, and DMFS. The role of additional chemotherapy included adjuvant chemotherapy or neoadjuvant chemotherapy is still controversial. Dueñas-Gonzaĺez et al. (30) investigated 515 patients with locally controlled cervical cancer in a randomized study. The results showed that the 3-year PFS of concurrent chemoradiotherapy following two adjuvant cycles of cisplatin plus gemcitabine was significantly improved compared with standard therapy (74.4% vs 65.0%, p=0.029); the same result was found for OS (log-rank p= 0.0224; HR, 0.68; 95% CI, 0.49 to 0.95). However, the intervention group had more grade 3 and 4 toxicities than the control group (p<0.001). Adjuvant chemotherapy has not been widely accepted because further studies are needed to demonstrate the contributions of multiagent chemoradiotherapy and adjuvant chemotherapy to survival outcomes, and toxicity cannot be ignored. Da Costa et al. (31) conducted a randomized phase II trial to evaluate the efficacy and safety of neoadjuvant chemotherapy (NAC) followed by concurrent chemoradiotherapy (CRT) and reported that the addition of NAC with cisplatin and gemcitabine to CRT is not superior to standard CRT alone for LACC. Additionally, a phase III multicenter trial of weekly induction chemotherapy consisting of paclitaxel and carboplatin followed by standard CRT versus standard CRT alone in patients with LACC is undergoing (NCT01566240).
Local recurrence in cervical cancer patients treated with concurrent chemoradiotherapy or radiotherapy was another pattern of treatment failure. MRI-guided adaptive brachytherapy, which plays an important role, increased the radiation dose to the tumor and led to a significant improvement in the local control rate while minimizing the radiation dose delivered to surrounding normal tissues (32,33). In our study, we also found that for the 22 patients with cervical TLG ≥ 113.4 who experienced treatment failure, disease recurrence occurred in all patients within 5 years after treatment. Thus, active follow-up for at least 5 years is essential. These findings may provide an early signal-individualized intensive therapeutic approach with either adjuvant chemotherapy or MRIguided adaptive brachytherapy.
The present study demonstrates the value of the metabolic parameters of pre-treatment 18 F-FDG PET/CT as prognostic factors in patients with LACC. However, this study has several limitations. Most notably, this is a retrospective study with a small number of included patients and baseline data are easily to be incomplete. Moreover, this study includes a long-time span and changes in the treatment strategies may affect the results. In addition, positive LNs are identified on PET/CT and not by histopathologic verification. We cannot confirm that all FDGavid LNs are histopathological lymphadenopathies. Finally, recently there are many promising methods such as radiomics, machine learning, and especially deep learning (34,35). Combining these promising image analysis techniques may have more significant predictive values for the prognosis of cervical cancer. Further prospective randomized clinical trials with a large number of patients are required to evaluate the value of the metabolic parameters in survival outcomes prediction.

CONCLUSION
Pre-treatment cervical and lymph node metabolic parameters were associated with survival outcomes in patients with LACC. In our study, we found that pre-treatment cervical TLG and lymph node SUVmax may be important prognostic biomarkers for OS, DFS, and DMFS in patients with LACC. However, further prospective studies with a large number of patients are required to evaluate the value of the metabolic parameters in survival outcomes prediction.

DISCLOSURE
The abstract of this paper was presented at the 2020 ASTRO Conference name "The Role of Metabolic Parameters of Pretreatment 18

DATA AVAILABILITY STATEMENT
Data used in this study are not publicly available and can only be accessed, with appropriate approvals from data custodians and ethical clearance, from Peking Union Medical College Hospital.
Requests to access the datasets should be directed to FZ, zhangfq@pumch.cn.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Peking Union Medical College Hospital Ethics Committee [Protocol number S-K 1645]. The patients/ participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
FZ and KH contributed to the conception and the design of the study. DW wrote the first draft of the manuscript. WW and XL contributed to the review of literatures. LH and QP organized the database. XR performed the statistical analysis. All authors contributed to the article and approved the submitted version.