Predicting survival and risk stratication based on nomogram for locally advanced cervical cancer patients of the 2018 FIGO staging system treated with radiochemotherapy in southern China: a retrospective study

Background: Concurrent chemoradiation has been the standard treatment for locally advanced cervical cancer patients worldwide. This study was conducted to construct the nomogram of prognosis and risk stratication for locally advanced cervical cancer in southern China Material and Methods: The medical records of 258 LACC patients who mostly received cisplatin-based concurrent chemoradiation and high-dose-rate intracavity brachytherapy at the rst aliated hospital of Guangxi Medical University were analyzed. Stepwise selection of 22 factors with the Elastic net method was used to obtain a predictive model and construct nomograms for both progression-free survival (PFS) and overall survival (OS). Results: The median follow-up period was 47 months. The 3-year and 5-year OS were 73.7% and 65.9%, respectively, while the 3-year and 5-year PFS were 68.1% and 60.2%. In the multivariable analysis, different cumulative cisplatin dose (<180mg/m 2 : HR: 0.099, 95% CI: 0.010–0.909, p=0.040; 180-200mg/m 2 : HR: 0.058, 95% CI: 0.006-0.545, p=0.012; >200mg/m 2 : HR: 0.050, 95% CI: 0.006–0.520, p=0.012), bulky tumors (HR: 1.869, 95% CI: 1.127–3.098, p=0.015), ICBT (HR: 0.385, 95% CI: 0.218–0.681, p=0.001), and thrombocytopenia pre-treatment (HR: 6.832, 95% CI: 1.169–39.902, p=0.032) had signicant effect on PFS, and OS was inuenced by bulky tumors (HR: 1.916, 95% CI: 1.100–3.336, p=0.021), ICBT (HR: 0.499, 95% CI: 0.252-0.986, p=0.045), thrombocytopenia pre-treatment (HR: 3.865, 95% CI: 1.367-10.922, p=0.010) and high neutrophilic granulocyte percentage level in early stages of the treatment (HR: 2.041, 95% CI: 1.172–3.552, p=0.011).Nomograms of the independent predictors were established (cross-validated concordance probability c=0.671 for PFS, and 0.673 for OS). Based on the scores of the nomogram, subgroups of risk stratication further constructed showed a statistically difference in survival (p<0.05). Conclusions: Our study shows the prognostic value of nomogram to predict survival and establish risk stratication for LACC patients from southern China. Future individualized tailoring of therapy should focus even further on the high-risk patient populations.


Introduction
Among the most common cancers in women, cervical cancer is one of the leading causes of cancerrelated deaths, ranking as the fourth in frequency and mortality worldwide and the second behind breast cancer in developing countries with approximately 604,000 global new cases and 342,000 deaths in 2020 1 . In China, owing to the imbalance of regional development, there are about 131,500 new cases and 53,000 deaths, with a crude incidence rate of 15.3 women per 100,000 annually 2 .
Five large prospective randomized clinical trials have revealed that concurrent chemoradiation (CCRT), especially with cisplatin-based chemotherapy, prolongs the disease-free survival while reducing mortality of patients with locally advanced cervical cancer (LACC) [3][4][5][6][7] . With developments in techniques sharply increasing the local control ratio of LACC for decades, different predictive and prognostic factors and reported most in common were as follows: FIGO stage (2009), histology, tumor size, pelvic and/or paraaortic lymph node status, age, delivery of chemotherapy. In 2018, the International Federation of Gynecology and Obstetrics (FIGO) revised its staging system for cervical cancer, taking into account the detrimental effect of regional lymph node metastases and other factors on cervical cancer survival. An important change from the previous staging system is that the new staging system designates patients with regional lymph node metastases as stage IIIC, whose aim was to provide more strati ed and accurate prognostic groups.
Nearly ten years, nomograms are universally used as prognostic devices in medicine and oncology in order to help clinical decision and prognostication. Several nomograms of prognostic factors for LACC patients in developed regions of the world have been created over the years 8 , However, there are few nomograms related to LACC patients of the 2018 FIGO staging system in less-developed areas recently.
In China, despite the signi cant improvements in screening techniques, diagnostic procedures, and treatment modalities to decrease the incidence and mortality of LACC, the mortality rate remains at a high level, and distant disease control, as well as overall survival, is still a concern, especially in comparatively less-developed southern China. With the high infection rate of HPV, the lack of effective screening and vaccination, a large number of patients in these areas presented at a relatively advanced stage with highrisk factors such as large tumor size and regional lymph node metastasis when rst diagnosed. In face of the increasing new cases, unsatisfactory survival, and the lack of research on the correlated crowd, this current study was conducted to construct the predictive model of prognosis and further risk strati cation through nomogram for LACC patients treated in a general hospital in southern China, hoping to help trigger more individualized and adequate treatment decisions in the future. and/or computed tomography (CT) were performed to help to detect the size of the primary lesion, and the bulky tumor was de ned as those whose maximum diameter on imaging is greater than or equal to 4cm. Under additional approval by the patient, the positron emission tomography/computed tomography (PET/CT) was applied after the previous MRI/CT examination as a supplement. For patients who received PET/CT, pelvic and para-aortic lymph nodes with FDG accumulation greater than the liver accumulation of standard uptake value (SUV) > 2.5 ng/ ml were diagnosed as lymph nodes metastasis. For those without PET/CT, lymph nodes with a short-axis diameter longer than 1 cm were de ned as positive lymph nodes. Once diagnosed, each patient's standard treatment was explained, followed by approval through written informed consent. We obtained other clinicopathological data such as age at diagnosis, the chief complaint from the electronic medical record.

Treatment
Enrolled patients received pelvic external beam radiotherapy (EBRT) with three-dimensional conformal radiotherapy (3DCRT) or intensity-modulated radiotherapy (IMRT) using fractions of 1.8 to 2 Gy, which were delivered up to ve times a week for a total dose of 45 to 50 Gy. Meanwhile, a part of patients irradiated by IMRT received simultaneous integrated boost (SIB), in which positive or suspicious positive lymph nodes were de ned as gross tumor volume (GTVnd) with a higher dose of 56-60Gy. During the late treatment period, High-dose-rate intracavity brachytherapy (ICBT) delivering a total dose of 30-36 Gy During the treatment, the patients underwent routine complete blood cell counts, blood-chemistry testing, and clinical assessment weekly. We recorded the important indicators, including hemoglobin (HLB), platelets (PLT), neutrophilic granulocyte percentage (NEU), at three stages: pre-treatment, in early stages of the treatment (after the beginning of the treatment about two weeks), and upon treatment completion respectively, as the potential prognostic factors. The evaluation and value of the main indicators were as follows: low HLB level, ≤90 g/L; low PLT level, 100×10 9 /L or high PLT level, 350×10 9 /L; high NEU level, ≥0.8, respectively.
Regarding treatment safety, the treatment was withheld in case of grade 4 hematologic toxicities or grade 3 to 4 nonhematologic toxicity, or fever >38℃, or severe liver and renal function damage. When the indicators mentioned above recovered to a normal level, the treatment was resumed.

Toxicities
We also collected clinical data regarding treatment-related complications and toxicities. Complications occurring within 90 days from the beginning of the primary treatment were de ned as acute complications, and those occurring later than 90 days were de ned as late. The severity of acute complications was classi ed according to the NCI Common Terminology Criteria for Adverse Events, Version 3.0, and the late was graded according to the Radiation Therapy Oncology Group (RTOG) Late Radiation Morbidity Scoring Scheme.

Follow-up
Once the treatment was completed, patients were followed up every 3 or 4 months for the rst 2 years, every 6 months for years 3 to 5, and annually thereafter. During the follow-up period, besides general examinations and biology, US, CT, MRI of the abdomen and pelvis as well as PET/CT were optionally used for assessment. Overall survival (OS) was de ned from the date of the treatment to death from any cause or the date of the last contact. Progression-free survival (PFS), on the other hand, was assessed from the completion of treatment to the date of disease progression or last contact.

Statistical Analysis
Survival rates were estimated and compared by the Kaplan-Meier method. A univariate Cox regression model was used to identify the predictors. The important variables were screened out with nonzero coe cients as potential predictors of this prediction model using the Elastic net method. Then multivariable Cox regression analysis was used to construct the predictive model based on the results of Elastic net regression, and a further nomogram was developed. Risk ratios and 95% con dence intervals (CIs) were calculated. The prediction e ciency of the predictive model was assessed by C-index with bootstrap veri cation (1000 bootstrap resampling), as well as calibration curves.
In addition, a risk strati cation based on the level of nomogram scores was conducted, in which the population was divided into three subgroups (averaged as much as possible). To demonstrate whether the nomogram distinguishes between people at different risks, Kaplan-Meier curves were performed and validated for these three groups. All statistical analyses were conducted using R version 4.0.3 and SPSS 26.0, and p < 0.05 was considered to be statistically signi cant.

Patient Characteristics
The tumor-associated and patients characteristics are summarized in   Table 2.A total of 22 factors were analyzed as predictive prognostic factors, and the univariate analysis was presented in Table 1.

Univariate and Multivariate Analysis of all Factors based on Elastic Net Regression
Elastic net regression analysis was performed to re-evaluate the variables in order to avoid the in uence of confounding factors. Finally, 9 variables were retained with nonzero coe cients for PFS ( Fig. 1 and Table 3) and 6 variables for OS ( Fig. 2 and Table 3) as potential predictors of the prediction model. To get a more comprehensive view of the association between the survival outcome and these predictors, we further performed a multivariable Cox regression analysis and constructed a predictive model.

Nomograms Establishment and Risk Strati cation
The results of the Cox regression analysis were given in Table 3 and visualized in the form of a nomogram plot to help practice in the clinic (Fig. 3) Moreover, we further took the nomogram score into the evaluation to perform different risk strati cation. The subsequent Kaplan-Meier analysis showed a signi cant difference in survival time between three groups of strati cation, suggesting that the nomogram can distinguish between patients at different risks. The 5-year PFS in the low-risk, intermediate-risk, and high-risk group were 80.0%, 60.1%, and 31.8%, and the 5-year OS were 83.1%, 51.8%, and 38.5%, respectively. Considering the obviously lower survival rate in the high-risk group, more individualized and intense treatment may be required for this group of patients.
As is reported, age is closely related to the prognosis of patients with LACC, especially patients younger than 35 years old, which were considered to be more susceptible to local recurrence or distant metastasis 9 . Nevertheless, our study didn't reveal that age has a statistically signi cant effect on OS or PFS, probably due to the lack of detailed strati cation and fewer patients in the younger age group. As for pathological type, with an increasing incidence in recent years, adenocarcinoma is considered a poor prognostic factor in patients with LACC 10 . Compared with SCC, adenocarcinoma is more likely to make larger primary lesions and lead to distant metastasis. However, Galic Vijaya et al. reported that despite the lower complete remission rate of 86.5%, patients with adenocarcinoma receiving CCRT had no apparent difference in long-term survival 11 . Likewise, our study found that patients who received CCRT have no statistical difference in OS and PFS between adenocarcinoma and SCC.
As is known, pelvic and para-aortic LNM plays a key role in the poor prognosis of cervical cancer. Liu et al. indicated that the overall survival was 91% for pelvic LNM negative cohort and 67% for that positive cohort 12 , Most nomograms estimating PFS or OS for LACC include pelvic lymph node status as a signi cant risk factor, regardless of diagnostic methods: lymphadenectomy, PET/CT, or multiple imaging 8, 13 . In 2018, FIGO revised the 2014 system into the 2018 staging system of cervical cancer 14 .
One crucial difference from the previous staging system is that the latest 2018 system designates patients with regional lymph node metastasis into new stage category, in which patients with pelvic lymph node metastasis only are allocated to stage C1, and with positive para-aortic lymph node (PALN) allocated to stage C2. Our study comprehensively evaluated patients diagnosed with the previous staging system before 2018 and restaged those who met the C category in 2018 FIGO. A total of 38 were diagnosed as C. However, we failed to analyze that the FIGO stage, the C category in a particular, is a signi cant and independent factor, and there is no signi cant difference in 5-year overall survival between C and B (59.9% vs. 54.8%, p>0.05). What calls for attention is that stage C of the new system doesn't take primary tumor size and extent as well as the characteristics of LNM into consideration. Liu et al. 15 reported that tumor size and number of PLNM are signi cant prognostic factors for DFS in patients with stage C1r while con rming the heterogeneity among this group of patients. Therefore, regarding patients with combined and complicated status of large tumor size, lymph node metastasis, or other potential risk factors, the reasonability of new FIGO staging system to evaluate prognosis is limited.
On the other hand, the comparability of clinical stages in different regions and countries in the world has decreased, because the clinical stage in some countries, developed countries especially, is affected by imaging like MRI or PET/CT. In a meta-analysis, PET/CT and DWI-MRI have a high accuracy in detecting LNM. Among several modalities, PET/CT has the highest speci city, and DWI-MRI has the highest sensitivity 16, 17 . What's more, compared to MRI on regional hypogastrium or pelvic, PET/CT has its advantages on detecting long-distant LNMs. Henrik Hansen et al. 16 reported that the inclusion of PET/CT in the pre-radiotherapy diagnostic protocol correlated with survival bene ts after CCRT in node-negative cervical cancer patients of 23 %, 19 % for OS, and disease-free survival (DFS). Due to the imbalance of regional development, the diagnoses of LACC patients in less developed areas were relatively limited. In our study, only 18 patients were diagnosed as stage C through PET/CT. A majority of patients refused to perform PET/CT or regional DWI-MRI because of high cost and personal reasons, which possibly lead to underestimates of clinical stage in patients with potential pelvic and para-aortic lymph node metastasis or worse distant organ metastasis. With developments in techniques, the failure pattern of LACC has changed. Lymph node failure has become a sort of main failure site remaining unsolved. It would be necessary to use high-precision imaging equipment like PET/CT or DWI-MRI to de ne the clinical metastases exactly in future practice.
Large tumor size was strongly suggested as a prognostic factor for LACC 18 . Similarly, our study demonstrated that patients with bulky tumors had an adverse effect on PFS as well as OS. It is acknowledged that tumor volume is the most direct indicator of tumor burden, representing the number of clones needed to be killed among the tumor 19 . In addition, large tumor size tends to give rise to a lack of oxygen, radiation resistance, and worse local control. In the relatively less-developed region of southern China, plenty of patients were diagnosed as advanced stages with the large primary site before treatment, leading to poor prognosis.
With advances in Image-guided radiation therapy (IGRT) and techniques, IMRT has gradually taken the place of traditional two-dimensional (2D) and three-dimensional (3D) techniques in EBRT procedure for better radiation protection to surrounding organs, bone marrow sparing as well as lower toxicity 20,21 . As shown in our study, patients with IMRT had a lower incidence of toxicity compared to those with four-box eld, especially severe and unacceptable rectum and sigmoid toxicity. Nevertheless, we failed to observe that IMRT improved overall survival signi cantly, which may be attributable to the slightly short time follow-up (median of 47 months), representing the low number of patients followed beyond 5 years.
For decades, it is universally acknowledged that the combination of radiotherapy with chemotherapy prolongs disease-free survival while reducing mortality. In our study, chemotherapy was performed using cisplatin alone or with several combinations, and most patients received standard chemotherapy performed at weekly or tri-weekly intervals with the cumulative cisplatin dose of 180 to 200 mg/m 2 .
Interestingly, we observed that not chemotherapy regimens but different cumulative cisplatin doses had an impact on PFS (p<0.05). Except for those who didn't complete chemotherapy as planned, patients who received higher doses of cisplatin had a better 5-year PFS (74.4% vs 67.0%). Due to the higher incidence of bulky tumor (50.8%) and parametrial involvement (75.2%) among the enrolled cohort, it seemed that dose-intensive or additional adjuvant cisplatin-based chemotherapy not only inhibits accelerated tumor cell repopulation in a period but also further reduces potential systemic micro-metastases, thereby probably increasing the actual local control rate. Likewise, Landoni et al. 22 suggested that additional chemotherapy could bene t those patients with intra-cervical residual and suboptimal responses. However, our study did not demonstrate that chemotherapy with higher doses of cisplatin bene ted the overall survival of LACC patients, which probably was associated with the higher incidence of hematological toxicity in additional chemotherapy. Therefore, among LACC patients at different risks, it is necessary initially to re-evaluated the best-t administration and delivery of chemotherapy.
Recently, indicators of blood cell counts have attracted attention as potential prognostic factors in cancer patients. In contrast to the nomograms performed by Rose et al 8 , we obtained three main complete indicators of blood cell counts at different periods during the treatment, aiming to investigate their dynamic variations and analyzed their prognostic predictive value, in consequence nding that thrombopenia before the treatment had adverse effects on both PFS (p=0.032) and OS (p=0.010), and high NEU level in early stages of the treatment was associated to poor OS (p=0.011). As is identi ed as a marker of poor prognosis in several forms of cancer, thrombocytopenia represents coagulation disorders and a higher incidence of bleeding risk, potentially leading to less tolerance of CCRT 23 . The tumor microenvironment plays a key role in growth and metastasis in present studies, which can be visualized by varieties of in ammation-related cells, such as neutrophils, platelets, and lymphocytes, especially neutrophils 24 . Continuous high level of NEU is one of the paracancerous syndromes of many malignant tumors, representing chronic in ammation in patients. In ammation could change the systemic microenvironment for inducing the angiogenesis of tumors. What's more, In ammation level positively correlated with granulocyte colony-stimulating factor (G-CSF) expression in tumor tissues, which stimulates the production of myeloid-derived suppressor cell (MDSC), thus leading to the rapid development of the tumor and radiation resistance 24 . Yet there is little high-quality evidence on how to eliminate such chronic in ammatory responses in oncology patients.
In the current study, multivariate analyses were performed based on the elastic net regression model as the best-t model to select potential risk factors. Moreover, nomograms and risk strati cation were established to verify our predictive model accuracy further. As a result, the calibration plots showed an acceptable consistency between the predictive and actual probability of the survivals (cross-validated concordance probability c=0.684 for PFS, and 0.654 for OS), which could be the study's strengths. However, as this is a retrospective observational study, some limitations need to be addressed. First, the follow-up time was slightly short (median of 47 months). Second, not all patients bene ted from a uniform and standard therapeutic scheme. Third, all patients were enrolled from a single institution. The factors mentioned above represent the fact that several biases would be inevitable to occur. Therefore, these factors can only be eliminated in prospective randomized studies with a more strati ed and larger population.

Conclusion
Cisplatin-based concurrent chemoradiation revealed a favorable survival of LACC in southern China. In addition to frequent factors (bulky tumors, ICBT), cumulative cisplatin dose and neutrophil percentage seem to be essential outcome predictors. The present study shows the prognostic value of nomogram to predict tumor-related survival and establish risk strati cation in a setting of LACC patients from southern China. Therefore, future individualized tailoring of therapy should focus even further on the high-risk patient populations.

Funding
This study was supported by the Guangxi Zhuang Autonomous Region Health and Wellness Committee Science and Technology Project (S2017017). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests
Con ict of interest relevant to this article was not reported.

Availability of data and material
The datasets generated during the current study are not publicly available due to ethical considerations, but are available from the corresponding author on reasonable request.

Ethics approval
The authors state that they have obtained appropriate institutional review board approval or have followed the principles outlined in the Declaration of Helsinki for all human or animal experimental investigations.
6. Consent to participate Written informed consent was obtained from the patients.  Tables   Table 1 Characteristics for patients and results of univariate regression analysis.