Can surgery boost the survival benefit of chemoradiotherapy in T1b1-T2a1 stage cervical cancer with lymph node metastasis? A population-based study

Cervical cancer is a common gynecological malignancy worldwide. It is the most commonly diagnosed gynecological cancer and the leading cause of cancer mortality among women in China [1]. According to the American Cancer Society, an estimated 14,100 new cases will be diagnosed, and 4,280 women will die of cervical cancer in 2022 [2]. The incidence and mortality rates of cervical cancer have substantially increased (2.9% and 5.4%, respectively) in recent decades in developing countries [1].

In 2018, the prior 2009 International Federation of Gynecology and Obstetrics (FIGO) staging system was revised to incorporate imaging and pathological findings into the new system [3, 4]. One of the main changes was the new classification of stage IB. Stage IB is now divided into three substages according to tumor size limits.

Another major change was the addition of stage IIIC, wherein stage IIIC1 refers to pelvic lymph node metastasis (LNM) and stage IIIC2 refers to para-aortic LNM. According to the National Comprehensive Cancer Network (NCCN) guidelines, the use of pelvic irradiation with concurrent chemotherapy (CCRT) is recommended as the primary therapy for stage IIIC cervical cancer [5].

A previous study noted that the revised stage IIIC guidelines provided superior cancer-specific survival compared with stage IIIA or stage IIIB, which fails to comply with the monotonicity and separability of the tumor staging system [6].

Local tumor size is one of the strongest factors adversely affecting the overall survival (OS) in patients suffering from stage IIIC cervical cancer. Before establishing the 2018 FIGO staging, surgical resection was the main mode of treatment for early-stage (T1-T2) cervical cancer, followed by adjuvant chemoradiotherapy if node positivity was pathologically verified after surgery. With the improvement of the sensitivity and specificity obtained by lymph node imaging, we questioned the optimal treatment for early-stage (T1-T2) cervical cancer with LNM.

Therefore, we conducted a retrospective study to assess the survival of patients under different treatment options for T1a2-T2a1 stage tumors with LNM based on the Surveillance, Epidemiology, and End Results (SEER) database. The aim of this study was to analyze the survival outcomes in a subcategory of FIGO 2018 stage IIIC cervical cancer (T1a2N1M0-T2a1N1M0) where surgery-based therapy or radiation-based therapy is provided.

We identified cervical cancer cases from the SEER program of the National Cancer Institute (http://seer.cancer.gov/). The detailed information of these patients was extracted using SEER*Stat (version 8.3.9.2, USA), which is a statistical software program. A SEER program data use agreement was confirmed before the acquisition of the data. The inclusion criterion was pathological confirmation of primary cervical cancer from 2000 to 2018. The exclusion criteria were as follows: unknown age of diagnosis; follow-up with incomplete data; undetermined grade; unspecified neoplasms; unknown tumor size; and unknown or incomplete lymph node status. Since the patient information in the SEER database was adjusted to the 3rd–7th American Joint Committee on Cancer (AJCC) staging system, we restaged the tumors according to the 9th AJCC staging system [7]. A total of 12,701 patients identified as FIGO 2018 stage IIIC were included in this study as shown in Fig. S1. Lymph node status was defined as category “N stage” in the SEER database and was divided into clinical evaluation or pathological evaluation. In our study, we excluded rare histology other than squamous cell neoplasm, adenocarcinoma and squamous adenocarcinoma. Grade was measured as three-tier system. Surgery was categorized as radical hysterectomy (RH; including modified radical, extended and RH) and total hysterectomy (TH). Local tumor destruction or excision is excluded in Surgery group. Radiotherapy (RT) included adjuvant chemoradiation or chemoradiation alone. Adjuvant chemoradiation was defined as categories “positive CCRT” and “positive beam radiation in the SEER database. Detailed T stages were categorized and described according to the 9th AJCC staging system if not considered. The outcomes of this study were OS and disease-specific survival (DSS), which were calculated in months from the time of diagnosis to death.

Due to the openness of the patient information in the SEER database, our study was exempt from Institutional Review Board approval.

The clinical characteristics among cervical cancer patients in multiple groups were compared with chi-square (χ²) or continuity correction tests. The Cox proportional hazards model was utilized to identify prognostic factors. Cox analyses of the covariates (age, race, grade, histology, T stage, and treatment) affected OS and DSS. Because the covariates were categorical variables, a specific value was chosen as the reference. Because there were significant differences in the baseline characteristics of whole cohort, there was a selection bias. We performed a propensity-score matched (PSM) analysis of detailed T stages, whether surgery or chemoradiation was used. The PSM model was based on age, race, histological subtypes, grade and tumor size (median size).

OS and DSS were determined by Kaplan–Meier curves. Survival curves were compared with the log-rank (Mantel–Cox) test. Statistical analyses were performed using the SPSS 22.0 software package and Prism 7.0 software.

A total of 66,873 patients with cervical cancer in the SEER database met our inclusion and exclusion criteria. First, we investigated the LNM rate in patients with different T stages as shown in Fig. 1A. Node-positive patients accounted for less than 2% of stage T1a patients. Increased T stage led to an increasing proportion of node-positive patients (stage T3a: 44%, stage T3b: 43%). Among those patients, 12,701 patients were diagnosed as 2018 FIGO stage IIIC. Then, we tried to observe whether different treatment strategies resulted in different survival rates, especially in early T stage node-positive patients.

Fig. 1
Distribution of patients with LNM. (A) Percentage of patients with LNM in the T category. (B) OS of patients with LNM in the T category.
LNM, lymph node metastasis; OS, overall survival.

• Click for larger image
• Download as PowerPoint slide

First, Table 1 provides an overview of the clinical characteristics of those patients. The median duration of follow-up was 35 months (range, 9–224). A total of 4,694 patients (37.0%) were treated with surgery, 4,930 patients (38.8%) received radiation therapy, and nearly 10,000 patients (79.4%) underwent chemotherapy. At the end of the observational period of the study, 6,845 patients died (53.9%). Moreover, the 5-year OS and DSS rates were 43.4% and 47.6%, respectively, among node-positive patients.

Table 1
Baseline characteristics of stage IIIC cervical cancer patients

• Click for larger image
• Click for full table
• Download as Excel file

Next, we demonstrated the detailed survival outcome of different T stage patients with LNM (2018 FIGO stage IIIC) regardless of treatment options. Interestingly, consistent with a previous study, the survival outcomes of stage IIIC demonstrate a paradoxical phenomenon, which varies from 26% (T3b) to 82% (T1a2) [6] (Fig. 1B).

Some factors were correlated with patient prognosis and could bring heterogeneity to the survival analysis. Then, we tried to incorporate six clinical factors (race, age, tumor grade, histology, T stage, and treatment) with Cox analysis for patients with positive lymph nodes. Taking OS and DSS as the dependent variables, factors including black participants (hazard ratio [HR]=1.40, p=0.006; HR=1.38, p=0.017), grade 3 and 4 (HR=1.31, p=0.006; HR=1.33, p=0.009), squamous adenocarcinoma (HR=1.67, p=0.001; HR=1.84, p<0.001), advanced T stage (T1b3-T2a1) and primary hysterectomy groups (PTH+RT and PRH+RT) were defined as independent prognostic factors for cervical cancer with LNM (Table 2). Then, for those surgery or chemoradiation was used, we performed PSM analyses to erase the significant difference of each variable (age, race, histological subtypes, grade and tumor size). After PSM, 1,361 cases from each T subgroup were well matched with matched mean bias of 0% from unmatched mean bias of 9.7%–37.3% (data not shown). In the matched dataset, we obtained 436 (38.8% of surgery group) patients with TH and 689 (61.2%) patients underwent RH (Table 3). In early-T stage, we included 12 (38.7%) T1a2, 75 (42.1%) T1b1 and 213 (38.1%) T1b2 patients in TH group. For RH surgery route, 19 (61.3%) T1a2, 103 (57.9%) T1b1 and 346 (61.9%) T1b2 patients were finally matched after PSM.

Table 2
Factors associated with survival outcomes in the whole cohort

• Click for larger image
• Click for full table
• Download as Excel file

Table 3
Treatments of T1a2-T2a1 cervical cancer in the propensity-score matched cohort

• Click for larger image
• Click for full table
• Download as Excel file

We next asked whether radiation alone versus surgery with postoperative chemoradiotherapy affected survival outcomes, especially in early T stage (T1a2-T2a1) patients with LNM. Compared with those treated with chemoradiotherapy alone, patients in the primary surgery group had a higher 5-year OS and 5-year DSS in the PSM cohort (T1a2-T2a1, OS: 79.5% vs. 66.0%, and DSS: 81.9% vs. 70.6%, p<0.001) (Fig. 2A and B). In addition, those early T stage patients received different surgical procedures. Thus, T1a2-T2a1 node-positive patients who received chemoradiotherapy were stratified into “primary RH followed by RT” (PRH+RT), “primary TH followed by RT” (PTH+RT), and “RT alone”. Interestingly, the PSM dataset showed that there was a tendency for a similar prognosis in patients who underwent primary RH and primary TH (OS: 79.1% and 80.3%, p=0.87, DSS: 81.8% and 82.4%, p=0.74) (Fig. 2C and D). However, both surgery groups showed a better survival outcome than the RT group.

Fig. 2
Comparisons of survival of patients with pro-surgery RT. (A) OS and (B) DSS according to whether surgery was applied. (C) OS and (D) DSS according to the type of surgery (RH or TH). Time 0 was defined as the date of the scheduled operation in each patient. Level of significance: NS.
DSS, disease-specific survival; NS, nonsignificant; OS, overall survival; P, primary; RH, radical hysterectomy; RT, radiotherapy; S, surgery; TH, total hysterectomy.

^*p<0.001.

• Click for larger image
• Download as PowerPoint slide

Once we found that the initial treatment would have resulted in a vastly different prognosis in T1a2-T2a1 patients with positive lymph nodes, we wondered whether tumor burden could affect the optimal management pathway. Because fewer than 10 patients underwent primary RT in T1a2 after PSM, this group was not discussed. Unexpectedly, as shown in Fig. 3A and B, patients with smaller tumors (T1b1-T1b2) had more favorable prognoses in the PRH+RT group and PTH+RT group (T1b1: 94.8% and 92.9%; T1b2: 84.2% and 80.9%), while the RT alone group showed a poorer prognosis (T1b1: 77.9%, p=0.008; T1b2: 64.6%, p<0.001). In regard to the advanced T group (T1b3-T2a1), the situation is changed. Patients who cannot gain DSS benefit from additional surgery compared with those who receive chemotherapy (p=0.07; p=0.63) (Fig. 3C and D). More specifically, T1b3-T2a1 patients who received surgery, whether TH or RH, showed similar survival outcomes.

Fig. 3
Comparisons of survival of patients with detailed treatments according to T category. (A) T1b1, (B) T1b2, (C) T1b3, and (D) T2a1. Time 0 was defined as the date of the scheduled operation for each patient. Level of significance: NS.
DSS, disease-specific survival; NS, nonsignificant; P, primary; RH, radical hysterectomy; RT, radiotherapy; TH, total hysterectomy.

^*p<0.05; ^†p < 0.001.

• Click for larger image
• Download as PowerPoint slide

In this study, we examined the largest retrospective cohort of node-positive cervical cancer patients (n=12,701). Regarding stages T1b1 and T1b2, there was a severe difference in DSS when comparing the primary surgery group (RH and TH) and the “RT alone” group. For a larger tumor burden (T1b3-T2a1), patients did not obtain better DSS outcomes in the primary surgery group than in the RT group. More nuanced research proved that primary RH management did not obtain better survival outcomes than primary hysterectomy in stage T1b1-T2a1.

OS in the specific T category stage from the whole cohort ranged from 26% to 82%, which agreed with a previous study [8]. Although the revised 2018 FIGO staging system has offered more realistic survival predictions in consideration of nodal status, there are also objections due to its heterogeneity. Recently, a multicenter retrospective study named ABRAX compared the survival outcomes of different strategies in 2009 FIGO stage IA-IIB patients with LNM detected intraoperatively who were stratified by completed surgery or abandoned [9]. In this trial, the abandonment of RH did not improve the risk of recurrence (HR = 1.15 [0.80–1.67], p=0.45), pelvic recurrence (HR = 0.84 [0.46–1.52], p=0.56) or death (HR = 1.06 [0.69–1.64], p=0.78) compared with RH. Herein, using a more detailed classification, our results seem to be consistent with the results from the ABRAX trial. In addition, the Suprasert et al. [10] group and the Gray et al. [11] group conducted similar studies in small-scale cohorts of patients with early-stage node-positive cervical cancer who underwent or abandoned RH. The results showed no significant differences in recurrence-free survival. Another study, conducted by Ziebarth et al. [12], evaluated the outcome of patients who received PRH+RT or RT. There were no significant differences between groups in the local recurrence rate, distant recurrence rate or recurrence-free survival.

Under the evidence of few cohorts, European guidelines suggest that no further surgery in cases of lymph node involvement occurs intraoperatively, aiming to avoid combining radical surgery and RT [13]. Meanwhile, NCCN guidelines recommend RT as the primary choice in cases of lymph node positivity [5]. Therefore, individual surgeons from different countries often face a dilemma of opinion in management regarding the role of tumor removal. Proponents of the combination of RH and RT believe better pelvic control and a lower risk of central pelvis recurrence and morbidity could be reached, which is supported by nearly sixty percent of centers in Europe [14, 15, 16]. In contrast, proponents for abandoning radical surgery believe it reduces morbidity associated with radical surgery [15]. The association between RH+RT and multiple complications has been confirmed by several retrospective studies emphasizing quality of life. There is a higher risk for cervical cancer survivors who receive RH+RT to have bladder and bowel dysfunction, sexual dysfunction and psychological consequences [17, 18].

One of the exciting points of our research is that we included a sufficiently large number of node-positive individuals to detect the survival outcomes after subdividing. The study evaluated the outcomes of 12,701 early-stage patients with lymph node involvement. In Fig. 1A, it is clear that the percentage of LNM increases with increasing stage. To avoid selection bias on performing treatment methods, we utilized PSM analysis to match each patient who underwent surgery with another patient who received chemoradiation.

According to the NCCN, concurrent chemoradiation is the principal therapy for stage IIIC [5]. This may cause those patients with stage T1a2 to stage T2a1 to lose the opportunities for surgery. Given doubt, all of the included patients underwent RT after surgery. It was hypothesized that performing RH before RT could be beneficial in patients with a larger tumor burden [19]. However, our study found that TH showed the same similar survival outcome as RH, which is partly contrary to a previous study [7]. Meanwhile, early-stage (T1b1 and T1b2) patients who received additional surgery, whether TH or RH, experienced a better survival outcome than those who only received RT. Thus, our results also imply that performing a less traumatic operation followed by adjuvant RT resulted in superior survival in T1b1-T1b2 node-positive patients. For stages T1b3 and T2a1, the use of any additional surgery (RH and TH) did not improve survival. Therefore, with regard to inferior quality of life due to the combination of injury from surgery and radiation, primary chemoradiation for stage T1b3 and T2a1 patients is optimal.

Although our study benefits from the inclusion of a large population of patients from the SEER database, we acknowledge a number of important limitations. First, due to the long duration of case recording, it is difficult to avoid patient selection bias. Second, some staging details were not accessible across all years of study. The SEER registry does not include data on the presence of bulky lymph node. Meanwhile, given that parametrial involvement and positive node status (pelvic and para-aortic lymph node) were not parts of pancancer clinical indicators, this information was not captured in the SEER database, which caused inaccurate classification. Other limitations included the failure access of surgery treatments (traditional or minimally invasive surgery). An randomized controlled trial trial (primary surgery vs. primary chemoradiation) for T1a2-T1b2 node-positive patients is needed in the future.

In summary, for early-T stage (T1a2-T1b2) tumors with LNM, the survival outcome resulting from additional primary surgery remains the best, whereas patients undergoing primary debulking surgery and chemoradiotherapy show a similar prognosis in T1b3-T2a1 stages. Therefore, further prospective studies comparing primary surgery with primary chemoradiation for T1a2-T1b2 patients with LNM should be performed.

Fig. S1

Flow chart illustrating the process of evaluating patients for inclusion in the study.