Assessment of local treatment modalities for FIGO stage IB-IIB cervical cancer: A propensity-score matched analysis based on SEER database

The aim of this study was to investigate the impact of local treatment modalities on the survival of patients with International Federation of Gynecology and Obstetrics (FIGO) stage IB-IIB cervical cancer, including cancer-directed surgery (CDS) alone and CDS combined with radiotherapy (RT). A total of 8,357 patients with cervical cancer between 1988 and 2013 were included in the final study cohort, including 4,298 (51.4%) patients who underwent CDS alone and 4,059 (48.6%) patients who received combination therapy. Univariate and multivariate analyses showed that local treatment modalities were prognostic factors for cause-specific survival (CSS). Patients who received combination therapy had worse CSS (HR = 1.38; 95% CI = 1.20–1.59; P < 0.001). Subgroup analyses showed the prognostic effect of local treatment modalities was significantly influenced by FIGO stage. In the propensity-score matched (PSM) dataset, CDS was associated with better CSS (P < 0.001) for patients with IB-IIA cervical cancer; nevertheless, no differences were observed in CSS (P = 0.639) for patients with IIB cervical cancer. In conclusion, radical surgery was the preferred treatment for patients with IB-IIA cervical cancer, and there was no difference between radical surgery alone and combination therapy for patients with IIB cervical cancer.


Impact of local treatment modalities on survival.
Kaplan-Meier analysis showed the impact of local treatment modalities on CSS for early stage cervical cancer patients (Fig. 1). Patients who received CDS alone had better CSS (P < 0.001; Fig. 1A) compared to patients who had CDS + RT. According to the NCCN guidelines, the preferred treatments for IB-IIB cervical cancer depend on disease stages. Thus, the prognostic effects of the different local treatment modalities according to FIGO stage were evaluated. For patients with IB and IIA cervical cancer, CDS was associated with better CSS (P < 0.001; Fig. 1B and C, respectively); nevertheless, for patients with IIB cervical cancer, no differences were observed in CSS according to different local treatment modalities (P = 0.259; Fig. 1D). According to this, we divided the data into two groups: one contains FIGO stage IB-IIA patients, the other FIGO stage IIB patients. Because there were both significant differences in the baseline characteristic of two groups, we performed two PSM analyses at a 1:1 ratio and a 5:1 ratio to erase significant difference of each variable, respectively (Table 3 and Table 4, respectively).  Table 2. Univariate and multivariate analysis of cause-specific survival. HR = hazard ratio; CI = confidence interval; FIGO: International Federation of Gynecology and Obstetrics; CDS: cancer-directed surgery; RT: radiotherapy.
(P = 0.259). In the matched dataset, we obtained similar results: in IB-IIA group, CDS was still associated with better CSS (P < 0.001; Fig. 2A) and multivariate analysis demonstrated patients who received CDS + RT treatment had worse CSS (HR = 1.50; 95% CI = 1.18-1.90; P = 0.001); in IIB group, there was no difference between two treatment groups in CSS (P = 0.639; Fig. 2B).

Discussion
In this study, we found that prognostic effects of the different local treatment modalities were greatly influenced by FIGO stage. On this basis, we further performed PSM analyses to adjust for biases from different baseline characteristics of two treatment groups. We demonstrated that CDS was associated with better CSS for patients IB-IIA cervical cancer. A retrospective study by Soisson et al. explained the reason why CDS outperformed CDS + RT. They reported that surgery alone had a better disease-free survival than combination therapy for patients with IB-IIA cervical cancer, but patients receiving adjuvant radiotherapy had several risk factors for recurrence, including lymphatic metastases, tumor involvement of the surgical margin, and large cervical lesions. Radiotherapy appears to reduce the incidence of pelvic recurrences but has no effect on distant failure of patients with high risks 7 . Besides, no difference was observed comparing CDS and CDS + RT for patients with IIB cervical cancer in our study. Several retrospective studies suggested that radiotherapy combined with radical surgery could prolong   the survival time for patients with IIB cervical cancer 8,9 . However, the previous studies may have a few shortcomings. First, these studies are composed of limited cases. Second, because multivariate analysis in our study showed that race, histological subtypes, grade, nodal status and tumor size were all affected prognosis of cervical cancer patients, it is important to adjust for differences between baseline characteristics; nevertheless previous studies have significant differences between baseline characteristics. There was also prospective study that indicated similar results 11 . Unfortunately, the significant differences between baseline characteristics were still observed in prospective studies. By contrast, our study has overcome these shortcomings. First, this study has a large sample size which included a total of 8,357 patients who met our inclusion criteria. In the matched dataset, there were still 2,998 and 1,140 patients in FIGO stage IB-IIA and IIB group respectively. Second, our current study is the first to adjust for imbalance in the baseline characteristics between two treatment groups using PSM analysis. Thus, our study is statistically significant and innovative. Zhou et al. 6 , using the SEER database, looked at the prognostic role of local treatment modalities in early stage small-cell carcinoma of the cervix (SCCC). They found that CDS was associated with better CSS in FIGO stage I SCCC; in FIGO stage II SCCC, no differences were observed in CSS. In our study, we adopted a more refined staging method to group patients into FIGO stage IB, IIA and IIB. We found that CDS was associated with better CSS in FIGO IB and IIA, and no differences were observed in FIGO IIB by subgroup analysis. Accordingly, we performed two PSM analyses regarding to IB-IIA and IIB respectively, which further confirmed our results. Furthermore, we investigate all subtypes of cervical cancer rather than simple subtype. Therefore, our study has larger application range and more reliability.
Although our study tried to overcome shortcomings of previous studies, it still had several potential limitations. First, SEER database lacks some important information regarding patient status, such as infection of human papilloma virus (HPV), comorbidities and economic conditions, that is apparently associated with prognosis 12,13 . Second, due to the limitations of SEER database, the information of chemotherapy regimens and dose is unknown, that limits our ability to assess the clinical outcome of local treatment modalities. Third, although our study has measure the effects of different surgery methods divided by resection scope, beyond that, how many patients had lymphadenecthomy and the number of lymphonodes removed both have effects on surgical outcomes. Nevertheless, the above information was not collected into SEER database. Fourth, the same condition for radiotherapy. The technique used and radiotherapy doses cannot be assessed. Fifth, adjuvant radiotherapy is recommended when postoperative pathological examinations reveal risk factors for recurrence, including lymph node metastasis, parametrial invasion, deep stromal invasion, lymph vascular space involvement, and bulky tumor (tumor size >4 cm) 14 . Unfortunately, the information about these risks is incomplete.
In conclusion, combination therapy has been investigated in several settings for decades [15][16][17] , but our study demonstrated that compared with radical surgery alone, radical surgery combined with radiotherapy did not improve survival in early stage cervical cancer patients. Given the excessive financial burden and side effects from radiotherapy 7,18,19 , radical surgery alone is preferred treatment for early stage cervical cancer patients. Further prospective studies designed adequately with larger sample size are needed to confirm the results of this study, as well as to define optimal local management in early stage cervical cancer.     We extracted cases of patients with a primary diagnosis of FIGO stage IB-IIB uterine cervical cancer (International Classification of Disease for Oncology, Third Edition) between 1988 and 2013. We excluded patients with more than one primary tumor, metastatic disease. Patients with unconfirmed or unknown tumor information were also excluded. The following covariates were collected from the database: age, race, marital status, histological subtypes, primary site, grade, FIGO stage, nodal status, size and surgical method. Relevant treatment-related data included cancer-directed surgery combined with radiotherapy (CDS + RT) and cancer-directed surgery (CDS) alone. In order to evaluate the overall effect of RT, we merge patients receiving either adjuvant or neoadjuvant RT into one group. Duration of follow-up, and cause of death described as due to cancer (CSS) were also included.
Statistical analysis. The chi-squared test (or Fisher's exact test, if appropriate) was used to analyze the differences between patients grouped by categorical variables. Clinical outcomes were compared between patients treated with CDS alone (referred to as CDS group) and CDS combined with RT (referred to as CDS + RT group). Survival curves were generated using Kaplan-Meier methods, and compared by log-rank test. Univariate and multivariate Cox regression analyses were used to analyze the independent risk factors for CSS. Only variables with statistical significance (P < 0.05) in univariate analysis were incorporated into multivariate analysis. Hazard ratios (HR) were calculated based on multivariable Cox proportional hazards models to estimate predictors of CSS. All CIs were stated at the 95% confidence level. Statistical significance was set at P < 0.05.
To adjust for differences between CDS + RT group and CDS group for FIGO stage IB-IIA and IIB patients, we performed two PSM analyses at a 1:1 ratio and a 5:1 ratio respectively. The PSM model was based upon age, race, marital status, histological subtypes, primary site, grade, nodal status, size and surgical method. The difference of each variable was considered significant if two-sided p-values less than 0.05. Statistical analyses were performed using SPSS statistical software package, version 22.0(SPSS Inc., Chicago, IL).