Cervical cancer prognosis and related risk factors for patients with cervical cancer: a long-term retrospective cohort study

This study aims to explore the recurrence rate and overall survival for patients with cervical cancer after the first treatment and the related risk factors. A retrospective cohort study was conducted on cervical cancer patients enrolled in a cancer specialist hospital in Hunan Province, China from January 1992 to December 2005 and followed up until December 2010. Kaplan–Meier survival analysis was used to estimate the cumulative recurrence rate, and Cox proportional hazards model was utilized to identify risk factors associated with prognosis. A total of 4358 patients were enrolled with a median follow-up of 7.4 years (range 5–19 years), and 372 (8.5%) patients had cancer recurrence. The cumulative recurrence rate showed a rapid increase from 3.8% in the first year after discharge to 8.0% in the fifth year, and the recurrence rate remained relatively stable afterward reaching 9.7% and 10.8% in the 10th and the 15th year, respectively. The median time to recurrence was 15.5 months with an IQR of 5.5–40.0 months. The Cox regression showed that miscarriage, clinical stage, and treatment received were significantly associated with cervical cancer recurrence after adjustment for confounders. Patients with recurrence showed a significantly higher risk for mortality than those without recurrence (HR 2.79, 95% CI 2.42–3.22). This study depicted the long-term recurrence rate and survival after recurrence for patients with cervical cancer after the first treatment, and reported time to recurrence and risk factors related to recurrence. These findings may provide important evidence for designing targeted interventions for the treatment of cervical cancer.

Risk factors for recurrence. We further analyzed the risk factors for cervical cancer recurrence by constructing Cox proportional hazards regression models (Table 3). We observed that miscarriage, clinical stage, and treatment received remained significant after adjustment for confounders. More specifically, the recurrence risk for subjects with three or more times of miscarriages was increased by 65% compared to those without a history of miscarriages (95% CI 1. 23-2.22); the HR for subjects with clinical stage II was 1.73 (95% CI 1.24-2.41) and clinical stage III or IV was 2.04 (95% CI 1.32-3.16) compared to those with stage I; subjects received chemotherapy only were 3.62 times more likely to had recurrence compared to those received surgery only (95% CI 1. 13-11.62). The sensitivity analysis that excluded subjects with recurrence within 3 months after discharge identified similar risk factors (Supplemental Table 3). When only including subjects with clinical stage I and II, we found that age greater than 60 years old was associated with a lower risk of recurrence (HR 0.53, 95% CI 0.30-0.94) than those with age younger than 40 years old; subjects with more than two times of miscarriages (HR 1.74, 95% CI 1. 26-2.39) and received chemotherapy only (HR 5.61, 95% CI 1.58-19.95) had a higher risk of recurrence (Supplemental Table 4).
Survival after recurrence. The total survival time was plotted for subjects with and without cancer recurrence (Fig. 3). The 1-year survival probability for those without cancer recurrence was 96.3%, the 5-year survival probability was 79.5%, and the survival probability dropped to 50.7% after 15 years of follow-up. The survival probability was significantly lower for those with cancer recurrence, which was 90.6% in the first year, 54.0% in the fifth year, 28.7% in the 10th year, and only 15.4% in the 15th year. The Cox regression showed that the HR of mortality was 2.79 (95% CI 2.42-3.22) for subjects with recurrence compared to those without recurrence after adjustment for confounders (Table 4). We also found that pathological type and year of admission were significantly associated with mortality in the adjusted model, and subjects with adenocarcinoma showed a higher risk for mortality compared to subjects with squamous cell carcinoma (HR 1.50, 95% CI 1.23-1.83). The sensitivity analysis including subjects with clinical stage I and II also found that recurrence, pathological type, and year of admission was associated with mortality (Supplemental Table 5).

Discussion
In this study, we reported the longitudinal pattern of cumulative recurrence rate over 15 years for patients with cervical cancer after the first treatment. We found that the overall crude recurrence rate was 8.5%, and the median time to recurrence was 15.5 months. Also, we found that the cumulative recurrence rate increased rapidly from 3.8% in the first year to 8.0% in the fifth year, and the increasing trend slowed down afterward where the rate was 9.7% in the 10th year and 10.8% in the 15th year. The recurrence rate reported in this study was consistent with a systematic review in 2009 which found that the recurrence rate for patients who completed primary therapy for cervical cancer ranged from 8.0 to 26%, and the median time to recurrence ranged from 7 to 36 months 10 12 . The variation of recurrence rate reported in different studies may be explained by the composition of patients with different disease severity in each study and different treatments received. Note that the major strength of this study was the analysis of the long-term recurrence rate pattern for a large population over 15 years of follow-up, and we included subjects ranging from FIGO stage Ia to stage IVb, based on medical records of patients admitted to the hospital during 1992 to 2005. These findings may provide evidence for designing treatment and follow-up guidelines for patients with cervical cancer. Our study reported several risk factors for cervical cancer recurrence, including younger age, lower parity, a greater number of miscarriages, higher clinical stage, lymph node metastasis, and received chemotherapy, and we found that miscarriage, clinical stage, and treatment received remained significant in the adjusted Cox regression model. These risk factors identified have also been reported by previous studies 4,20 . Clinical stage was the most commonly reported risk factor for cancer recurrence 19,21,22 , our study results showed that patients with clinical stage II had 1.73 times the risk of recurrence, and patients with stage III or IV had 2.04 times the risk compared to those with clinical stage I. Patients with advanced clinical stages had a greater range of cancer lesions, and a higher probability of peripheral invasion and lymph node metastasis, thus even if the patient was systematically treated, the possibility of recurrence for these patients was relatively higher 19,23 . In this study, we observed that patients with younger age had a higher risk of recurrence in the unadjusted model, while an insignificant association was observed in the adjusted model. It is still controversial whether age is related to recurrence with some studies reported positive associations 6,24 and some reported negative associations 14,25 . It's possible that patients with younger age had better adherence to the follow-up schedule, thus they were more likely to have a recorded recurrence. Unfortunately, we did not have detailed records on the adherence to follow-up, and it is important to further explore the reason between age and cancer recurrence. In addition, we did not include some risk factors that have been reported in previous studies, such as smoking, age of first intercourse, HPV infection, parametrial invasion, and vascular invasion [26][27][28] , and future investigations should take a more comprehensive considerations on related risk factors.
In this study, we observed that the treatment received varied by clinical stages, which jointly influenced the cancer recurrence rate. We found that surgery with adjuvant therapy was the most common treatment received for patients with clinical stage I, and more than half of the patients in clinical stage II and above were mainly treated with radiotherapy alone or chemotherapy plus radiotherapy. The treatment received in this study was generally consistent with the treatment guidelines indicating that surgery was the main treatment option for patients with early-stage cervical cancer 29 , and supplemental adjuvant therapy was added based on whether the patient has a risk factor for recurrence 30 . It's also important to note that this study included a heterogeneous population with large variations in clinical stages and treatment plans, and we observed that the treatment plan was not consistent with the standard guidelines for a very small number of patients. For example, a few patients in Clinical stage I and II received chemotherapy only without surgery or radiotherapy. One possible explanation was that these patients were not able to receive surgery due to other complications or economic reasons. It is also possible that some patients were transferred to other local hospitals of whom the treatment regimen was not recorded in great detail. In addition, as this study collected data over a long period, we showed that the recurrence rate and survival rate increased during the follow-up period. The increase in recurrence rate may be due to the  0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18  www.nature.com/scientificreports/ improved patient management system so that an increased number of subjects were followed up routinely and recurrence was more likely to be recorded. The improved survival rate may result from the improved standard of care over the years [31][32][33][34] . Note that in recent years, cisplatin-based chemotherapy and radiotherapy for advanced cervical cancer and inoperable early cervical cancer have been widely used in clinical practice, which has been shown to improve patient prognosis 35,36 .
In this study, we also demonstrated the overall survival rate for subjects with and without cancer recurrence separately and showed that cancer recurrence significantly reduced the overall survival rate, especially over the long term. The 5-year survival probability for patients with and without recurrence was 54.0% and 79.5%, and the 10-year survival probability was 28.7% and 63.6%, respectively. Such differences in survival between patients with and without recurrence were observed in many previous studies 6,37 , and these findings re-emphasized the importance of prevention of cancer recurrence in the management of cervical cancer. www.nature.com/scientificreports/ Our study had several limitations. First, it was performed on patients in one hospital in South China, which may limit the ability to extrapolate these findings to the different settings. Also, the database is only available until 2010, and it would be necessary to conduct follow-up studies to update the recurrence rate data using more recent and comprehensive data records. Despite that, this study had a relatively large sample size with 5-19 years of follow-up, which provided a rare opportunity to study the overall trend of long-term recurrence rate for cervical cancer patients. Second, a small number of patients were lost to follow-up, which may lead to loss to follow-up bias, although the number has been kept to minimal as we have conducted several rounds of follow-up with multiple tracing methods. Third, despite the large overall sample size, we were not able to conduct multiple stratification analyses due to limited sample size after stratification, and we combined patients in clinical stages III and IV due to the small number of subjects in Stage IV. Fourth, we found that a very small proportion of subjects had a short disease-free time of fewer than 3 months after completion of the primary treatment, and these patients may have a disease progression rather than cancer relapse. However, we conducted a sensitivity analysis that excluded these subjects and reported consistent findings in the Cox regression.
In this study, we reported the pattern of cumulative recurrence rate using a study with 5-19 years of followup. The overall cumulative recurrence rate was 3.8% in the first year and increased to 8.0% in the fifth year, then the recurrence rate stayed at 9.7% in the 10th year and 10.8% in the 15th year. The median time to recurrence in this study was 15.5 months. Also, we reported that younger age, lower parity, a greater number of miscarriages, higher clinical stage, lymph node metastasis, and received chemotherapy were significantly associated with cancer recurrence. Patients with recurrence have a much poorer prognosis in terms of survival probability compared to those without recurrence. This study may provide important evidence for designing targeted interventions for cervical cancer treatment.     Data collection and follow-up. Baseline information was collected by trained doctors through face-toface interviews, phone calls, and the extraction of medical records. The trained doctor was responsible for filling in a questionnaire that included baseline characteristics (i.e. age, gravidity, parity, and history of miscarriage), clinicopathological characteristics (i.e. duration of discomfort before diagnosis, clinical stage, pathological type, and level of differentiation), and treatment plan (i.e. surgical method, radiotherapy method, and chemotherapy method). Clinical stage was classified according to the 2009 International Federation of Gynecology and Obstetrics (FIGO 2009) staging criteria, and lymph node metastasis was classified into Yes or No based on postoperative pathological biopsy results. The pathological type and degree of differentiation were obtained according to the pathological examination report. Treatment received referred to the treatment received after the first diagnosis, which was classified into the following categories: surgery only, surgery plus adjuvant therapy, radiotherapy only, concurrent chemotherapy and radiotherapy, and chemotherapy only. The treatment plan was decided by the physician-in-charge according to the standard treatment guidelines based on the patient's condition and willingness for treatment. The post-treatment follow-up for subjects was planned for every three to 6 months during the first 5 years, and once a year afterward according to treatment guidelines in China 38 . Follow-up for subjects' recurrence in this study was conducted every year through telephone, mail, and extraction of medical records by trained researchers after the patient was discharged from the hospital until December 31, 2010. Subjects lost to follow-up were censored, and subjects completely lost to follow-up were excluded from the data analysis. Cervical cancer recurrence was defined based on clinical-pathological diagnosis results, and this study was focused on the first recurrence of cervical cancer after the first treatment.
Statistical analysis. Descriptive statistics were used for patient characteristics and summarized as mean ∓ SD or percentages. Differences in baseline characteristics and clinicopathological characteristics and treatment plan of the study participants with and without recurrent cervical cancer during follow-up were examined using student t-tests or chi-square tests, as appropriate. Average time to recurrence was presented with median and interquartile range (IQR) and a Log-rank test was conducted. Kaplan-Meier survival curves were used to plot the recurrence rate and survival rate after recurrence, stratified by different risk factors. Cox proportional hazards regression was used to calculate the hazard ratio (HR) and 95% confidence intervals (95% CIs) for cervical cancer recurrence and mortality. Proportional hazards assumptions were tested using Schoenfeld tests (Supplemental Table 2). Univariate analyses were conducted to evaluate patient characteristics associated with cervical cancer recurrence and mortality. The final multivariate model was selected based on the univariate analysis and biological significance, and variance inflation factors were used to test the collinearity between variables. Age, clinical stage, and treatment received were excluded from Cox regression for mortality due to violation of proportional hazards assumptions. A sensitivity analysis was conducted using Cox regression on cancer recurrence by excluding subjects with recurrence within 3 months as these subjects may have a disease progression rather than cancer relapse. We also conducted sensitivity analyses among subjects with clinical stage I and II using Cox regressions to explore factors associated with cancer recurrence and overall mortality. All statistical analyses were conducted using R version 4.0.5.

Data availability
The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.