The main treatment approach for early-stage liver cancer patients is liver cancer resection. Patients with early, small-sized liver tumor who either reject or cannot be operated on can be treated with local external radiation [7]. Liver tumors are the second most sensitive tumors to radiotherapy, with moderate to high radiosensitivity, after tumor tissues that have extremely high radiosensitivity to bone marrow and lymphatic tissues [8]. EBRT mainly includes stereotactic body radiotherapy (SBRT), three-dimensional conformal radiotherapy (3DCRT), intensity modulated radiotherapy (IMRT), and volume modulated arc radiotherapy (VMAT). Protons and carbon ions are also used to treat patients with liver cancer, but are more expensive [9]. SBRT is the most successful approach for the treatment of small-sized liver cancer, with a main focus on early HCC [10]. For lesions of all sizes, SBRT showed a relatively high 1-year OS and a low incidence of acute grade 3 + complications in HCC [11]. The significance of SBRT in HCC treatment has always been proven, regardless of the size of the study population in well-designed phase II trials, being either a relatively small retrospective cohort or a large series [12]. Currently, radiation therapy has become an indispensable part of the comprehensive treatment of liver cancer [13]. In the guidelines of some countries such as South Korea [14], radiotherapy is listed as the priority treatment for patients with early, middle, and late stage HCC. However, it is used as a treatment alternative when other standard treatments in some guidelines, such as NCCN (National Committee on Computer Network), are not feasible [15]. Our results (78% 1-year OS after external beam radiation) are close to the 1-year overall survival rate of 73.6–81.1% reported in previous studies for the treatment of early-stage liver cancer. This indicates that EBRT has a good therapeutic effect on early-stage liver cancer [16]. Therefore, the aim of this study was to further evaluate the prognosis of patients with early-stage liver cancer undergoing either surgery or external radiation, to guide clinical decision-making.
The competitive risk model established in this study suggests that age, disease, time of diagnosis, tumor tissue grade, and treatment are all independent factors for the prognosis of liver cancer patients. This study also reported that patients < 60 years old, with delayed diagnosis, low histological grade, and who had undergone surgical treatment showed good prognosis. Surgery showed a higher survival rate than EBRT (HR = 0.224, 95% CI [0.139–0.36], p < 0.001). After PSM, subgroup analysis and interaction test found that all patients can benefit from surgery and external radiation. The benefit of chemotherapy in female, white, widowed/other, < 60 years old patients, and those with a diagnosis time during or after 2014 is greater than that of other groups.
Our study reports that surgery shows a higher survival rate among HCC patients than EBRT. After PSM, subgroup analysis and interaction test found that all HCC patients can benefit from surgery and external radiation. The group of white, widowed/other, female, < 60 years old patients, with a diagnosis time during or after 2014, and receiving chemotherapy showed a greater benefit than that of other groups.
Through the conditional survival rate, we found that as the survival time of liver cancer patients increased, the survival curve showed a slower decline in surgery than that of radiotherapy patients, and stabilized after about 3 years. The survival curve of radiotherapy patients dropped significantly within 4 years, and then stabilized thereafter.
The propensity ratio scoring method is one of the commonly used methods to control confounding factors in real-world research. Its basic principle is to express the influence of multiple confounding factors in a comprehensive propensity score to correct the imbalance of data between groups [17]. Previous studies have shown that significant risk factors for HCC recurrence mainly include HCC lesion size (especially > 3 cm), etiology, serum albumin levels, and serum alpha-fetoprotein levels [18]. Factors affecting survival include patient’s age and Child Pugh staging. The patient’s age has also been considered to be related to the incidence of the disease and post-treatment complications [19]. Our study was a retrospective large-sample study, and used PSM to enhance credibility and reduce potential confounding factors and selection bias between the case group and the control group. Our study findings conformed with those of the previous studies, where PSM, subgroup analysis and interaction test of competitive risk model showed that the treatment benefit was higher in the group < 60 years old, female, widowed/others, white patients who underwent chemotherapy and were diagnosed during or after 2014.
By comparing the results of the traditional survival analysis COX model and the competitive risk model, a previous study [20]showed that the use of the COX model could not provide an accurate estimate of the impact value, because it only considered the results of a single factor, and thus might overestimate or underestimate the impact of the independent risk factors. If the proportion of competition events was > 10%, using traditional methods could cause serious bias. On the other hand, a proportion of competition events < 10% may have false positives or false negatives [21]. For survival analysis, competitive risk model can divide the end points of survival data into multiple categories and eliminate the impact of competitive events on prognosis research [22]; thus it is considered a more effective survival analysis model. Previous studies mostly used traditional survival analysis methods, ignoring the existence of competing events, and hence, the risk of death from cancer may be overestimated [23]. This study was a multi-center, large sample study of patients with stage I liver cancer based on the SEER database. The study included a total of 2155 cases and had a strong statistical power, which compensated for the shortcomings of the small sample size studies of general clinical research, and had high clinical reference value. In this study, 523 people died in the OS analysis, of which 300 people died in the DSS analysis, accounting for 57.36% of total deaths. Therefore, the proportion of deaths due to other competitive events was 42.64%, which was suitable for analysis using a competitive risk model to incorporate multiple factors, as shown in this study. The cumulative risk curve showed that there were significant differences between the two groups at different ages and with different treatments. However, there was no significant difference between the groups with respect to receiving chemotherapy, having different time of diagnosis, or belonging to different races and sexes, which further supported the superiority of our usage of the competitive risk model.
Generally, the survival rate post liver cancer resection is evaluated based on the date of surgery. However, this traditional survival curve may not provide an accurate prediction of long-term survival, mainly because the recurrence and mortality rates are generally the highest in the first few years after surgery. Survival rate is dynamic and is directly related to the duration of time between the beginning of the treatment course and the time of evaluation [24]. Estimates of survival time vary over time, thus conditional survival evaluation is a more meaningful way to evaluate long-term prognosis. It is calculated based on the patient’s survival time and is more relevant to the actual clinical practice because it can measure the patient’s survival risk over a certain period. Conditional probability analysis is increasingly used to analyze the long-term effects of tumor characteristics [25]and is particularly suitable for comparing the difference between immediate and late-stage survival benefits after treatment [26].
In this study, we evaluated DSS and OS in early-stage liver patients after surgery or external irradiation based on a large, multi-center database. The results showed that after 3 years of surgery or 4 years of external radiation, the decline in patients' DSS and OS tends to be more stable as the survival time increases, which means that generally long-term cancer patients have better prognosis than newly diagnosed patients. This was similar to other studies [27], which showed that the impact of tumor-related factors on the long-term survival of liver cancer patients was diminished starting from the third year post surgery. This also showed that patients with early-stage liver cancer had the most survival-related events such as recurrence and metastasis within 3 years post-surgery or 4 years post external radiation. Close follow-up and re-examination are needed for timely treatment to reduce mortality. Previous studies [26] had shown that approximately 7% of patients undergoing surgical treatment for HCC with mild gross vascular infiltration were expected to be cured, and conditional survival probability analysis could also be used to guide the follow-up of such patients after surgery [15]. These findings were different from our study because their research object was liver cancer patients with mild gross vascular infiltration who showed a worse prognosis than that of stage I liver cancer patients. In addition, the number of cases in their study was limited to only a few hundred.
The current HCC NCCN guidelines recommend that all liver cancer patients should be followed up every 3–6 months for the first 2 years after surgery, but there is no optimal postoperative/post-radiotherapy follow-up strategy for patients with early-stage liver cancer. Conditional survival evaluation can provide more valuable information for the determination of subsequent strategies. On the other hand, DSS survival rate showed a slower decrease than OS after surgery and external radiation, indicating that other factors that affect the survival of liver cancer patients 3 years after surgery or 4 years after external radiation have become more important. Therefore, this study not only compares the difference in survival of patients with early-stage liver cancer after surgery and external radiation and the relationship between clinicopathological characteristics and prognosis, but also conducts for the first time a dynamic analysis of the survival of patients with early-stage liver cancer.
One of the limitations of this study is that most of the data collected by the SEER database were clinical indicators, excluding laboratory examinations, imaging examinations and others, thus the accuracy of the clinical TNM staging of patients without surgery needed to be improved. Second, there were some missing clinical indicators in the SEER database. Third, this study only conducted internal verification of the prognostic model, and we hoped to conduct further external verification to evaluate the applicability of the predictive model in the population in a more comprehensive manner. All of these shortcomings might affect the accuracy of the prognostic prediction model constructed in this study. In the future, more clinical studies are needed for further verification of the results.
In summary, this study is the first to analyze and study the prognosis model of early-stage liver cancer patients established in the SEER database based on the competitive risk model and conditional survival analysis, which introduces new ideas and methods for the study of predictive models. It can more comprehensively and accurately predict the prognosis of patients with early-stage liver cancer, provide important parameters for the prevention of tumor recurrence, and provide personalized treatment plans and prognostic judgments.
