Introduction
Over the past few decades, substantial improvements in cancer survival rates have led to a growing population of cancer survivors (Felicetti et al.
2018). These individuals face a heightened risk of developing new malignant tumors (Araujo-Filho et al.
2021; Bostrom and Soloway
2007), attributable to a combination of lifestyle factors, genetic predisposition, and previous cancer treatments (Weir et al.
2013). Radiation therapy (RT) is a fundamental treatment modality for pelvic tumors, including malignancies of the cervix, rectum, and ovary. While RT effectively reduces tumor recurrence and significantly enhances prognosis (De Sanctis et al.
2013), it also presents long-term risks for patients, such as the progression of secondary primary malignancies (SPMs)–a rare but consequential late complication of cancer treatment (Berrington de Gonzalez et al.
2011; Cuccia et al.
2020). Recent studies, however, have demonstrated that RT may not invariably increase the risk of SPMs and could even exert a protective effect in certain cases (Wiltink et al.
2015).
In the context of Surveillance, Epidemiology, and End Results (SEER) analyses, the risk of RT-associated SPMs remains relatively high (Conway et al.
2017; Guan et al.
2021; Li et al.
2022; Moschini et al.
2019; Yi et al.
2023; Yu et al.
2022). Several recent studies have investigated the effect of RT on SPM risk in relation to primary malignant neoplasms. For instance, Wen et al. performed a retrospective review of prior studies to assess the influence of RT on the risk of secondary bladder cancer and the clinical endpoint in patients diagnosed with gynecological cancer (Wen et al.
2022b). In another investigation, Rombouts et al. examined the association between pelvic RT and the development of rectal cancer as an SPM. Their findings indicated an elevated risk of rectal cancer in patients who had previously received RT for pelvic cancer, a risk particularly pronounced in individuals treated for prostate and endometrial cancers (Rombouts et al.
2020).
The association between RT for gynecological malignant neoplasms (GMNs) and an increased risk of secondary malignant tumors at a population-based level remains a contentious issue. Consequently, this study aims to explore the impact of GMN radiotherapy on the risk of secondary malignant tumors. By utilizing the SEER database, we conducted a comprehensive analysis of SPM characteristics in GMN patients and further examined the risk factors for SPM occurrence at a population-based level. We also established a competing-risk nomogram visualization tool to aid clinicians in identifying GMN patients at a high risk of developing SPMs, thereby facilitating closer monitoring and timely treatment.
Materials and methods
Data sources
SEER database is the largest and most authoritative cancer database in North America (Yu et al.
2009), encompassing cancer data for nearly 30% of the population across diverse geographic regions of the United States. We extracted data for patients with GMN who underwent surgery between 1973 and 2015 using the SEER database and SEER*Stat software (version 8.4.0,
http://seer.cancer.gov/). Patients diagnosed with gynecological cancer included cervical cancer (site codes C53.0–C53.9), uterine and uterus cancer (site codes C54.0–54.3, C54.8, C54.9, and C55.9), ovarian cancer (site code C56.9), and other female genital cancers (site codes C51.0-C51.9, C52.9, C57.0–C57.9, C58.9). The follow-up period for SPMs commenced six months after the diagnosis of GMNs and concluded upon the diagnosis of any SPMs, death from any cause, or after 30 years of follow-up, whichever occurred first. We set the follow-up deadline as January 1, 2016. As the extracted data are publicly accessible and de-identified, approval from the Institutional Review Committee is not required in accordance with the Human Research Protection Office.
Data collection
In this study, nine variables were analyzed, including marital status, age, race, primary site, stage, differentiation grade, chemotherapy, tumor size (mm), and RT. Patients who met any of the following criteria were excluded: (1) non-histological diagnoses; (2) autopsy or death certificate diagnosis; (3) no history of surgical resection; (4) age less than 18 years old; (5) no racial information provided for the patient; and (6) no complete prognostic information available. To establish and validate the nomogram, we divided enrolled patients into a training cohort and a validation cohort randomly. The detailed flow chart is illustrated in Fig S1. The final study sample comprised 109,537 patients. Using a 7:3 ratio, we assigned GMN patients to either a development group (n = 76,675) or a validation group (n = 32,862) randomly. The development group was employed to determine risk prediction factors and construct models, while the validation group was used for internal model validation.
Statistical analysis
Categorical variables were compared as percentages using the Fisher's exact test and Chi-square test. In this study, the Mann–Whitney test was employed to analyze continuous variables with both normal and non-normal distributions. SEER*Stat 8.4.1 was used to estimate standardized incidence rates (SIRs), and SIR represented the ratio of observed SPMs to expected cases in the general population of the United States. Results were stratified by radiotherapy, age, and calendar time. The cumulative incidence of SPM development was evaluated using Fine-Gray competitive risk regression analysis. We utilized the Kaplan–Meier curve to depict the survival characteristics of patients with GMNs. A multivariate Fine and Gray proportional competing risk model was employed to identify relevant risk factors for SPM occurrence, subsequently constructing a risk prediction nomogram. The nomogram was verified by evaluating its discrimination and calibration capabilities using internal (training) and external (validation) sets, respectively. To assess prognostic accuracy, receiver operating characteristic (ROC) curves and calibration curves based on time and area under curves (AUCs) were generated at 5, 10, and 20 years.
Discussion
This study aims to explore the correlation between the development of SPMs following GMN resection and radiotherapy, as well as the subsequent impact on the prognosis of SPMs. The current data reveals that the cumulative incidence of SPMs in GMN patients who previously received radiotherapy is evidently higher than in those who did not undergo radiotherapy. Radiotherapy has been established as an independent risk characteristic for SPMs among GMN survivors. Additionally, this population-based cohort study identified the characteristics and risk factors of SPMs in GMNs.
An increase in cancer survivorship, the long-term adverse reactions of radiotherapy, advancements in early screening and diagnostic technologies, and the ongoing influence of risk factors have all contributed to a substantial rise in the incidence of various malignant tumors (Viyanant and Upatham
1988). An increasing amount of research is concentrating on the location of SPMs after the first primary malignant tumor. Ding et al. conducted a retrospective study on 11,017 patients with colorectal neuroendocrine neoplasms, discovering that the most common sites of SPM in males were the prostate, lungs (bronchi), rectum, and kidneys. In females, the most prevalent locations were the breasts, lungs (bronchi), rectum, and uterine body (Ding et al.
2023). Conversely, Warschkow et al. examined a cohort of 77,484 patients who had undergone resection of localized or locally advanced rectal adenocarcinoma, finding that prostate, breast, and lung cancers were the three most common types of SPMs (Warschkow et al.
2017). By analyzing extensive data from the SEER database, focusing on primary GMNs, it was determined that the three most frequent SPM sites in female GMN patients were the breast, colorectal, and lung (bronchus) regions. This underscores the importance of emphasizing cancer screening in these areas during GMN patient follow-up.
In summary, our study highlights that research on the risk factors for SPMs in GMNs is limited, but our analysis identified age, primary site, grade, stage, chemotherapy, and radiotherapy as significant risk factors for developing SPMs in GMNs. Among these factors, radiotherapy was found to be a key contributor to the occurrence of SPMs in GMNs. Our finding is consistent with past studies which reported a higher risk of developing SPMs in patients undergoing radiotherapy for other tumor types such as gynecological, rectum, and prostate cancers (Guan et al.
2021; Li et al.
2022; Moschini et al.
2019; Wen et al.
2022a). Generally, secondary cancers tend to occur in irradiated or adjacent areas, with the probability of random effects such as carcinogenesis increasing with the increasing dose of radiotherapy (Brown et al.
2010; Rombouts et al.
2018). The introduction of intensity-modulated RT (IMRT) has shown benefits in various tumor sites (Klem et al.
2008), as it allows for different doses to be delivered to different structures under the same irradiation (Wang et al.
2006). However, there is limited data supporting the relationship between the distribution of radiotherapy doses and the increased risk of secondary malignant tumors (Berrington de Gonzalez et al.
2015; Lonn et al.
2010). Our study also found that chemotherapy acted as a protective factor for GMNs to develop SPMs, consistent with previous research (Guan et al.
2021). Poor histopathology grading and distant metastasis were negatively related to the occurrence of SPMs in GMNs, which aligns with the results of Bateni SB et al. for patients with neuroendocrine tumors (Bateni et al.
2021). Our study also revealed that the presence of primary sites in the cervix reduced the risk of developing SPMs in patients with ovarian cancer. This difference may be attributed to the distinct clinical pathological characteristics of different primary organs (Lazzaroni et al.
2022), further emphasizing the diversity of different primary lesions.
This study presents several advantages and limitations. The advantages include the use of the SEER database, which provides detailed clinicopathological information about GMNs, and the construction of a competitive risk nomogram based on the multivariate Fine and Gray proportional sub-distribution risk model. This nomogram serves as a practical tool for improving guidance, monitoring, and managing GMN survivors and is the first model to assess the probability of SPMs occurring in survivors of GMN at 5, 10, and 20 years after the first diagnosis. The effectiveness of the nomogram was validated using ROC curves and calibration curves, allowing for proactive screening methods and close follow-up for potential SPMs in individuals with GMNs with a total risk point of 208 or above.
However, there are some limitations to consider. Firstly, there is a selection bias in the retrospective study, and future prospective studies will be necessary to verify the nomogram. Secondly, the occurrence of SPMs may be influenced not only by radiotherapy but also by other key risk factors, including lifestyle, environmental factors, genetic background, and other treatments that are not considered in our work owing to the unavailability of relevant data in the SEER dataset. Lastly, the SEER dataset only records the initial treatment strategy of GMN, and it is unclear whether GMN patients will receive delayed radiation exposure in further treatment, which could lead to misclassification of patients in the RT group as non-RT group. Despite these limitations, the main conclusion of our study remains valid, although the increased risk of RT may be underestimated.
Conclusion
Utilizing a large population from SEER database, our work identified primary site, age, grade, stage, chemotherapy, and radiotherapy as independent risk factors for SPMs in GMN patients. These characteristics were found to be correlated with the progression of SPMs in GMN patients. We constructed a competitive risk nomogram based on a competing-risk model to predict the 5-, 10-, and 20-year probabilities of SPMs in GMN patients, demonstrating strong predictive capabilities.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.