Primary bone lymphoma (PBL) is a relatively uncommon type of neoplastic disease, accounting for approximately 1% of all lymphomas, 3-7% of extranodal lymphomas, and 7% of principal malignant bone tumors (1–3). And the estimated 5-year survival rate (OS) for PBL patients has been reported to fall within a range of approximately 55%-75% (4, 5). Non-Hodgkin lymphoma (NHL) constitutes the predominant category of PBLs with diffuse large B-cell lymphoma (DLBCL) accounting for more than 80% of all instances (2, 6). PBL can occur across all age groups, with a prevalent diagnostic age range of 45-60 years, and males are predominantly affected compared to females (2, 7, 8). PBL of the spine is an uncommon entity, typically arising as metastatic disease from additional sites (2). The clinical manifestations of spinal PBL are commonly non-specific, often resulting in a diagnostic delay, and usually necessitating a biopsy for a definitive diagnosis (5, 9–11). Owing to the scarcity of the PBL of the spine, there is a scarcity of comprehensive data on the optimal treatment strategy and prognosis. The prevailing modalities encompass chemotherapy or immunochemotherapy with or without radiotherapy; while surgical intervention are generally restricted (2). Furthermore, the late complications after chemotherapy in such patients are currently under scrutiny, including bone loss and the development of fragility fractures (12). And survival outcomes of these patients are yet to be investigated.

With the progression of medical technology, the substantial augmentation in the proportion of tumor patients with enduring survival heightens the necessity to comprehend the dynamic survival probability. Traditional evaluations intended for patient survival prediction, which are frequently from the time of diagnosis, may unable to present updated survival information for individuals who have survived several years post-diagnosis. As a varietal of extensively utilized statistical methodologies, conditional survival (CS) analysis offers an immediate recalculation of survival outcomes, delineating the evolution of survival probabilities across different time periods (13). And these dynamic and precise prognostic data can aid in optimizing the clinical management and follow-up strategies, facilitating enhanced physician-patient communication, and providing psychological support for these patients, which hold significant clinical value.

On the other hand, accurate prognosis evaluation necessitates integrating the clinicopathological characteristics and therapeutic status of patients. Despite traditional nomogram’s ability to personalize prognosis prediction via taking individual characteristics into account (14, 15), this methodology remains limited in its capacity to delineate the dynamic change in survival over time (16). Thus, incorporating CS within a nomogram holds potential for customized, timely and dynamic prognostic prediction.

Due to its rarity, our comprehension of PBL of the spine predominantly rests on case studies, with study in a substantial population remains deficient. Hence, in this current study, utilizing the Surveillance Epidemiology and End Results (SEER) database (17), we aimed to investigate survival outcomes of patents with PBL of the spine and to describe the CS profile of these patients at a population-based level. And we also developed and validated a novel CS-nomogram model for dynamic survival prognosis quantification for spinal PBL patients.

While several articles have explored spinal intramedullary lymphomas (20), research specific to PBL of the spine remains substantially lacking. This is the first study to analyze the CS trend of spinal PBLs, and we observed that that the instantaneous survival rates of PBL of the spine progressively augmented over time. Meanwhile, through three different prognostic factor selection methods, we identified the best predicter subset for survival prediction model construction. Finally, we constructed and verified a unique CS-centered nomogram model for instantaneous and dynamic survival evaluation. Furthermore, based on this nomogram, we further designed a risk stratification system, thereby enhancing the efficacy of therapeutic strategies for these patients.

PBL typically portends a favorable prognosis (2, 10). Indeed, it is deemed to have the most favorable prognosis among all primary malignant bone neoplasms and better prognosis than secondary bone lymphoma (1, 21). And it is reported that tumors within the spinal column demonstrated substantially diminished survival compared to those in extremity bones, potentially due to nerve compression and the ensuing complications of this phenomenon (9), indicating that this subset of PBL requires further independent analysis. Instead of traditional survival analysis, we employed CS analysis to describe the survival pattern of spinal PBL and demonstrated that survival rate of PBLs of the spine progressively elevated over time with the 5-year survival probability evolving from 63% at the original diagnosis to 66.4%, 84%, 89%, 93% and 96% annually. This real-time feedback of survival and prognostic data will contribute to augmenting the survival reassurance of patients, mitigating psychological stress, and positively influencing the patients’ prognosis. In addition, we integrated CS into a nomogram model which considered the patient’s individualized clinicopathological factors to establish a novel CS-based prognosis prediction model with the ability of dynamic and personalized survival estimation.

Univariate Cox Regression, LASSO and BSR were selected as potential variable screening methodologies. These methods were selected to guarantee a comprehensive and robust examination, enabling us to address the complexity of our data effectively and build a model with more stable and excellent performance. Through a series of screening, our model finally utilized age, tumor histology, tumor stage, chemotherapy and marital status as predictive factors, demonstrating substantial correlation with prognosis. Advanced age and tumor stage represented significant and unfavorable prognostic indicators for survival among patients affected with PBL (2, 8, 11, 22, 23). Chemotherapy is the optimal therapeutic approach utilized for PBL patients, with rituximab in combination with anthracycline-based chemotherapy presently being endorsed (9, 22, 24). And our study also found that chemotherapy constituted the paramount factor influencing the outcomes. Moreover, the efficacy of surgical intervention in the management of PBL remains limited, and it usually applied for diagnostic purposes (2, 9, 24, 25). Despite the observation that our study failed to uncover a significant correlation between surgery and PBL prognosis, the therapeutic merit of surgical intervention warrants further exploration in relation to improving spinal PBL patient’s quality of life. Furthermore, the efficacy of radiotherapy remains contentious in the management of PBL patients. Some studies have indicated that radiotherapy may augment the local tumor control rate, thus potentially enhancing patient survival (9, 26); while some research reported that consolidation radiotherapy cannot improve PBL outcomes after receiving chemotherapy (22, 27, 28). A unanimous consensus has not yet been established for radiation treatment; hence, an intensified exploration of PBL subsets suitable for radiotherapy is indispensable. And our risk stratification system could aid in identifying individuals at elevated risk, facilitating exploration into potential applications of radiotherapy.

In the clinical practice, by inputting individual patient characteristics into our nomogram, clinicians can generate specific risk assessments that forecast outcomes with superior accuracy compared to traditional techniques. This precision permits a more refined understanding of each patient’s prognosis. Additionally, through our risk classification system, patients identified as high-risk may benefit from more aggressive therapies, whereas low-risk individuals might be exempted from superfluous treatments, thereby reducing potential adverse effects and optimizing resource utilization. And by providing a visual representation of risk, these tools can help demystify complex medical information, making it easier for patients to understand their prognosis and treatment options.

Our study was subjected to certain limitations. Firstly, this retrospective research was inherently biased. Secondly, the SEER database was devoid of imperative details, including the precise chemotherapy regimen, surgical procedure and radiotherapy dose, thereby potentially restricting our analysis. Thirdly, we failed to obtain sufficient cases for our model external validation. Fourth, while the study presents a novel CS-based nomogram for PBL of the spine, exploring the external validity of this model in different clinical settings could further validate its applicability. Future investigation could encompass multi-center studies to assess the performance of the nomogram within varying populations and healthcare systems. This method would aid in ascertaining any requisite modifications to assure its utility across various clinical contexts. And The progression of therapeutic modalities for PBL of the spine necessitates ongoing modifications to our model. Furthermore, employing more advanced machine learning methodologies can facilitate real-time modifications to the model predicated upon existing therapeutic protocols.

In summary, based on the SEER database, this study estimated the CS trend of PBL of the spine and demonstrated that the real-time survival rate of PBL of the spine dynamically increased over time. We also successfully established and validated a novel CS-based prognosis prediction model with strong performance for survival estimation. Furthermore, a nomogram-based risk classification system was also designed for risk stratification for these patients, thereby aiding in optimization of clinical management of this disease. Future research demands further advancement in the formulation of predictive models with superior quality and performance for extensive applications, facilitating more meticulous therapeutic and follow-up strategies customization for patients afflicted with these rare tumor types.

The original contributions presented in the study are included in the article/ Supplementary Materials , further inquiries can be directed to the corresponding author/s.

Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the patients/participants or patients/participants’ legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.

GZ: Formal analysis, Methodology, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. ZY: Formal analysis, Validation, Writing – review & editing. HQ: Validation, Writing – review & editing. HH: Methodology, Writing – review & editing. ZG: Methodology, Supervision, Validation, Writing – review & editing.