Nomograms use a straightforward graphical presentation to provide individual predictions of survival outcomes and have grown as a key part of recent decisions [17]. Recent studies have confirmed a higher predictive power capacity of nomograms for various cancers than the traditional AJCC TNM staging system [18–20]. Lymphoma is the most commonest primary malignant tumor of the orbit [21]. POAL represents about 8% of all extranodal lymphomas and is most frequently observed in elderly patients [22–23]. Compared to younger patients, older patients have a significantly low survival rate [10]. Additionally, the suitable treatment for young and old POAL patients is not the same [24]. Therefore, it is critical to create specific nomograms to distinguish elderly POAL patients into prognostically distinct groups or to predict a categorical staging system. Herein, we established OS and CSS prediction models integrating only four predictors. Their excellent performance was confirmed by ROC and calibration curves.
To our knowledge, there is only one nomogram for POAL patients. Peng et al. constructed and validated a POAL-specific nomogram for CSS and OS [10]. The risk factors included in their models were treatment strategy, pathological type, laterality, primary site, marital status, year of diagnosis, gender, and age. Our study analyzed 827 POAL patients aged 60 or over. After univariate and multivariate Cox regression analysis, nomograms were successfully established using independent age, surgery treatment, M stage, sequence number, and primary site.
The significant connection of cancer to aging was considered as the center of research, as aging represents the single most significant predictor for most cancers [25–27]. In the present study, age was a vital risk factor in predicting OS and CSS. Although only patients aged 60 or over were included, the points of adults aged ≥ 87 were significantly higher than 60–86, indicating that being older significantly increases the risk of death. We need to pay more attention to older patients with POAL. Hence, more elderly-specific research such as ours is needed.
Lymphocytes comprise B, T, and natural killer (NK) cells, for example, and have been closely associated with tumor immunity [28]. Most lymphoma cases are of B-cell origin; the others are NK cell or T-cell lymphomas. As a rare type of non-Hodgkin lymphoma, NK/T cell lymphoma presents worse clinical outcomes [29]. Few studies have focused on the effects of the grade associated with survival in POAL because of the rarity of NK/T cell POAL patients. Although the vast majority were B-cell in this study, the grade was essential in POAL patient survival, especially in CSS. POAL derived from B-cell might have a better prognosis than T- or NK-cell. Observations from our study include that scholarships should pay more attention to differentiating the cell origin of POAL, which might dictate treatment selection for elderly patients.
The differentiation of histological types is also crucial for diagnosing lymphoma, as different lymphomas exhibit different behaviors and outcomes [30]. Along with previous reports, our current results showed that the main composition of POAL was MALT, followed by DLBCL [31–33]. MALT lymphoma is a type of B-cell in the marginal zone of the marginal location, resulting from a chronic proliferation of lymphocytes [34]. This tumor is usually painless, slow-growing, and has a good prognosis [6, 35]. DLBCL is common in older adults and has a considerably poor prognosis [36]. Moreover, FL is a kind of indolent B-cell lymphoma, much less malignant than others [37].
Similarly, we revealed that DCBCL patients might have higher survival rates and shorter survival times than others. MALT patients represented an intermediate prognosis, and FL patients exhibited the best prognosis among the main histological types. Various new immunological and molecular techniques can help distinguish histological types [3]. Therefore, it is crucial to identify the histological type of elderly POAL patients to make a suitable clinical decision as early as possible.
As the main characteristics of tumors, the invasion and metastasis are significantly associated with the survival of patients [38]. The metastatic and M stages were considered as the risk factors of the survival in old POAL patients and independent risk factors. Regarding the M stage, patients classified as M1 at the first consultation might have poorer prognoses than those classified as M0, consistent with other studies conducted on tumors [39–40]. As the POAL progressed, a few patients had a regional or distant tumor. The OS rate of the regional tumor was higher than localized and distant ones. However, this conclusion is still controversial. Many studies have shown that patients suffering from multiple tumors have worse long-term survival rates than those with a single tumor [41–42]. Furthermore, we demonstrated that sequence number was a significant predictor of CSS and OS in elderly POAL patients. Patients with only one primary tumor might have a substantially greater outcome than those with more than one primary tumor. This topic has received little attention until now in the POAL group but deserves more attention in the future.
The primary site was another predictor influencing the tumor's outcome. POAL is found in the eye in various locations, including the orbit, lacrimal gland, conjunctiva, and eyelids [6]. Only four primary sites were included in our analysis due to data incompleteness. The sites were divided into two groups: ocular (eyelid, conjunctive and lacrimal gland) and orbital. In the categorization, the orbital was more likely to occur than the ocular. Moreover, a significant difference was found in CSS for elderly POAL patients. So far, there is still a lack of unified therapy for POAL. Surgical excision, radiation therapy, chemography, or a combination of these modalities are the most used treatments for POAL patients. Our results showed that approximately half of the patients received radiation treatment. Radiotherapy for ocular adnexa patients achieved reasonable local control with acceptable toxicity [3], and the dose was associated with the survival of POAL patients [43]. However, only the surgery receipt was considered as an predictive factor of OS independently. Patients tended to have a better prognosis after receiving surgery in this research. This result demonstrated that surgery is an essential procedure in treating POAL. Based on the SEER database, patients who did not know if they had adopted treatment and those who had not received it were in the same group, and this lack of definitive grouping might have influenced the results. Therefore, more research is required to understand better the associations and impact of the treatment strategy on patient outcomes.
Recently, nomograms have been commonly applied for evaluating cancer prognosis [44]. As far as we know, this was the first nomogram built on an extensive population database to predict the survival of elderly POAL patients. The calibration and DCA curves revealed good consistency and the potential to be applied in clinical practice. By applying these prediction models to POAL elderly patients, a distinct separation for survival was observed between the K-M curves of high-risk and low-risk patients. In routine clinical investigations, TNM staging is typically used to predict the survival of cancer patients. The AUCs of our models were higher than TNM staging for OS and CSS, showing their better performance. Hence, our nomograms might effectively supplement the TNM staging system. Additionally, we built web-based nomograms that clinicians can easily use to provide a quantitative prognosis for each patient.
Nevertheless, our research also has some shortcomings. First, this research might have a risk of selection bias since we used retrospective data from the SEER database. Additionally, because of the lack of data, some significant factors, such as tumor size or genetic mutations, were not specific. Besides, the risk factors in our study were censored a lot as the AJCC staging system data was blank. Considering the limitations above, incorporating other risk factors is encouraged to improve this model in future research.