The choice between observation, systemic combination therapy, and aggressive surgery should be based on an accurate assessment of tumor size, function, location, and degree of invasiveness. It should not be made solely based on a critical dimension of 2 cm. According to the results of the RF model, aggressive surgical treatment should be preferred if the tumor is assessed to be at risk of invasive and malignant transformation after opting for conservative treatment in patients with small NF-PNET.
The optimal management of NF-PNETs smaller than or equal to 2 cm remains challenging, not least due to the heterogeneity of the biological and clinical characteristics of these tumors. Although these tumors are usually considered to be of low malignant potential, the optimal management of these tumors still needs to be clarified and consistent due to the lack of sufficient evidence and well-defined treatment options(18, 19).
Preoperative tumor size has been proposed to predict malignant potential and aid clinical decision-making. However, the European Neuroendocrine Tumor Society (ENETS) consensus guidelines and the National Comprehensive Cancer Network (NCCN) guidelines suggest a conservative, watchful waiting strategy for patients with low-grade tumors less than 2 cm and emphasize the importance of a thorough assessment of individual patient characteristics (4, 20). Lymph node metastasis, distant metastasis, and disease recurrence have been documented in small tumors, suggesting that even with small tumor size, the malignant potential cannot be ignored(21, 22).
Therefore, more than relying solely on tumor size as the sole indicator for treatment decisions may be required. In addition to tumor size, other clinical characteristics such as individual patient characteristics, comorbidities, and other survival risk factors must be considered(23). In our study, many clinicopathological variables were associated with overall survival in NF-PNETs smaller than 2 cm, including age, gender, race, grade, tumor number, and tumor stage. Therefore, clinicians should consider these risk factors as references when making decisions about observation or surgical resection.
There is some inconsistency in the results of multiple retrospective analyses(24–26), possibly due to relatively small sample sizes and lack of multicentre prospective studies. In this study, we found that patients who underwent surgical resection of the primary site were strongly associated with an OS benefit. To further validate the reliability of this result, we also performed a multivariate Cox analysis, adjusted for all covariates. We observed consistent results, suggesting that the impact of surgical treatment on patient prognosis is trustworthy and reliable.
Numerous treatment guidelines have been published internationally for small NF-PNETs (4, 20, 27, 28), in which conservative observation is usually recommended as a treatment option. Patients may be hesitant to undergo primary tumor resection, considering their age, health status, preferences, and values. To promote patient compliance, the study hypothesized that patients would benefit from non-surgical treatment if they survived more than 15 months after the observation period, which is the lower limit of the 95CI% for median survival time. The study then used four different machine learning methods (LR, RF, SVM, KNN, AdaBoost, and MLP) to predict which patients would be suitable for surgical resection based on independent risk factors affecting prognosis. Upon comparison, the RF model performed better than LR, SVM, KNN, AdaBoost, or MLP. This means that the RF model can more accurately predict whether non-surgical patients need more appropriate treatment options and can more accurately differentiate between patients who need more appropriate treatment options and those who do not. Thus, RF can be a valuable tool for predicting which patients will likely benefit from an actively observed treatment modality. More validation and optimization of these machine learning models is needed to ensure they are sufficiently reliable and stable to achieve better results in clinical practice. For patients with small, non-functioning neuroendocrine tumors under active observation, enhanced monitoring and reassessment of the need for surgical treatment at 15 months is a reasonable strategy. This individualized monitoring and decision-making process can help identify patients who may benefit from surgical treatment and reduce the unnecessary risks and burdens imposed on those who do not require surgery.
This study inevitably has several limitations. Firstly, the design of the retrospective analysis made it impossible to control for coding errors, data incompleteness, and inaccuracies in the data sources. This may have led to the omission or misclassification of some essential variables, affecting the accuracy and reliability of the results. Second, some accepted prognostic parameters, such as the Ki-67 index, CgA, and surgical complications, may be missing from the SEER database. These factors may have a substantial impact on the assessment of patient prognosis and treatment outcome. Third, excluding patients with missing information on selected variables may lead to selection bias. Despite the above limitations, our study contributes valuable information for small, non-functioning pancreatic neuroendocrine tumors, given the breadth of demographic information in this study and the availability of long-term follow-up data. However, because of these limitations, we still need to conduct more large-sample, multicentre, prospective studies to validate these findings. Such study designs can better control for potential bias and provide more reliable data to support treatment decisions and guideline development.
According to the results of this study, patients diagnosed with small, non-functioning pancreatic neuroendocrine tumors demonstrated improved overall survival (OS) outcomes after undergoing surgical intervention, implying that surgical intervention may have a positive impact on survival in such patients. Reassessment and adjustment of treatment regimens after 15 months of expectant observation for small NF-PNETs in highly selected patients may be justified. Further, large-sample, multicentre, prospective studies and reliable data are needed to validate the benefits of such a proposal.