Comment on “Usefulness of artificial intelligence for predicting recurrence following surgery for pancreatic cancer: Retrospective cohort study”

Abstract

Dear Editor, A retrospective cohort study was recently published to investigate the effectiveness and feasibility of artificial intelligence (AI) in predicting recurrence of pancreatic cancer in patients after surgical resection [1]. This multicenter study, which recruited 4846 participants, found AI could significantly predict recurrence after surgery for pancreatic cancer patients, suggesting that AI can provide critical prediction in support of recurrence of pancreatic cancer in clinical practice. The publication of this study will undoubtedly help clinicians to better apply AI in clinical practice to further improve prognosis of patients with pancreatic cancer. However, the following issues need to be further interpreted: First, the purpose of this study as described by the authors was to predict recurrence of pancreatic cancer after surgery using AI. However, the preoperative treatment strategy is missing. Nowadays, although the mortality rate after pancreatic cancer resection has improved, the overall survival rate has not remarkably been improved. The majority of pancreatic cancer patients still die from local and/or distant recurrence [2,3]. Previous studies have revealed that the potential advantages of preoperative treatment for pancreatic cancer patients comprise of improving the surgical procedures of patients, increasing the rates of negative-margin (R0) resection and timely management of invisible micro-metastasis [4]. Furthermore, some studies have shown that a standardized preoperative treatment could significantly reduce the local recurrence rate in patients with pancreatic cancer when compared with adjuvant therapy [5,6]. A retrospective study of 309 patients with pancreatic cancer found that preoperative treatment was associated with significantly longer disease-free survival, lower local recurrence and lower liver metastasis, indicating that preoperative treatment should be used to reduce the recurrence rates of pancreatic cancer patients [7]. Although AI can be used as a novel method in predicting pancreatic cancer recurrence, important known prognostic factors in clinical practice should not be ignored. Second, as stated in the abstract that “The C-Index averages of the random forest and the Cox model were 0.6805 and 0.7738, respectively”. As the C-index was only 0.6805, the prediction ability is limited, which might be related to not using important serological biomarkers. Currently, actively looking for the most appropriate method in detecting cancer recurrence is a hot topic in clinical research. Conventionally, tumor sizes of greater than 1 cm could be detected radiologically, thus it is unlikely to detect tumor recurrence at very early tumor stages [8]. Previous studies have shown that a sudden increase in serum tumor biomarkers to be associated with occult cancer recurrence [9,10]. A study by Azizian and colleagues found that a 2.45-fold increase in CA19-9 value indicated recurrence of pancreatic cancer [9]. The sensitivity and specificity were 90% and 83,33%, respectively, and the area under the curve was 95%. Similarly, a meta-analysis demonstrated that the sensitivity and specificity of CA19-9 in predicting recurrence of pancreatic cancer were 0.73 (95% CI 0.66–0.80) and 0.83 (95%: 0.73–0.91), respectively [10]. All these evidences suggest that CA19-9 plays an essential role in predicting recurrence of pancreatic cancer. Thus, it is speculated that adding CA19-9 to this AI model will further improve the C indexes. Provenance and peer review Commentary, internally reviewed. Funding for your research None. Ethical approval This is a letter to the editor. Author contribution Yu Cai participates in study design and writing. Guarantor The Guarantor is the one or more people who accept full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish. Please note that providing a guarantor is compulsory. Yu Cai. Declaration of competing interest None.