The disease progression of PDAC is extremely rapid that many patients cannot undergo surgery because they are in the advanced stage with loss of 98% healthy life expectancy at the time of diagnosis [14]. Multiple studies have proposed the concept of early recurrence to characterize recurrence that occur within few months of pancreatic cancer resection; however, they did not reach an agreement on cut-off values (varying from 6 to 12 months) and ignored the certain group that might have showed relapse before the first time of their follow-up or postoperative adjuvant therapy. The concept, hyperprogression HPD was mostly referred to in the field of immunotherapy, however, growth-stimulating effects from surgical trauma-induced immunosuppression were also mentioned in recent studies [5, 15]. The inflammatory response played a critical role in tumor invasion, progression, and metastasis by promoting tumor angiogenesis and decreasing anticancer effects [15]. According to previous studies, systemic response to tissue damages led to surgeries and the subsequent wound healing triggered a cascade alteration in cellular immunity. The high level of circulating damage-associated molecular patterns induced upregulating inflammation. This surgery-induced immunosuppressive status, which might last from several days to 6 months, was related to cancer outcomes since it created a window for cancer cell proliferation and dormant cancer cell revival, resulting in rapid recurrences [5]. In presented study, the total incidence of complications is highest in HPD group, though without statistical difference, which might indicate an increased trend of postoperative inflammation level in HPD patients.
In this study, we proposed the concept of PO-HPD in PDAC to characterize extremely rapid recurrence that occurred within 2 months following curable-intent resection of PDAC. The presented results demonstrated that 73.6% patients suffered recurrence following resection, of which 14.1% showed relapse within 2 months. The prognosis in PO-HPD group was extremely poor with 1- and 2-year survival rates at 36.6 and 5.9%, respectively, and median OS of 9.8 months, which is shorter than that of ER (14.7 months) and LR (30.8 months) groups. Considering the median survival time of unresectable pancreatic cancer that is reported to be 15–17 months in some clinical studies of non-surgical therapy [16, 17] and the median total survival time of pancreatic cancer patients at stage-IV disease from survival data collected from SEER is calculated to be 10.7 months, the unfavorable outcome suggested that PO-HPD patients rarely derived improvement in survival from surgery. In addition, pancreatectomy is one of the most challenging types of surgery with high morbidity and mortality due to its technical difficulty. Serious complications not only prolonged hospital stay, but also decreased patients’ quality of life, leading to delay or intolerance of systemic therapy [18, 19]. Evidence from recent clinical trials have stated a non-negligible position of systemic therapy for PDAC, and in view of the unsatisfactory survival benefit and severe surgical trauma-induced immunosuppression status, the role of pancreatic surgery for PO-HPD patients should be prudently assessed.
We identified four independent preoperative risk factors for PO-HPD group, consisting of large tumor size, elevated CA19–9 and CA125, and anemia. The AUC of final logistic regression was 0.739. Tumor size was confirmed to be a significant predictive factor of ER, and 3.0 cm was recommended as an optimal cut-off in some studies [9, 17], with the definition of ER varying from 6 to 12 months postoperatively. ROC curve and associated AUC analysis in our study revealed a similar optimal tumor size threshold of 3.45 cm for prediction of PO-HPD.
Despite approximately 5 to 10% population with no or scarce secretion [20], CA19–9 was considered the most studied and well-known biomarker for PDAC [21]. Several studies have demonstrated that increased preoperative CA19–9 was associated with short post-pancreatectomy DFS and decreased life expectancy. Previous studies have explored the threshold of CA19–9 for early recurrence prediction but did not attain consensus. Viencent P. et al. [9]’s study found a preoperative CA19–9 of > 210 U/mL as optimal cut-off to predict recurrence within 12 months while Kim et al. and Sugiura et al. reported favorable predictive capability of preoperative CA19–9 > 100 U/mL for recurrence within 6 months. Another multi-center study which analyzed resectable PDAC patients identified preoperative CA19–9 > 300 U/ml as predictive risk factor for recurrence within 6 months. Similarly, elevated CA19–9 was regarded as a risk factor of PO-HPD in this study and the cut-off was set at 288.6 U/mL according to ROC curve analysis, with an AUC of 0.628.
CA125 was employed as a biomarker for numerous cancers, especially for ovarian cancer, and its serum level would not be influenced by serum bilirubin levels [22, 23]. Elevated CA125 was observed in approximately 45% patients with pancreatic cancer [24]. Studies by Xianjun Y [20] found that CA125 was a potential biomarker in Lewis-negative patient with pancreatic cancer, and studies by Chan A et al. [25] found integrating CA125, CA19–9, and LAMC2 in one panel could improve the detection ability of CA19–9, implying CA125 could serve as a supplement for CA19–9 in pancreatic cancer monitoring. However, few studies found correlation between preoperative CA125 and recurrence. The results of this study verified preoperative CA125 ≥ 22.3 U/mL as a risk factor of PO-HPD.
Preoperative anemia (RBC < 3.94 × 109/L), seldom mentioned in other studies, was an independent risk factor of PO-HPD. The explanation might be that preoperative anemia patients were more probable to undergo perioperative blood transfusion. In this study, the rates of blood transfusion for patients with RBC < 3.94 × 10 [9]/L was 79.9% (258/323), and 71.8% (469/653) for patients with normal RBC count. Although few randomized trials explored the correlation between blood transfusion and cancer relapse, it was implied in earlier retrospective studies that perioperative blood transfusion, especially allogeneic blood transfusion, was associated with poor postoperative prognosis in several kinds of solid cancers including pancreatic, liver, colorectal and prostate, and head and neck cancers [26,27,28,29,30]. Evidence derived from a Cochrane Group meta-analysis yielded an odds ratio of 1.42 for the effect of perioperative blood transfusion on cancer recurrence; however, the author did not establish a clear causal relationship considering the heterogeneity [31].
Besides preoperative factors, we explored the molecular features of PO-HPD patients. Previous studies reported KRAS, TP53, SMAD4, and CDKN2A as the four major driver genes associated with poor prognosis of pancreatic cancer [32]. In the present study, we observed higher mutation frequencies of these four genes in PO-HPD without significant difference, which might be attributed to the small sample size. It was unexpected that mutations of CEBPA, ATR, and JAK1 were only identified in PO-HPD patients. The regulation of CEBPA accelerated cancer progression by disrupting circadian rhythm-signaling pathway [33]. Jiren Zhang et al. [34] proposed a risk score system for pancreatic adenocarcinoma based on CEBPA and other 11 methylation genes (HIST1H4E, STAMBPL1, PLD3, CEP55, SSBP4, GRIA1, SWAP70, ADCYAP1R1, YPEL3, HOXC4, and IGFBP1), suggesting that CEBPA might be critical for the survival of PDAC. ATR was identified to be involved in homologous recombination repair, and its mutation might lead to homologous recombination deficiency (HRD) [35]. JAK1 was known to drive cancers with microsatellite instability (MSI) and was often seen in mismatch repair deficient (MMRD) PDAC [36]. The certain mutation distribution pattern of PO-HPD offered alternative therapeutic options for patients that might rarely benefit from surgery, considering that the evidence from preclinical experiments and phase-II clinical trials suggested sensitivity to poly (ADP-ribose) polymerase (PARP) inhibitor of pancreatic cancer patients with HRD [37, 38], and guidelines recommended pancreatic cancer patients with MSI as candidates for immune-checkpoint inhibition [39].
CL Wen et al.’s study [40] proved high CNI as an independent predictive biomarker for early recurrence in PDAC patients. Our results showed higher CNI in PO-HPD patients than ER + LR group, but was not statistically significant. PO-HPD patients showed lower level of CN gain, which was not mentioned in previous studies and need further exploration.
Both in PO-HPD and ER groups, more than half of the patients showed systemic recurrence, supporting the hypothesis that occult micro metastases existed before surgery. It was suggested that the timing of recurrence was important for OS, and systemic therapy was of potential importance for patients at high risk of rapid recurrence [3, 9, 41, 42]. Recent studies have emphasized the role of neoadjuvant therapy in borderline/locally advanced PDAC [43,44,45] and confirmed that neoadjuvant therapy allows initial treatment of occult metastases, downstage large tumors, and improves rates of negative margin, thereby prolonging life expectancy in patients with advanced disease [42]. The results from a meta-analysis comparing upfront surgery with neoadjuvant treatment in patients with resectable or borderline resectable pancreatic cancer emphasized that neoadjuvant treatment appeared to improve OS, which is in accordance with the current result from ESPAC-5F trial (Four arm, international randomised phase II trial of immediate surgery compared with neoadjuvant gemcitabine plus capecitabine (GEMCAP) or FOLFIRINOX or chemoradiotherapy (CRT) in patients with borderline resectable pancreatic cancer) that neoadjuvant therapy group showed a significant survival advantage at 1 year (77% vs. 42%) [46].
This study has several limitations. First, CA19–9 could be affected by serum bilirubin level and 29% patients in our study had elevated bilirubin, which might detriment the accuracy of CA19–9. Second, the retrospective nature of the study may have induced bias. Third, we excluded few patients with complete clinical data, and the sample size for genomic investigation was small, which might limit the generalizability of our conclusion.