In this study, we develop and validate a nomogram predicting 1- and 2-year OS and 1-year PFS for patients with LAPC receiving SBRT and chemotherapy. Notably, this pilot study focused on a major population of pancreatic cancer and evaluated outcomes of an intensified treatment modality with hypofractionated radiotherapy and standard chemotherapy. Additionally, the nomograms were beneficial for patient-specific estimated of OS and PFS that can be used for stratifications of survival probability and assessment of prognosis with patients.
Although previous studies have investigated nomograms for advanced pancreatic cancer [7, 8, 20, 21], chemotherapy regimens in most studies were heterogeneous, which may influence outcomes [7, 20, 21]. Furthermore, patients in those studies underwent chemotherapy alone other than combination therapy, which may provide better local control and more survival benefits [20, 21]. Besides, no external validations were performed in two nomograms [7, 21]. Additionally, Vernerey et al. included patients of LAP07 as the development cohort, where patients either received chemoradiotherapy or chemotherapy [8]. Moreover, three-dimensional conformal radiation therapy was applied in this study, which was not a mainstay modality of radiotherapy in the case of intensified modulated radiotherapy or SBRT commonly used in LAPC [8]. Most importantly, SBRT has already been recommended in the management of pancreatic cancer according to American Society for Radiation Oncology clinical practice guideline for pancreatic cancer [22]. Hence, evaluations of patients’ outcomes after standard chemotherapy regimens with the novel radiotherapy technique are required.
Compared with those published nomograms, our internally and externally validated nomograms have some similarities and potential advantages. Several predictive factors in our nomogram were common with them, including performance status, tumor size, radiation dose and surgical resection. However, our larger sample size of LAPC, longer follow-up and prospective data collection allow us to investigate contributions of other factors to survival and develop separate nomograms for 1- and 2-year OS and 1-PFS, while most of the studies only predicted short-term OS [7, 8, 20, 21]. In addition, clinical net benefits were estimated by decision curve analysis and patients were classified into four groups of different OS and PFS probability with recursive partitioning analysis, which may be in favor of accurate personalized evaluations of prognosis and decision making of treatment. Unfortunately, this had not been discussed in previous studies.
It was demonstrated that CA19-9 level correlated with survival of patients with pancreatic cancer, which was confirmed in published nomograms where CA19-9 baseline level was included as a predictor [7, 8, 20]. However, dynamic changes of CA19-9 level after treatment are usually considered as the surveillance of pancreatic cancer. A significant decrease may indicate effective treatment while a dramatic increase may imply disease progressions, which may be more sensitive than imaging examinations. Therefore, CA19-9 response was included in our nomogram. It was clarified that significant decrease of CA19-9 baseline level after treatment was predictive of better outcome. As a result, it suggested that patients could still achieve survival benefits despite high pre-treatment CA19-9 levels if they responded well to treatment. This may result in a discrepancy between previous and our nomograms. For the former ones, patients with elevated baseline CA19-9 levels were all stratified into inferior survival groups, which may reduce the predictive accuracy of prognostication for an individual patient.
Furthermore, we identified that some of patients in our study become candidates for surgical resection after chemoradiotherapy, which could lead to longer OS. This was only proven in a previous study, where chemoradiotherapy was delivered [7]. However, for the rest studies with systemic therapy alone, surgical resection was not performed and included in the nomogram [8, 20, 21]. In a previous study with review of The National Cancer Data Base, it was demonstrated that the addition of radiotherapy to neoadjuvant chemotherapy was associated with higher downstaging and complete pathologic response rates than chemotherapy alone [23]. Additionally, there was about 6.0-14.4% downstaging rate after chemoradiotherapy for LAPC [23–25], which was similar with that (9.3%) in our study. Nevertheless, chemoradiation did not confer a survival benefit compared with chemotherapy alone [23]. This may be ascribed to conventional fractionation of radiotherapy and relatively low dose. Hence, chemoradiation should be considered as multimodality treatment for LAPC but requires further investigations.
In our nomogram, a high dose may provide favorable OS and PFS, which was proved in previous studies [14, 26, 27]. Similarly, the radiation dose was also a predictor in the nomogram in a published one [7]. The median OS and 2-year OS rates of patients with BED10 > 70Gy in our center (data not published) were consistent with the ones in Krishnan et al [27] (median OS: 20.3 months vs. 17.8 months, 2-year OS rate: 36.8% vs. 36.0%, 3-year OS rate: 18.7% vs. 31.0%). However, there were controversial results from the meta-analysis [28]. The median OS of two included studies with the most weight employing BED10 > 70Gy was 12.5 months and 10.3 months, while the median OS with the most weight employing a BED10 of 60-70Gy from another two studies were found to be 13.9 months and 15.0 months, respectively. The controversial result could be attributed to the different radiosurgery platforms and chemotherapy regimens. Hence, the interpretations of the results may not negatively impact clinical practice with dose escalation as a therapeutic paradigm for LAPC.
We acknowledge several limitations. First, patients in our study did not receive FOLFIRINOX, which may be given the first priority in chemotherapy regimens. However, patients receiving FOLFIRINOX may experience more severe adverse effects than gemcitabine-based chemotherapy. Furthermore, S-1 is not recommended in the guideline but it has been proven effective in Asians [3–6]. Therefore, the nomogram could not be used for patients with FOLFIRINOX. Second, molecular profiles were not available for much of the study period and could not be investigated as predictors of outcomes. Nonetheless, our recent study showed that high signal intensity of Ki-67, P53 and PD-L1 were associated with worse prognosis [17]. Hence, these signatures would be assessed and included in our next-generation nomogram. Moreover, we cannot exclude all possibility of residual confounding after internal validation as a result of possible overfitting from variable and threshold selection for these models. However, internal validation with bootstrapping and external validation could address the concerns.