Validated Nomogram Predicting 6-Month Survival in Pancreatic Cancer Patients Receiving First-Line 5-Fluorouracil, Oxaliplatin, and Irinotecan
Abstract
Background
FOLFIRINOX, a potent combination chemotherapy regimen comprising leucovorin, 5-fluorouracil, irinotecan, and oxaliplatin, has emerged as a significant therapeutic option for a select group of patients diagnosed with advanced pancreatic cancer. This includes individuals with metastatic pancreatic cancer (MPC), where the disease has spread beyond the pancreas, and those with locally advanced unresectable pancreatic cancer (LAPC), where the tumor is confined to the pancreas and surrounding major blood vessels but cannot be surgically removed. While FOLFIRINOX has demonstrated improved outcomes in terms of survival compared to other regimens in fit patients, a crucial clinical challenge remains: there is currently a lack of consistently reliable and objective criteria that can accurately identify which specific patients within this diverse group are most likely to derive substantial benefit and experience better clinical outcomes from this intensive chemotherapy. This absence of robust predictive markers makes personalized treatment decisions challenging and highlights a significant unmet need in the management of advanced pancreatic cancer.
Patients And Methods
In response to this critical clinical need, our study was designed as a comprehensive investigation to identify putative prognostic factors among a cohort of 137 patients diagnosed with either metastatic pancreatic cancer (MPC) or locally advanced unresectable pancreatic cancer (LAPC). All these patients had received treatment with the FOLFIRINOX triplet chemotherapy regimen. The primary endpoint for this investigation was the assessment of 6-month survival status, providing a crucial early indicator of treatment efficacy and patient prognosis. To determine the association of various clinical and laboratory parameters with this primary endpoint, multivariate logistic regression models were meticulously employed. This statistical approach allowed for the simultaneous evaluation of multiple variables, accounting for their potential confounding effects, and identifying those independently associated with survival at 6 months.
Building upon the insights gained from the multivariate analysis, a novel prognostic tool, specifically a nomogram, was constructed. This nomogram was designed to predict the individual patient’s risk of death within the first 6 months following the initiation of FOLFIRINOX chemotherapy. The nomogram was developed by assigning a precise numeric score to each identified prognostic variable, with these scores meticulously weighted based on their demonstrated level of association with survival, as determined by the logistic regression models. This systematic approach ensured that variables with stronger predictive power contributed more significantly to the overall risk score. To rigorously assess the reliability and generalizability of the constructed nomogram, external validation was performed. This crucial step involved testing the nomogram on an entirely independent dataset comprising 206 additional patients, ensuring that its predictive accuracy was not limited to the original development cohort. The entire study protocol, including its objectives and methodology, was formally registered at ClinicalTrials.gov under the identification number NCT03590275, ensuring transparency and adherence to ethical research guidelines.
Results
Through rigorous multivariate analysis, our investigation successfully identified four key variables that were significantly associated with the 6-month survival status of patients receiving FOLFIRINOX chemotherapy for metastatic or locally advanced unresectable pancreatic cancer. These identified factors were: the patient’s performance status, indicating their general well-being and functional capacity; the presence or absence of liver metastases, a critical indicator of disease dissemination; the baseline carbohydrate antigen 19-9 (CA19-9) level, a tumor marker commonly elevated in pancreatic cancer; and the neutrophil-to-lymphocyte ratio (NLR), an inflammatory biomarker derived from routine blood tests. These variables were either found to be statistically significant in the multivariate analysis or possessed sufficient clinical plausibility to warrant their inclusion in the developed prognostic nomogram, ensuring its practical relevance.
The accuracy and predictive power of the nomogram were subsequently confirmed through meticulous external validation in an independent cohort of 206 patients. The C-index, a statistical measure of discrimination that assesses the nomogram’s ability to differentiate between patients who will survive and those who will not, was calculated to be 0.762. This value, with a 95% confidence interval of 0.713 to 0.825, signifies a strong discriminative ability, indicating that the nomogram can reliably predict patient outcomes. Following the comprehensive analysis, which involved grouping all cases from both the development and validation cohorts, we successfully identified four distinct patient subsets based on the cumulative number of “poor prognostic features” present (0, 1, 2, or >2). Crucially, these subsets exhibited statistically significant differences in their outcomes (P < .0001), demonstrating the nomogram’s capacity to stratify patients into meaningful risk categories. This stratification provides a clear and actionable framework for understanding varying prognoses.
Conclusion
In conclusion, the nomogram meticulously constructed in this study offers a highly accurate and reliable method for predicting the risk of death within the initial 6 months following the commencement of FOLFIRINOX chemotherapy in patients diagnosed with metastatic or locally advanced unresectable pancreatic cancer. This innovative prognostic tool holds significant potential to be remarkably useful in multiple clinical and research contexts. Clinically, it could serve as an invaluable aid for healthcare professionals to guide more precise and empathetic communication with patients and their families regarding their individual prognosis, fostering realistic expectations and informed decision-making. Furthermore, for the design and interpretation of future clinical trials, this nomogram can provide a standardized and objective framework for patient stratification, allowing for the creation of more homogeneous study populations and enabling more accurate assessment of novel therapeutic interventions. Ultimately, this tool contributes to a more personalized approach in the complex management of advanced pancreatic cancer.
Introduction
Pancreatic cancer (PC) continues to pose a formidable challenge within the landscape of oncology, representing one of the most aggressive and lethal malignancies. It currently stands as the fourth leading cause of cancer-related deaths globally, and alarming projections indicate that it is poised to ascend to the position of the second most lethal cancer by the year 2030. Despite the commendable advancements made in systemic treatment strategies over recent years, the prognosis for patients afflicted with metastatic pancreatic cancer (MPC), where the disease has disseminated, or locally advanced unresectable pancreatic cancer (LAPC), where the tumor cannot be surgically removed despite being confined to the pancreas and surrounding major blood vessels, remains profoundly grim. The dismal reality is reflected in a 5-year overall survival (OS) rate that languishes at less than 5%.
In the realm of first-line systemic treatment for fit patients with MPC, two prominent regimens have emerged as current standards of care, rigorously established through pivotal Phase 3 clinical trials: FOLFIRINOX and gemcitabine plus nab-paclitaxel (Gem-Nab). FOLFIRINOX, an intensive multi-drug chemotherapy regimen, comprises a combination of leucovorin, 5-fluorouracil, irinotecan, and oxaliplatin. Both FOLFIRINOX and Gem-Nab have also demonstrated promising levels of efficacy in the treatment of patients presenting with LAPC. Notably, FOLFIRINOX is increasingly regarded as a highly suitable therapeutic option for LAPC cases and has, more recently, been established as the new reference standard in the adjuvant setting, administered after surgical removal of the primary tumor to prevent recurrence.
Despite its efficacy, the FOLFIRINOX triplet chemotherapy regimen is unfortunately associated with a considerable burden of potentially severe adverse events. These toxicities primarily manifest as digestive and hematologic complications. Specifically, grade 3 or 4 neutropenia, a severe reduction in a type of white blood cell, occurs in a significant 46% of patients treated with FOLFIRINOX, with a notable 5.4% experiencing febrile neutropenia, a life-threatening complication characterized by fever in the setting of severe neutropenia. Furthermore, even in highly selected patient populations enrolled in rigorous randomized studies—typically characterized by an excellent performance status (PS of 0-1), bilirubin levels less than 1.5 times the upper limit of normal, and age generally below 75 years—the median overall survival barely surpasses 11 months. In routine clinical practice, the stringent eligibility criteria mean that only approximately 25% of patients with MPC are deemed sufficiently fit to receive FOLFIRINOX. Consequently, various strategies, broadly categorized as “modified FOLFIRINOX,” have been explored with the explicit aim of improving the tolerability of this intensive regimen. These modifications often involve omitting the 5-fluorouracil bolus component, reducing the irinotecan dose, or prophylactically administering growth factors to mitigate hematologic toxicities. This modified approach appears to successfully reduce the incidence of grade 3 or 4 gastrointestinal or hematologic events, while reportedly maintaining comparable overall survival outcomes, as evidenced in trials such as PRODIGE4-ACCORD11.
The Gem-Nab combination represents a widely accepted alternative first-line treatment option. While this regimen is associated with an overall similar incidence of hematologic toxicities when compared to FOLFIRINOX (with grade 3 or 4 neutropenia occurring in 38% of patients and febrile neutropenia in 3%), it presents a distinct toxicity profile. Gem-Nab results in a higher rate of grade 3 or higher peripheral neuropathy (17% versus 9% with FOLFIRINOX) but a lower rate of severe diarrhea (6% versus 12.7%). Consequently, Gem-Nab is generally considered a suitable option for a broader percentage of MPC patients in routine clinical practice, owing to its distinct side-effect profile that may be more manageable for certain individuals.
In a concerted effort to enhance risk stratification and optimize patient selection for both routine clinical decision-making and the design of future clinical trials, numerous researchers have meticulously investigated a wide array of clinical and laboratory factors that are putatively linked with patient outcomes. For instance, Goldstein et al. comprehensively analyzed data from the MPACT study, successfully identifying several variables significantly associated with overall survival. Building on these findings, they developed a nomogram capable of predicting patient survival probability at various time points when treated with gemcitabine, with or without nab-paclitaxel. Predictive algorithms, and nomograms in particular, have recently gained significant momentum in clinical practice due to their demonstrated accuracy and inherent ease of use. Prior studies utilizing modified FOLFIRINOX regimens have reported that the presence of liver metastases, the patient’s performance status, and the neutrophil-to-lymphocyte ratio (NLR) are independently associated with overall survival. However, crucially, no validated tool is currently available to predict the prognosis for an individual patient specifically when treated with the intensive FOLFIRINOX triplet regimen.
Given the availability of distinct first-line treatment options for advanced pancreatic cancer and the absence of direct head-to-head comparative trials between FOLFIRINOX and Gem-Nab, discussing the relative benefits and risks of each regimen with patients remains a complex and challenging endeavor. Therefore, the overarching aim of our study was to develop and rigorously validate a simple yet accurate nomogram. This tool is designed to predict the 6-month survival probability in patients with metastatic pancreatic cancer (MPC) and locally advanced unresectable pancreatic cancer (LAPC) who are treated with first-line triplet chemotherapy, encompassing both the classic and modified FOLFIRINOX schedules. Such a nomogram would provide a valuable, patient-specific prognostic assessment, aiding both clinicians and patients in making more informed treatment decisions.
Patients And Methods
Patient Selection And Data Collection
The development set (DS) for this study comprised consecutive patients diagnosed with metastatic pancreatic cancer (MPC) and locally advanced unresectable pancreatic cancer (LAPC) who received treatment at a single institution, Azienda Ospedaliero-Universitaria Pisana in Pisa, Italy. These patients were treated between January 2008 and December 2014, and their cases were thoroughly discussed by a dedicated multidisciplinary team specializing in pancreatic malignancies.
Eligible patients for inclusion in the developing set met the following stringent criteria: they were over 18 years of age; they had a cytologically or histologically confirmed diagnosis of pancreatic carcinoma; their disease was classified as non-resectable, stage III or IV according to the American Joint Committee on Cancer (AJCC) staging system; comprehensive clinical data collected prior to the initiation of first-line chemotherapy was accessible; pre-treatment laboratory information was available; objective tumor response evaluations were performed; and complete survival data were obtainable. It is important to note that the FOLFOXIRI schedule (comprising leucovorin, 5-fluorouracil, oxaliplatin, and irinotecan) employed in Pisa represented an alternative to the standard FOLFIRINOX regimen. This modified regimen was derived from extensive experience in colorectal cancer and has demonstrated apparently superimposable efficacy compared to FOLFIRINOX in MPC/LAPC cases. Detailed information regarding this modified regimen has been described in previous publications.
A comprehensive array of putative predictors was investigated for their potential association with patient outcomes. These included demographic variables such as age and gender. Clinical status indicators included Eastern Cooperative Group (ECOG) Performance Status (categorized as 0 vs. 1). Disease characteristics comprised AJCC stage (III vs. IV), tumor location within the pancreas (head vs. body/tail), history of prior surgery of the primary tumor (yes vs. no), and previous adjuvant chemotherapy (yes vs. no). Additional factors included the presence of biliary drainage (yes vs. no), the total number of disease sites involved, and the specific presence (yes vs. no) of metastases at distinct anatomical sites such as the liver, lung, peritoneum, or bone. Furthermore, a range of laboratory parameters were collected: neutrophil, lymphocyte, and platelet counts, as well as calculated ratios such as the neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR), all obtained before the first cycle of treatment. Pretreatment serum levels of lactate dehydrogenase, carcinoembryonic antigen, and carbohydrate antigen 19-9 (CA19-9) were also included. Age, the number of disease sites, and all laboratory parameters were recorded and analyzed as continuous variables to capture their full variability.
For the crucial step of external validation, an independent cohort of MPC/LAPC patients was assembled from different Italian and French institutions, treated between January 2011 and June 2017. The inclusion criteria for this validating set (VS) were identical to those used for the developing set (DS), ensuring consistency in patient characteristics. Similarly, the same variables were collected for analysis in the VS. All patients included in the VS received FOLFIRINOX strictly according to the PRODIGE4-ACCORD11 schedule, which is considered the standard classic regimen.
All analyses conducted in this study adhered strictly to the principles outlined in the Declaration of Helsinki and received full approval from the ethics committee of the coordinating center (Azienda Ospedaliero-Universitaria Pisana, Pisa, Italy). Prior to their inclusion in the investigation, written informed consent was obtained from all patients, authorizing the research use of their de-identified data. The entire study protocol was formally registered on ClinicalTrials.gov under the identifier NCT03590275, ensuring transparency and accountability.
Statistical Analyses
A crucial binary variable, termed “early death,” was meticulously calculated from patient survival times to indicate their 6-month survival status. A value of 1 was assigned if a death event occurred within the first 180 days from the start of treatment, and 0 was assigned otherwise. The association of various covariates with this 6-month survival status was initially assessed by constructing univariate unconditional logistic regression models. In these models, each potential prognostic variable was individually modeled against the 6-month survival status. Statistical significance for these associations was evaluated using the Wald test, with a two-tailed P-value of less than 0.05 considered statistically significant. Given the inherent high variability of CA19-9 levels, this covariate was logarithmically transformed prior to conducting any analyses, which helps to normalize its distribution and improve the robustness of statistical models.
Covariates that demonstrated statistical significance in the univariate analyses were then used to develop various multivariate logistic regression models. Both forward and backward selection methods were employed to identify the most parsimonious and predictive models. The significance of each individual covariate within the multivariate model was again assessed using the Wald test. The overall global fit of the models was comprehensively evaluated using several statistical metrics, including Nagelkerke R^2, Somer D, and the model log-likelihood ratio chi-square. To rigorously address potential multicollinearity among covariates, a suite of tests was utilized, including t-tests, Mann-Whitney tests, Fisher exact tests, ANOVA, linear regressions, and the calculation of variance inflation factors, with the specific test chosen based on the nature and characteristics of the covariates (e.g., binary, categorical, or continuous). These same tests were also applied to assess any statistically significant differences in baseline clinical characteristics between patients included in the developing set (DS) and the validating set (VS) cohorts.
The final decision regarding the inclusion of a specific variable into the ultimate prognostic model was based on a holistic assessment. This included its statistical significance, the consistency of its significance across a high percentage of the models explored, the overall global fit of the chosen model, and critically, the clinical plausibility and biological relevance of the covariates. The probabilities predicted by the developed nomogram were then rigorously tested against the observed probabilities in the external validating set (VS) to assess its generalization capability. The nomogram’s discriminative ability, which refers to its capacity to differentiate between patients who will and will not experience the outcome, was evaluated using Somer D, the C-index (equivalent to the Area Under the Curve, AUC, for binary outcomes), and the Spiegelhalter Z test. The Brier score was used to quantify the accuracy of predictions. The 95% confidence intervals (CIs) for the C-index were robustly calculated using a bootstrap resampling method. The calibration plot was visually assessed to examine the agreement between predicted and observed probabilities, providing insight into how well the model’s predicted probabilities matched the actual event rates. Survival analyses were performed using the Kaplan-Meier method, with differences assessed by the log-rank test, and by constructing Cox regression models. Median follow-up times were calculated using the reverse Kaplan-Meier method, which accounts for censoring.
Tumor response rate was evaluated according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1, providing a standardized measure of tumor shrinkage or growth. Progression-free survival (PFS) was meticulously defined as the time interval from the initiation of FOLFIRINOX chemotherapy until the first occurrence of clinical or radiological disease progression, or death from any cause, whichever event occurred first. For patients who did not experience progression or death, data were censored at the date of their last follow-up. Overall survival (OS) was defined as the time from the start of FOLFIRINOX treatment until death from any cause. Survival data were censored at the date of the last follow-up for living patients. Receiver operating characteristic (ROC) curves were utilized to determine the optimal cutoff values for the categorization of continuous variables, maximizing their predictive utility. All statistical analyses were comprehensively performed using the “Survival” and “rms” packages within the R statistical software environment.
Results
Patient Characteristics And Treatment Outcome
A total of 343 patient data sets were meticulously analyzed in this study, with 137 cases constituting the developing set (DS) and 206 cases forming the validating set (VS). A detailed comparison of the patient characteristics between these two cohorts is presented, highlighting areas of similarity and difference. Notably, a higher proportion of patients in the validating set exhibited an ECOG Performance Status of 1 compared to the developing set (54.9% vs. 32.8%), a statistically significant difference (P < .001). The neutrophil-to-lymphocyte ratio (NLR) was also significantly higher among patients in the validating set, with a median of 3.2 compared to 2.3 in the developing set (P < .001). Despite these differences in performance status and NLR, no other significant disparities were observed between the two cohorts in terms of the number and anatomical location of metastases, baseline CA19-9 serum levels, or any other known established prognostic factors (all P > .1), indicating a general comparability across many key clinical features.
The median follow-up period for patients in the developing set was 30 months, while for the validating set, it extended to 35 months. The clinical outcomes achieved in the two cohorts were largely similar, reinforcing the generalizability of the findings. The overall response rate was 38.6% in the developing set and 31.4% in the validating set, indicating comparable rates of tumor shrinkage. Median progression-free survival (PFS) was 8.0 months (with a 95% confidence interval of 6.7-9.2 months) in the developing set and 7.2 months (95% CI, 5.6-8.2 months) in the validating set, demonstrating similar periods of disease control. Similarly, the median overall survival (OS) was 11.6 months (95% CI, 10.5-13.9 months) in the developing set and 10.5 months (95% CI, 9.2-12.1 months) in the validating set, indicating comparable long-term survival. Death events were observed in the vast majority of patients across both cohorts, with only 8.8% of patients in the developing set and 9.7% in the validating set being censored for OS (i.e., still alive at the last follow-up). Importantly, there were no censored observations within the first 180 days, confirming that complete 6-month survival status was known for all patients, which was crucial for the primary endpoint.
Prognostic Nomogram Development
All the meticulously collected variables were comprehensively analyzed to determine their association with 6-month survival. From this extensive analysis, four distinct variables were ultimately selected for inclusion in the final multivariable model, owing to their significant prognostic value. These critical factors were: the ECOG Performance Status (PS), the pretreatment Neutrophil-to-Lymphocyte Ratio (NLR), the presence of liver metastases, and the baseline serum Carbohydrate Antigen 19-9 (CA19-9) level. While initial collinearity analyses indicated a slight correlation between CA19-9 levels and the presence of liver metastases, and also between ECOG PS and NLR, the variance inflation factor (VIF) for all variables consistently remained below 2. This low VIF value indicated that the level of multicollinearity was minimal and unlikely to significantly compromise the stability or interpretation of the model; thus, the model was retained without further modifications. Other variables, such as pretreatment platelet-to-lymphocyte ratio, the number of involved disease sites, and disease stage, although demonstrating statistical significance or borderline significance in univariate analyses, were ultimately not retained in the final model. This decision was primarily driven by an excessive degree of collinearity with the selected core variables (NLR and liver metastases), which would have introduced redundancy and potentially instability into the model.
The overall global fit of the developed model was rigorously evaluated using several robust statistical metrics. The model exhibited a good global fit, as evidenced by a Nagelkerke R^2 of 0.283, a Somer D statistic of 0.592, and a C-index of 0.796. Furthermore, the model’s statistical significance was overwhelmingly confirmed by a highly significant log-likelihood (P < .0001), indicating that the model provides a far better fit to the data than a null model. The culmination of this analytical process is the resulting nomogram, a visual representation of the predictive model, which allows for the straightforward calculation of individual patient risk based on the identified prognostic factors.
Prognostic Nomogram Validation
The probabilities of death predicted by the nomogram were rigorously tested against the observed probabilities within the independent validating set (VS) to ascertain its real-world applicability and robustness. The nomogram’s discriminative ability, which measures its capacity to correctly distinguish between patients who will and will not experience the outcome (death within 6 months), was confirmed as satisfactory. This was quantified by a Somer D value of 0.524, which corresponds to a C-index of 0.762. The 95% confidence interval for this C-index was calculated to be 0.713 to 0.825, affirming a strong and reliable discriminative performance. The Brier score, a measure of prediction accuracy, resulted in a value of 0.16, indicating good accuracy. Furthermore, the Spiegelhalter Z test, used to assess calibration (how well the predicted probabilities match the observed event rates), was not statistically significant (P = .087), suggesting a good overall calibration. Visual inspection of the calibration plot reinforced this, demonstrating a good overlap between the predicted and observed probabilities. However, it was noted that there was a slight underestimation of risk for patients who were truly at a very high risk of early death, suggesting a minor area for potential refinement in the extreme risk category.
Survival Analysis Based On Prognostic Factors
As an ancillary and highly practical analysis, we proceeded to categorize the continuous variables included in the final prognostic model. This categorization aimed to assess the utility of these factors in stratifying patients into distinct risk groups, which would facilitate immediate clinical application. To achieve this, patients from both the developing set (DS) and the validating set (VS) were combined to form a larger, more robust cohort. Four distinct risk categories were then defined based on the cumulative number of “poor prognostic features” present in each patient. These poor prognostic features were identified as: an ECOG Performance Status of 1, the presence of liver metastases, a logCA19-9 level above a predefined threshold value, and an NLR above a predefined threshold value.
To establish the optimal threshold values for the continuous variables, logCA19-9 and NLR, Receiver Operating Characteristic (ROC) curves were developed. These analyses returned Area Under the Curve (AUC) values of 0.641 for logCA19-9 and 0.676 for NLR, indicating moderate discriminative abilities for predicting early death. Based on these ROC analyses, a threshold of 6.75 was set for logCA19-9, which corresponds to a basal CA19-9 serum level of 845 U/mL. This threshold yielded a sensitivity of 0.64 and a specificity of 0.62 for predicting early death. For NLR, the optimal threshold was determined to be 2.46, providing a sensitivity of 0.78 and a specificity of 0.52.
Upon categorizing patients according to the number of these poor prognostic features (0, 1, 2, or >2), remarkably distinct survival outcomes were observed across the four identified subgroups. The median overall survival (OS) significantly differed between these subgroups (P < .0001 for the overall comparison), demonstrating the profound prognostic power of these combined factors. Patients with more than 2 poor prognostic features exhibited the shortest median OS at 7.2 months (95% CI, 5.6-8.7 months). This progressively improved to 10.8 months (95% CI, 9.4-12.9 months) for those with 2 risk factors, 13.9 months (95% CI, 12.5-16.6 months) for patients with 1 risk factor, and reached its longest at 18.3 months (95% CI, 14.5-23.5 months) for the most favorable subgroup, possessing 0 risk factors. This clear stratification underscores the practical utility of the nomogram in identifying patient populations with vastly different clinical trajectories.
Discussion
The life expectancy for patients afflicted with unresectable pancreatic cancer (PC), even when treated with first-line chemotherapy, unfortunately remains persistently poor, despite the relatively recent introduction of more active and intensive chemotherapy regimens such as FOLFIRINOX and Gem-Nab. In light of the significant toxicity profile associated with an intensive triplet schedule like classic or modified FOLFIRINOX, the ability to accurately anticipate the prognosis for an individual patient becomes exceptionally valuable. Such foresight allows for a more meaningful discussion regarding the benefit-to-risk ratio of this particular regimen, enabling a more informed and shared decision-making process concerning different first-line therapeutic options. Leading authorities in oncology have consistently advocated for the development of alternative measures to enhance our understanding and communication of treatment impact on overall survival. Evidence from the existing literature robustly demonstrates that open and transparent discussions about prognosis during clinical encounters actively strengthen the patient-oncologist relationship. This further emphasizes the critical need for validated and easily implementable instruments that can clearly communicate specific risks at defined time points throughout the disease course. While similar instruments have been proposed for second-line therapy in metastatic PC, they notably lack the inclusion of the triplet chemotherapy regimen commonly used in the first-line setting. Furthermore, in the context of clinical trials, prognostic nomograms can serve as invaluable tools for achieving better stratification of enrolled patients and facilitating a more nuanced interpretation of results across diverse subgroups, thereby enhancing the scientific rigor and clinical relevance of research findings.
Our study successfully identified several readily available and easily measurable clinical and laboratory parameters as major determinants of prognosis within this patient population. These critical factors include ECOG Performance Status, Neutrophil-to-Lymphocyte Ratio (NLR), the presence of liver metastases, and baseline CA19-9 levels. The accessibility of these parameters in routine clinical settings makes our developed nomogram highly practical and implementable. A substantial body of existing literature consistently supports the prognostic importance of these variables in PC patients treated with gemcitabine-based chemotherapy, particularly regarding performance status and CA19-9 values. However, to our current knowledge, this study represents the first attempt to integrate these factors into a rigorously validated model specifically designed to predict early deaths among metastatic or locally advanced unresectable PC patients treated with a more modern and intensive regimen like FOLFIRINOX. Intriguingly, our ancillary analysis, which stratified patients based on the presence of each single identified determinant, revealed distinct populations with significantly different survival outcomes. Specifically, in the most favorable risk subgroup (i.e., patients with no identified poor prognostic features), the median overall survival was nearly three times longer than that observed for the worst risk category (i.e., patients with more than two poor prognostic features present). This striking difference underscores the practical utility of the information retrieved by the nomogram for both clinical practice and future research endeavors. Given the current lack of validated predictive biomarkers in this setting, prognostic stratification is essential for guiding discussions about alternative treatment options for individual cases. Therefore, our nomogram could serve as a highly suitable tool for the precise identification of different patient subgroups and could additionally stimulate further research into the underlying biological bases explaining the influence of these clinical variables on survival outcomes.
The variables deliberately included in our nomogram were either definitively confirmed as independent prognostic determinants through rigorous multivariate analysis or were retained due to the overwhelming and robust evidence of their prognostic value established in the existing scientific literature. Notably, while previous studies have consistently shown tumor stage to be an important prognostic determinant, its statistical significance was not formally demonstrated in our specific datasets (P = .067 at univariate analysis). This lack of formal significance could potentially be attributed to the relatively limited number of locally advanced unresectable pancreatic cancer (LAPC) cases included in our study, as well as the potential presence of other confounding poor prognostic features within the LAPC cohort itself. Consequently, we made the considered decision not to retain disease stage in the final nomogram. This decision was further supported by the observed high degree of collinearity between disease stage and the presence of liver metastases, which could introduce redundancy and instability into the model.
It is particularly noteworthy that three out of the four factors ultimately included in our nomogram were also incorporated into the nomogram developed from the MPACT database. The MPACT nomogram included performance status, neutrophil-to-lymphocyte ratio (NLR), liver metastases, serum albumin, the sum of the largest lesion dimensions, receipt of analgesics, and treatment arm. This significant overlap in key prognostic factors further underscores the external validity and robustness of our current work. While the relative contribution of CA19-9 to the performance of the MPACT nomogram was limited when added to other factors like PS, NLR, liver lesions, and serum albumin, in our study, CA19-9 emerged as the strongest predictor of overall survival among all analyzed variables. This finding is highly consistent with previous literature, which has convincingly established this serum tumor marker as a main and confirmed determinant of patient outcome in this challenging population. Moreover, certain limitations of the MPACT nomogram—such as its lack of external validation, the relatively few points assigned to several variables (e.g., receipt of analgesics and treatment arm), and the inclusion of a highly selected patient population from a registering Phase 3 trial—might constrain its immediate utility in routine clinical practice and prevent its broad applicability to other populations or alternative treatment regimens. However, in light of the partial overlap in factors included in both nomograms and the inherent ease of use of our prognostic variables, it would be of significant interest to rigorously test the performance of this FOLFIRINOX nomogram in patients receiving other regimens, particularly Gem-Nab. In this regard, one recent study has already confirmed the prognostic role of performance status, NLR, and CA19-9 in a prognostic nomogram developed from a retrospective series of 210 patients treated with first-line Gem-Nab, further highlighting the consistency of these key markers across different regimens.
When strategically planning the design of this study, we consciously decided to construct an instrument specifically addressing the risk of early death, rather than focusing on general overall survival. This deliberate decision was made in light of the considerable risks of toxicity inherently associated with FOLFIRINOX chemotherapy, as well as other practical issues that profoundly affect a patient’s daily life, such as the requirement for implanting a central venous catheter for prolonged infusions. Indeed, there is little doubt that FOLFIRINOX, regardless of the specific schedule employed, remains a challenging treatment option. Ideally, this intensive regimen should be selectively provided to patients who are most likely to experience the greatest benefit in terms of overall survival, while simultaneously sparing those individuals who are highly unlikely to derive significant advantage due to a very short anticipated overall survival probability. The choice of a 6-month period as the primary outcome measure was directly informed by the results of the PRODIGE4-ACCORD11 trial, which reported a median progression-free survival (PFS) of 6.4 months and a median overall survival (OS) of 11.1 months in the experimental arm. In our considered opinion, the probability of experiencing an early death within the first 6 months after treatment initiation (i.e., a time period shorter than the median PFS typically expected with FOLFIRINOX) can be regarded as an acceptable and clinically meaningful criterion to discuss openly with a patient when evaluating various treatment alternatives to intensive triplet chemotherapy.
While our study offers significant contributions, it is important to acknowledge its main limitation: its retrospective design and the non-exhaustive nature of data collection regarding all potentially prognostic parameters. However, it is crucial to note that the patient characteristics were generally well-balanced between the developing set and the validating set, which enhances the generalizability of our findings. Furthermore, when the treatment activity and efficacy were comprehensively investigated, no statistically significant differences were reported in terms of response rate, progression-free survival, and overall survival between the two cohorts (and, by extension, between the FOLFIRINOX and FOLFOXIRI regimens). Moreover, these results were highly comparable with those described in the broader literature, reinforcing the consistency of our patient population and treatment outcomes. As previously discussed, our study was not designed to evaluate the outcome of the different risk categories with other alternative treatment options. Therefore, it should be explicitly kept in mind that our study was not intended to validate the developed nomogram as a predictive tool for anticipating the specific benefit from a particular regimen (e.g., FOLFIRINOX) when compared directly to other available options (such as Gem-Nab, single-agent chemotherapy, or purely supportive care). The information gleaned from our nomogram provides a more granular and detailed prognostic assessment for individual cases, offering valuable insights into their likely trajectory. Importantly, FOLFIRINOX (or its modified treatment schedules) remains a highly valid and important therapeutic option for fit metastatic or locally advanced unresectable pancreatic cancer patients who meet the appropriate clinical criteria.
Conclusion
In conclusion, this study unequivocally demonstrates that it is indeed possible to accurately and reliably predict the risk of death within the first 6 months following the initiation of FOLFIRINOX therapy for patients suffering from metastatic or locally advanced unresectable pancreatic cancer. This prediction can be achieved using a limited set of easily accessible, reproducible, and cost-effective clinical and laboratory parameters. MK-8245 The innovative nomogram we have constructed, along with the distinct risk categories it defines, enables immediate and straightforward prognostic stratification. It provides data that are inherently easy to interpret for both clinicians and patients, fostering greater clarity in complex clinical discussions. Ultimately, this robust instrument holds the potential to significantly facilitate effective patient-physician communication in routine clinical practice and to markedly improve prognostic stratification in the context of clinical research, leading to more informed decisions and more precise study designs. Furthermore, the validation of this powerful tool for its applicability with other prevalent treatment regimens, such as Gem-Nab, is a warranted and critical next step for future research.