Additive Value of Preoperative Sarcopenia and Lymphopenia for Prognosis Prediction in Localized Pancreatic Ductal Adenocarcinoma

Background Surgical resection with adjuvant chemotherapy is the only treatment that can provide long term survival in localized pancreatic ductal adenocarcinoma (LPDAC). Notwithstanding, recurrence occurs in the vast majority of patients and a better stratification of preoperative therapies is required. This study aimed to investigate preoperative immunological and nutritional factors to predict relapse-free survival (RFS) in patients with LPDAC. Methods Analyses were derived from all consecutive LPDAC patients treated with surgical resection at Besancon University Hospital, France, between January 2006 and December 2014 (n=146). Biological and nutritional parameters were recorded before and after surgery. The association of 24 baseline parameters with RFS was evaluated using univariate and multivariate Cox analyses. Based on the final model, a prognostic score was developed. Results Lymphocyte count and body composition were available for 94 patients. In multivariate analysis, preoperative lymphopenia and sarcopenia (or a low muscle mass) were identified as independent prognostic factors for RFS. The score determined three groups with a median RFS of 5.6 months (95% confidence interval [CI] = 4.3 to 9.6 months) for high-risk group, corresponding to patients with lymphopenia; 11.5 months (95%CI = 9.8 to 13.9 months), and 21.2 months (95%CI = 9.9 to 55.3 months), for intermediate-(patient with sarcopenia without lymphopenia), and low-risk groups (no risk factor), respectively (p <0.001). Preoperative sarcopenia predicts the occurrence of postoperative lymphopenia in patients with a preoperative lymphocyte count above 1,000/mm3 (p = 0.0029). Conclusions Preoperative lymphopenia and sarcopenia are pejorative prognostic factors in LPDAC and should be considered in the preoperative evaluation to stratify death risk in patients with LPDAC.


Patients
All consecutive patients with histologically proven LPDAC treated by surgical resection at Besancon University Hospital, France, between January 2006 and December 2014 were involved.
Pancreatectomy and systematic lymphadenectomy were performed as a curative intent in all patients. A relapse of the disease was defined radiologically with RECIST v1.1 criteria (17). Patients could have received adjuvant chemotherapy. All therapeutic decisions were discussed and validated during digestive oncology-dedicated multidisciplinary meetings. Followup of patients was performed every three months with clinical examination, blood analysis (including carbohydrate antigen  [CA19-9] and carcinoembryonic antigen [CEA]), and computed tomography (CT) scan. The study is in accordance with standard procedures in France with approval from the relevant institutional review boards. The database was registered and declared to the National French Commission for bioinformatics data and patient liberty (CNIL; No. of CNIL declaration: 1906173 v 0). A general informed consent was signed by all patients at the time of their first visit to the university hospital. This consent allows the use of their clinical, radiological, and biological data in the cohort study. The database was locked on November 3, 2017.
Demographics, cancer history, clinical, pathological, radiological parameters, as well as treatment outcomes, were retrospectively collected from medical records. Preoperative and postoperative (one month after surgery) biological (CRP, albumin, lymphocytes, neutrophils, CA19-9, CEA) and nutritional parameters were recorded, including body composition parameters (skeletal muscle) by CT scan. According to our previous research, lymphopenia was defined as a lymphocyte count below 1,000/mm 3 (10). An underweight was defined by body mass index <18.5 kg/m² or <21 kg/m² over 70 years. For the assessment of skeletal muscle area, CT Digital Imaging and Communication in Medicine (DICOM) images at the third lumbar (L3) level were analyzed using NIH Image J1.47 to determine the indexed muscle area (IMA) excluding L3, by a single operator, blinded to patient information. Muscle area was normalized by height in squared meters (m²) and reported as the IMA (cm²/m²). The thresholds for defining sarcopenia (or low skeletal muscle mass) were 38.5 cm²/m² for women and 52.4 cm²/m² for men, according to Prado et al. (18).

Statistical Analysis
Median value (interquartile range [IQR]) and frequency (percentage) were provided for the description of continuous and categorical variables, respectively. Medians and proportions were compared using Student's t-test and Chi-square test (or Fisher's exact test, as appropriate), respectively. RFS was calculated from the date of surgery to the date of postoperative tumor relapse or death from any cause, or the date of the last follow-up, at which point data were censored. Overall survival (OS) was calculated from the date of surgery to the date of death from any cause. Survival data were censored at the last follow-up. OS and RFS were estimated using the Kaplan-Meier method and described using median or rate at specific time points with 95% confidence intervals (95%CI). Follow-up duration was calculated using a reverse Kaplan-Meier estimation when feasible (19).
Cox-proportional-hazard models were performed to estimate the hazard ratio (HR) and 95%CI for factors associated with RFS. The association of 24 baseline parameters with RFS was first assessed using univariate Cox analyses and then parameters with p <0.05 were entered into a final multivariate Cox regression Abbreviations: 95%CI, 95% confidence intervals; CA19-9, carbohydrate antigen 19-9; CEA, carcinoembryonic antigen; CNIL, National French Commission for bioinformatics; CT, computed tomography; HR, hazard ratio; IL-1, interleukin 1; IL-6, interleukin 6; IMA, index muscular area; IQR, interquartile range; L3, third lumbar vertebral level; LPDAC, localized pancreatic ductal adenocarcinoma; NA, not available; NF-kb, nuclear factor-kappa b; NLR, neutrophil to lymphocyte ratio; OS, overall survival; CRP, C-reactive protein; PDAC, pancreatic ductal adenocarcinoma; RFS, relapse-free survival; SM, skeletal muscle mass; TNF-a, tumor necrosis factor alpha. model, after considering collinearity among variables with a correlation matrix. When used in continuous in the Cox model, a potential non-linear relationship between predictors and RFS was first investigated using the fractional polynomials method to determine the best transformation for continuous variables (20)(21)(22) and validated by the restricted cubic splines method with graphical evaluation. The assumption of proportionality was checked by plotting log-minus-log survival curves and by the cumulative martingale process plots.
The accuracy of the final model was verified regarding two parameters: discrimination and calibration. The predictive value and the discrimination ability of the final model were assessed with the Harrell's concordance index (C-index) (22). Random samples of the population were used to derive 95%CI bootstrap percentile for the C-statistic. Calibration was assessed by visual examination of calibration plot at 6, 12, 24, and 48 months. Internal validation of the final model was performed with a bootstrap sample procedure.
The final model was used to establish a prognostic score allowing the preoperative estimation of RFS. To give a reasonable spread of risk, we chose to distinguish two levels of sarcopenia using IMA, according to their risk score level, which were identified based on cut points determined following two methods: the median value and the Cox's method (23). Patient characteristics were compared between prognostic risk groups using Fisher-exact test and Kruskal-Wallis test for categorical and quantitative parameters, respectively. The prognostic score developed to estimate preoperative RFS was applied in the same population to evaluate preoperative OS, and RFS with postoperative parameters.
All analyses were performed using SAS version 9.4 (SAS Institute) and R software version 2.15.2 (R Development Core Team; http://www.r-project.org). Values of p <0.05 were considered statistically significant and all tests were two-sided. Details on the interpretation of important statistical concepts are given in the Supplementary Methods.

Population Characteristics
From January 2006 to December 2014, 146 patients who underwent surgery for a LPDAC were included in this cohort ( Figure 1). Patient characteristics are described in Table 1

Preoperative Prognostic Factors of RFS
We identified six preoperative parameters as prognostic factors for RFS, in the univariate analyses (p <0.05): tumor size, IMA, sarcopenia, lymphopenia, NLR, and CA19-9 ( Table 2). Other nutritional factors (weight loss or albumin level) were not statistically associated with RFS. The transformations used for continuous variables are summarized in Supplementary  Figure 1. A square root transformation was applied for NLR and CA 19-9, while an inverse square root transformation was necessary for lymphocyte count. All other continuous variables were considered without any transformation.
A correlation matrix was used to detect statistically significant correlations between investigated parameters (Supplementary Figure 2). Significant correlations were defined by a correlation coefficient ≥ 0.4 associated with a p-value <0.001. A correlation was identified between "Lymphopenia" and "Neutrophil-tolymphocyte ratio", and between "Sarcopenia" and "Index Muscle Area". We selected as most clinically relevant variables "Lymphopenia" and "Sarcopenia".
Finally, the multivariable Cox analysis showed two independent risk factors for RFS: sarcopenia (HR = 1.78, 95% CI= 1.01 to 3.14, p = 0.0469) and lymphopenia (HR = 4.57, 95% CI = 2.24 to 9.34, p <0.0001; Table 3). Among the 146 patients operated for LPDAC, lymphocyte count and body composition calculation were available for 94 patients ( Figure 1). However, the two groups with or without complete information displayed similar RFS (Supplementary Figure 3), and patient characteristics were well-balanced between them, except for the median time between diagnosis and surgery ( Table 1).

Performance Assessment and Internal Validation of the Final Model
The multivariable model exhibited good discrimination ability (C-index = 0.67, 95% CI = 0.57 to 0.77). The calibration plots showed an optimal agreement between model prediction and actual observation for predicting RFS probability at 6, 12, 24, and 48 months (Supplementary Figure 4). In the internal validation, uncertainties around hazard ratio measured with a bootstrapping procedure reflected the robustness of the final model ( Table 3).

Preoperative Scoring System to Predict RFS
A prognostic score integrating the two independent factors for RFS was built. Kaplan-Meier curves of RFS according to preoperative lymphopenia and sarcopenia showed four groups (Supplementary Figure 5). The survival of patients with lymphopenia (corresponding to the highest HR), with or without sarcopenia, was similar (median of 6.6 months, 95% CI = 4.4 to 12.6, and 5.6 months, 95%CI = 3.4 to 9.6, respectively). Thus, overall patients with lymphopenia were grouped together, and patients were categorized into three risk groups (high, intermediate, and low risk; Figure 2). The highrisk group is constituted by patients with lymphopenia and/or sarcopenia. Patients with sarcopenia without lymphopenia were classified in the intermediate group, while patients without any risk factor were in the low-risk group. Risk groups had median RFS of 5.6 months (95% CI = 4.3 to 9.6 months), 11.5 months (95% CI = 9.8 to 13.9 months), and 21.2 months (95% CI = 9.9 to 55.3 months), respectively (p <0.001).
Patient characteristics in each risk group are described in Table 4. The parameters were similar in the three groups, especially hypoalbuminemia. Patients with preoperative lymphopenia and/or sarcopenia had sarcopenia after surgery in 76.9%, 80.0%, and 27.3%, respectively in high, intermediate, and low risk groups.
The discriminative ability of the three-group model was confirmed in OS analysis ( Figure 3). Of note, the adjuvant chemotherapy administration was homogeneous in the three risk groups regardless of lymphopenia or sarcopenia levels (p = 0.1557; Supplementary Table 1).

Defining a Threshold of Preoperative Sarcopenia Correlated With PDAC Patients' Risk of Death
In order to determine which threshold of sarcopenia might influence PDAC patients' clinical outcomes, we explored the prognosis of patients who had no baseline lymphopenia and were clustered in two different groups according to sarcopenia levels. In a first analysis, we observed that the median value for sarcopenia measures could not distinguish different risk groups (Supplementary Figure 6). However, using the Cox's method, two degrees of sarcopenia were associated with prognosis using thresholds of 36.1 cm²/m² for women and 45.7 cm²/m² for men. Thus, different risk groups for RFS were distinguished ( Figure 4), with a median RFS of 11.4 months (95%CI = 8.4 to 13.1), and 28.3 months (95% CI = 3.2 to NA), respectively (p <0.0001; Figure 4). The two risk groups displayed similar patient characteristics ( Table 4), suggesting that in the absence of lymphopenia, sarcopenia is one of the major determinants to predict the risk of death for patients eligible for PDAC surgery.

Evaluation of Postoperative Lymphopenia
We have previously shown that lymphopenia exhibits a better accuracy when monitored one month after rather than before PDAC surgery (10). Then, we decided to analyze the impact of sarcopenia on the incidence of postoperative lymphopenia.    Figure 5). Furthermore, among patients with baseline lymphocyte count ≥ 1,000/mm 3 , the risk of postoperative lymphopenia was significantly enhanced in the presence of sarcopenia measured at diagnosis (77.8% versus 22.2%, p = 0.0029; Table 1).

DISCUSSION
Preoperative lymphopenia and sarcopenia were identified as independent prognostic factors for RFS in LPDAC. The additive value of baseline sarcopenia and lymphopenia allows the proposal of a prognostic score where LPADC patients are classified into three risk groups. These results highlight considerable heterogeneity in LPDAC patients' survival.
Our study confirmed that preoperative lymphocyte count is an independent prognostic factor in LPDAC (HR = 4.57, p <0.0001). The use of a threshold offered better discrimination than the use of lymphocyte count because it allows relapse-risk stratification. Both lymphocyte count and NLR are recognized independent prognostic factors in pancreatic adenocarcinoma and are used in clinical practice (24,25).
In our cohort, the prevalence of sarcopenia was high (60%), as observed in previous studies (12,26). Median BMI was 24.8 kg/m² (IQR, 22.2 -27.7) and more than half of patients (52.1%) were overweight or obese. Importantly, almost 40% of these individuals had also preoperative sarcopenia. Severe depletion of skeletal muscle detection remains a challenge in patients with obesity (27). BMI and clinical evaluation are not enough, a radiological assessment of skeletal muscle area is necessary to identify sarcopenic obesity. We highlighted that preoperative sarcopenia is the only nutritional independent prognostic factor for RFS in LPDAC (HR = 1.78, p = 0.0469). Few studies have analyzed the prognostic value of preoperative sarcopenia in LPDAC on OS and RFS and showed contradictive results (26,(28)(29)(30), probably because of the lack of consensus on the definition of sarcopenia. Among many definitions, we chose to use the threshold defined by Prado et al. (31), validated for gastrointestinal cancer and in the western population, as ours. Nevertheless, we pointed out a new cut off value of sarcopenia more accurately correlated to prognosis with thresholds corresponding to 36.1 cm²/m² for women and 45.7 cm²/m² for men to distinguish high-risk group with a median RFS of 11.4 months and low-risk group with a median RFS of 28.3 months (P <0.0001).
Prognostic factors that can be identified before surgery and chemotherapy are mandatory to stratify the treatment decisionmaking process in current clinical practice and for the development of more personalized neoadjuvant strategies. In the  multivariate analysis performed in the present study, only preoperative lymphocyte count and sarcopenia were independent prognostic factors. Thus, combining both variables allowed to elaborate a preoperative prognostic score and identified several subgroups of patients with different prognoses. Patients with the worse prognosis were those with lymphopenia (RFS of 5.6 months, 95%CI = 4.3 to 9.6 months). Similarly, postoperative lymphopenia is also an independent negative prognostic factor (HR = 2.50, 95%CI = 1.53 to 4.09, p = 0.0003). In postoperative time, the score confirmed that  patients with lymphopenia had the worst clinical outcomes, with a median RFS of 9.0 months (95%CI = 4.3 to 10.3 months). In our previous study, none of postoperative lymphopenic patients had long-term survival (10). Similarly, in the study of Tsujita et al. (32), the 3 years survival rate after pancreatectomy was 33.9% in patients with a postoperative NLR of less than 3 at one month and 7.3% in those with a postoperative NLR of 3 or more (p <0.001). Interestingly, in our study, 77.8% of patients with postoperative lymphopenia had preoperative sarcopenia suggesting the predictive value of this factor. Our results suggest that sarcopenia measured at the baseline might be a predictive factor for the occurrence of postoperative lymphopenia. Several mechanisms are probably involved in sarcopenia (a decreased of skeletal muscle mass). Inadequate intake due to anorexia, increased energy expenditure, systemic inflammation, and abnormal metabolism result in muscle wasting and body weight loss (33). In addition, tumor cells product pro-cachectic factors such as proteolysis-inducing factor (34) and also interact with host cells to produce inflammatory cytokines, such as TNF-a, IL-1, and IL-6 which activate muscular nuclear factor-kappa b (NF-kb) and cause wasting of skeletal muscle (35,36) notably in pancreatic cancer (37). Some measures that have been proposed to treat sarcopenia have not been supported by evidence and currently, no study has shown an increase in lean mass nor OS following the usual nutritional treatments in pancreatic cancer (38). However, according to Sandini et al., after neoadjuvant chemotherapy some patients with primary unresectable pancreatic cancer who underwent resection had experienced a 5.9% skeletal muscle area increase during treatment, whereas those who did not undergo resection had a 1.7% decrease (p <0 .001) (39).
Decreased lymphocyte count results from an inadequate immunologic reaction and is a valuable biomarker for identifying cancers associated with an increased risk of tumor immune evasion and poor prognosis. The role of the immune system in cancer was highlight by studies investigating the prognostic influence of Tumor Infiltrating Lymphocytes (TIL). Indeed, in pancreatic cancer elevated CD8 + T lymphocytes in tumor stroma is a favorable prognostic factor influencing OS. Conversely, an increasing rate of FOXP3 + lymphocytes reflects immunological tolerance and correlates with decreased survival rates (40,41). Interstingly accumulating data in immunology attested that chemotherapy might improve anti-tumor immunity (42). In breast cancer, Goto et al. point out the predictive value of change in the CD8 + TIL levels and the CD8/FOXP3 ratio (p <0.001) after neoadjuvant therapy (43). In pancreatic cancer, after neoadjuvant chemotherapy, the median OS of patients with a high CD8 + /FOXP3 + lymphocyte ratio was longer than that of patients with a low CD8 + /FOXP3 + lymphocyte ratio (p=0.01) (44).
These data suggest the potential utility of neoadjuvant strategy in LPDAC patients with preoperative lymphopenia and/or sarcopenia. Some prehabilitation studies including nutrition and exercise are in progress and may impact sarcopenia, lymphopenia, and probably quality of life (45). In addition, the available data suggest a potential anti-tumor effect of the practice of physical activity and a benefit on survival, which could be mediated in particular by the decrease in insulin resistance, the modulation of the secretion of adiponectins, the decrease of the inflammatory syndrome, a modulating effect of intratumoral signaling pathways, a decrease in the toxicity of the treatments and therefore a better dose-intensity, and the reduction of sarcopenia (46)(47)(48). Systemic inflammation can be also reduced by pharmacological agents (such as corticosteroids or nonsteroidal anti-inflammatory drugs) as well as specific nutrients enriched with fatty acids. Particularly, some protocols with omega-3 fatty acids are under investigation in elderly patients (49).  There are some limitations in our study. There are some missing data due to the retrospective design of the study, but the two groups with or without complete information displayed similar RFS (Supplementary Figure 3). Our results have to be confirmed using a validation cohort. From a statistical point of view, the assessment of model performance measures such as discrimination, calibration, and internal validation strengthen the present investigation. The multivariate analysis significantly improved the model discrimination capacity because the C statistic increased significantly from 0.60 to 0.67 (bootstrap mean difference = 0.07, 95% CI = 0.57 to 0.77) demonstrating the additive value of lymphopenia and sarcopenia for death risk stratification. Moreover, the assessment of skeletal muscle area is only quantitative. The quality of the muscle (skeletal muscle density)  and muscle function (handgrip strength are not evaluated, but these measurements are strongly correlated with muscle mass and associated with survival in digestive cancers (50,51). Finally, our results may provide evidence for appropriate lymphocyte count and sarcopenia cut-off definition in order to better select PDAC patients eligible for neoadjuvant therapy. Preoperative lymphopenia and sarcopenia are pejorative independent prognostic factors for RFS and OS in LPDAC. Assessment of these factors at baseline may be relevant in current clinical practice for death risk stratification.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by National French Commission for bioinformatics data and patient liberty (CNIL). The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
Conception and design: Cd'E, JG, JR, CB, and AV. Administrative support: DV, CB, and AV. Provision of study materials or patients: Cd'E, JG, JR, BH, CB, and AV. Collection and assembly of data: Cd'E, JG, JR, and AV. Data analysis and interpretation: Cd'E, JG, JR, DV, CB, and AV. Manuscript writing: All authors. All authors contributed to the article and approved the submitted version.