A clinically applicable muscular index predicts long-term survival in resectable pancreatic cancer

ARTICLE IN PRESS

A clinically applicable muscular index predicts long-term survival in resectable pancreatic cancer Daniel Delitto, MD, PhD,a Sarah M. Judge, PhD,b Thomas J. George, Jr, MD,c George A. Sarosi, MD,a,d Ryan M. Thomas, MD,a,d Kevin E. Behrns, MD,a Steven J. Hughes, MD,a Andrew R. Judge, PhD,b and Jose G. Trevino, MD,a Gainesville, FL

Background. The relationship between myopenia, nutritional status, and long-term oncologic outcomes remains poorly characterized in patients with clinically resectable pancreatic cancer. We sought to reliably quantify prognostic indicators of preoperative cachexia in a manner applicable to any clinical setting. Methods. Preoperative computed tomographies were available electronically and suitable for analysis in 73 of 82 consecutive patients with pancreatic cancer undergoing pancreatoduodenectomy between November 2010 and February 2014. The psoas index was computed from the cross-sectional area of the psoas muscles normalized to vertebral body area at the third lumbar vertebra. Correlation and proportional hazards analyses were performed to identify relationships between muscularity, preoperative nutritional markers, clinicopathologic parameters, and long-term survival. Results. The psoas index correlated strongly with preoperative hemoglobin and albumin levels (P = .001 and .014, respectively) identifying a pattern of preoperative frailty. High psoas index and the albumin and hemoglobin levels were associated with improved long-term survival (hazard ratio 0.014, P < .001; hazard ratio 0.43, P < .001; and hazard ratio = 0.80, P = .014); however, on multivariate analysis, the psoas index proved to be the only independent predictor of survival (hazard ratio 0.021; P = .003). Rapid decreases in the psoas index during neoadjuvant chemotherapy were associated with poor postoperative outcomes, as were decreases in the psoas index during the postoperative period. Conclusion. The data indicate that the psoas index, a calculation derived from a clinically mandated, preoperative computed tomography, is a statistically powerful and easily calculated predictor of survival in pancreatic cancer when compared to tumor grade and stage as well as previously validated nutritional parameters. (Surgery 2016;j:j-j.) From the Department of Surgery,a College of Medicine, the Department of Physical Therapy,b and the Department of Medicine,c College of Medicine, University of Florida Health Science Center, Gainesville, FL; and the North Florida/South Georgia Veterans Health System,d Department of Surgery, University of Florida College of Medicine, Gainesville, FL

PANCREATIC CANCER (PC) is the fourth leading cause of cancer death in the United States and by 2030 is projected to be second only to lung cancer.1 Local approaches to curative therapy encounter high recurrence rates, indicative of the systemic nature of early stage disease.2 Therefore, it is not surprising that patients with clinically resectable disease demonstrate highly variable long-term outcomes.

Accepted for publication September 7, 2016. Reprint requests: Jose G. Trevino, MD, Department of Surgery, University of Florida College of Medicine, PO Box 100286, 1600 SW Archer Rd, Gainesville, FL 32610. E-mail: jose.trevino@ surgery.ufl.edu. 0039-6060/$ - see front matter Ó 2016 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.surg.2016.09.038

Patients predisposed to poor outcomes after resection, however, remain indistinguishable clinically from long-term survivors in the preoperative setting.3-5 The morbidity and recovery time associated with pancreatic operation underscores the need for better clinical tools to recognize those patients at greatest risk for early disease progression postoperatively. To this effect, a large body of literature has focused on preoperative risk assessment in surgical patients. Poor nutritional status emerges consistently as a powerful predictor of poor outcomes, particularly in cancer-based investigations.6-10 In addition to a compromised ability to convalesce from operative trauma, this phenomenon also appears to reflect an association between poor enteral intake (anorexia) and an advanced oncologic disease state. Physiologically, this anorexia SURGERY 1

ARTICLE IN PRESS 2 Delitto et al

combined with concurrent abnormal metabolic/ catabolic state results in cachexia, characterized by profound muscular atrophy.11,12 Classically, preoperative hemoglobin and albumin have been associated with advanced states of cachexia as well as poor perioperative outcomes.13-15 The evolutionary pressure to conserve serum levels of these factors for essential processes in times of physiologic stress led us to hypothesize that localized muscle wasting may serve as a more reliable indicator of a cachectic state, ie, sarcopenia. Further, early clinical recognition of muscle wasting in patients with PC may convey vital clinical and prognostic information prior to initiating therapy.16,17 Estimates of sarcopenia from radiologic data, however, mandate the use of expensive, densitometric software and correction factors derived empirically for sex and body surface area, thereby hindering widespread clinical use. To address this, we provide proof of principle for a quick calculation to simplify radiologic estimates of sarcopenia, thereby establishing a measure that is performed easily and interpreted by the clinician in the preoperative setting. Indeed, we show that the psoas index, defined as the ratio of the psoas area to vertebral body area measured from the clinically mandated computed tomography (CT), is a statistically powerful predictor of survival superior to tumor grade and lymphatic stage as well as previously validated preoperative biomarkers. These data indicate that the careful quantification of muscularity may have a role in the clinical staging of clinically resectable PC. METHODS Patient population. A prospectively maintained, institutional review board–approved clinical database at the University of Florida was used to identify all consecutive patients with PC who underwent pancreatoduodenectomy (PD) between November 2010 and February 2014. Of the 82 patients identified, 73 had preoperative CT available electronically and suitable for analysis. The time frame facilitated a minimum postoperative follow-up of 6 months to ensure that early and late complications as well as oncologic outcomes were evaluable. Preoperative data collected included patient age, sex, body mass index, Charlson comorbidity status, completion of gemcitabine-based neoadjuvant chemotherapy regimens, and serum hemoglobin and albumin levels. Postoperative pathologic data obtained included lymph node ratio, defined as the number of lymph nodes

Surgery j 2016

positive for malignancy divided by the total number of evaluated lymph nodes, tumor grade, tumor size, and margin status. Postoperative complications were graded according to the Clavien-Dindo index,18 and pancreatic fistulas were categorized using the definition of a pancreatic fistula by the International Study Group of Pancreatic Fistula (ISGPF).19 CT analysis. Images were stored and analyzed using the Philips iSite Enterprise System (Philips, Eindhoven, the Netherlands). As depicted in Fig 1, A, psoas muscles were outlined manually at the level of the inferior portion of the third lumbar vertebra (L3) with both transverse processes in view, as described previously.20-25 The L3 vertebral body was outlined in a similar manner and the cross-sectional area computed automatically by the iSite software. The psoas index was defined as the average psoas cross-sectional area divided by the area of the L3 vertebral body. Statistical analysis. All statistical analyses were performed using the SPSS version 22.0 statistical software package (IBM SPSS statistics for Windows; IBM Corp, Armonk, NY) and GraphPad Prism version 5.02 (GraphPad Software, San Diego, CA). Continuous variables were analyzed using the independent samples t test to compare 2 groups or 1-way analysis of variance to compare groups of $3. Kaplan-Meier survival curves were generated using dichomotized continuous variables at median values and evaluated using the log-rank (or Mantel-Cox) test. Median values were used as cutoffs for dichotomization only in Kaplan-Meier analyses to generate consistently equal groups of patients for each variable. Univariate analysis was performed using a Cox proportional hazards model. All variables significant (P < .05) on univariate analysis were included in multivariate Cox regression. Receiver operating characteristic (ROC) curves were generated using GraphPad Prism. Sensitivity and specificity for each variable to predict death within one year of PD was evaluated in the cohort of 68 patients with at least one year of clinical follow-up. ROC curves were plotted against lines of identity, and the area under the curve (AUC) was calculated. Post hoc power calculations were performed to determine statistical robustness using G*Power version 3.9.1.2 as described previously setting a to 0.05.26,27 RESULTS Low psoas index correlates with older age, preoperative anemia, and hypoalbuminemia. Of the 82 consecutive patients undergoing PD for PC,

ARTICLE IN PRESS Surgery Volume j, Number j

Delitto et al 3

Fig 1. (A) Calculation of the psoas index. Preoperative computed tomography scans were analyzed from 73 patients undergoing pancreatic head resection for pancreatic cancer. The psoas index was calculated from the average crosssectional area of each iliopsoas muscle divided by the cross-sectional area of the vertebral body. All computed tomography scans were analyzed at the level of the third lumbar vertebra with each transverse process in view. The psoas index correlates with (B) preoperative serum albumin and (C) hemoglobin levels, producing a model of preoperative “health” using surrogate measures of nutrition and catabolism. P values were calculated using Pearson correlations. The relationship between (D) the psoas index, (E) albumin, and (F) hemoglobin concentrations with postoperative morbidity was evaluated. Bars represent mean ± standard error of the mean. P values were calculated using independent samples t tests. *P < .05.

electronically stored CTs suitable for analysis were available for 73 patients. Mean psoas indices are shown for each clinicopathologic subgroup in Table I. Men tended to have greater mean psoas

indices compared to women (0.63 vs 0.57; P = .083). The psoas index decreased with increasing age (r =
0.27; P = .019) in addition to demonstrating correlations with both

ARTICLE IN PRESS 4 Delitto et al

Surgery j 2016

Table I. Patient demographics and clinicopathologic parameters Variable

Count (%)

Sex Male 43 (58.9) Female 30 (42.9) Age, y <65 37 (50.7) 65–74 22 (30.1) $75 14 (19.2) Body mass index <25 33 (45.2) 25–30 22 (30.1) >30 18 (24.7) Hemoglobin, g/dLy <12 29 (40.3) $12 43 (59.7) Serum albumin, g/dL <3.5 23 (31.5) $3.5 50 (68.5) Neoadjuvant therapy Yes 13 (17.8) No 60 (82.2) Major morbidity (Clavien 3–5) Yes 14 (19.2) No 59 (80.8) Tumor sizez <3 cm 31 (43.1) $3 cm 41 (56.9) Tumor differentiationx Well/moderate 39 (55.7) Poor 31 (44.3) N stage N0 15 (20.5) N1 58 (79.5) Positive operative margin Yes 25 (34.2) No 58 (65.8)

Mean psoas index (SE)

P value

0.63 (.025) 0.57 (.021)

.083

0.66 (.024) 0.55 (.028) 0.55 (.035)

.007*

0.57 (.021) 0.64 (.032) 0.62 (.040)

.137

0.55 (.017) 0.64 (.025)

.002*

0.55 (.027) 0.63 (.021)

.035*

0.58 (.034) 0.61 (.020)

.515

0.61 (.042) 0.61 (.019)

.935

0.64 (.026) 0.58 (.023)

.066

0.63 (.022) 0.59 (.028)

.253

0.64 (.040) 0.59 (.019)

.289

0.61 (.023) 0.59 (.025)

.488

*Indicates a significant result. yPreoperative hemoglobin was unavailable for one patient. zOne cancer was excluded from tumor size analyses due to the presence of scattered microscopic foci status post–neoadjuvant therapy. xThree adenosquamous cancers were excluded from the analysis of tumor differentiation. Groups were compared using the independent samples t test for dichotomous variables and 1-way analysis of variance for $3 groups.

preoperative albumin and hemoglobin levels (r = 0.29, P = .014; and r = 0.41, P < .001; Fig 1, B and C). Low psoas index, therefore, defines a population characterized loosely by older, anemic, and hypoalbuminemic patients. Because age, anemia, and hypoalbuminemia are all associated with operative morbidity,28 we asked whether a low psoas index also correlated with postoperative complications. Unlike anemia and

the trend observed with hypoalbuminemia, however, a low psoas index was not predictive of major morbidity as defined by the Clavien-Dindo classification (Table I, Fig 1, D–F).18 Specifically, the psoas index did not correlate with duration of stay in days (r = –0.09; P = .446) and was not significantly different in patients with clinically important pancreatic leaks (Grades B–C by ISGPF definition19) (P = .244), postpancreatectomy hemorrhage (P = .107), or wound infections (P = .536). Therefore, despite demonstrating correlations with known risk factors, a low psoas index was not associated with postoperative morbidity. Low psoas index correlates with increased preoperative serum CA 19-9 levels and a high degree of lymphatic metastasis. The relationship between oncologic-specific factors and the psoas index is shown in Table I and Fig 2. Tumor mass and histology were not associated with differences in mean psoas indices, suggesting that local tumor burden per se did not significantly influence muscle mass. A high lymph node ratio, however, correlated with low psoas indices, establishing a link between metastatic spread and myopenia (r =
0.24; P = .039; Fig 2, A–C). Additionally, patients with high preoperative serum CA 19-9 levels, shown previously to significantly correlate with PC disease burden,29 presented with lesser psoas indices (r =
0.24; P = .048; Fig 2, D–F). Low psoas index is associated with poor longterm survival. To examine the prognostic value of the psoas index compared to hemoglobin and albumin levels, dichotomized groups of patients were generated based on median values. A Kaplan-Meier analysis was performed in this manner due to the lack of a normal cutoff value for the psoas index, because this work represents the first investigation to analyze this particular method. Both the psoas index (P = .001) and albumin (P < .001) proved to be highly predictive of long-term survival when analyzed in this fashion (Fig 3, A). To better evaluate the effect of continuous variables on long-term survival, we employed a Cox proportional hazards model (Table II). Univariate analysis revealed strong correlations between overall survival and the psoas index (hazard ratio [HR] 0.014; P < .001), serum albumin levels (HR 0.43; P < .001), hemoglobin (HR 0.80; P = .014), lymph node ratio (HR 4.92; P = .017), poor tumor differentiation (HR 1.96; P = .022), and a microscopically positive operative margin (HR 2.02; P = .016). In addition, a trend was observed between major morbidity (Clavien 3–5) and poor survival

ARTICLE IN PRESS Surgery Volume j, Number j

Delitto et al 5

Fig 2. The psoas index correlates with both lymph node ratio and preoperative serum CA 19-9 levels. Scatter plots of (A, D) the psoas index, (B, E) albumin, and (C, F) hemoglobin were produced with respect to both (A–C) the degree of lymphatic metastasis expressed as the positive lymph node ratio and (D–F) serum CA 19-9 levels. Preoperative serum CA 19-9 levels were not performed in 5 patients. P values were calculated using Pearson correlations. *P < .05.

(HR 1.95; P = .053). Incorporating variables demonstrating significant associations with survival into a multivariate analysis demonstrated that, in contrast to albumin and hemoglobin, the psoas index proved to be an independent predictor of long-term survival (HR 0.021; P = .003). Low psoas index is predictive of death within one year of resection. We next sought to examine specifically the potential effectiveness of a low psoas index as a clinical test to predict death within one year of PD. ROC curves were generated for each continuous variable significant on univariate analysis to predict long-term survival (Fig 3, B). Indeed, the psoas index demonstrated both high sensitivity and specificity in predicting at least one year of postoperative survival compared to serum albumin, hemoglobin, and lymph node ratio (AUC = 0.778 vs 0.696, 0.651, and 0.652, respectively); however, application of the McNemar test using median values to separate high and low categories demonstrated that the superiority of the psoas index as a clinical test to predict death within one year of resection did not reach statistical significance compared to either albumin (P = .77) or hemoglobin (P = .75). Therefore, a low psoas index is predictive of death within one year of pancreatic resection for cancer but not necessarily more predictive than other known risk factors. A decrease in the psoas index during therapy is associated with a poor prognosis. In order to examine changes in muscularity during times of

chemotherapy, the psoas index was evaluated both before and after gemcitabine-based, neoadjuvant chemotherapy. Pretherapy CTs were available and suitable for analysis in 8 of 13 patients who underwent neoadjuvant therapy. Two patients experienced a decrease of at least 20% of their psoas index over the course of the neoadjuvant chemotherapy, which corresponded to postoperative survival times of 2 and 3 months (Fig 3, C). In contrast, all patients who underwent neoadjuvant chemotherapy surviving at least 6 months postoperatively did not experience this decrease in the psoas index during the course of therapy. CTs were also available approximately 3 months postoperatively for 4 patients experiencing death within 6 months of PD; this cohort demonstrated a mean decrease of 34% in their psoas index within the 3month postoperative period (Fig 3, D). Conversely, 8 patients surviving at least one year postoperatively demonstrated a mean decrease of 11% on their 6-month scan. While the small sample size and incongruent timing of patient imaging are confounding factors, the data, nonetheless, illustrate the capacity for the psoas index to change considerably with physiologic decompensation over only a few months’ time, suggesting it may be useful to monitor ongoing care. DISCUSSION The relationship between preoperative nutritional status, tumor burden, and long-term

ARTICLE IN PRESS 6 Delitto et al

Surgery j 2016

Fig 3. (A) High psoas indices and serum albumin levels associate with improved long-term survival. Kaplan-Meier survival curves were generated for psoas index, serum albumin, and hemoglobin levels. Groups were dichotomized based on median preoperative values and compared using the log-rank test. (B) The psoas index is both sensitive and specific in predicting 1-year survival postoperatively from pancreatic head resection for ductal adenocarcinoma. Patients remaining alive with between 6 months and 1 year of clinical follow-up (n = 5) were excluded. Receiver operating characteristic curves were then generated for the psoas index, serum albumin, and hemoglobin levels as well as the pathologically assessed lymph node ratio. (C–D) A decline in muscularity is associated with a poor prognosis. (C) The psoas index was calculated from patients undergoing neoadjuvant chemotherapy both prior to and after completion of gemcitabine-based regimens. Indices were normalized as a percentage of pretherapy values. Bars indicate mean ± standard error of the mean. (D) The psoas index was calculated at the indicated times for patients undergoing curative resection for PC and normalized as a percentage of preoperative values. Groups were generated based on postoperative overall survival. Data are presented as mean ± standard error of the mean. *P < .05.

postoperative outcomes in PC are well known. Our present study demonstrates the prognostic value of the psoas muscle index, a statistically powerful tool to identify patients at high risk of early postoperative mortality. Indeed, the psoas index reported herein represents a straightforward calculation available to any clinician with access to a patient’s imaging CT. A low psoas index correlates with poor preoperative nutritional parameters, advanced age, lymphatic metastasis, and poor long-term survival after a PD for PC.

The association between the psoas muscle area and survival in PC has been established previously by Peng et al.22 The key difference in our work includes the normalization of psoas area to vertebral body area rather than both patient sex and height. Peng et al22 described a mean adjusted psoas area for men (611 mm2/m2) and women (454 mm2/ m2) in a large series of patients, demonstrating ultimately a correction factor of 611/454 = 1.35 for men versus women. Similarly, in our series, the average vertebral body areas for men (13.4 cm2)

ARTICLE IN PRESS Delitto et al 7

Surgery Volume j, Number j

Table II. The preoperative psoas index is the single best predictor of long-term survival in resected pancreatic head cancer HR Univariate analysis Psoas index Preoperative albumin (g/dL) Preoperative hemoglobin (g/dL) Age (y) Female sex BMI Charlson comorbidity index Neoadjuvant chemotherapy Major morbidity (Clavien 3–5) Tumor size (cm) Positive lymph node ratio Poor tumor differentiation Positive operative margin Multivariate analysis Psoas index Preoperative albumin (g/dL) Preoperative hemoglobin (g/dL) Positive lymph node ratio Poor tumor differentiation Positive operative margin

P value

95% CI for HR

0.014 0.43 0.80 1.004 0.97 0.98 0.96 1.85 1.95 1.07 4.92 1.96 2.02

0.002 0.26 0.67 0.98 0.74 0.94 0.82 0.94 0.99 0.84 1.33 1.10 1.14

0.12 0.69 0.96 1.03 1.29 1.03 1.14 3.64 3.85 1.37 18.26 3.50 3.57

<.001* <.001* .014* .783 .847 .487 .963 .077 .053 .590 .017* .022* .016*

0.021 0.68 1.04 1.48 2.12 2.34

0.002 0.40 0.85 0.29 1.11 1.20

0.26 1.15 1.27 7.55 4.04 4.57

.003* .145 .719 .635 .023* .012*

*Indicates a significant result. A total of 73 patients undergoing pancreatoduodenectomy for pancreatic cancer were followed postoperatively. Survival analysis was performed using a univariate Cox proportional hazards model. Variables demonstrating significant associations with overall survival on univariate analysis (P < .05) were incorporated into multivariate analysis. BMI, Body mass index; CI, confidence interval; HR, hazard ratio.

and women (10.1 cm2) demonstrated a very similar correction factor of 13.4/10.1 = 1.33 for men versus women to that of Peng et al,22 suggesting the vertebral body area may be a suitable candidate for efficient normalization of psoas area between sexes. Our index effectively avoids the need to artificially dichotomize each sex to “sarcopenic” in our statistical analysis. Our work also characterizes novel trends between muscularity and oncologic factors such as lymphatic metastasis but not tumor size or histology, further defining a relationship between advanced systemic disease and myopenia. This link was reported recently by Malietzis et al13 in colorectal cancer, which established a greater proportion of myopenic patients in stage II versus stage I disease. Recent analyses in PC by Tan et al30 and Cooper et al31 take advantage of software algorithms to quantify total skeletal muscle at L3 based on radiodensity measurements normalized to patient height in meters squared. The prognostic value of normalized muscular area as defined by these investigations, however, is less apparent, because neither study demonstrated a direct relationship

between preoperative skeletal muscle area and long-term outcomes in patients with PC. More importantly, the software required to perform the image-based analysis utilized in these investigations is not available in the majority of clinical settings. Limitations of the current study include its retrospective nature and a relatively small sample size. Every attempt was made to remove bias from this cohort by establishing prospectively a time frame in which patients were likely to have stored electronic CTs and including all consecutive patients undergoing PD for PC within this time frame. Additionally, we included only patients with pancreatic head cancer in an effort to analyze a more homogenous population. To address our small sample size, we performed a post hoc power analyses using resultant Pearson correlation coefficients observed in this study between the psoas index and clinicopathologic parameters (;0.3), demonstrating a power level (1 - b) of 0.83 for our sample size. Alternatively, using psoas indices from patients surviving <1 year (0.54 ± 0.11) vs those surviving >1 year postoperatively (0.67 ± 0.15), as evaluated in Fig 3, D, the estimated power from our sample

ARTICLE IN PRESS 8 Delitto et al

size was 0.988. Therefore, despite a smaller sample size than other series, the power of our study appears to be sufficient to support our conclusions. The potential error introduced by defining the psoas area manually represents another potential criticism of this work; however, the relative ease of defining psoas anatomy as well as vertebral body area provides a simple methodology for any practicing clinician to employ, because the majority of imaging software enables the user to measure the single image, cross-sectional area relatively easily. Further, this methodology continues to be supported by a wealth of data investigating the link between muscle wasting and clinical outcomes.20-24,32 The theoretic basis for the prognostic superiority of clinically measurable myopenia over serum markers of malnutrition stems from known evolutionary pressures as a result of chronic physiologic stress. The cachectic process in cancer entails a functional redistribution of whole-body protein stores to maintain host energy requirements and provide precursors for hepatic acute-phase protein synthesis.33,34 The preservation of essential physiologic functions, therefore, requires the host to maintain serum levels of mobilized protein and hemoglobin at the expense of skeletal muscle stores. Thus, the direct measurement of skeletal muscle stores may represent a more sensitive method of detecting PC-associated cachexia, which is supported by this work. We therefore propose the psoas index as a potential prognostic tool in resectable pancreatic cancer. The data presented herein reinforce a widely reported link between nutritional status, muscle wasting, the systemic effects of advanced cancer, and overall survival. While the predictive nature of the psoas index must be confirmed prospectively, the capacity to identify patients with clinically resectable PC at high risk for poor survival postoperatively would represent a major advance in the field. Further, emerging evidence indicates that cancer cachexia may be modifiable through multidisciplinary interventions entailing nutritional support, exercise, and targeted systemic therapies.11 In addition to enhancing physiologic reserve with nutrition and exercise programs, therapies designed to slow or reverse cancer cachexia are evolving. A combination of these modalities may benefit cachectic patients, but defining this population and measuring cachexia-specific responses is essential. By defining muscle wasting in PC as a powerful prognostic factor, this work serves as a strong foundation to further support efforts to

Surgery j 2016

understand the biology and potential interventions associated with cachexia in PC.

REFERENCES 1. Rahib L, Smith BD, Aizenberg R, Rosenzweig AB, Fleshman JM, Matrisian LM. Projecting cancer incidence and deaths to 2030: the unexpected burden of thyroid, liver, and pancreas cancers in the United States. Cancer Res 2014;74:2913-21. 2. Hishinuma S, Ogata Y, Tomikawa M, Ozawa I, Hirabayashi K, Igarashi S. Patterns of recurrence after curative resection of pancreatic cancer, based on autopsy findings. J Gastrointest Surg 2006;10:511-8. 3. Ansari D, Rosendahl A, Elebro J, Andersson R. Systematic review of immunohistochemical biomarkers to identify prognostic subgroups of patients with pancreatic cancer. Br J Surg 2011;98:1041-55. 4. Cleary SP, Gryfe R, Guindi M, Greig P, Smith L, Mackenzie R, et al. Prognostic factors in resected pancreatic adenocarcinoma: analysis of actual 5-year survivors. J Am Coll Surg 2004;198:722-31. 5. Qin R, Smyrk TC, Reed NR, Schmidt RL, Schnelldorfer T, Chari ST, et al. Combining clinicopathological predictors and molecular biomarkers in the oncogenic K-RAS/Ki67/ HIF-1alpha pathway to predict survival in resectable pancreatic cancer. Br J Cancer 2015;112:514-22. 6. Wu N, Chen G, Hu H, Pang L, Chen Z. Low pretherapeutic serum albumin as a risk factor for poor outcome in esophageal squamous cell carcinomas. Nutr Cancer 2015;67:4815. 7. Park JW, Jang JY, Kim EJ, Kang MJ, Kwon W, Chang YR, et al. Effects of pancreatectomy on nutritional state, pancreatic function and quality of life. Br J Surg 2013;100:1064-70. 8. La Torre M, Ziparo V, Nigri G, Cavallini M, Balducci G, Ramacciato G. Malnutrition and pancreatic surgery: prevalence and outcomes. J Surg Oncol 2013;107:702-8. 9. Tu MY, Chien TW, Chou MT. Using a nutritional screening tool to evaluate the nutritional status of patients with colorectal cancer. Nutr Cancer 2012;64:323-30. 10. Poziomyck AK, Weston AC, Lameu EB, Cassol OS, Coelho LJ, Moreira LF. Preoperative nutritional assessment and prognosis in patients with foregut tumors. Nutr Cancer 2012;64:1174-81. 11. Fearon K, Arends J, Baracos V. Understanding the mechanisms and treatment options in cancer cachexia. Nat Rev Clin Oncol 2013;10:90-9. 12. Judge SM, Wu CL, Beharry AW, Roberts BM, Ferreira LF, Kandarian SC, et al. Genome-wide identification of FoxOdependent gene networks in skeletal muscle during C26 cancer cachexia. BMC Cancer 2014;14:997. 13. Malietzis G, Johns N, Al-Hassi HO, Knight SC, Kennedy RH, Fearon KC, et al. Low muscularity and myosteatosis is related to the host systemic inflammatory response in patients undergoing surgery for colorectal cancer. Ann Surg 2016;263:320-5. 14. Johns N, Hatakeyama S, Stephens NA, Degen M, Degen S, Frieauff W, et al. Clinical classification of cancer cachexia: phenotypic correlates in human skeletal muscle. PLoS One 2014;9:e83618. 15. Blum D, Stene GB, Solheim TS, Fayers P, Hjermstad MJ, Baracos VE, et al. Validation of the consensus-definition for cancer cachexia and evaluation of a classification model–a study based on data from an international multicentre project (EPCRC-CSA). Ann Oncol 2014;25:1635-42.

ARTICLE IN PRESS Surgery Volume j, Number j

16. Fearon KC, Baracos VE. Cachexia in pancreatic cancer: new treatment options and measures of success. HPB (Oxford) 2010;12:323-4. 17. Wigmore SJ, Plester CE, Richardson RA, Fearon KC. Changes in nutritional status associated with unresectable pancreatic cancer. Br J Cancer 1997;75:106-9. 18. Dindo D, Demartines N, Clavien PA. Classification of surgical complications: a new proposal with evaluation in a cohort of 6336 patients and results of a survey. Ann Surg 2004;240:205-13. 19. Bassi C, Dervenis C, Butturini G, Fingerhut A, Yeo C, Izbicki J, et al. Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery 2005;138:813. 20. Sheetz KH, Zhao L, Holcombe SA, Wang SC, Reddy RM, Lin J, et al. Decreased core muscle size is associated with worse patient survival following esophagectomy for cancer. Dis Esophagus 2013;26:716-22. 21. Sabel MS, Terjimanian M, Conlon AS, Griffith KA, Morris AM, Mulholland MW, et al. Analytic morphometric assessment of patients undergoing colectomy for colon cancer. J Surg Oncol 2013;108:169-75. 22. Peng P, Hyder O, Firoozmand A, Kneuertz P, Schulick RD, Huang D, et al. Impact of sarcopenia on outcomes following resection of pancreatic adenocarcinoma. J Gastrointest Surg 2012;16:1478-86. 23. Sabel MS, Lee J, Cai S, Englesbe MJ, Holcombe S, Wang S. Sarcopenia as a prognostic factor among patients with stage III melanoma. Ann Surg Oncol 2011;18:3579-85. 24. Peng PD, van Vledder MG, Tsai S, de Jong MC, Makary M, Ng J, et al. Sarcopenia negatively impacts short-term outcomes in patients undergoing hepatic resection for colorectal liver metastasis. HPB (Oxford) 2011;13:439-46. 25. Goodpaster BH, Kelley DE, Thaete FL, He J, Ross R. Skeletal muscle attenuation determined by computed tomography is associated with skeletal muscle lipid content. J Appl Physiol (1985) 2000;89:104-10.

Delitto et al 9

26. Faul F, Erdfelder E, Lang AG, Buchner A. G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav Res Methods 2007;39: 175-91. 27. Faul F, Erdfelder E, Buchner A, Lang AG. Statistical power analyses using G*Power 3.1: tests for correlation and regression analyses. Behav Res Methods 2009;41:1149-60. 28. Khuri SF, Daley J, Henderson W, Hur K, Demakis J, Aust JB, et al. The Department of Veterans Affairs’ NSQIP: the first national, validated, outcome-based, risk-adjusted, and peercontrolled program for the measurement and enhancement of the quality of surgical care. National VA Surgical Quality Improvement Program. Ann Surg 1998;228:491-507. 29. Ballehaninna UK, Chamberlain RS. The clinical utility of serum CA 19-9 in the diagnosis, prognosis and management of pancreatic adenocarcinoma: an evidence based appraisal. J Gastrointest Oncol 2012;3:105-19. 30. Tan BH, Birdsell LA, Martin L, Baracos VE, Fearon KC. Sarcopenia in an overweight or obese patient is an adverse prognostic factor in pancreatic cancer. Clin Cancer Res 2009;15:6973-9. 31. Cooper AB, Slack R, Fogelman D, Holmes HM, Petzel M, Parker N, et al. Characterization of anthropometric changes that occur during neoadjuvant therapy for potentially resectable pancreatic cancer. Ann Surg Oncol 2015;22: 2416-23. 32. Hasselager R, Gogenur I. Core muscle size assessed by perioperative abdominal CT scan is related to mortality, postoperative complications, and hospitalization after major abdominal surgery: a systematic review. Langenbecks Arch Surg 2014;399:287-95. 33. Norton JA, Stein TP, Brennan MF. Whole body protein synthesis and turnover in normal man and malnourished patients with and without known cancer. Ann Surg 1981;194: 123-8. 34. Delano MJ, Moldawer LL. The origins of cachexia in acute and chronic inflammatory diseases. Nutr Clin Pract 2006; 21:68-81.