1 INTRODUCTION

Cancer cachexia is a state of progressive weight loss and decrease in body composition caused by various factors, including reduced food intake, alteration of metabolic networks, excessive catabolism, and inflammation caused by cancer.1 Cachexia is commonly assessed by body composition parameters, including subcutaneous fat area (SFA), visceral fat area (VFA), and skeletal muscle index (SMI) measured from computed tomography (CT) images.2-4 There is increasing evidence that chemotherapy induces cachexia, weight loss, and decrease in body composition parameters,5-7 and that cachexia progresses during chemotherapy.8-11 As a result, progression of cachexia affects survival of malignant disease.12-15

Pancreatic cancer (PC) has poor prognosis. Roughly 80% of patients with PC have unresectable, locally advanced, or metastatic disease at diagnosis.16 The 5-year survival rate of patients with unresectable PC is only 3%.17, 18 The poor outcome of PC is mainly attributable to cachexia, which is observed in >85% of patients, and this is a higher rate than for other types of malignancy.19 Nearly one-third of PC deaths are because of cachexia rather than tumor burden.20, 21 A study that reviewed advanced pancreatic cancer (APC) patients who underwent FOLFIRINOX (FFX) demonstrated that skeletal muscle loss during chemotherapy is significantly related to shorter overall survival (OS).22

For unresectable cases with adequate performance status (PS), combination chemotherapy regimens, modified FFX (mFFX),23, 24 and gemcitabine/nab-paclitaxel (GEM/nab-PTX; GnP)24 have been shown to be effective as first-line therapy (L1). However, median progression-free survival (PFS) was only around 6 months. Extension of OS after L1 is important to improve prognostic outcome for patients with APC. A consensus on second-line therapy (L2) started to be built by the CONKO-03 trial, which showed that oxaliplatin, 5-fluorouracil (5-FU), and folinic acid improved survival compared with best supportive care.25 Recently, nanoliposomal irinotecan, 5-FU, and folinic acid were established as the standard L2 regimen in a global phase III trial (NAPOLI-1).26 National Comprehensive Cancer Network guidelines recommend use of several regimens as L2 on the condition that these are administered to patients with good PS, which is the only reliable prognostic factor.27, 28 However, there are no clear indications of suitable patients for L2. Changes in body composition during L1, which reflect the patient's condition, might be a prognostic factor for L2. The main aim of this study was to evaluate the impact of cachexia progression during L1 on survival of L2 and establish prognostic factors for APC refractory to GnP.

2 METHODS 2.1 Patients

We reviewed patients diagnosed with unresectable PC between January 2015 and December 2019 at the Department of Hepato-Biliary-Pancreatology, National Hospital Organization Kyushu Cancer Center, who received GnP as L1 and mFFX or S-1 as L2. To analyze the real-world data, we included patients who discontinued L1 because of disease progression or adverse events. We excluded patients who had a history of surgical resection or received chemotherapy in other institutions. We also excluded patients who, for any reason, discontinued L2 within two courses. All patients underwent CT scan within 4 weeks before the start of L1 and 2 weeks before the start of L2.

The present study was conducted according to the guidelines laid down in the Declaration of Helsinki and was approved by the Ethics Board of Kyushu Cancer Center (approval number: 2019-40).

2.2 Treatment and assessment

For L1, all patients received intravenous infusion of 125 mg/m2 nab-PTX, followed by infusion of 1 g/m2 GEM on days 1, 8, and 15 every 4 weeks. For L2, patients treated with mFFX received intravenous infusion of oxaliplatin (85 mg/m2), irinotecan (180 mg/m2), and leucovorin (400 mg/m2) on day 1, followed by 2.4 g/m2 5-FU as 46-h continuous infusion every 2 weeks. For the patients treated with S-1, 40–60 mg was administered orally according to body surface area, twice daily on days 1–14 of a 21-day cycle. Dose reduction was decided upon by the physician, based on toxicity. We assessed the effect of chemotherapy by CT scan every 2–3 months according to the Response Evaluation Criteria in Solid Tumors (RECIST) version 1.0. L1 was discontinued if there was disease progression or any unacceptable adverse events. L2 regimen was chosen based on the physician's decision or patient's demand after being offered both regimens.

2.3 3 Body composition analysis

Cross-sectional CT images of the third lumbar vertebra were used to evaluate the detailed parameters of body composition, including SFA, VFA, and SMI analyzed by SYNAPSE VINCENT® software (Fuji Medical Systems, Tokyo, Japan). The abdominal muscular fat boundaries were manually sketched at the level of the lower part of the third lumbar vertebra, as described previously.29 To calculate cross-sectional area, we used predefined Hounsfield unit (HU) ranges as follows: −190 to −30 for subcutaneous fat, −150 to −50 for visceral fat, and −29 to 150 for skeletal muscle. SMI, normalized for height, was calculated as follows: SMA (cm2)/height (m2).

2.4 Data collection

In addition to disease status, response to chemotherapy, and body composition data, we collected clinical data at the start of L1 and L2, including age, sex, height, body weight (BW), Eastern Cooperative Oncology Group (ECOG) PS, laboratory data, and tumor markers. We also collected survival data. The primary endpoint of this study was OS from the start of L2 to death (OS2). In the absence of confirmed death, OS2 was censored at the last date the patient was known to be alive.

2.5 Statistical analysis

Categorical variables are expressed as frequencies and proportions, and group data were compared using Fisher's exact test. Continuous variables with normal distribution are expressed as medians and interquartile ranges (IQRs), and values were compared using the Mann–Whitney U test. Cutoff values for each body composition parameter were explored using receiver operating characteristic (ROC) curves. Correlation between rates of change in body composition parameters was calculated by Pearson correlation. Survival curves were estimated using the Kaplan–Meier method, and the differences were analyzed using the log-rank test. Univariate and multivariate hazard ratios (HRs) were calculated using Cox proportional hazards regression, and all significant variables in univariate analysis were included in multivariate analysis for adjustment of confounding factors. A two-sided P < 0.05 was considered significant. All statistical analyses were performed using Easy R version1.3.6.30

3 RESULTS 3.1 Patient demographics

From 440 patients with unresectable PC diagnosed during 2015–2019, we identified 59 patients in the present study (Figure 1). The characteristics at the start of L2 after GnP therapy are shown in Table 1. Median age was 67 years, and 32 patients (54.2%) were male. Most patients (96.6%) had metastasis, and 44 (74.6%) had liver metastasis. The response rate to GnP was 33%, and PFS was 5.25 months. GnP was discontinued because of disease progression in 50 patients (84.7%) and severe adverse events in nine patients (16.3%). Median changes during L1 in BW SMI, SFA, and VFA were 0.19%, −4.17%, and −18.39%, respectively. Median follow-up time of the entire cohort was 13.4 months.

Patient selection flow

TABLE 1. Patient characteristics at the start of second line therapy n/median % [IQR] Age (years) 67 [60, 70] Sex Male 32 (54.2) Primary tumor location Head 23 (39.0) Disease status Locally advance 2 (3.4) Metastatic 57 (96.6) Metastatic disease Liver 44 (74.6) Peritoneal dissemination 21 (35.6) Lung 10 (17.0) ECOG PS 0 28 (48.3) 1 22 (37.9) 2 8 (13.8) Best effect of the first line therapy Partial response 20 (33.9) Stable disease 24 (40.7) Progressive disease 15 (24.4) RDI of the first line therapy (%) 76 [59, 93] PFS of the first line therapy (months) 5.25 [2.92, 7.30] Body composition at the start of second line BMI 21.00 [19.70, 22.50] SMI (cm2/ m2) 39.38 [35.06, 44.07] SFA (cm2) 7869 [3547.00, 10,757.50] VFA (cm2) 5597 [3524.50, 10,618.50] Change rate of Body composition during the first line therapy BMI (%) –2.30 [–7.93, 1.92] SMI (%) 0.19 [–7.53, 6.48] SFA (%) –4.17 [–35.91, 22.16] VFA (%) –18.39 [–37.91, 14.37] Second-line regimen mFFX 29 (49.2) S-1 30 (50.8) Abbreviations: BMI, body mass index; mFFX, modified FOLFIRINOX; PFS, progression-free survival; PS, performance status; RDI, relative dose intensity; SFA, subcutaneous fat area; SMI, skeletal muscle index; VFA, visceral fat area. 3.2 Cutoff value and correlations between changes in body composition

Cutoff values for each body composition parameter were established using ROC curves. The cutoff values were selected on the basis of best accuracy in relation to 6-month mortality, which was almost the same as median OS. The cutoff values for changes in SFA, VFA, and SMI were −8.469 (area under the curve [AUC] = 0.625), −34.09 (AUC = 0.543), and −8.615 (AUC = 0.508), respectively (Figure 2). Therefore, we defined loss of SFA, VFA, and SMI as decreases greater than 8.5%, 34.1%, and 8.7%, respectively. Thirty patients had no loss of adipose tissue (AT) or SMI. Among the remaining 29 patients, the number with loss of both AT and SMI, SMI only, and AT only were 10, four, and 15, respectively (Figure S1).

Receiver operating characteristic curves of (A) subcutaneous fat area (SFA), (B) visceral fat area (VFA), and (C) skeletal muscle index (SMI) as predictors of overall survival after initiation of second-line therapy

3.3 Correlation among changes in SFA, VFA, SMI, and BW

There was a moderate correlation between change in SFA and VFA (r = 0.668, P < 0.00001) (Figure 3A ). According to the correlation between change in SFA and VFA, AT loss (SFA or VFA loss) was used for the following analysis. There was no significant correlation between change in SFA and SMI (r = 0.162, P = 0.219) (Figure 3B), and between VFA and SMI (r = 0.142, P = 0.285).

Correlation between changes in body composition parameters of (A) subcutaneous fat area (SFA) and visceral fat area (VFA), and (B) SFA and skeletal muscle index (SMI). Data of all 59 patients were used in this analysis [Colour figure can be viewed at wileyonlinelibrary.com]

Survival curve. Kaplan–Meier curves for overall survival after initiation of second-line therapy. Survival in patients with advanced pancreatic cancer according to loss of body composition. (A) Adipose tissue loss (subcutaneous fat area [SFA] loss ≥8.5% and/or visceral fat area [VFA] loss ≥34.1%), (B) SMI loss ≥8.7%, and (C) loss of any body composition parameter versus no loss

Moderate correlation between change in BW and SFA (r = 0.42, P < 0.0008), BW and VFA (r = 0.41, P = 0.0014), and BW and SMI (r = 0.36, P = 0.005) were also observed (Figure S2).

3.4 Correlation between changes in inflammatory markers and changes in body composition parameters

We next investigated whether change rate of systematic inflammatory markers (neutrophil–lymphocyte ratio [NLR] and C-reactive protein/albumin ratio [CAR]) affect change in body composition parameters (SFA, VFA, and SMI) during L1. An inverse correlation was observed between change in NLR and SFA change (Figure S3), whereas all other combinations showed no significant correlation (data not shown).

3.5 Survival

The median OS from diagnosis and median OS2 were 14.3 and 6.3 months (Figure S4), respectively. The survival of patients with AT loss (SFA or VFA) was significantly worse than that of patients without AT loss (median OS2: 3.8 vs. 8.6 months, P = 0.005) (Figure 4A). Survival of patients with SMI loss was significantly worse than that of patients without SMI loss (median OS2: 3.0 vs. 7.3 months, P = 0.007) (Figure 4B).

Patients with loss in any of the three composition parameters had poorer OS than patients without loss (median OS2: 3.83 vs. 8.73 months, P = 0.0009) (Figure 4C). Although there was a significant difference, patients with both AT and SMI loss tended to have shorter OS than patients with either AT or SMI loss (data not shown).

Cox regression analysis for OS with respect to factors at the start of L2 and change in body composition parameters are summarized in Table 2. In univariate analysis, ECOG PS ≥ 2 (HR = 3.51; 95% confidence interval [CI] = 1.58–7.79; P = 0.002), CAR > 0.16 (HR = 3.26; 95% CI = 1.80–5.89; P = 0.0001), CA19-9 > 1 μg/ml (HR = 2.10; 95% CI = 1.04–4.26; P = 0.038), and AT or SMI loss during L1 (HR = 2.57; 95% CI = 1.44–4.60; P = 0.0014) were negative prognostic factors for survival. As mentioned earlier, moderate correlation between BW change and changes in each body composition was observed. Thus, we conducted multivariate analysis with two models, which included BW loss (model 1) and loss in any of body composition parameter (model 2). In model 1, ECOG PS ≥ 2 (HR = 2.61; 95% CI = 1.13–6.07; P = 0.025) and CAR > 0.16 (HR = 3.03; 95% CI = 1.57–5.87; P = 0.0009) were independent negative prognostic factors for OS2, although BW loss was not. In model 2, ECOG PS ≥ 2 (HR = 2.35; 95% CI = 1.02–5.42; P = 0.044), CAR > 0.16 (HR = 3.47; 95% CI = 1.76–6.84; P = 0.0003), and loss in any body composition parameter (HR = 2.29; 95% CI = 1.20–4.37; P = 0.010) were identified as independent negative prognostic factors for OS2.

TABLE 2. Survival analysis No. of patients Univariate Multivariate-model1 Multivariate-model2 HR 95% CI P-value HR 95% CI P-value HR 95% CI P-value Factors at the start of the second line therapy Age ≥65 19 1.01 (0.56–1.84) 0.97 <65 40 1 Sex Male 32 0.98 (0.55–1.72) 0.96 Female 27 BMI <22 19 0.69 (0.37–1.26) 0.22 ≤22 40 1 ECOG PS ≥2 8 3.51 (1.58–7.79) 0.002* 2.61 (1.13–6.07) 0.025* 2.35 (1.02–5.42) 0.044* <2 51 1 1 1 Tumor location Head 24 1.18 (0.67–2.08) 0.58 Body and Tail 35 1 CAR ≻0.16 25 3.25 (1.80–5.89) 0.0001* 3.03 (1.57–5.87) 0.0009* 3.47 (1.76–6.84) 0.0003* ≤ 0.16 34 1 1 1 NLR ≻4 16 1.67 (0.91–3.06) 0.11 ≤4 43 1 Ca19-9 >1000 ng/ml 37 2.10 (1.04–4.26) 0.038* 2.06 (0.98–4.32) 0.055 1.86 (0.88–3.95) 0.10 ≤1000 ng/ml 16 1 1 1 Second line therapy mFFX 29 1.01 (0.57–1.79) 0.96 S-1 30 1 Decrease rate in body weight ≥5% 1.92 (1.08–3.42) 0.027* 1.28 (0.68–2.43) 0.45 <5% 1 1 Either Fat or SMI loss during the first line therapy Present 30 2.57 (1.44–4.60) 0.0014* 2.29 (1.20–4.37) 0.010* Absent 29 1 1 Note: Fat loss means decrease in visceral fat area (>34.1%) or subcutaneous fat area (>8.5%). Multivariate analysis was conducted between factors with P-value <0.10 in univariate analysis. Abbreviations: BMI, body mass index; CAR, CRP–albumin ratio; NLR, neutrophil–lymphocyte ratio, SMI, skeletal muscle index. 4 DISCUSSION

Several recent studies have reported that development of cancer cachexia, represented by loss in AT and skeletal muscle, has a significant impact on survival in patients with various types of malignant tumors.12-15 Patients with PC are reported to have a higher risk of developing cachectic body composition than patients with other types of malignancies.31 The current study demonstrated that 51% of patients with APC were defined as having cachexia during the therapeutic course,1 strongly suggesting that cachexia is one of the major complications among these patients. Thus, assessment of change in body composition during and after cancer treatment is becoming increasingly important, particularly in patients with APC. The prognostic value of body composition assessment in APC has already been reported by several groups32-34; however, data are scarce for the true effect on the outcome of L2.

Here, we demonstrated for the first time that loss in any of SFA, VFA, and SMI during L1 is a novel negative prognostic factor for L2 in patients with APC. Our finding is in line with a previous study reporting that APC patients with decreased SMI during FFX treatment had significantly shorter OS.22 Also, other studies showed that SMI reduction was a negative prognostic factor in patients with colorectal cancer undergoing L18 and urothelial cancer undergoing L2.35 Several studies have indicated a strong association between AT loss and cancer mortality.36, 37 Accelerated rates of VFA loss have been shown to be associated with reduced survival.38 Another study revealed that prognosis of patients with AT loss was poorer than in patients without AT loss, regardless of the presence or absence of skeletal muscle loss among APC patients undergoing treatment with FFX.39 The results of our study showed that we should focus not only on skeletal muscle loss but also AT loss during L1 when we predict the outcome of APC patients.

In this study, we performed further detailed analysis of the correlation between changes in these parameters. The change in SFA was significantly associated with that of VFA, whereas no significant correlation was observed between AT and SMI. This indicates that decreases in AT and skeletal muscle begin at different phases of cancer progression and proceed at different rates. Although 15 patients (25.