MCO Molecular and Clinical Oncology 2049-9450 2049-9469 D.A. Spandidos 10.3892/mco.2017.1177 MCO-0-0-1177 Articles Non-alcoholic fatty liver disease fibrosis score predicts hematological toxicity of chemotherapy including irinotecan for colorectal cancer YahagiMasashi TsurutaMasashi HasegawaHirotoshi OkabayashiKoji KitagawaYuko Department of Surgery, Keio University School of Medicine, Tokyo 160-8582, Japan Correspondence to: Dr Masashi Tsuruta, Department of Surgery, Keio University School of Medicine, 35 Shinanomachi, Shinjyuku-ku, Tokyo 160-8582, Japan, E-mail: championtsuru@yahoo.co.jp 04 2017 01 03 2017 6 4 529 533 17112016 13012017 Copyright © 2017, Spandidos Publications 2017

Liver dysfunction that may affect drug metabolism is a major concern in patients treated with chemotherapy. Thus, assessment of the degree of liver dysfunction is crucial for predicting the adverse events of chemotherapy. The non-alcoholic fatty liver disease fibrosis score (NFS) is a non-invasive clinical scoring system constructed from routine clinical and laboratory variables. The aim of this study was to evaluate whether NFS was useful for predicting the adverse events of chemotherapy including irinotecan (CPT-11) for colorectal cancer. Between January, 2007 and May, 2013, a total of 87 patients with unresectable/recurrent colorectal cancer who received first-line chemotherapy including CPT-11 were reviewed. Demographic variables, including pretreatment NFS, were retrospectively collected from medical records. The primary outcome was the association between pretreatment NFS and adverse events, such as hematological and non-hematological toxicity, of chemotherapy including CPT-11. The median pretreatment NFS was 1.302 (range, 5.158–2.620). Pretreatment NFS was an independent risk factor for hematological toxicity in a multivariate analysis (coefficient=0.932, 95% CI: 0.083–1.781; P=0.031). Receiver operating characteristic curve analysis identified 0.347 as the optimal cut-off value associated with hematological toxicity. Using this cut-off, high NFS was found to be a significant risk factor for hematological toxicity (coefficient=2.019, 95% CI: 0.239–3.798, P=0.026), but not for non-hematological toxicity (P=0.546). Therefore, based on these results, NFS appears to be a significant predictor of hematological adverse events in chemotherapy including CPT-11 for colorectal cancer and it is a non-invasive, useful tool that may be used for determining regimens or doses of chemotherapy including CPT-11.

 non-alcoholic fatty liver disease fibrosis score colorectal cancer hematological toxicity non-hematological toxicity irinotecan Introduction

Liver dysfunction, which may affect drug metabolism, is a major concern in patients receiving chemotherapy for colorectal cancer. Thus, the assessment of the degree of any liver dysfunction, such as hepatic fibrosis and inflammation, is crucial for predicting the effects or adverse events of chemotherapy. Currently, the gold standard method for the assessment of liver dysfunction is liver biopsy; however, this is a costly and invasive method, with a high risk of complications. In addition, the typical sample is small and may result in sampling error (1,2). Certain benefits of non-invasive clinical scoring systems, constructed from routine clinical and laboratory variables, have been reported for several liver conditions, including non-alcoholic fatty liver disease (NAFLD) (3). Among those, the NAFLD fibrosis score (NFS) has been recommended even for Japanese populations, in which the incidence of obesity is relatively low (4,5).

Irinotecan (CPT-11) is a key drug used to treat unresectable/recurrent colorectal cancer. Polymorphisms in the uridine diphosphate-glucuronosyltransferase (UGT)1A1 gene, which is involved in the metabolism of CPT-11 in the liver, may cause severe adverse events, such as neutropenia (6). However, testing for such polymorphisms is time-consuming and costly. In addition, the number of UGT1A1*28 homozygous patients who experience severe adverse events due to CPT-11 is quite small in the Japanese population. Thus, it is important to identify a more appropriate surrogate marker reflecting liver dysfunction in chemotherapy with CPT-11. The aim of this study was to evaluate whether NFS is useful for predicting the effects and adverse events of chemotherapy including CPT-11 for colorectal cancer.

 Materials and methods <sec> <title>Patients and clinicopathological characteristics

Between January, 2007 and May, 2013, 118 patients were diagnosed with unresectable/recurrent colorectal cancer at Keio University Hospital (Tokyo, Japan), of whom 87 patients who underwent first-line chemotherapy including CPT-11 were enrolled in this study.

The tumor location, from the vermiform appendix (VA) to the rectum, was classified according to the 7th edition of the TNM Classification of Malignant Tumours (http://www.uicc.org/sites/main/files/private/TNM_Classification_of_Malignant_Tumours_Website_15%20MAy2011.pdf). The colon was defined as spanning from the cecum to the sigmoid colon, while ‘other’ was defined as the rectum; thus, tumors of the VA were excluded. In terms of histological types, well- and moderately differentiated tubular adenocarcinomas were defined as highly differentiated carcinomas, whereas poorly differentiated, mucinous adenocarcinomas and signet-ring cell carcinomas were defined as poorly differentiated carcinomas.

Demographic variables were retrospectively collected from medical records and included age, gender, tumor location, histological type, number of metastatic sites (single or multiple), temporal association of metastasis (synchronous or metachronous), cycles of chemotherapy, use of bevacizumab, pretreatment total bilirubin serum levels and pretreatment NFS. NFS was calculated as follows: NFS=1.675 + 0.037x age (years) + 0.094x BMI (kg/m2) + 1.13x impaired fasting glucose/diabetes (yes=1, no=0) + 0.99x AST/ALT ratio-0.013x platelet (x109/l)-0.66x albumin (g/dl), where BMI, body mass index; AST, aspartate aminotransferase; and ALT, alanine aminotransferase.

 Treatment and evaluation

Our regimen including CPT-11 was FOLFIRI or uracil/ftorafur/leucovorin combined with CPT-11 (TEGAFIRI); thus, the participants received FOLFIRI (CPT-11 150 mg/m2 on day 1, leucovorin 200 mg/m2 on day 1, 5-fluorouracil 400 mg/m2 on day 1, 5-fluorouracil 2,400 mg/m2 as a 46-h infusion, every 2 weeks) or TEGAFIRI (uracil/ftorafur 300 mg/m2/day on days 1–21, leucovorin 75 mg/day on days 1–21, CPT-11 150 mg/m2/day on days 1 and 15, every 4 weeks). The National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE), version 4.0, were used to assess toxicity (7).

Treatment was administered until disease progression, development of unacceptable toxicity, or the patient's wish to withdraw. The doses of CPT-11 were lowered by 20% if severe toxic effects (CTCAE grade ≥3) occurred. Patients were evaluated for progression every four cycles by computed tomography. During the follow-up period, any continuing adverse effects were closely monitored.

The primary outcome was the association between pretreatment NFS and adverse events, such as hematological and non-hematological toxicity, of chemotherapy including CPT-11.

This study was approved by the Ethics Review Board of our institute (no. 20140459).

 Statistical methods

Logistic regression analysis was used to identify univariate and multivariate risk factors for hematological and non-hematological toxicity (CTCAE grade ≥3). The coefficient (coef) was used to estimate the impact on hematological and non-hematological toxicity. All predictors with P<0.05 in univariate analysis and clinically important factors were retained in the multivariate models.

A receiver operating characteristic (ROC) curve analysis was performed to determine a cut-off point for NFS as a continuous variable. Youden's index was calculated and the maximum value was used as the optimum cut-off point. Using this cut-off value, the impact on adverse events was also assessed. Each risk factor was assessed using the χ2 test, Mann-Whitney U test, Fisher's exact test and logistic regression analysis. All statistical analyses were two-sided and statistical significance was set at P<0.05. Statistical analyses were performed using the STATA software, version 11; (StataCorp LP, College Station, TX, USA).

 Results <sec> <title>Patient background

The clinicopathological demographics of the 87 patients are summarized in Table I. The median age was 67 years (range, 36–82 years) and 59 of the participants (67.8%) were men. The tumors were located in the colon in 59 patients (67.8%). The median pretreatment total bilirubin serum levels were 0.6 mg/dl (range, 0.3–1.5 mg/dl) and the median NFS was 1.302 (range, 5.158–2.620). Grade ≥3 hematological and non-hematological toxicities occurred in 6 (6.9%) and 13 (14.9%) patients, respectively.

 Association between NFS and adverse events

In the univariate analysis, pretreatment NFS was significantly associated with hematological toxicity (Table II). Furthermore, pretreatment NFS was an independent risk factor for hematological toxicity in a multivariate analysis (coef=0.932, 95% CI: 0.083–1.781; P=0.031), adjusted for gender, age, temporal association of metastasis, cycle number and pretreatment total bilirubin serum levels. ROC analysis, including Youden's index of NFS, identified 0.347 as the optimal cut-off value associated with hematological toxicity (Fig. 1). Using this cut-off, 21 (24.1%) patients were classified as high NFS, while 66 (75.9%) patients had a low NFS (Table III). High NFS was found to be a significant risk factor for hematological toxicity (coef=2.019, 95% CI: 0.239–3.798; P=0.026) but not for non-hematological toxicity (P=0.546). This model exhibited a sensitivity of 83.3%, a specificity of 79.3%, a likelihood ratio for a positive finding of 3.97, and a likelihood ratio for a negative finding of 0.21. However, pretreatment NFS was not identified as a risk factor for non-hematological toxicity in any univariate or multivariate analysis (Table IV). The results were similar even when the investigation was limited to diarrhea among non-hematological toxicities (coef=0.214, 95% CI: −0.514 to 0.942; P=0.564).

 Discussion

Our findings indicate that NFS may be a promising marker for predicting the hematological adverse events of chemotherapy including CPT-11 for colorectal cancer. Liver dysfunction, such as fibrosis, may cause severe CPT-11-related hematological adverse events by hampering its metabolism. The metabolism of CPT-11 occurs via a complex cascade, with hydrolysis of the parent prodrug by hepatic carboxylesterases into 7-ethyl-10-hydroxycamptothecin (SN-38) as an active metabolite, detoxification of SN-38 into an inactive glucuronidated form (SN-38G) by UGT1A1, and excretion of SN-38G in the urine and bile (8). A fraction of SN-38 is also transported directly via ATP-binding cassette transporters into the bile and released in the intestinal lumen (9). In the intestine, SN-38 may undergo glucuronidation by intestinal UGT1A1 or be excreted unchanged in the feces (10). Thus, reduced SN-38G formation is closely associated with severe toxicities. As UGT1A1 is considered to be primarily responsible for the formation of SN-38G, genetic polymorphisms of UGT1A1, such as UGT 1A1*6 and UGT 1A1*28, increase CPT-11-related neutropenia (6,11–13). UGT1A1 is also responsible for the detoxification of bilirubin, and a correlation between pretreatment bilirubin levels and UGT1A1 polymorphisms has been reported (14,15). Thus, high levels of total bilirubin in the serum prior to treatment have been reported to be correlated with neutropenia (12,13,16,17). However, this study did not determine the impact of pretreatment total bilirubin on the adverse effects of chemotherapy including CPT-11; instead, NFS was found to be a robust predictor of hematological adverse events. As the enzyme activity of UGT1A1 was reported to be largely unchanged by liver fibrosis in rats (18), other mechanisms through which liver dysfunction may cause CPT-11-related adverse effects have been suggested. However, CYP3A, one of the most important P450 enzymes, was dysregulated in those rats (19). Considering that bilirubin was not a significant predictor in our study, liver fibrosis may affect the metabolism of CPT-11, through mechanisms other than via UGT1A1, such as CYP3A4 or a bile duct transporter, and result in higher concentrations of SN-38.

Regarding non-hematological toxicity, the inactive metabolite SN-38G is secreted into the duodenum and deconjugated into SN-38 by intestinal bacterial β-glucuronidase, which may cause diarrhea (20,21). SN-38 is also converted into SN-38G, catalyzed by the UGT enzyme in the intestine (10). Thus, polymorphisms of UGT1A1 have been reported to increase not only neutropenia but also delayed diarrhea (22). NFS was not significantly associated with non-hematological toxicity in our findings, even when the investigation was limited to diarrhea, as liver fibrosis may not affect the activity of UGT or bacterial β-glucuronidase, at least in the intestine.

Our study revealed that NFS was a significant predictor of hematological toxicity in chemotherapy including CPT-11 for CRC. However, while some studies have pointed out the significance of chemotherapy-related adverse events, such as hematological toxicity, for the prognosis of colorectal cancer (23,24), our study observed no correlation between NFS and prognosis (data not shown). This may explain why the impact of NFS was limited and why CPT-11 chemotherapy alone does not determine the prognosis of colorectal cancer patients, nor do pretreatment serum bilirubin levels or UGT1A1 polymorphisms (16,25). Further investigation is required.

This study has several limitations. Although the majority of studies on CPT-11 have included information on UGT1A1 polymorphisms, this was lacking in our medical records, as it is not routinely investigated at our institution. Another limitation was weakness of the statistical power of our study to a certain extent, as this was a retrospective study with a small sample size and data from a single institution. Thus, a prospective study with information on UGT1A1 polymorphisms should be conducted.

In conclusion, NFS is a significant predictor of hematological adverse events in chemotherapy including CPT-11 for colorectal cancer and its use may represent a non-invasive method useful for determining regimens or doses of chemotherapy including CPT-11.

 References Al KnawyBShiffmanMPercutaneous liver biopsy in clinical practiceLiver Int2711661173200710.1111/j.1478-3231.2007.01592.x17919227 RatziuVCharlotteFHeurtierAGombertSGiralPBruckertEGrimaldiACapronFPoynardTLIDO Study GroupSampling variability of liver biopsy in nonalcoholic fatty liver diseaseGastroenterology12818981906200510.1053/j.gastro.2005.03.08415940625 McPhersonSStewartSFHendersonEBurtADDayCPSimple non-invasive fibrosis scoring systems can reliably exclude advanced fibrosis in patients with non-alcoholic fatty liver diseaseGut5912651269201010.1136/gut.2010.21607720801772 AnguloPBugianesiEBjornssonESCharatcharoenwitthayaPMillsPRBarreraFHaflidadottirSDayCPGeorgeJSimple noninvasive systems predict long-term outcomes of patients with nonalcoholic fatty liver diseaseGastroenterology145782789.e4201310.1053/j.gastro.2013.06.05723860502 YonedaMImajoKEguchiYFujiiHSumidaYHyogoHOnoMSuzukiYKawaguchiTAokiNNoninvasive scoring systems in patients with nonalcoholic fatty liver disease with normal alanine aminotransferase levelsJ Gastroenterol4810511060201310.1007/s00535-012-0704-y23184095 MinamiHSaiKSaekiMSaitoYOzawaSSuzukiKKaniwaNSawadaJHamaguchiTYamamotoNIrinotecan pharmacokinetics/pharmacodynamics and UGT1A genetic polymorphisms in Japanese: Roles of UGT1A1*6 and *28Pharmacogenet Genomics17497504200710.1097/FPC.0b013e328014341f17558305 Cancer Therapy Evaluation Program (CTEP)N.C.I. Adverse Events (CTCAE) Version 4.02009 June 14, 2010 [cited 2009 May 28]http://ctep.cancer.gov/protocolDevelopment/electronic_applications/ctc.htm#ctc_40 KleinCEGuptaEReidJMAthertonPJSloanJAPitotHCRatainMJKastrissiosHPopulation pharmacokinetic model for irinotecan and two of its metabolites, SN-38 and SN-38 glucuronideClin Pharmacol Ther72638647200210.1067/mcp.2002.12950212496745 IyerLKingCDWhitingtonPFGreenMDRoySKTephlyTRCoffmanBLRatainMJGenetic predisposition to the metabolism of irinotecan (CPT-11). Role of uridine diphosphate glucuronosyltransferase isoform 1A1 in the glucuronidation of its active metabolite (SN-38) in human liver microsomesJ Clin Invest101847854199810.1172/JCI9159466980 HiroseKKozuCYamashitaKMaruoEKitamuraMHasegawaJOmodaKMurakamiTMaedaYCorrelation between plasma concentration ratios of SN-38 glucuronide and SN-38 and neutropenia induction in patients with colorectal cancer and wild-type UGT1A1 geneOncol Lett3694698201222740978 InnocentiFUndeviaSDIyerLChenPXDasSKocherginskyMKarrisonTJanischLRamírezJRudinCMGenetic variants in the UDP-glucuronosyltransferase 1A1 gene predict the risk of severe neutropenia of irinotecanJ Clin Oncol2213821388200410.1200/JCO.2004.07.17315007088 OnoueMTeradaTKobayashiMKatsuraTMatsumotoSYanagiharaKNishimuraTKanaiMTeramukaiSShimizuAUGT1A1*6 polymorphism is most predictive of severe neutropenia induced by irinotecan in Japanese cancer patientsInt J Clin Oncol14136142200910.1007/s10147-008-0821-z19390945 RamchandaniRPWangYBoothBPIbrahimAJohnsonJRRahmanAMehtaMInnocentiFRatainMJGobburuJVThe role of SN-38 exposure, UGT1A1*28 polymorphism, and baseline bilirubin level in predicting severe irinotecan toxicityJ Clin Pharmacol477886200710.1177/009127000629506017192505 SaiKSaekiMSaitoYOzawaSKatoriNJinnoHHasegawaRKaniwaNSawadaJKomamuraKUGT1A1 haplotypes associated with reduced glucuronidation and increased serum bilirubin in irinotecan-administered Japanese patients with cancerClin Pharmacol Ther75501515200410.1016/j.clpt.2004.01.01015179405 BosmaPJSaekiMSaitoYOzawaSKatoriNJinnoHHasegawaRKaniwaNSawadaJKomamuraKBilirubin UDP-glucuronosyltransferase 1 is the only relevant bilirubin glucuronidating isoform in manJ Biol Chem269179601796419948027054 MeyerhardtJAKwokARatainMJMcGovrenJPFuchsCSRelationship of baseline serum bilirubin to efficacy and toxicity of single-agent irinotecan in patients with metastatic colorectal cancerJ Clin Oncol2214391446200410.1200/JCO.2004.10.04315084617 RaymondEBoigeVFaivreSSanderinkGJRixeOVernilletLJacquesCGatineauMDucreuxMArmandJPDosage adjustment and pharmacokinetic profile of irinotecan in cancer patients with hepatic dysfunctionJ Clin Oncol2043034312200210.1200/JCO.2002.03.12312409328 HaoHZhangLJiangSSunSGongPXieYZhouXWangGThioacetamide intoxication triggers transcriptional up-regulation but enzyme inactivation of UDP-glucuronosyltransferasesDrug Metab Dispos3918151822201110.1124/dmd.111.03917221733883 XieYWangGWangHYaoXJiangSKangAZhouFXieTHaoHCytochrome P450 dysregulations in thioacetamide-induced liver cirrhosis in rats and the counteracting effects of hepatoprotective agentsDrug Metab Dispos40796802201210.1124/dmd.111.04353922266841 IkunoNSodaHWatanabeMOkaMIrinotecan (CPT-11) and characteristic mucosal changes in the mouse ileum and cecumJ Natl Cancer Inst8718761883199510.1093/jnci/87.24.18767494232 TakasunaKHagiwaraTHirohashiMKatoMNomuraMNagaiEYokoiTKamatakiTInvolvement of beta-glucuronidase in intestinal microflora in the intestinal toxicity of the antitumor camptothecin derivative irinotecan hydrochloride (CPT-11) in ratsCancer Res563752375719968706020 LiMWangZGuoJLiuJLiCLiuLShiHLiuLLiHXieCClinical significance of UGT1A1 gene polymorphisms on irinotecan-based regimens as the treatment in metastatic colorectal cancerOnco Targets Ther716531661201425285015 RambachLBertautAVincentJLorgisVLadoireSGhiringhelliFPrognostic value of chemotherapy-induced hematological toxicity in metastatic colorectal cancer patientsWorld J Gastroenterol2015651573201410.3748/wjg.v20.i6.156524587632 ShitaraKMatsuoKTakahariDYokotaTInabaYYamauraHSatoYNajimaMUraTMuroKNeutropaenia as a prognostic factor in metastatic colorectal cancer patients undergoing chemotherapy with first-line FOLFOXEur J Cancer4517571763200910.1016/j.ejca.2009.01.01919217278 LiuCYChenPMChiouTJLiuJHLinJKLinTCChenWSJiangJKWangHSWangWSUGT1A1*28 polymorphism predicts irinotecan-induced severe toxicities without affecting treatment outcome and survival in patients with metastatic colorectal carcinomaCancer11219321940200810.1002/cncr.2337018300238

Receiver operating characteristic curve (ROC) analysis based on the association of pretreatment non-alcoholic fatty liver disease fibrosis score with hematological toxicity (Common Terminology Criteria for Adverse Effects grade ≥3). Arrow, optimum cut-off point. In this model, the sensitivity was 83.3% and the specificity was 79.5%.

Patient characteristics.

Variables All (n=87)
Gender (male/female) 59/28
Age, years 67 (36–82)a
Tumor location (C/A/T/D/S/R) 5/15/9/6/24/28
Histological type (well/poorly differentiated) 77/10
Single/multiple metastasis 49/38
Synchronous/metachronous metastasis 37/50
Number of cycles 13 (1–126)a
Use of bevacizumab (yes/no) 45/42
Total serum bilirubinb, mg/dl 0.6 (0.3–1.5)a
NFSb −1.302 (−5.158 to 2.62)a
Hematological toxicityc 6
  Decreased platelet count 1
  Decreased neutrophil count 5
Non-hematological toxicityc 13
  Vomiting 5
  Diarrhea 4
  Nervous system disorders 1
  Increased serum bilirubin 1
  Cough 2

Median (range).

Pretreatment.

Common Terminology Criteria for Adverse Effects grade ≥3. C, cecum; A, ascending colon; T, transverse colon; D, descending colon; S, sigmoid colon; R, rectum; NFS, non-alcoholic fatty liver disease fibrosis score.

Uni- and multivariate analysis for hematological toxicity.

Variables Unadjusted coef (95% CI) P-value Adjusted coef (95% CI) P-value
Male gender −1.558 (−3.321 to 0.205) 0.083 −1.247 (−3.480 to 0.985) 0.274
Age 0.173 (−0.068 to 0.103) 0.691 0.008 (−0.131 to 0.147) 0.907
Rectum 0.056 (−1.704 to 1.816) 0.950
Multiple metastasis −0.470 (−2.223 to 1.283) 0.599
Synchronous metastasis −1.386 (−3.578 to 0.805) 0.215 −0.664 (−3.504 to 2.175) 0.647
Number of cycles 0.016 (−0.016 to 0.049) 0.320 0.022 (−0.015 to 0.058) 0.243
Use of bevacizumab −0.074 (−1.733 to 1.584) 0.930
Total serum bilirubina 1.248 (−1.619 to 4.116) 0.394 −0.434 (−4.516 to 3.648) 0.835
NFSa 0.957 (0.257 to 1.658) 0.007 0.932 (0.083 to 1.781) 0.031

Pretreatment. Coef, coefficient; CI, confidence interval; NFS, non-alcoholic fatty liver disease fibrosis score.

Patient characteristics by NFS.

Variables High NFS (n=21) Low NFS (n=66) P-value
Gender (male/female) 13/8 46/20 0.506b
Age, years 73 (47–82)a 65 (36–80)a 0.007c
Tumor location (colon/rectum) 16/5 43/23 0.346b
Histological type (well/poorly differentiated) 20/1 57/9 0.267b
Single/multiple metastasis 14/7 35/31 0.272b
Synchronous/metachronous metastasis 7/14 30/36 0.328b
Number of cycles 17 (3–87)a 12 (1–126)a 0.042c
Use of bevacizumab (yes/no) 12/9 33/33 0.568b
Total serum bilirubine, mg/dl 0.7 (0.5–1.3)a 0.6 (0.3–1.5)a 0.163c
NFSe 0.823 −1.692 <0.001c
(−0.268 to 2.62)a (−5.158 to −0.347)a
Hematological toxicityf 4 2 0.028d
  Decreased platelet count 1 0
  Decreased neutrophil count 3 2
Non-hematological toxicityf 4 9 0.546d
  Vomiting 1 4
  Diarrhea 1 3
  Nervous system disorders 0 1
  Increased serum bilirubin 1 0
  Cough 1 1

Median (range).

χ2 test.

Mann-Whitney U test.

Logistic regression analysis.

Pretreatment.

Common Terminology Criteria for Adverse Effects grade ≥3. NFS, non-alcoholic fatty liver disease fibrosis score.

Uni- and multivariate analysis for non-hematological toxicity.

Variables Unadjusted coef (95% CI) P-value Adjusted coef (95% CI) P-value
Male gender −0.326 (−1.547 to 0.895) 0.600 −0.745 (−2.109 to 0.619) 0.284
Age −0.007 (−0.064 to 0.049) 0.799 −0.044 (−0.116 to 0.027) 0.227
Rectum 0.326 (−0.895 to 1.547) 0.600
Multiple metastasis 0.481 (−0.703 to 1.666) 0.426 0.823 (−0.588 to 2.234) 0.253
Synchronous metastasis 0.537 (−0.648 to 1.722) 0.374 1.146 (−0.426 to 2.719) 0.153
Number of cycles −0.094 (−0.186 to −0.003) 0.043 −0.127 (−0.238 to −0.017) 0.024
Use of bevacizumab −0.262 (−1.444 to 0.920) 0.664
Total serum bilirubina 0.508 (−1.538 to 2.554) 0.627 1.377 (−1.298 to 4.053) 0.313
NFSa 0.053 (−0.313 to 0.419) 0.776 0.405 (−0.101 to 0.910) 0.117

Pretreatment. Coef, coefficient; CI, confidence interval; NFS, non-alcoholic fatty liver disease fibrosis score.