Equivalent survival in patients with and without steatosis undergoing resection for colorectal liver metastases following pre-operative chemotherapy

Available online at www.sciencedirect.com

ScienceDirect EJSO 40 (2014) 1436e1444

www.ejso.com

Equivalent survival in patients with and without steatosis undergoing resection for colorectal liver metastases following pre-operative chemotherapy* E. Parkin a,b,*, D.A. O’Reilly a,b, R. Adam c, G.M. Kaiser d, C. Laurent e, D. Elias f, L. Capussotti g, A.G. Renehan a, the LiverMetSurvey Centres a Institute of Cancer Sciences, University of Manchester, Manchester, UK Department of Hepatobiliary Surgery, North Manchester General Hospital, Manchester, UK c Centre Hepato-Biliaire, AP-HP H^opital Paul Brousse, Universite Paris-Sud, Villejuif, France d Department of General, Visceral and Transplantation Surgery, Essen University Hospital, Essen, Germany e Service de Chirurgie Digestive, H^opital Saint Andre, Bordeaux, France f Institut Gustave Roussy, Cancer Centre, Villejuif, France g Department of HPB and Digestive Surgery, Ospedale Mauriziano Umberto I, Torino, Italy b

Accepted 24 July 2014 Available online 21 August 2014

Abstract Background: We previously reported that the presence of steatosis did not adversely influence survival in patients undergoing resection for colorectal liver metastases (CLM) without pre-operative chemotherapy. Here, this hypothesis is tested in patients undergoing resection for CLM following pre-operative chemotherapy. Methods: We assessed the effects of background liver pathology, categorized as ‘normal’, ‘steatosis’ and ‘other’, on perioperative mortality, overall survival (OS) and cancer-specific survival (CSS) in LiverMetSurvey patients. Survival analyses included log-rank tests and multivariate Cox models, incorporating well-established prognosticators. In secondary analyses, re-populating the model with non-chemotherapy patients, the effect modification of chemotherapy on the impact of steatosis on survival was tested. Results: Of 4329 patients undergoing first-time liver resection following pre-operative chemotherapy, histologies were normal in 1913 (44%), steatosis in 1675 (39%), and other abnormal pathologies in 741 (17%). For normal, steatosis and other, 90-day mortalities were 2.1%, 2.3%, and 3.5% (P ¼ 0.103). For the three histo-pathological groups, 5-year OS rates were 39%, 42%, and 36% (Plogrank ¼ 0.363); 5-year CSS rates were 43%, 45% and 41% (Plogrank ¼ 0.496), respectively. The associations of steatosis with OS and CSS were materially unchanged in the multivariate models. Chemotherapy did not interact with the effect of steatosis on survival. Conclusion: The findings of equivalent survivals challenge the common perception that steatosis in CLM patients after pre-operative chemotherapy is associated with increased peri-operative mortality and poorer long-term survival. Ó 2014 Elsevier Ltd. All rights reserved.

Keywords: Steatosis; Liver resection; Colorectal cancer; Chemotherapy; Survival

Introduction

*

Previous presentation: Presented to the Association of Surgeons of Great Britain and Ireland, Glasgow, May 2013. * Corresponding author. Institute of Cancer Sciences, University of Manchester, Manchester Academic Health Science Centre, The Christie NHS Foundation Trust, Wilmslow Road, Manchester M20 4BX, United Kingdom. Tel.: þ44 161 446 3157. E-mail address: [email protected] (E. Parkin). http://dx.doi.org/10.1016/j.ejso.2014.07.040 0748-7983/Ó 2014 Elsevier Ltd. All rights reserved.

Hepatic steatosis is the pathological accumulation of intra-hepatic triglycerides,1 and is estimated to occur in one third of the Western adult population.2,3 The diagnosis is conventionally made, at the time of histological examination, when 5 per cent of liver tissue contains fat.4 The presence of steatosis mirrors the presence of excess body adiposity, and in turn, the latter is a risk factor for incident

E. Parkin et al. / EJSO 40 (2014) 1436e1444

colorectal cancer.5,6 Therefore, it is reasonable to speculate that steatosis might be an adverse factor for outcome after resection of colorectal liver metastases (CLM). In patients undergoing resection of CLM, background liver steatosis may additionally be caused by preoperative chemotherapy (notably but not exclusively irinotecancontaining regimens),7,8 in which the histological appearance is indistinguishable from that observed in a state of excess body adiposity. Steatosis is a relatively common finding in liver resection specimens, with rates varying from 18 to 56 per cent, depending on the diagnostic criteria used.9e11 Several reports have described an association between steatosis and increased risk for perioperative morbidity, particularly in patients undergoing a major liver resection9,12e14 (removal of three or more segments according to the Brisbane 2000 Terminology of Liver Anatomy and Resections15). A metaanalysis16 concluded that moderate/severe steatosis (30 per cent) was associated with an increased risk for perioperative mortality, but apart from the study by Behrns et al.12 (published at a time when it was uncommon to administer pre-operative chemotherapy), the other studies in that analysis lumped together patients with steatosis regardless of whether they had or had not received preoperative chemotherapy.16 For long-term survival, to-date four studies have evaluated the impact of steatosis or steatohepatitis e one included cancer types other than CLM13; two were small CLM cohorts, and compared with patients without steatosis, reporting non-significant reductions in median overall survival (OS) and recurrence-free survival in patients with steatosis11 and severe grade steatohepatitis,17 respectively; and one large study, using a propensity score matching process to reduce confounding, reported significant reductions in local recurrence-free survival, but no change in OS, associated with the presence of steatosis.18 If the confounding effects of preoperative chemotherapy are not taken into account, it is unclear whether obesity-related steatosis per se or the factors that select for preoperative chemotherapy impact adversely on early and long-term outcomes. We previously reported, from LiverMetSurvey, that the presence of steatosis did not adversely influence survival in patients undergoing resection for CLM without preoperative chemotherapy. Here, this hypothesis is tested in patients undergoing resection for CLM following preoperative chemotherapy. Methods Database LiverMetSurvey (http://www.livermetsurvey.org) is a prospective international database of patients undergoing surgery for CLM that comprises data voluntarily registered by 483 centres across 69 countries.19e24 Details about the primary tumour, number, size, and location of liver metastases, chemotherapy history, hepatic resection, postoperative complications and survival are entered using a standardized online

1437

questionnaire. Data quality control is assessed by online queries and a bi-annual review by the survey data manager. Histological categorization The local clinical team uses the online questionnaire to document whether background non-tumour liver histology was normal or abnormal. If abnormal, a tick-box list is completed to indicate whether this reflected ‘congestion’, ‘fibrosis’, ‘nodular regenerative hyperplasia’, ‘steatosis’ or ‘other’. More than one box may be ticked and free text used. For the purpose of the primary analysis and to facilitate comparison with results for non-chemotherapy patients in the same LiverMetSurvey cohort,22 background (nontumour parenchymal) liver histology was categorized into three: “normal”; “steatosis”; and “other” (other hepatic pathologies apart from steatosis e.g. ‘congestion’, ‘fibrosis’, ‘nodular regenerative hyperplasia’). Combinations of more than one category were re-allocated such that steatosis entries were given hierarchical preference to the steatosis category. For example, ‘steatosis & fibrosis’ or free-text documentation of ‘steatohepatitis’ were placed in the “steatosis” category; free-text phrases such as ‘sinusoidal obstruction syndrome’ and ‘vascular congestion’ were recoded as congestion and placed in the “other” category. As a secondary analysis, background liver histology was reclassified (denoted by “R-” prefix) to explore other histological abnormalities of clinical relevance in CLM patients treated with pre-operative chemotherapy e namely pure steatosis (i.e. not in combination with any other histological abnormalities); steatosis with fibrosis (used here as a surrogate marker of steatohepatitis); and sinusoidal congestion (Fig. S1, supplemental material). Inclusion criteria We included patients undergoing first-time resection between 1990 and 2011 with information on background liver histology, treated with pre-operative chemotherapy. Patients with the following were excluded: (i) non-surgical interventions; (ii) repeat resections; and (iii) absent reporting of background liver histology. For the main analysis, all patients recorded in the registry as ‘Chemotherapy pre-operative e yes’ were included. Sensitivity analyses were subsequently performed to explore the effect of steatosis on survival stratified by the timing of pre-operative chemotherapy and the number of recorded cycles. Duration of pre-operative and number of cycles were dichotomized at cut-off points of six months and three cycles, respectively, based on definitions used in the New EPOC trial25 and previous LiverMetSurvey studies.19,23 Follow-up and outcome measures Data recorded on the ‘latest news’ survey page were extracted to ascertain follow-up and vital status; recorded as alive, dead, and whether the death was cancer-related.

1438

E. Parkin et al. / EJSO 40 (2014) 1436e1444

The primary outcome measures for this analysis were: 90day peri-operative mortality (a preferred time frame, compared with traditional 30-day mortality, in patients after hepatic surgery25); OS and cancer-specific survival (CSS). Statistical analysis Differences in baseline characteristics across the background liver histology categories were explored using standard approaches for continuous (KruskaleWallis test) and categorical (chi-squared test for multiple categories) variables. Peri-operative mortality was expressed as a crude rate, and for comparative tests, we derived odds ratios (OR) and 95 per cent confidence intervals (CI) using logistic regression models adjusted for age, gender and year of surgery. These models were extended to test for interactions between steatosis and extent of resection as multiplicative products. KaplaneMeier curves were constructed for all time-toevent outcome measures and compared across groups using the log-rank test. Time zero was date of liver resection. OS was time from first resection to death from any cause; and CSS as time from first resection to death from colorectal cancer.26 Cox models were generated to assess whether steatosis was an independent prognostic factor on OS and CSS, as we have previously described.22,27 To test for confounding, covariates shown to independently prognosticate for survival after resection for CLM according to other models28 were included: node positive primary; number of hepatic metastases >3; carcinoembryonic (CEA) antigen level >60 ng/mL; tumour diameter 5 cm; positive resection margin; and extra-hepatic disease. This prognostic modelling system was chosen because (i) six of its seven factors are available in LiverMetSurvey (primary tumour differentiation is unavailable) and (ii) it was developed during a period contemporaneous with that for LiverMetSurvey. The robustness of these findings were further tested by patient age, year of resection and random sub-sampling without replacement. To test the effect modification of chemotherapy on the impact of steatosis on survival, the dataset was re-populated with patients undergoing liver resection for CLM and without preoperative chemotherapy i.e. the first study dataset.22 A multiplicative interaction term e [steatosis* treatment] e was generated and included in the Cox model. To take account of multiple testing, a P value of <0.01 was taken as statistically significant. All statistical analyses were performed using STATA version 11.1 (College Station, Tx, USA). Results

operative chemotherapy from 1990 to 2011 and with known background liver histology were identified. The proportion of missing data on histology was highest in the early period of study (Table S1 and S2), but for statistical analyses, these missing data are considered missing at random (MAR).29 A flow diagram is shown in Fig. 1. Baseline characteristics Table 1 shows the baseline characteristics of the cohort for analysis, categorized according to background (nontumour parenchymal) liver histology: ‘normal’, ‘steatosis’ and ‘other pathologies’. The median age was 62.1 (interquartile range: 54e68) year; the male: female ratio was approximately 3: 2.

LiverMetSurvey Until December 2011 N =16779

Incomplete pre-operative chemotherapy variable, n = 32 Non-surgical, n = 1092 Repeat hepatectomy, n = 1736 Background histology unknown, n = 3503 Resection pre-1990, n = 234

First-time hepatic resection with known background histology 1990 to 2011 N =10182 Without pre-operative chemotherapy, n = 5853

With pre-operative chemotherapy N = 4329

Background liver histology Normal: n = 1913 (44.2%) Steatosis: n = 1675 (38.7%) Other pathologies: n = 741 (17.1%)

Overview The LiverMetSurvey database included 16,779 procedure cases to December 2011. From these, 4329 patients who underwent first-time liver resections with pre-

Tests for interaction model N = 10182 Figure 1. Flow diagram deriving number of patients in main analysis.

E. Parkin et al. / EJSO 40 (2014) 1436e1444

1439

Table 1 Baseline characteristics of patients undergoing first-time hepatic resection for colorectal cancer metastases with pre-operative chemotherapy, LiverMetSurvey 1990e2011. Total n ¼ 4329 Median age (IQR) Gender Year band

Type of resection

Prognostic factors Nodal status of primary tumour

Median CEA at liver resection (IQR) Number of hepatic metastases

Diameter of largest hepatic metastasis

Minimum margin of hepatic resection

EHD at hepatectomy

Chemotherapy agentsc 5-FU

Oxaliplatin

Irinotecan

Capecitabine Bevacizumab Cetuximab Other agents Number of chemotherapy cycles

Chemotherapy end to hepatectomy interval

Background histology

P

Normal n ¼ 1913

Steatosis n ¼ 1675

Other n ¼ 741

All Male Female 1990e2000 2001e2005 2006e2011 Minor (<3) Major (3) Unknown

62.1 (55e69) 2686 1642 419 747 3163 1532 2748 49

61.4 1155 757 209 324 1380 743 1144

62.7 1069 606 143 283 1249 552 1108

61.7 462 279 67 140 534 237 496

N0 N1/N2 unknown All 3 >3 unknown <5 cm 5 cm Unknown <1 mm 1 mm unknown yes no Unknown

1188 2357 784 6.4 (2.7e23.6) 2940 1265 124 3089 992 248 861 2955 513 565 1085 2679

514 (33.6) 1016 (66.4)

Yes No Unknown Yes No Unknown Yes No Unknown Yes No Yes No Yes No Yes No 1e3 4-8 >8 Unknown 0e6 months >6 months Unknown

3291 950 88 2416 1826 87 1283 3003 43 612 3717 1034 3295 331 3998 240 4089 549 2095 959 726 2747 653 929

(53e69) (60.4) (39.6) (10.9) (16.9) (72.1) (39.4) (60.6)

(56e69) (63.8) (36.2) (8.5) (16.9) (74.6) (33.3) (66.7)

475 (33.8) 932 (66.2)

(54e68) (62.3) (37.7) (9.0) (18.9) (72.1) (32.3) (67.7)

0.001a 0.108b 0.096b

<0.001b

199 (32.7) 409 (67.3)

0.041b

6.0 (2.6e23.0) 1343 (72.4) 513 (27.6)

6.8 (2.7e23.3) 1124 (69.3) 499 (30.7)

7.28(2.8e26.0) 473 (65.2) 253 (34.8)

0.300a 0.003b

1359 (76.1) 427 (23.9)

1221 (85.4) 370 (23.3)

509 (72.3) 195 (27.2)

0.029b

336 (20.2) 1326 (79.8)

346 (23.0) 1157 (77.0)

179 (27.5) 472 (72.5)

<0.001b

241 (36.5) 419 (63.5)

203 (30.3) 468 (69.7)

121 (37.9) 198 (62.1)

<0.001b

1419 (76.0) 447 (24.0)

1287 (78.1) 361 (21.9)

585 (80.5) 142 (19.5)

0.049b

1043 (55.6) 833 (44.4)

876 (53.4) 763 (46.6)

497 (68.4) 230 (31.6)

<0.001b

554 (29.2) 1344 (70.8)

529 (31.9) 1128 (68.1)

200 (27.4) 531 (72.6)

0.097b

286 1627 448 1465 135 1778 108 1805 262 912 374

218 1457 437 1238 126 1549 96 1579 197 818 403

108 633 149 592 70 671 36 705 90 365 182

0.235

(15.0) (85.0) (23.4) (76.6) (7.1) (92.9) (5.6) (94.4) (16.9) (58.9) (24.2)

1195 (81.3) 275 (18.7)

(13.0) (87.0) (26.1) (73.9) (7.5) (92.5) (5.7) (94.3) (13.9) (57.7) (28.4)

1081 (80.4) 264 (19.6)

Unless otherwise stated, values in parentheses are percentages. IQR: interquartile range. CEA: Carcinoembryonic antigen. EHD: Extra-hepatic disease. 5-FU: 5 fluorouracil. Missing data: age in 11 patients; gender data in one patient; CEA at time of hepatectomy in 1192 patients. a KruskaleWallis test. b Chi-squared test for multiple categories, excluding unknown categories. c Chemotherapy agents as single agent and in combination.

(14.6) (85.4) (20.1) (79.9) (9.4) (90.6) (4.9) (95.1) (14.1) (57.3) (28.6)

471 (80.5) 114 (19.5)

0.005b 0.112b 0.665b <0.001b

0.145b

1440

E. Parkin et al. / EJSO 40 (2014) 1436e1444

The proportion of patients with steatosis following preoperative chemotherapy was 38.7 per cent. This was significantly higher compared with 30.6 per cent previously reported by the authors in chemotherapy-na€ıve patients undergoing resection for CLM (P < 0.001).22 The proportion of patients with steatosis after pre-operative chemotherapy was uninfluenced by gender, but the median age was greater in patients with steatosis compared with those with normal histology (median age: 62.7 versus 61.4 years, P ¼ 0.001). The proportion of patients undergoing resection with steatosis was consistent across period bands (P ¼ 0.096). Almost twice as many patients underwent major compared with minor resections e patients undergoing major resection had a higher proportion of steatosis and other non-tumour parenchymal pathologies compared with normal histology (P < 0.001). By established prognosticators,28 there were significant differences across the three histology categories by: number of hepatic metastases (P ¼ 0.003); minimum resection margin (P < 0.001); and extra-hepatic disease at the time of resection (P < 0.001). Additionally, there were significant differences across the histological categories according to preoperative chemotherapy regimens: oxaliplatin (mainly increased other pathologies, which included sinusoidal congestion, P < 0.001); bevacizumab (mainly reduced other pathologies, P ¼ 0.005); and >8 versus fewer chemotherapy cycles (increased steatosis and other pathologies, P ¼ 0.001). Peri-operative mortality A total of 105 deaths occurred within 90 days of first resection, giving a crude mortality rate of 2.43 per cent (Table 2). Rates of 90-day peri-operative mortality in patients with normal background histology, steatosis, and

other pathologies were 2.10 per cent, 2.33 per cent and 3.51 per cent, respectively (P ¼ 0.103). Peri-operative mortality rates were higher in those undergoing major versus minor resections. Compared with normal histology, there was no difference in peri-operative mortality for patients with steatosis, but the former was increased in patients with other pathologies (OR ¼ 1.682, 95 per cent CI: 1.017e2.779; P ¼ 0.043). Stratification by extent of surgery did not materially influence associations between steatosis and perioperative mortality (test for interaction: P ¼ 0.148). Long-term survival Median follow-up in the entire cohort was 17.3 months (8569 person-years). The 5-year OS rate was 39.4 per cent and 5-year CSS was 43.5 per cent. By background liver histology, 5-year OS rates in the normal histology, steatosis and other pathology groups were 38.6 per cent, 41.9 per cent and 36.0 per cent, respectively (Fig. 2a). The apparent improvement in OS in patients with steatosis compared with normal histology and other pathologies was nonsignificant (log rank test, P ¼ 0.363). The 5-year CSS rates in patients with normal histology, steatosis and other pathologies were 42.9 per cent, 45.4 per cent and 41.2 per cent, respectively (Fig. 2b). Again, these differences were non-significant (log rank test, P ¼ 0.496). Cox models were used to test whether the apparent improved CSS observed for steatosis was statistical significant after adjustment for potential confounding factors (Table 3). In the minimally adjusted model (adjusted for aged, gender and year of surgery), the apparent CSS advantage associated with steatosis was non-significant (HR ¼ 0.957, 95 per cent CI: 0.844e1.086). In the maximally adjusted model, extended to incorporate known prognostic factors28; associations for steatosis were materially unchanged.

Table 2 Crude rates and logistic regression models for 90-day peri-operative mortality by background liver histology in patients undergoing first-time hepatic resection for colorectal cancer metastases with pre-operative chemotherapy, according to extent of resection, LiverMetSurvey 1990e2011. Total

Background histology Normal

Numbers (crude rates) All resections 105/4329 (2.43) Minor resection 30/1532 (1.96) Major resection 74/2748 (2.69) Logistic regression (odds ratio, 95% CI)b All resections Minor resection Major resection

40 (2.10) 17 (2.30) 23 (2.02) 1.000 (referent) 1.000 (referent) 1.000 (referent)

P Steatosis 39 (2.33) 7 (1.27) 31 (2.80) 1.070 (0.684, 1.674) 0.523 (0.413, 1.272) 1.349 (0.780, 2.332)

Other pathologies 26 (3.51) 6 (2.53) 20 (4.03) 1.682 (1.017, 2.779) 1.062 (0.412, 2.731) 2.047 (1.111, 3.770) Test for interactionc

0.103a 0.332a 0.068a Other v referent 0.043 0.901 0.022 0.148

For crude mortality rates, values in parentheses are percentages. CI: confidence intervals. Missing data: number of segments resected in 49 patients. a Chi-squared test for multiple categories. b All logistic regression models adjusted for age, gender, and year of surgery. c For the test for interaction, other pathologies were excluded i.e. the question was an interaction between steatosis and extent of resection.

E. Parkin et al. / EJSO 40 (2014) 1436e1444

a)

Overall survival

Survival(%)

1.00 Log rank test, P = 0.363

0.75

with a better CSS compared with those with normal histology but there was no interaction between the two terms, i.e. steatosis together with pre-operative chemotherapy did not equal a worse prognosis (Pinteraction ¼ 0.143). Reclassified histology sub-categories

0.50 Survival (%)

0.25 Normal Steatosis Other pathologies

0.00 0

12

b)

3 years

5 years

56.7 58.4 55.2

38.7 41.9 36.0

24

36

48

60

Cancer-specific survival 1.00

Survival(%)

1441

Log rank test, P = 0.496

0.75

0.50

Reclassification of background liver histology resulted in 1097 cases of R-steatosis, 246 of R-steatosis & fibrosis and 243 of R-congestion (Fig. S1, supplemental material). Compared with patients with a normal background histology, rates of peri-operative mortality were unaffected by histology reclassification (Table S5). For long-term survival, taking into account established prognostic factors,28 R-steatosis was associated with borderline improved CSS (HR ¼ 0.863, 95 per cent CI 0.746, 0.998); but no significant effects were observed for R-steatosis and fibrosis or Rcongestion (Table S6).

Survival (%)

0.25

Normal Steatosis Other pathologies

0.00 0 Number at risk Steatosis 1675 Other pathologies 741 Normal 1913

12

1085 448 1165

3 years

5 years

61.4 62.4 59.0

42.9 45.4 41.2

24 36 Time in months 632 257 653

369 145 397

Sensitivity analyses

48

60

207 79 254

129 44 163

Figure 2. Survival curves for (a) overall survival and (b) cancer-specific survival by background histological types.

Internal validation To test for robustness of the model results, the associations between steatosis and CSS were tested stratified by age groups and year-bands (Table S3); hazard ratios were similar across these strata. Additionally, random subsamplings without replacement into three datasets from the whole data were undertaken, and again, there were no material influences on the hazard ratio estimates for associations of steatosis with CSS. Pre-operative chemotherapy steatosis interaction model To investigate whether the administration of chemotherapy modified the effect of steatosis on survival, data on patients without pre-operative chemotherapy (n ¼ 5853) from the authors’ previous study22 was added back into the current analysis of patients with preoperative chemotherapy patients (n ¼ 4329). A multiplicative interaction term was generated (Table S4). Taken together, patients who received pre-operative chemotherapy had worse CSS compared with those not receiving preoperative chemotherapy (HR ¼ 1.222, 95 per cent CI: 1.096e1.356). For all patients, steatosis was associated

Sensitivity analyses were performed to test whether: (i) the time between chemotherapy end date and resection; and (ii) the number of cycles of chemotherapy, might affect results (p7 supplemental material). No material differences were noted. A pathology reporting bias was tested for over the study period (p7 supplemental material), and no material differences were for pre-operative chemotherapy versus chemotherapy-na€ıve groups. Discussion Main findings Our analysis confirmed that in patients undergoing resection for CLM following pre-operative chemotherapy, background steatosis was significantly more prevalent than in chemotherapy-na€ıve patients. However, contrary to general perception, steatosis was not associated with an increased risk of peri-operative mortality, even after major resection. Long-term survival was equivalent in patients with background steatosis compared with those with a normal non-tumour parenchymal histology, an observation that remained robust after adjusting for known prognostic factors. Secondary findings included that (i) other liver pathology was associated with a non-significant but modestly increased risk of peri-operative mortality; and (ii) that subclassification of the steatosis category found no significant differential effects on short- or long-term outcomes. Comparison with published literature For short-term outcome, the present study found no deleterious effect of steatosis on 90-day peri-operative mortality. This contrasts to a meta-analysis reported by de Meijer et al,16 reporting a 2.5-fold increase in peri-

1442

E. Parkin et al. / EJSO 40 (2014) 1436e1444

Table 3 Cox proportional hazard model testing the impact of steatosis on cancer-specific survival in minimally and maximally adjusted models. Covariates Minimally adjusted modela Background histology Normal (referent) Steatosis Other pathologies Maximally adjusted modelb Background histology Normal (referent) Steatosis Other pathologies Gender (male v female) Age (continuous variable: per year) Other prognosticators Nodal status of primary tumour (N1/N2 v N0) CEA at hepatectomy (>60 ng/ml v  60 ng/ml) No. of hepatic metastases (>3 v  3) Diameter of largest hepatic metastasis (5 cm v < 5 cm) Minimum margin of hepatic resection (1 mm v < 1 mm) EHD (yes v no)

Hazard ratio

95%CI

P

1.000 0.957 1.044

(0.844, 1.086) (0.997, 1.228)

0.496 0.606

1.000 0.904 0.918 1.057 1.003

(0.796, (0.778, (0.939, (0.998,

1.026) 1.083) 1.190) 1.009)

0.119 0.309 0.357 0.214

1.504 1.499 1.393 1.501 1.227 1.403

(1.299, (1.270, (1.227, (1.312, (1.066, (1.181,

1.741) 1.769) 1.580) 1.716) 1.412) 1.667)

<0.001 <0.001 <0.001 <0.001 0.004 <0.001

CI: confidence interval. CEA: carcinoembryonic antigen. EHD: extra-hepatic disease. In sensitivity analyses, we tested dropping each prognostic factor one at a time e this made no material difference to associations for background histology. For minimally adjusted model, Harrell’s C-statistic ¼ 0.5336. For maximally adjusted model, Harrell’s C-statistic ¼ 0.6275. a Minimally adjusted model included age, gender, and year of surgery. b Maximally adjusted model include as above plus all other listed variables, entered simultaneously.

operative mortality associated with steatosis of greater than 30 per cent. Several reasons may explain these differences: (i) the de Meijer16 analysis grouped chemotherapy-treated and non-chemotherapy patients together9,12e14; (ii) definitions of peri-operative mortality in studies included in the meta-analysis varied from within 30 days,12 within 60 days or in-hospital mortality,13 within 90 days,14 and death before discharge9; (iii) LiverMetSurvey does not grade steatosis, whereas de Meijer16 found significant differences in peri-operative mortality but only for steatosis greater than 30 per cent, and (iv) peri-operative mortality after resection for CLM is relatively rare, such that deaths reported in the small-sample studies included by de Meijer et al.16 may have occurred by chance. Notably, two series published subsequent to de Meijer et al16 e one of 2715 patients from Leeds and Basingstoke18 (neither centres in LiverMetSurvey); the other of 506 patients from the SloanKettering Memorial Cancer Center30 e found no differences in peri-operative mortality between patients with and without steatosis. Data on the impact of steatosis and/or steatohepatitis on long-term survival is limited to six studies. Four analyzed data combined for chemotherapy-treated and chemotherapy-na€ıve patients, reporting no significant effect on recurrence-free survival17 or OS.11,13,18 One study limited their analysis to patients following pre-operative chemotherapy, and paradoxically, reported a better OS associated with grade 0e1 steatosis compared with normal

background histology.24 The sixth study, from the present authors,22 limited analyses to patients without preoperative chemotherapy, and showed steatosis was paradoxically associated with marginally improved cancerspecific survival (CSS) compared with that for patients with a normal background histology.

Strengths and limitations There are several strengths to this study. First, the large size of the LiverMetSurvey database allowed this analysis restricted to only those patients treated with pre-operative chemotherapy, undergoing first-time liver resection. Second, the database captures many well-established prognostic factors e these were adjusted for in the survival analyses. Third, the labelling of steatosis diagnoses throughout LiverMetSurvey was demonstrated to be consistent according to patient age, gender, and year of resection, and was proportionate by pre-operative chemotherapy category with time. Fourth, internal validation was performed to compare findings for CSS across age groups, year of surgery and random sub-samples of the main dataset and produced consistent results. Finally, pre-operative chemotherapy details facilitated sensitivity analyses by chemotherapy to surgery interval and number of chemotherapy cycles received. Findings were again broadly consistent.

E. Parkin et al. / EJSO 40 (2014) 1436e1444

1443

This study has potential limitations. First, a large international clinical database is reliant upon the accuracy of data entry by multiple users e pathology reporting is not standardized across LiverMetSurvey centres and different institutions may apply different criteria to diagnose abnormal liver pathology or to define “pre-operative chemotherapy”. It was not possible to assess the impact of centre-level variation. Second, steatosis was captured as either present or absent, rather than grade or percentage. Nonetheless, when hepatic fat is assessed using visual techniques, it may be more appropriate to refer simply to its presence or absence because of the lack of agreement between pathologists on percentage steatosis, grade, micro-versus macroscopic steatosis and features of steatohepatitis.4,31 Third, there may be a pathology report bias over the study period as the literature learnt of associations between chemotherapy and steatosis. This was tested for and no clear evidence of a reporting bias identified. Finally, LiverMetSurvey lacks data on BMI, diabetes, alcohol consumption history and comorbidities. It would have been useful to examine how these correlated with the presence or absence of steatosis and to have investigated whether excess BMI and the presence of diabetes impacted upon survival.

demonstrated a positive association between pre-operative chemotherapy and background liver fat,8,30 others have not.35,36 Chemotherapy end date to surgery interval and dosing regimens may be important. Finally, future studies examining post-resection outcome need to incorporate a more reliable and reproducible method for intra-hepatic fat quantification, such as digital histological quantification or non-invasive imaging methods. Protocols for magnetic resonance imaging modalities such as chemical shift37 and proton spectroscopy38 are currently being developed. Given the future projected increase in the numbers of patients eligible for resection of CLM, and a parallel increase in the prevalence of steatosis, better understanding of this topic will benefit many patients.

Clinical implications

Source of funding

Findings from the present analysis, and a recent Memorial Sloan-Kettering study,30 do not support the view, held in the wider surgical literature,32e34 that fat in the background liver following pre-operative chemotherapy has a negative effect upon short- and long-term post-resection outcome. These findings suggest that a change in current clinical practice is not warranted. Steatosis is more prevalent in patients receiving preoperative chemotherapy (39 per cent versus 31 per cent22). In turn, patients receiving pre-operative chemotherapy are, on average, at higher risk for death compared with patients not receiving pre-operative chemotherapy, as evident in the LiverMetSurvey data (5-year CSS was 43.5 per cent versus 51.8 per cent,22 respectively). The apparent worse outcome in patients with chemotherapy-associated steatosis reported by previous studies probably reflects poor prognosis disease, rather than steatosis per se.

EP is funded by HALT (Help Against Liver Tumours), registered UK charity number 1054556. LiverMetSurvey is supported by an unrestricted grant from Sanofi-Aventis.

Key unanswered questions and directions for future research Three key areas for future research result from this study. First, there could be residual confounding e for example, BMI and alcohol consumption, are not reported in LiverMetSurvey. The present findings should be replicated in other settings e for instance, observational analyses within large trials with standardized treatment protocols. Second, there is a need to study the mechanisms underlying the development of steatosis and steatohepatitis in patients with CLM. Whilst some studies have

Acknowledgements The authors wish to thank: David Delvart, data manager at LiverMetSurvey for his help in transferring the data; Lee Malcomson, former data manager in the department of surgery at Christie NHS Trust for his assistance in the statistical analysis; and all the 415 centres across 67 countries that contribute to LiverMetSurvey.

Conflicts of interest The authors declare no conflict of interest. Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.ejso.2014.07.040.

References 1. Fabbrini E, Sullivan S, Klein S. Obesity and nonalcoholic fatty liver disease: biochemical, metabolic, and clinical implications. Hepatology 2010;51:679–89. 2. Bellentani S, Saccoccio G, Masutti F, et al. Prevalence of and risk factors for hepatic steatosis in Northern Italy. Ann Intern Med 2000;132: 112–7. 3. Szczepaniak LS, Nurenberg P, Leonard D, et al. Magnetic resonance spectroscopy to measure hepatic triglyceride content: prevalence of hepatic steatosis in the general population. Am J Physiol Endocrinol Metab 2005;288:E462–8. 4. Kleiner DE, Brunt EM, Van Natta M, et al. Design and validation of a histological scoring system for nonalcoholic fatty liver disease. Hepatology 2005;41:1313–21. 5. Renehan A, Tyson M, Egger M, Heller RF, Zwahlen M. Body mass index and incidence of cancer: a systematic review and meta-analysis of prospective observational studies. Lancet 2008;371:569–78.

1444

E. Parkin et al. / EJSO 40 (2014) 1436e1444

6. Harriss DJ, Atkinson G, George K, et al. Lifestyle factors and colorectal cancer risk (1): systematic review and meta-analysis of associations with body mass index. Colorectal Dis 2009;11:547–63. 7. Fernandez FG, Ritter J, Goodwin JW, Linehan DC, Hawkins WG, Strasberg SM. Effect of steatohepatitis associated with irinotecan or oxaliplatin pretreatment on resectability of hepatic colorectal metastases. J Am Coll Surg 2005;200:845–53. 8. Vauthey JN, Pawlik TM, Ribero D, et al. Chemotherapy regimen predicts steatohepatitis and an increase in 90-day mortality after surgery for hepatic colorectal metastases. J Clin Oncol 2006;24:2065–72. 9. Gomez D, Malik HZ, Bonney GK, et al. Steatosis predicts postoperative morbidity following hepatic resection for colorectal metastasis. Br J Surg 2007;94:1395–402. 10. Jarnagin WR, Gonen M, Fong Y, et al. Improvement in perioperative outcome after hepatic resection: analysis of 1,803 consecutive cases over the past decade. Ann Surg 2002;236:397–406. discussion -7. 11. Pathak S, Tang JM, Terlizzo M, Poston GJ, Malik HZ. Hepatic steatosis, body mass index and long term outcome in patients undergoing hepatectomy for colorectal liver metastases. Eur J Surg Oncol 2010; 36:52–7. 12. Behrns KE, Tsiotos GG, DeSouza NF, Krishna MK, Ludwig J, Nagorney DM. Hepatic steatosis as a potential risk factor for major hepatic resection. J Gastrointest Surg 1998;2:292–8. 13. Kooby DA, Fong Y, Suriawinata A, et al. Impact of steatosis on perioperative outcome following hepatic resection. J Gastrointest Surg 2003;7:1034–44. 14. McCormack L, Petrowsky H, Jochum W, Furrer K, Clavien PA. Hepatic steatosis is a risk factor for postoperative complications after major hepatectomy: a matched case-control study. Ann Surg 2007;245: 923–30. 15. Terminology Committee of the International Hepato-Pancreato-Biliary Association. Brisbane 2000 terminology of liver anatomy & resections. HPB (Oxford) 2000;2:333–9. 16. de Meijer VE, Kalish BT, Puder M, Ijzermans JN. Systematic review and meta-analysis of steatosis as a risk factor in major hepatic resection. Br J Surg 2010;97:1331–9. 17. Tamandl D, Klinger M, Eipeldauer S, et al. Sinusoidal obstruction syndrome impairs long-term outcome of colorectal liver metastases treated with resection after neoadjuvant chemotherapy. Ann Surg Oncol 2011;18:421–30. 18. Hamady ZZ, Rees M, Welsh FK, et al. Fatty liver disease as a predictor of local recurrence following resection of colorectal liver metastases. Br J Surg 2013;100:820–6. 19. Adam R, Bhangui P, Poston G, et al. Is perioperative chemotherapy useful for solitary, metachronous, colorectal liver metastases? Ann Surg 2010;252:774–87. 20. Adam R, Frilling A, Elias D, et al. Liver resection of colorectal metastases in elderly patients. Br J Surg 2010;97:366–76. 21. Andres A, Toso C, Adam R, et al. A survival analysis of the liver-first reversed management of advanced simultaneous colorectal liver metastases: a LiverMetSurvey-based study. Ann Surg 2012;256:772–8. discussion 8e9. 22. Parkin E, O’Reilly DA, Adam R, et al. The effect of hepatic steatosis on survival following resection of colorectal liver metastases in

23.

24.

25.

26.

27.

28.

29. 30.

31.

32.

33. 34.

35.

36.

37. 38.

patients without preoperative chemotherapy. HPB (Oxford) 2013;15: 463–72. Vigano L, Capussotti L, Barroso E, et al. Progression while receiving preoperative chemotherapy should not be an absolute contraindication to liver resection for colorectal metastases. Ann Surg Oncol 2012;19: 2786–96. Vigano L, Capussotti L, De Rosa G, De Saussure WO, Mentha G, Rubbia-Brandt L. Liver resection for colorectal metastases after chemotherapy: impact of chemotherapy-related liver injuries, pathological tumor response, and micrometastases on long-term survival. Ann Surg 2013;258:731–42. Mullen JT, Ribero D, Reddy SK, et al. Hepatic insufficiency and mortality in 1,059 noncirrhotic patients undergoing major hepatectomy. J Am Coll Surg 2007;204:854–62. discussion 62e4. Punt CJ, Buyse M, Kohne CH, et al. Endpoints in adjuvant treatment trials: a systematic review of the literature in colon cancer and proposed definitions for future trials. J Natl Cancer Inst 2007;99:998–1003. Crosbie EJ, Roberts C, Qian W, Swart AM, Kitchener HC, Renehan AG. Body mass index does not influence post-treatment survival in early stage endometrial cancer: results from the MRC ASTEC trial. Eur J Cancer 2012;48:853–64. Rees M, Tekkis PP, Welsh FK, O’Rourke T, John TG. Evaluation of long-term survival after hepatic resection for metastatic colorectal cancer: a multifactorial model of 929 patients. Ann Surg 2008;247:125–35. Little RJA, Rubin DB. Statistical analysis with missing data. 2nd ed. Hoboken: NJ: Wiley; 2002. Wolf PS, Park JO, Bao F, et al. Preoperative chemotherapy and the risk of hepatotoxicity and morbidity after liver resection for metastatic colorectal Cancer: a single institution experience. J Am Coll Surg 2013;216:41–9. El-Badry AM, Breitenstein S, Jochum W, et al. Assessment of hepatic steatosis by expert pathologists: the end of a gold standard. Ann Surg 2009;250:691–7. Belghiti J. Fatty liver disease as a predictor of local recurrence following resection of colorectal liver metastases (Br J Surg 2013: 100: 820-826). Br J Surg 2013;100:827. Brouquet A, Nordlinger B. Metastatic colorectal cancer outcome and fatty liver disease. Nat Rev Gastroenterol Hepatol 2013;10:266–7. Lehmann K, Rickenbacher A, Weber A, Pestalozzi BC, Clavien PA. Chemotherapy before liver resection of colorectal metastases: friend or foe? Ann Surg 2012;255:237–47. Ryan P, Nanji S, Pollett A, et al. Chemotherapy-induced liver injury in metastatic colorectal cancer: semiquantitative histologic analysis of 334 resected liver specimens shows that vascular injury but not steatohepatitis is associated with preoperative chemotherapy. Am J Surg Pathol 2010;34:784–91. Brouquet A, Benoist S, Julie C, et al. Risk factors for chemotherapyassociated liver injuries: a multivariate analysis of a group of 146 patients with colorectal metastases. Surgery 2009;145:362–71. Raptis DA, Fischer MA, Graf R, et al. MRI: the new reference standard in quantifying hepatic steatosis? Gut 2012;61:117–27. Urdzik J, Bjerner T, Wanders A, et al. The value of pre-operative magnetic resonance spectroscopy in the assessment of steatohepatitis in patients with colorectal liver metastasis. J Hepatol 2012;56:640–6.