- Email: [email protected]
Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection
Available online at www.sciencedirect.com
ScienceDirect EJSO xx (2017) 1e8
www.ejso.com
Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection D. Reim a,b,d, Y.-S. Choi a,d, H.M. Yoon a, B. Park c, B.W. Eom a, M.-C. Kook a, K.W. Ryu a, I.J. Choi a, J. Joo c,**,e, Y.-W. Kim a,*,e a
Gastric Cancer Branch, Research Institute & Hospital, National Cancer Center, Ilsan-ro 323, Ilsandong-gu, Goyang-si Gyeonggi-do, 410-769, Republic of Korea b Klinikum Rechts der Isar der Technischen Universit€at M€unchen, Department of Surgery, Ismaninger Strasse 22, 81675 Munich, Germany c Biometric Research Branch, Clinical Research Coordination Center, National Cancer Center, Ilsan-ro 323, Ilsandong-gu, Goyang-si Gyeonggi-do, 410-769, Republic of Korea Accepted 21 April 2017 Available online - - -
Abstract Background: Prognosis of alpha-fetoprotein positive gastric cancer (AFPP-GC) remains elusive so far due to disparities in cohort size and baseline characteristics in previous studies. A propensity score matching (PSM) analysis as well as multivariable model was performed for unbiased evaluation of the outcome in AFPGC. Methods: Among 3034 gastric cancer patients who underwent curative gastric cancer surgery (R0, M0) at the National Cancer Center, Korea between 2002 and 2007, we identified 97 patients being positive for AFP either by elevation of serum-AFP levels >10 mg/L or by immunohistochemical staining. Due to marked disparities in baseline characteristics and cohort size, propensity-score-matching was performed which matched 87 AFPP-GC patients to the same number of AFP-negative gastric cancer (AFPN-GC) patients. Baseline characteristics were compared using c2-test. Survival curves were compared using the Kaplan-Meier-method and multivariable regression analysis was performed to evaluate the effect of AFP-positivity while adjusting the effects of confounding variables. Results: AFPP-GC and AFPN-GC patients revealed marked disparities in patient cohorts. After PSM, groups were balanced for age, sex, tumor size, BMI, tumor location, grade of differentiation, presence of lymphatic vessel infiltration (LVI), Lauren histologic type and stage distribution. In multivariable regression analysis of the PSM-groups, only AFP-positivity and pathologic stage were predictive for overall survival (HR 2.98, CI 95% {1.7e5.1}, p < 0.0001). Five-year-survival rates were significantly worse for AFPP-GC patients (57.9% vs. 76.1%, p ¼ 0.014). Recurrence was significantly more frequent in AFPP-GC patients (p ¼ 0.003). Conclusion: AFP can be considered as an independent negative predictor of overall and recurrence-free survival in patients with gastric cancer. Ó 2017 Elsevier Ltd, BASO ~ The Association for Cancer Surgery, and the European Society of Surgical Oncology. All rights reserved.
Keywords: Gastric cancer; AFP; Surgery
Introduction Several case series and retrospective analyses on alphafetoprotein (AFP)-positive gastric cancer were published * Corresponding author. Fax: þ82 31 920 0696. ** Corresponding author. E-mail addresses: [email protected] (J. Joo), [email protected] (Y.-W. Kim). d DR and YSC contributed equally to this study as a first author. e JJ and YWK contributed equally to this study as a Corresponding author.
over the recent years.1e4 AFP-positive gastric cancer (AFPP-GC) appears to be a rare entity compared to common gastric cancer. The incidence is reported to be 1e15% among gastric cancer patients. AFPP-GCs combined with liver metastasis were first reported in 1970.5 Further reports concluded that AFPP-GC revealed poor prognosis, high recurrence rates and high potential to metastasize to the liver and lymph nodes.1e5 This appears to be related to a histologically more aggressive tumor phenotype. AFPP-GCs tend to show higher rates of
http://dx.doi.org/10.1016/j.ejso.2017.04.005 0748-7983/Ó 2017 Elsevier Ltd, BASO ~ The Association for Cancer Surgery, and the European Society of Surgical Oncology. All rights reserved. Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
2
D. Reim et al. / EJSO xx (2017) 1e8
lymphatic and blood vessel infiltration,1 which may lead to higher recurrence rates and shorter postoperative survival. Most of the previous reports could not provide data for direct comparisons between AFPP-GC and AFP-negative gastric cancer (AFPN-GC) due to small and heterogeneous cohort sizes and marked differences in baseline characteristics.2 Only one Chinese analysis reported a matched cohort result.3 However, the matching algorithm was not reported which leads to the notion that selection bias may not be completely excluded. Therefore the purpose of this analysis was to examine prognosis of AFPP-GC from a large patient population after propensity score matching (PSM) which attempts to eliminate the disparities in cohort size and baseline characteristics. Patients and methods The screening of the prospectively documented database for gastric cancer at the National Cancer Center Korea (NCCK) revealed 3034 patients who were subjected to primary curative oncologic resection (R0) for gastric cancer between January 2002 and August 2007 without undergoing preoperative chemo- or radiochemotherapy. All patients in whom metastatic disease was ruled out, received oncologic resection according to Japanese guidelines including D2 lymph node dissection.6 All specimens were examined postoperatively by specialized pathologists and staged according to the 6th UICC edition TNM staging system.7 Serum AFP was tested in all patients preoperatively and were considered as AFP-positive when elevated preoperative serum AFP levels (>10 mg/L) were detected. IHC was only done in those patients with preoperatively elevated AFP-levels and hepatoid features in the preoperative biopsy. Finally, 97 patients were identified as AFP-positive. Patients received adjuvant chemotherapy in case of metastatic lymph nodes and/or from pathologic stages II-IV. Patients were followed postoperatively via the outpatient clinic in regular intervals (every 6 months for the first 24 months and then every 12 months until at least 60 months postoperatively for early gastric cancer patients and every 3 months until 3 years postoperatively and then every 6 months until the sixth postoperative year for advanced gastric cancer patients). The follow-up period was calculated from the day of surgery to the last follow-up date. During this five year period 187 patients were diagnosed with gastric cancer but didn’t receive curative resections. Among them, 174 patients were tested for AFP and ten of them were AFP-positive (AFP>10 mg/L). Overall survival (OS) was defined as the time from surgery to death and otherwise the patients were censored. Recurrence-free survival (RFS) was defined as the time from surgery to recurrence evaluation.
Statistical analysis Intergroup comparisons were analyzed by c2-testing for categorical, t-test or Wilcoxon-test for continuous variables as appropriate. Categorical variables were presented as proportions and continuous variables as mean standard deviation. Survival curves were estimated by the KaplaneMeier method and differences were evaluated by the log-rank test. Univariable and multivariable regression analysis for overall survival were performed by using the Cox proportional hazards model. Propensity score matching (PSM) for reduction of intergroup disparities was performed by logistic regression analysis on all variables with possible influence on the outcome variable (i.e. OS and RFS).Variable with the greatest estimated influence were selected using backward variable selection with an elimination criterion of p-value >0.2, and the propensity score computed.8,9 Shortly, multivariable logistic regression was performed on AFP-positivity (positive vs. negative) using all variables with possible influence on the patients’ survival. Variables finally included in the multivariable logistic regression after backward variable selection were sex, histology, Lauren’s classification, and UICC stages. Due to some missing data in these variables, only 2838 patients were used to generate the propensity score. A propensity score was then estimated for all subjects using this logistic regression, and AFP-positive patients (n ¼ 87) were matched to AFP-negative patients (n ¼ 87) using the nearest neighbor matching within a caliper of 0.20 times the standard deviation of the propensity score. Matching was performed without replacement. P-values of <0.05 were considered as statistically significant. All data were analyzed using SAS version 9 (SAS Institute Inc., Cary, NC, USA). This retrospective analysis was approved by the local institutional review board (IRB-Nr: NCC2014-0036). Results Baseline characteristics Direct comparison between AFPP-GC (n ¼ 97) and AFPN-GC (n ¼ 2937) patients revealed significant differences in preoperative AFP-levels (p < 0.0001), gender distribution (p ¼ 0.0013), tumor size (p ¼ 0.04), Lauren classification (p ¼ 0.0006), presence of lymphatic vessel infiltration (LVI) (p ¼ 0.0003), pT/pN/UICC-stage distribution (p < 0.0001 respectively) and recurrence rates (p < 0.0001). For AFP-positive patients no statistical difference in OS between patients with positive IHC or just AFP elevation over 10 mg/L (p ¼ 0.79) was found. There were no differences for age, location, histology (differentiation according to Japanese guidelines) and the number of retrieved lymph nodes. There was no statistically
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
D. Reim et al. / EJSO xx (2017) 1e8
significant difference for postoperative mortality between the two cohorts. Three patients died during the postoperative course, two in the AFP-negative and one in the AFPpositive cohort (p ¼ 0.09). After propensity score matching, all baseline characteristics used for matching were well-balanced. In the unmatched cohort, 841 AFPN-GC patients (36.7%) and 49 AFPP-GC patients (59.0%) received adjuvant chemotherapy (p < 0.0001). After matching, the proportion of patients receiving adjuvant chemotherapy was not statistically different anymore (37.9% AFPN-GC patients vs. 48.3% AFPP-GC patients, p ¼ 0.168). The results are summarized in Table 1.
3
Follow-up, recurrence and overall survival Median follow up for all patients was 52 [1e103] months and 61 [18e103] months for survivors. During follow-up, 44 patients (45.6%) in the AFPP-GC group died compared to 604 patients (20.6%) in the AFPN-GC group (p < 0.0001). OS was significantly different between two groups (log-rank, p < 0.0001). Five year survival-rates (FYSR) in the unmatched cohort were 79.4% for AFPNGC and 57.0% for AFPP-GC patients (p < 0.0001). The recurrence rate was significantly higher in the AFPP-GC group (45.4% vs. 20.4%, p < 0.0001). The predominant
Table 1 Baseline characteristics of AFP-positive and -negative patients before and after propensity score matching (PSM). Characteristics
Age (year) AFP (mg/L) Sex Tumor size (cm)
Location
Histology type Lauren classification
LVI pT
pN
Retrieved LN UICC-stage
Recurrence Recurrence site
Chemotherapy
Subgroup
Mean SD Mean SD Male Female Mean SD 5 >5 Upper Middle Lower Diffuse Differentiated Undifferentiated Intestinal Diffuse Mixed Absent Present pT1 pT2 pT3 pT4 pN0 pN1 pN2 pN3 Median [range] UICC I UICC II UICC III UICC IV Negative Positive Liver Other distant organ Peritoneum Regional LN No Yes
Total Patients (n ¼ 3034)
After PSM (n ¼ 174)
AFP-positive (%) (n ¼ 97)
AFP-negative (%) (n ¼ 2937)
P
AFP-positive (%) (n ¼ 87)
AFP-negative (%) (n ¼ 87)
P
57.6 12.9 91.4 167.8 80 (82.5) 17 (17.5) 5.6 2.8 53 (54.6) 44 (45.4) 17 (17.5) 21 (21.7) 54 (55.7) 5 (5.2) 43 (44.3) 54 (55.7) 62 (68.1) 22 (24.2) 7 (7.7) 39 (41.1) 56 (58.9) 23 (23.7) 44 (45.4) 19 (19.6) 11 (11.3) 31 (32.0) 36 (37.1) 22 (22.7) 8 (8.3) 41 [14e118] 33 (34.0) 27 (27.8) 19 (19.6) 18 (18.6) 63 (65.0) 34 (35.1) 15 (44.1) 4 (11.8) 4 (11.8) 11 (32.4) 34 (41.0) 49 (59.0)
57.7 11.8 3.2 1.5 1966 (66.9) 971 (33.1) 5.0 3.0 1840 (62.9) 1087 (37.1) 490 (16.7) 819 (27.9) 1534 (52.2) 94 (3.2) 1181 (40.2) 1756 (59.8) 1413 (49.8) 1264 (44.5) 162 (5.7) 1699 (59.6) 1153 (40.4) 1369 (46.7) 955 (32.6) 538 (18.4) 68 (2.3) 1672 (57.0) 724 (24.7) 305 (10.4) 235 (8.0) 41 [8e131] 1754 (59.7) 460 (15.7) 403 (13.7) 320 (10.9) 2479 (84.4) 458 (15.6) 98 (21.4) 120 (26.2) 116 (25.3) 124 (27.1) 1450 (63.3) 841 (36.7)
0.9066 <0.0001 0.0013
58.7 12.1 82.2 145.5 72 (82.8) 15 (17.2) 5.5 2.8 50 (57.5) 37 (42.5) 13 (14.9) 20 (23.0) 50 (57.5) 4 (4.6) 40 (46.0) 47 (54.0) 61 (70.1) 20 (23.0) 6 (6.9) 36 (41.4) 51 (58.6) 22 (25.3) 42 (48.3) 14 (16.1) 9 (10.3) 29 (33.3) 30 (34.5) 22 (25.3) 6 (6.9) 41 [14e118] 32 (36.8) 24 (27.6) 16 (18.4) 15 (17.2) 57 (65.5) 30 (34.5) 13 (43.3) 4 (13.3) 3 (10.0) 10 (33.3) 45 (51.7) 42 (48.3)
60.5 10.3 3.2 1.8 76 (87.4) 11 (12.6) 5.4 2.6 46 (52.9) 41 (47.1) 21 (24.1) 17 (19.5) 46 (52.9) 3 (3.5) 40 (46.0) 47 (54.0) 59 (67.8) 21 (24.1) 7 (8.1) 40 (46.0) 47 (54.0) 26 (29.9) 39 (44.8) 20 (23.0) 2 (2.3) 35 (40.2) 27 (31.0) 13 (15.0) 12 (13.8) 43 [20e94] 33 (37.9) 24 (27.6) 15 (17.2) 15 (17.2) 74 (85.1) 13 (14.9) 1 (7.7) 4 (30.8) 5 (38.4) 3 (23.1) 54 (62.1) 33 (37.9)
0.2841 <0.0001 0.3950
0.0400 0.0996 0.4454
0.4159 0.0006
0.0003 <0.0001
<0.0001
0.6593 <0.0001
<0.0001 0.0060
<0.0001
0.7340 0.5420 0.4841
1 0.9349
0.5409 0.1137
0.1693
0.8460 0.9973
0.0028 0.0212
0.1680
# Histologic type: According to the Japanese Gastric Cancer Classification, 14th edition: Differentiated ¼ papillary adenocarcinoma, tubular adenocarcinoma (well/moderately differentiated), Undifferentiated ¼ tubular adenocarcinoma (poorly/undifferentiated), signet ring cell carcinoma, mucinous carcinoma. & Staging: According to TNM classification system, 6th edition. $ LVI: Lymphatic vessel infiltration. Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
4
D. Reim et al. / EJSO xx (2017) 1e8
location of recurrence in AFPP-GC patients was the liver compared to AFPN-GC patients (44.1% vs. 21.4%, p ¼ 0.006). RFS was significantly different between the two groups (log-rank p-value <0.0001). Five year RFS rate was 59% for AFPN-GC and 52% for AFPP-GC patients (p < 0.0001). After propensity score matching (PSM), median OS was still significantly different between groups (log-rank, p ¼ 0.0121). FYSRs were 57.9% vs. 76.1% (AFPP-GC vs. AFPN-GC, p ¼ 0.014). The recurrence rate still was significantly higher in the AFPP-GC group (34.5% vs. 16.9%, p ¼ 0.003) without a predominant organ location. RFS was still significantly different between AFPN-GC and AFPP-GC patients (p ¼ 0.002). Five year RFS rate was 76% for AFPN-GC and 53% for AFPP-GC patients (p ¼ 0.0029). OS and RFS curves for the unmatched and PSM-groups are shown in Fig. 1. Cox regression analysis on OS and RFS of the unmatched and PSM-cohorts In univariable analysis of the unmatched groups, older age, AFP positivity, tumor size over 5 cm, BMI less than 25 kg/m2 tumor location (lower or diffuse), undifferentiated
histology, presence of lymphovascular invasion (LVI), diffuse/mixed types according to Lauren’s classification and advanced UICC-stage were significantly negatively related to overall survival. In the multivariable model, older age, AFP positivity, tumor size over 5 cm, BMI less than 25 kg/m2 tumor location (lower or diffuse), presence of LVI and advanced UICC-stage were significantly related to overall survival. After PSM, AFP-positivity was one of the strongest predictors of OS in the multivariable model (HR 2.98, CI 95% (1.7e5.1), p < 0.0001). Except AFP positivity, only age and pathologic tumor stage were significantly related to OS. The results are displayed in Table 2. Similar patterns were observed for RFS (Table 3). Discussion This PSM analysis on AFPP-GC patients demonstrates that AFP-positivity may be a relevant prognostic factor for OS and RFS. Several analyses investigated on this matter before,1e5 reporting on dismal outcome for this rare tumor entity. In this cohort the prevalence of AFPP-GC was only 3.2% in a period of five years. Former series reported on prevalence rates between 1 and 15%.2e4 The reasons for the relatively
Figure 1. Overall survival (OS) and recurrence-free survival (RFS) of the unmatched cohorts and after propensity score matching (PSM) related to AFPpositive gastric cancer. Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
Characteristics
Subgroup
Before PSM (n ¼ 2838)
After PSM (n ¼ 174)
Univariable
Age (year) AFP (mg/L) Sex Tumor size (cm) BMI (kg/m2) Location
Histology type LVI Lauren classification UICC-stage
(per 5 year) <10 10 Male Female 5 >5 <25 25 Upper Middle Lower Diffuse Differentiated Undifferentiated Absent Present Intestinal Diffuse Mixed UICC I UICC II UICC III UICC IV
Multivariable
Univariable
Multivariable
Hazard ratio
P
Hazard ratio
P
Hazard ratio
P
Hazard ratio
P
1.16 (1.12, 1.20) 1 2.69 (1.98, 3.65) 1 0.88 (0.74, 1.04) 1 4.03 (3.42, 4.75) 1 0.60 (0.50, 0.73) 1 0.59 (0.46, 0.77) 1.04 (0.84, 1.29) 4.24 (3.11, 5.78) 1 1.50 (1.27, 1.77) 1 5.08 (4.23, 6.10) 1 1.28 (1.09, 1.51) 1.22 (0.87, 1.72) 1 3.14 (2.38, 4.15) 8.12 (6.41, 10.28) 26.53 (21.21, 33.20)
<0.0001
1.17 (1.13, 1.21) 1 2.52 (1.81, 3.49)
<0.0001
1.12 (0.99, 1.26) 1 1.90 (1.14, 3.16) 1 1.29 (0.67, 2.48) 1 2.27 (1.37, 3.75) 1 0.61 (0.34, 1.11) 1 1.00 (0.43, 2.31) 1.31 (0.67, 2.57) 2.76 (0.88, 8.73) 1 1.02 (0.62, 1.67) 1 3.33 (1.84, 6.05) 1 0.94 (0.52, 1.71) 1.06 (0.42, 2.66) 1 1.35 (0.51, 3.58) 7.50 (3.31, 16.96) 14.93 (6.70, 33.26)
0.0770
1.27 (1.13, 1.44) 1 2.98 (1.75, 5.07)
<0.0001 <0.0001
1 1.44 (0.54, 3.85) 11.63 (4.99, 27.11) 24.41 (10.55, 56.47)
<0.0001 0.4666 <0.0001 <0.0001
<0.0001
<0.0001
0.1351 <0.0001 <0.0001 <0.0001 <0.0001 0.705 <0.0001
1 1.37 (1.12, 1.66) 1 0.66 (0.53, 0.81) 1 1.20 (1.00, 1.43) 1.96 (1.44, 2.65)
0.0018 <0.0001 <0.0001 0.0476 <0.0001
<0.0001 <0.0001 0.0112 0.0030 0.2546 <0.0001 <0.0001 <0.0001 <0.0001
1 1.38 (1.10, 1.73)
0.0049
1 2.20 (1.60, 3.03) 5.90 (4.39, 7.93) 17.19 (12.72, 23.24)
<0.0001 <0.0001 <0.0001 <0.0001
0.0136 0.4417
D. Reim et al. / EJSO xx (2017) 1e8
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
Table 2 Uni- and multivariable regression analysis for factors related to overall survival before and after PSM.
0.0015 0.1080 0.2990 0.9993 0.4292 0.0830 0.9543 <0.0001 0.9688 0.8380 0.9103 <0.0001 0.5538 <0.0001 <0.0001
# Histologic type: According to the Japanese Gastric Cancer Classification, 14th edition: Differentiated ¼ papillary adenocarcinoma, tubular adenocarcinoma (well/moderately differentiated), Undifferentiated ¼ tubular adenocarcinoma (poorly/undifferentiated), signet ring cell carcinoma, mucinous carcinoma. & Staging: According to TNM classification system, 6th edition. $ LVI: Lymphatic vessel infiltration.
5
6
Characteristics
Subgroup
Before PSM (n ¼ 2838)
After PSM (n ¼ 174)
Univariable
Age (year) AFP (mg/L) Sex Tumor size (cm) BMI (kg/m2) Location
Histology type LVI Lauren classification UICC-stage
(per 5 year) <10 10 Male Female 5 >5 <25 25 Upper Middle Lower Diffuse Differentiated Undifferentiated Absent Present Intestinal Diffuse Mixed UICC I UICC II UICC III UICC IV
Multivariable
Univariable
Multivariable
Hazard ratio
P
Hazard ratio
P
Hazard ratio
P
Hazard ratio
P
1.09 (1.06, 1.13) 1 2.83 (2.09, 3.85) 1 0.92 (0.78, 1.08) 1 4.04 (3.43, 4.76) 1 0.7 (0.58, 0.84) 1 0.63 (0.49, 0.82) 1.02 (0.82, 1.26) 3.81 (2.78, 5.23) 1 1.61 (1.36, 1.9) 1 5.18 (4.31, 6.22) 1 1.32 (1.12, 1.55) 1.08 (0.76, 1.56) 1 3.83 (2.88, 5.08) 10.44 (8.18, 13.32) 30.37 (24, 38.43)
<0.0001
1.1 (1.06, 1.14) 1 2.65 (1.91, 3.69)
<0.0001
1.08 (0.96, 1.22) 1 2.29 (1.33, 3.92) 1 1.26 (0.64, 2.48) 1 2.37 (1.4, 4) 1 0.62 (0.33, 1.14) 1 1.1 (0.47, 2.58) 1.37 (0.68, 2.77) 1.95 (0.53, 7.11) 1 1.23 (0.73, 2.06) 1 3.37 (1.82, 6.26) 1 1.08 (0.58, 1.98) 1.52 (0.64, 3.6) 1 1.2 (0.44, 3.31) 7.6 (3.36, 17.22) 13.13 (5.81, 29.65)
0.2111
1.23 (1.09, 1.38) 1 3.85 (2.16, 6.86)
0.0008
1 1.22 (0.44, 3.38) 11.37 (4.87, 26.53) 22.52 (9.5, 53.4)
<0.0001 0.6959 <0.0001 <0.0001
<0.0001
<0.0001
0.3120 <0.0001 0.0001 <0.0001 0.0005 0.8606 <0.0001
1 1.32 (1.09, 1.6)
0.0049
1
0.0095
1.11 (0.93, 1.33) 1.63 (1.19, 2.22)
0.2494 0.0023
<0.0001 <0.0001 0.0042 0.0010 0.6626 <0.0001 <0.0001 <0.0001 <0.0001
1 3.15 (2.33, 4.27) 9.2 (7.04, 12.04) 24.5 (18.71, 32.09)
<0.0001 <0.0001 <0.0001 <0.0001
0.0027
<0.0001
0.5128
D. Reim et al. / EJSO xx (2017) 1e8
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
Table 3 Uni- and multivariable regression analysis for factors related to recurrent free survival before and after PSM.
0.0013 0.1241 0.6687 0.8361 0.3806 0.3121 0.4371 0.0001 0.6355 0.8176 0.3417 <0.0001 0.7261 <0.0001 <0.0001
# Histologic type: According to the Japanese Gastric Cancer Classification, 14th edition: Differentiated ¼ papillary adenocarcinoma, tubular adenocarcinoma (well/moderately differentiated), Undifferentiated ¼ tubular adenocarcinoma (poorly/undifferentiated), signet ring cell carcinoma, mucinous carcinoma. & Staging: According to TNM classification system, 6th edition. $ LVI: Lymphatic vessel infiltration.
D. Reim et al. / EJSO xx (2017) 1e8
low AFP-positivity rate in this study may be attributed to the circumstance that only curatively resected patients without any residual disease (R0, M0) were included. Interestingly there is no standardized available definition for AFP-positivity for gastric cancer so far. Other series included only patients with a preoperative AFP level >40 mg/L and hepatoid features in the preoperative workup of the tumor-biopsy.2 Others defined AFP positivity only depending on the immunohistochemical result during postoperative histological workup irrespective of AFP levels.4 An analysis from Liu et al. combined AFP-positivity and preoperative serum AFP levels.3 Applying these different definitions of AFPP-GC, the prevalence in this cohort would have been even lower (1.8%). This analysis included all patients with an AFP level of 10 mg/L and over, irrespective of the preoperative biopsy result. This analysis, demonstrates no statistical difference in OS between patients with positive IHC or just AFP elevation over 10 mg/L. Therefore, we concluded that AFP elevation over 10 mg/L alone without IHC confirmation conveys unfavorable prognosis for the respective patient. The cutoff of 10 mg/ml was described in the literature. Therefore, any elevation of AFP above 10 mg/ml in gastric cancer patients was considered pathological irrespective of the individual value of each patient.18e21 Furthermore, direct comparisons to former analyses are not only hampered by different definitions of AFPGC but also by substantially different baseline characteristic of the respective patient cohorts. The most recent case series4 included patients in metastasized stages and patients undergoing palliative resections. In contrast, this analysis reports only on curatively resected patients without proof of metastases (R0, UICC-IV stages are derived from N3status in the 6th edition of TNM classification system7). This is also reflected by a higher FYSR of 78.7% in this analysis compared to former publications (FYSR 34e50%1e5). Previous publications consisted either of only case series or caseecontrol studies in which a selection bias may not be excluded.1e5 Here we introduced PSM which is a recognized method for balancing pronounced differences in baseline characteristics between two groups. PSM is a statistical method applied to reduce the possible selection bias in observational studies, which was initially introduced by Rubin and Rosenbaum in 1983.8 There are even hints that PSM may create results as if a randomized controlled trial would have been performed.10 The simple comparisons between the differently sized groups revealed marked differences between AFPP-GC and AFPN-GC patients. The AFPP-GC group revealed more advanced pT-stages, a higher rate of lymph node metastasis and a higher rate of lymphatic vessel infiltration. Survival analysis of the unmatched groups revealed dismal prognosis for AFPpositivity. Those marked differences in pathological findings might explain why survival rates were significantly lower in the AFP-positive group. This analysis
7
demonstrates that the PSM created well-balanced cohorts. Pathological tumor properties such as pT-/pN/-UICC stages, Lauren classification and LVI were comparable after the matching. However, survival analysis revealed that survival differences still exist between the two groups and multivariable regression analysis demonstrated that, besides the pathologic tumor stage, AFP positivity was a prognostic factor for OS and RFS. The liver as primary recurrence location in AFPP-GC was significantly predominant before and after PSM. Other studies showed similar results11,12 and concluded that predominantly occurring liver metastasis might be caused by early venous invasion in AFPPGC.11 There are several studies which investigated on the biologic properties of AFPP-GC.12e15 Compared to common gastric cancers AFPP-GC was reported to have a high proliferative, pro-angiogenetic and reduced apoptotic activity.12 It was described that the aggressive phenotype of AFPP-GC may be related to c-Met overexpression.14,15 The c-Met proto-oncogene encodes the c-Met receptor which is known to regulate cell proliferation or migration.14 Hepatocyte growth factor (HGF) has been identified as its ligand,16 is considered an important regulator of liver regeneration in response to injury and promotes epithelialcell mobilization.17 It was reported that AFPP-GC cells with higher c-Met expression behave more invasively in response to HGF.16 The higher recurrence rates were explained to be related to HGF-release during abdominal surgery, which might affect proliferation/migration of AFPP-GC cells.1 Therefore, meticulous surgery and reduction of perioperative trauma should be of highest interest for patients with AFPP-GC from the surgical perspective. Moreover, immunotherapy targeting c-Met and/or HGF could be a promising treatment of this clinically distinct entity. Further investigation for proteogenomic profiles underlying AFPP-GC is required to identify the tentative target. There are several limitations for this analysis: Although PSM is an accepted tool to overcome selection-bias, the data were analyzed retrospectively. Another limitation is that there is a huge disparity of the patient-numbers between AFP-positive and AFP-negative gastric cancer patients. Therefore no clear conclusions can be drawn on those many patients not having been included in the PSM-cohort. Moreover, PSM has some limitations which are the inability to account for unmeasured factors like surgical quality, biologic and genetic differences and the need for the statistical analysis to account for the paired nature of the matched samples, possibly explaining the remaining differences in outcome. Further, some patients with incomplete datasets had to be omitted from the analysis due to the reasons described above, which may have led to a certain kind of selection bias. Besides that, among the 3034 patients, 194 patients either had liver cirrhosis, or hepatitis, or both, and 14 of them were AFP-positive (AFP >10 mg/L), which may have affected the final analysis.
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
8
D. Reim et al. / EJSO xx (2017) 1e8
However, none of the patients had HCC. Due to the fact that AFP is usually affected mostly by HCC, none of the patients were excluded from the analysis. In conclusion, this PSM analysis demonstrates that a preoperative elevation of AFP (>10 mg/L) can be considered as an independent poor prognostic predictor of OS and RFS in gastric cancer patients. It was shown that biological differences in common gastric cancers are in fact related to more aggressive behavior of this malignancy. Meticulous surgery with perioperative reduction of surgical trauma may improve prognosis for those patients. Due to high recurrence rates, more frequent follow-up visits and more aggressive adjuvant therapy including targeted immunotherapy should be considered. Further we propose and encourage over-regional co-operations to define a clear definition of AFPP-GC as a distinct clinical entity. Conflict of interest Statement This manuscript fully consists of original material, which has not been published or submitted to another journal. All authors declare that they participated in the analysis and that they have seen and approved the final version. None of the authors has any conflicts of interest. No funds or grants or company gifts have been received, nor has the article been written by a third party. Data acquisition and analysis was performed according to local ethical standards and GCP guidelines. Acknowledgements This work was supported by the National Cancer Center grant NCC-1410130-1. We thank Hyun Jung Park (Research Nurse) and Soo Hee Kim (Research Nurse) for the extraordinary support in data collection and preparation of data.
References 1. Kono K, Amemiya H, Sekikawa T, et al. Clinicopathologic features of gastric cancers producing alpha-fetoprotein. Dig Surg 2002;19(5): 359–65. discussion 365. 2. Inoue M, Sano T, Kuchiba A, Taniguchi H, Fukagawa T, Katai H. Long-term results of gastrectomy for alpha-fetoprotein-producing gastric cancer. Br J Surg Jul 2010;97(7):1056–61. 3. Liu X, Cheng Y, Sheng W, et al. Clinicopathologic features and prognostic factors in alpha-fetoprotein-producing gastric cancers: analysis of 104 cases. J Surg Oncol Sep 1 2010;102(3):249–55. 4. Hirajima S, Komatsu S, Ichikawa D, et al. Liver metastasis is the only independent prognostic factor in AFP-producing gastric cancer. World J Gastroenterol Sep 28 2013;19(36):6055–61. 5. Bourreille J, Metayer P, Sauger F, Matray F, Fondimare A. Existence of alpha feto protein during gastric-origin secondary cancer of the liver. Presse Med Jun 6 1970;78(28):1277–8.
6. Japanese Gastric Cancer Association. Japanese gastric cancer treatment guidelines 2010 (ver. 3). Gastric Cancer Jun 2011;14(2):113–23. 7. UICC. TNM classification of malignant tumours. In: Sobin LH, Wittekind C, editors. 6th ed. New York, NY: Wiley-Liss; 2002. 8. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika April 1, 1983; 70(1):41–55. 9. Ho D, Imai K, King G, Stuart E. Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Polit Anal 2007;15:199–236. 10. Kuss O, Legler T, Borgermann J. Treatments effects from randomized trials and propensity score analyses were similar in similar populations in an example from cardiac surgery. J Clin Epidemiol Oct 2011; 64(10):1076–84. 11. Chang YC, Nagasue N, Abe S, Taniura H, Kumar DD, Nakamura T. Comparison between the clinicopathologic features of AFP-positive and AFP-negative gastric cancers. Am J Gastroenterol Mar 1992; 87(3):321–5. 12. Koide N, Nishio A, Igarashi J, Kajikawa S, Adachi W, Amano J. Alpha-fetoprotein-producing gastric cancer: histochemical analysis of cell proliferation, apoptosis, and angiogenesis. Am J Gastroenterol Jun 1999;94(6):1658–63. 13. Ishigami S, Natsugoe S, Nakashima H, et al. Biological aggressiveness of alpha-fetoprotein (AFP)-positive gastric cancer. Hepatogastroenterology MayeJun 2006;53(69):338–41. 14. Amemiya H, Kono K, Mori Y, et al. High frequency of c-Met expression in gastric cancers producing alpha-fetoprotein. Oncology Aug 2000;59(2):145–51. 15. Kaji M, Yonemura Y, Harada S, Liu X, Terada I, Yamamoto H. Participation of c-met in the progression of human gastric cancers: anti-cmet oligonucleotides inhibit proliferation or invasiveness of gastric cancer cells. Cancer Gene Ther NoveDec 1996;3(6):393–404. 16. Matsumoto K, Nakamura T. Hepatocyte growth factor: molecular structure, roles in liver regeneration, and other biological functions. Crit Rev Oncog 1992;3(1e2):27–54. 17. Weidner KM, Behrens J, Vandekerckhove J, Birchmeier W. Scatter factor: molecular characteristics and effect on the invasiveness of epithelial cells. J Cell Biol Nov 1990;111(5 Pt 1):2097–108. 18. Akuta N, Suzuki F, Kawamura Y, et al. Substitution of amino acid 70 in the hepatitis C virus core region of genotype 1b is an important predictor of elevated alpha-fetoprotein in patients without hepatocellular carcinoma. J Med Virol [Internet] 2008 Aug;80(8):1354–62. 19. Ball D, Rose E, Alpert E. Alpha-fetoprotein levels in normal adults. Am J Med Sci [Internet] 1992 Mar;303(3):157–9. 20. Li P, Wang S-S, Liu H, et al. Elevated serum alpha fetoprotein levels promote pathological progression of hepatocellular carcinoma. World J Gastroenterol [Internet] 2011 Nov 7;17(41):4563–71. 21. Liu X, Cai H, Wang Y. Prognostic significance of tumor markers in T4a gastric cancer. World J Surg Oncol [Internet] 2012 Apr 27;10:68.
Abbreviations AFPN-GC: AFP-negative gastric cancer. AFPP-GC: AFP-positive gastric cancer. BMI: body mass index. FYSR: five year survival rate. HR: hazard ratio. ICH: immunohistochemistry. LVI: lymphatic vessel infiltration. OS: overall survival. PSM: propensity score matching. RFS: recurrence free survival. UICC: Union internationale contre le cancer.
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
ScienceDirect EJSO xx (2017) 1e8
www.ejso.com
Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection D. Reim a,b,d, Y.-S. Choi a,d, H.M. Yoon a, B. Park c, B.W. Eom a, M.-C. Kook a, K.W. Ryu a, I.J. Choi a, J. Joo c,**,e, Y.-W. Kim a,*,e a
Gastric Cancer Branch, Research Institute & Hospital, National Cancer Center, Ilsan-ro 323, Ilsandong-gu, Goyang-si Gyeonggi-do, 410-769, Republic of Korea b Klinikum Rechts der Isar der Technischen Universit€at M€unchen, Department of Surgery, Ismaninger Strasse 22, 81675 Munich, Germany c Biometric Research Branch, Clinical Research Coordination Center, National Cancer Center, Ilsan-ro 323, Ilsandong-gu, Goyang-si Gyeonggi-do, 410-769, Republic of Korea Accepted 21 April 2017 Available online - - -
Abstract Background: Prognosis of alpha-fetoprotein positive gastric cancer (AFPP-GC) remains elusive so far due to disparities in cohort size and baseline characteristics in previous studies. A propensity score matching (PSM) analysis as well as multivariable model was performed for unbiased evaluation of the outcome in AFPGC. Methods: Among 3034 gastric cancer patients who underwent curative gastric cancer surgery (R0, M0) at the National Cancer Center, Korea between 2002 and 2007, we identified 97 patients being positive for AFP either by elevation of serum-AFP levels >10 mg/L or by immunohistochemical staining. Due to marked disparities in baseline characteristics and cohort size, propensity-score-matching was performed which matched 87 AFPP-GC patients to the same number of AFP-negative gastric cancer (AFPN-GC) patients. Baseline characteristics were compared using c2-test. Survival curves were compared using the Kaplan-Meier-method and multivariable regression analysis was performed to evaluate the effect of AFP-positivity while adjusting the effects of confounding variables. Results: AFPP-GC and AFPN-GC patients revealed marked disparities in patient cohorts. After PSM, groups were balanced for age, sex, tumor size, BMI, tumor location, grade of differentiation, presence of lymphatic vessel infiltration (LVI), Lauren histologic type and stage distribution. In multivariable regression analysis of the PSM-groups, only AFP-positivity and pathologic stage were predictive for overall survival (HR 2.98, CI 95% {1.7e5.1}, p < 0.0001). Five-year-survival rates were significantly worse for AFPP-GC patients (57.9% vs. 76.1%, p ¼ 0.014). Recurrence was significantly more frequent in AFPP-GC patients (p ¼ 0.003). Conclusion: AFP can be considered as an independent negative predictor of overall and recurrence-free survival in patients with gastric cancer. Ó 2017 Elsevier Ltd, BASO ~ The Association for Cancer Surgery, and the European Society of Surgical Oncology. All rights reserved.
Keywords: Gastric cancer; AFP; Surgery
Introduction Several case series and retrospective analyses on alphafetoprotein (AFP)-positive gastric cancer were published * Corresponding author. Fax: þ82 31 920 0696. ** Corresponding author. E-mail addresses: [email protected] (J. Joo), [email protected] (Y.-W. Kim). d DR and YSC contributed equally to this study as a first author. e JJ and YWK contributed equally to this study as a Corresponding author.
over the recent years.1e4 AFP-positive gastric cancer (AFPP-GC) appears to be a rare entity compared to common gastric cancer. The incidence is reported to be 1e15% among gastric cancer patients. AFPP-GCs combined with liver metastasis were first reported in 1970.5 Further reports concluded that AFPP-GC revealed poor prognosis, high recurrence rates and high potential to metastasize to the liver and lymph nodes.1e5 This appears to be related to a histologically more aggressive tumor phenotype. AFPP-GCs tend to show higher rates of
http://dx.doi.org/10.1016/j.ejso.2017.04.005 0748-7983/Ó 2017 Elsevier Ltd, BASO ~ The Association for Cancer Surgery, and the European Society of Surgical Oncology. All rights reserved. Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
2
D. Reim et al. / EJSO xx (2017) 1e8
lymphatic and blood vessel infiltration,1 which may lead to higher recurrence rates and shorter postoperative survival. Most of the previous reports could not provide data for direct comparisons between AFPP-GC and AFP-negative gastric cancer (AFPN-GC) due to small and heterogeneous cohort sizes and marked differences in baseline characteristics.2 Only one Chinese analysis reported a matched cohort result.3 However, the matching algorithm was not reported which leads to the notion that selection bias may not be completely excluded. Therefore the purpose of this analysis was to examine prognosis of AFPP-GC from a large patient population after propensity score matching (PSM) which attempts to eliminate the disparities in cohort size and baseline characteristics. Patients and methods The screening of the prospectively documented database for gastric cancer at the National Cancer Center Korea (NCCK) revealed 3034 patients who were subjected to primary curative oncologic resection (R0) for gastric cancer between January 2002 and August 2007 without undergoing preoperative chemo- or radiochemotherapy. All patients in whom metastatic disease was ruled out, received oncologic resection according to Japanese guidelines including D2 lymph node dissection.6 All specimens were examined postoperatively by specialized pathologists and staged according to the 6th UICC edition TNM staging system.7 Serum AFP was tested in all patients preoperatively and were considered as AFP-positive when elevated preoperative serum AFP levels (>10 mg/L) were detected. IHC was only done in those patients with preoperatively elevated AFP-levels and hepatoid features in the preoperative biopsy. Finally, 97 patients were identified as AFP-positive. Patients received adjuvant chemotherapy in case of metastatic lymph nodes and/or from pathologic stages II-IV. Patients were followed postoperatively via the outpatient clinic in regular intervals (every 6 months for the first 24 months and then every 12 months until at least 60 months postoperatively for early gastric cancer patients and every 3 months until 3 years postoperatively and then every 6 months until the sixth postoperative year for advanced gastric cancer patients). The follow-up period was calculated from the day of surgery to the last follow-up date. During this five year period 187 patients were diagnosed with gastric cancer but didn’t receive curative resections. Among them, 174 patients were tested for AFP and ten of them were AFP-positive (AFP>10 mg/L). Overall survival (OS) was defined as the time from surgery to death and otherwise the patients were censored. Recurrence-free survival (RFS) was defined as the time from surgery to recurrence evaluation.
Statistical analysis Intergroup comparisons were analyzed by c2-testing for categorical, t-test or Wilcoxon-test for continuous variables as appropriate. Categorical variables were presented as proportions and continuous variables as mean standard deviation. Survival curves were estimated by the KaplaneMeier method and differences were evaluated by the log-rank test. Univariable and multivariable regression analysis for overall survival were performed by using the Cox proportional hazards model. Propensity score matching (PSM) for reduction of intergroup disparities was performed by logistic regression analysis on all variables with possible influence on the outcome variable (i.e. OS and RFS).Variable with the greatest estimated influence were selected using backward variable selection with an elimination criterion of p-value >0.2, and the propensity score computed.8,9 Shortly, multivariable logistic regression was performed on AFP-positivity (positive vs. negative) using all variables with possible influence on the patients’ survival. Variables finally included in the multivariable logistic regression after backward variable selection were sex, histology, Lauren’s classification, and UICC stages. Due to some missing data in these variables, only 2838 patients were used to generate the propensity score. A propensity score was then estimated for all subjects using this logistic regression, and AFP-positive patients (n ¼ 87) were matched to AFP-negative patients (n ¼ 87) using the nearest neighbor matching within a caliper of 0.20 times the standard deviation of the propensity score. Matching was performed without replacement. P-values of <0.05 were considered as statistically significant. All data were analyzed using SAS version 9 (SAS Institute Inc., Cary, NC, USA). This retrospective analysis was approved by the local institutional review board (IRB-Nr: NCC2014-0036). Results Baseline characteristics Direct comparison between AFPP-GC (n ¼ 97) and AFPN-GC (n ¼ 2937) patients revealed significant differences in preoperative AFP-levels (p < 0.0001), gender distribution (p ¼ 0.0013), tumor size (p ¼ 0.04), Lauren classification (p ¼ 0.0006), presence of lymphatic vessel infiltration (LVI) (p ¼ 0.0003), pT/pN/UICC-stage distribution (p < 0.0001 respectively) and recurrence rates (p < 0.0001). For AFP-positive patients no statistical difference in OS between patients with positive IHC or just AFP elevation over 10 mg/L (p ¼ 0.79) was found. There were no differences for age, location, histology (differentiation according to Japanese guidelines) and the number of retrieved lymph nodes. There was no statistically
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
D. Reim et al. / EJSO xx (2017) 1e8
significant difference for postoperative mortality between the two cohorts. Three patients died during the postoperative course, two in the AFP-negative and one in the AFPpositive cohort (p ¼ 0.09). After propensity score matching, all baseline characteristics used for matching were well-balanced. In the unmatched cohort, 841 AFPN-GC patients (36.7%) and 49 AFPP-GC patients (59.0%) received adjuvant chemotherapy (p < 0.0001). After matching, the proportion of patients receiving adjuvant chemotherapy was not statistically different anymore (37.9% AFPN-GC patients vs. 48.3% AFPP-GC patients, p ¼ 0.168). The results are summarized in Table 1.
3
Follow-up, recurrence and overall survival Median follow up for all patients was 52 [1e103] months and 61 [18e103] months for survivors. During follow-up, 44 patients (45.6%) in the AFPP-GC group died compared to 604 patients (20.6%) in the AFPN-GC group (p < 0.0001). OS was significantly different between two groups (log-rank, p < 0.0001). Five year survival-rates (FYSR) in the unmatched cohort were 79.4% for AFPNGC and 57.0% for AFPP-GC patients (p < 0.0001). The recurrence rate was significantly higher in the AFPP-GC group (45.4% vs. 20.4%, p < 0.0001). The predominant
Table 1 Baseline characteristics of AFP-positive and -negative patients before and after propensity score matching (PSM). Characteristics
Age (year) AFP (mg/L) Sex Tumor size (cm)
Location
Histology type Lauren classification
LVI pT
pN
Retrieved LN UICC-stage
Recurrence Recurrence site
Chemotherapy
Subgroup
Mean SD Mean SD Male Female Mean SD 5 >5 Upper Middle Lower Diffuse Differentiated Undifferentiated Intestinal Diffuse Mixed Absent Present pT1 pT2 pT3 pT4 pN0 pN1 pN2 pN3 Median [range] UICC I UICC II UICC III UICC IV Negative Positive Liver Other distant organ Peritoneum Regional LN No Yes
Total Patients (n ¼ 3034)
After PSM (n ¼ 174)
AFP-positive (%) (n ¼ 97)
AFP-negative (%) (n ¼ 2937)
P
AFP-positive (%) (n ¼ 87)
AFP-negative (%) (n ¼ 87)
P
57.6 12.9 91.4 167.8 80 (82.5) 17 (17.5) 5.6 2.8 53 (54.6) 44 (45.4) 17 (17.5) 21 (21.7) 54 (55.7) 5 (5.2) 43 (44.3) 54 (55.7) 62 (68.1) 22 (24.2) 7 (7.7) 39 (41.1) 56 (58.9) 23 (23.7) 44 (45.4) 19 (19.6) 11 (11.3) 31 (32.0) 36 (37.1) 22 (22.7) 8 (8.3) 41 [14e118] 33 (34.0) 27 (27.8) 19 (19.6) 18 (18.6) 63 (65.0) 34 (35.1) 15 (44.1) 4 (11.8) 4 (11.8) 11 (32.4) 34 (41.0) 49 (59.0)
57.7 11.8 3.2 1.5 1966 (66.9) 971 (33.1) 5.0 3.0 1840 (62.9) 1087 (37.1) 490 (16.7) 819 (27.9) 1534 (52.2) 94 (3.2) 1181 (40.2) 1756 (59.8) 1413 (49.8) 1264 (44.5) 162 (5.7) 1699 (59.6) 1153 (40.4) 1369 (46.7) 955 (32.6) 538 (18.4) 68 (2.3) 1672 (57.0) 724 (24.7) 305 (10.4) 235 (8.0) 41 [8e131] 1754 (59.7) 460 (15.7) 403 (13.7) 320 (10.9) 2479 (84.4) 458 (15.6) 98 (21.4) 120 (26.2) 116 (25.3) 124 (27.1) 1450 (63.3) 841 (36.7)
0.9066 <0.0001 0.0013
58.7 12.1 82.2 145.5 72 (82.8) 15 (17.2) 5.5 2.8 50 (57.5) 37 (42.5) 13 (14.9) 20 (23.0) 50 (57.5) 4 (4.6) 40 (46.0) 47 (54.0) 61 (70.1) 20 (23.0) 6 (6.9) 36 (41.4) 51 (58.6) 22 (25.3) 42 (48.3) 14 (16.1) 9 (10.3) 29 (33.3) 30 (34.5) 22 (25.3) 6 (6.9) 41 [14e118] 32 (36.8) 24 (27.6) 16 (18.4) 15 (17.2) 57 (65.5) 30 (34.5) 13 (43.3) 4 (13.3) 3 (10.0) 10 (33.3) 45 (51.7) 42 (48.3)
60.5 10.3 3.2 1.8 76 (87.4) 11 (12.6) 5.4 2.6 46 (52.9) 41 (47.1) 21 (24.1) 17 (19.5) 46 (52.9) 3 (3.5) 40 (46.0) 47 (54.0) 59 (67.8) 21 (24.1) 7 (8.1) 40 (46.0) 47 (54.0) 26 (29.9) 39 (44.8) 20 (23.0) 2 (2.3) 35 (40.2) 27 (31.0) 13 (15.0) 12 (13.8) 43 [20e94] 33 (37.9) 24 (27.6) 15 (17.2) 15 (17.2) 74 (85.1) 13 (14.9) 1 (7.7) 4 (30.8) 5 (38.4) 3 (23.1) 54 (62.1) 33 (37.9)
0.2841 <0.0001 0.3950
0.0400 0.0996 0.4454
0.4159 0.0006
0.0003 <0.0001
<0.0001
0.6593 <0.0001
<0.0001 0.0060
<0.0001
0.7340 0.5420 0.4841
1 0.9349
0.5409 0.1137
0.1693
0.8460 0.9973
0.0028 0.0212
0.1680
# Histologic type: According to the Japanese Gastric Cancer Classification, 14th edition: Differentiated ¼ papillary adenocarcinoma, tubular adenocarcinoma (well/moderately differentiated), Undifferentiated ¼ tubular adenocarcinoma (poorly/undifferentiated), signet ring cell carcinoma, mucinous carcinoma. & Staging: According to TNM classification system, 6th edition. $ LVI: Lymphatic vessel infiltration. Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
4
D. Reim et al. / EJSO xx (2017) 1e8
location of recurrence in AFPP-GC patients was the liver compared to AFPN-GC patients (44.1% vs. 21.4%, p ¼ 0.006). RFS was significantly different between the two groups (log-rank p-value <0.0001). Five year RFS rate was 59% for AFPN-GC and 52% for AFPP-GC patients (p < 0.0001). After propensity score matching (PSM), median OS was still significantly different between groups (log-rank, p ¼ 0.0121). FYSRs were 57.9% vs. 76.1% (AFPP-GC vs. AFPN-GC, p ¼ 0.014). The recurrence rate still was significantly higher in the AFPP-GC group (34.5% vs. 16.9%, p ¼ 0.003) without a predominant organ location. RFS was still significantly different between AFPN-GC and AFPP-GC patients (p ¼ 0.002). Five year RFS rate was 76% for AFPN-GC and 53% for AFPP-GC patients (p ¼ 0.0029). OS and RFS curves for the unmatched and PSM-groups are shown in Fig. 1. Cox regression analysis on OS and RFS of the unmatched and PSM-cohorts In univariable analysis of the unmatched groups, older age, AFP positivity, tumor size over 5 cm, BMI less than 25 kg/m2 tumor location (lower or diffuse), undifferentiated
histology, presence of lymphovascular invasion (LVI), diffuse/mixed types according to Lauren’s classification and advanced UICC-stage were significantly negatively related to overall survival. In the multivariable model, older age, AFP positivity, tumor size over 5 cm, BMI less than 25 kg/m2 tumor location (lower or diffuse), presence of LVI and advanced UICC-stage were significantly related to overall survival. After PSM, AFP-positivity was one of the strongest predictors of OS in the multivariable model (HR 2.98, CI 95% (1.7e5.1), p < 0.0001). Except AFP positivity, only age and pathologic tumor stage were significantly related to OS. The results are displayed in Table 2. Similar patterns were observed for RFS (Table 3). Discussion This PSM analysis on AFPP-GC patients demonstrates that AFP-positivity may be a relevant prognostic factor for OS and RFS. Several analyses investigated on this matter before,1e5 reporting on dismal outcome for this rare tumor entity. In this cohort the prevalence of AFPP-GC was only 3.2% in a period of five years. Former series reported on prevalence rates between 1 and 15%.2e4 The reasons for the relatively
Figure 1. Overall survival (OS) and recurrence-free survival (RFS) of the unmatched cohorts and after propensity score matching (PSM) related to AFPpositive gastric cancer. Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
Characteristics
Subgroup
Before PSM (n ¼ 2838)
After PSM (n ¼ 174)
Univariable
Age (year) AFP (mg/L) Sex Tumor size (cm) BMI (kg/m2) Location
Histology type LVI Lauren classification UICC-stage
(per 5 year) <10 10 Male Female 5 >5 <25 25 Upper Middle Lower Diffuse Differentiated Undifferentiated Absent Present Intestinal Diffuse Mixed UICC I UICC II UICC III UICC IV
Multivariable
Univariable
Multivariable
Hazard ratio
P
Hazard ratio
P
Hazard ratio
P
Hazard ratio
P
1.16 (1.12, 1.20) 1 2.69 (1.98, 3.65) 1 0.88 (0.74, 1.04) 1 4.03 (3.42, 4.75) 1 0.60 (0.50, 0.73) 1 0.59 (0.46, 0.77) 1.04 (0.84, 1.29) 4.24 (3.11, 5.78) 1 1.50 (1.27, 1.77) 1 5.08 (4.23, 6.10) 1 1.28 (1.09, 1.51) 1.22 (0.87, 1.72) 1 3.14 (2.38, 4.15) 8.12 (6.41, 10.28) 26.53 (21.21, 33.20)
<0.0001
1.17 (1.13, 1.21) 1 2.52 (1.81, 3.49)
<0.0001
1.12 (0.99, 1.26) 1 1.90 (1.14, 3.16) 1 1.29 (0.67, 2.48) 1 2.27 (1.37, 3.75) 1 0.61 (0.34, 1.11) 1 1.00 (0.43, 2.31) 1.31 (0.67, 2.57) 2.76 (0.88, 8.73) 1 1.02 (0.62, 1.67) 1 3.33 (1.84, 6.05) 1 0.94 (0.52, 1.71) 1.06 (0.42, 2.66) 1 1.35 (0.51, 3.58) 7.50 (3.31, 16.96) 14.93 (6.70, 33.26)
0.0770
1.27 (1.13, 1.44) 1 2.98 (1.75, 5.07)
<0.0001 <0.0001
1 1.44 (0.54, 3.85) 11.63 (4.99, 27.11) 24.41 (10.55, 56.47)
<0.0001 0.4666 <0.0001 <0.0001
<0.0001
<0.0001
0.1351 <0.0001 <0.0001 <0.0001 <0.0001 0.705 <0.0001
1 1.37 (1.12, 1.66) 1 0.66 (0.53, 0.81) 1 1.20 (1.00, 1.43) 1.96 (1.44, 2.65)
0.0018 <0.0001 <0.0001 0.0476 <0.0001
<0.0001 <0.0001 0.0112 0.0030 0.2546 <0.0001 <0.0001 <0.0001 <0.0001
1 1.38 (1.10, 1.73)
0.0049
1 2.20 (1.60, 3.03) 5.90 (4.39, 7.93) 17.19 (12.72, 23.24)
<0.0001 <0.0001 <0.0001 <0.0001
0.0136 0.4417
D. Reim et al. / EJSO xx (2017) 1e8
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
Table 2 Uni- and multivariable regression analysis for factors related to overall survival before and after PSM.
0.0015 0.1080 0.2990 0.9993 0.4292 0.0830 0.9543 <0.0001 0.9688 0.8380 0.9103 <0.0001 0.5538 <0.0001 <0.0001
# Histologic type: According to the Japanese Gastric Cancer Classification, 14th edition: Differentiated ¼ papillary adenocarcinoma, tubular adenocarcinoma (well/moderately differentiated), Undifferentiated ¼ tubular adenocarcinoma (poorly/undifferentiated), signet ring cell carcinoma, mucinous carcinoma. & Staging: According to TNM classification system, 6th edition. $ LVI: Lymphatic vessel infiltration.
5
6
Characteristics
Subgroup
Before PSM (n ¼ 2838)
After PSM (n ¼ 174)
Univariable
Age (year) AFP (mg/L) Sex Tumor size (cm) BMI (kg/m2) Location
Histology type LVI Lauren classification UICC-stage
(per 5 year) <10 10 Male Female 5 >5 <25 25 Upper Middle Lower Diffuse Differentiated Undifferentiated Absent Present Intestinal Diffuse Mixed UICC I UICC II UICC III UICC IV
Multivariable
Univariable
Multivariable
Hazard ratio
P
Hazard ratio
P
Hazard ratio
P
Hazard ratio
P
1.09 (1.06, 1.13) 1 2.83 (2.09, 3.85) 1 0.92 (0.78, 1.08) 1 4.04 (3.43, 4.76) 1 0.7 (0.58, 0.84) 1 0.63 (0.49, 0.82) 1.02 (0.82, 1.26) 3.81 (2.78, 5.23) 1 1.61 (1.36, 1.9) 1 5.18 (4.31, 6.22) 1 1.32 (1.12, 1.55) 1.08 (0.76, 1.56) 1 3.83 (2.88, 5.08) 10.44 (8.18, 13.32) 30.37 (24, 38.43)
<0.0001
1.1 (1.06, 1.14) 1 2.65 (1.91, 3.69)
<0.0001
1.08 (0.96, 1.22) 1 2.29 (1.33, 3.92) 1 1.26 (0.64, 2.48) 1 2.37 (1.4, 4) 1 0.62 (0.33, 1.14) 1 1.1 (0.47, 2.58) 1.37 (0.68, 2.77) 1.95 (0.53, 7.11) 1 1.23 (0.73, 2.06) 1 3.37 (1.82, 6.26) 1 1.08 (0.58, 1.98) 1.52 (0.64, 3.6) 1 1.2 (0.44, 3.31) 7.6 (3.36, 17.22) 13.13 (5.81, 29.65)
0.2111
1.23 (1.09, 1.38) 1 3.85 (2.16, 6.86)
0.0008
1 1.22 (0.44, 3.38) 11.37 (4.87, 26.53) 22.52 (9.5, 53.4)
<0.0001 0.6959 <0.0001 <0.0001
<0.0001
<0.0001
0.3120 <0.0001 0.0001 <0.0001 0.0005 0.8606 <0.0001
1 1.32 (1.09, 1.6)
0.0049
1
0.0095
1.11 (0.93, 1.33) 1.63 (1.19, 2.22)
0.2494 0.0023
<0.0001 <0.0001 0.0042 0.0010 0.6626 <0.0001 <0.0001 <0.0001 <0.0001
1 3.15 (2.33, 4.27) 9.2 (7.04, 12.04) 24.5 (18.71, 32.09)
<0.0001 <0.0001 <0.0001 <0.0001
0.0027
<0.0001
0.5128
D. Reim et al. / EJSO xx (2017) 1e8
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
Table 3 Uni- and multivariable regression analysis for factors related to recurrent free survival before and after PSM.
0.0013 0.1241 0.6687 0.8361 0.3806 0.3121 0.4371 0.0001 0.6355 0.8176 0.3417 <0.0001 0.7261 <0.0001 <0.0001
# Histologic type: According to the Japanese Gastric Cancer Classification, 14th edition: Differentiated ¼ papillary adenocarcinoma, tubular adenocarcinoma (well/moderately differentiated), Undifferentiated ¼ tubular adenocarcinoma (poorly/undifferentiated), signet ring cell carcinoma, mucinous carcinoma. & Staging: According to TNM classification system, 6th edition. $ LVI: Lymphatic vessel infiltration.
D. Reim et al. / EJSO xx (2017) 1e8
low AFP-positivity rate in this study may be attributed to the circumstance that only curatively resected patients without any residual disease (R0, M0) were included. Interestingly there is no standardized available definition for AFP-positivity for gastric cancer so far. Other series included only patients with a preoperative AFP level >40 mg/L and hepatoid features in the preoperative workup of the tumor-biopsy.2 Others defined AFP positivity only depending on the immunohistochemical result during postoperative histological workup irrespective of AFP levels.4 An analysis from Liu et al. combined AFP-positivity and preoperative serum AFP levels.3 Applying these different definitions of AFPP-GC, the prevalence in this cohort would have been even lower (1.8%). This analysis included all patients with an AFP level of 10 mg/L and over, irrespective of the preoperative biopsy result. This analysis, demonstrates no statistical difference in OS between patients with positive IHC or just AFP elevation over 10 mg/L. Therefore, we concluded that AFP elevation over 10 mg/L alone without IHC confirmation conveys unfavorable prognosis for the respective patient. The cutoff of 10 mg/ml was described in the literature. Therefore, any elevation of AFP above 10 mg/ml in gastric cancer patients was considered pathological irrespective of the individual value of each patient.18e21 Furthermore, direct comparisons to former analyses are not only hampered by different definitions of AFPGC but also by substantially different baseline characteristic of the respective patient cohorts. The most recent case series4 included patients in metastasized stages and patients undergoing palliative resections. In contrast, this analysis reports only on curatively resected patients without proof of metastases (R0, UICC-IV stages are derived from N3status in the 6th edition of TNM classification system7). This is also reflected by a higher FYSR of 78.7% in this analysis compared to former publications (FYSR 34e50%1e5). Previous publications consisted either of only case series or caseecontrol studies in which a selection bias may not be excluded.1e5 Here we introduced PSM which is a recognized method for balancing pronounced differences in baseline characteristics between two groups. PSM is a statistical method applied to reduce the possible selection bias in observational studies, which was initially introduced by Rubin and Rosenbaum in 1983.8 There are even hints that PSM may create results as if a randomized controlled trial would have been performed.10 The simple comparisons between the differently sized groups revealed marked differences between AFPP-GC and AFPN-GC patients. The AFPP-GC group revealed more advanced pT-stages, a higher rate of lymph node metastasis and a higher rate of lymphatic vessel infiltration. Survival analysis of the unmatched groups revealed dismal prognosis for AFPpositivity. Those marked differences in pathological findings might explain why survival rates were significantly lower in the AFP-positive group. This analysis
7
demonstrates that the PSM created well-balanced cohorts. Pathological tumor properties such as pT-/pN/-UICC stages, Lauren classification and LVI were comparable after the matching. However, survival analysis revealed that survival differences still exist between the two groups and multivariable regression analysis demonstrated that, besides the pathologic tumor stage, AFP positivity was a prognostic factor for OS and RFS. The liver as primary recurrence location in AFPP-GC was significantly predominant before and after PSM. Other studies showed similar results11,12 and concluded that predominantly occurring liver metastasis might be caused by early venous invasion in AFPPGC.11 There are several studies which investigated on the biologic properties of AFPP-GC.12e15 Compared to common gastric cancers AFPP-GC was reported to have a high proliferative, pro-angiogenetic and reduced apoptotic activity.12 It was described that the aggressive phenotype of AFPP-GC may be related to c-Met overexpression.14,15 The c-Met proto-oncogene encodes the c-Met receptor which is known to regulate cell proliferation or migration.14 Hepatocyte growth factor (HGF) has been identified as its ligand,16 is considered an important regulator of liver regeneration in response to injury and promotes epithelialcell mobilization.17 It was reported that AFPP-GC cells with higher c-Met expression behave more invasively in response to HGF.16 The higher recurrence rates were explained to be related to HGF-release during abdominal surgery, which might affect proliferation/migration of AFPP-GC cells.1 Therefore, meticulous surgery and reduction of perioperative trauma should be of highest interest for patients with AFPP-GC from the surgical perspective. Moreover, immunotherapy targeting c-Met and/or HGF could be a promising treatment of this clinically distinct entity. Further investigation for proteogenomic profiles underlying AFPP-GC is required to identify the tentative target. There are several limitations for this analysis: Although PSM is an accepted tool to overcome selection-bias, the data were analyzed retrospectively. Another limitation is that there is a huge disparity of the patient-numbers between AFP-positive and AFP-negative gastric cancer patients. Therefore no clear conclusions can be drawn on those many patients not having been included in the PSM-cohort. Moreover, PSM has some limitations which are the inability to account for unmeasured factors like surgical quality, biologic and genetic differences and the need for the statistical analysis to account for the paired nature of the matched samples, possibly explaining the remaining differences in outcome. Further, some patients with incomplete datasets had to be omitted from the analysis due to the reasons described above, which may have led to a certain kind of selection bias. Besides that, among the 3034 patients, 194 patients either had liver cirrhosis, or hepatitis, or both, and 14 of them were AFP-positive (AFP >10 mg/L), which may have affected the final analysis.
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005
8
D. Reim et al. / EJSO xx (2017) 1e8
However, none of the patients had HCC. Due to the fact that AFP is usually affected mostly by HCC, none of the patients were excluded from the analysis. In conclusion, this PSM analysis demonstrates that a preoperative elevation of AFP (>10 mg/L) can be considered as an independent poor prognostic predictor of OS and RFS in gastric cancer patients. It was shown that biological differences in common gastric cancers are in fact related to more aggressive behavior of this malignancy. Meticulous surgery with perioperative reduction of surgical trauma may improve prognosis for those patients. Due to high recurrence rates, more frequent follow-up visits and more aggressive adjuvant therapy including targeted immunotherapy should be considered. Further we propose and encourage over-regional co-operations to define a clear definition of AFPP-GC as a distinct clinical entity. Conflict of interest Statement This manuscript fully consists of original material, which has not been published or submitted to another journal. All authors declare that they participated in the analysis and that they have seen and approved the final version. None of the authors has any conflicts of interest. No funds or grants or company gifts have been received, nor has the article been written by a third party. Data acquisition and analysis was performed according to local ethical standards and GCP guidelines. Acknowledgements This work was supported by the National Cancer Center grant NCC-1410130-1. We thank Hyun Jung Park (Research Nurse) and Soo Hee Kim (Research Nurse) for the extraordinary support in data collection and preparation of data.
References 1. Kono K, Amemiya H, Sekikawa T, et al. Clinicopathologic features of gastric cancers producing alpha-fetoprotein. Dig Surg 2002;19(5): 359–65. discussion 365. 2. Inoue M, Sano T, Kuchiba A, Taniguchi H, Fukagawa T, Katai H. Long-term results of gastrectomy for alpha-fetoprotein-producing gastric cancer. Br J Surg Jul 2010;97(7):1056–61. 3. Liu X, Cheng Y, Sheng W, et al. Clinicopathologic features and prognostic factors in alpha-fetoprotein-producing gastric cancers: analysis of 104 cases. J Surg Oncol Sep 1 2010;102(3):249–55. 4. Hirajima S, Komatsu S, Ichikawa D, et al. Liver metastasis is the only independent prognostic factor in AFP-producing gastric cancer. World J Gastroenterol Sep 28 2013;19(36):6055–61. 5. Bourreille J, Metayer P, Sauger F, Matray F, Fondimare A. Existence of alpha feto protein during gastric-origin secondary cancer of the liver. Presse Med Jun 6 1970;78(28):1277–8.
6. Japanese Gastric Cancer Association. Japanese gastric cancer treatment guidelines 2010 (ver. 3). Gastric Cancer Jun 2011;14(2):113–23. 7. UICC. TNM classification of malignant tumours. In: Sobin LH, Wittekind C, editors. 6th ed. New York, NY: Wiley-Liss; 2002. 8. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika April 1, 1983; 70(1):41–55. 9. Ho D, Imai K, King G, Stuart E. Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference. Polit Anal 2007;15:199–236. 10. Kuss O, Legler T, Borgermann J. Treatments effects from randomized trials and propensity score analyses were similar in similar populations in an example from cardiac surgery. J Clin Epidemiol Oct 2011; 64(10):1076–84. 11. Chang YC, Nagasue N, Abe S, Taniura H, Kumar DD, Nakamura T. Comparison between the clinicopathologic features of AFP-positive and AFP-negative gastric cancers. Am J Gastroenterol Mar 1992; 87(3):321–5. 12. Koide N, Nishio A, Igarashi J, Kajikawa S, Adachi W, Amano J. Alpha-fetoprotein-producing gastric cancer: histochemical analysis of cell proliferation, apoptosis, and angiogenesis. Am J Gastroenterol Jun 1999;94(6):1658–63. 13. Ishigami S, Natsugoe S, Nakashima H, et al. Biological aggressiveness of alpha-fetoprotein (AFP)-positive gastric cancer. Hepatogastroenterology MayeJun 2006;53(69):338–41. 14. Amemiya H, Kono K, Mori Y, et al. High frequency of c-Met expression in gastric cancers producing alpha-fetoprotein. Oncology Aug 2000;59(2):145–51. 15. Kaji M, Yonemura Y, Harada S, Liu X, Terada I, Yamamoto H. Participation of c-met in the progression of human gastric cancers: anti-cmet oligonucleotides inhibit proliferation or invasiveness of gastric cancer cells. Cancer Gene Ther NoveDec 1996;3(6):393–404. 16. Matsumoto K, Nakamura T. Hepatocyte growth factor: molecular structure, roles in liver regeneration, and other biological functions. Crit Rev Oncog 1992;3(1e2):27–54. 17. Weidner KM, Behrens J, Vandekerckhove J, Birchmeier W. Scatter factor: molecular characteristics and effect on the invasiveness of epithelial cells. J Cell Biol Nov 1990;111(5 Pt 1):2097–108. 18. Akuta N, Suzuki F, Kawamura Y, et al. Substitution of amino acid 70 in the hepatitis C virus core region of genotype 1b is an important predictor of elevated alpha-fetoprotein in patients without hepatocellular carcinoma. J Med Virol [Internet] 2008 Aug;80(8):1354–62. 19. Ball D, Rose E, Alpert E. Alpha-fetoprotein levels in normal adults. Am J Med Sci [Internet] 1992 Mar;303(3):157–9. 20. Li P, Wang S-S, Liu H, et al. Elevated serum alpha fetoprotein levels promote pathological progression of hepatocellular carcinoma. World J Gastroenterol [Internet] 2011 Nov 7;17(41):4563–71. 21. Liu X, Cai H, Wang Y. Prognostic significance of tumor markers in T4a gastric cancer. World J Surg Oncol [Internet] 2012 Apr 27;10:68.
Abbreviations AFPN-GC: AFP-negative gastric cancer. AFPP-GC: AFP-positive gastric cancer. BMI: body mass index. FYSR: five year survival rate. HR: hazard ratio. ICH: immunohistochemistry. LVI: lymphatic vessel infiltration. OS: overall survival. PSM: propensity score matching. RFS: recurrence free survival. UICC: Union internationale contre le cancer.
Please cite this article in press as: Reim D, et al., Alpha-fetoprotein is a significant prognostic factor for gastric cancer: Results from a propensity score matching analysis after curative resection, Eur J Surg Oncol (2017), http://dx.doi.org/10.1016/j.ejso.2017.04.005