[Skip to Navigation]
Sign In
Figure 1.  Overall Survival Among Patients With Hepatoid Adenocarcinoma of the Stomach (HAS)
Overall Survival Among Patients With Hepatoid Adenocarcinoma of the Stomach (HAS)

pM indicates pathological metastasis; pN, pathological node; pT, pathological tumor; and pTNM, pathological tumor-node-metastasis.

Figure 2.  Nomogram Based on Results of Multivariable Cox Analysis of Derivation Cohort
Nomogram Based on Results of Multivariable Cox Analysis of Derivation Cohort

Points were assigned for pathological node category, perineural invasion, and carcinoembryonic antigen by drawing a line upward from the corresponding values to the Points line. The sum of these 3 points, plotted on the Total Points line, corresponds to estimates of the overall survival of patients with hepatoid adenocarcinoma of the stomach at 1 year and 3 years. CEA indicates carcinoembryonic antigen (to convert to micrograms per liter, multiply by 1.0); and pN, pathological node.

Figure 3.  Overall Survival of Low-Risk vs High-Risk Groups
Overall Survival of Low-Risk vs High-Risk Groups

B, P value derived from exact log-rank test.

Table 1.  Demographic, Clinical, Treatment, and Pathological Characteristics of Study Population
Demographic, Clinical, Treatment, and Pathological Characteristics of Study Population
Table 2.  Concordance Index and Akaike Information Criterion for Nomogram, pTNM Staging, and Clinical Models by Cohort
Concordance Index and Akaike Information Criterion for Nomogram, pTNM Staging, and Clinical Models by Cohort
Supplement.

eTable 1. Number of Patients With HAS at Each Center

eTable 2. Demographic and Clinicopathological Characteristics of Derivation and Validation Cohorts

eTable 3. Univariable and Multivariable Analyses of Risk Factors Associated With Overall Survival in Derivation Cohort Using Cox Mixed-Effects Model

eTable 4. Comparison of Concordance Index and Akaike Information Criterion of Nomogram, AJCC pTNM Staging System, and Clinical Model Between Validation Cohort (FMUUH) and Derivation Cohort (Other 13 Centers)

eTable 5. Comparison of Concordance Index and Akaike Information Criterion of Nomogram, AJCC pTNM Staging System, and Clinical Model Between Simple and Mixed HAS Groups

eTable 6. Demographic and Clinicopathological Characteristics of Low-Risk and High-Risk Groups

eFigure 1. Study Flow Diagram

eFigure 2. Histopathological Components of HAS

eFigure 3. Calibration Curves for Overall Survival Nomogram Models in Derivation, Validation, and Whole Cohorts

eFigure 4. Time-Dependent ROC Curve Comparing AUCs of Nomogram, AJCC pTNM Staging System, and Clinical Model in Derivation, Validation, and Whole Cohorts for Overall Survival of HAS

eFigure 5. Decision Curve Analysis of Each Model for Estimating 3-Year Overall Survival in Derivation, Validation, and Whole Cohorts

eFigure 6. Calibration Curves for Overall Survival Nomogram Models in Simple and Mixed HAS Groups

eFigure 7. Time-Dependent ROC Curve Comparing AUCs of Nomogram, AJCC pTNM Staging System, and Clinical Model in Simple and Mixed HAS Groups for Overall Survival

eFigure 8. Decision Curve Analysis of Each Model for Estimating 3-Year Overall Survival in Simple and Mixed HAS Groups

eFigure 9. Division of Patients by Cutoff Points Produced by X-Tile Plot

eFigure 10. Kaplan-Meier Analysis of Overall Survival Between Low-Risk and High-Risk Groups by Histological Type

1.
Su  JS, Chen  YT, Wang  RC, Wu  CY, Lee  SW, Lee  TY.  Clinicopathological characteristics in the differential diagnosis of hepatoid adenocarcinoma: a literature review.   World J Gastroenterol. 2013;19(3):321-327. doi:10.3748/wjg.v19.i3.321PubMedGoogle ScholarCrossref
2.
Ishikura  H, Kirimoto  K, Shamoto  M,  et al.  Hepatoid adenocarcinomas of the stomach: an analysis of seven cases.   Cancer. 1986;58(1):119-126. doi:10.1002/1097-0142(19860701)58:1<119::AID-CNCR2820580121>3.0.CO;2-UPubMedGoogle ScholarCrossref
3.
Zhang  JF, Shi  SS, Shao  YF, Zhang  HZ.  Clinicopathological and prognostic features of hepatoid adenocarcinoma of the stomach.   Chin Med J (Engl). 2011;124(10):1470-1476.PubMedGoogle Scholar
4.
Wang  Y, Sun  L, Li  Z,  et al. Hepatoid adenocarcinoma of the stomach: a unique subgroup with distinct clinicopathological and molecular features. Gastric Cancer. 2019;22(6):1183-1192. doi:10.1007/s10120-019-00965-5
5.
Zeng  XY, Yin  YP, Xiao  H,  et al.  Clinicopathological characteristics and prognosis of hepatoid adenocarcinoma of the stomach: evaluation of a pooled case series.   Curr Med Sci. 2018;38(6):1054-1061. doi:10.1007/s11596-018-1983-1PubMedGoogle ScholarCrossref
6.
Liu  X, Cheng  Y, Sheng  W,  et al.  Analysis of clinicopathologic features and prognostic factors in hepatoid adenocarcinoma of the stomach.   Am J Surg Pathol. 2010;34(10):1465-1471. doi:10.1097/PAS.0b013e3181f0a873PubMedGoogle ScholarCrossref
7.
Zhou  K, Wang  A, Ao  S,  et al.  The prognosis of hepatoid adenocarcinoma of the stomach: a propensity score–based analysis.   BMC Cancer. 2020;20(1):671. doi:10.1186/s12885-020-07031-9PubMedGoogle ScholarCrossref
8.
World Medical Association. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA. 2013;310(20):2191-2194. doi:10.1001/jama.2013.281053
9.
Cheon  SH, Rha  SY, Jeung  HC,  et al.  Survival benefit of combined curative resection of the stomach (D2 resection) and liver in gastric cancer patients with liver metastases.   Ann Oncol. 2008;19(6):1146-1153. doi:10.1093/annonc/mdn026PubMedGoogle ScholarCrossref
10.
Kakar S, Pawlik TM, Allen PJ, et al. AJCC Cancer Staging Manual. 8th ed. Springer-Verlag; 2017.
11.
Nagai  E, Ueyama  T, Yao  T, Tsuneyoshi  M.  Hepatoid adenocarcinoma of the stomach: a clinicopathologic and immunohistochemical analysis.   Cancer. 1993;72(6):1827-1835. doi:10.1002/1097-0142(19930915)72:6<1827::AID-CNCR2820720606>3.0.CO;2-8PubMedGoogle ScholarCrossref
12.
Schemper  M, Smith  TL.  A note on quantifying follow-up in studies of failure time.   Control Clin Trials. 1996;17(4):343-346. doi:10.1016/0197-2456(96)00075-XPubMedGoogle ScholarCrossref
13.
Gainor  JF, Rizvi  H, Jimenez Aguilar  E,  et al.  Clinical activity of programmed cell death 1 (PD-1) blockade in never, light, and heavy smokers with non–small-cell lung cancer and PD-L1 expression ≥50.   Ann Oncol. 2020;31(3):404-411. doi:10.1016/j.annonc.2019.11.015PubMedGoogle ScholarCrossref
14.
Bustoros  M, Sklavenitis-Pistofidis  R, Park  J,  et al.  Genomic profiling of smoldering multiple myeloma identifies patients at a high risk of disease progression.   J Clin Oncol. 2020;38(21):2380-2389. doi:10.1200/JCO.20.00437PubMedGoogle ScholarCrossref
15.
Harrell  FE  Jr, Califf  RM, Pryor  DB, Lee  KL, Rosati  RA.  Evaluating the yield of medical tests.   JAMA. 1982;247(18):2543-2546. doi:10.1001/jama.1982.03320430047030PubMedGoogle ScholarCrossref
16.
Akaike  H.  A new look at the statistical model identification.   IEEE Trans Automat Contr. 1974;19(6):716-723. doi:10.1109/TAC.1974.1100705Google ScholarCrossref
17.
Heagerty  PJ, Lumley  T, Pepe  MS.  Time-dependent ROC curves for censored survival data and a diagnostic marker.   Biometrics. 2000;56(2):337-344. doi:10.1111/j.0006-341X.2000.00337.xPubMedGoogle ScholarCrossref
18.
Vickers  AJ, Cronin  AM, Elkin  EB, Gonen  M.  Extensions to decision curve analysis, a novel method for evaluating diagnostic tests, prediction models and molecular markers.   BMC Med Inform Decis Mak. 2008;8:53. doi:10.1186/1472-6947-8-53PubMedGoogle ScholarCrossref
19.
Therneau  T. Mixed effects Cox models. Version 2.2-16. CRAN Repository, R Project. January 14, 2020. Accessed July 15, 2021. https://cran.r-project.org/web/packages/coxme/index.html
20.
Grambsch  PM, Therneau  TM.  Proportional hazards tests and diagnostics based on weighted residuals.   Biometrika. 1994;81(3):515-526. doi:10.1093/biomet/81.3.515Google ScholarCrossref
21.
X-tile. Version X. Rimm Lab, Yale School of Medicine; 2019. Accessed May 13, 2021. https://medicine.yale.edu/lab/rimm/research/software
22.
Ishikura  H, Fukasawa  Y, Ogasawara  K, Natori  T, Tsukada  Y, Aizawa  M.  An AFP-producing gastric carcinoma with features of hepatic differentiation: a case report.   Cancer. 1985;56(4):840-848. doi:10.1002/1097-0142(19850815)56:4<840::AID-CNCR2820560423>3.0.CO;2-EPubMedGoogle ScholarCrossref
23.
Xiao  C, Wu  F, Jiang  H,  et al.  Hepatoid adenocarcinoma of the stomach: nine case reports and treatment outcomes.   Oncol Lett. 2015;10(3):1605-1609. doi:10.3892/ol.2015.3430PubMedGoogle ScholarCrossref
24.
Liu  X, Sheng  W, Wang  Y.  An analysis of clinicopathological features and prognosis by comparing hepatoid adenocarcinoma of the stomach with AFP-producing gastric cancer.   J Surg Oncol. 2012;106(3):299-303. doi:10.1002/jso.23073PubMedGoogle ScholarCrossref
25.
Tsuruta  S, Ohishi  Y, Fujiwara  M,  et al.  Gastric hepatoid adenocarcinomas are a genetically heterogenous group; most tumors show chromosomal instability, but MSI tumors do exist.   Hum Pathol. 2019;88:27-38. doi:10.1016/j.humpath.2019.03.006PubMedGoogle ScholarCrossref
26.
Fakhruddin  N, Bahmad  HF, Aridi  T,  et al.  Hepatoid adenocarcinoma of the stomach: a challenging diagnostic and therapeutic disease through a case report and review of the literature.   Front Med (Lausanne). 2017;4:164. doi:10.3389/fmed.2017.00164PubMedGoogle ScholarCrossref
27.
Yang  J, Wang  R, Zhang  W, Zhuang  W, Wang  M, Tang  C.  Clinicopathological and prognostic characteristics of hepatoid adenocarcinoma of the stomach.   Gastroenterol Res Pract. 2014;2014:140587. doi:10.1155/2014/140587PubMedGoogle Scholar
28.
Lin  CY, Yeh  HC, Hsu  CM, Lin  WR, Chiu  CT.  Clinicopathological features of gastric hepatoid adenocarcinoma.   Biomed J. 2015;38(1):65-69. doi:10.4103/2319-4170.126860PubMedGoogle Scholar
29.
Markar  SR, Mikhail  S, Malietzis  G,  et al.  Influence of surgical resection of hepatic metastases from gastric adenocarcinoma on long-term survival: systematic review and pooled analysis.   Ann Surg. 2016;263(6):1092-1101. doi:10.1097/SLA.0000000000001542PubMedGoogle ScholarCrossref
30.
Japanese Gastric Cancer Association. Japanese gastric cancer treatment guidelines 2018 (5th edition). Gastric Cancer. 2021;24(1):1-21. doi:10.1007/s10120-020-01042-y
31.
Ajani  JA, D’Amico  TA, Almhanna  K,  et al.  Gastric cancer, version 3.2016, NCCN Clinical Practice Guidelines in Oncology.   J Natl Compr Canc Netw. 2016;14(10):1286-1312. doi:10.6004/jnccn.2016.0137PubMedGoogle ScholarCrossref
32.
Kamoshida  S, Suzuki  M, Sakurai  Y,  et al.  Expression of chemoresistance-related proteins in alpha-fetoprotein– producing adenocarcinoma of the digestive organs.   Oncol Rep. 2006;16(4):721-727.PubMedGoogle Scholar
33.
Hwang  JE, Hong  JY, Kim  JE,  et al.  Prognostic significance of the concomitant existence of lymphovascular and perineural invasion in locally advanced gastric cancer patients who underwent curative gastrectomy and adjuvant chemotherapy.   Jpn J Clin Oncol. 2015;45(6):541-546. doi:10.1093/jjco/hyv031PubMedGoogle Scholar
34.
Xia  Q, Chen  L.  [Cellular and molecular mechanisms of perineural invasion in gastric cancer].   Zhonghua Wei Chang Wai Ke Za Zhi. 2015;18(2):198-200.PubMedGoogle Scholar
35.
Liebig  C, Ayala  G, Wilks  JA, Berger  DH, Albo  D.  Perineural invasion in cancer: a review of the literature.   Cancer. 2009;115(15):3379-3391. doi:10.1002/cncr.24396PubMedGoogle ScholarCrossref
36.
Wang  Y, Li  J, Xia  Y,  et al.  Prognostic nomogram for intrahepatic cholangiocarcinoma after partial hepatectomy.   J Clin Oncol. 2013;31(9):1188-1195. doi:10.1200/JCO.2012.41.5984PubMedGoogle ScholarCrossref
37.
Konishi  T, Shimada  Y, Hsu  M,  et al.  Association of preoperative and postoperative serum carcinoembryonic antigen and colon cancer outcome.   JAMA Oncol. 2018;4(3):309-315. doi:10.1001/jamaoncol.2017.4420PubMedGoogle ScholarCrossref
38.
Kochi  M, Fujii  M, Kanamori  N,  et al. Evaluation of serum CEA and CA19-9 levels as prognostic factors in patients with gastric cancer. Gastric Cancer. 2000;3(4):177-186. doi:10.1007/pl00011715
39.
Kinoshita  T, Ohtsuka  T, Yotsukura  M,  et al. Prognostic impact of preoperative tumor marker levels and lymphovascular invasion in pathological stage I adenocarcinoma and squamous cell carcinoma of the lung. J Thorac Oncol. 2015;10(4):619-628. doi:10.1097/JTO.0000000000000480
Original Investigation
Oncology
October 5, 2021

Assessment of Clinicopathological Characteristics and Development of an Individualized Prognostic Model for Patients With Hepatoid Adenocarcinoma of the Stomach

Author Affiliations
  • 1Department of Gastric Surgery, Fujian Medical University Union Hospital, Fuzhou, China
  • 2Key Laboratory of Ministry of Education of Gastrointestinal Cancer, Fujian Medical University, Fuzhou, China
  • 3Department of Gastrointestinal Oncology Surgery, Xiamen Cancer Center, The First Affiliated Hospital of Xiamen University, Xiamen, China
  • 4Department of Gastrointestinal Surgery, Union Hospital, Tongji Medical College, Huazhong University of Science and Technology, Wuhan, China
  • 5Department of Gastrointestinal Surgery, Ren Ji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
  • 6Department of Gastrointestinal Surgery, First Hospital of Jilin University, Changchun, China
  • 7Department of Gastrointestinal Surgery, Shandong Provincial Hospital Affiliated to Shandong First Medical University, Jinan, China
  • 8The First Affiliated Hospital of Nanjing Medical University, Nanjing, China
  • 9Department of Gastrointestinal Surgery, The Second Affiliated Hospital of Fujian Medical University, Quanzhou, China
  • 10Gastrointestinal Surgery Unit 1, Teaching Hospital of Putian First Hospital of Fujian Medical University, Putian, China
  • 11Gastrointestinal Department of The Sixth Affiliated Hospital of Sun Yat-sen University, Guangzhou, China
  • 12General Surgery Department, The First Hospital of Lanzhou University, Lanzhou, China
  • 13Department of General Surgery, Zhongshan Hospital Affiliated with Xiamen University, Xiamen, China
  • 14Department of Gastrointestinal Surgery, Affiliated Quanzhou First Hospital to Fujian Medical University, Quanzhou, China
  • 15Department of Gastrointestinal Surgery, Longyan First Hospital Affiliated with Fujian Medical University, Longyan, China
JAMA Netw Open. 2021;4(10):e2128217. doi:10.1001/jamanetworkopen.2021.28217
Key Points

Question  What are the clinicopathological characteristics and prognoses of patients with hepatoid adenocarcinoma of the stomach (HAS), and what is an accurate and efficient method of predicting overall survival (OS) among these patients?

Findings  In this prognostic study of 315 patients with HAS, 3 risk factors (perineural invasion, preoperative carcinoembryonic antigen levels ≥5 ng/mL, and pathological node category 3b) were independently associated with worse OS among patients with HAS. Based on these risk factors, a nomogram to predict postoperative OS among patients with HAS was developed.

Meaning  This study found that, among patients with HAS, perineural invasion, preoperative carcinoembryonic antigen levels, and pathological node category 3b were independently associated with OS; an individualized nomogram that was developed to predict postoperative OS had good prognostic value and may be useful as a supplement to the current tumor-node-metastasis staging system.

Abstract

Importance  Few studies have examined the clinicopathological characteristics and prognoses of patients with hepatoid adenocarcinoma of the stomach (HAS).

Objective  To explore the clinicopathological characteristics and prognoses of patients with HAS and develop a nomogram to predict overall survival (OS).

Design, Setting, and Participants  This prognostic study involved a retrospective analysis of data from 315 patients who received a diagnosis of primary HAS between April 1, 2004, and December 31, 2019, at 14 centers in China.

Main Outcomes and Measures  OS and prognostic factors. Patients were randomly assigned to a derivation cohort (n = 220) and a validation cohort (n = 95). A nomogram was developed based on independent prognostic factors identified through a multivariable Cox mixed-effects model.

Results  Among 315 patients with HAS (mean [SD] age, 61.9 [10.2] years; 240 men [76.2%]), 137 patients had simple HAS (defined as the presence of histologically contained hepatoid differentiation areas only), and 178 patients had mixed HAS (defined as the presence of hepatoid differentiation areas plus common adenocarcinoma areas). Patients with simple HAS had a higher median preoperative α-fetoprotein level than those with mixed HAS (195.9 ng/mL vs 48.9 ng/mL, respectively; P < .001) and a higher rate of preoperative liver metastasis (23 of 137 patients [16.8%] vs 11 of 178 patients [6.2%]; P = .003). The 3-year OS rates of patients with simple vs mixed HAS were comparable (56.0% vs 60.0%; log-rank P = .98). A multivariable Cox analysis of the derivation cohort found that the presence of perineural invasion (hazard ratio [HR], 2.13; 95% CI, 1.27-3.55; P = .009), preoperative carcinoembryonic antigen levels of 5 ng/mL or greater (HR, 1.72; 95% CI, 1.08-2.74; P = .03), and pathological node category 3b (HR, 3.72; 95% CI, 1.34-10.32; P = .01) were independent risk factors for worse OS. Based on these factors, a nomogram to predict postoperative OS was developed. The concordance indices of the nomogram (derivation cohort: 0.72 [95% CI, 0.66-0.78]; validation cohort: 0.72 [95% CI, 0.63-0.81]; whole cohort: 0.71 [95% CI, 0.66-0.76]) were higher than those derived using the American Joint Committee on Cancer’s AJCC Cancer Staging Manual (8th edition) pathological tumor-node-metastasis (pTNM) staging system (derivation cohort: 0.63 [95% CI, 0.57-0.69]; validation cohort: 0.65 [95% CI, 0.56-0.75]; whole cohort: 0.64 [95% CI, 0.59-0.69]) and those derived using a clinical model that included pTNM stage and receipt of adjuvant chemotherapy (derivation cohort: 0.64 [95% CI, 0.58-0.69]; validation cohort: 0.65 [95% CI, 0.56-0.75]; whole cohort: 0.64 [95% CI, 0.59-0.69]). Based on the nomogram cutoff of 10 points, the whole cohort was divided into high-risk and low-risk groups. The 3-year OS rate of patients in the high-risk group was significantly lower than that of patients in the low-risk group (29.7% vs 75.9%, respectively; log-rank P < .001), and the 3-year prognosis of high-risk and low-risk groups could be further distinguished into pTNM stage I to II (33.3% vs 80.2%; exact log-rank P = .15), stage III (34.3% vs 71.3%; log-rank P < .001), and stage IV (15.5% vs 70.3%; log-rank P = .009).

Conclusions and Relevance  This study found that perineural invasion, preoperative carcinoembryonic antigen levels of 5 ng/mL or greater, and pathological node category 3b were independent risk factors associated with worse OS. An individualized nomogram was developed to predict OS among patients with HAS. This nomogram had good prognostic value and may be useful as a supplement to the current American Joint Committee on Cancer TNM staging system.

Introduction

Hepatoid adenocarcinoma of the stomach (HAS) is a unique subtype of gastric cancer (GC), accounting for 0.3% to 1.0% of GC cases.1 The pathological diagnosis of HAS is based on histological features similar to those of hepatocellular carcinoma.2 Based on the different proportions of hepatoid differentiation areas, cases can be classified as simple HAS, which is defined as the presence of histologically contained hepatoid differentiation areas only, or mixed HAS, which is defined as histologically contained hepatoid differentiation areas plus common adenocarcinoma areas.3,4

Given the low incidence of HAS, limited data exist regarding the clinicopathological characteristics and prognoses of patients with this cancer subtype. Most previous studies have reported that patients with HAS have an unfavorable prognosis and a low 5-year survival rate.1,5,6 Zhou et al7 recently compared the clinicopathological data of 75 patients with HAS and 722 patients with common GC and found differences between the 2 cancer types with regard to tumor invasion depth, human epidermal growth factor receptor levels, human epidermal growth factor receptor 2 expression, antigen Ki-67 levels, preoperative serum carcinoembryonic antigen (CEA) levels (to convert to micrograms per liter, multiply by 1.0), carbohydrate antigen 19-9 levels, and proportion of patients receiving neoadjuvant chemotherapy. The survival analysis after 1:2 propensity matching revealed that 3-year overall survival (OS) rates of patients with HAS compared with common GC were 79.9% vs 96.9%, respectively (log-rank P = .20).7 However, these studies1,5-7 were limited by small samples (≤75 participants) and may not accurately reflect the clinicopathological characteristics and prognoses of patients with HAS. An individualized prognostic model has not yet been developed. Therefore, the current study aimed to explore the clinicopathological characteristics and prognoses of those with HAS and to develop and validate an individualized prognostic nomogram to predict OS among patients with HAS from 14 centers in China.

Methods

This prognostic study was approved by the human ethics review committee of each participating center. All participants provided written informed consent to use and publish their deidentified medical data at the time of their first hospitalization. This study adhered to the Declaration of Helsinki8 and followed the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) reporting guideline

.

Patients and Definitions

We retrospectively analyzed the clinicopathological data of patients with HAS who were admitted to 14 centers affiliated with the China Hepatoid Adenocarcinoma of the Stomach Study Group between April 1, 2004, and December 31, 2019. Patients were included in the study if they had pathologically confirmed HAS (including both simple and mixed HAS). Patients were excluded if they had remnant GC (n = 5), other cancers (n = 2), and/or an unknown HAS subtype (n = 3). In total, 315 patients with primary HAS were included; of those, 137 patients had simple HAS, and 178 patients had mixed HAS (eTable 1 and eFigure 1 in the Supplement).

For patients without distant metastasis, curative surgery was defined as microscopic tumor-free margin (R0) resection of the primary tumor plus standard lymphadenectomy (limited dissection [D1 or D1-plus] or extended dissection [D2]). For patients with liver metastases, curative surgery was defined as R0 resection of the primary tumor and liver metastases combined with a standard D2 lymphadenectomy9; palliative surgery was considered for patients receiving any other type of resection. The resected specimens were carefully examined by 2 experienced pathologists in each center, and tumor staging was performed based on the AJCC Cancer Staging Manual (8th edition) tumor-node-metastasis (TNM) staging system for GC from the American Joint Committee on Cancer (AJCC).10 For patients with advanced HAS without contraindications, we recommended treatment with 6 to 8 cycles of postoperative adjuvant chemotherapy. Because no international guideline regarding adjuvant chemotherapy for the treatment of those with HAS currently exists, the chemotherapy regimen administered to patients in this study was the same as that used for the treatment of patients with common GC (ie, 5-fluorouracil–based chemotherapy).

Pathological Diagnosis

Pathological examination is the criterion standard for the diagnosis of HAS, and morphological features are the main basis for pathological diagnosis.11 No quantity requirement for the diagnosis of histological hepatoid differentiation was used, and some patients with focal differentiation could have also received a diagnosis of HAS.4 Most cases of HAS presented histologically contained hepatoid differentiation areas plus common adenocarcinoma areas, whereas some cases presented only hepatoid differentiation areas.3,4 In the areas of hepatoid differentiation, tumor cells were arranged in a trabecular manner or in solid nests separated by narrow fibrous stroma composed of sinusoid-like capillaries. The tumor cells were cuboidal or polygonal with eosinophilic or clear cytoplasm and occasionally contained hyaline globules. The nucleus was large and ovoid and contained 1 to 2 nucleoli.6

If histological examination identified only a hepatoid differentiation area, the HAS was defined as simple (eFigure 2 in the Supplement); if both hepatoid differentiation and common adenocarcinoma areas were identified (eFigure 2 in the Supplement), the HAS was defined as mixed. Immunohistochemical staining was helpful for the diagnosis of HAS. More than 90% of patients with HAS expressed α-fetoprotein (AFP) (to convert to micrograms per liter, multiply by 1.0),1 which is often positive or highly positive in the hepatoid differentiation area (eFigure 2 in the Supplement), whereas in the common adenocarcinoma area, AFP is generally negative (eFigure 2 in the Supplement). However, a positive AFP result on immunohistochemical staining is not necessary for the diagnosis of HAS.

Follow-up

The strategy for follow-up of patients with HAS was the same as that used for patients with common GC. Participants were followed up through outpatient visits and telephone calls. Outpatient visits included physical examination, laboratory examination (routine blood test, blood biochemical examination, and measurement of AFP, CEA, and carbohydrate antigen 19-9 levels), chest radiography, abdominal ultrasonography or computed tomography, and annual esophagogastroduodenoscopy examination. All patients were followed up once every 3 to 6 months during the first 2 years, once every 6 to 12 months during the following 3 to 5 years, and once every year thereafter. Overall survival was recorded from the date of surgery or pathological diagnosis to the date of last follow-up or death. Because patients from multiple centers were included in this study, the last follow-up date of each center was different. A median follow-up period of 35.3 months (IQR, 16.0-58.4 months) was calculated using the reverse Kaplan-Meier method.12-14

Construction and Validation of Nomogram

Patients were randomly assigned to a derivation cohort (220 patients) and a validation cohort (95 patients) through a random number list generated by SPSS software, version 22.0 (SPSS Statistics). In the derivation cohort, independent prognostic risk factors among patients with HAS were examined using univariable and multivariable Cox mixed-effects models.

Based on these risk factors, we used R software, version 4.0.3 (R Foundation for Statistical Computing), to generate a nomogram to predict OS at 1 and 3 years. Because the current clinical judgment of prognosis among patients with GC is mainly based on TNM stage and the receipt of postoperative adjuvant chemotherapy, we compared the prognostic ability of the nomogram with both the AJCC pathological TNM (pTNM) staging system and a clinical model that included pTNM stage and receipt of adjuvant chemotherapy among the different cohorts.

The performance evaluation of the nomogram included calibration and discrimination. Internal validation was performed using bootstrapping with the 1000 resampling method. Performance was further verified using the Harrell concordance index15 and the calibration curve generated by the validation cohort and the whole cohort. In addition, we calculated the Akaike information criterion between the 3 models (derivation, validation, and whole cohorts) to assess the risk of overfitting.16 Time-dependent receiver operating characteristic curves were used to compare the dynamic difference of the area under the curve between the 3 models.17 A decision curve analysis was performed to determine the clinical usefulness of the nomogram by quantifying the net benefit across a range of risk thresholds.18

Statistical Analysis

Statistical analysis was performed using SPSS statistical software, version 22.0 (SPSS Statistics), and R software, version 4.0.3 (R Foundation for Statistical Computing). Categorical data were summarized as frequencies with percentages, and statistical comparisons for categorical variables were performed using the χ2 or Fisher exact test. Continuous variables were reported as means with SDs for variables with normal distribution and medians with IQRs for variables without normal distribution. Between-group means were compared using a t test, and medians were compared using the Mann-Whitney U test. The Kaplan-Meier method was used to calculate time-specific survival probabilities. A log-rank test was used for statistical comparisons between the survival curves. To account for possible differences in patient characteristics and survival outcomes at different centers, a Cox mixed-effects model19 was used to identify independent prognostic factors associated with survival. Variables with P < .05 in the univariable analysis were included in the multivariable model. Proportional hazards were evaluated using a proportional hazards assumption test and a Schoenfeld residual test.20 All statistical tests were 2-sided with a significance threshold of P < .05. Each patient (excluding 5 of 315 patients who received biopsies only) was assigned a risk score based on the nomogram. X-tile software, version 3.6.1 (Rimm Lab, Yale School of Medicine),21 was used to examine the diagnostic thresholds for the high-risk and low-risk scores for all patients. Patients with risk scores higher and lower than the diagnostic threshold were classified into high-risk and low-risk groups, respectively.

Results
Clinicopathological Characteristics

Among 315 patients with HAS included in the study, the mean (SD) age was 61.9 (10.2) years; 240 patients (76.2%) were male, and 75 patients (23.8%) were female (Table 1). The mean (SD) tumor size was 5.4 (2.6) cm, and the median preoperative serum AFP value was 73.0 ng/mL (IQR, 4.2-798.3 ng/mL). A total of 139 of 223 patients (62.3%) had elevated serum AFP levels (≥20 ng/mL). Among all patients, 117 (37.1%) had gastric antrum tumors, and 34 (10.8%) had preoperative liver metastases. Of the 34 patients with liver metastases, 3 also had distant metastases (2 patients had pulmonary metastases, and 1 patient had pancreatic head metastases). Among 31 patients with synchronous liver metastases only, 26 (83.9%) received curative surgery. Of 315 total patients, 295 (93.7%) received curative surgery, 15 (4.8%) received palliative surgery, and 5 (1.6%) received biopsies. Lymphovascular invasion was noted in 179 patients (56.8%) and perineural invasion in 127 patients (40.3%).

Based on the AJCC pTNM staging system, 24 patients (7.6%) had early HAS (pathological tumor category 1 [pT1]), and 286 patients (90.8%) had advanced HAS (pT2-pT4), with 140 patients (44.4%) having pT4a and 23 patients (7.3%) having pT4b HAS. A total of 57 patients (18.1%) had pathological node-negative (pN0) cancer, and 253 patients (80.3%) had pathological node-positive (pN1-pN3) cancer. Comparison of the clinicopathological data of patients with simple vs mixed HAS revealed that preoperative AFP levels in the simple HAS group were higher than those of the mixed HAS group (median, 195.9 ng/mL [IQR, 8.2-1416.1 ng/mL] vs 48.9 ng/mL [IQR, 3.2-237.3 ng/mL], respectively; P < .001), and those with simple HAS vs mixed HAS had a higher rate of preoperative liver metastasis (23 of 137 patients [16.8%] vs 11 of 178 patients [6.2%]; P = .003). A total of 17 of 21 patients (81.0%) with simple HAS and synchronous liver metastasis received curative surgery. Patients with simple vs mixed HAS also had different pTNM staging, particularly regarding later stages of disease (eg, 33 of 137 patients [24.1%] with simple HAS vs 13 of 178 patients [7.3%] with mixed HAS had stage IV cancer; P = .004).

AJCC TNM Staging Survival Analysis

The 3-year OS rate in the total sample of 315 patients with HAS was 58.1% (Figure 1). No significant difference in the 3-year OS rate was found between the simple and mixed HAS groups (56.0% vs 60.0%, respectively; log-rank P = .98). Based on pT category, the 3-year OS rates were 73.0% for those with T1 cancer, 80.5% for those with T2 cancer, 65.3% for those with T3 cancer, 46.0% for those with T4a cancer, and 53.9% for those with T4b cancer (log-rank P < .001). Based on pN category, the 3-year OS rates were 83.1% for those with N0 cancer, 67.2% for those with N1 cancer, 55.1% for those with N2 cancer, 53.9% for those with N3a cancer, and 15.3% for those with N3b cancer (log-rank P < .001). Based on pathological metastasis (pM) category, the 3-year OS rates were 60.8% for those with M0 cancer and 35.9% for those with M1 cancer, which was not a statistically significant survival difference (log-rank P = .07). Based on AJCC pTNM stage, the 3-year OS rates were 88.5% for those with stage I cancer, 71.1% for those with stage II cancer, 52.1% for those with stage III cancer, and 35.9% for those with stage IV cancer (log-rank P < .001).

Prognostic Risk Factors and Nomogram Construction

To evaluate the nomogram, the total sample of 315 patients with HAS was randomly divided into a derivation cohort (220 patients [69.8%]) and a validation cohort (95 patients [30.2%]). The derivation and validation cohorts were balanced and comparable in terms of clinicopathological data (eg, mean [SD] age, 62.0 [10.5] years vs 61.7 [9.6] years; 168 men [76.4%] vs 72 men [75.8%]; 128 patients [58.2%] vs 51 patients [53.7%] with lymphovascular invasion; 88 patients [40.0%] vs 39 patients [41.1%] with perineural invasion, respectively) (eTable 2 in the Supplement). In the derivation cohort, the univariable Cox analysis revealed that the presence of lymphovascular invasion (hazard ratio [HR], 2.79; 95% CI, 1.64-4.75; P < .001), the presence of perineural invasion (HR, 2.88; 95% CI, 1.81-4.60; P < .001), a preoperative CEA level of 5 ng/mL or greater (HR, 2.33; 95% CI, 1.50-3.64; P < .001), 1 pT category (pT4a: HR, 3.57; 95% CI, 1.09-11.72; P = .04), and 3 pN categories (pN2: HR, 2.62 [95% CI, 1.12-6.15; P = .03]; pN3a: 3.46 [95% CI, 1.46-8.21; P = .005]; pN3b: HR, 7.78 [95% CI, 3.21-18.90; P < .001]) were all associated with OS among patients with HAS. The multivariable Cox analysis found that the presence of perineural invasion (HR, 2.13; 95% CI, 1.27-3.55; P = .009), a preoperative CEA level of 5 ng/mL or greater (HR, 1.72; 95% CI, 1.08-2.74; P = .03), and 1 pN category (pN3b: HR, 3.72; 95% CI, 1.34-10.32; P = .01) were independently associated with OS (eTable 3 in the Supplement). The variables associated with OS in the multivariable analysis were incorporated into the prognostic model to develop an individualized nomogram of OS at 1 and 3 years (Figure 2).

Nomogram Validation

The calibration curve of the nomogram model had good consistency in each cohort (eFigure 3 in the Supplement). Among all cohorts, the concordance indices of the nomogram (derivation cohort: 0.72 [95% CI, 0.66-0.78]; validation cohort: 0.72 [95% CI, 0.63-0.81]; whole cohort: 0.71 [95% CI, 0.66-0.76]) were higher than those of the AJCC pTNM staging system (derivation cohort: 0.63 [95% CI, 0.57-0.69]; validation cohort: 0.65 [95% CI, 0.56-0.75]; whole cohort: 0.64 [95% CI, 0.59-0.69]) and the clinical model (derivation cohort: 0.64 [95% CI, 0.58-0.69]; validation cohort: 0.65 [95% CI, 0.56-0.75]; whole cohort: 0.64 [95% CI, 0.59-0.69]) (Table 2). The difference in the concordance index of the nomogram in the derivation and whole cohorts was significantly higher than that of the AJCC pTNM staging system (derivation cohort: 0.09 [95% CI, 0.03-0.14; P = .004]; whole cohort: 0.08 [95% CI, 0.04-0.11; P < .001]) and the clinical model (derivation cohort: 0.08 [95% CI, 0.02-0.14; P = .009]; whole cohort: 0.07 [95% CI, 0.03-0.11; P < .001]). The nomogram had a lower Akaike information criterion score than both the AJCC pTNM staging system and the clinical model, independent of cohort (Table 2). For example, among the whole cohort, the Akaike information criterion score was 1071.65 for the nomogram, 1110.69 for the AJCC staging system, and 1112.53 for the clinical model.

The time-dependent receiver operating characteristic curve found that the nomogram predicted OS at each postoperative point better than the AJCC pTNM staging system and the clinical model, independent of cohort (eFigure 4 in the Supplement). A decision curve analysis evaluating the clinical usefulness of each model revealed that the nomogram had a higher net benefit than the other 2 models with the same risk thresholds, independent of cohort (eFigure 5 in the Supplement).

We also performed a sensitivity analysis in which 79 patients from 1 center (Fujian Medical University Union Hospital) were assigned to the validation cohort and 236 patients from the other 13 centers were assigned to the derivation cohort. The results revealed that the concordance indices of the nomogram among the derivation and validation cohorts (0.71 [95% CI, 0.65-0.77] and 0.74 [95% CI, 0.62-0.85], respectively) were higher than those of the AJCC pTNM staging system (0.63 [95% CI, 0.57-0.68] and 0.69 [95% CI, 0.57-0.81]) and the clinical model (0.62 [95% CI, 0.57-0.68] and 0.69 [95% CI, 0.57-0.81]). The nomogram had a lower Akaike information criterion score in both the derivation and validation cohorts (818.06 and 149.99, respectively) than the AJCC pTNM staging system (841.72 and 156.80) and the clinical model (843.72 and 158.80) (eTable 4 in the Supplement). An additional stratified analysis was performed based on different histological types, and the calibration curve of the nomogram model revealed good consistency between the simple and mixed HAS groups; the prognostic prediction performance of the nomogram was also better than that of the other 2 models in both the simple and mixed HAS groups (eTable 5 and eFigures 6-8 in the Supplement).

Risk Grouping and Nomogram Survival Analysis

After exclusion of 5 patients who received biopsies only, and without knowledge of patient pN category and perineural invasion status, we used the nomogram to assign risk scores to 310 patients, with scores ranging from 0 to 21 points. Based on the 3-year OS rate, the minimum P value of the log-rank χ2 test and cutoff values of the high-risk and low-risk groups were calculated using X-tile software. When the risk score was 10 points, the corresponding χ2 value was 49.23 (P < .001). Based on this cutoff value, patients with a risk score higher than 10 points were defined as high risk, and patients with a risk score lower than 10 points were defined as low risk (eFigure 9 in the Supplement). Compared with the tumors of patients in the low-risk group, those of patients in the high-risk group were larger and more often located in the upper stomach; had higher preoperative CEA levels, higher rates of total gastrectomy, and a larger number of metastatic lymph nodes; commonly presented with lymphovascular and perineural invasion; and were classified in a more advanced postoperative pathological stage (eTable 6 in the Supplement).

The 3-year OS rate of patients in the high-risk group was 29.7%, which was significantly lower than the 75.9% OS rate found among low-risk patients (P < .001) (Figure 3). An additional stratified analysis based on AJCC pTNM stage revealed that the 3-year OS rate of high-risk and low-risk patients could be distinguished as stage I to II (33.3% vs 80.2%, respectively; exact log-rank P = .15), stage III (34.3% vs 71.3%; log-rank P < .001), and stage IV (15.5% vs 70.3%; log-rank P = .009) (Figure 3). A stratified survival analysis based on histological type found that regardless of simple or mixed HAS status, the 3-year OS rate of patients in the high-risk group was significantly lower than that of the low-risk group (simple HAS: 21.0% vs 73.7%, respectively; log-rank P < .001; mixed HAS: 34.8% vs 77.9%; log-rank P < .001) (eFigure 10 in the Supplement).

Discussion

This prognostic study examined the clinicopathological characteristics and prognoses of patients with HAS, a distinct type of cancer with a low incidence. The general characteristics of HAS were first identified by Ishikura et al22 in 1985, and studies of HAS have been performed since then.23,24 Hepatoid adenocarcinoma of the stomach occurs more often in middle-aged and older men, is more commonly located in the gastric antrum, presents varying levels of elevated serum AFP, and is more likely to be accompanied by lymph node and liver metastasis.4,25,26 The mean (SD) age in our sample of 315 patients with HAS was 61.9 (10.2) years, the male to female ratio was approximately 3:1, gastric antrum tumors accounted for 37.1% of cases, and 81.6% of patients had lymph node metastases, consistent with previous reports.4,25,26 In addition, to our knowledge, this study was the first to compare the clinicopathological characteristics of patients with simple vs mixed HAS, finding that the 2 groups had similarities in most clinicopathological characteristics; however, patients with simple HAS had higher preoperative serum AFP levels and a higher rate of liver metastasis compared with those with mixed HAS. These findings suggest that monitoring of serum AFP levels and the performance of liver imaging may be more important for patients with simple HAS.

Previous studies have found that, among patients with HAS, the 1-year survival rate was 30% to 44%, the 3-year survival rate was 13% to 23%, and the 5-year survival rate was 9% to 20%.3,24,27,28 The 3-year OS rate in the current study was 58.1%, which is higher than the rates reported in most of those previous studies and may be associated with the fact that most patients in the current study received curative surgery. Similar to a previous study,3 we did not find a difference in OS between those with simple vs mixed HAS, suggesting that the hepatoid differentiation area proportion may not be an independent risk factor associated with HAS prognosis. Although patients with simple HAS had more liver metastases, 17 of 21 patients (81.0%) with simple HAS and synchronous liver metastasis in this study received curative surgery, suggesting that among those with HAS and liver metastasis only, the receipt of radical resection of primary and metastatic lesions may still improve the prognosis; however, further studies with larger samples and longer follow-up periods are needed to confirm this finding. In addition, although surgical resection of liver metastases in patients with GC has not been explicitly recommended by guidelines, a systematic review reported long-term survival benefits of surgical resection of liver metastases among those with GC.29 Therefore, it is likely that when liver metastases meet the technical requirements for resectability and when hepatectomy can be tolerated (as determined by liver reserve function assessment), patients with HAS and liver metastasis can be treated with R0 resection of the primary tumor and liver metastases combined with a standard D2 lymphadenectomy, which can achieve the purpose of curative surgery. Because of the lack of high-level evidence, among patients with HAS and other distant metastases, the best treatment options may be chemotherapy, radiotherapy, palliative care, or best supportive care, as recommended in current treatment guidelines for patients with GC.30,31

Although most HAS is accompanied by an increase in preoperative serum AFP levels, we found that elevated AFP (≥20 ng/mL) was not associated with prognosis among those with HAS, which is consistent with previous results3,27 and suggests that the diagnostic value of preoperative AFP level among patients with HAS may be greater than its prognostic value. However, because of the lack of data on postoperative serum AFP levels, we were unable to examine the association of postoperative AFP levels and the changes between preoperative and postoperative AFP levels with the prognosis of patients with HAS. In addition, we found no significant difference in survival between patients who did and did not receive adjuvant chemotherapy. Kamoshida et al32 found that patients with HAS lacked thymidine phosphorylase, which has an important role in the activation of fluorouracil but is high in dihydropyrimidine dehydrogenase, which is responsible for degrading fluorouracil. This factor may be one of the reasons for the insufficient response of HAS to traditional chemotherapy regimens. Therefore, further assessment of appropriate adjuvant treatment regimens for patients with HAS is warranted.

Previous studies have reported that, among patients with advanced GC, the presence of perineural invasion is associated with local recurrence and worse prognosis.33 Perineural invasion has been found to be an independent factor associated with GC metastasis in addition to direct invasion, lymphatic metastasis, hematological metastasis, and peritoneal metastasis.34,35 Our results also revealed that perineural invasion was an independent risk factor associated with worse prognosis in patients with HAS. In addition, an elevated preoperative serum CEA level has been reported to be independently associated with the prognosis of various cancers.36-39 We found that a preoperative CEA level of 5 ng/mL or greater was also independently associated with HAS. The prognosis of patients with elevated CEA levels was significantly worse than that of patients with normal CEA levels; therefore, clinical attention to CEA levels is warranted.

To better predict the long-term prognosis of patients with HAS and develop targeted treatment and follow-up strategies, we developed an individualized nomogram based on 3 factors independently associated with prognosis (perineural invasion, preoperative CEA level, and pN category 3b). To our knowledge, the current study is the first to use a nomogram to predict the OS of patients with HAS. The nomogram had good performance in the derivation, validation, and whole cohorts and was suitable for patients with both simple and mixed HAS. Furthermore, according to the risk scores generated by the nomogram for the classification of high-risk and low-risk patients, there was a significant difference in survival between the 2 groups, independent of histological classification. The prognosis of high-risk and low-risk patients could be further distinguished, especially in patients with stage III and IV cancer, suggesting that the nomogram developed in this study may be used as a supplement to the current AJCC TNM staging system.

Limitations

This study has several limitations. It is a multicenter study from China, with a relatively small number of patients with HAS in each center, and the results lack verification of data from Western populations. Nevertheless, the nomogram we developed is based on simple parameters that are easy to use. With the improvement of pathological diagnosis and the increase in clinicians’ understanding of HAS, we hope that, in the future, we can replicate our findings in centers with large samples and include data from Western countries to verify the applicability of the nomogram. In addition, this study uses a retrospective design, and the number of patients with HAS in each center was unevenly distributed; therefore, bias was inevitable. Moreover, to conduct a study among a large sample of patients with HAS, we retrospectively collected data from 14 centers. However, it is difficult for us to obtain complete data on patients with common GC, who comprise a substantially larger population than those with HAS. Therefore, we cannot compare the clinicopathological characteristics and prognoses of those with HAS vs common GC. Nevertheless, the main purpose of this study was to develop an individualized prognostic model for HAS and perform a comparative study of clinicopathological characteristics and prognoses among patients with simple vs mixed HAS, which has not, to our knowledge, been previously investigated.

Conclusions

This study combined HAS data from 14 centers in China to explore the clinicopathological characteristics and prognostic risk factors of patients with HAS. Consistent with previous studies, this study found that the preoperative serum AFP levels of those with HAS were elevated and that HAS easily metastasized to the liver. In addition, the current study compared patients with simple vs mixed HAS for the first time to date, finding that those with simple HAS had a significantly higher level of serum AFP and a higher rate of liver metastases than those with mixed HAS; however, the long-term prognoses of patients with the 2 types of HAS were similar. This study also developed an individualized nomogram based on risk factors that were independently associated with worse prognosis (perineural invasion, preoperative CEA levels of 5 ng/mL or greater, and pN category 3b); the nomogram had good performance in predicting OS among patients with HAS and may be useful as a supplement to the current AJCC TNM staging system.

Back to top
Article Information

Accepted for Publication: July 25, 2021.

Published: October 5, 2021. doi:10.1001/jamanetworkopen.2021.28217

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2021 Lin JX et al. JAMA Network Open.

Corresponding Authors: Chang-Ming Huang, MD (hcmlr2002@163.com) and Chao-Hui Zheng, MD, PhD (wwkzch@163.com), Department of Gastric Surgery, Fujian Medical University Union Hospital, No. 29, Xinquan Road, Fuzhou 350001 Fujian Province, China.

Author Contributions: Dr Huang had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs J.-X. Lin, Z.-K. Wang, Hong, P. Zhang, Z.-Z. Shang, and He contributed equally to this work and should be considered co–first authors.

Concept and design: J.-X. Lin, Z.-K. Wang, Hong, P. Zhang, Z.-Z. Shang, He, L.-J. Wang, Sun, Z.-X. Li, E.-D. Lin, Fu, S.-M. Lin, Xie, P. Li, Zheng, Huang.

Acquisition, analysis, or interpretation of data: Z.-K. Wang, Hong, P. Zhang, Z.-Z. Shang, He, Q. Wang, Shang, L.-J. Wang, Sun, Z.-X. Li, Liu, Ding, E.-D. Lin, Fu, S.-M. Lin.

Drafting of the manuscript: J.-X. Lin, Z.-K. Wang, Hong, P. Zhang, He, L.-J. Wang, Sun, Z.-X. Li, Liu, Fu, S.-M. Lin.

Critical revision of the manuscript for important intellectual content: Z.-K. Wang, Hong, P. Zhang, Z.-Z. Shang, He, Q. Wang, Shang, L.-J. Wang, Sun, Z.-X. Li, Ding, E.-D. Lin, Fu, S.-M. Lin, Xie, P. Li, Zheng, Huang.

Statistical analysis: J.-X. Lin, Z.-K. Wang, Hong, P. Zhang, Z.-Z. Shang, He, Shang, Sun, Z.-X. Li, E.-D. Lin, Fu, S.-M. Lin.

Obtained funding: Z.-Z. Shang, Fu, P. Li, Huang.

Administrative, technical, or material support: J.-X. Lin, Z.-Z. Shang, Q. Wang, L.-J. Wang, Liu, Fu.

Supervision: Hong, P. Zhang, Z.-Z. Shang, He, Sun, Z.-X. Li, Fu, S.-M. Lin, Zheng, Huang.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by grants 2017Y9011, 2018Y9008, and 2019Y9098 from Fujian Province Joint Funds for the Innovation of Science and Technology and grant 2016B013 from Fujian Province Innovation and Entrepreneurship Talents.

Role of the Funder/Sponsor: The funding organizations had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References
1.
Su  JS, Chen  YT, Wang  RC, Wu  CY, Lee  SW, Lee  TY.  Clinicopathological characteristics in the differential diagnosis of hepatoid adenocarcinoma: a literature review.   World J Gastroenterol. 2013;19(3):321-327. doi:10.3748/wjg.v19.i3.321PubMedGoogle ScholarCrossref
2.
Ishikura  H, Kirimoto  K, Shamoto  M,  et al.  Hepatoid adenocarcinomas of the stomach: an analysis of seven cases.   Cancer. 1986;58(1):119-126. doi:10.1002/1097-0142(19860701)58:1<119::AID-CNCR2820580121>3.0.CO;2-UPubMedGoogle ScholarCrossref
3.
Zhang  JF, Shi  SS, Shao  YF, Zhang  HZ.  Clinicopathological and prognostic features of hepatoid adenocarcinoma of the stomach.   Chin Med J (Engl). 2011;124(10):1470-1476.PubMedGoogle Scholar
4.
Wang  Y, Sun  L, Li  Z,  et al. Hepatoid adenocarcinoma of the stomach: a unique subgroup with distinct clinicopathological and molecular features. Gastric Cancer. 2019;22(6):1183-1192. doi:10.1007/s10120-019-00965-5
5.
Zeng  XY, Yin  YP, Xiao  H,  et al.  Clinicopathological characteristics and prognosis of hepatoid adenocarcinoma of the stomach: evaluation of a pooled case series.   Curr Med Sci. 2018;38(6):1054-1061. doi:10.1007/s11596-018-1983-1PubMedGoogle ScholarCrossref
6.
Liu  X, Cheng  Y, Sheng  W,  et al.  Analysis of clinicopathologic features and prognostic factors in hepatoid adenocarcinoma of the stomach.   Am J Surg Pathol. 2010;34(10):1465-1471. doi:10.1097/PAS.0b013e3181f0a873PubMedGoogle ScholarCrossref
7.
Zhou  K, Wang  A, Ao  S,  et al.  The prognosis of hepatoid adenocarcinoma of the stomach: a propensity score–based analysis.   BMC Cancer. 2020;20(1):671. doi:10.1186/s12885-020-07031-9PubMedGoogle ScholarCrossref
8.
World Medical Association. World Medical Association Declaration of Helsinki: ethical principles for medical research involving human subjects. JAMA. 2013;310(20):2191-2194. doi:10.1001/jama.2013.281053
9.
Cheon  SH, Rha  SY, Jeung  HC,  et al.  Survival benefit of combined curative resection of the stomach (D2 resection) and liver in gastric cancer patients with liver metastases.   Ann Oncol. 2008;19(6):1146-1153. doi:10.1093/annonc/mdn026PubMedGoogle ScholarCrossref
10.
Kakar S, Pawlik TM, Allen PJ, et al. AJCC Cancer Staging Manual. 8th ed. Springer-Verlag; 2017.
11.
Nagai  E, Ueyama  T, Yao  T, Tsuneyoshi  M.  Hepatoid adenocarcinoma of the stomach: a clinicopathologic and immunohistochemical analysis.   Cancer. 1993;72(6):1827-1835. doi:10.1002/1097-0142(19930915)72:6<1827::AID-CNCR2820720606>3.0.CO;2-8PubMedGoogle ScholarCrossref
12.
Schemper  M, Smith  TL.  A note on quantifying follow-up in studies of failure time.   Control Clin Trials. 1996;17(4):343-346. doi:10.1016/0197-2456(96)00075-XPubMedGoogle ScholarCrossref
13.
Gainor  JF, Rizvi  H, Jimenez Aguilar  E,  et al.  Clinical activity of programmed cell death 1 (PD-1) blockade in never, light, and heavy smokers with non–small-cell lung cancer and PD-L1 expression ≥50.   Ann Oncol. 2020;31(3):404-411. doi:10.1016/j.annonc.2019.11.015PubMedGoogle ScholarCrossref
14.
Bustoros  M, Sklavenitis-Pistofidis  R, Park  J,  et al.  Genomic profiling of smoldering multiple myeloma identifies patients at a high risk of disease progression.   J Clin Oncol. 2020;38(21):2380-2389. doi:10.1200/JCO.20.00437PubMedGoogle ScholarCrossref
15.
Harrell  FE  Jr, Califf  RM, Pryor  DB, Lee  KL, Rosati  RA.  Evaluating the yield of medical tests.   JAMA. 1982;247(18):2543-2546. doi:10.1001/jama.1982.03320430047030PubMedGoogle ScholarCrossref
16.
Akaike  H.  A new look at the statistical model identification.   IEEE Trans Automat Contr. 1974;19(6):716-723. doi:10.1109/TAC.1974.1100705Google ScholarCrossref
17.
Heagerty  PJ, Lumley  T, Pepe  MS.  Time-dependent ROC curves for censored survival data and a diagnostic marker.   Biometrics. 2000;56(2):337-344. doi:10.1111/j.0006-341X.2000.00337.xPubMedGoogle ScholarCrossref
18.
Vickers  AJ, Cronin  AM, Elkin  EB, Gonen  M.  Extensions to decision curve analysis, a novel method for evaluating diagnostic tests, prediction models and molecular markers.   BMC Med Inform Decis Mak. 2008;8:53. doi:10.1186/1472-6947-8-53PubMedGoogle ScholarCrossref
19.
Therneau  T. Mixed effects Cox models. Version 2.2-16. CRAN Repository, R Project. January 14, 2020. Accessed July 15, 2021. https://cran.r-project.org/web/packages/coxme/index.html
20.
Grambsch  PM, Therneau  TM.  Proportional hazards tests and diagnostics based on weighted residuals.   Biometrika. 1994;81(3):515-526. doi:10.1093/biomet/81.3.515Google ScholarCrossref
21.
X-tile. Version X. Rimm Lab, Yale School of Medicine; 2019. Accessed May 13, 2021. https://medicine.yale.edu/lab/rimm/research/software
22.
Ishikura  H, Fukasawa  Y, Ogasawara  K, Natori  T, Tsukada  Y, Aizawa  M.  An AFP-producing gastric carcinoma with features of hepatic differentiation: a case report.   Cancer. 1985;56(4):840-848. doi:10.1002/1097-0142(19850815)56:4<840::AID-CNCR2820560423>3.0.CO;2-EPubMedGoogle ScholarCrossref
23.
Xiao  C, Wu  F, Jiang  H,  et al.  Hepatoid adenocarcinoma of the stomach: nine case reports and treatment outcomes.   Oncol Lett. 2015;10(3):1605-1609. doi:10.3892/ol.2015.3430PubMedGoogle ScholarCrossref
24.
Liu  X, Sheng  W, Wang  Y.  An analysis of clinicopathological features and prognosis by comparing hepatoid adenocarcinoma of the stomach with AFP-producing gastric cancer.   J Surg Oncol. 2012;106(3):299-303. doi:10.1002/jso.23073PubMedGoogle ScholarCrossref
25.
Tsuruta  S, Ohishi  Y, Fujiwara  M,  et al.  Gastric hepatoid adenocarcinomas are a genetically heterogenous group; most tumors show chromosomal instability, but MSI tumors do exist.   Hum Pathol. 2019;88:27-38. doi:10.1016/j.humpath.2019.03.006PubMedGoogle ScholarCrossref
26.
Fakhruddin  N, Bahmad  HF, Aridi  T,  et al.  Hepatoid adenocarcinoma of the stomach: a challenging diagnostic and therapeutic disease through a case report and review of the literature.   Front Med (Lausanne). 2017;4:164. doi:10.3389/fmed.2017.00164PubMedGoogle ScholarCrossref
27.
Yang  J, Wang  R, Zhang  W, Zhuang  W, Wang  M, Tang  C.  Clinicopathological and prognostic characteristics of hepatoid adenocarcinoma of the stomach.   Gastroenterol Res Pract. 2014;2014:140587. doi:10.1155/2014/140587PubMedGoogle Scholar
28.
Lin  CY, Yeh  HC, Hsu  CM, Lin  WR, Chiu  CT.  Clinicopathological features of gastric hepatoid adenocarcinoma.   Biomed J. 2015;38(1):65-69. doi:10.4103/2319-4170.126860PubMedGoogle Scholar
29.
Markar  SR, Mikhail  S, Malietzis  G,  et al.  Influence of surgical resection of hepatic metastases from gastric adenocarcinoma on long-term survival: systematic review and pooled analysis.   Ann Surg. 2016;263(6):1092-1101. doi:10.1097/SLA.0000000000001542PubMedGoogle ScholarCrossref
30.
Japanese Gastric Cancer Association. Japanese gastric cancer treatment guidelines 2018 (5th edition). Gastric Cancer. 2021;24(1):1-21. doi:10.1007/s10120-020-01042-y
31.
Ajani  JA, D’Amico  TA, Almhanna  K,  et al.  Gastric cancer, version 3.2016, NCCN Clinical Practice Guidelines in Oncology.   J Natl Compr Canc Netw. 2016;14(10):1286-1312. doi:10.6004/jnccn.2016.0137PubMedGoogle ScholarCrossref
32.
Kamoshida  S, Suzuki  M, Sakurai  Y,  et al.  Expression of chemoresistance-related proteins in alpha-fetoprotein– producing adenocarcinoma of the digestive organs.   Oncol Rep. 2006;16(4):721-727.PubMedGoogle Scholar
33.
Hwang  JE, Hong  JY, Kim  JE,  et al.  Prognostic significance of the concomitant existence of lymphovascular and perineural invasion in locally advanced gastric cancer patients who underwent curative gastrectomy and adjuvant chemotherapy.   Jpn J Clin Oncol. 2015;45(6):541-546. doi:10.1093/jjco/hyv031PubMedGoogle Scholar
34.
Xia  Q, Chen  L.  [Cellular and molecular mechanisms of perineural invasion in gastric cancer].   Zhonghua Wei Chang Wai Ke Za Zhi. 2015;18(2):198-200.PubMedGoogle Scholar
35.
Liebig  C, Ayala  G, Wilks  JA, Berger  DH, Albo  D.  Perineural invasion in cancer: a review of the literature.   Cancer. 2009;115(15):3379-3391. doi:10.1002/cncr.24396PubMedGoogle ScholarCrossref
36.
Wang  Y, Li  J, Xia  Y,  et al.  Prognostic nomogram for intrahepatic cholangiocarcinoma after partial hepatectomy.   J Clin Oncol. 2013;31(9):1188-1195. doi:10.1200/JCO.2012.41.5984PubMedGoogle ScholarCrossref
37.
Konishi  T, Shimada  Y, Hsu  M,  et al.  Association of preoperative and postoperative serum carcinoembryonic antigen and colon cancer outcome.   JAMA Oncol. 2018;4(3):309-315. doi:10.1001/jamaoncol.2017.4420PubMedGoogle ScholarCrossref
38.
Kochi  M, Fujii  M, Kanamori  N,  et al. Evaluation of serum CEA and CA19-9 levels as prognostic factors in patients with gastric cancer. Gastric Cancer. 2000;3(4):177-186. doi:10.1007/pl00011715
39.
Kinoshita  T, Ohtsuka  T, Yotsukura  M,  et al. Prognostic impact of preoperative tumor marker levels and lymphovascular invasion in pathological stage I adenocarcinoma and squamous cell carcinoma of the lung. J Thorac Oncol. 2015;10(4):619-628. doi:10.1097/JTO.0000000000000480
×