Intestinal-type gastric cancer carries a better prognosis but a higher risk of lymph node metastasis than diffuse-type: a SEER study of nearly 28,000 patients

Background: The impact of Lauren type, namely intestinal and diffuse types, on prognosis and lymph node metastasis (LNM) for gastric cancer (GC), requires further exploration, since current samples are decient and the results are inconsistent.We aimed to translate the widely used WHO classications into Lauren types and analyze the impact with the largest sample size available. Methods: Corresponding Lauren types of the WHO classication system in the Surveillance, Epidemiology, and End Results (SEER) database were included to identify all patients with histological diagnoses of intestinal-type or diffuse-type GC. Results: 4,338 intestinal-type and 22,990 diffuse-type GC were included in our study. Compared with intestinal-type, diffuse-type had a relatively poor prognosis after adjustment for other risk factors (HR, 1.236; P<0.001). Similar to patients with all stages of GC, the prognosis of patients with early diffuse GC was also poor (HR, 1.295; P=0.001). Surgical operation and radiotherapy markedly improved the DSS for patients (P<0.05). To our surprise, the risk of intestinal LNM in all stages of GC was higher than that in the diffuse type (OR=0.891; P=0.036). Although, there was no positive correlation between LNM and Lauren’s GC (OR=1.096; P=0.467) in early stage. Conclusion: Intestinal-type GC carries a better prognosis but a higher risk of LNM than diffuse-type. Patients with intestinal-type GC are likely to gain a greater benet from receiving surgical treatment, but it is worth noting that clinicians should pay more attention to their lymphatic metastasis. *Contributed


Introduction
Gastric cancer (GC) is regarded as one of the common malignant tumours of the digestive system. The incidence of GC is different all over the world, and this trend is particularly evident in developing countries, especially in East Asia, where GC is highly prevalent [1] . At the moment, the diagnosis of GC mainly relies on imaging and pathology. There are different standards and controversy in the clinical staging and pathological diagnosis of GC. For the pathological classi cation of GC, the WHO and Lauren classi cations are most commonly used [2][3] . The WHO classi cation is more detailed and considers the degree of tumour differentiation, but it is limited by the different morphological manifestations of GC and the subjective assessments of pathologists, resulting in low reproducibility. Lauren classi cation is relatively simple and easy and has been favoured by researchers for many years [4][5] .
Lauren classi cation divides the histomorphology and cell characteristics of GC into diffuse-type, intestinal-type and mixed-type. Intestinal-type GC presents a highly differentiated adenocarcinoma structure, diffuse-type GC cells have gastric mucous cell characteristics and poor differentiation and show invasive growth and rare glandular cavity formation. The mixed type obtains the above two features, which combines multiple classi cations of GC biological behaviour and provides a reference for epidemiological characteristics [6][7] .
Lauren's is usually regarded as a great method of classi cation but there still exist different opinions.
Researchers have various attitudes to the difference of prognosis between intestinal and diffuse gastric cancer [8][9][10] . In addition, the sample size of the largest research is only 5593. There are lots of similarities between Lauren's method and WHO, that is, most of them is correspond with each other. In order to nd out the difference between them and deal with the contradiction, we attempt to combine them and then do some research. However, there is not a single conclusion due to the limited amount of cases.
The SEER database contains codes for both the Lauren and WHO classi cation. Therefore, we also brought corresponding intestinal and diffuse cases of the WHO classi cation system into the study to enlarge sample size, as to learn the clinical feature and prognosis of Lauren type GC.

Methods
We selected and analysed data from 27,328 patients with GC from the SEER database.
In both WHO and Lauren systems, the basic principle is to include GC with obvious glandular formation in the category of intestinal type. Cases in which there are isolated or small-stripe cancer cells scattered in the gastric wall during GC in ltration are categorized as diffuse. In addition to typical cases (intestinal and diffuse, ICD-O-3/WHO 2008 morphology codes 8144 and 8145, respectively), intestinal GC also includes papillary adenocarcinoma, highly differentiated tubular adenocarcinoma and medium differentiated tubular adenocarcinoma (ICD-O-3/WHO 2008 morphology codes 8260 and 8211, respectively). For the diffuse type of gastric carcinoma, signet ring cell carcinoma and poorly differentiated adenocarcinoma were also included (ICD-O-3/WHO 2008 morphology codes 8490 and 8140, respectively).
The selected primary tumour sites were classi ed as cardia, fundus of stomach, body of stomach, gastric antrum, pylorus, lesser curvature of stomach NOS, greater curvature of stomach NOS, overlapping lesion of stomach, stomach, or NOS (C16.0-16.9).
We selected patients with GC who met the above conditions between 2004 and 2013 from the SEER database, analysing the clinicopathological and prognostic variables.

Statistical analysis
We used the Chi-square test to compare the differences in clinicopathological characteristics between patients in two groups. Univariate and multivariate analysis of COX proportional risk regression model was carried out for both types. The survival curve was plotted by Kaplan-Meier method, and the differences among factors were calculated by Log-rank test. For the sake of adjusting potential confounding factors, we established a multivariate logistic model for the analysis of LNM. SPSS 21.0 (IBM Corporation) was used for statistical analysis. The statistical signi cance was determined as P < 0.05. Independent prognostic risk factors obtained by COX multivariate regression analysis were used to develop a Nomogram graph and calculate the index of concordance (C-index), which was utilized to evaluate the predictive performance. We build and verify the nomograms on statistical R software (3.3.1). P < 0.05 was considered statistically signi cant.

Incidence and trends
From 2004 to 2013, the age standardized incidence rate of every 100 thousand person years showed a distinct upward trend (Fig. 1). The incidence was signi cantly higher in men than in women when considering total GC cases or cases of intestinal or diffuse types. The age-adjusted incidence rates of the diffuse-type declined slowly over time. The incidence of intestinal-type GC remained the same.
Demographic and clinicopathological characteristics Table 1 provides a list of the main characteristics of the intestinal-type and diffuse-type of GCs. A total of 27,328 patients were brought into our study, including 4,338 intestinal-type and 22,990 diffuse-type. By comparing the clinicopathological features of different Lauren classi cation GC, it turned out that patients with intestinal-type and diffuse-type GC have signi cant differences in age, sex, tumour location, tumour size, lymph node metastasis, tumour stage, distant metastasis, and degree of tissue differentiation (P < 0.001).

Univariate and Multivariable Survival Analysis
To identify prognostic factors in the intestinal type of GC, we conducted univariate and multivariate Cox regression analysis. By univariate analysis, the intestinal-type showed that the factors in Table 2 were signi cantly different, except for sex and age, and the P values of the remaining factors were < 0.001. Multivariate analysis showed that age > 65, poorly differentiated, more advanced TNM stage, upper stomach tumour site, larger tumour size, T3, N3, no surgery, and no radiation were signi cantly associated with poorer DSS. But results revealed that M status (P = 0.761) and surgery type (P = 0.286) were not associated with DSS. The results of the multivariate analysis of diffuse-type GC are shown in Table 3. The variables included in the univariate analysis were age, race, pathological differentiation, TNM stage, tumour location, tumour size, regional node positivity, LN metastases, distant metastases, summary stage, history of surgery, surgery type and history of radiation. By multivariate analysis, signi cantly poorer survival were seen in age ≥ 65, African American, more advanced TNM stage, larger tumour size, N3, distant metastasis, were found to increase the risk of death for patients with diffuse-type GC. Furthermore, history of surgery was markedly bound up with a decreased hazard of DSS, and partial /subtotal surgery type and history of radiation. Interestingly, we found that T status was not signi cantly related to DSS. As shown in Fig. 2, the results showed that the survival rate of GC patients with intestinal-type disease was markedly higher than that of patients with diffuse-type(P < 0.001). Furthermore, the Kaplan-Meier analysis, as shown in Fig. 3, illustrated that patients with advanced-stage disease had signi cantly less survival time than that of patients with other diseases (P < 0.001). As the size of the tumour increased, the survival time of the patients similarly decreased. However, we also found that survival was worse for patients with T4, N3, and M1 disease than for patients with other stages of disease (P < 0.001) (Fig. 3, 4). Additionally, Fig. 5 displays the Kaplan-Meier analysis of DSS, which indicated that the survival outcome was improved by receiving radiation and undergoing surgery.

Forest plot
We performed forest plot to show the results of the multivariate Cox regression model (Table 4), Lauren classi cation, race, age, tumour grade, tumour size, invasion depth, regional lymph node depth, distant metastasis, surgery, radiation, primary tumour site, and local tumour destruction were all independent predictors in all stages of GC.
Furthermore, Table 5 showed that diffuse type, age ≥ 65, poorly differentiated, large tumour size, no history of surgery, and no history of radiation, were signi cantly related with less DSS bene ts in the early GC.
undifferentiated state, young age, upper stomach tumour site, deep invasion depth, and distant metastasis were obviously related toa higher hazard of LNM in GC.
To make further discussion about whether intestinal-type EGC revealed as high of a risk as diffuse-type EGC, we adjusted the potential confounding factors with multivariate logistic model to analyse the correlation between the histological type of EGC and the hazard of LNM. No positive association was found (P = 0.467) between Lauren type and LNM (Table 7). We also carried out a subgroup analysis, which demonstrated tumour size and depth of invasion were also considered as independent risk factors of LNM in early-stage GC.
Moreover, to make a prediction of the DSS of patients, the signi cant variables obtained from multivariate Cox regression model were used to generate nomogram (Figs. 6, 7). The C-index of the intestinal type was estimated to be 0.784, 95% CI = 0.774-0.794, while the C-index of the diffuse type was estimated to be 0.739, 95% CI = 0.735-0.743.

Discussion
Lauren classi cation is widely used in scienti c research, but there are still different opinions on this method. One reason is that the sample size is too small, with most of the cases included in the study range from several hundred to thousand, Therefore, there is no substantial progress owing to the less representativeness of these research results. Although Li et al. recently reported a study of Lauren classi cation in early GC based on the SEER database [11] , they included 5593 cases and only focused on early GC. In our study, we have a lot of data available with 28000 patients, so we can get the desired and more reliable results theoretically.
Analysis of the clinicopathological features showed that the age of onset of intestinal-type GC was older (≥ 65) (P < 0.001) than that of diffuse-type GC; additionally, when comparing intestinal-type GC with diffuse-type GC, there were more males affected than females (P < 0.001), there was a higher degree of pathological differentiation (P < 0.001), disease occurred at an earlier stage (P < 0.001), disease was more likely to be distributed in the lower stomach (P < 0.001), tumour size was more likely to be 4-8 cm (P < 0.001), there was more invasion into the muscularis propria (P < 0.001), there was less lymph node metastasis (P < 0.001), and there was less distant metastasis (P < 0.001). Compared with diffuse-type, the prognosis of intestinal-type was signi cantly better. Different from intestinal-type group, patients in diffuse-type group have relatively poor pathological differentiation (P < 0.001), and diffuse cases were often at an advanced stage when they were diagnosed, to be located in the upper stomach, and to have distant metastasis (P < 0.001) compared with intestinal cases.
LNM is often considered an important determinant of outcomes and treatment strategy choices in EGC [12][13] . While many studies have reported a high rate of diffuse lymph node metastasis, our study concluded that two groups didn't signi cantly differed in lymph node metastasis; Thus, the LNM risk of intestinal and diffuse GC is considered similar. Our ndings indicated that there was no positive correlation between LNM and Lauren type (OR = 1.096; 95% CI: 0.856-1.403; P = 0.467), but we demonstrated that the hazard of intestinal lymph node metastasis in all stages of GC was higher than that in the diffuse type (OR = 0.891; 95% CI, 0.801-0.992; P = 0.036).
The cause of these contradictory results may be the fact that the patients who basically register the information in the SEER database are mainly postoperative pathology. Moreover, the general surgery patients were mainly without metastasis and peripheral lymph node metastasis, while the proportion of distant metastasis in the surgical enrollment group was relatively small. In this study, the proportion of patients with intestinal type GC undergoing surgery was 68.07%, and the proportion of patients with diffuse GC undergoing surgery was 44.87%. As a result, we obtained this result when we estimated the risk of lymph node metastasis. However, theconclusion of this large sample of studies also remind us that we should pay more attention to lymphatic metastasis of intestinal type GC than before.
Another reason may be that our de nitions of intestinal and diffuse types differ from those of other studies. In addition to Lauren classi cation, we also used WHO classi cation. Intestinal GC also includes papillary adenocarcinoma, highly differentiated/moderately differentiated tubular adenocarcinoma, diffuse GC including signet-ring cell carcinoma, and low-differentiated adenocarcinoma. Different inclusion criteria may have resulted in inconsistent analysis results. Based on the conclusion that the hazard of lymph node metastasis in intestinal-type GC is high, clinicians must be cautious when choosing surgical treatment for intestinal-type GC.
The most obvious nding to emerge from the analysis is that the prognosis of diffuse-type GC was worse than that of intestinal-type GC. The median DSS time of intestinal-type patients was 29.79 months, compared with 10.41 months in diffuse-type patients, and the 5-year survival DSS rate was 42.8% and 21.4%, respectively. These ndings are consistent with previous reports [14] and may be partially explained by the tendency of the diffuse-type to present at more advanced T and N stages.
Previous research ndings related to GC are inconsistent and contradictory with our results, and the studies have showed that the prognosis of intestinal-type GC is worse than that of diffuse-type GC. In particular, studies by Li et.al [11] , which are based on the SEER database, analysed only early GC patients. The results showed that patients of different GC subtypes had similar prognoses, or even that the prognosis of intestinal-type GC was worse than that of diffuse-type GC. Therefore, we further analysed the differences in prognosis between patients with early-stage GC and diffuse-type GC. it was concluded that the prognosis of diffuse-type GC was worse than that of intestinal-type GC in both early stage and all-stage GC populations.
Most studies have concluded that intestinal-type GC is common in older men, has a low degree of invasion, has a high level of lymph node metastases, and has vascular invasion [15] . Diffuse GC is common in young and female patients and has serous invasion; additionally, lymph node metastasis occurs very early [16][17][18][19] .
In intestinal-type GC, the median DSS time decreased according to tumour in ltration depth (T1, T2, T3, and T4), indicating that an increase in tumour in ltration depth was negatively correlated with the prognosis of patients. Analysis of lymph node metastasis revealed that patient survival time decreased with the increase of the number of lymph node metastasis. The survival time of patients with distant metastases was evidently shorter than that of patients without distant metastases. The relationship between tumour invasion depth, lymph node metastasis, distant metastasis and prognosis in diffuse GC also showed the same trends. Studies have shown that lymph node metastasis is an independent factor in the prognosis of patients [20][21] . Patients with node-negative disease have a higher 5-year survival rate than that of patients with lymph node metastasis, and the survival time of patients with lymph node metastasis decreases with an increase in the number of metastatic lymph nodes. The larger extent of metastasis from lymph nodes in diffuse GC than intestinal GC suggests that the prognosis of diffuse GC is worse than that of intestinal-type GC. Invasive depth of diffuse patients The proportion of GC patients with T3 and T4 disease in diffuse-type is higher than that in intestinal-type.
Previous researches have shown that the depth of invasion is an independent factor in the prognosis of GC [22][23][24] . According to our study, the prognosis of diffuse GC is worse than that of intestinal-type GC. In patients with diffuse TNM, the proportion of patients with stage I and II disease was lower than that of patients with intestinal-type disease (P = 0.004), while the percentage of patients with stage IV was signi cantly higher than that of patients with intestinal-type disease (P = 0.008). Previous studies have shown that patients with stage III and IV GC have poor prognosis, and the 5-year survival rate is signi cantly lower in these patients than in patients with stage I and II disease [25][26] . The prognosis of diffuse GC is worse than that of intestinal-type GC.
According to follow-up statistics, the difference in survival rate 5 years after surgery for intestinal GC and diffuse GC was statistically signi cant (P < 0.05), and the survival rate of diffuse GC was lower than that of intestinal GC after 5 years (see Fig. 2). The 5-year survival rate of patients receiving surgery was signi cantly higher than that of patients without surgery P < 0.05) (see Fig. 5A). In patients with diffuse GC, there was a statistically signi cant difference between the 5-year survival rate of those receiving radiotherapy and the 5-year survival rate of patients who did not receive radiotherapy (P > 0.05), and the 5-year survival rate of patients receiving radiotherapy was higher than that of patients who did not receive radiotherapy (see Fig. 5B).
We found that there was a recent article that was also based on the SEER database and Lauren classi cation of GC, but its conclusions were very inconsistent [11] . This article suggests that diffuse-type has a better prognosis than intestinal-type GC. The results also showed that diffuse types occur in younger female patients and tend to occur in upper gastric areas, with smaller sizes and more lymph nodes involved than those in intestinal types. Table 1 detailed the clinical characteristics of both types, however, the differences between them may be due to several reasons. Firstly, early GC was included in this study, it meant all stages of GC were contained. Secondly, in addition to the intestinal and diffuse coding, the coding of the WHO classi cation was also included in this research.
At the same time, we should admit that the study has some shortcomings. First, some important clinical information is missing from the SEER database, such as chemotherapy used, omics data for bioinformatics analysis, and the data of modern prognostic factors such HER-2 status were not complete in SEER. Hence, a few potential prognostic factors cannot be applied into our study. In addition, a majority of cases were not updated to the newest AJCC staging system in the database. Our results suggested that diffuse-type carried a worse prognosis but a lower risk of LNM than intestinal-type. It is remarkable that clinicians should take their lymphatic metastasis seriously.

Declarations
Authors' contribution: Chanqiong Zhang and Guangyu Chen: data collection, writing-original draft; Haofeng Hong and Jinbo Zhu: analyzing data; Hongyuan Zhang, Chongan Huang,and Dan Pan: statistical analysis and making gures; Huajun Ye: study design, writing-review and editing. All authors read and approved the nal manuscript.  Nomograms predicting 1-, 3-, 5-, and 8-year DSS of patients with diffuse type GC.