Nomogram Prediction for Gastric Cancer Development

Abstract

INTRODUCTION:

Helicobacter pylori (Hp) and gastric atrophy represent significant risk factors for gastric cancer (GC). Nevertheless, to date, no nomogram has been developed to predict GC based on the specific combination of risk factors present in individual cases.

METHODS:

A retrospective cohort study was conducted using health screening data collected between 2003 and 2018. Subjects with positive results for anti-Hp antibody were enrolled. Individuals were classified into 4 groups: low-B (low titer without atrophy), high-B (high titer without atrophy), high-C (high titer with atrophy), and low-C (low titer with atrophy). Nomogram prediction models were developed for overall GCs as well as intestinal and diffuse cancers, with each type considered a competing event, by using both Cox proportional and subdistribution hazard models. Prediction performance was evaluated using the concordance index (c-index) and the area under the curve through 10-fold cross-validation.

RESULTS:

During a median follow-up period of 5.7 years, 231 new GC cases developed among the total cohort of 28,311 subjects, including 159 intestinal type, 68 diffuse type, and 4 cases of unknown type. Multivariable analyses indicated that age, body mass index, family history, smoking, and classification into the high-C or low-C group were significant predictors of GC. The nomograms for intestinal type, diffuse type, and total GC demonstrated area under the curve values of 0.82, 0.62, and 0.75, respectively, and c-indices of 0.85, 0.54, and 0.76, respectively.

DISCUSSION:

The nomograms for GC prediction would be useful in identifying high-risk individuals, particularly for intestinal type. This would facilitate the implementation of personalized eradication and intensive screening strategies to target those at higher risk for GC.

INTRODUCTION

Gastric cancer (GC) is the fifth most common cancer worldwide and the fifth most frequent cause of cancer-related death (1). GC incidence is particularly elevated in East Asia. Notably, early GC has relatively favorable prognosis, with a 5-year survival rate exceeding 95%, whereas that of advanced disease is only 63%. Furthermore, less invasive treatment can be applied for early GC, emphasizing the importance of early diagnosis (2).

Helicobacter pylori (Hp), a Gram-negative bacillus colonizing the gastric mucosa, is the single most significant risk factor for GC (3,4). It is well established that Hp can induce gastric adenocarcinoma through the stages of atrophic gastritis, intestinal metaplasia, and dysplasia, through chronic inflammation (5–7). Meanwhile, once atrophic gastritis develops, it can deplete the normal epithelium, which, in turn, can induce self-regression of Hp (8,9). Thus, those with severe atrophy are likely to carry a lower Hp load. Therefore, to investigate the risk of GC, it is necessary to take the atrophy into account in combination with the Hp load.

A previous study has indicated that diffuse type GC (DTGC) is associated with current or recent Hp infection, contrary to intestinal type gastric cancer (ITGC) which arises from a prolonged past Hp infection (10). Another retrospective study demonstrated that DTGC exhibited a higher anti-Hp antibody titer compared with ITGC (11). Several studies have reported that higher titers in individuals without atrophy are associated with a higher risk of DTGC, whereas individuals with lower titers in the presence of atrophy are at a higher risk of ITGC (12,13). Until now, the Hp serology test has been regarded as a categorical variable that can indicate whether a patient is infected or not. However, given the above findings, the antibody titer should be considered a predictor of GC risk. In addition, the combination of atrophy and the antibody titer may also be used to predict GC risk. To date, there have been no prediction models including the combination of risk factors. This study was designed to develop a nomogram to predict GC risk according to the individual combination of risk factors, including anti-Hp antibody titer and gastric atrophy.

METHODS

Study population

We retrospectively reviewed prospectively collected data from the cohort who underwent screening endoscopy and Hp serology test in the Seoul National University Hospital Gangnam Center between 2003 and 2018. Among a total of 122,892 individuals, those who are 18 years or older with a positive result for the Hp serology were enrolled. From the total of 61,238 enrolled, 31,277 subjects with <1-year of follow-up duration, 1,508 subjects of non-Korean ethnicity, 131 subjects who had undergone previous gastrectomy, 8 subjects who had undergone endoscopic resection, 1 subject previously diagnosed with GC who was under observation, and 1 subject who developed metastatic GC of another origin during the follow-up were excluded. Thus, the remaining 28,311 individuals were included in this study (Figure 1). During the study period, the date of the initial screening was designated as the index date. The baseline characteristics of the individuals were obtained from the index date. The follow-up duration was calculated from the index date until the occurrence of newly developed GC, death, the last follow-up visit, or the end of the study date, whichever came first. Participants with atrophic gastritis were recommended to undergo endoscopy annually, and those without atrophic gastritis were recommended every other year.

Figure 1.

Study flow diagram showing patient enrolment.

This study was approved by the Institutional Review Board of the Seoul National University Hospital (IRB No. H-2209-024-1355). This study was conducted in accordance with the tenets of the 1964 Declaration of Helsinki. Informed consent was waived because of the retrospective nature of the study; however, this study did not include any identifiable personal information, and the waiver of informed consent was approved by the Institutional Review Board.

Clinical parameters and biochemical analysis

Information on previous gastric surgery or endoscopic resection, or previously diagnosed of gastric neoplasm, family history, and health behaviors was obtained from a self-reported questionnaire, and in case of unclear data, it was clarified by oral questioning conducted by well-trained interviewers. From this information, we extracted following details: history of GC diagnosis, history of gastric surgery or endoscopic resection, family history of GC in first-degree relatives, smoking history (never/former/current), and alcohol consumption (none/<30 g/d/≥30 g/d). The body mass index (BMI) was calculated using the measurement at the index date. Anti-Hp antibody titer was examined through the HPG kit (Immulite 2000 CMIA; Siemens, Munich, Germany), which uses a chemiluminescent enzyme immunoassay with 91% of sensitivity and 100% of specificity (14). Levels below 0.90 IU/mL were considered negative and excluded from this study. The degree of gastric atrophy was measured using the Kimura-Takemoto classification, which was designed for the categorization of endoscopic gastric atrophy (15), by 2 expert endoscopists (J.H.L. and S.J.C.) with more than 10,000 cases of endoscopy under mutual discussion. As there was no available tissue sample preserved from this population, histological evaluations using Sydney classification or Operative Link on Gastritis Assessment could not be made.

Pathology was reported following the World Health Organization criteria (16), and GC cases were classified into ITGC or DTGC according to the Lauren classification (17). When both types were mixed, the predominant histology was taken.

Statistical analysis

Participants' baseline characteristics were summarized using descriptive statistics. t tests and χ² tests were used to compare the differences between population groups.

Endoscopic results and anti-Hp antibody titers were converted into binary categories according to clinical opinion (0-C2 [no atrophy; B group] and C3-O3 [atrophy; C group]) and according to a statistical cutoff point (<5.21514 [low], ≥5.21514 [high]), respectively. The B and C groups were named according to the conventional “ABC method” (18), which will be discussed in detail in the Discussion section. The cutoff point was taken as the value showing the largest hazard ratio (HR) in the univariate analysis. Two factors were then combined into 4 categories as low B, low C, high B, and high C for the ABC group (Figure 2).

Figure 2.

Group classification according to the ABC method.

The subdistribution hazard ratio (SHR) of ITGC and DTGC was estimated and presented with the 95% confidence intervals (CIs) using the competing risk model of Fine and Gray (19), considering the 2 types of cancer as competing events. The HR of total GC was calculated using Cox proportional hazard regression analysis (20). Age, sex, BMI, family history, smoking, alcohol, and ABC group at the baseline were selected as risk factors to be considered in the prediction model, and corresponding nomograms were constructed.

The predictive performance of the nomogram was assessed by the concordance index (c-index) and area under the curve (AUC) values through 10-fold cross-validation by randomly partitioning the data set into 10 equal-sized subsamples. A calibration curve at 15 years for each model was also presented graphically by plotting the observed risks against the predicted risks for validation.

All analyses were performed using R version 4.1.3, the R packages.

RESULTS

Baseline characteristics

Table 1 presents the baseline characteristics of our study population both overall and by each type of GC. Overall, the mean age at enrollment was approximately 49 years, but the age at cancer diagnosis was higher in ITGC compared with DTGC (63.1 vs 56.6, P < 0.001). Both types of GC showed male predominance, but it was greater in ITGC (80.5% vs 57.4%, P < 0.001). BMI (24.6 vs 24.0, P < 0.176) and family history of GC (19.4% vs 25.0%, P = 0.374) were not significantly different between the 2 types, although DTGC had a higher proportion with a family history. The proportion of current smokers was greater in ITGC (28.8% vs 22.1%, P = 0.044), but alcohol consumption habits were not statistically different between the 2 groups (P = 0.619).

Table 1.

Baseline characteristics

	Overall (n = 28,311)	Gastric cancer (n = 231)	P value (normal vs gastric cancer)	Intestinal type (n = 159)	Diffuse type (n = 68)	P value (intestinal vs diffuse)
Age (yr)	49.3 ± 10.0	54.7 ± 9.4	<0.001	56.3 ± 9.2	50.8 ± 8.7	<0.001
Age at diagnosis (yr)	NA	61.2 ± 10.0		63.1 ± 9.4	56.6 ± 9.9	<0.001
Sex			<0.001			<0.001
Female	12,444 (44.0)	60 (26.0)		31 (19.5)	29 (42.6)
Male	15,867 (56.0)	171 (74.0)		128 (80.5)	39 (57.4)
Body mass index (kg/m2)	23.5 ± 3.0	24.4 ± 3.0	<0.001	24.6 ± 2.8	24.0 ± 3.5	0.176
Family history of gastric cancer			0.001			0.374
No	24,279 (86.9)	179 (79.2)		125 (80.6)	51 (75.0)
Yes	3,646 (13.1)	47 (20.8)		30 (19.4)	17 (25.0)
Smoking habit			<0.001			0.044
Never	14,962 (53.3)	80 (35.2)		48 (30.8)	32 (47.1)
Former	7,349 (26.2)	85 (37.4)		63 (40.4)	21 (30.9)
Current	5,763 (20.5)	62 (27.3)		45 (28.8)	15 (22.1)
Alcohol consumption			0.026			0.619
None	8,155 (29.1)	59 (25.9)		40 (25.5)	19 (27.9)
Social (<30 g/d)	16,438 (58.7)	128 (56.1)		88 (56.1)	38 (55.9)
Heavy (≥30 g/d)	3,401 (12.1)	41 (18.0)		29 (18.5)	11 (16.2)
Anti-Hp antibody titer (IU/mL)	5.0 ± 2.4	5.1 ± 2.2	0.764	4.8 ± 2.2	5.8 ± 2.2	0.003
Atrophic gastritis ≥C3	3,052 (10.8)	117 (50.6)	<0.001	108 (67.9)	7 (10.3)	<0.001
Follow-up duration (mo)	81.1 ± 52.4	78.9 ± 46.9	0.296	82.2 ± 47.2	71.1 ± 44.2	0.100

ontinuous variables are mean ± SD; categorical variables are n (%).

Bold entries represent statistical significance.

Hp, Helicobacter pylori.

Survival analysis

A total of 231 participants developed GC during the study period, including 159 with ITGC, 68 with DTGC, and 4 type-unknown. In the hazard analysis, those with type-unknown GC were censored.

Table 2 presents the results of survival modelling for GC associated with each factor. For GC overall, the following factors were identified as independent risk factors in the Cox regression model: age (HR 1.596, 95% CI 2.235–1.811), BMI (HR 1.259, 95% CI 1.047–1.515), family history of GC (HR 1.474, 95% CI 1.068–2.036), current smoking (HR 1.758, 95% CI 1.140–2.710), and the high C (HR 6.681, 95% CI 4.435–10.063) and low C (HR 7.161, 95% CI 4.898–10.468) groups. A similar pattern was observed for ITGC in the results from the subdistribution hazard models, except for the family history of GC (SHR 1.250, 95% CI 0.839–1.870), with age (SHR 1.040, 95% CI 1.022–1.060), BMI (SHR 1.070, 95% CI 1.012–1.130), current smoking (SHR 1.780, 95% CI 1.070–2.960), and high C group (SHR 10.560, 95% CI 6.428–17.330) and low C group (SHR 11.550, 95% CI 7.262–18.360) as risk factors. Meanwhile, family history (SHR 2.091, 95% CI 1.215–3.600) and high B group (SHR 1.712, 95% CI 1.031–2.840) were found to be associated with DTGC.

Table 2.

Adjusted hazard ratios for each type of gastric cancer by the Cox proportional hazard model

	Gastric cancer (n = 227)a		Intestinal type (n = 159)		Diffuse type (n = 68)
	aHR	95% CI	aHR	95% CI	aHR	95% CI
Age (continuous, yr)	1.496	1.235–1.811	1.040	1.022–1.060	1.018	0.993–1.040
Sex (male over female)	1.009	0.663–1.536	1.300	0.758–2.220	0.633	0.359–1.120
Body mass index (continuous, kg/m2)	1.259	1.047–1.515	1.070	1.012–1.130	1.056	0.963–1.160
Family history of gastric cancer	1.474	1.068–2.036	1.250	0.839–1.870	2.091	2.091–3.600
Smoking habit
Never	1	Ref	1	Ref	1	Ref
Former	1.373	0.918–2.053	1.320	0.822–2.130	1.492	0.810–2.750
Current	1.758	1.140–2.710	1.780	1.070–2.960	1.690	0.803–3.560
Heavy alcohol consumption (≥30 g/d)
No	1	Ref	1	Ref	1	Ref
Yes	1.320	0.919–1.896	1.280	0.840–1.950	1.450	0.684–3.080
ABC group
Low B	1	Ref	1	Ref	1	Ref
High B	1.255	0.863–1.827	0.840	0.475–1.490	1.712	1.031–2.840
High C	6.681	4.435–10.063	10.560	6.428–17.330	1.383	0.469–4.070
Low C	7.161	4.898–10.468	11.550	7.262–18.360	0.808	0.237–2.760

aHR, adjusted hazard ratio; CI, confidence interval; Ref, reference.

Bold entries represent statistical significance.

^aType-unknown gastric cancers were censored.

Nomogram prediction models

The results of the multivariable analysis were used to develop nomogram prediction models for the probability of each type of GC at 5, 10, and 15 years. These models were presented to facilitate prediction interpretation in Figure 3a and b for each type and in Figure 3c for overall cancer. For example, a score of 200 on the nomogram indicates a 4.5% risk of developing ITGC within 5 years and just over 25% probability of developing the disease within 15 years. For DTGC, a person with the same score is predicted to have a very low probability of developing the disease, estimated at less than 4% within 15 years. For overall GC, the 5-year incidence rate is predicted to be just over 5% and the 15-year incidence rate is predicted to be just over 25%.

Figure 3.

Nomograms for gastric cancer probability. (a) Nomogram for intestinal type gastric cancer. (b) Nomogram for diffuse type gastric cancer. (c) Nomogram for overall gastric cancer. BMI, body mass index; Prob., probability.

Predictive performance of nomograms

AUC plots for the 15-year prediction performance through 10-fold cross-validation are presented in Figure 4. The nomogram for ITGC showed excellent prediction performance with both AUC values and c-index >0.8 (AUC = 0.82; c-index = 0.85), whereas that for DTGC showed poor performance (AUC = 0.61; c-index = 0.54). Including both, the nomogram for overall GC showed fair performance (AUC = 0.75; c-index = 0.76).

Figure 4.

AUC plot for 15-year prediction performance through 10-fold cross-validation for intestinal type gastric cancer (dotted line), diffuse type gastric cancer (dashed-dotted line), and overall gastric cancer (solid line). AUC, area under the curve.

DISCUSSION

This study developed and validated nomograms predicting GC in healthy subjects who had current or previous Hp infection by using a large-scale health surveillance cohort. In 10-fold cross-validation, the prediction performance of the nomogram for overall GC was found to be fair. However, although the nomogram for ITGC demonstrated better prediction performance, that for the DTGC seemed to be less reliable. Owing to the very low incidence of DTGC, these results are perhaps not surprising.

As is known, those with atrophic gastritis and negative Hp serology are highly likely to develop GC. Therefore, the so-called “ABC method,” which stratifies the risk for GC according to the anti-Hp serology and the presence of atrophy, has been suggested (18) (Figure 5a). According to the ABC method, the risk for GC can be stratified into 4 groups: group A [Hp (−) atrophy (−)], group B [Hp (+) atrophy (−)], group C [Hp (+) atrophy (+)], and group D [Hp (−) atrophy (+)]. In a prospective cohort study, the HR for groups B, C, and D compared with group A was found to be 1.1, 6.0, and 8.2, respectively (19). However, it has been recently argued that the ABC method should be refined to classify risk according to the level of anti-Hp antibody titer (20) (Figure 5b). In this theory, groups A, B, and C can be subdivided into 2 groups based on antibody titers. In group A, GC risk in the “high-negative antibody” subgroup is higher than that in the “low-negative” subgroup. Similarly, in group B, GC risk in the “high-positive antibody” subgroup is higher than that in the “low-positive” subgroup, mostly due to DTGC. However, in group C, GC risk is higher in the low-positive antibody subgroup than in the high-positive subgroup, typically due to ITGC.

Figure 5.

The “ABC method” which discriminates the risk for gastric cancer according to the anti-H. pylori serology and the presence of atrophy. (a) Conventional “ABC method” (15). (b) Updated ABC method which subclassifies the risk according to the anti-H. pylori antibody titer (12). HR, hazard ratio.

In this study, DTGC showed significantly different characteristics from ITGC. DTGC was not associated with older age, male sex, greater BMI, smoking, or even atrophic gastritis, unlike ITGC. Such characteristic differences have been reported in previous studies (21,22). DTGC is known to develop without passing through the precancerous mucosal change induced by long-term Hp infection (4,6,19,23,24), therefore developing in the earlier ages (17,25). In addition, DTGC is known to show no gender predominance or even female predominance and to have more aggressive traits than ITGC (26). In our previous research, most GC cases exceeding the indication for endoscopic resection, though under regular endoscopy within 2-year intervals, were DTGC, and among them, significantly younger age, current Hp infection, and less than mild degree of atrophic gastritis, were more frequent compared with endoscopically treatable ones (27), showing rapid progression of DTGC. As for smoking, several previous studies have found that ITGC is more likely to have a smoking history than DTGC (11,22). This might be because DTGC is less affected by the environmental factors than ITGC (28).

Interestingly, family history was shown to be highly related to DTGC but not with ITGC. Traditionally, patients with GC are reported to have 2- to 3-fold greater likelihood of a family history, showing that family history of GC is a significant risk factor (29–31). The reason why our study failed to show the risk of family history for ITGC is unclear. Nonetheless, the fact that DTGC has higher rate of family history is an interesting finding. There exists hereditary diffuse GC which is an autosomal-dominant cancer syndrome arising from a germline CDH1 gene mutation (32). However, hereditary diffuse GC accounts for only 1%–3% of GC cases, and germline mutations were found to be very rare in Korea according to a previous study (33). Nevertheless, there might be other unknown germline mutations or hereditary traits associated with DTGC. This should be confirmed in future studies.

The anti-Hp antibody titer represents the severity of gastritis (34–36). This study demonstrated that high titers are related to DTGC, which is in line with the previous studies. Previously, a Japanese study group showed a higher incidence of DTGC among those with higher anti-Hp antibody titer (13,34); however, those studies were only case-control studies without consideration of other baseline characteristics besides serum antibody titer and pepsinogen status. Another cohort study has shown that among those without gastric mucosal atrophy high antibody titer was associated with a higher risk of GC (37). This study included only male subjects, did not distinguish between types of GC, and observed only a small number of GC cases (n = 7). Similarly, another study also reported the association between high antibody titer and the risk of DTGC among those without gastric atrophy (38); however, it was also conducted exclusively with male population. Therefore, our study is the first to investigate the risk of each type of GC in relation to anti-Hp antibody titer along with other variables among large-scale prospective cohort.

As for “ABC group” in our study, we confirmed the hypothesis that even in group B, the risk of GC increases as the anti-Hp antibody titer increases. In this study, the HR of high B group for DTGC was about 1.7. Considering the HRs of high C and low C groups for ITGC was greater than 10, the risk of high B group for DTGC may seem insignificant; however, given that the anti-Hp antibody titer is the only modifiable risk factor for DTGC, and considering the high impact of DTGC on premature mortality despite its low incidence, this result has great implications for the concept of early diagnosis and prevention of DTGC.

DTGC develops relatively rarely compared with ITGC and thus has been less studied; however, it is known to have greater malignant potential and to occur at younger ages with less male predominance (39–41). Although there have been many studies aimed at identifying high-risk group for ITGC, DTGC has been often overlooked because of its relatively low incidence. However, although the incidence of ITGC is predominant in East Asia, DTGC occurs almost uniformly across regions geographically and shows little variation in incidence (42). In addition, the incidence of DTGC is relatively stable or increasing, whereas that of ITGC is decreasing (43), thereby increasing the significance of DTGC. This study reaffirms that DTGC is less attributed to environmental factors and is more likely induced by genetic factors. However, as DTGC was found to be more affected by the intensity of inflammation caused by Hp infection, greater focus should be placed on eradicating Hp in the young population with high anti-Hp antibody titers.

As for overall GC, the nomogram demonstrated a fair prediction performance, with an AUC and a c-index >0.7. Notably, the nomogram for ITGC showed excellent performance, with both the AUC and the c-index >0.8. The utilization of a large-scale cohort screening data set proved advantageous in this study. Nevertheless, it is probable that data comprising larger sample sizes and extended follow-up periods than those in this study would prove useful for the construction of more informative prediction models for DTGC. This is due to the markedly low probability of DTGC over a 10- or 15-year period. Indeed, when evaluated within realistic intervals of incidence, the model predictions were found to align with the observed probabilities. As for DTGC, it may be feasible to expand the current data set for model enhancement with longer follow-up periods in the future. This study is significant in that it established a predictive model for GC that quantitatively estimates the probability of occurrence with excellent performance, particularly for ITGC.

The limitation of this study is that we did not include those with negative Hp serology in this study. There may be individuals with a past infection with Hp that disappeared because of severe atrophy or past eradication treatment, making it difficult to distinguish them from a purely never-infected population. Thus, we could not include them in the discussion. This needs to be investigated in a further study involving a cohort including both sero-positive and sero-negative populations with careful consideration of eradication history. Another limitation is that the immunoassay analysis used here is not the only existing modality to measure the antibody titer, and thus, the cutoff value presented in this study may not be universally applicable. However, similar results are expected when using the same method as in this study, even with other serology test modalities and their corresponding cutoff levels. In addition, in this study, histological findings such as the degree of atrophy, intestinal metaplasia, and dysplasia were not included. Including these factors would likely enhance prediction effectiveness. However, as the purpose of this study was to develop a model to distinguish high-risk individuals within a general healthy populations, we believe that those histological evaluations requiring invasive biopsy would be less acceptable and less feasible due to costs and resource utilization in some settings. Meanwhile, including further risk factors such as dietary habit, socioeconomic factor, relevant medical history, and serum markers such as pepsinogen would contribute to a more detailed risk prediction.

By using a large-scale health surveillance cohort, we successfully developed and validated nomograms predicting GC development. In particular, the nomogram for ITGC showed excellent prediction performance. Our nomogram for GC probability is expected to be a valuable tool for identifying high-risk individuals among healthy populations. With this, it is expected that the prevention or early diagnosis of GC will become more achievable. For the prediction of DTGC, further studies involving larger populations are necessary.

In conclusion, this study developed and validated nomograms for predicting GC, demonstrating strong performance, which can facilitate high-risk identification and support early diagnosis. Efforts to incorporate a wider range of variables will likely be necessary in the future to further improve predictive accuracy.

CONFLICTS OF INTEREST

Guarantor of the article: Soo Jeong Cho, MD, PhD.

Specific author contributions: J.H.L. and S.J.C.: study conception and design, data analysis/interpretation. A.H. and S.H.: data analysis/interpretation. J.H.L.: manuscript drafting. S.H. and S.G.K.: critical revision of manuscript. All authors read and approved the final manuscript and the authorship list. All authors fulfill the ICMJE criteria for authorship.

Financial support: This research was supported by a grant (04-2022-0300) from the SNUH Research Fund and a grant (2022R1A2C101043711, 2022R1A2B5B01001430 and IRB # H-2104-165-1214) from the National Research Foundation of Korea, and the Korean Society of Gastrointestinal Endoscopy (2021). The work was independent of these fundings.

Potential competing interests: None to report.

Study Highlights.

WHAT IS ALREADY KNOWN

• ✓ It is well established that H. pylori can lead to gastric adenocarcinoma through atrophic gastritis.

• ✓ Once atrophic gastritis develops, it can induce self-regression of H. pylori by reducing the normal gastric epithelium.

• ✓ To evaluate the risk for gastric cancer, taking the atrophy into account in combination with the degree of inflammation by H. pylori is necessary.

WHAT IS NEW HERE

• ✓ Risk for intestinal type gastric cancer development can be well predicted by using a nomogram including anti-H. pylori antibody titer and degree of atrophy.