Serum pepsinogen I and anti-Helicobacter pylori IgG antibodies as predictors of gastric cancer risk in Finnish males

SUMMARY

Background

Serum pepsinogen (SPG) and anti-Helicobacter pylori serology have been used for gastric risk stratification in Asia.

Aim

To assess utility of these markers in a Western population.

Methods

SPGI measurements were available for 21,895 Finnish male smokers in the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study. We used Cox proportional hazards models adjusted for potential confounders to estimate gastric cancer hazard ratios (HR) and 95% confidence intervals (95% CI) for low SPGI (<25μg/l). In a subset (n=3,555) with anti-H. pylori serology, these markers jointly defined the following: Group A (H. pylori[−], SPGI[normal]; reference group), Group B (H. pylori[+], SPGI[normal]), Group C (H. pylori[+], SPGI[low]) and Group D (H. pylori[−], SPGI[low]). Odds ratios (ORs) and 95% CI were calculated using multivariate logistic regression.

Results

There were 329 gastric cancers diagnosed an average of 13.9 years after baseline. Prediagnostic low SPGI was significantly associated with increased gastric cancer risk (HR 2.68, 95% CI 1.99–3.61). Among subjects with both SPGI and H. pylori serology, groups B, C and D had increased gastric cancer ORs (95% CI) of 1.79 (1.21–2.64), 3.85 (2.36–6.28) and 6.35 (2.20–18.34) respectively. CagA seropositives had significantly higher ORs than CagA seronegatives within group B (P_{heterogeneity}=0.01). For groups B and C, repeat SPGI level at 3 years did not further stratify gastric cancer risk.

Conclusions

Low SPGI was associated with increased gastric cancer risk in our large Finnish cohort. A single measurement of SPGI along with H. pylori whole cell and CagA serology provides potentially useful prediction of gastric cancer risk.

INTRODUCTION

Chronic infection with Helicobacter pylori (H. pylori) is well-known to be a strong risk factor for gastric cancer ^1–3. H. pylori colonization of the gastric mucosa induces inflammation that causes chronic gastritis and mucosal atrophy that may eventually lead to gastric cancer ^{4, 5}. Serologic response to infection by H. pylori can be assessed by measuring anti-H. pylori immunoglobulin G (IgG) antibodies using assays based on whole cell sonicate or one or more individual bacterial antigens ⁶.

Altered levels of serum pepsinogens (SPG), which are mainly produced by the chief cells of the fundic glands of the stomach, reflect the atrophic status (i.e., gland loss) of gastric mucosa ^{7, 8}. SPG levels not only reveal the past infection status or current atrophy of the stomach, respectively, but have also been shown to be predictive of gastric cancer risk^{9, 10}. However, previous studies in Western populations are limited, failing to adjust for potential confounders ^11–13 and/or to quantitate risk ^14–16, mostly because of small sample sizes. Even in Asian populations, there are only a few prospective studies investigating the main effect of SPGI with gastric cancer risk ^{17, 18}.

Anti-H. pylori antibodies may undergo seroreversion with time and/or progression of disease, and may be undetectable later in the course of disease ^{19, 20}. SPG levels are normal among H. pylori- infected individuals without atrophic gastritis ²¹ as well as in some cases of gastric cancer, particularly with diffuse-type histology ^{17, 22}. Thus, the combination of the two markers has been suggested to overcome the limits of each, and this has been applied in Japan as a screening tool for gastric cancer, an approach known as the “ABC(D) method” ^{21, 23}. A recent meta-analysis of East Asian studies reported a gastric cancer meta-HR as high as 13 times in the highest risk group ²⁴. Previous studies in Western populations that examined the joint association of serum pepsinogen and anti-H. pylori seropositivity with gastric cancer risk have been limited in sample size (less than 100 cases) ^{11, 13, 25}, did not provide overall risk estimates for the combined effects ^26–28, and/or were not adjusted for possible confounders ¹².

Gastric cancer is a heterogenous disease, with important epidemiologic differences in among subtypes. For instance, with regards to anatomical subsites, H. pylori is a major risk factor for noncardia but not for cardia gastric cancer ²⁹. Divergent incidence trends have been reported for these subtypes in different populations ^{30, 31}. While intestinal-type gastric cancer is often related to environmental factors such as H. pylori or diet, diffuse-type cancer is more closely associated with genetic predisposition ³². Furthermore, some studies report stronger associations with higher anti-H. pylori antibody titer and/or infection with cytotoxin-associated gene A (CagA) virulence factor-positive strains ^{33, 34}.

The high mortality rate of gastric cancer is mostly a consequence of late detection, stemming from the lack of specific symptoms of the disease ^{35, 36}. Gastric cancer when found early may be curable by endoscopic or minimally invasive surgery ³⁷. In countries of high gastric incidence where general population screening by endoscopy is not routinely conducted, triaging high risk individuals for definitive evaluation by endoscopy would be useful. However, the utility of non-invasive risk stratification by blood tests has not been evaluated outside of a few high-income Asian countries. Therefore, the aims of this study are to evaluate the association of low serum pepsinogen I (SPGI) with gastric cancer risk overall and by subtypes and to assess the combination of H. pylori serology and SPGI as a joint predictor of gastric cancer risk, in a prospective cohort study conducted in a Western population.

METHODS

Study Population

The current analysis represents an extension of prior reports ^{14, 15, 38}, with inclusion of additional cancer cases and consideration of repeated SPGI measurements. Subjects were from the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study, a randomized, double-blinded, placebo-controlled, 2 × 2 factorial trial of daily supplementation of alpha-tocopherol (50 mg) and/or beta-carotene (20 mg) for the primary prevention of lung cancer ^{39, 40}. A total of 29,133 Caucasian male smokers aged 50–69 years were originally recruited between 1985 and 1988 in southwestern Finland. At baseline, study participants completed questionnaires on demographic characteristics, self-reported medical history, life-style factors and dietary history. Fasting blood samples were collected at baseline and after 3 years’ intervention, stored in serum aliquots at −70 °C until testing. The study was approved by the Institutional Review Boards of both the National Cancer Institute, Bethesda, Maryland, USA and the National Public Health Institute, Helsinki, Finland. All participants provided written informed consent.

SPGI was measured at baseline and follow-up for ATBC participants who continued in the study for more than 3–5 years (n=21,895, 75% of the original cohort). Low SPGI in either blood sample triggered referral for upper gastrointestinal endoscopy, as previously reported ^{14, 15}. In the current study, baseline measurements were used for determining the association between SPGI and incidence of gastric cancer.

Anti-H. pylori antibody status determined in prior nested-case control analyses of pancreas, biliary tract, esophagus, lung, colorectal, and gastric cancers ^{38, 41–48} was used to evaluate the joint effect of SPGI and anti-H. pylori antibodies on gastric cancer risk for a total of 3,555 subjects with both measures.

Identification and Classification of Cancer Cases

The intervention phase ended in April, 1993, but subjects have been passively followed-up using the Finnish Cancer Registry which has nearly 100% case coverage of the ATBC cohort ⁴⁹. Diagnoses of gastric cancer were classified according to the International Classification of Diseases, Ninth Revision (ICD-9) as cardia (ICD-9 code 151.0) or noncardia (ICD-9 codes 151.1–151.9). Two histological subtypes according to Lauren classification, intestinal- and diffuse-types, were separately assessed ⁵⁰. The median time to gastric cancer diagnosis was 13.9 (Standard Deviation, SD, 6.8) years.

Laboratory Analysis

SPGI analyses were performed by radioimmunoassay in two laboratories. Serum samples were assayed at the laboratory of I. M. Samloff, Sepulveda, California, USA during 1989–1991 ⁵¹. After that facility was damaged by an earthquake, the remaining samples were assayed at the laboratory of M. Härkönen, Helsinki, Finland from 1992–1993^{14, 15}. The SPGI measurements from the two laboratories were standardized and transformed for compatibility, with low SPGI defined as 25 μg/L or less, as previously reported ^{14, 15, 38}.

Anti-H. pylori IgG antibodies were measured by whole cell enzyme-linked immunosorbent assays (ELISA) ⁵² or multiplex bead-based assays ^{53, 54}, as previously described ^{38, 41–48}. The two methods were standardized to create an indicator variable for seropositivity and the continuous values dichotomized into low vs. high titer among seropositives, as previously reported ⁵⁵. Antibodies to the H. pylori virulence factor CagA antigen were also measured by ELISA or bead-based assays and classified for statistical analysis as present vs. absent, as previously reported ⁵⁵. Furthermore, positivity for anti-CagA antibodies was considered indicative of anti-H. pylori seropositivity regardless of other test results, since anti-CagA antibodies can remain positive relatively longer than anti-H. pylori antibodies ⁵⁶. Antibody titer and anti-CagA seropositivity were evaluated for further refinement of anti-H. pylori associations with gastric cancer risk.

Statistical Analyses

Gastric cancer hazard ratios (HRs) and 95% confidence intervals (95% CI) associated with low SPGI were estimated using Cox proportional hazards models. For each participant, follow-up time was calculated from the date of randomization until the diagnosis of cancer, death or December 31, 2014. In the subset of subjects (n=3,555) with both baseline SPGI and H. pylori serology information available, odds ratios (ORs) and 95% CI were estimated using logistic regression models for the following 4 categories: Group A (H. pylori[−], SPGI[normal]; reference group), Group B (H. pylori[+], SPGI[normal]), Group C (H. pylori[+], SPGI[low]) and Group D (H. pylori[−], SPGI[low]). Subgroup analyses were performed to evaluate variation by anti-H. pylori antibody titer and anti-CagA seropositivity within anti-H. pylori seropositive groups. We calculated the ratio of the ORs and p-heterogeneity for subtypes of gastric cancer in case-case comparisons ⁵⁷. The associations with histologic subtypes were assessed in both the Cox proportional hazards model and in the logistic models.

The effects of changes between baseline and 3-year follow-up SPGI within each ABCD group were analyzed using logistic regression. The 3-year follow-up measurements were available for 3,462 subjects, approximately 97.4% of the participants with both baseline SPGI and H. pylori serology information. Lag analyses were conducted to estimate marker associations with incident gastric cancer occurring less than 5 years, 5–10 years, and more than 10 years after enrollment. Sensitivity analyses were performed by: 1) excluding overlapping and unspecified subsites for noncardia cancer, 2) limiting to only cancer-free controls and 3) using a previously defined gastric cancer nested case-control set ⁴⁵. To address the concern for potential bias in case ascertainment due to pepsinogen measurement, two sensitivity analyses were performed. First, a log-rank test assessed whether gastric cancer incidence overall differed between the participants with vs. without SPGI. Second, a chi-square test assessed difference in early (IA, IB) vs. late (II–IV) stage distributions of these incident cancers.

For all analyses, except for the log-rank model, minimally adjusted models included age at randomization and type of assigned intervention in order to account for the experimental study design. Based on known or suspected associations with gastric cancer risk, additional covariates for the full models included Body Mass Index (BMI; kg/m², continuous), pack years of smoking (continuous), alcohol drinking (g/day, continuous), highest level of education (categorical), fruit intake (g/day), vegetable intake (g/day). The fully adjusted models excluded approximately 5% of subjects who did not have information on dietary factors.

Statistical analyses were conducted using SAS 9.3 (SAS Institute Inc, Cary, NC). All P values were two-sided, and were considered significant for P <0.05.

RESULTS

The baseline characteristics of the total cohort and gastric cancer cases, and subjects with both H. pylori serology and pepsinogen measurements in subset analysis are presented in Table 1. Within the cohort analysis, cancer cases had more years of smoking and lower consumption of fruit and vegetables. Low SPGI was more common among gastric cancer cases than controls. In the subset analysis, gastric cancer cases consumed less vegetables and showed a higher prevalence of low SPGI. All other variables showed no difference between the cases and controls.

Table 1.

Baseline characteristics for ATBC study participants who did or did not develop gastric cancer in the overall study cohort and in the subgroup with both H. pylori serology and serum pepsinogen I measurements

	Cohort analysis (n=21,895)				Subset analysis (n=3,555)
	entire cohort		gastric cancer		controls		gastric cancer
N	21,895		329		3,291		264
Age, y (SD)	57.0	(5.0)	57.3	(4.9)	58.2	(4.9)	58.0	(4.9)
Body mass index, kg/m2 (SD)	26.3	(3.7)	26.5	(3.7)	26.2	(3.7)	26.5	(3.7)
Pack years of smoking (SD)	36.3	(18.0)	37.6	(19.0)	36.8	(17.9)	38.7	(19.9)
Alcohol intake, g (SD)	17.3	(20.3)	17.3	(20.5)	17.3	(21.1)	18.1	(21.7)
Daily fruit intake, g (SD)	221.4	(196.5)	205.0	(161.8)	217.9	(185.9)	201.3	(157.6)
Daily vegetable intake, g (SD)	297.7	(113.8)	289.2	(106.3)	291.5	(111.2)	284.9	(103.8)
Some high school, college, or technical school (%)	14,772	67.5	223	67.8	2,222	67.5	177	67.1
Low SPGI (< 25μg/l), n (%)	1,791	8.2	58	17.6	298	9.1	51	19.3
H. pylori seropositivity, n (%)					2,536	77.1	224	84.9

Abbreviations: ATBC - Alpha-Tocopherol, Beta-Carotene Cancer Prevention, SD - Standard deviation, SPGI – serum pepsinogen I

Table 2 shows the association of low baseline SPGI with subsequent gastric cancer risk. In the fully adjusted Cox model (model 2), low SGPI at baseline conferred 2.7-fold higher risk for gastric cancer (95% CI 1.99–3.61) as compared to normal SPGI. Analysis by anatomical subsite showed HRs of 2.95 (95% CI 2.11–4.12) for noncardia and 2.01 (95% CI 1.05–3.83) for cardia gastric cancers. HRs were significantly elevated for intestinal-type gastric cancer (HR 2.57, 95% CI 1.57–4.21), but not for diffuse-type gastric cancer (HR 0.92, 95% CI 0.33–2.55).

Table 2.

Association of prediagnostic low serum pepsinogen I and subsequent gastric cancer risk in the ATBC Study

	number of cases	person years	model 1		model 2
			HR	95% CI	HR	95% CI
Gastric cancer total	329	5,724	2.53	1.89–3.37	2.68	1.99–3.61
Anatomical subsite
Noncardia	245	4,274	2.76	1.99–3.83	2.95	2.11–4.12
Cardia	84	1,449	1.93	1.04–3.59	2.01	1.05–3.83
Lauren classification
Intestinal	115	1,759	2.33	1.43–3.81	2.57	1.57–4.21
Diffuse	63	1,101	1.04	0.41–2.61	0.92	0.33–2.55
Others	151	2,864	3.49	2.35–5.17	3.74	2.49–5.61

Abbreviations: ATBC - Alpha-Tocopherol, Beta-Carotene Cancer Prevention, HR – hazard ratio, CI – confidence interval

model 1: adjusted for age at randomization and type of intervention

model 2: model 1 + body mass index, pack years of smoking, alcohol drinking, education, fruit intake, and vegetable intake

In the subset with both SPGI and H. pylori serology information available, Groups B, C, D had significantly elevated risks for total gastric cancer as compared to group A, with ORs (95% CI) of 1.79 (1.21–2.64), 3.85 (2.36–6.28), and 6.35 (2.20–18.34), respectively, P_trend <0.0001 (Table 3). ORs (95% CI) restricted to noncardia gastric cancer were higher for groups B, C and D (P_trend <0.0001), 3.59 (2.02–6.39), 7.49 (3.84–14.61), and 16.55 (5.26–52.04) respectively. Associations with cardia gastric cancer were not statistically significant for any group, nor were P_trends statistically significant.

Table 3.

Combination of anti-H. pylori seropositivity and pepsinogen I on risk of gastric cancer in ATBC Study

	control (n=3,291)		case (n=264)		model 1		model 2
	N	%	N	%	OR	95% CI	OR	95% CI
Total Gastric Cancer (n=264)
Group A	736	22.4	35	13.3	ref		ref
Group B	2,257	68.6	178	67.4	1.70	1.17–2.48	1.79	1.21–2.64
Group C	279	8.5	46	17.4	3.71	2.32–5.93	3.85	2.36–6.28
Group D	19	0.6	5	1.9	5.94	2.09–16.92	6.35	2.20–18.34
p-trend					<0.0001		<0.0001
Anatomical subsite
Noncardia (n=197)
Group A	736	22.4	14	7.1	ref		ref
Group B	2,257	68.6	143	72.6	3.49	2.00–6.09	3.59	2.02–6.39
Group C	279	8.5	35	17.8	7.40	3.89–14.07	7.49	3.84–14.61
Group D	19	0.6	5	2.5	15.62	5.07–48.12	16.55	5.26–52.04
p-trend					<0.0001		<0.0001
Cardia (n=67)
Group A	736	22.4	21	31.3	ref		ref
Group B	2,257	68.6	35	52.2	0.53	0.31–0.92	0.58	0.33–1.03
Group C	279	8.5	11	16.4	1.30	0.61–2.77	1.40	0.63–3.11
Group D	19	0.6	0	0.0	-		-
p-trend					0.65		0.80

Abbreviations: ATBC - Alpha-Tocopherol, Beta-Carotene Cancer Prevention, OR – odds ratio, CI – confidence interval, SPGI- Serum pepsinogen I,

Group A: H. pylori (−) and normal SPGI, Group B: H. pylori (+) and normal SPGI, Group C: H. pylori (+) and low SPGI, Group D: H. pylori (−) and low SPGI

model 1: adjusted for age at randomization and type of intervention

model 2: model 1 + body mass index, pack years of smoking, alcohol drinking, education, fruit intake, and vegetable intake

H. pylori titer was not associated with risk of noncardia gastric cancer either within group B, the normal SPGI and H. pylori seropositive group (P_{heterogeneity}=0.23), or within group C, the low SPGI and anti-H. pylori seropositive group (P_{heterogeneity}=0.52) (Table 4). On the other hand, the CagA seropositive group showed significantly higher OR (4.34, 95% CI 2.41–7.79) than the CagA seronegative group OR (2.45, 95% CI 1.23–4.88) within group B (P_{heterogeneity} =0.01). There was no significant difference by CagA status in group C (P_{heterogeneity} =0.06).

Table 4.

Combination of anti-H. pylori titer and CagA seropositivity with pepsinogen I on risk of noncardia gastric cancer in ATBC Study

	control (n=3,291)		case (n=197)		model 1		model 2
	N	%	N	%	OR	95% CI	OR	95% CI
Anti-H. pylori antibody titer in all noncardia cancer (n=197)
Group A	736	22.4	14	7.1	ref		ref
Group B + low titer H. pylorib	1,571	47.7	92	46.7	3.21	1.82–5.68	3.34	1.85–6.03
Group B + high titer H. pylori	686	20.8	51	25.9	4.15	2.27–7.58	4.17	2.23–7.79
Group C + low titer H. pylorib	186	5.7	25	12.7	7.96	4.03–15.73	8.09	4.00–16.37
Group C + high titer H. pylori	93	2.8	10	5.1	6.34	2.72–14.78	6.31	2.59–15.35
Group D	19	0.6	5	1.9	15.66	5.08–48.25	16.57	5.27–52.13
p-trend					<0.0001		<0.0001
Anti-CagA seropositivity in all noncardia cancer (n=186)a
Group A	736	24.1	14	7.5	ref		ref
Group B + CagA (−)c	590	19.4	25	13.4	2.37	1.22–4.60	2.45	1.23–4.88
Group B + CagA (+)	1,451	47.6	110	59.1	4.18	2.38–7.36	4.34	2.41–7.79
Group C + CagA (−)c	46	1.5	11	5.9	14.14	6.04–33.11	13.33	5.47–32.47
Group C + CagA (+)	207	6.8	21	11.3	6.03	2.99–12.15	6.28	3.04–12.98
Group D	19	0.6	5	2.7	15.71	5.09–48.42	16.73	5.31–52.68
p-trend					<0.0001		<0.0001

Abbreviations: ATBC - Alpha-Tocopherol, Beta-Carotene Cancer Prevention, OR – odds ratio, CI – confidence interval, SPGI- Serum pepsinogen I,

Group A: H. pylori (−) and normal SPGI, Group B: H. pylori (+) and normal SPGI, Group C: H. pylori (+) and low SPGI, Group D: H. pylori (−) and low SPGI

model 1: adjusted for age at randomization and type of intervention

model 2: model 1 + body mass index, pack years of smoking, alcohol drinking, education, fruit intake, and vegetable intake

^aexcluded with missing CagA status (11 cases & 242 controls) or H. pylori negative/CagA positive cases (157 controls)

^bWithin group B high titer vs. low titer H. pylori (P_{heterogeneity} = 0.23), group C high titer vs. low titer H. pylori (P_{heterogeneity} =0.52) in model 2

^cWithin group B CagA (+) vs. CagA (−) (P_{heterogeneity} = 0.01), group C CagA (+) vs. CagA (−) (P_{heterogeneity} =0.06) in model 2

The trend across groups was statistically significant in intestinal-type gastric cancer (P_trend <0.0001) but not in diffuse-type gastric cancer in multivariate analysis (P_trend =0.25) (Table 5).

Table 5.

Combination of anti-H. pylori seropositivity and pepsinogen I on risk of noncardia gastric cancer by Lauren classification in ATBC Study

	control (n=3,291)		case (n=197)		model 1		model 2
	N	%	N	%	OR	95% CI	OR	95% CI
Intestinal Type Gastric Cancer (n=69)
Group A	736	22.4	5	7.3	ref		ref
Group B	2,257	68.6	51	73.9	3.47	1.37–8.73	4.23	1.52–11.80
Group C	279	8.5	12	17.4	7.02	2.42–20.39	8.65	2.72–27.50
Group D	19	0.6	1	1.5	8.86	0.98–80.45	11.95	1.25–114.23
p-trend					<0.0001		<0.0001
Diffuse Type Gastric Cancer (n=49)
Group A	736	22.4	5	10.2	ref		ref
Group B	2,257	68.6	40	81.6	2.69	1.05–6.86	2.57	1.00–6.59
Group C	279	8.5	4	8.2	2.27	0.60–8.64	1.74	0.41–7.46
Group D	19	0.6	0	0.0	-		-
p-trend					0.15		0.25
Other Type Gastric Cancer (n=79)
Group A	736	22.4	4	5.1	ref		ref
Group B	2,257	68.6	52	65.8	4.53	1.63–12.58	4.25	1.52–11.84
Group C	279	8.5	19	24.1	14.69	4.90–44.01	13.92	4.61–42.09
Group D	19	0.6	4	5.1	46.21	10.60–201.47	43.81	9.89–194.11
p-trend					<0.0001		<0.0001

Abbreviations: ATBC - Alpha-Tocopherol, Beta-Carotene Cancer Prevention, OR – odds ratio, CI – confidence interval, SPGI- Serum pepsinogen I

Group A: H. pylori (−) and normal SPGI, Group B: H. pylori (+) and normal SPGI, Group C: H. pylori (+) and low SPGI, Group D: H. pylori (−) and low SPGI

model 1: adjusted for age at randomization and type of intervention

model 2: model 1 + body mass index, pack years of smoking, alcohol drinking, education, fruit intake, vegetable intake

Gastric cancer risks associated with 3-year change of SPGI are shown in Figure 1. Within groups B and C, gastric cancer ORs did not significantly differ between individuals with normal vs. low SPGI at follow-up.

Figure 1.

Association of prediagnostic and 3-year follow-up measurement of serum pepsinogen I and subsequent gastric cancer risk in the ATBC Study

Abbreviations: ATBC – Alpha-Tocopherol, Beta-Carotene Cancer Prevention, f/u – follow-up, OR – odds ratio, CI – confidence interval, SPGI - Serum pepsinogen I,

Group A: H. pylori (−) and normal SPGI, Group B: H. pylori (+) and normal SPGI, Group C: H. pylori (+) and low SPGI, Group D: H. pylori (−) and low SPGI

Model adjusted for age at randomization, type of intervention, pack years of smoking, alcohol drinking, education, fruit intake, vegetable intake

* Numbers in parentheses indicate gastric cancer cases in each group at follow-up

Cases diagnosed before follow-up measurements (≤3 years) were excluded from the analysis (n=5)

In a lag analysis of prediagnostic SPGI and subsequent gastric cancer risk, HR was highest within 5 years after baseline (4.73, 95% CI 1.88–11.88). The estimate is accentuated in noncardia and intestinal-type gastric cancers with HR (95% CI) of 5.97 (2.15–16.56) and 5.63 (1.30–24.32), respectively (Supporting Information Table 1).

Similarly, in the combined analysis of SPGI and anti-H. pylori seropositivity, group D, which is mainly the association of low SPGI (anti-H. pylori seronegative group), had the highest OR when restricted to cancers diagnosed < 5 years after enrollment (OR 16.93, 95% CI 2.85–100.43). The OR decreased to 5.98 (95% CI 1.24–28.93) in lag 5–10 years and 3.16 (95% CI 0.39–25.67) in lag > 10 years (Supporting Information Figure 1). On the other hand, when SPGI was normal the association of anti-H. pylori seropositivity was lowest in lag < 5 years with OR 0.64 (95% CI 0.22–1.91) and increased to OR 1.58 (95% CI 0.86–2.90) in lag 5–10 years and OR 2.43 (95supp% CI 1.38–4.30) in lag > 10 years.

Three separate sensitivity analyses restricting the noncardia cases excluding overlapping and unspecified site (Supporting Information Table 2), using only non-cancer controls from subset of the cohort with both SPGI and H. pylori serology information available from previous nested case-control sets (Supporting Information Figure 2), and testing within one nested case-control set for gastric cancer risk (Supporting Information Figure 3) did not change the risk estimates or the trend of risk.

Based on the log-rank test, there was no significant difference in gastric cancer incidence between those with SPGI testing and those without (p=0.38). Furthermore, the fraction of early gastric cancer did not differ between these two groups (p=0.19).

DISCUSSION

We report the largest study of the association of SPGI and H. pylori serology with gastric cancer risk in a Western population. Our analyses were based on combined marker categories similar to the Japanese “ABC(D)” method which defines low pepsinogen as PGI ≤ 70ng/ml and PGI/PG II ≤ 3.

Despite the difference in definition of low pepsinogen, the risk estimates for noncardia gastric cancer derived from our study are similar to estimates for overall gastric cancer from previous studies in Asia, where cardia cancer is relatively uncommon ^58–62 (Supporting Information Table 3).

There are only limited studies that further stratify H. pylori seropositivity within ABCD groups in relation to gastric cancer risk. In our study, titer of anti-H. pylori antibodies was not associated with level of risk among participants with normal SPGI status (i.e., group B). Among individuals with low SPGI status (i.e., group C), the non-significantly higher OR we found for low anti-H. pylori antibody titer may reflect longer exposure to H. pylori infection and more severe disease progression leading to diminished antibody production ⁶³. Previous reports in Asian populations have been inconsistent, associating increased gastric cancer risk with high antibody titer in some studies ^{59, 64} and with low titer in others ^{33, 65}. To the best of our knowledge, this study is the first to investigate H. pylori antibody titer in relation to SPGI status in a Western population.

Mucosal atrophy represents the intermediate outcome in the causal pathway from H. pylori infection to gastric cancer. Antibodies to CagA have been associated with higher risk of gastric cancer among H. pylori seropositive populations ³⁴, presumably because CagA positive H. pylori strains cause more severe mucosal damage than negative strains. We found anti-CagA seropositivity was associated with gastric cancer risk among individuals with normal pepsinogen (group B), but not with low pepsinogen (group C). Thus, our data implies that CagA-positive infection confers no additional risk once mucosal damage has occurred.

Another point to note is the assessment of the usefulness of repeated measurement of SPGI in predicting gastric cancer risk (Figure 1). Our analysis shows that a 3-year follow-up of SPGI does not differentiate risk within baseline categories of B and C. The current analysis cannot provide evidence for other follow-up intervals but, in our lag analysis for prediagnostic SPGI, the association of SPGI with gastric cancer was highest in the first five years prior to diagnosis (Supporting Information Table 1). Therefore, we cautiously conclude that repeated measurement may not be necessary for SPGI and H. pylori serology assessed at baseline. For individuals with normal levels of SPGI, measurement of anti-CagA antibodies may provide better discrimination of gastric cancer risk.

The current Cox regression analysis associated low SPGI with gastric cancer risk, in line with our previous nested case-control analysis of cases diagnosed through April 2006 ³⁸. We found a stronger association within shorter time intervals with the highest HR observed for cancers occurring within 5 years after baseline (Supporting Information Table 1). Our findings replicate previous studies (Supporting Information Table 4), not only regarding the significant association of low SPGI with gastric cancer overall, but also higher risk estimates for noncardia subsites and intestinal-type histology.

The major strength of our study is the use of prediagnostic samples analyzed for SPGI and anti-H. pylori seropositivity at baseline, with a long follow-up period. Another strength is that we have the largest number of cases (n=329) ever studied in a Western population, allowing us for a more comprehensive analysis across group categories and anatomical and histological subtypes, as well as accounting for potential confounder effects. While this study has a large number of gastric cancer cases, sub-group analyses are likely to be underpowered.

Although both men (as compared to women) and smokers (as compared to non-smokers) have higher risks of gastric cancer, our findings based on male smokers may have limited generalizability. Nonetheless, the risk estimates were similar to the previous studies. Another possible limitation of our study is that the SPGI measurement selected high-risk individuals for further screening with endoscopy, but we did not find evidence of excess cancer diagnoses among the screened fraction of the cohort compared to those who were unscreened. Moreover, the analysis incorporating H. pylori serology information was based on availability of data from prior nested case-control datasets for various cancers, but the results were unchanged by restriction to the matched gastric cancer case-control set. We also did not gather information about possible H. pylori eradication therapy post-enrollment.

In conclusion, we found that the joint consideration of SPGI and anti-H. pylori whole cell and CagA seropositivity is a potentially useful predictor for the development of gastric cancer, especially noncardia gastric cancer, in a large population study of a Western population. This noninvasive and relatively inexpensive method could be used to identify high risk individuals for definitive evaluation by endoscopy and/or to advise on H. pylori eradication. Risk stratification could potentially be improved by incorporating other factors, such as an individual’s genetic and epigenetic information ⁶⁶. Further studies of cost-effectiveness and of risk-specific screening intervals are warranted for development of personalized screening guidelines.