Infection with the gastric bacterium Helicobacter pylori (in particular infection with strains possessing the cytotoxin-associated gene [cagA], which encodes a highly immunogenic high molecular weight [120 kDa] protein [CagA]) and smoking have been identified as risk factors for the development of gastric cancer.1–4 Both risk factors start to act early in life, when the gastric mucosa is probably particularly vulnerable to the influence of carcinogens: H. pylori infection is typically acquired in early childhood,5 and smoking careers are typically initiated in adolescence. Furthermore, both risk factors typically prevail over decades if not for life, and they contribute to the development of precancerous lesions and gastric cancer among adults.6 It may therefore be hypothesized that co-occurrence of both risk factors might lead to synergistic effects on gastric cancer risk. The aim of this analysis was therefore to assess the risk of gastric cancer according to the individual and joint impact of H. pylori infection and smoking.
Infection with the gastric bacterium Helicobacter pylori (in particular infection with CagA-positive strains) and smoking have been identified as risk factors for the development of gastric cancer. Both risk factors are typically acquired early in life and prevail over decades if not for life. We assessed the individual and joint impact of both risk factors on gastric cancer risk in a population-based case-control study from Germany including 71 patients with histologically verified gastric cancer and 363 patients with colorectal cancer who served as controls. Information on smoking and potential confounding factors was collected by standardized interviews. H. pylori infection was measured serologically by immunoglobulin G antibody titers against H. pylori. In addition, antibodies against the CagA antigen were determined by Western blot. Twenty-seven percent of cases compared with 15% of controls were smokers, and 43% of cases compared with 23% of controls were infected with CagA-positive H. pylori strains. After control for potential confounders, the relative risk of gastric cancer was 2.6 (95% CI 1.2-5.7) for nonsmoking subjects with CagA-positive H. pylori infections and 7.2 (95% CI 2.2–23.6) for smoking subjects with CagA-positive H. pylori infections compared with subjects without these risk factors. The corresponding relative risks for noncardia gastric cancer were 6.1 ( 95% CI 2.3–16.5) and 16.6 (95% CI 4.3–64.2). We conclude that smoking subjects with CagA-positive H. pylori infections have a strongly increased risk of gastric cancer and may be an important group for targeting efforts of prevention and early detection. © 2002 Wiley-Liss, Inc.
MATERIAL AND METHODS
Study design and study population
A state-wide population-based study of risk factors, patterns of diagnostic process, and prognosis was conducted among patients with various forms of cancer in Saarland, a state with about 1 million inhabitants in Western Germany. Patients with a first diagnosis of gastric, colorectal or breast cancer between November 1996 and February 1998 who were younger than 80 years at the time of diagnosis were eligible for participation. Thirty-four of 36 hospitals in Saarland and the adjacent counties participated in recruitment. Among eligible subjects reported to the study center by these hospitals, 2% died before an interview could be conducted, and another 3% refused to participate. Based on estimates from the Saarland cancer registry, overall between 50–60% of eligible patients were enrolled. The study was approved by the Ethics Committees of the University of Ulm and of the Physicians Boards of the State of Saarland and the adjacent State of Rheinland Pfalz.
We report on a case-control analysis of the relation among H. pylori infection, smoking and gastric cancer, conducted among a subsample of the participants who agreed to provide a blood sample (79%). The case group included patients with gastric cancer; patients with colorectal cancer served as controls. In a previous analysis from the same study that had focused on the joint impact of family history and H. pylori infection on gastric cancer risk, H. pylori infection had only been measured in a subsample of patients with colorectal cancer who were matched to the cases with respect to age and sex.7 In the mean time, measurement of H. pylori infection was extended to all patients with gastric and colorectal cancer (n = 71 and 363, respectively) who agreed to give a blood sample (79% of study participants) to increase the power of the study, and all these subjects were included in the current analysis.
Cases and controls were recruited for participation and reported to the study center by their clinicians during first hospitalization due to the cancer. Informed consent was obtained in each case. Subsequently, personal interviews based on standardized questionnaires were conducted by a team of 4 trained interviewers. The questionnaires included detailed information about sociodemographic characteristics, medical history, family history and lifestyle habits. In particular, patients were asked whether they had smoked at least 100 cigarettes during lifetime, and if so, at what age they had started smoking, whether and at what age they had stopped smoking and how many cigarettes they smoked (or had smoked) on average. After completion of the interview, a blood sample was drawn and stored at −80°C. Sera were analyzed for H. pylori immunoglobulin G antibodies (GAP; Bio-Rad Laboratories Diagnostic Group, München, Germany). In case of a positive result, further analyses of antibodies against the CagA antigen (Helicobacter pylori Western blot; manufactured by AID, Strasberg, Germany, and distributed by DPC, Biermann, Germany) were performed. All laboratory analyses were done at a central laboratory at the University of Ulm after completion of the field work, as previously described.7
Cases and controls were described according to basic sociodemographic factors and family history of gastric cancer. In addition, detailed descriptive analyses were carried out on the smoking habits of former and current smokers. Patients who indicated that they had stopped smoking less than 1 year previously (n = 17) were classified as current smokers, because their smoking exposure is much more like the average exposure of current smokers than that of former smokers, and because some of these patients might have given up smoking after (and possibly as a consequence of) the onset of cancer symptoms. We then assessed the association between smoking habits (never/former/current) and gastric cancer and the association between H. pylori infection (uninfected/CagA-negative infection/CagA-positive infection) and gastric cancer, first looking at 1 risk factor at a time, and then looking at patterns of joint occurrence of both risk factors. We used multiple logistic regression to adjust associations of risk factors with gastric cancer for the following potential confounders: age (in years), sex, years of school education (≤9 years/>9 years), family history of gastric cancer (history of gastric cancer in at least 1 first-degree relative) and alcohol consumption (consumption of at least 1 alcoholic beverage per week on average). Because H. pylori infection has been found to selectively increase the risk of distal gastric cancer without increasing the risk of cardia cancer,1 the analyses were repeated after restriction of the cases to patients with noncardia gastric cancer only.
Basic characteristics of cases and controls are shown in Table I. Median age was 64 years in cases and 66 years in controls, and slightly over 60% of both cases and controls were men. Controls tended to be more highly educated and to drink alcohol more often, whereas a family history of gastric cancer was about 3 times as common among cases than among controls.
|Cases (n = 71)||Controls (n = 363)|
|Age (median yrs)||64||66|
|Males||43 (61)||225 (62)|
|>9 years school education||12 (17)||89 (24)|
|Alcohol consumption2||47 (66)||261 (72)|
|Family history of gastric cancer3||10 (14)||19 (5)|
Table II provides details of the smoking habits of cases and controls. The majority of both cases and controls had ever smoked, but the proportion of smokers, in particular the proportion of current smokers, was higher among cases than among controls. The majority of ever smokers had given up smoking. Former smokers had smoked as heavily as current smokers in the past, but most of them had given up smoking many years ago (median: 20 and 19 years ago in cases and controls, respectively; median age at cessation: 45.5 and 46 years in cases and controls, respectively).
|Cases (n = 71)||Controls (n = 363)|
|Never smokers (no.)||21 (30%)||156 (43%)|
|Current smokers (no.)||19 (27%)||54 (15%)|
|Age at initiation||18 (15–20)||20 (17–24)|
|Cigarettes/day||20 (15–30)||15 (8–20)|
|Pack years||38 (21–45)||29 (16–48)|
|Former smokers (no.)||31 (44%)||153 (42%)|
|Age at initiation||18 (17–21)||18 (17–21.5)|
|Age at cessation||45.5 (39–54)||46 (38–55)|
|Years since cessation||20 (10–22)||19 (11–29)|
|Cigarettes/day||20 (15–25)||20 (10–30)|
|Pack years||22 (16–39)||28 (10–42)|
The association between smoking status and gastric cancer is shown in Table III. After control for potential confounders, current smoking was associated with an approximately 3-fold increase in risk of gastric cancer, and a similar result was obtained when the analysis was restricted to cases with noncardia cancer only. There was also a positive albeit much weaker association between former smoking and gastric cancer (adjusted odds ratios: 1.9 for total gastric cancer, 1.7 for noncardia gastric cancer).
|Smoking||Cases [no. (%)]||Controls [no. (%)]||Odds ratio (95% CI)|
|All gastric cancers||Never||21 (30)||156 (43)||1.02||1.02|
|Former||31 (44)||153 (42)||1.5 (0.8–2.7)||1.9 (1.0–3.8)|
|Current||19 (27)||54 (15)||2.6 (1.3–5.2)||2.9 (1.4–6.1)|
|Noncardia cancer only||Never||19 (32)||156 (43)||1.02||1.02|
|Former||22 (37)||153 (42)||1.2 (0.6–2.3)||1.7 (0.8–3.6)|
|Current||19 (32)||54 (15)||2.9 (1.4–5.9)||3.4 (1.6–7.4)|
H. pylori infection, in particular infection with a CagA-positive strain, was strongly associated with noncardia gastric cancer (adjusted odds ratio for CagA-positive and -negative infections: 5.5 and 2.5, respectively), whereas these associations were weaker but still pronounced in analyses including all gastric cancers (adjusted odds ratios: 3.1 and 1.5, respectively) (Table IV).
|H. pylori infection||Cases [no. (%)]||Controls [no. (%)]||Odds ratio (95% CI)|
|All gastric cancers||No||17 (24)||135 (37)||1.02||1.02|
|CagA−||23 (33)||145 (40)||1.3 (0.6–2.5)||1.5 (0.8–3.0)|
|CagA+||30 (43)||85 (23)||2.8 (1.5–5.4)||3.1 (1.6–6.1)|
|Noncardia cancer only||No||10 (17)||135 (37)||1.02||1.02|
|CagA−||20 (34)||145 (40)||1.9 (0.8–4.1)||2.5 (1.1–5.8)|
|CagA+||29 (49)||85 (23)||4.6 (2.1–9.9)||5.5 (2.4–12.4)|
Tables V and IV show the risk of total gastric cancer and of noncardia gastric cancer, respectively, according to the joint distribution of smoking and H. pylori infection. In these analyses, never and former smokers were combined into a common group of current nonsmokers due to sample size limitations. Former smokers were classified as nonsmokers rather than smokers because most of them had given up smoking a very long time ago (Table II) and because their risk of gastric cancer, in particular their risk of noncardia gastric cancer, appeared to be closer to the risk of never smokers than to that of current smokers (Table III).
|Smoking||H. pylori infection||Cases [no. (%)]||Controls [no. (%)]||Odds ratio (95% CI)|
|No||No||13 (19)||108 (30)||1.02||1.02|
|No||CagA−||15 (21)||126 (35)||1.0 (0.5–2.2)||1.2 (0.5–2.7)|
|No||CagA+||23 (33)||75 (21)||2.5 (1.2–5.3)||2.6 (1.2–5.7)|
|Yes||No||4 (6)||27 (7)||1.2 (0.4–4.1)||1.2 (0.3–4.0)|
|Yes||CagA−||8 (11)||17 (5)||3.9 (1.4–10.8)||3.7 (1.3–10.8)|
|Yes||CagA+||7 (10)||10 (3)||5.8 (1.9–17.9)||7.2 (2.2–23.6)|
|Smoking||H. pylori infection||Cases [no. (%)]||Controls [no. (%)]||Odds ratio (95% CI)|
|No||No||6 (10)||108 (30)||1.02||1.02|
|No||CagA−||12 (20)||126 (35)||1.7 (0.6–4.7)||2.4 (0.8–6.9)|
|No||CagA+||22 (37)||75 (21)||5.3 (2.0–13.6)||6.1 (2.3–16.5)|
|Yes||No||4 (7)||27 (7)||2.7 (0.7–10.1)||2.9 (0.7–11.4)|
|Yes||CagA−||8 (14)||17 (5)||8.5 (2.6–27.4)||9.2 (2.7–31.9)|
|Yes||CagA+||7 (12)||10 (3)||12.6 (3.5–44.8)||16.6 (4.3–64.2)|
Both CagA-negative and -positive infections were associated with an increased risk of gastric cancer among nonsmokers, but the risk was much stronger for CagA-positive infections (adjusted odds ratios: 2.6 for total gastric cancer, 6.1 for noncardia gastric cancer) than for CagA-negative infections (adjusted odds ratios: 1.2 and 2.4, respectively) (Tables V, VI). In the absence of H. pylori infection, smoking was associated with a moderate (albeit not statistically significant) increase in the risk of noncardia gastric cancer (adjusted odds ratios for total and noncardia gastric cancer: 1.2 and 2.9, respectively). However, gastric cancer risk, and particularly risk of noncardia gastric cancer, was very strongly increased among smokers who were infected with H. pylori, particularly among smokers with CagA-positive infections. Compared with uninfected nonsmokers, the risk of noncardia gastric cancer (total gastric cancer) was increased about 9-fold (4-fold) in smokers infected with CagA-negative H. pylori strains and about 17-fold (7-fold) in smokers infected with CagA-positive H. pylori strains.
In our population-based case-control study from Germany, we found both smoking and H. pylori infection to be risk factors for gastric cancer. Strong risk elevations were found in case of joint presence of both risk factors. In particular, we found a very strong increase in risk of noncardia gastric cancer among smokers infected with CagA-positive H. pylori infections.
Our results concerning the individual impact of smoking or H. pylori infection on gastric cancer risk are consistent with results of previous studies addressing these questions.1–4, 6, 8–10 Despite the large number of such studies, the joint impact of both risk factors has only recently been addressed in 1 case-control study from Sweden (56 cases, 224 controls), which found that tobacco smoking is still a risk factor among H. pylori-infected individuals and that the risk of gastric cancer among H. pylori-infected current smokers was 11 times that of noninfected individuals not currently smoking.11 Although this result is entirely consistent with our finding, no distinction was made according to site of gastric cancer and according to presence of CagA antibodies in the study from Sweden. In agreement with previous studies,1–3 we found that H. pylori infection selectively increased the risk of noncardia cancer and that the risk increase was particularly strong for CagA-positive infections. CagA-positive infections and smoking increased the risk of noncardia gastric cancer in a close to multiplicative way, leading to an almost 20-fold risk increase of noncardia gastric cancer among smoking subjects with a CagA-positive H. pylori infection.
A strength of our study was its population-based approach. An unselected sample of patients was recruited from all types of hospitals from the entire state of Saarland. Thirty-four of 36 hospitals in Saarland and the adjacent counties participated in recruitment. Detailed information on smoking habits and important potential confounding factors was obtained in a standardized manner in all hospitals by the same team of 4 trained interviewers.
Our study also has limitations. The limited number of cases (n = 71) prohibited a more comprehensive assessment of gastric cancer risk according to joint classification of subjects by smoking habits and H. pylori infection and led to quite wide confidence intervals for some of the risk estimates. For example, because of the small number of subjects in individual cells, never smokers and former smokers were combined in the analyses assessing the joint impact of smoking and H. pylori infection. Although this appears to be justified given that former smokers typically had given up smoking a very long time ago (median: 19 years), exsmokers still appear to have a somewhat increased risk of gastric cancer. Therefore, the risk increase due to the joint impact of H. pylori infection and smoking would be expected to be even higher if never smoking, noninfected subjects were used as a reference group.
We used patients with colorectal cancer rather than healthy subjects as controls. The validity of our results is therefore contingent on the assumption that these controls adequately reflect the prevalence of H. pylori infection and smoking patterns in the population. Despite control for a variety of potential confounders in multivariable analysis, we cannot rule out residual confounding by factors that may differentially affect the risk of gastric cancer and colorectal cancer and that are also related to smoking or H. pylori infection, such as dietary influences, which were not controlled for in our analysis. Although results have been12 and continue to be inconsistent,13 there is now increasing evidence from recent large-scale epidemiologic studies that long-term cigarette smoking is associated with a weakly to moderately increased risk of colorectal cancer.14–21 A positive association between smoking and colorectal cancer would have worked against the strong patterns found in our study, however, i.e., it would have led to underestimation rather than overestimation of the effects of smoking. Although some reports have suggested a possible weak association between H. pylori infection and colorectal cancer,22, 23 this association was not confirmed in other, partly much larger studies,24, 25 suggesting a role of chance and possible publication bias and the absence of a major effect. Again, a positive association between H. pylori infection and colorectal cancer would have worked against the strong patterns found in our study, i.e., it would have led to underestimation rather than overestimation of the role of H. pylori infection on gastric cancer risk.
In our study, H. pylori infection was measured serologically by commercial tests. Although the GAP BioRad IgG test kit used in our study performs well compared with other commercial test kits, specificity is limited (median sensitivity = 94% and median specificity = 79% in published evaluations among 20 populations including 2877 participants).26 However, limited specificity (which was determined by comparison with gold standard methods for assessing current infection) may partly reflect the fact that patients who have been infected in the past and who have lost the infection only recently are still serologically positive.27 Because these subjects are likely to have been exposed to H. pylori infection for most of their life, their serologic classification as infected appears to be justified from an etiologic point of view. A more serious concern may be that some of the gastric cancer patients may have been seronegative, even though they had been infected in the past, due to a loss of infection after development of advanced gastric disease.28 Taking these factors together, some underestimation of associations is probably due to imperfect classification of infection status.29
In summary, we found a strongly increased risk of gastric cancer in smokers infected with H. pylori, which is more likely to be underestimated than overestimated due to data limitations. These findings may have important public health implications, because they may help to identify subjects at highest risk who might potentially benefit most from eventual measures of prevention or early detection of gastric cancer. Our results also underline once more the public health importance of prevention of smoking and of assistance in smoking cessation. This particularly applies to developing countries, in which H. pylori infection remains very common, and which currently experience a most unfortunate increase in smoking rates.30
The authors thank Drs. C. Hetke, W. Michaels, A. Becker, and M. Schramm for conducting the interviews.