Gastric cancers in Finnish patients after cure of Helicobacter pylori infection: A cohort study
Helicobacter pylori infection is associated with gastric cancer. A total of 97% of the infected subjects have elevated levels of H. pylori antibodies. The antibody titers have been shown to decline rapidly (40–60% within 4–12 months) only after successful eradication therapy. We allocated 26,700 consecutive patients tested during 1986–1998 for H. pylori antibodies to 3 subcohorts: seropositive patients with rapidly falling antibody titers (Hp+CURED, n = 3,650), seropositive patients where no serological information indicating cure was obtained (Hp+NoInfo, n = 11,638) and seronegative patients (Hp–, n = 11,422). In the subcohorts, the standardised incidence ratios (SIRs) with 95% confidence intervals (CI) were defined for subsequent cancers of stomach, pancreas, colon, rectum, breast and prostate separately and for all cancers except stomach combined. The mean follow-up time was 10.1 years and the number of gastric cancers was 72. For the Hp+CURED, the SIR for gastric cancers for the first 5 follow-up years was 1.62 but decreased from the sixth follow-up year thereon to 0.14 (CI: 0.00–0.75). Likewise, the risk ratio, defined in a Poisson regression analysis using the Hp+NoInfo group as the reference, decreased from 1.60 to 0.13 (CI: 0.02–1.00, p = 0.049). The SIR for Hp– was not significantly higher than that for Hp+NoInfo for any of the cancers analysed. To conclude, cured H. pylori infection led to a significantly decreased incidence of gastric cancers from the sixth follow-up year. Advanced atrophic gastritis would be a plausible contributor to the elevated SIR in elderly Hp– patients.
Chronic gastritis caused by Helicobacter pylori is associated with peptic ulcer disease,1 noncardia gastric cancer,2–5 primary low-grade B-cell gastric lymphoma of mucosa-associated lymphoid tissue type (MALT lymphoma)6 and in some patients with cardia cancer in association with atrophic gastritis.7 The annual incidence of gastric cancer in patients with premalignant lesions is 0.1% for atrophic gastritis, 0.25% for intestinal metaplasia, 0.6% for mild-to-moderate dysplasia and 6% for severe dysplasia within 5 years from the baseline according to a Dutch nationwide quantification.8
Successful eradication of H. pylori infection leads to, besides healing of gastritis, permanent cure of peptic ulcer disease9 and regression of the majority of MALT lymphomas.6 Follow-up studies have also shown the regression of precancerous lesions after eradication therapy10–12 and limited success in the prevention of the development of gastric cancer,13–19 even in patients with resected early gastric cancer.20 The general problems of therapeutic trials and the relatively long time needed to collect evidence for the beneficial effect of successful therapy on carcinogenesis and gastric cancer have been reviewed by Fuccio et al.21, 22 The latter review includes a meta-analysis of 6 randomised trials aimed to reduce the risk for gastric cancer by the eradication of H. pylori infection. Some of the studies were still ongoing.
A total of 97% of the subjects with H. pylori infection have elevated levels of specific serum IgG and/or IgA antibodies when highly accurate and specific quantitative enzyme immunoassay based tests are used.23–26 Successful eradication of the infection is followed by an at least 40–50% decrease of antibody titers within 4–12 months.23–25
To study the effect of eradication of H. pylori infection on the subsequent development of gastric cancers, we compared the standardised incidence ratios (SIRs) and risk ratios (RRs) of gastric cancers between 3 groups of patients: initially H. pylori seropositive patients with serologically verified cure of infection (Hp+CURED), initially H. pylori seropositive individuals for whom no serological information on a possible cure was obtained (Hp+NoInfo) and initially H. pylori seronegative subjects (Hp–).
Material and Methods
The study material included H. pylori antibody data from 26,700 consecutive patients whose serum samples were in 1986–1998 sent to the diagnostic service laboratory at the Department of Bacteriology and Immunology, University of Helsinki (after administrative changes in 1997: Helsinki University Central Hospital Laboratory Diagnostics) for H. pylori antibody testing from several Finnish hospitals and laboratories.
The patients were divided according to the H. pylori antibody status at baseline and the possible change in it during follow-up. Firstly, there were 3,650 initially seropositive patients (mean age 56.4 years, SD 13.8 years) whose rapidly falling antibody titers indicated successful eradication therapy (Hp+CURED), including 239 patients who had not responded to previous treatment. Secondly, there were 11,628 initially seropositive patients (mean age 55.0, SD 15.3 years) where subsequent serological information regarding cure was lacking (Hp+NoInfo), including 649 patients in whom treatment failure was verified by no or minor changes in antibody titers after eradication therapy. Thirdly, 11,422 initially seronegative patients, mean age 43.7, SD 17.2 years (Hp–) were included.
Serum samples were tested for H. pylori IgG- and IgA-antibodies by an in-house enzyme immunoassay. During the study period, the sensitivity and specificity of the IgG test were 94% and 93%, respectively, and the sensitivity and specificity of the IgA test 73% and 95%, respectively. As determined in patients whose H. pylori infection had been verified by culture and histology of gastric biopsies, 3% of the infected patients had a raised IgA titer only; thus, the sensitivity of the combined IgG and IgA tests was 97%.23 Success of H. pylori eradication therapy was defined by at least a 40% decrease of the IgG titer within 6–12 months after therapy, or in patients with a raised IgA antibody titer only, by at least a 40% decline.23–25 The sensitivity and specificity of decreased IgG titers were 97–98% and 90–100%, respectively, and those of decreased IgA titers 91–99% and 90–100%, respectively.23, 25 The pre- and post-treatment samples were always run in parallel.
Data on deaths and emigrations from the national Statistics Finland data base and National Population Registry were linked to the cohort until 31st December, 2006. The accumulation of the observation years in the Hp+CURED cohort began on the actual date of the serum sample showing the success of eradication (median interval from the baseline sample 7 months); in the two other cohorts the follow-up started 7 months after the date of the baseline serum sample. The accumulation of the person-years at risk ended on the date of emigration, death, or on 31st December, 2006, whichever was first. The mean follow-up time in the Hp+CURED cohort was 10.4 years, in Hp+NoInfo 9.3 years and in the Hp– group 10.9 years.
Gastric and other cancers diagnosed until the end of 2006 were identified from the Finnish Cancer Registry. Eleven-digit personal identity codes were used in all linkages. In each cohort, the numbers of patients and person-years at risk, and observed numbers of stomach, pancreas, colon, rectum, breast and prostate cancers separately and all cancers except stomach combined were calculated, stratified by gender, calendar period (1986–2006) and 5-year age groups. Further stratification was made according to time since the beginning of follow-up (<1,1–2, 3–4, 5–9 and 10–20 completed years). The expected numbers of cases were calculated by multiplying the number of person-years in each stratum by the respective cancer incidence rate in the Finnish population, also derived from the Finnish Cancer Registry. The relative risks in the cohorts were expressed as observed to expected ratios (SIR), and exact 95% confidence intervals (CI) for the SIRs were defined assuming Poisson distribution. The information collected by the Finnish Cancer Registry has been shown to cover over 99% of all malignant solid tumors diagnosed in Finland.27
Numbers of observed gastric cancers (including non-cardia and cardia cancers) and person-years at risk were further used in a Poisson regression analysis to calculate the age and gender adjusted RRs with 95% CI of the Hp+CURED and Hp– cohorts using the Hp+NoInfo cohort as the reference. Poisson regression analysis was also applied to the numbers of gastric cancers categorised by age groups. All p values presented are two sided. Statistical analyses were carried out using the SPSS 15.0 software package (SPSS, Chicago, IL).
The study was approved by the Ethical Committee for Epidemiology and Public Health of the Helsinki University Hospital District. Permission to link pre-existing laboratory data and the data from national registries was obtained from the Ministry of Social and Health Affairs, Helsinki University Central Hospital and the Department of Bacteriology and Immunology, University of Helsinki.
A total of 26,700 patients and 296,696 person-years at risk were included in the study (Table 1). The number of subsequent gastric cancers during the follow-up was 11 in the Hp+CURED, 37 in the Hp+NoInfo and 24 in the Hp– subcohort (Tables 2 and 3). Of the total of 72 cases, 61 were noncardia gastric cancers and 11 cardia gastric cancers.
Table 1. The number of patients (n), gastric cancers (Ca) and person-years at risk by gender and completed years of follow-up in the 3 subcohorts defined by Helicobacter pylori antibody status at the baseline and the changes in it
Table 2. The number of patients with noncardia gastric cancer (nc) and cardia cancer (c) observed in different years of the follow-up period
Table 3. The number (N obs) of patients with gastric cancers (noncardia gastric cancer and cardia cancer) and the age and gender adjusted risk ratios (RR) with 95% confidence intervals (95% CI) by Helicobacter pylori antibody category and completed years of follow-up time
For the Hp+CURED gastric cancers (cardia cancers included), the SIR for the first 5 follow-up years was 1.62 (CI: 0.78–2.98) but decreased from the sixth follow-up year thereon to 0.14 (CI: 0.00–0.75). The person-years at risk were rather evenly distributed (46.5 and 53.5%, respectively) between these follow-up periods (Table 1). For the total follow-up the SIR was 0.81 (CI: 0.40–1.45) (Table 4). For noncardia cancers in the Hp+CURED the SIRs were 1.56 (CI: 0.67–3.07), 0.16 (CI: 0.00–0.92) and 0.80 (CI: 0.37–1.52), respectively.
Table 4. The observed and expected numbers of patients with cancers of stomach, pancreas, colon together with rectum, breast, and prostate separately and all cancers except stomach combined and the age and gender adjusted standardised incidence ratios (SIR) with 95% confidence intervals (95% CI) by Helicobacter pylori antibody category for the total follow-up time
For the total follow-up time, the age and gender adjusted RRs of gastric cancers of the Hp+CURED and Hp– cohorts were slightly below the reference value 1 of the Hp+NoInfo cohort (Table 3).
During the first 5 years of follow-up, the RR of gastric cancers for the Hp+CURED group was 1.60 (CI: 0.75–3.45) as compared to the Hp+NoInfo group. From the sixth year thereon only 1 case of gastric cancer (a noncardia cancer) was found among the Hp+CURED group (RR 0.13, CI: 0.02–1.00, p = 0.049) (Tables 2 and 3). During this particular period, a similar trend was also seen when the RRs for noncardia cancers of the Hp+CURED and Hp+NoInfo groups were compared (0.15, CI: 0.02–1.13, p = 0.066) (data not shown). The RR of gastric cancers for the Hp– was close to unity throughout the follow-up period (Table 3).
In the age groups of <50, 50–69 and ≥70 years, the SIRs increased by age in the Hp– (0.51, CI: 0.25–5.57, 1.01, CI: 0.51–2.14 and 1.34, CI: 0.77–2.35, respectively), whereas they decreased in the Hp+NoInfo group (6.58, CI: 2.80–14.42, 1.03, CI: 0.58–1.85 and 0.78, CI: 0.47–1.30, respectively). The numbers of gastric cancers in these subcohorts showed a significant interaction between the age groups and H. pylori antibody status (Poisson regression analysis, RR 12.77, CI: 1.57–103.78; p = 0.017); this indicates that the cancer patients in the Hp– cohort were older than those in the Hp+NoInfo cohort. (Data not shown.)
The SIRs of subsequent cancers of pancreas, colon, rectum, breast, prostate separately and all cancers except stomach combined were determined to analyse the risks of other than stomach cancers. In all the 3 subcohorts, all the SIRs for the whole follow-up period were either on the level of or above the national background (Table 4) and the same was true both for the first 5 years of the follow-up and the later years (data not shown). The SIR for Hp– was not significantly higher than that for Hp+NoInfo for any of the cancers analysed.
Since the discovery of the association of H. pylori infection with gastric malignancies in the early 1990s,2–6 there has been a need to define the role of eradication therapy in the prevention of these malignant processes. In our study, a markedly lower SIR of noncardia gastric cancers was found from the sixth follow-up year on among patients successfully treated for H. pylori infection (Hp+CURED) as compared with the SIR for the first 5 years. Beginning from the sixth follow-up year also a markedly lower age and gender adjusted RR of gastric cancers (including cardia cancers) was detected in the Hp+CURED cohort compared with H. pylori positive subjects where information on eradication of the infection was lacking (Hp+NoInfo). This long time interval from the commencement of treatment to a demonstrable preventive effect implies that the point of no return in the development of gastric cancers must have been a few years before the day the malignancy was diagnosed.
Our findings are consistent with the recent reports based on gastroendoscopical findings, which indicate lower gastric cancer rates in patients with a confirmed cure of H. pylori infection than in those who had not been treated or who had failed eradication therapy.13–17, 20, 28 In all these studies, several years of follow-up were needed to confirm that successful eradication therapy significantly prevents the development gastric cancers. In the randomized, placebo-controlled, population-based study Wong et al.13 found during 7.5 years no gastric cancers after successful eradication therapy in 485 patients without premalignant lesions (atrophic gastritis, intestinal metaplasia or dysplasia) at baseline, but 6 cases appeared among 503 control patients (Kaplan-Meier analysis, log rank test, p = 0.02). However, in patients with precancerous lesions at baseline and in the whole study material, this difference was not significant. Further proof of the posteradication cancer risk associated with premalignant lesions has been provided by Kamada et al.,29 Yanaoka et al.14 and de Vries et al.15 Kamada et al.29 carried out a prospective study on the baseline features in patients who despite successful eradication therapy developed gastric cancer within 9 years. Of the total of 1,787 patients all the 20 cancer patients had been diagnosed to have severe atrophic gastritis in the corpus at the baseline.29 Yanaoka et al.14 conducted a cohort study in asymptomatic H. pylori infected males categorized in groups with and without extensive atrophic gastritis by serum pepsinogens (in the latter group pepsinogen I was > 70 ng/ml and pepsinogen I/II >3.0). Eradication of the infection resulted in a significant reduction in cancer incidence within the mean observation period of 9.3 years in those without extensive atrophic gastritis (2 of 318 vs. 3 of 155 in men with extensive atrophic gastritis, p < 0.05).
de Vries et al.15 described two patients with gastric ulcer and intestinal metaplasia; despite successful eradication therapy, dysplasia developed in both of them, and gastric cancer in one of them after 4 and in the other one after 14 years. In a retrospective Japanese multicenter study, gastric cancer developed within a median follow-up period of 5.9 years in 1% of 1,788 patients with eradicated infection compared with 4% of 1,244 patients with persisting infection within 7.7 years (OR, 0.36, CI: 0.22–0.62).16 Takenaka et al.17 enrolled 1,807 patients to receive eradication therapy. At the latest follow-up endoscopy 12–118 months later, they recorded 6 cases of gastric cancers among 1,519 patients with a successful eradication of H. pylori, which was significantly less than the 5 cases found among 288 patients with failed eradication therapy (Kaplan-Meier analysis, log rank test, p < 0.01, median follow-up times 39 and 35 months, respectively). In a report published by Ogura et al.18 gastric cancer developed in 13 of 304 infected patients with persisting infection, but in only 6 of 404 patients with cured infection within slightly over a 3.2-year time period (Kaplan-Meier analysis, log rank test, p = 0.02). Two studies have been limited to H. pylori infected patients with peptic ulcer disease.19, 28 Take et al.19 found all cancers in patients with gastric ulcers; 8 of the 592 gastric ulcer patients cured of the infection developed gastric cancer during the follow-up extending to 8.6 years (mean 3.4 years), while significantly more gastric cancers were found among the patients with persisting infection (4 of 111; Kaplan-Meier analysis, log rank test, p = 0.04). Wu et al.28 gave eradication therapy to patients hospitalised for peptic ulcer disease; 54,576 patients were treated early after the admission (median 14 days from it) and 25,679 patients at least a year later (median 1,053 days after the baseline admission). The success of the therapy was not confirmed. The follow-up was extended up to 10 years (mean 5.9 and 7.2 years for the early and late subgroups, respectively); cancer cases recorded within the first 2 years from the baseline admission were excluded from the analysis. Early treatment was associated with lower hazard ratio (HR) for gastric cancers than the late one (HR 0.77, CI: 0.60–0.99, p = 0.038).
In the meta-analysis of 6 randomised studies (including 3 studies referred here10, 12, 13) Fuccio et al.22 calculated that eradication of the infection lowered the RR for gastric cancer to 0.70 (CI: 0.46–1.08; corrected values, see Ref. 22) in subjects followed for 4–10 years.
The detection of H. pylori antibodies offer a sound and proven method for epidemiological studies when highly sensitive and specific enzyme immunoassay based tests are applied. Our in-house test has filled these requirements both in primary diagnosis and post-therapy follow-up.21, 23 Although in the present study the proportion of infected patients with a confirmed cure of the infection was limited to one-fourth of all H. pylori positive ones, the incidence of subsequent gastric malignancies in this cohort followed a pattern significantly different from that seen in the Hp+NoInfo and Hp– cohorts. It cannot be ruled out that some patients in the Hp+NoInfo cohort may have been treated successfully but not followed-up serologically. However, the number of treated subjects is likely to be small, since there were no generally accepted international or local recommendations for the treatment of infected patients before 1997, when the first European guidelines on the management of H. pylori infection were published by the European Helicobacter pylori Study Group (The Maastricht Consensus Report).30
In the Finnish population, subjects from lower social classes have been shown to have a higher incidence of noncardia gastric cancer.31 Considering that H. pylori infection is the main risk factor of gastric cancer, these data are in line with the finding that low socioeconomic status is a general risk factor for H. pylori infection.32 In our results, the role of social class would be critical if the acceptance and the use of eradication therapy were dependent on socioeconomic situation and therefore would have differed between the Hp+NoInfo and Hp+CURED cohorts. This does not seem to be likely, as, to reduce the influence of poverty on the level of healthcare, all Finnish patients are entitled to low-cost municipal health services and nearly 50% reimbursement of the costs for prescribed medicines. Furthermore, the population has remained ethnically homogenous without significant immigration from any subgroups or minorities during the study period.
The Hp-seronegative cohort consists of two main subgroups of patients: those who have never been infected with H. pylori and those who have been infected but who have lost the infection spontaneously, due to antimicrobial therapy or advanced atrophic gastritis. A third minor subgroup, 2–3% of Hp infected patients, do not produce circulating IgA or IgG antibodies detectable by the current routine methods.23 The incidence of gastric cancers observed in Hp– patients was as high as that of the Hp+NoInfo cohort. In this respect, the results of our patient material, tested for H. pylori infection for clinical indications, differ strongly from the epidemiological findings presented previously, for example, in the meta-analysis of serological case–control studies that revealed an approximately 6-fold risk of seropositivity for subsequent noncardia gastric cancer in comparison with seronegativity.5 The equally high RR of gastric cancers we observed in Hp– and Hp+NoInfo patients could be partly caused by the fact that some Hp+NoInfo patients may have received eradication therapy but not been tested later serologically for changes in the level of antibodies. Seronegativity at the baseline does not, however, exclude the role of H. pylori infection as a risk factor in the Hp– subcohort. The older the subjects are when initially tested negative for H. pylori infection, the more likely the Hp– subcohort will include subjects who initially have been infected, but due to advanced atrophy, their gastric mucosa does not support the growth of H. pylori any longer and even the longest lasting signs of the infection, H. pylori antibodies, have disappeared.33–35 Actually, both negative and positive H. pylori antibody tests in combination with low serum pepsinogen I levels (and low pepsinogen I/II ratio) indicating atrophic gastritis have been shown to be associated with an equally high RR for gastric cancer.36, 37 The with age increased incidence of gastric malignancies observed among our Hp– patients is biologically consistent with the possibility that many of these gastric cancers could have been associated with advanced atrophic gastritis. This would be in line with the chain of events leading from chronic helicobacter gastritis via the precancerous lesions atrophic gastritis, small intestinal metaplasia, colonic metaplasia and dysplasia to a gastric cancer of the most common type, the intestinal adenocarcinoma, presented by Correa.38
To study the possibility that the relatively high SIR for gastric cancers in the Hp– subcohort could be explained by factors (e.g. environmental) other than H. pylori infection, we analysed the SIRs for some common cancers of the gastrointestinal tract (pancreas, colon and rectum), some common cancers outside the gastrointestinal tract (breast and prostate) and all cancers except stomach combined. The SIR for Hp– was not significantly higher than that for Hp+NoInfo for any of the cancers analysed, which is consistent with the role of past H. pylori infection in Hp– gastric cancers. Interestingly, for several cancers the SIRs were clearly above the population level at each follow-up phase irrespective of the subcohort. This may be associated with the specific quality of the patient material, which is likely to consist of subjects who have come to clinical investigations for a wide variety of mainly gastrointestinal symptoms and many of them at an age when malignant neoplasms are not exceptional. It is of note that in the Hp+CURED subcohort the only SIR lower than background value for any follow-up phase and for any of the cancers analysed was found for gastric cancers. This is a further indication of the specific preventive effect on gastric cancers achievable by eradication of H. pylori infection.
To conclude, the present study and the others collectively provide evidence that eradication of H. pylori infection will prevent later development of gastric cancers, but the benefit is seen only after a delay of several years. This implies that at some stage the carcinogenous process becomes independent of helicobacter infection and thus, uninfluenced by eradication therapy. Until the point of no return in the carcinogenesis has been well-defined, it seems advisable to treat patients infected with H. pylori, to ensure successful eradication and, according to risk estimates, arrange further follow-up for those with precancerous lesions.