None of the previous studies evaluated patients with duodenal or gastric ulcers before they were diagnosed or hospitalized with their disease. The usual temporal sequence is that patients with peptic ulcer disease are examined to detect H. pylori infection at the time the diagnosis is made or after they have had treatment to eradicate the organism. Consequently, peptic ulceration could possibly predispose persons to colonization by H. pylori [9].

We did a prospective study in a large population-based cohort using serum specimens that were obtained from study participants before they were diagnosed with peptic ulcer. We tried to determine whether H. pylori infection, as shown by the presence of specific IgG antibodies, is a risk factor for the subsequent development of either duodenal or gastric ulceration.

Methods

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Study Population

Japanese-American men born from 1900 to 1919, who were identified by the Honolulu Heart Program in 1965 by using the comprehensive 1942 Selective Service draft registration files [10], composed the study population. Of 11 148 eligible men, 8006 (72%) were interviewed and examined from 1965 to 1968, 180 (2%) died before they could be examined, and 2962 (26%) did not participate in the program. Study participants ranged in age from 45 to 68 years. The data collected included birthplace, marital status, history of alcohol use, blood pressure, and body mass index (the weight in kilograms divided by the square of the height in meters). Serum cholesterol values were determined by the Auto Analyzer N-24A method, and serum glucose values were determined by the Auto Analyzer N-2B method 1 hour after a 50-g glucose load had been given [11].

A total of 7498 (94%) men returned for a second examination between 1967 and 1970, and a serum specimen was obtained at this time. Serum specimens for a 20% random sample of the men were sent to the U.S. Public Health Service Hospital in San Francisco, whereas specimens for the remaining 5924 men were stored at –20°C at the study site. Four hundred eighty-one patients with previous gastrectomy or a previous diagnosis of peptic ulcer disease were excluded from the study. The average age of the remaining 5443 patients at the time of their second examination was 56.6 years.

Surveillance Methods

Surveillance of the cohort to identify incident patients with peptic ulcer was done by a continuous review of discharge records of all general hospitals on Oahu. Based on a 19-year follow-up survey of the study patients from the time of their examination in 1965 to 1968, only 1.3% of the men could not be located on Oahu. Thus, surveillance was nearly complete.

Two hundred fifty-eight patients were hospitalized with peptic ulcer disease from 1968 to 1989. One hundred sixty (62%) patients had their diagnosis confirmed by examination of tissue obtained by either surgery or biopsy, 36 (14%) were diagnosed by radiologic examination, and 62 (24%) were clinically diagnosed based on the endoscopic or surgical report of the presence of an ulcer.

One hundred sixty-nine incident patients had gastric ulcer, 73 had duodenal ulcer, and 16 had gastric and duodenal ulcers. Seventeen of the 169 patients with gastric ulcer, 6 of the 73 patients with duodenal ulcer, and 2 of the 16 patients with gastric and duodenal ulcers were removed from study because they had an insufficient amount of serum in the freezer repository.

Selection of Controls

Each of the remaining patients was matched with one control from the study cohort based on age at examination (47 to 70 years) and date of serum collection. If a potential control had a diagnosis of gastric cancer before or after the serum was obtained, he was excluded from the study. As a consequence, 160 patients (3.1%) were removed from the control pool of 5185 men because of the reported association between H. pylori infection and gastric cancer [12, 13]. Of the remaining 5025 men, 336 (6.7%) were excluded because they previously had cardiovascular disease or other cancer, and 1532 (30.5%) were excluded because they were diagnosed with cardiovascular disease or other cancer after their serum collection. This exclusion was done because the serum specimens from these patients are going to be used for other studies. A total of 3157 patients remained in the pool of controls from which 233 (7.4%) were matched to incident case-patients with peptic ulcer. Each control participant was alive at the time of hospitalization of the matched case-patient, so that death was not a competing factor.

Serologic Testing

The presence of serum IgG antibodies to H. pylori was determined by enzyme-linked immunosorbent assay (ELISA), using the Pyloristat kit (Whittaker Bioproducts, Inc., Walkersville, Maryland). As validated by the manufacturer, the results of this assay closely mirror those of a previously described IgG ELISA [13-18]. In brief, serum specimens from patients were diluted 1:20 for use in the kit, and IgG levels of sera were determined according to the manufacturer's instructions. The serum specimens were coded so that the laboratory technician could not distinguish case-patients from controls. A ratio of 1.00 or greater was considered positive, a ratio of less than 0.80 was considered negative, and a ratio of 0.80 to 0.99 was considered equivocal, as calibrated in the kit. Two patients with gastric ulcer and two patients with duodenal ulcer or their controls had equivocal values, so they were excluded from the study.

Statistical Analysis

Top Methods Statistical Analysis Results Discussion Author & Article Info References

A matched case–control study design was used to identify the patients and controls for serum tests. As a result, odds ratios, based on the results of the H. pylori IgG antibody test, were determined using conditional logistic regression methods [19]. When odds ratios were indeterminable, approximate confidence intervals (CIs) were determined by the method of Breslow and Day [19]. Tests for trend in the logit of risk were derived from conditional logistic regression models by using grouped H. pylori test results (coded as 1, 2, 3, and 4). All conditional logistic regression models were fitted using iterative maximum likelihood methods and a special application of the proportional hazards regression model [20].

Results

Top Methods Statistical Analysis Results Discussion Author & Article Info References

The characteristics of the 229 patients with peptic ulcer and of their matched controls are presented in Table 1. The two groups of men were similar with respect to demographic characteristics and laboratory values.

The mean age at diagnosis was 67.5 years (range, 52.7 to 87.2 years) for the 150 patients with gastric ulcer, 64.5 years (49.4 to 80.0 years) for the 65 patients with duodenal ulcer, and 63.5 years (52.4 to 83.1 years) for the 14 patients with gastric and duodenal ulcers.

Table 2 shows the association between the H. pylori test result and peptic ulcer by specific type. Ninety-three percent (139 of 150) of the patients with gastric ulcer and 78% (117 of 150) of the matched controls had a positive H. pylori-specific IgG antibody level, yielding an odds ratio of 3.2 (P = 0.001). (The odds ratio is determined by dividing 32 ± pairs by 10 -/+ pairs.) Ninety-two percent of the case-patients with duodenal ulcer and 78% of the matched controls had a positive test result, yielding an odds ratio of 4.0 (P = 0.03). If a patient had either a gastric or duodenal ulcer, then the odds ratio was 3.4 (P = 0.0001). Only 14 patients had both gastric and duodenal ulcers. The odds ratio was 1.3 (P > 0.2), based on these patients.

Because some of the hospitalized patients with peptic ulcer may have developed their ulcers after the use of adrenocorticosteroids or nonsteroidal anti-inflammatory drugs, we systematically reviewed the hospital records of the patients. A similar review could not be done for the controls because they were not hospitalized. We found that 18 of the 150 patients with gastric ulcer and 9 of the 65 patients with duodenal ulcer had taken these medications at the time of hospitalization. If these patients and their matched controls were excluded, the odds ratio would have been 3.3 (95% CI, 1.6 to 7.0) for gastric ulcer, 3.3 (CI, 0.9 to 12.1) for duodenal ulcer, and 3.3 (CI, 1.7 to 6.4) for either gastric or duodenal ulcer.

Of the 215 patients with either gastric or duodenal ulcer, 100 also had a hospital-based diagnosis of cardiovascular disease or cancer. If these patients were excluded from the study, as were controls with these diseases, 115 patients with gastric or duodenal ulcer would have remained in the study. Ninety-four percent of these case-patients and 77% of their controls had a positive test result for H. pylori antibody. The odds ratio for peptic ulcer was 4.3 (CI, 1.8 to 10.5) in this group of patients.

When the patients with gastric ulcer who had positive antibody results were separated into tertile groupings, based on the distribution of antibody levels among men who were controls, a trend (P < 0.001) was noted in the odds ratios (Table 3). The same analysis was done for the patients with duodenal ulcer Table 4 and a linear trend was also found (P = 0.014). The highest tertile group had an odds ratio of 6.8 when compared with patients who had a negative test result for H. pylori antibody.

The median and mean time intervals from phlebotomy to hospitalization were 10.7 and 10.2 years, respectively, for the 150 patients with gastric ulcer and were 8.5 and 9.0 years, respectively, for the 65 patients with duodenal ulcer. We did an analysis to determine whether the association of antibody positivity with ulcer was affected as the time interval from phlebotomy to hospitalization increased (Table 5). Patients diagnosed 0 to 9 years after phlebotomy or 10 or more years after phlebotomy had a statistically significant increased risk for gastric ulcer. For duodenal ulcer, a statistically significant association was observed mainly for patients hospitalized 10 or more years after phlebotomy.

Discussion

Top Methods Statistical Analysis Results Discussion Author & Article Info References

Our study sought to determine whether H. pylori infection preceded the diagnosis of duodenal or gastric ulcer in a cohort of Japanese-American men. Although the results are affirmative, several limitations in our study design should be considered.

First, this study identified hospitalized patients with peptic ulcer but did not identify nonhospitalized patients who were diagnosed during the surveillance period. Because of practical considerations, surveillance was limited to only the review of hospital records of study participants and did not include other sources of medical records. If nonhospitalized patients with peptic ulcer are included in the control group and if a positive association with H. pylori infection also exists among these patients with less complicated disease, then the findings of our study underestimate the strength of the association between peptic ulcer disease and previous infection with H. pylori.

Second, some of the patients may have been diagnosed with peptic ulcer before they were hospitalized with their condition. This would likely apply more to patients with duodenal ulcer than to those with gastric ulcer. However, each hospital record was carefully reviewed and the date of earliest diagnosis of peptic ulcer, even without hospitalization, was recorded. As a result, 481 patients (170 patients with peptic ulcer who did not have gastrectomy and 311 who had had a gastrectomy) were excluded from the study because their condition existed before study entry. This is standard procedure for cohort studies, which identify characteristics that precede the appearance of the disease under investigation [21].

Third, we could not completely assess the effects of the use of adrenocorticosteroids or nonsteroidal anti-inflammatory drugs on the diagnosis of peptic ulcer in this cohort because information on the use of these medications was not available for controls. However, exclusion of those patients whose hospital records indicated that they had recently used the medications did not substantially change the results shown in Table 2. These findings are consistent with the observation that the use of nonsteroidal anti-inflammatory drugs does not increase susceptibility to H. pylori infection [22].

Lastly, although 37% of the potential controls in our study were excluded because of cancer or cardiovascular disease, their exclusion did not appreciably affect the findings. Some of the serum specimens had been thawed previously for use in other studies, but we reported previously that multiple cycles of freezing and thawing do not affect the results of H. pylori serologic determinations [23].

More patients had gastric ulcer than had duodenal ulcer in this investigation. This predominance of gastric ulceration among Japanese-Americans has been shown in an earlier study [24], which also reported that the average age at diagnosis was higher among patients with gastric ulcer (53.8 years) than among those with duodenal ulcer (48.3 years). This difference would favor the diagnosis of more patients with gastric ulcer in our study because the average age of the patients at study entry was 56.6 years (age range, 47 to 70 years).

Our finding that H. pylori infection preceded the diagnosis of peptic ulcer is strengthened by the observation that patients hospitalized 10 or more years after phlebotomy still showed a statistically significant association (Table 5). This supports the view that infection with H. pylori contributes to the clinical manifestation of peptic ulcer and makes it less likely that ulceration predisposes patients to colonization by H. pylori. As suggested by Rabeneck and Ransohoff [25], establishing that H. pylori infection precedes duodenal ulcer makes it less plausible that H. pylori is a commensal organism. Based on our findings, 44% of the total risk for hospitalization with either gastric or duodenal ulcer can be attributed to a previous H. pylori infection [26].

An increase was noted in the risk for peptic ulcer with increasing antibody levels for H. pylori, especially for duodenal ulcer. In vitro studies indicate that soluble H. pylori products recruit [27] and activate macrophages [28]. The height of the IgG immune response may be related to the extent of inflammation in the gastric mucosa or of metaplastic gastric epithelium in the duodenum [29]; thus, the intensity of inflammation might contribute to the risk for ulcer development. In an analysis of gastric cancer in the same cohort of men [13], high antibody levels also were associated with increased risk; thus, the results of these two studies are parallel.

Despite the evidence that an H. pylori infection precedes the diagnosis of peptic ulcer, we found that 77% of the population-based controls in this study also were infected with this organism. Although 12.5% of the 5924 participants in the study have been diagnosed with peptic ulcer disease (481 before study entry and 258 after study entry), most of the H. pylori-infected participants will probably never develop peptic ulcer. These data indicate that other factors in addition to H. pylori infection have an important role in the development of peptic ulcer. Duodenal ulceration may ultimately depend on the synergy between H. pylori infection and the presence of gastric metaplasia in the duodenum [30]. The organism does not colonize the small intestinal mucosa but has been identified in areas of gastric metaplasia within the duodenum [31, 32].

Helicobacter pylori is the major causative agent for chronic superficial (antral and fundic) gastritis [7, 8]. The presence of this type of gastritis is strongly associated with the presence of duodenal and gastric ulcers [33]. In a recent prospective study in Finland, Sipponen and colleagues [34] found that the evidence of chronic gastritis in biopsy specimens was associated with a 14-fold increase in the risk for peptic ulcer during the next decade. That study provides further evidence that pre-existing H. pylori infection is a risk factor for peptic ulceration through its effects leading to chronic gastritis.

Our study is consistent with other reports [2-5, 7, 35] and shows for the first time that H. pylori infection is associated with an increased risk for the subsequent development of peptic ulcer disease. This association is probably mediated at least in part by the chronic inflammation induced by H. pylori [36]. Whether this process is attributable to direct bacterial injury [37, 38], facilitation of mucosal injury by host aggressive factors [33], host immune and inflammatory response to the organisms [27, 28], or a combination [36] remains to be determined.