The risk of stomach cancer in patients with gastric or duodenal ulcer disease

Authors: Hansson L-E, Nyren O, Hsing AW, Bergstrom R, Joseffson S, et al.
Source: New England Journal of Medicine. 335:242-9. July 25, 1996.
Institutions: Uppsala University and Mora Hospital, Sweden; National Cancer Institute, International Epidemiology Institute and Harvard School of Public Health, USA.
Financial support: National Cancer Institute.

Summary

Background

Helicobacter pylori infection is associated with gastritis, gastric ulcer, duodenal ulcer and gastric cancer. There has been some evidence, however, that although gastric ulcer disease is associated with an increased risk for gastric cancer, this relationship does not hold between duodenal ulcer and gastric cancer. This study was designed to clarify the relationship between gastric ulcer, duodenal ulcer and gastric cancer.

Methods

In this retrospective cohort study from Sweden, a group of patients with documented gastric or duodenal ulcer disease was selected using hospital discharge data. This group was then analyzed for the subsequent development of gastric cancer. The incidence of cancer was then compared to that of the Swedish population at large.

Results

The cohort consisted of 29,287 patients with gastric ulcer only, 24,456 with duodenal ulcers and 4.193 with both (analyzed separately). Endoscopy was performed in about 20% of patients between 1965 and 1975, in about 40% between 1976 and 1983.

Authors' Discussion

The main finding of the study was the divergent effect of gastric and duodenal ulcers on the incidence of gastric cancer. The authors note that although cohort studies such as this are subject to problems of selection bias (a difference in the population sampled and the population to which it is compared), any selection bias operating here would have to explain the difference they found between gastric and duodenal ulcers. This difference is less likely to occur purely as a result of selection bias.

The authors also note that H. pylori infection is present in nearly all patients with duodenal ulcer, and is a causative factor in the pathogenesis of gastric cancer. They conclude that "some processes at work in patients with duodenal ulcers may profoundly modify the carcinogenic effect of H. pylori infection".

Editorial

In an editorial, Dr. Julie Parsonnet from Stanford provides a plausible explanation for the paradox that is revealed here - the reduction in the incidence of gastric cancer in patients with duodenal ulcer despite the fact that H. pylori is a causative agent of both. H. pylori leads to gastric cancer by the development of atrophic gastritis, which causes a decrease in gastric acid output. Since "no acid, no duodenal ulcer", those patients who develop atrophic gastritis are less likely to have duodenal ulcers but more likely to develop gastric cancer.

Why some patients infected with H. pylori progress to duodenal ulcer and some to atrophic gastritis may be related to the age at which the infection is acquired. Acquisition in childhood is more likely to lead to gastritis, later infection more likely to lead to duodenal ulcer. This argument is expounded in greater detail in the editorial, and makes for fascinating reading. Dr. Parsonnet also alludes to the problem of gastric acid suppression and infection with H. pylori, which is the subject of another article reviewed here.

Comment

This study adds another piece to the emerging story of H. pylori and GI pathology. H. pylori infection contributes to atrophic gastritis, gastric ulcer disease, gastric cancer and duodenal ulcer. The development of atrophic gastritis may protect against duodenal ulcer disease, however, which would explain the negative correlation found here between duodenal ulcer and gastric cancer. The editorial by Dr. Parsonnet helps put the results in perspective and provides a very plausible explanation for them.

Comparisons between groups that are not randomly selected present a significant potential for unrecognized differences between the groups and/or confounding. As the authors of this study point out, however, the main comparison is not really between the ulcer cohort and the population at large, but between the two different types of ulcers within the ulcer cohort. This reduces the potential for confounding and bias.

This sort of study also makes clear the power of a centralized health care system such as Sweden's when it comes to gathering and analyzing epidemiologic data.

8/20/96


Reader comments

I received the following letter from Dr. Olof Nyren, co-author of this study:


UPPSALA UNIVERSITET
September 5, 1996

Dear Dr Jacobson:

Thank you for your letter and for the flattering attention to our recent article on the risk of stomach cancer in patients with gastric and duodenal ulcer disease. Your summary is well written and in good agreement with what we wished to convey in our article. I have only a few comments:

Firstly, in your first paragraph it is stated that gastric ulcer disease is known to be associated with an increased risk of gastric cancer. This is not entirely correct. Although it is quite clear that many gastric cancers are initially misinterpreted as benign gastric ulcers, there has been no firm proof of a long-term risk of gastric cancer among gastric ulcer patients. Our study was the first to demonstrate such an association.

Secondly, the rationale for our study was the following: Since Helicobacter pylori is considered to be a causal agent both in duodenal ulcer and gastric cancer, one would expect at least a positive association between the two diseases. Any other finding would require a reconsideration of the role of Helicobacter pylori in either of the diseases.

The third comment that I would like to make relates to the incidence peak in the first years after entry in the cohorts. Although earlier detection could be one contributory explanation, the most important factor is probably selection bias. Such bias occurs when patients with peptic ulcer and yet unrecognized subclinical cancer are more likely to be admitted to hospital care compared to peptic ulcer patients without such latent cancer.

My fourth point relates to the comment by Julie Parsonnet: Although she provides good arguments for considering timing of the infection as the critical modifier of the gastric cancer risk, this presupposes that the long-term course of the infection is quite different in those early infected compared to the course in subjects who contract the infection in adult life. If it is only a matter of duration of the infection, then we would expect to find a successively increasing risk among very old duodenal ulcer patients with long duration. We did not find any indication of such a pattern in our data.

Again, thank you very much for your kind letter and for advising me about the very interesting Journal Club on the Web.

Yours, sincerely,

Olof Nyren
Department of Cancer Epidemiology


December 12, 1996

Letters to the Editor about this article from the NEJM website.

In particular, more about Dr. Parsonnet's editorial on the timing of H. Pylori acquisition and its role in gastric cancer vs. duodenal ulcer. 


Date: Mon, 23 June 1997
From: crespi <roma99chair@uni.net>

Dear Doctor Jacobson,

I would like to comment further on the article by Hansson LE et al and on the comment about the editorial by Doctor Julie Parsonnet. In a previous article (Scand J Gastroenterol 1996; 31:1041-1046) we already highlighted several inconsistencies of the studies trying to establish Hp as the main causative agent of stomach cancer. We well know and admire the efforts by Dr. Parsonnet to explain the rising evidence casting doubts on the prevailing role of Hp in carcinogenesis.

In fact, her last hypothesis is, again, flawed by a review of existing data on the prevalence of infection in young age groups and mortality for gastric cancer in several countries. As may be seen from the Table below, the prevalence of Hp infection before the age of 20 (and in some cases, before the age of 5), ranging from 4.2% in Belgium to 83% in India, doesn't correlate at all with cancer mortality rates found in all the listed countries. In addition, the similar prevalence of Hp infection in both sexes doesn't correlate with the gender differences in cancer mortality.

The listed countries include areas with medium/low to high life expectancy to avoid the bias of low frequency of cancer due to premature death. Data on Hp prevalence are drawn from published studies on asymptomatic subjects, whereas the gastric cancer mortality was obtained from the IARC.

Yours sincerely,

Massimo Crespi
Francesco Citarda
National Cancer Institute "Regina Elena"
Roma, Italy
 
Prevalence of Hp infection in young age and gastric cancer
 
% HP prevalence

(< 20 years)

Gastric cancer

mortality (x 100.000)

   
Males
Females
Australia
14.3
9.6
4.2
Bangladesh
48.0 (< 5 years)
2.6
1.7
Belgium
4.2 to 7.0
14.4
7.5
China (Guangdong)
23 (< 5 years)
11.4
5.1
China (Henan)
85.6
29
15
Colombia
69.0 (< 10 years)
23.1
14.6
Ethiopa
60.0 (< 5 years)
12.0
11.4
Finland
10.2
17.2
9.9
Gambia
50 (< 5 years)
2.6
0.9
Germany
21.0
18.5
10.5
Iceland
10.2
25.5
7.3
India
83.8
9.1
4.7
Italy (Puglia)
28.3
14.6
10.5
Italy (Sardinia)
34.0
15.7
10.8
Nigeria
50.0 (< 5 years)
4.7
3.7
Panama
82.0
12.3
4.2
Sweden
7.0
11.6
6.5
Turkey
41.7
10.0
9.4
USA
31.0
5.3
2.7


Journal Club on the Web main screen

Submit a comment about this article

Site Meter