A prospective study to determine if Helicobacter pylori infection is strongly associated with gastric and duodenal ulcers

Reference:  Schottker, B., et al., Helicobacter pylori Infection Is Strongly Associated With Gastric and Duodenal Ulcers in a Large Prospective Study. Clin Gastroenterol Hepatol, 2012. 10(5): p. 487-493 e1.

A study out of Germany on the effect of Heliocobacter pylori Infection in patients with either gastric or duodenal ulcers. 

Brief Summary:   It is known that Heliocobacter pylori infection (a type of bacteria which infects the stomach) is a risk factor and strongly associated with peptic ulcer disease (PUD), but it is un-known if it is associated specifically with gastric and duodenal ulcers – two specific sub-types of peptic ulcer disease (PUD).  The main objective of this prospective study was to determine the degree of association between Heliocobacter pylori infection with both gastric and duodenal ulcer disease separately.  Furthermore, patients with Heliocobacter pylori infections were further delineated between those which had expression of the cytotoxin-associated gene A (cagA+) and those that did not express the gene (cagA-).  A series of demographic factors along with whether the patient had an infection of Helicobacter pylori was collected over a period of 2 to 5 years from a German population of about 10000 individuals.   The authors also conducted a cross-sectional study to identify the risk factors of lifetime history of peptic ulcer disease.  

Results: For those patients who had a cagA positive strain (cagA+) and a Helicobacter pylori infection, there was a significant association between the presence of both the gastric and duodena ulcers separately and the Heliocobacter infection.  For those patients who did not have the cagA positive strain (cagA-) and a Heliocobacter pylori infection, the association with a Heliocobacter infection was not significant.

 

Implications For Practice: Heliocobacter infection is a strong risk factor for both gastric and duodena ulcers – especially if the patient has a positive expression for the cytotoxin-associated gene (cagA+).

Discussion: Really great investigative work here – especially the study design which I will get to in a bit.  From a biological perspective, I wonder why the patients which had an expression for the cytotoxin-associated gene A also had a much higher likelihood of association than those which did not.  Obviously, since I’m not the main author and it’s hard for me to find time to divulge myself into an investigation like this, questions like this are probably straight forward for the authors (but not to me).

Also, curious   

Commentary and Statistics and Design:  aww man, where do I start?  I really like the statistical analysis and study design of this investigation.  First, the authors were correct in the statement that prospective studies are better able to detect a true causal relationship for a given observational factor than a cross-sectional study.  The reason being is that prospective studies are better able to delineate out any potential confounding factors by the means of excluding certain patients who could confound the investigation in the first place.  This is especially true with medical based studies and is the reason why prospective studies should be chosen over cross-sectional studies for the purpose of identifying risk factors of a given disease.

           Also, I liked how the authors performed both a cross-sectional and longitudanal study with essentially the same data.  Often times in the search for risk factors, the authors will only conduct a prospective study for the risk factors, but there is a-lot of other useful information that the authors could extract initially with a cross-sectional study.    

Also, I liked how the authors included both the results from the logistic regression in both the univariate and multi-variate perspective.  Often times, I see authors include results only from the univariate perspective, but it is often best to report the results of both, because it gives a reader a sense of how the associations change from non-control to a control based setting based on the other variables.   Also, the tables were made very nicely for visualization by the reader.    The authors definitely did their job in searching for any and all potential variables to include in the model – good job here. 

My only question is the following.  In the cross-sectional study, why did the authors decide to separate the categorical and continuous variables for the use of the full multi-variate model?  I couldn’t find an answer from this in the paper, but I’m sure the authors have good reason.  I see no reason why the full model could not have included both. 

A big beer at the Hofberhaus for our German friends!  This paper will definitely be one of the references I go back to when it comes to statistics and study design.