Rapid Responses to:

PAPERS: Paul Moayyedi, Shelly Soo, Jonathan Deeks, David Forman, James Mason, Michael Innes, and Brendan Delaney Systematic review and economic evaluation of Helicobacter pylori eradication treatment for non-ulcer dyspepsia BMJ 2000; 321: 659-664 [Abstract] [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Why no remission in the Markov model?

Graham Byrnes   (18 September 2000)

Rapid response-Helicobacter eradication in non-ulcer dyspepsia

A Poullis   (19 September 2000)

Re: Why no remission in the Markov model?

Brendan Delaney   (27 September 2000)

Why no remission in the Markov model? 18 September 2000
Graham Byrnes, Biostatistician Victorian Infectious Diseases Reference Lab Send response to journal: Re: Why no remission in the Markov model?	Using a Markov model to describe the evolution of dyspepsia seems like a reasonable strategy. However it is unfortunate that the authors do not discuss their motives for the structure of the model chosen. This is more serious given that the schematic used in the e-BMJ appendix is presented in an obscure format particular to a particular software package. After downloading the manual for the Tree Age package, I was able to decipher it and was surprised to find that the Markov model these authors employ treats dyspepsia as an absorbing state. In other words, the model permits no remission of symptoms. Moreover, all subjects are assumed to start free of symptoms. These would seem to be important points in deciding the validity of the model. In the special case of a 2-state Markov model where one state is absorbing and all subjects start in the other, the Markov model is in fact just an ordinary exponential or geometric failure model. Describing such a simple model as Markov seems unnecessarily obscure. Sincerely, Graham Byrnes
Rapid response-Helicobacter eradication in non-ulcer dyspepsia 19 September 2000
A Poullis Send response to journal: Re: Rapid response-Helicobacter eradication in non-ulcer dyspepsia	Editor-The review by Moayyedi and colleges (1) suggests that H pylori eradication in non-ulcer dyspepsia may be cost effective in terms of dyspepsia treatment and quality of life but they neglect to mention another important area in this problem. Many patients with non-ulcer dyspepsia will be prescribed proton pump inhibitors (PPIs) and may remain on these drugs for many years. Studies have demonstrated that PPI therapy results in more rapid development of gastric atrophy in H. pylori infected individuals and it has been suggested that these histological changes may lead to gastric cancer (2). As H.pylori is also a carcinogen in its own right it would seem advantageous to eradicate H. pylori in patients with non-ulcer dyspepsia for more than just symptom control. A.Poullis SpR Gastroenterology Mayday University Hospital, Thornton Heath CR4 7YE 1. Moayyedi P, Soo S, Deeks J, Forman D, Mason J, Innes M, Delaney B. Systematic and economic evaluation of Helicobacter pylori eradication treatment for non-ulcer dyspepsia. BMJ 2000;321:659-664. 2. Genta RM. Acid suppression and gastric atrophy: sifting fact from fiction. Gut 1998;43:35-8.
Re: Why no remission in the Markov model? 27 September 2000
Brendan Delaney, Senior Lecturer Dept of Primary Care and General Practice, The University of Birmingham, UK Send response to journal: Re: Re: Why no remission in the Markov model?	Dear Editor We thank Byrnes for his interest in our model evaluating the cost- effectiveness of H.pylori eradication in non-ulcer dyspepsia. We agree that a geometric failure model is an equally valid description of our simple Markov model. Different audiences will be familiar with different terms. We feel that a clinical audience would be more familiar with the term Markov model. The BMJ web site contains a description of the model and the rationale behind it. The diagram is in the format of a Data 3.5 decision tree (TreeAge software, Williamstown, MA). This is a commonly used software package in the health economic literature, and a recognised format for the presentation of Markov models. We agree that dyspepsia is a relapsing and remitting condition, whereas the model assumes that patients remain symptomatic once they relapse. The Markovian assumption is that all individuals within a cohort are at equal risk. This simplifying assumption is valid at the population level if the overall proportions of individuals relapsing and remitting follow an exponential pattern. Data from the randomised controlled trials in the systematic review would support this assumption for symptoms of non -ulcer dyspepsia. The model assumed all patients were asymptomatic after therapy, whereas the trial data indicated that only 50% were asymptomatic at one month. We did construct a more complex Markov model to reflect this, which gave almost identical results. We therefore presented the more parsimonious model in the paper. Sincerely Paul Moayyedi Brendan Delaney