Article Text

Download PDFPDF

Response to letter to the editor by Moayyedi et al
  1. Ka-Shing Cheung,
  2. Wai Keung Leung
  1. Department of Medicine, University of Hong Kong, Hong Kong, Hong Kong
  1. Correspondence to Dr Wai Keung Leung, Departmenyt of Medicine, Queen Mary Hospital, Hong Kong, Hong Kong; waikleung{at}

Statistics from

We thank Moayyedi et al 1 for their letter on our recent study investigating proton pump inhibitors (PPIs) on gastric cancer (GC) risk after Helicobacter pylori (HP) eradication.2 Although the association detected by an observational study may not mean causation, the possibility of causality can be strengthened by fulfilling the Bradford Hill criteria. In our study, these include strength (HR: 2.43), specificity (stomach is the only organ that PPIs may impose a cancer risk), temporality (all GC cases developed after triple therapy), biological gradient (dose and duration response relationship shown), plausibility (worsening of preneoplastic gastric changes and bacterial overgrowth under profound acid suppression), experiment (as illustrated in animal model studies) and analogy (achlorhydria due to autoimmune gastritis causes GC).

Concerns were raised over inadequate adjustment for socioeconomic status and comorbidities. Hong Kong is a highly urbanised city with excellent sanitation and easily accessible healthcare services. Hence, socioeconomic status is likely a negligible concern. Our study included a wide array of variables such as cardiovascular, neurological, respiratory, renal, liver diseases and medications (including aspirin, non-steroidal anti-inflammatory drugs, statins and metformin), which may serve as surrogate markers for unmeasured variables and minimise residual confounding.3

Although post-treatment HP status was inferred indirectly from need of second-line or third-line therapies, retreatment rate of 13% was consistent with that observed from local studies.4 Regarding indication bias, if present, an increased GC risk among histamine-2 receptor antagonist (H2RA) users (negative control exposure) would also be expected. Ideally, comparison should be made with HP-negative subjects who were prescribed PPIs. However, as HP status was not directly available, we instead compared GC risk in HP-eradicated cohort with another cohort of PPI users without prior HP therapy, comprising both HP-negative and HP-untreated patients (table 5 in our original article).2 If significant confounding with PPI indications did exist, one would expect a similar increase in GC risk among this cohort of PPI users without prior HP therapy. However, GC risk among PPI users without prior HP therapy was much lower than HP-eradicated non-PPI users (0.8 vs 2.9 per 10 000 person-years), suggesting prior HP infection is indeed the most important determinant of GC risk. PPIs likely promote GC development only in the context of HP-induced mucosal changes. We speculate that when used in individuals without prior HP infection and underlying gastric mucosal abnormalities, PPIs will likely impose a neutral effect on GC development.

In response to the request for additional analysis, we compared GC risk among three groups of patients (non-PPI/H2RA use, PPI with or without H2RA use, and H2RA use alone). When compared with those with neither PPI nor H2RA use, PPI users (with or without H2RA) had a higher GC risk (HR 2.11; 95% CI 1.23 to 3.64) but not H2RA users (table 1). Excluding non-PPI/non-H2RA users shows the HR of GC with PPI use (with or without H2RA) was 2.4 (95% CI 1.38 to 4.16) as compared with H2RA use alone.

Table 1

Comparison of gastric cancer risk among different groups

We agree with their comments that causality is best achieved by randomised clinical trials. However, it is unethical to randomise patients to receive PPIs to observe for potential side effects. Given the long time lag and low incidence for GC in the West,5 the Cardiovascular Outcomes for People Using Anticoagulation Strategies (COMPASS) study1 is likely underpowered by recruiting only 17 500 patients with a follow-up of 3 years. Second, a secondary analysis of a randomised trial is not equivalent to a randomised trial, which is also subject to various biases. Third, failure to ascertain HP status in their study precludes the establishment of causality—the same caveat of previous studies. Fourth, all participants received aspirin, a potential chemopreventive agent that may negate the harmful effect of PPIs on GC.6


View Abstract


  • Contributors Both K-SC and WKL contributed to this response to the letter to the editor.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests WKL has received honorarium for attending advisory board meetings of Takeda and Abbott Laboratories. There are no competing interests for the other author.

  • Patient consent Obtained.

  • Ethics approval HKU IRB.

  • Provenance and peer review Not commissioned; internally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.