Sir,

We read with great interests the retrospective case–control study by Wei et al (2015). As the authors Wei et al introduced that epidemiological study is the first one, which found a potential association between hepatitis B virus (HBV) serology and gastric cancer risk. This main finding is indeed surprising to readers. On the basis of the literature from Chinese CNKI journal database, the prevalence of HBV DNA in gastric cancer tissues is only 0–3% by PCR test. Therefore, to evidence the causality between HBV infection and gastric cancer risk, a qualified study with adequate statistical power requires a dramatically larger scale of sample size than that of the study by Wei et al (2015). In particular, direct detection of HBV DNA in gastric cancer cells by in situ hybridisation is the most convincing evidence to confirm that association.

As known, WHO has defined Helicobacter pylori as a class I human carcinogen for gastric cancer development (Fock et al, 2013). Besides, Epstein–Barr virus infection is also found to be associated with around 10% of gastric cancer (Murphy et al, 2009). However, in the study by Wei et al (2015), these two critical confounders were not considered in the logistic regression models. The investigated population in the study by Wei et al (2015) is also collected from an endemic region (Guangzhou Province) of both Helicobacter pylori and Epstein–Barr virus infections in mainland China (Wang and Chen, 2014). Therefore, the results are unable to rule out the confounding effects from these two kinds of infections.

Probably, the association between HBV infection and gastric cancer risk might be biased by chance, imbalance of prevalence of H. pylori and/or Epstein–Barr virus infection in stomach, or potentially indirect linkage between HBV and those two pathogens. Without the adjustment for those two co-infections, the results may have a risk of misleading readers. Thus, we would like to underline these pitfalls behind interpreting the results to readers and practitioners.

Critically, the epidemiological study Wei et al provides some information about the potential association between HBV infection and gastric cancer risk, but the obvious defect in covariate modelling makes the results still far from public health policy and clinical practice. Despite of that, the interesting findings also suggest further investigations with large scale and well-adjusted model to rule out potential biases.