Gastrointestinal and liver involvement in patients with COVID-19 – Authors' reply

We thank Yongxing Xu and colleagues and Francisco J Barrera and colleagues for their comments on our meta-analysis about gastrointestinal and liver involvement in patients with COVID-19.¹

We agree with Xu and colleagues that when assessing the prevalence of liver injury, a subgroup analysis based on the severity of COVID-19 is more valuable than is a pooled analysis. As shown in figure 6 in our study,¹ there was a higher risk of increased alanine aminotransferase (odds ratio [OR] 1·89; 95% CI 1·30–2·76; p=0·0009) in patients with severe COVID-19 than in patients with non-severe disease. Similarly, the OR for increased aspartate aminotransferase in patients with severe COVID-19 was 3·08 (95% CI 2·14–4·42; p<0·00001) compared with the group with non-severe COVID-19.¹

We concur that abnormal liver function tests do not always equate to liver injury, and that the incidence of liver injury might be overestimated in patients with COVID-19 in the currently published literature as our understanding of COVID-19 is evolving rapidly and SARS-CoV-2 has been shown to induce myocardial injury.² Of the 12 studies included in our meta-analysis of liver injury, only two gave a clear definition for liver injury. Thus, we adopted a loose definition based on abnormal liver function tests. For this reason, we used the terms “abnormal liver function” and “abnormal liver chemistry” in our study.¹ Additionally, we analysed other liver injury indices such as bilirubin and albumin, which are rarely influenced by myocardial injury and muscle injury.

Barrera and colleagues raise the concern that some patients might have been included in multiple studies included in our analysis. We agree that a few patients might have been included in more than one study, which has been documented during the pandemic.³ Avoidance of this limitation would have required us to obtain individual patient data to ensure that patients were not included in more than one publication. This process was difficult to do during the unprecedented outbreak of COVID-19. In our study, our literature search was restricted to articles published in English only, partly to try to avoid inclusion of data from patients who had been reported both in English and in Chinese, or other languages. In another meta-analysis, Sultan and colleagues⁴ used a hierarchical model of data extraction to minimise double counting of patients across similar institutions with coinciding dates of study inclusion, and their findings were similar to ours.

We agree that there is heterogeneity in the definitions used for severe COVID-19. In our meta-analysis, disease severity was defined per study; in some cases, patients with pulse oxygen saturation of less than 90%, in need of intensive care unit care, or with acute respiratory distress syndrome were also classified as having severe disease.¹ In subgroup analyses by COVID-19 severity, there was no substantial heterogeneity for gastrointestinal symptoms (I ²=0–24% except for loss of appetite [I ²=64%, indicating modest heterogeneity]) or liver chemistry (I ²=0–10%). Subgroup analyses including only those studies with similar severity definitions were not done because only a few studies used the same definitions.