We thank Drs. Kuriki and Nishida for their interest in and insightful comments on our article, in which we investigated the gastrointestinal symptoms in Japanese patients with coronavirus disease 2019 (COVID-19) [1, 2]. Diarrhea was found to be a good prognostic factor in our nationwide registry-based study among 74 facilities (n = 3498).

The authors’ letter mentioned the possibility that diarrhea was underestimated because of the retrospective nature of the study. Although diarrhea was less common in our study than in that by Kuriki et al. was this really caused by the difference in the study design? If a systematic error exists, other gastrointestinal symptoms should also be less common in our study. However, nausea/vomiting and abdominal pain were observed in 8.9 and 3.5% of patients in our study, whereas they were observed in 5.9 and 1.3% of patients in the study by Kuriki et al. This inconsistency prevents us from concluding that the difference in the prevalence of diarrhea between the two studies was solely due to underestimation. Furthermore, a systematic review of 55,245 patients with COVID-19 from 55 studies showed that the incidence of diarrhea ranged from 4.7 to 45.8% with a cumulative incidence of 16.0% [3], similar to our data.

Regarding outcomes, we analyzed intensive care unit (ICU) admission, the need for mechanical ventilation, and death. The criteria for ICU admission varied among the facilities; in this regard, a strength of our study is its inclusion of 74 facilities. Kuriki et al. classified the severity of COVID-19 according to the definition established by the Ministry of Health, Labour and Welfare in Japan. In this classification, severe disease is defined as the need for ICU admission or mechanical ventilation [2]. We re-analyzed our data using this classification. Among 3498 patients, the incidence of severe disease was 20.4%. Severe disease was less frequent in patients with diarrhea (16.2%, 94/581) than in those without (21.3%, 620/2917) (p = 0.006). In the multivariable analysis using the same models, the multivariable-adjusted odds ratio (95% confidence interval) of diarrhea for severe disease was 0.70 (0.54–0.92) and 0.69 (0.53–0.92) in Models 1 and 2, respectively.

We revealed a robust association between diarrhea and better outcomes in Japanese patients with COVID-19 in our multivariable analyses. We understand that the retrospective design was an important limitation. We could not determine whether the timing or severity of diarrhea affected the results. Unidentified factors may have caused effect modification, preventing extrapolation. Further investigation is warranted to explore the reason for the discrepancies among the studies.