22 April 2020
(CWI Amsterdam & Leiden University)
BCG (anti-tuberculosis) vaccin is known to have a "non-specific temporary immune-system boosting" effect. The BCG CORONA randomized clinical trial (approved within 3 days - normally this takes at least 6 months!) tests whether BCG vaccinations reduce sickness (due to corona and otherwise) amongst hospital workers. With standard, p-value based methods, it is neither possible to stop a trial early if preliminary results indicate success or futility (early stopping) nor to keep adding new data if results are promising but nonconclusive (optional continuation) - but this is what we want here. At CWI we have recently been very much involved with safe testing and always-valid confidence intervals, new techniques for hypothesis testing that, unlike earlier approaches (group sequential testing, alpha spending) do allow to stop/continue at every data point. Our methods also provide new ways to combine data from different sources (different hospitals or even different clinical trials - meta-analysis) in an error-proof way.
We are helping University Medical Center Utrecht to make their BCG-CORONA trials more flexible in this way.