So, it's not a strong enough or consistent enough difference between treatment and placebo to be certain that it isn't just chance. (this could happen if for example if out of our 1000 data points, 2 people on placebo were hospitalised but only 1 on the treatment was hospitalised. That'd be "a 50% reduction in risk", but not be statistically significant since it could easily have just been random chance when we're talking about such a small difference).
Probably needs an even larger sample population to really nail down whether it's real or not. Seems suggestive that all the various reported tests showed (inconclusive) benefits when you'd expect an actually-ineffective treatment to show results varying randomly in both "effective" and "not-effective" directions, though that could certainly be positive reporting bias, especially since it's being run and reported by a company with a financial stake in the results.