Meaning of "Cox model censoring at relapse" in simple words

Question

I'm translating into Russian a news report on the results of a clinical trial:

Quote:

Bruce Cree, MD, PhD, MAS, lead author, and neurologist, UCSF Multiple Sclerosis Center, and colleagues found that siponimod reduced 3-month CDP by 14% to 20% and 6-month CDP by 29% to 33% in non-relapsing patients across the Month 12, Month 18, and Month 24 timepoints.

Cox model censoring at relapse confirmed these beneficial effects, reaching nominal statistical significance for 6-month CDP (HR, 0.77; 95% CI, 0.62–0.96).

I did not know the meaning of "Cox model" and "censoring" so I looked up a variety of simplified explanations. The Cox model allows us to investigate factors that may have contributed to the results of a trial. Censoring is the use of data for patients who have not yet died or had a relapse or achieved any other endpoint relevant to the trial. They might have been lost to follow-up.

Still I have no ideas about the meaning of the phrase cox model censoring at relapse.

I don't understand even the use of the term "censoring" in this particular case. Someone performed "censoring" using the Cox model? What may that mean? And why "at relapse" if the patients were "non-relapsing"?

This is the trial being reported upon.

The paper describing the obtained results is probably here: Long-term Efficacy and Safety of Siponimod in Patients with SPMS: EXPAND Extension Analysis Up to 5 Years (4128)

Answer 1

There are two uses of 'non-relapsing'. One looks like it refers to a period of time before the start of the trial.

The risk reductions in non-relapsing patients in the 1 and 2 years before the study were 18% [...] and 13% [...], respectively

That is, patients who didn't have a relapse in the 1 year before the study had a risk reduction of 18%, and those who didn't have a relapse in the 2 years before the study had a 13% risk reduction. That's a potentially sensible subgroup comparison, trying to look at people with more frequent or less frequent relapses.

'Cox model censoring at relapse' is something different. It looks like it means an analysis of "6-month CDP" found it was 29% to 33% lower in the treatment group if you stop counting patients as soon as they have a relapse. At any given time after the start of the study you can compare patients in the treatment and control group who haven't yet had a relapse, and see at what rate they are getting CDP. The claim is that this is lower in the treatment group.

The rationale for this might be that they don't claim to prevent relapses in relapsing/remitting MS, just to slow disease progression during remissions. I suspect this is likely to be a bit dodgy as an outcome measure, but I'm not an MS expert.