I propose a principle. Kaplan Meier reliability heuristic. The core rule: If majority of censoring occurs prior to 67% events, the data is immature and needs more followup before jumping to conclusions. Below this event fraction, sample size is too small to anchor the curves well
A rough bullsh*t detector when p values look amazing. For HRs look for CI. If upper 95th/lower 95th bound CI ratio is <2 = strong finding. 2-3=ok. 3-4=‘wobbly’ HR. >4 = useless
@rajshekharucms@NEJM@TheLancet@VPrasadMDMPH True, semi parametric estimates would provide better answers or parametric even better. Kaplan Meier is non parametric so it is in the nature of KM to plateau.Although, Cartitude-4 has right shifted censoring vs Majestec-9 with the left, a structural difference so hard to compare
@rajshekharucms@NEJM@TheLancet No im not worried about the initial difference. Even if the censoring is administrative, if majority of censoring precedes events, the remaining at risk pool is enriched making the tail a mathematical artifact regardless
@bdermanmd Low numbers, the CI is far too wide. May be a signal there but the HR is unstable. Same story with prothena. Vital to affirm al. I wonder if history will repeat
@TheDesKelly If that is not a foul, I don't know what is? You were expecting a chokeslam on Raya before it is a foul? It was a perfectly fair decision.
@Tripple____M@El_Inshai If you think about it- We fell apart after our main wing backs went out on injuries along with merino. Merino, calafiori and timber are critical (central, left and right attack).
@ZanwarSaurabh@HiraSMian@RahulBanerjeeMD@AJPortuguese@BloodCancerJnl Agree with saurabh. Median follow up 7.5 mos. One of the limitations of KM (non parametric)modeling and focus on medians. (Not that the paper was incorrevt in doing it). Semi parametric or parametric methods give better answers for these questions
@rajshekharucms@bdermanmd@CoMMitTrials Voted for cart but a Lot of options beyond cart. Pt is still PI sensitive, (carfilzomib combos)linker-mm2 data looks quite strong (PI+ TCE when apptoved). I believe TCE with imids or celmods (even if imid resistant) would be the way in the near future
@grok@utdshows The two domestic cups make sense. Historically arsenal wear down around this time of the season and lose the pace for an EPL win. UCL might be doable but gut feeling they wont win it. Today’s win against chelsea will be decisive