Thread by @DrAndyMatthews

1/12 Tweetorial on initial therapy in older adults with #AML at #ASH21. We presented on CPX-351 vs ven/aza. There was no difference in OS. HR 0.88 with 95%CI crossing 1. OS at 2yrs 28% both arms and median OS 13 vs 11 mos tinyurl.com/2p95bn78. #leusm

2/12 We looked at 656 patients at Penn and across the nation with Flatiron. Ven/aza patients were older, more likely to be treated in community rather than academic centers and more had de novo AML.

3/12 There was no difference in OS among univariate subgroups and no difference after: unadjusted, multivariate Cox (MVC), restriction to CPX-351 eligible patients only, multiple imputation (MI) and MVC, and MI & inverse probability of treatment weighting.

4/12 For those distrusting statistics as chicanery. Strap in. Confounding by indication and missing data are huge factors in comparing treatments like this!! Consider the avg age in VIALE-A and Lancet JCO & effect of age alone on induction survival (tinyurl.com/rbaejpna)

5/12 IPTW likely isn’t that different from MVC but missing data matters here. I won’t get deep on MI but watch how much the n drops and the 95% CIs spread when similar studies go from unadjusted Kaplan Meir to multivariate cox.

6/12 So if overall survival is similar what can we learn here? Transplant matters. A lot. Median OS was 4x higher in transplanted vs non for both txs (of course that’s stratifying on a collider but HR with HSCT as a time varying covariate is still 0.3!!)

7/12 What about safety? This is sticky in retrospective setting. But early morality similar. CPX-351 pts had 2x documented infections, abx use and length of stay (? beating up CPX-351 not usually outpt but some have tried! tinyurl.com/y85hfw88 tinyurl.com/2c22f5wy).

8/12 Justin Grenet, @DrPinkalDesai et al also presented on this question and found OS favored CPX-351 vs any HMA + ven in multivariate adjustment. I can’t wait to see this paper printed out next to ours to really see the similarities and differences.
tinyurl.com/2h7at5dn

9/12 A few caveats for all of us: (1) How do you compare response rates if CPX-351 patients are re-induced? (2) mind difference in transplant rates (mediator for OS) (3) are tx groups similar and multivariate Cox can eat power!

10/12 I think both of our studies show there is, at most, a marginal difference in CPX-351 vs ven/aza. And, again, transplant matters!! Need prospective like this: tinyurl.com/2p9yecnz and we need novel approaches!

11/12 Where 7+3 fits in or how triplets (ASH 367, 370, 696, 701) change calculus is unclear. Many ppl have gripes on CPX-351 (see strong takes by the amazing @Berninini

https://twitter.com/berninini/status/1470054513180651525?s=21

and @Plenary_Session episode 3.27). Cooperative grps are planning trials here too

12/12 Others have been looking if less is more in #AML, check it out:
tinyurl.com/2h7at5dn
tinyurl.com/ycksnz89
tinyurl.com/2p8mve2k

For our study grab it via ^ASH link up top & pdf tinyurl.com/2p9c7da7

Appreciate the academy. Thx esp to Keith, Sasha & @jrladd

Some extras to be added here at some point from @OncLive & @ASCOPost