Replies to post #123915 on Biotech Values

[NP1986]

Shkrell's article had multiple cites, which may not be wart free (when are they ever?) but were are at least reasonable. Last week's SA article had no such cites - just blithely ingesting an assumption of 20 weeks for control arm of START trial.

The bullish article actually did list 12 studies towards the end of the article. I think the premise of that article is wrong, but I think it did a slightly better job of at least pointing out some of the risks (even if there was some degree of handwaving).

I agree that it is possible that the control arm of START will have a survival time greater than 20 months. I don't agree with anyone - bull or bear - who claims with certitude that the median survival will be X months. There are little data to go on, and there is bound to be variation between trials. Individual trials tend to have wide confidence intervals, so even if we could find a trial that replicated the same design and inclusion criteria as START, a substantial difference in survival times between the two trials would be plausible purely from a statistical point of view.

Net/net I'd say the Kelly trial is a very reasonable comparison and it had a much longer time to death than assumed by the ONTY bulls.

Relatively speaking, it might be. But just to address a couple of points that you raised:
- Regarding entry criterion of PS 0/1 for START, if you look at the baseline characteristics of the Kelly trial, the vast majority of patients were PS 0,1. The remaining ones are listed as unknown. So even if it wasn't explicitly a criterion for inclusion, virtually all of the patients ended up being PS 0/1.
- regarding pleural effusions, these patients were also excluded from the Kelly trial
- regarding exclusion of N1 patients in the Kelly trial, you should note that these patients are usually candidates for surgery, and that patients are not eligible for START if they undergo surgery prior to randomization

In the post hoc subgroup picked out of the ph iib trial it was always known that the control group did much more poorly than they SHOULD have (in the ph iib the Stage iii Locoregional control group survival was the same as the remaining stage iii and stage iv control group at about 13 months - which is completely non-credible)

That's a fair point, although I don't think it is completely incredible since there is also the possibility that the stage IIIB patients with pleural effusion and stage IV patients did better than expected (in fact, in the discussion of the original JCO publication, the authors mention that they were expecting a 7-month survival in the whole control arm). I think it is plausible that part of the difference between the two arms in the stage IIIB LR subset is attributable to imbalances in performance status and other prognostic characteristics. However, I don't believe that it can account for the whole difference between the two arms - which leaves chance as the alternative. Since the finding was not statistically significant (and this is even without adjusting for multiple comparisions), it is quite possible that it was a fluke.