Comment by tominous | Hacker News Reader

Comment by tominous | Hacker News Reader

Parent Full thread

I'm biased. My wife was diagnosed with metastatic cancer 6 months ago. With current standard treatments there is a 3-year survival rate of practically zero, so to me the most important thing is getting new effective treatments developed and brought into clinical practice as soon as possible.

I haven't seen the presentation about the study (a related news briefing is here [1]) so I don't know whether you're correct to say that the researchers haven't shared their methods and data. They seem quite happy to share technical details at the news briefing.

However, regardless of that, I still think it is a good thing if researchers talk about their findings at an early stage, especially if the findings are as dramatic and positive as these. If a treatment is available and you have patients out there dying with no other options, then it is just wrong to hide that away.

In my opinion we treat these things far too cautiously. It is not optimal for the individual and it is not optimal for society. By sharing /and/ opening up options like this to terminally ill patients we can (a) save individual lives and (b) speed up development and refinement of new treatments to benefit everyone.

The problem is that the incentives we have in place treat a cancer death as acceptable, while a treatment-related death is unacceptable for researchers, doctors and drug companies (unless they are shielded by layers of clinical-trial bureaucracy or years of standard practice). The result is that the course of treatment doesn't properly weigh the different sources of risk. Whether the goal is quality-adjusted life years for an individual patient or faster development of effective and safe treatments, the current system is way too conservative.

[1] http://www.eurekalert.org/aaasnewsroom/2016/webcast/?b=38

For what it's worth, I used to say I would take some experimental treatment when the time came. Not after doing research myself and seeing first hand what "experimental" really means when it comes to medicine. It means: no one has any way of really knowing what is going on. For example,

1) There will be some studies where a measurement was higher in the treatment group compared to controls, but it is not 100% clear what combination of things is actually being measured.

2) There will be other studies you hear about that got different results, but didn't get published because publishing "negative results" is discouraged and bad for careers.

3) The criteria for publishing a positive result is a p-value below an arbitrary threshold, and most people researching medical treatments really do not understand the meaning of a p-value to begin with. There are all sorts of myths like it is the probability the treatment didn't work, etc. So (at least the implementation of) the basic criteria used to judge success is fatally flawed.

It is quite scary and I don't see anything I can do about it but make people aware there is a problem here.

From a sociology point of view, yes you should distrust any signs of p-hacking or selective reporting, especially when there is a low effect size.

From a technology point of view, I think we are already well past "no one.. really knowing what is going on." Take the current study. Past researchers had identified the CD19 antigen found (often exclusively) on B cells, and others had found that autologous T-cells could persist in a patient for many years, potentially providing a durable response. The current researchers built on this knowledge to design and engineer T-cells that could wipe out a patient's B cells.

So it's a very planned, targeted approach, more akin to engineering than the old approach of testing thousands of random molecules as potential treatments. When the targeted approach works, and especially when it has a large and consistent effect on patients as in this case, then you can be very confident at an early stage.

This needs to be recognized and medical practice needs to change as a result. We're entering an era of incredibly detailed knowledge and precision tools for research and development. The pace of development will continue to increase and treatments will become more patient-specific as individual targets are identified.

You could consider the large, slow, blunt clinical trials as being like optimization for a particular treatment. When new drugs are rare or not well understood it makes sense to spend time to optimize their use. But when new drugs come frequently and are well targeted this premature optimization costs lives.

>"Past researchers had identified the CD19 antigen found (often exclusively) on B cells, and others had found that autologous T-cells could persist in a patient for many years, potentially providing a durable response."

I didn't watch the video, so I'm not sure where this info is coming from, but providing such references is one reason why they should write a paper up before talking to the media.

Also, I am fairly certain if we do follow up on these claims (I know nothing about them specifically) we will find the actual evidence is much more ambiguous, there has been no direct replication attempted, the references do not actually contain the evidence claimed, etc.

See the problem? I can claim anything I want about this, be it optimistic or pessimistic, and no one can check. There is no paper (or equivalent source of information), so no way to assess the evidence.

One way to approach an announcement like this is to wait until everything is wrapped up nicely in a neat paper.

Another approach is to investigate further using the tools available: look up the authors' past work on Google Scholar, check clinicaltrials.gov, look for third party opinions, even ask the authors directly. I'd argue that the scientists and journalists at the AAAS conference where this was presented are capable of this second approach, and so are many participants on HN.

I think if you try it out, after a few you'll find this is naive. Further, you are doing the job of the person making the claim!

It is too time consuming to research a claim indirectly like that. That is what you do after reading the paper which contains (usually most of) the required information and (usually a biased sample of) citations to the other research that were found relevant.

Reading the paper is only a first filtering step to evaluating a claim, a conference presentation simply doesn't merit that effort because it contains far too little information. A conference presentation is more like "hey, check out what I'm working on".

Now, that doesn't mean people don't try to do this, it is one of the primary uses for misinterpretation of p-values.

> I'm biased. My wife was diagnosed with metastatic cancer 6 months ago My sincerest condolences. Even attempting to imagine what that must be like is beyond painful; my mind refuses to contemplate it for more than a few moments.

I hope that you're able to find something that works out.

If you know the experimental drug\treatment you want to try check overseas. Don't wait for local (I'm assuming) docs to give you the info. Contact centers with a known rep. Good luck!

Thanks, I certainly have my eyes peeled for news about metastatic ACUP (or any related diagnoses) but as far as I'm aware there's no silver bullet anywhere yet.

Right now the strategy is to kick the can down the road until there is a solution: traditional chemo, then the blunt immunotherapy of ipilimumab + nivolumab/pembrolizumab.

Also (with oncologist's knowledge) adding in very speculative but reasonably safe extras like a COX-2 inhibitor, anti-histamine, and melatonin. (Certainly not the well-targeted approach I talk about in my other comment, unfortunately!)

Reply on news.ycombinator.com