I took care to read the fine submitted article (based on a press release) and most of the preceding twenty-some comments before replying here. Comments to the effect that we often see stories like this based on press releases that don't translate into clinically useful treatments are correct. Anyone who has been on Hacker News as long as I have has seen dozens upon dozens of stories about breakthrough medical treatments that don't turn out to be breakthroughs.
As usual, here in this thread I recommend Peter Norvig's article "Warning Signs in Experimental Design and Interpretation" about how to read research reports in general.
Medical research is slow and painstaking. There has been genuine progress in medical research in my lifetime. (My mother was a nurse in our state's largest research hospital, and medical research was a subject of dinner table conversation in my home.) But false starts have been numerous. There are a great many medical treatments that have been proposed in my lifetime that have not been proven to be both safe and effective for treating what they purport to treat. Peter Norvig's article provides a good checklist of all the ways that optimistic initial reports can turn out to be wrong, of which "too few subjects" especially applies here.
I reply over and over and over again on threads like this here on Hacker News because I did grow up with a strong interest in medical research--and some anecdotes about medical research that went awry--and because I know that legitimate medical researchers sometimes find their work hyped by the university press office or funding agencies far beyond their own more cautious statements. But here on Hacker News we may as well learn how to be discerning readers. Let's not just believe what we want to believe, which is the cognitive bias all human beings have, but let's test factual statements for strong evidence and make sure that the experimental procedures reported in some new study mentioned in a press release really support a grand conclusion. It's too early here to say "Drug made from toxic weed kills cancer" in a way that is safe and effective for a broad range of human patients for even the one kind of cancer that is mentioned in the article submitted here. I'm glad someone is working on this, but it's way too early to declare that we've found a cancer cure through this work.
Upvoted! What I really miss (in general, not just here) is a certain level-headedness when it comes to things like this. I know it is hard to avoid the hype since it is everywhere and omnipresent. but as technical people I still want to believe that we can avoid it and think things through before we open the champagne. We can do it in our own domains, why not in others too?
As usual, here in this thread I recommend Peter Norvig's article "Warning Signs in Experimental Design and Interpretation" about how to read research reports in general.
http://norvig.com/experiment-design.html
Medical research is slow and painstaking. There has been genuine progress in medical research in my lifetime. (My mother was a nurse in our state's largest research hospital, and medical research was a subject of dinner table conversation in my home.) But false starts have been numerous. There are a great many medical treatments that have been proposed in my lifetime that have not been proven to be both safe and effective for treating what they purport to treat. Peter Norvig's article provides a good checklist of all the ways that optimistic initial reports can turn out to be wrong, of which "too few subjects" especially applies here.
I reply over and over and over again on threads like this here on Hacker News because I did grow up with a strong interest in medical research--and some anecdotes about medical research that went awry--and because I know that legitimate medical researchers sometimes find their work hyped by the university press office or funding agencies far beyond their own more cautious statements. But here on Hacker News we may as well learn how to be discerning readers. Let's not just believe what we want to believe, which is the cognitive bias all human beings have, but let's test factual statements for strong evidence and make sure that the experimental procedures reported in some new study mentioned in a press release really support a grand conclusion. It's too early here to say "Drug made from toxic weed kills cancer" in a way that is safe and effective for a broad range of human patients for even the one kind of cancer that is mentioned in the article submitted here. I'm glad someone is working on this, but it's way too early to declare that we've found a cancer cure through this work.