Curves and spin-offs

Posted on January 27, 2010 by Cath@VWXYNot?

A number of “my” PIs are trying to develop novel predictive and prognostic biomarkers specific to different tumour subtypes. We have excellent resources for this kind of work, including extensive outcomes-linked tissue microarrays for immunohistochemical assessment of marker status and subsequent statistical analysis. However, the representation of some of the rarer tumour subtypes in these arrays is something we still need to work on.
At a meeting today, one of the stats guys handed the PI a group of Kaplan-Meier survival curves based on the expression of a candidate biomarker in different tumour subtypes. Some of the curves looked very promising indeed.
One of them looked like this:
kaplan
The PI said, “you know, I’ve seen spin-off biotech companies formed in response to curves like this. Personally, I think I’ll hold off”.
I want to see a Kaplan-Meier curve showing bankruptcy-free survival of the companies using such data, compared to the companies using actual good data.

About Cath@VWXYNot?

"one of the sillier science bloggers [...] I thought I should give a warning to the more staid members of the community." - Bob O'Hara, December 2010
This entry was posted in Uncategorized. Bookmark the permalink.

19 Responses to Curves and spin-offs

  1. Bob O'Hara says:
    January 27, 2010 at 9:12 pm

    Brilliant! My wife wishes I could make her laugh like that. Let’s see if anyone else can work out the sample size for the blue line.
    I guess the sample in the red line was a statistician who stormed out as soon as they found out the details of the experimental design.

  2. Cath Ennis says:
    January 27, 2010 at 9:21 pm

    I’d guess a single digit, even, square number.
    The statistician didn’t storm out, he just chose to focus on the tumour subtypes with better n values!
    Glad you and the wife enjoyed this… it made my day!

  3. Stephen Curry says:
    January 27, 2010 at 9:24 pm

    4 – or there could be some weird synchronicity effect going on in a very large population…

  4. Cath Ennis says:
    January 27, 2010 at 9:37 pm

    Well, the units aren’t given on the x axis. It’s conceivable that you’d see these effects in a large population if the units are actually in decades…

  5. Bob O'Hara says:
    January 27, 2010 at 10:16 pm

    But not from 1 to 0.75 to 0.5. I’m trained to spot these things. It’s part of our Dark Arts, against which even Prof. Snape knows no defence.

  6. Cath Ennis says:
    January 27, 2010 at 10:25 pm

    Can one be trained in these Dark Arts, or is it an ability that only A Chosen Few are born with?
    I wish the x axis units were in decades. There’d be less work for me if they were, but for the best possible reason.

  7. Heather Etchevers says:
    January 28, 2010 at 9:10 am

    I like the idea that the line might just represent one individual, half-dead from exhaustion.

  8. Cath Ennis says:
    January 28, 2010 at 3:35 pm

    LOL!
    Or realising “there’s only four people in the other cohort? SRSLY?! Why bother? I withdraw consent for you to analyse me”

  9. Bob O'Hara says:
    January 28, 2010 at 7:32 pm

    Ah, one can be trained in our Dark Arts, but it also requires great practice and the study of arcane texts.

  10. Cath Ennis says:
    January 28, 2010 at 10:36 pm

    Oh, I won’t bother then.

  11. Kausik Datta says:
    February 2, 2010 at 7:18 pm

    May I, with trepidation, put in a cent’s worth? IMHO, these KM curves are difficult to compare in analysis – for two reasons: (a) One cannot compare the median outcome (survival) with these curves, since only one reached the 50% outcome and not the other, and (b) the lagging curve has a few early events and then stabilized. A log rank test to compare these two will need special mathematical manipulations (such as the Gehan-Breslow-Wilcoxon test).
    Sorry for speaking out of turn. I now await chastisement.

  12. Cath Ennis says:
    February 2, 2010 at 7:27 pm

    “May I, with trepidation, put in a cent’s worth?”
    Of course you may! Welcome to the blog.
    I’m sure you’re right about the types of analysis needed if we wanted to proceed. (Bob’s a statistician, but I’m not: I just work with statisticians (not Bob, other ones), and post their unwanted results on the internet for Bob’s amusement). But luckily, I can assure you that we won’t be pursuing this particular analysis any further unless we can increase our sample size by at least a factor of ten. We have better numbers for other subtypes of this type of disease, and our stats team are working on those subtypes while ignoring this one.
    “Sorry for speaking out of turn. I now await chastisement.”
    You’re not out of turn at all. I hope you’ll hang around and check out some of the other blogs on here!

  13. Kausik Datta says:
    February 2, 2010 at 8:21 pm

    Thank you, Cath. I shall add your blog to my to-read list (though I don’t know when I would read it… perhaps only occasionally; I apologize in advance for that). I wish reading blogs and responding could be a full time profession – that would be so much fun! Something like being a science journalist, I guess. Right now, the few blogs I read and comment in take up all my spare time! 🙁
    Re: stats. The K-M caught my attention. I work in Infectious Diseases, and use animal models all the time. Therefore, K-M curves and the log rank test have perforce become my staples. The kind of curves you put up is one of the nightmarish ones, from an analysis perspective, because it provides no meaningful conclusion. Perhaps statisticians (such as Bob; I am not) are able to better deal with these.
    My fear of chastisement comes (rather irrationally, perhaps) from the recent brouhaha involving Henry Gee and a few of the sciencebloggers. It unsettled me. It was as if my two separate worlds, NN and SB, intersected and crashed into each other. I hope this uneasiness will pass in time.
    All the best for your work and blog.

  14. Cath Ennis says:
    February 2, 2010 at 8:41 pm

    “The kind of curves you put up is one of the nightmarish ones, from an analysis perspective, because it provides no meaningful conclusion.”
    Nightmarish from an analysis perspective, yes, but good fodder for bloggers!
    “My fear of chastisement comes (rather irrationally, perhaps) from the recent brouhaha involving Henry Gee and a few of the sciencebloggers. It unsettled me.”
    I wondered if that might be the case 😉
    The types of comments you’ve left on this post would be welcome anywhere on NN, any time, and the same almost certainly goes for ScienceBlogs too (I only read a few of their blogs so I can’t speak for the whole network).
    “I hope this uneasiness will pass in time.”
    Oh, it will.
    Thank you for your best wishes, and the same to you!

  15. Henry Gee says:
    February 2, 2010 at 9:28 pm

    I’m a palaeontologist, so any number greater than three means my knees go all wobbly and I have to sit down. Back in the day, though, I could wrangle Michaelis-Menten whatsits like a pro.

  16. Cath Ennis says:
    February 3, 2010 at 1:23 am

    “ny number greater than three means my knees go all wobbly and I have to sit down.”
    Seven!
    Twenty nine!
    A thousand and five!

  17. Henry Gee says:
    February 3, 2010 at 7:26 am

    OOOh! Stop! STOP!!!

  18. Bob O'Hara says:
    February 3, 2010 at 3:55 pm

    bq. The kind of curves you put up is one of the nightmarish ones, from an analysis perspective, because it provides no meaningful conclusion.
    In my professional capacity, I’d say the analysis of these was easy. You’re not going to get anything from them, so don’t bother trying. It’s the sort of statistical problem I like.

  19. Kausik Datta says:
    February 5, 2010 at 3:06 am

    Bob, now that‘s the kind of analysis I like!! 😀

Comments are closed.