Understanding Test Data

[Clays711]

Is there a website that can teach me how to read & understand the the studies' statistics?

An example:

survival 0·50, 99·9% CI 0·34–0·71; p<0·0001; overall survival 0·75, 95·1% CI 0·59–0·95; p=0·017)

I don't understand the info being conveyed in CI, P and HR etc.

Thanks!

[Tall_Allen]

You can take classes in medical statistics, or you can just ask plain English questions.

[OzzieJ]

You can copy it into Chat GPT which gives a good explanation.

[maley2711]

Here is a definition of HR.. " A hazard ratio (HR) is the probability of an event in a treatment group relative to the control group probability over a unit of time." so, if two men in the control group died over a 2 yr span, and one man died in the new treatment group, the HR would be 0.5 for that period of time. The CI is the confidence interval, and the narrower the CI, the more confidence that the stated HR is the true and accurate HR.

"p" also a measure of confidence I believe. The low p in your example is favorable for a valid study conclusion . Google for better explanation...abundant resources on web !!!

we need to be careful when looking at HR numbers. Eg., if control group has 96% survival over 10 yearsm and latest and greatest alternative treatment results in 98% survival, the HR would be 0.5 on a mortality basis...dramatic, but is 98 vs 96% survival a dramatic difference ....especially if the improved 98% survival comes with worse side effects????? each man will have a different perspective probably. I find that Docs are very poor when it comes to such discussions....but maybe I've just beeen unlucky?

[Hidden]

in addition to Allen’s excellent comment, I was told by my research Professor that do to its complexity. Trial data is often misunderstood by many; including the medical community. You can get basic definitions, however the interpretation by a qualified medical professional is essential. Ask them after you start to understand what it is that you are asking.

GD

[Teacherdude72]

TA has it right. Ask your doctor to explain in simple terms until you actually understand. You are paying for their time so don't give up until you really understand. And have another person with you to hear it. Plus take notes.

[lokibear0803]

Understanding the numbers is not as hard as it looks, but you do have to be in the mood.

Understanding what it means for your treatment decisions is best done speaking with your MO.

So, if you’re in the mood — I’ll try to decipher the statistics in a somewhat imprecise (but, for me, simple and useful) way:

survival 0·50, 99·9% CI 0·34–0·71; p<0·0001

CI means “Confidence Interval”. So researchers have 99.9% confidence that the actual range of survival would be 0.50 in the general population, with the same treatment.

But this number is not exact. Since the experimental range of survival was determined in a trial population that was limited in size, they say “it would be 0.50 plus or minus a certain amount”. So they give their 0.50 number a lower limit of 0.34 and an upper limit of 0.71. The 0.34 to 0.71 part is the interval.

The whole reason to do the trial is to find out, was there really a benefit or not? Or is it possible that the patients who saw benefit would have seen that benefit regardless of any treatment? The p-value gives us a quantitative way to answer this question — lower p-values are better indicators that it was not just coincidence, that yes the treatment really worked.

This p-value of 0.0001 is very good — values less than 0.005 are considered to be statistically significant, i.e. it’s unlikely to have been coincidence.

I wrote this up in something of a rush, so it’s probably somewhat sloppy. Hope this helps.

[Clays711]

Thank you for your reply. I went on line and found a medical statistics for dummies type on publication. I found it helpful for now. It's "Medical Statistics Made Easy" by M Harris and G Taylor.

[Clays711 in reply to Clays711]

I was able to download the book as a PDF. I can now reference the stats to the book to better understand the significance of the numbers from a study.