Artist Mark said...
Oh yes ….those wonderful # 5's. I have way too many of those.
Could someone explain this to me. I have read many studies and am not quite sure how to read the hazard ratio and what it means. I understand the abbreviations FFM....CSS....and OS . It's these numbers that get me
confused (p < 0.002; hazard ratio [HR] 3.4 [1.7–7.1])
.The presence of GP5 predicted lower FFM (p < 0.002; hazard ratio [HR] 3.4 [1.7–7.1]); CSS (p < 0.0001; HR 12.9 [5.4-31]); and OS (p < 0.0001; HR 3.6 [2.0-6.5]) in comparison with GS8 (without GP5).
The hazard ratio is just the ratio of two outcomes. It's a comparative thing. So, if treatment A has 5 out of 100 fail, that may be set to 1. Then, if treatment B has a 10 out of 100 fail, the Hazard Ratio is 2 (10/5). So, double the risk of whatever. Conversely, if you set B as the 1.0 baseline, then treatment A has an HR of 0.5 (5/10). It's to clarify the relative risk of a group of things. In that study, they're giving a range of hazard ratios for some of those things.
This is terribly abused in media reporting. They say things like, "Doing Thing X doubles the risk of Bad Thing Y! Never do Thing X!! Film at 11". You've heard that countless times. What they don't say is the baseline risk may be 1 out of a billion. And yes, 2 out of a billion is an HR of 2, and is in fact double the risk. But come on, it's only 2 out of a billion! Statistics are so ripe for misuse, either intentionally or inadvertently.
The p value is sort of how "good" the conclusion is. Very small p values indicate strong statistical support. It's too deep to go much further than that here, and you can find out more with some web searches. It gets fairly obscure quickly.