some results

Yesterday we did some real analysis of the data about statins. I learned quite a bit about statistics and especially how to tell if your results are significant – it turns out the bar is pretty high. For instance, one factor we investigated was how long patients survived after their initial diagnosis. In comparing the patients who took statins with those who didn’t, it turned out those on the drugs lived for an average of five months longer than those not on the drugs, after checking to make sure age wasn’t a factor. That seems pretty important when the prognosis is rarely even two years, but then we had to perform a bunch of statistical tests. The “T Test” gives a “confidence interval,” in this case, it was a range of times such that there was a 95% chance a given patient would live that long. The confidence interval for non-statin patients and statin patients overlapped, meaning that even though there seemed to be a big difference in the average survival, the results didn’t indicate any correlation whatsoever.

It was a little frustrating to find out that after all that, there was not a statistically significant result and we’re not at all closer to finding the causes of brain tumors. The idea was that GBMs, the nastiest kind of brain tumor and the one this study used, are highly vascularized, meaning they involve the blood vessels in the brain to a great extent. This is partly why they’re so hard to completely remove. The researcher thought that statins, which lower cholesterol levels and have some effect on blood vessels, might somehow interfere with the cancer’s vascularization or perhaps increase it. Either way, it was interesting to learn the extremely high standards that truly scientific data must reach, and it did make me trust the statistics you hear from different studies a little bit more. There’s a lot that goes into collecting that data and analyzing it.