You may have heard of a recent study showing Splenda causes leukemia. What it actually shows is mice fed the equivalent of ten cans a day showed "a significant dose-related increased incidence (p < 0.01) of neoplasias in males, in particular at the dose levels of 2,000 ppm (p < 0.01) and 16,000 ppm (p < 0.01)." Nothing in the abstract mentions the p-value of the middle dose 8,000 ppm, which seems odd if they are claiming dose dependence.
If you don't know what a p value is, let me explain: no one really knows what a p-value is. That's not entirely true. Its basic definition is that it is a statistical measure of how likely a result is and one of the most often cited values in these types of studies. In this case, researchers gave Splenda to about 1,000 mice and measured cancer rates and compared to the control group. They start with a null hypothesis, "there is no link between Splenda and cancer." They compare the cancer rate with the Splenda-fed animals to what you would expect in a normal mouse population. A p-value of 0.01 means that there is a 1% chance that you would randomly see this cancer rate in the normal population that wasn't scarfing down Splenda.
Anyway, the test is problematic because the values set are somewhat arbitrary and the null hypothesis has a high percentage of being wrongly rejected. Basically, everyone is confused about p-values, so you really have to dive into a study to see what the data means. (Note: I did not dive into this study).
I didn't read this article (only the summary) because you have to pay for it, and it's the weekend so I didn't want to harass my friends at universities to send me a copy (Note to self: is that technically illegal? should I be blogging about that? Check later.), but once again I feel let down by science news. They always have a scandalous headline with very few real details or helpful interpretation.
No comments:
Post a Comment