Sitting in Israel writing this blog post, drinking a diet coke, and wondering what they were thinking.
You have read the headline, or heard it on TV or the radio – diet sodas cause dementia and strokes.
That was the headline, but any careful analysis of the study suggests that they did not PROVE anything.
Aaron Carroll – the Incidental Economist – has a great analysis – They did not prove that diet soda causes Alzheimer’s Disease. THEY DID NOT!
For a profane, but funny take – DIET COKE WON’T CAUSE STROKE, BUT READING SENSATIONALISTIC HEADLINES MIGHT
The problems in short:
- Too many comparisons without adjustments. When you do multiple comparisons in an observational study, you must adjust your p value for the multiple comparisons. I learned Bonferroni’s correction (divide p by the number of comparisons. No such correction occurred in this study.
- All retrospective observational analyses should raise question marks.
- The headline writers know neither medicine nor statistics. The headlines scare many people reading only the headlines.
This story is really a non-story.
Vinay Prasad said it best in a tweet:
Diet soda study,
Adjust for the 90+ comparisons you provide, and a significant p is 0.0005
Nothing here is significant.
This is BAD paper
The diet coke tasted great.