3 ms·

Apollo astronauts dying of heart disease at 4-5X the rate of counterparts

8 points by shawndumas 10y ago

nkurz 10y ago
At a quick read, the study this article is based on seems like it's either a parody, a pointed critique of abusive statistical methods, or a terrible misunderstanding. On the bright side, the paper seems clearly written and is freely available[1]. On the not-so-bright side, they conclusion of the study is based on the a sample of 7 lunar astronauts, of whom 3 died of cardiovascular disease. 3/7 is 43%. Surely they can't be claiming to have a statistically significant result based on the having 3 rather than 2 lunar astronauts? Yup, that's exactly what they are claiming. 3 of the 35 non-flight astronauts died of heart disease, which is 8.5%.[2] Since there were 5 times as many in this group as the lunar-flight group, that's the upper multiple of 5X. For the low-earth-orbit-only group, 4 out of 35 (coincidental same number) died of heart disease. 4/35 is 11%, which gives us the lower 4x. Maybe as so often happens, the author of the Ars Technica article putting more weight on this minor detail than the authors, while simultaneously removing the carefully phrased caveats? Given that the title of the article is "Apollo Lunar Astronauts Show Higher Cardiovascular Disease Mortality: Possible Deep Space Radiation Effects on the Vascular Endothelium", I think we can eliminate this excuse. I'm hope I'm being too harsh.[3] Perhaps someone with a better grounding in statistics than me can defend their methods and show me what I've missed? [1] http://www.nature.com/articles/srep29901 http://www.nature.com/articles/srep29901 [2] http://www.nature.com/articles/srep29901/tables/2 http://www.nature.com/articles/srep29901/tables/2 [3] How did this get published in Nature? How can this possibly have passed statistical review? Wait, you mean "peer review" doesn't imply "statistical review"? Well then why should we trust their peers to vouch for the statistics any more than we trust the authors?
- bbctol 10y ago
  The main goal of the paper was to explore a possible mechanism for the finding in a mouse model. The differences between Apollo and non-Apollo incidences of heart disease are ~technically~ significant, but look to be there more as motivation for the experiment than a conclusion in their own right--their analysis of astronauts in general vs. the rest of the US population is also new, and more interesting.

8 points by shawndumas 10y ago

nkurz 10y ago

At a quick read, the study this article is based on seems like it's either a parody, a pointed critique of abusive statistical methods, or a terrible misunderstanding. On the bright side, the paper seems clearly written and is freely available[1]. On the not-so-bright side, they conclusion of the study is based on the a sample of 7 lunar astronauts, of whom 3 died of cardiovascular disease. 3/7 is 43%. Surely they can't be claiming to have a statistically significant result based on the having 3 rather than 2 lunar astronauts? Yup, that's exactly what they are claiming. 3 of the 35 non-flight astronauts died of heart disease, which is 8.5%.[2] Since there were 5 times as many in this group as the lunar-flight group, that's the upper multiple of 5X. For the low-earth-orbit-only group, 4 out of 35 (coincidental same number) died of heart disease. 4/35 is 11%, which gives us the lower 4x. Maybe as so often happens, the author of the Ars Technica article putting more weight on this minor detail than the authors, while simultaneously removing the carefully phrased caveats? Given that the title of the article is "Apollo Lunar Astronauts Show Higher Cardiovascular Disease Mortality: Possible Deep Space Radiation Effects on the Vascular Endothelium", I think we can eliminate this excuse. I'm hope I'm being too harsh.[3] Perhaps someone with a better grounding in statistics than me can defend their methods and show me what I've missed? [1] http://www.nature.com/articles/srep29901 http://www.nature.com/articles/srep29901 [2] http://www.nature.com/articles/srep29901/tables/2 http://www.nature.com/articles/srep29901/tables/2 [3] How did this get published in Nature? How can this possibly have passed statistical review? Wait, you mean "peer review" doesn't imply "statistical review"? Well then why should we trust their peers to vouch for the statistics any more than we trust the authors?

bbctol 10y ago
The main goal of the paper was to explore a possible mechanism for the finding in a mouse model. The differences between Apollo and non-Apollo incidences of heart disease are ~technically~ significant, but look to be there more as motivation for the experiment than a conclusion in their own right--their analysis of astronauts in general vs. the rest of the US population is also new, and more interesting.