1/ Thread: The evidence I’m looking for.
One thing that I realized I haven’t taken the time to lay out is the evidence I’ve been searching for — literally since Nov’ 15 when I started this journey:
*Genetically low LDL associating with lower all cause mortality*
One thing that I realized I haven’t taken the time to lay out is the evidence I’ve been searching for — literally since Nov’ 15 when I started this journey:
*Genetically low LDL associating with lower all cause mortality*
2/ That was the most intuitive way to test the lipid hypothesis fully as it would suggest reduction of LDL associating with lower CVD had a net benefit to longevity when compared to the general population.
Genetically less LDL => longer lifespan than overall population
Genetically less LDL => longer lifespan than overall population
3/ There are many factors I’m particularly interested in that make a study (*any* study) more compelling for me:
(A) Minimal modification. Generally, the less one-off modeling, extrapolating, and unusual exclusions being applied, the more I like it.
(A) Minimal modification. Generally, the less one-off modeling, extrapolating, and unusual exclusions being applied, the more I like it.
4/
(B) A broad and categorically inclusive criteria for both the population being looked at and exposure duration. This is somewhat intuitive, as we should avoid novel inclusions and exclusions where it is simply making the case through selection bias.
(B) A broad and categorically inclusive criteria for both the population being looked at and exposure duration. This is somewhat intuitive, as we should avoid novel inclusions and exclusions where it is simply making the case through selection bias.
5/
(C) A shared, transparent dataset. I really like large, publicly available datasets like NHANES, of course, because there’s both higher accountability and replication compared to a proprietary dataset.
(C) A shared, transparent dataset. I really like large, publicly available datasets like NHANES, of course, because there’s both higher accountability and replication compared to a proprietary dataset.
6/
(D) If making a claim of mortality benefit — publication of all cause mortality outcome as well. No need to expand on this... it’s already covered many times over in my plethora of tweets for the topic.
(D) If making a claim of mortality benefit — publication of all cause mortality outcome as well. No need to expand on this... it’s already covered many times over in my plethora of tweets for the topic.
7/ There are actually more factors than (A), (B), (C), and (D), but that convers the major categories.
Which brings me around to the study @AviBittMD posted yesterday and why I want to pursue it more than any I’ve seen to date on the topic of risk...
Which brings me around to the study @AviBittMD posted yesterday and why I want to pursue it more than any I’ve seen to date on the topic of risk...
8/ The abstract listed sounds attractive to me if I’m interpreting it correctly so far.
This appears to fulfill (C) given it is a shared dataset and (D) ACM is measured (yay ACM!).
Yes, there may be one-off modeling/adjusting with regard to (A) & (B), but maybe not...
This appears to fulfill (C) given it is a shared dataset and (D) ACM is measured (yay ACM!).
Yes, there may be one-off modeling/adjusting with regard to (A) & (B), but maybe not...
9/ If I had a hold of the Copenhagen data, I’d want to run an analysis similar to the one described, using as minimal adjusting as possible and reaching age parity for comparison groupings with maximum exposure time available..In other words, as little exclusion as is reasonable
10/ If those with high PCSK9 allele exposures (categorically) had stepwise lower ACM than the general population, I’d consider that very meaningful!
Bookmark this tweet. We may be returning here some day. :)
Bookmark this tweet. We may be returning here some day. :)
11/ As mentioned in the earlier thread, I’m engaging a biostatistian soon and this might be something they could help us replicate.
Make no mistake, if we can show stepwise lower ACM than average pop with the methodology listed above, I’ll definitely be sharing it far and wide.
Make no mistake, if we can show stepwise lower ACM than average pop with the methodology listed above, I’ll definitely be sharing it far and wide.
12/ I’ll concede there is some uncertainty at the moment regarding the abstract @AviBittMD linked where we don’t have the paper, but he’s contacting the author on our behalf to better understand the analysis. (And perhaps how we can reproduce it)
13/ Again, I’m calling special attention to this thread well before I have the data (or author’s answer) in hand as I’m genuinely very interested if this is the real deal and want to maximize self-accountability. This is something I’ve been in pursuit of for years (literally).
