49 comments

[ 1.8 ms ] story [ 103 ms ] thread
I have to agree. The lower end of the 95% confidence interval on the hazard ratio is 1.01, which means this result is just barely significant at a p-value of 0.05. (The null hypothesis is a hazard ratio of 1.) After any kind of multiple testing correction, this result is not significant.
I'm sure that we'll discover many new things to study with proper dredging. In fact under some conditions it is called Unsupervised Machine Learning and viewed in a positive light.

However, as stated, these findings could be used to form hypothesis and then to study them. These findings are _not_ conclusions.

Of course, Randal Munroe says it better than I could: https://xkcd.com/882/

To be fair, the authors are open about the shortcomings:

> further research is needed to replicate these findings, determine their significance, and define the underlying mechanism.

I don't know that the level of significance reported in the paper even justifies further research.
> To be fair, this is more or less "boiler-plate text" added at the end of every research paper

There is usually something discussing future research needs, but it's not always oriented toward “replicate and determine significance”.

Though something very much like the note on my this paper is common, for good reason, on papers reporting a novel, borderline significant association with no clear mechanism.

>To be fair, the authors are open about the shortcomings:

>> further research is needed to replicate these findings, determine their significance, and define the underlying mechanism.

To be fair, this is more or less "boiler-plate text" added at the end of every research paper, sometimes I would like to see a research paper that ends with "... and that's all about it".

So there really is something about the blood of virgins?
(comment deleted)
I hate to break it to you, but never pregnant != virgin
Never pregnant is a superset of virgin. According to most
Welllll....take a gold star lesbian. She is technically a virgin if you measure virginity by penis-in-vagina sex. She still can become pregnant through artificial insemination. Plus the Virgin Mary!
Don't forget Schmi Skywalker. ;-)
Their comment seemed to be partially in jest... I don't imagine they believe never pregnant = virgin.
Or maybe only very pragmatic, how would you make sure - should you want to procure never-pregnant female blood - to select the right donours?
Not completely surprised.

not exactly the same, but in rats young blood makes older rats more healthy and the younger ones less so due to stem cells: https://www.nytimes.com/2014/05/05/science/young-blood-may-h...

(making the assumption more pregnancies = higher age)

If it was related to higher age, wouldn't blood donations from older males also be problematic?
yeah, but as far as I saw they didn't distinguish age at all. they did men, ever pregnant women, and non ever pregnant women.
Has anyone ever studied this stuff before? You think they'd have noticed that if true...
This is an interesting study, and I highly encourage everyone to also read the accompanying editorial, which raises some important methodological shortcomings. See here: https://jamanetwork.com/journals/jama/fullarticle/2657356

The reason this study was done because there has been a very well replicated association between the plasma (not red cells) donated from previously pregnant women and the risk of TRALI, a very rare but deadly transfusion reaction. After this association was made clear, blood banks no longer collect plasma from previously pregnant ("ever pregnant") women, and the overall rate of TRALI has dramatically dropped (from about 35 fatalities in the US per year to ~15).

Source: MD/lab medicine resident

Do you have any insight into why, when they were looking at men vs. two types of women, 88% of the blood donors were men? Seems like you would want at least 50% women.
I don't think an ethics committee would approve of exposing a study group to a known or suspected higher risk of death without significant safeguards.

Using a small sample exposed to the suspected risk and the vast majority not exposed sounds like responsible ethics.

It is not stated, but there should also have been additional safeguards and monitoring. It is normal for studies to be terminated early if they are found to expose participants to either too much danger or if the study group is found to be significantly advantaged (in which case, for example, the whole group is given the beneficial treatment instead of just the small study group).

50% may sound 'fair' or 'equal', but this is about researching ways of saving life, some social notion of fairness or gender equality is not relevant and, if this study is correct, could have caused excessive death. Is it worth people being killed to get the gender balance into the shape you prefer?

> Is it worth people being killed to get the gender balance into the shape you prefer?

This is obviously a tough question to answer, and maybe the answer here is no. Further, I wouldn't say "worth people being killed" as much as I'd say it's "worth risking people dying". Obviously this is context sensitive and is different for each treatment and study.

Despite this, as the devils advocate, you could say that given that men and women can respond to different treatments in a range of ways, the ideal distribution would be diverse enough to accommodate for those differences, which would provide a better statistical subset and by extension producing "better" data. Same caveats apply, where you mentioned "too much danger" and if the study is "significantly advantaged". In short, it could be less about fairness (equity) or equality and more about collecting a broader or more diverse range of data.

Correct answer is always no. No one should be knowingly killed or endangered to suit a biased view.
It seems as though people are getting the idea that this was a randomized trial, where people were assigned to receive units of blood from various types of donors. However, it was a retrospective case-control study, where records from everyday clinical practice were analyzed after the fact to look for patterns or, as in this case, try to answer questions. Did people who had received units of PRBCs from various types of donors fare any differently? I tend to view this kind of study as valuable for hypothesis generation, but not for solid conclusions, because it's impossible to control for every possible confounding factor. So the notion of ethics committees being concerned about patient's mortality doesn't apply here. All patients in the study have already either survived the study period or not. Same for safeguards, monitoring, and early termination.
Not OP, but have sold plasma before. The typical person selling plasma is at the bottom 1% of the income distribution, turns out there are a lot more homeless men than women so I would venture that it's an availability issue.
thanks for defining "ever pregnant", I thought this was some weird birthcontrol thing as in faking the body to thinking its already pregnant perpetually. "Ever pregnant" as in "has been pregnant at least once" or as you say "previously pregnant".

Also for those who want to know https://en.wikipedia.org/wiki/Transfusion-related_acute_lung...

Can I ask where you're from? I don't personally see a difficulty using "ever pregnant" to mean "has ever been pregnant".
from the states, but I guess my thoughts landed on phrases like "ever more", which makes me wonder if that's actually two words... nope it's a single word: evermore. Looks like I was interpreting the word in its first definition, but it was meant in the 2a definition

Definition of ever

1 :always; eg * ever striving to improve * the ever-increasing population 2a :at any time * more than ever before 2b :in any way * how can I ever thank you 3 —used as an intensive * looks ever so angry * am I ever happy to see you

Hmmm... I don't like definition 2b. I think senses (1) and (2a) as given here are both related to the core "ever" concept "encompassing all of time", and "how can I ever thank you" rhetorically emphasizes that no matter how long you spend trying, you'll never be able to thank the person properly.
If one cannot exhaustively find a sufficient reply to give in thanks after so much time than I'd say that's logically identical to being unable to reply "in any way". the emphasis of the former is on the search and it's ultimate futility, the emphasis of the latter for the existence lacking in the first place which precludes any successful search.
> If one cannot exhaustively find a sufficient reply to give in thanks after so much time than I'd say that's logically identical to being unable to reply "in any way".

That is true, but that's not evidence that "ever" can mean "in any way". "You can count on me" and "I won't let you down" are precisely equivalent at the full-sentence level, but there are no equivalences between the individual words (barring "you"/"you" and "I"/"me").

I'm a native English speaker and I read "ever pregnant" like you would read "evergreen".

Evergreen trees are always green, so ever pregnant woman are...

Ditto. Native English speaker. “Ever-Pregnant” definitely sounds more like “Permanently Pregnant” than “Has Been Pregnant” to me.
Same here. Was perplexed at the title until I read the article.
I had the same problem but the abstract uses the term (so I guess it's legit) and it was hard to make a shorter title without it.
Should be able to detect some rogue protien.
As a guess, a follow up study should maybe consider that chimerism is a factor. IIRC, a woman remains a chimera for about 27 (or 27.5?) years after giving birth (ie her blood contains DNA from the child she carried). So, you could look at data for female blood donors whose last pregnancy was more than 27 years ago and those whose last pregnancy was less than 27 years ago.
I wonder if that data is even captured in the donation process?

Aren’t donations pooled? Can’t imagine this would leave very many pools without a single ever-pregnant donor.

Ya got me. I don't know much about blood donation. I'm just a mom who has reason to be curious about medical side effects of having had children.
As far as I know only Plasma doses are pooled because of the washing process.

Pooling is also done for blood testing they’ll pool samples from a batch together and if the pooled sample comes back positive they’ll discard the entire batch of part of it if subpooling was used.

Red blood cells doses come from a single donor.

I’m not sure about platelets.

There is no way that the number of years fetal DNA remains in a mother's system is precise to 3 significant figures.
Please use the original title.
I tried to, but the original title "Association of Blood Transfusion From Female Donors With and Without a History of Pregnancy With Mortality Among Male and Female Transfusion Recipients" was approximately double the maximum length allowed by HN. I actually spent about 10 minutes trying to figure out the smallest change that was still grammatically correct and fit the 80 char limit while being as accurate as possible.

That being said, if you have an improved version that fits, I'd be happy to support that.

Likewise. That was a particularly hard title to crack.
<title> says: "Association of Female-Donor Blood Transfusion and Recpient Mortality"
Probably better. I almost used it but thought people would complain about it burying the lede.
I only opened this because I thought "Ever-Pregnant" meant never not pregnant.