Skeptics Guide to the Universe Forums

General Discussions => Health, Fitness, Nutrition, and Medicine => Topic started by: lonely moa on February 12, 2019, 10:03:44 PM

Title: processed food and mortality
Post by: lonely moa on February 12, 2019, 10:03:44 PM
Just an observational study, but, something to consider.

https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2723626
Title: Re: processed food and mortality
Post by: bachfiend on February 12, 2019, 10:14:48 PM
Just an observational study, but, something to consider.

https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2723626

Without being able to read the full article (it’s behind a paywall) it looks like a pretty good study.  And of course there’s the almost mandatory authors’ caveat that there’s a need for further prospective studies.  Although, how you’d do a prospective trial is a mystery to me.

I certainly like the study results because it validates my prejudice against processed foods.
Title: Re: processed food and mortality
Post by: lonely moa on February 13, 2019, 12:20:12 AM

I certainly like the study results because it validates my prejudice against processed foods.

Snap.
Title: Re: processed food and mortality
Post by: fred.slota on February 13, 2019, 04:27:02 AM
Two comments.

1) The study wasn't on "processed food", but the more specific "ultraprocessed food" category, defined as " (from the NOVA food classification system), characterized as ready-to-eat or -heat formulations made mostly from ingredients usually combined with additives".  I politely suggest renaming the thread subject.  (also, gotta love the phrase "made mostly from ingredients usually combined with additives", as opposed to, you know, "made from ingredients unusually combined with additives".  Or just "made from ingredients".  Or possibly "made from additives".  "Made without ingredients"?)

2)
I certainly like the study results because it validates my prejudice against processed foods.

Does it?  Correlation is not causation.  Observed differences might have no link to the diet.
Title: Re: processed food and mortality
Post by: bachfiend on February 13, 2019, 06:22:56 AM
Two comments.

1) The study wasn't on "processed food", but the more specific "ultraprocessed food" category, defined as " (from the NOVA food classification system), characterized as ready-to-eat or -heat formulations made mostly from ingredients usually combined with additives".  I politely suggest renaming the thread subject.  (also, gotta love the phrase "made mostly from ingredients usually combined with additives", as opposed to, you know, "made from ingredients unusually combined with additives".  Or just "made from ingredients".  Or possibly "made from additives".  "Made without ingredients"?)

2)
I certainly like the study results because it validates my prejudice against processed foods.

Does it?  Correlation is not causation.  Observed differences might have no link to the diet.

Well, they tried to control for confounding factors, such as the subjects consuming increased ultraprocessed foods being younger, less affluent, less educated, etc.  At the risk of resuscitating a dead horse, it’s ‘correlation does not prove causation,’ rather than ‘correlation is not causation.’  Actually, if there is causation, there will be correlation, necessarily. 

Again, it would be extremely difficult to do a prospective controlled study.

I probably expressed myself badly.  I should have written ‘I certainly like the study results because it validates (in my own mind) my prejudice against processed foods.’
Title: Re: processed food and mortality
Post by: fred.slota on February 13, 2019, 08:10:22 AM
Two comments.

1) The study wasn't on "processed food", but the more specific "ultraprocessed food" category, defined as " (from the NOVA food classification system), characterized as ready-to-eat or -heat formulations made mostly from ingredients usually combined with additives".  I politely suggest renaming the thread subject.  (also, gotta love the phrase "made mostly from ingredients usually combined with additives", as opposed to, you know, "made from ingredients unusually combined with additives".  Or just "made from ingredients".  Or possibly "made from additives".  "Made without ingredients"?)

2)
I certainly like the study results because it validates my prejudice against processed foods.

Does it?  Correlation is not causation.  Observed differences might have no link to the diet.

Well, they tried to control for confounding factors, such as the subjects consuming increased ultraprocessed foods being younger, less affluent, less educated, etc.  At the risk of resuscitating a dead horse, it’s ‘correlation does not prove causation,’ rather than ‘correlation is not causation.’  Actually, if there is causation, there will be correlation, necessarily. 

Again, it would be extremely difficult to do a prospective controlled study.

I probably expressed myself badly.  I should have written ‘I certainly like the study results because it validates (in my own mind) my prejudice against processed foods.’

Sorry for my brevity/misquote.  My intention was "Correlation is not necessarily causation."  Thank you.

Couple of added thoughts.

1) Self-reported data, not the greatest - its what they had, I know.

2) I'm unclear when they said "The ultraprocessed foods group (from the NOVA food classification system), characterized as ready-to-eat or -heat formulations made mostly from ingredients usually combined with additives." if they were only counting ready-to-eat or -heat, or the entirety of NOVA group 4, which includes your carbonated drinks, snack foods, breakfast cereals, most diet items, etc.  Oh, and whisky, gin, rum and vodka.

3) The mortality is referred to in the paper several times as all-cause mortality.  That is, 602 deaths from an initial population of 44,551, but no differentiation of cause of death, it sounds like.  I'm curious if there's a more nuanced answer if there was a bit of granularity in that.  Is it death overall, or is there more correlation to certain types of death - heart attack, choking, food poisoning, auto accident, or even a stampede of wild elephants in your own home between 3:55 and 4:00 p.m. on the Fourth of July during a hailstorm (with one baby zebra).


Even with their dealing with the possible confounding factors that they had data for, plus your irrelevant reminder of "but Causation is Correlation", and the addition that your approval was based on your opinion, I'm still left with the conclusion that this study should not validate anything.  Correlation is not necessarily causation.


Title: Re: processed food and mortality
Post by: lonely moa on February 13, 2019, 01:29:27 PM
Death is clearly the easiest definite endpoint to identify.  A fairly important one as well.

I thought the authors made the limitations of the study fairly clear.
Title: Re: processed food and mortality
Post by: bachfiend on February 13, 2019, 02:32:20 PM
Death is clearly the easiest definite endpoint to identify.  A fairly important one as well.

I thought the authors made the limitations of the study fairly clear.

And of course, there also the mandatory caveat that there’s a need for further studies.
Title: Re: processed food and mortality
Post by: daniel1948 on February 13, 2019, 03:46:09 PM
I share the opinion that ultra-processed foods (using my own, intuitive but undefined, meaning for the term) to be unhealthy. Some ready-to-eat foods do not fit my definition: fruits and veggies eaten raw are ready to eat, but not ultraprocessed.

OTOH, bread seems to fit their definition of untra-processed: it is sold ready to eat, and is generally made with ingredients plus additives. A few regulars here regard bread as bad, but I feel safe in saying that most here, along with the medical and dietary communities, regard 100% whole-grain bread as healthy. I certainly do.

So I’m not clear on what ultra-processed foods are. I usually speak of junk foods, which I define as any finished product, intended to be eaten, in which there is a significant excess of sugar, fat, or salt. (Sugar, fats, and salt are not themselves intended to be eaten separately, so I don’t consider them as junk foods as sold.)
Title: Re: processed food and mortality
Post by: Harry Black on February 13, 2019, 05:36:35 PM
I agree that some foods, cooked or prepared by some methods are not optimal for health.
Title: Re: processed food and mortality
Post by: jt512 on February 13, 2019, 06:01:34 PM
Just an observational study, but, something to consider.

https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2723626 (https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2723626)

Without being able to read the full article (it’s behind a paywall) it looks like a pretty good study.  And of course there’s the almost mandatory authors’ caveat that there’s a need for further prospective studies.  Although, how you’d do a prospective trial is a mystery to me.

The authors conclude that more prospective studies are needed, not necessarily prospective trials.  The study itself is a prospective study.
Title: Re: processed food and mortality
Post by: jt512 on February 13, 2019, 06:15:50 PM
I'm still left with the conclusion that this study should not validate anything.  Correlation is not necessarily causation.

The point is that the reported relation between processed food and mortality is not mere correlation.  Rather, it is a relation observed after controlling for numerous potentially confounding variables.  This control can never be perfect, so it is rare that even the best prospective cohort study can be considered conclusive.  But you can't just brush the results off as "correlation."
Title: Re: processed food and mortality
Post by: bachfiend on February 13, 2019, 07:10:36 PM
Just an observational study, but, something to consider.

https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2723626 (https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2723626)

Without being able to read the full article (it’s behind a paywall) it looks like a pretty good study.  And of course there’s the almost mandatory authors’ caveat that there’s a need for further prospective studies.  Although, how you’d do a prospective trial is a mystery to me.

The authors conclude that more prospective studies are needed, not necessarily prospective trials.  The study itself is a prospective study.

OK, I stand corrected.  It’s not a controlled study though.  People have to choose themselves to consume ultra-processed food, and they may differ in some significant way difficult to control for, although the authors tried to control for the factors they could think of.  People eating ultra-processed food may be under stress that can’t be quantified.

I haven’t heard from you recently.  How is Hamburg?
Title: Re: processed food and mortality
Post by: arthwollipot on February 13, 2019, 09:15:53 PM
Did we already cover the difference between an "ingredient" and an "additive"? Isn't an additive just another ingredient?
Title: Re: processed food and mortality
Post by: bachfiend on February 13, 2019, 09:43:25 PM
Did we already cover the difference between an "ingredient" and an "additive"? Isn't an additive just another ingredient?

Well, an ingredient is a necessary part of a recipe.  An additive isn’t necessary, and can be deleted if wanted.  Additives are for taste or to preserve the food.

A meat pie wouldn’t be a meat pie if it doesn’t include meat as an ingredient.  A meat pie still would be a meat pie if it doesn’t include extra added salt as an additive.  It’s not just the quantity that’s important.  Yeast (or sourdough culture) are ingredients.  Omit them, and you don’t get bread.  Or at least, leavened bread.
Title: Re: processed food and mortality
Post by: fred.slota on February 13, 2019, 10:39:39 PM
I'm still left with the conclusion that this study should not validate anything.  Correlation is not necessarily causation.

The point is that the reported relation between processed food and mortality is not mere correlation.  Rather, it is a relation observed after controlling for numerous potentially confounding variables.  This control can never be perfect, so it is rare that even the best prospective cohort study can be considered conclusive.  But you can't just brush the results off as "correlation."
Yes, I can, and with greater strength than just a brush.  "numerous potentially confounding variables"? Really? They controlled for six possible confounding variables.  Six.  By my count, that leaves, ummm, all the rest.

You can't just jump to "causation".

Title: Re: processed food and mortality
Post by: lonely moa on February 14, 2019, 01:43:05 AM
There's this:

https://www.sciencedaily.com/releases/2019/01/190123191637.htm?fbclid=IwAR2qt8Ixp1Jv745LlaOzkKw7bFJCgr-debSnzFkcVZrXgHRYSrgyAzFIqFM

Title: Re: processed food and mortality
Post by: fred.slota on February 14, 2019, 07:34:31 AM
8 covariables.  Okay, now we're talking no other choice but to say this proves correlation.  I stand corrected.
Title: Re: processed food and mortality
Post by: jt512 on February 14, 2019, 07:55:37 AM
Quote
You can't just jump to "causation".

You‘re right. You can’t „jump“ to conclusion. Nobody is. But the study is evidence in favor of causation. And due to the prospective cohort design and the statistical control, it is stronger evidence than mere correction.


Sent from my iPhone using Tapatalk
Title: processed food and mortality
Post by: jt512 on February 14, 2019, 08:00:44 AM

I certainly like the study results because it validates my prejudice against processed foods.

Snap.

At least he admits it.


Sent from my iPhone using Tapatalk
Title: processed food and mortality
Post by: jt512 on February 14, 2019, 08:10:51 AM
Just an observational study, but, something to consider.

https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2723626 (https://jamanetwork.com/journals/jamainternalmedicine/article-abstract/2723626)

Without being able to read the full article (it’s behind a paywall) it looks like a pretty good study.  And of course there’s the almost mandatory authors’ caveat that there’s a need for further prospective studies.  Although, how you’d do a prospective trial is a mystery to me.

The authors conclude that more prospective studies are needed, not necessarily prospective trials.  The study itself is a prospective study.

OK, I stand corrected.  It’s not a controlled study though.  People have to choose themselves to consume ultra-processed food, and they may differ in some significant way difficult to control for, although the authors tried to control for the factors they could think of.  People eating ultra-processed food may be under stress that can’t be quantified.

Prospective cohort studies, like the present, are classified as controlled studies.  The quality of the evidence is much greater than would be from an uncontrolled study, like a cross-sectional study.

Quote
I haven’t heard from you recently.  How is Hamburg?

Hamburg is fantastic. I love it here. I will not be happy when I have to leave.

Sent from my iPhone using Tapatalk on a ferry on the Elbe.
Title: Re: processed food and mortality
Post by: fred.slota on February 14, 2019, 09:15:21 AM
Prospective cohort studies, like the present, are classified as controlled studies.  The quality of the evidence is much greater than would be from an uncontrolled study, like a cross-sectional study.

You‘re right. You can’t „jump“ to conclusion. Nobody is. But the study is evidence in favor of causation. And due to the prospective cohort design and the statistical control, it is stronger evidence than mere correction.

Yes, they are greater than uncontrolled.  And yes, they are stronger evidence.  But you are vastly overselling the improvement.

The base information in the study shows a linkage between ultraprocessed food and death.   Pure correlation.  Might be the cause, might be a side effect of some other cause.

Is this effect because UPF causes death, or because something else causes death and is linked with UPF? So you look at overweight people.  The study shows there is also a linkage between overweight people and death.  But control for weight, and you find that the UPF link is still present, independent of weight.  So the conclusion is stronger that UPF can cause death, independent of age.

You control for age, income, educational level, living alone and physical activity level, all of which the study shows are also linked with death, and you find that the UFP effect is still there despite all of those other influences.  Even stronger conclusion that UPF can cause death.

You've controlled for 6 factors.  6 very important factors.  You've got a stronger case.  But, there's still so much else in the world to consider.

Pre-existing conditions.  Depression.  Whether you have kids.  Left-handedness.  Being ginger.  Alcoholism. 

You are right to say that this study that attempted to control for some factors is better than an uncontrolled study.  I applaud them for doing that work.
You are right to say that this study better shows the possibility that UPF increases mortality.
You are right to say it is evidence in favor of causation.

But it is still a long way from being strong evidence, from being persuasive evidence. 




You shouldn't conclude aliens when you prove that a UFO isn't a satelite, weather balloon or bird.
You shouldn't conclude Nessie when you prove that a photo isn't shopped, a croc or a dolphin.
And you shouldn't conclude UPF causes death when you prove its not age, income or weight.

It might be an alien, Nessie or UPF.  But you still don't know.

This study should not validate your prejudice against UPF.
Title: Re: processed food and mortality
Post by: jt512 on February 14, 2019, 12:48:09 PM
Prospective cohort studies, like the present, are classified as controlled studies.  The quality of the evidence is much greater than would be from an uncontrolled study, like a cross-sectional study.


You‘re right. You can’t „jump“ to conclusion. Nobody is. But the study is evidence in favor of causation. And due to the prospective cohort design and the statistical control, it is stronger evidence than mere correction.

Yes, they are greater than uncontrolled.  And yes, they are stronger evidence.  But you are vastly overselling the improvement.

Please reread what I have written without reading anything into what I've written.  If after doing that you still think I'm "overselling" the study, please quote the statements that you feel constitute "overselling."

Quote
You've controlled for 6 factors.  6 very important factors.  You've got a stronger case.  But, there's still so much else in the world to consider.

The investigators controlled for sex, age, income, education, marital status, urbanization, BMI, physical activity, smoking, alcohol intake, energy intake, family cancer history, family CVD history, overall diet quality, number of diet reports, reporting season, and death within three years of enrollment.  That's 17 factors.  I have no idea where you got 6 (or 8 ) from.

Quote
You are right to say that this study that attempted to control for some factors is better than an uncontrolled study.  I applaud them for doing that work.
You are right to say that this study better shows the possibility that UPF increases mortality.
You are right to say it is evidence in favor of causation.

But it is still a long way from being strong evidence, from being persuasive evidence.

I doubt that the investigators care whether you applaud them or not.  Other than that, I agree with everything you wrote above.  This is not, however, the only evidence of processed foods increasing mortality.  The case is stronger when the totality of the evidence is considered.  That said, I agree with you that it is not completely convincing.  But that said, I think it is sufficient to warrant limiting one's consumption of processed foods.

Quote
This study should not validate your prejudice against UPF.

Huh?  What makes you think I have a prejudice against UPF?  Are you confusing me with bachfiend?
Title: Re: processed food and mortality
Post by: fred.slota on February 14, 2019, 01:17:44 PM
Sorry, linguistically my 'you's were/are a bit fluid.  I appear to have been you'ing directly to you, you'ing at other specific yous, and at other times you'ing at a rhetoric generic 2nd person you.

As to where I got 6, the full paper is behind a paywall.  The 6 I mentioned were the six mentioned in the Results section of the publicly readable section.