I already hear all the health influencers of the world doubling down on their recommendations of affiliated omega-3s supplements taking this very specific population as an example !!
Given the poor prognosis of dialysis patients and the relative benignity of fish oils: I agree with the concept to recommend them until opposing data suggests otherwise.
1. Kidneys are unique in the body in that they are a vast bed of microvessels, have high blood flow, and don't kill you as soon as similar impairments in other microvascular beds. Kidney dialysis implies atherosclerosis is worse in the kidneys than elsewhere. My suggestion is that diseased kidneys are uniquely positioned to be microclot factories. These wouldn't cause CV events themselves, but they could cause a spillover of pro-thrombotic chemicals which circulate and cause clots elsewhere. Fish oil preventing microclots in the kidney microvessels could thus have systemic effects that are both large and broad.
2. Efficacy of fish oil and harm of corn oil and not mutually exclusive. If these are both acting at once then more uniform benefit/harm across outcomes might be expected.
No specificity on the oils used. What type of fish oil, from what source, farmed fish get a lot of antibiotics and tend to be unhealthy, how is the oil extracted --as in could there be excipients that might mess with kidneys further. Similarly for corn oil--corn is festooned with toxic pesticides--are the oils tested for toxins? Also, again the extraction process is important. If it is poor quality oil, it is toxic to begin with--not likely to help.
If a deficiency in omega 3 is a known phenomenon in dialysis patients, and a known effect modifier, this might explain the large effect (particularly if the fish oil grp has more participants with a deficiency in omega 3 as a result of imbalanced randomization on this variable). Interestingly table 1 does not mention omega 3 as a baseline variable, and randomization did not stratify by omega 3 deficiency (i.e, baseline lab testing for n−3 fatty acids in plasma phospholipids). For more, see Nutrition Users’ Guides on RCTs (based on JAMA Users’ Guides framework): https://pmc.ncbi.nlm.nih.gov/articles/PMC11773643/
My healthcare focus largely begins and ends with all-cause mortality. That said, quality of life and disease/cost burdens are real, too. As a pharmacist pointed out below, the risks are few to give fish oil--the costs next to nothing. Several have pointed out that corn oil may be a huge negative--which would, for me, be almost as big of a headline. On the other hand, anything unexplained in medicine now tends to get assigned to "inflammation" without much afterthought. Obviously inflammation is problematical, but it is also evolutionary. It is not a pure negative. The science in this small study seems plain. Give the fish oil until contra-indicated. All cause mortality suggests no big deal. As a person who has worked many years with complex data and who has taught statistics, I don't get too excited about the smooth lines or early demarcation, but they are indeed "notable." Red flags might be too strong.
Contradictory study results don't cause me to scratch my head. I just figure the factor under investigation is probably insignificant. I also give biological probability a lot of weight in my assessment. Given the hundreds or thousands of biochemical transformations of molecules that occur in the digestive tract and then further changes within the cells of the intestinal lining, and further in the liver and other organs, it has hard to imagine there is much significance in the foods we eat. Consult any reputable textbook of nutritional biochemistry and draw your own conclusions.
As a relatively new conspiracy theorist, I remember when we were all told to stop taking fish oil supplements due to potential for mercury contamination.
As others have noted, the immediate separation of event curves is a bit curious. OTOH, the endpoints were of the harder-outcome variety…the only one that might be a bit clinician-dependent is PVD leading to amputation…so it would seem the likelihood of malfeasance or bias is lower.
I'm not a researcher or in the medical field, but I am a statistician. I've done many trials in the corporate world (marketing, effectiveness of process changes, etc.) over the years. Two things about the curves jump out to me.
The first is that the curves separate immediately. It's possible that results are immediate, but still I would have expected more "jumpiness" in the curves early. There are so few events in the first few months that a relatively smooth curve is unlikely. I realize they've done some smoothing, but the immediate separation of curves is suspicious.
The other issue is the curves maintain roughly the same slope for the entire range. Maybe that's an artifact of the population (hemodialysis patients), but it's still interesting. I would have expected both curves to bend downwards over time as the really sick people have "events", not maintain a roughly straight line. Maybe also this is an artifact of what the UK researcher mentioned as the "multiple recurrent events".
Looking at this from my vantage (corporate issues, looser controls than medical research) my first thought was that we're looking at different populations. Yes, I know this is double blinded with extensive controls to make sure the treatment and control groups were identical. Just sayin', that's the first place I would look for answers.
Not a clinician or a statistician, but the immediate curve separation is a massive red flag. I've seen that in other studies, and the typical reason for that is because there's something fundamentally different about the control vs. treatment groups that wasn't taken into account.
Was there an increased risk of bleeding for patients in the fish oil group who were also on blood thinners?
Seems obvious but maybe corn oil is bad for these patients. Is there a study showing corn oil is neutral? If not, why was it used for the control arm?
I already hear all the health influencers of the world doubling down on their recommendations of affiliated omega-3s supplements taking this very specific population as an example !!
Given the poor prognosis of dialysis patients and the relative benignity of fish oils: I agree with the concept to recommend them until opposing data suggests otherwise.
Two hypotheses not yet clearly spelled out:
1. Kidneys are unique in the body in that they are a vast bed of microvessels, have high blood flow, and don't kill you as soon as similar impairments in other microvascular beds. Kidney dialysis implies atherosclerosis is worse in the kidneys than elsewhere. My suggestion is that diseased kidneys are uniquely positioned to be microclot factories. These wouldn't cause CV events themselves, but they could cause a spillover of pro-thrombotic chemicals which circulate and cause clots elsewhere. Fish oil preventing microclots in the kidney microvessels could thus have systemic effects that are both large and broad.
2. Efficacy of fish oil and harm of corn oil and not mutually exclusive. If these are both acting at once then more uniform benefit/harm across outcomes might be expected.
No specificity on the oils used. What type of fish oil, from what source, farmed fish get a lot of antibiotics and tend to be unhealthy, how is the oil extracted --as in could there be excipients that might mess with kidneys further. Similarly for corn oil--corn is festooned with toxic pesticides--are the oils tested for toxins? Also, again the extraction process is important. If it is poor quality oil, it is toxic to begin with--not likely to help.
If a deficiency in omega 3 is a known phenomenon in dialysis patients, and a known effect modifier, this might explain the large effect (particularly if the fish oil grp has more participants with a deficiency in omega 3 as a result of imbalanced randomization on this variable). Interestingly table 1 does not mention omega 3 as a baseline variable, and randomization did not stratify by omega 3 deficiency (i.e, baseline lab testing for n−3 fatty acids in plasma phospholipids). For more, see Nutrition Users’ Guides on RCTs (based on JAMA Users’ Guides framework): https://pmc.ncbi.nlm.nih.gov/articles/PMC11773643/
Calling “Corn oil” a placebo is dishonest.
It could be that the corn oil is the problem.
Draws from urns with red and white beads.
fish oil Urn A: red = 127 out of 610, 20.82%
corn oil Urn B: red = 208 out of 618, 33.66%
If we start from idea that there is no difference [FN1]; the p-bar is 27.28% [FN2]
Draw 610 from Urn A should have no more than 32.69% red, nor less than 21.87% (3 std)
Draw 618 from Urn B should have no more than 32.66% red, nor less than 21.91% (3 std)
Yes, fish oil Urn has less events than we should see. But it is also true that corn oil Urn has MORE events than we should see.
So it makes just as much sense to look for why corn oil was apparently unhealthy as it does to look for why fish oil was apparently salubrious.
There should have been a 3rd arm where patients get neither.
[FN1] Unless there is a good mechanism, I think this should be the presumption.
[FN2] Unless you have other data showing some different expected result, the p-bar must be derived from the data at hand.
JDK
My healthcare focus largely begins and ends with all-cause mortality. That said, quality of life and disease/cost burdens are real, too. As a pharmacist pointed out below, the risks are few to give fish oil--the costs next to nothing. Several have pointed out that corn oil may be a huge negative--which would, for me, be almost as big of a headline. On the other hand, anything unexplained in medicine now tends to get assigned to "inflammation" without much afterthought. Obviously inflammation is problematical, but it is also evolutionary. It is not a pure negative. The science in this small study seems plain. Give the fish oil until contra-indicated. All cause mortality suggests no big deal. As a person who has worked many years with complex data and who has taught statistics, I don't get too excited about the smooth lines or early demarcation, but they are indeed "notable." Red flags might be too strong.
Contradictory study results don't cause me to scratch my head. I just figure the factor under investigation is probably insignificant. I also give biological probability a lot of weight in my assessment. Given the hundreds or thousands of biochemical transformations of molecules that occur in the digestive tract and then further changes within the cells of the intestinal lining, and further in the liver and other organs, it has hard to imagine there is much significance in the foods we eat. Consult any reputable textbook of nutritional biochemistry and draw your own conclusions.
As a relatively new conspiracy theorist, I remember when we were all told to stop taking fish oil supplements due to potential for mercury contamination.
As others have noted, the immediate separation of event curves is a bit curious. OTOH, the endpoints were of the harder-outcome variety…the only one that might be a bit clinician-dependent is PVD leading to amputation…so it would seem the likelihood of malfeasance or bias is lower.
Was there any difference in atrial fibrillation between the groups??
Have you been in communication with the authors?
I'm not a researcher or in the medical field, but I am a statistician. I've done many trials in the corporate world (marketing, effectiveness of process changes, etc.) over the years. Two things about the curves jump out to me.
The first is that the curves separate immediately. It's possible that results are immediate, but still I would have expected more "jumpiness" in the curves early. There are so few events in the first few months that a relatively smooth curve is unlikely. I realize they've done some smoothing, but the immediate separation of curves is suspicious.
The other issue is the curves maintain roughly the same slope for the entire range. Maybe that's an artifact of the population (hemodialysis patients), but it's still interesting. I would have expected both curves to bend downwards over time as the really sick people have "events", not maintain a roughly straight line. Maybe also this is an artifact of what the UK researcher mentioned as the "multiple recurrent events".
Looking at this from my vantage (corporate issues, looser controls than medical research) my first thought was that we're looking at different populations. Yes, I know this is double blinded with extensive controls to make sure the treatment and control groups were identical. Just sayin', that's the first place I would look for answers.
Not a clinician or a statistician, but the immediate curve separation is a massive red flag. I've seen that in other studies, and the typical reason for that is because there's something fundamentally different about the control vs. treatment groups that wasn't taken into account.