The following article by Dr David Morris unravels and corrects what lies behind the headline:
“High levels of B12 were associated with an increased risk of early death in a new study”
Dr David Morris.
Dr Morris has been a qualified doctor since 1994 – after training in internal general medicine he moved into General Practice in 2000 and practices as freelance GP and Integrative Physician. He has trained and practices in a wide range of integrative modalities including acupuncture, functional nutritional medicine and homeopathy. He holds Membership of the Royal College of Physicians – MRCP(UK) – and has a Postgraduate Diploma in the Study of Integrative Medicine.
B12 supplement study rebuttal
A recent paper published in the Journal of the America Medical Association (JAMA) 1 has been reported by various sources as indicating the B12 supplementation may be harmful and that “this paper adds to the growing body warning people that vitamin supplements may not have as much benefit as they claim.” Insider Health leads with the headline “High levels of B12 were associated with an increased risk of early death in a new study” and the web page address for this article is actually titled “B12 supplements could be lethal in high doses, study finds.”
The title of the study is the “Association of Plasma Concentration of Vitamin B12 With All-Cause Mortality in the General Population in the Netherlands” and a cursory look at the abstract indicates no such suggestion that B12 supplementation is any way harmful and this was not even a consideration of the study. It is tempting to simply dismiss the false interpretations and polemic claims against supplementation and simply move on. However, in this current age of increasing censorship of all that is not mainstream pharmaceutical based medicine it is clearly important we robustly challenge bad science and bad reporting.
So here follows a step by step appraisal of the study in an attempt to elucidate what it actually does or does not say on this subject. My apologies for some of the technicalities that I will use but this is necessary in this context.
What was the study?
This was a “post hoc longitudinal cohort study” which simply means that the authors looked back at a population they had studied, split the population into groups (cohorts) and performed an analysis -in this case B12 levels and overall mortality.
The population they were looking at was from an ongoing renal and vascular end stage disease study in the Netherlands and they analysed all the participants in whom there was a plasma Vitamin B12 level and looked at the overall mortality (death) rate over 10 years. In total there were 5571 patients studied after they had excluded people that were not suitable. They were followed up for median (average) time of 8.2 years the range of follow up being 7.7.to 8.9 years, so nearly a decade.
They split the population into four quartiles – i.e. ranked all the participants according to their B12 level and then split them into four groups – the lowest quarter of B12 levels, 26-50 % level , 51 -75 % and 76-100%.
Using statistical techniques they “corrected” for any differences in these groups that were not simply due to B12 e.g. age, sex or smoking and also any other diseases. They also corrected for various laboratory measurements and blood pressure etc.
Correction is very important – if, for example, one of the cohorts by coincidence had many more smokers this would distort the results and invalidate the study. To be fair, the correction appears to have been done thoroughly.
What were the quartile ranges?
||B12 range ( pg/ml)
||Less than 338.85
||338.85 – 397.13
||More than 455.41
There are two comments here:
Firstly the differences are quite small between the groups even from the lowest to highest quartile which makes it questionable how you would expect to see a difference in outcomes.
Secondly we have no idea of the distribution of B12 levels in the highest quartile as they do not give the range – there may be any number of patients in there with extremely high B12 levels which as I will come to later may well be very relevant and skew the results. They acknowledge that there was “a right skew” to the B12 levels – i.e. the levels in the 4th quartile were a wider range – and they used a mathematical technique to adjust this. For the purposes of mathematics and statistics this might be fine but the real life issue of some patients having potentially very high levels is not corrected by this. As I discuss later very high levels of B12 are caused by a number of serious diseases, not vice versa.
How were the results presented?
They compared the mortality difference between the quartiles as a ratio – they used the bottom quartile as the baseline and defined its mortality rate as 1.0. The other quartiles were then measured against the lowest quartile death rate i.e. if the death rate in quartile one is 20% and in quartile 4 is 40% then the mortality ratio is 2.0. The ratio of death rates is called the “Hazard Ratio.”
Here there is an important observation- although they split the population into the 4 quartiles they presented the results with the lowest and highest quartile but merged the second and third quartile. They state they did this “to improve the graphic presentation of the Hazard Ratios.”
I strongly suspect this is a sleight of hand to try and justify their conclusions. The way the results are presented it shows the lowest mortality in the lowest quartile, then it rises in the merged 2nd and 3rd quartile and then rises further in the upper quartile. This nicely presents a rising mortality rate with each rise in B12. However if we consider the possibility that the 3rd quartile had a lower mortality than the 2nd (or indeed than the 1st) then this nice tidy observation that higher B12 equals higher mortality is harder to justify.
The reason I strongly suspect this is what happened is that I can see no other reason when you have the data for all quartiles to merge two of them. In particular when you look at the graphs of survival (figure 3 in the paper for those that wish to look it up) then the overall survival rate in quartile 2 and 3 is better than quartile 1 until just before 5 years in the study and then it starts to creep just below quartile 1 – I can think of no biological reason why this should be so.
They used Kaplan-Meier curves and Kruskal-Wallis test, chi squared and various other statistical techniques to adjust for the variables between the groups, compute the mortality and calculate so called statistical significance in the result. I am not an mathematician, in particular not a statistician, so I have to accept their word for it that these were the correct statistical techniques – it is however true to say (I am not claiming this for this study) that incorrectly applied statistical techniques are sometimes used in studies and distort the results.
What else was done statistically?
This study was in patients in a renal study and so inevitably had more patients in it with renal impairment- for this reason there was an over representation of people with excessive protein in the urine. They used “statistical weighting method which they claimed “allows conclusions to be generalised into the general population.”
As stated already I am not a statistician but I am afraid I do not buy this and I think it is important to clarify what the population is actually being studied –
This is part of the ongoing PREVEND study (Prevention of REnal and Vascular ENd stage Disease), running in the city of Groningen, the Netherlands2 .The participants were selected by asking all inhabitants of the city of Groningen between the ages of 28 and 75 years (85,421 subjects) to send in a morning urine sample and to fill out a short questionnaire on demographics and cardiovascular history. 40,856 subjects (47.8%) responded – 30,890 subjects had a urinary albumin concentration of <10 mg/L and 9,966 subjects had a urinary albumin concentration of > 10 mg/L.
Pregnant women and people with diabetes were excluded and then all subjects with a urinary albumin concentration of >10 mg/L (n 7768) together with a randomly selected control group with a urinary albumin concentration of <10 mg/L (n 3395) were invited for further investigations in an outpatient clinic (total number 11163). Finally, 8592 subjects completed the total screening program- the actual study group.
The original PREVEND study has shown that elevated urinary microalbuminuria is a marker for renal dysfunction in the general population not just in those with diabetes. You can see from the figures above that the population level of high urinary albumin is 24.3% but in the study population the rate was 70% (of the 11,163 – we cannot determine the actual percentage of the final 8,592 but it is unlikely to be less.) In other words the level of potential kidney dysfunction in the B12 study was nearly 3 times that than in the general population.
I think it is a fairly reasonable rule of thumb that the more statistical adjustment that is required in a study then the less confidence we can have that the results are meaningful in real life.
More importantly we have good evidence that renal problems are a cause of elevated B12. Studies have shown that raised albuminuria, even without overt renal function decline, is associated with elevated B12 3.
The title of the paper is at best disingenuous – it states “the General Population” when it is anything but that. A statistical adjustment does not turn a renal/vascular study into the general population. The only reason I can think for this claim would be to try to increase the number of people picking up the study rather than it remaining in some more obscure specialist journal.
What about supplementation?
This is particularly interesting given the interpretation people have placed on the study. Firstly the study excluded anyone on B12 injections- this is clearly important because if supplementation is harmful then would be the ones we should most interested in. Secondly, with regard to oral or sublingual supplementation, prescribed or over the counter, they simply had no way of knowing. In other words the study simply has no data on supplementation.
What did the results show?
After fully correcting for any differences in the quartiles the results are as followed –
There were 226 deaths in total (4.1%) of the population. They converted the death rate in each quartile into the rate of deaths per 10,000 patient years – i.e. if there are 100 people in a group and over 10 years 5 people die then this is 5 deaths in 1,000 patient years which becomes 0.5 deaths per 10,000 patient years.
In the table below I have presented the results from the analysis of the fully “corrected” data –
||Quartiles 2 and 3
||1.16 – 2.97
A reminder the Hazard Ratio is the ratio of death rate (per 10,000 patient years) comparing to the rate in quartile one.
The confidence interval is a statistical calculation which indicates whether a result is simply considered a chance finding – in this context it the confidence interval is less than 1.0 then we consider this result to be likely by chance alone – hence the results between quartiles 2/3 and quartile 1 is considered to be by chance.
P value is the probability value – this indicates the odds of something being by chance or a real difference. The lower the P value the less likely the result is by chance.
Note – convention says that a P value of 0.05 (1 in 20) or less is “statistically significant” – this means only a 1 in 20 chance of being a fluke finding. It is important to be aware that 1 in 20 is simply an agreed convention and if I was told to cross the road with a 1 in 20 chance of getting run over I would be a little hesitant to take a step…
In this case the results suggests that the probability of the finding of difference between quartile 1 and 4 being a coincidence is 1 in a 1000 – which on the face of it seems pretty clear cut.
What does this actually mean?
The simple interpretation is this that having a higher level of serum B12 (above 455.41) means you are 1.85 times more likely to die over a time period of 8.2 years than someone who’s level is less than 338.85. Hence we see the polemic statements “high B12 levels nearly double your death rate.”
However it is a favourite trick of authors of studies to use relative risk rather than absolute risk –absolute risk is much more meaningful in real life.
To explain this it is useful to consider playing the lottery – in this case winning is a benefit not a risk (at least in theory… )
The relative “risk” of winning the lottery if you buy two lottery tickets instead of one is 100% higher or 2.0 times more likely.
However, the chance of winning the lottery with one ticket is 1 in 45,057,474 i.e. 0.0000002219% chance – so buying two tickets, despite relatively increasing your chances by 100%, in absolute terms improves your chances of winning by another 0.0000002219%. So now your chances are a stunning 0.0000004438%!
Back to the study
Over 8.2 years, in quartile one the odds of dying are 2.95% (nearly 3/100) and in quartile 4 the odds are 5.23% (just over 5/100) – as outlined this is a relative risk increase of 85%, which sounds very scary, but in fact is an absolute risk increase of 2.28%. To be clear that actual increased risk of dying if you have high B12 levels (as defined by this study) is 2.28% NOT 85%.
Correlation versus Causation
Having put the actual risk into some sort of perspective there is still more to consider – correlation (association) does not necessarily indicate causation.
If we screen the population for yellow stained fingers then we will find a correlation with an increased risk of lung cancer. However no one believes that yellow fingers cause lung cancer – the common link clearly is smoking.
This example is pretty clear, but there many cases where there has been misleading confusion between correlation and causation. For example we were advised for many years that HRT was a good protector for heart disease because women on HRT, in the largest observational study at the time, had less heart disease – many years on further analysis indicated that this was because healthier women were more likely to seek HRT and if anything once this was corrected for there was an increase in heart disease form taking HRT. (Note -I am aware that there are ongoing debates and studies in this areas that might indicate benefit but the point is that the evidence we were using at the time was flawed.)
Even when all factors are corrected however we must take care not to presume causation and for this reason Sir Austin Bradford Hill, who was a British epidemiologist, set out nine criteria to prove causation – he is best known for his collaborative research with Sir Richard Doll, which linked smoking with cancer and other chronic illness. They did the very first work proving smoking caused cancer.
The criteria are as follows –
1 Strength – death rate from lung cancer in smokers was 9-10 times that of non-smokers. This is why Hill and Doll reported that smoking caused lung cancer, not just that they were associated. It is worth noting that Hill did not consider that a correlation factor (i.e. Hazard Ratio) below two times was sufficient to claim causality.
2 Consistency – do we find the association in different people, in different circumstances at different times? Is the observation consistent? In the case of elevated B12 the answer is no – various studies once properly corrected have not shown an association with high levels of B12 and mortality 4,5.
3 Specificity –this is asking very precisely “what specifically/exactly is the relationship?” e.g. can we show that high levels of B12 for a specific number of years or at a particular level are tightly linked to mortality and the answer is clearly no.
4 Temporality – perhaps best called “direction of causation.” With an association – which is the cart and which is the horse? I.e. what comes first? This is an extremely important point with respect to high B12 levels because we know that a significant number of serious diseases (see below) cause high B12 levels.
5 Biological gradient –this an alternative way of saying “dose response”. In the case of smoking the longer and the more you smoke the higher the risk. In the case of B12 in this study we are not able to determine a dose response and indeed the merging of quartiles 2 and 3 strongly suggests otherwise as I have already commented.
6 Plausibility – this is easy to understand – when we observe smoking and lung cancer together it is plausible that one causes the other. Of course in the case of yellow fingers there is no plausibility of them directly being the cause of lung cancer.
So is there a plausible mechanism by which high B12 levels might be a cause of harm? The simple answer is no and I will detail this below.
7 Coherence – does the association fit with other facts? If we see an association between smoking and lung cancer and we have seen an increase in population smoking and an increase in lung cancer cases then this is coherent. In the case of B12 levels and mortality we have no evidence to support this.
8 Experiment – Hill suggested that any experimental evidence would be useful to determine whether association was indeed causation. As per “consistency” there have been numerous studies of B12 supplementation that have not shown increased mortality 4,5 .
9 Analogy – have we seen similar before? For example, we have seen evidence that some drugs are associated with birth defects, so if we see an association between a new drug and birth defects we can see an analogy. In the case of B12 this does not exist.
In summary this study does not in fact meet any of the Braford Hill criteria that would suggest causation rather than association.
What are the causes of high Vitamin B12 other than supplementation?
It is well known that a number of serious diseases are a cause of raised vitamin B12 so this really asks us question whether the study has simply put the cart before the horse.
A review paper titled “The Pathophysiology of Elevated Vitamin B12 in Clinical Practice”6 is an excellent summary:
Solid tumours and non-malignant liver disease
Solid tumours, particularly liver but also breast, colon, stomach and pancreatic are strongly associated with high B12. In primary liver cancer the odds ratio is 4.7 and for cancers with liver metastases the odd ratio is as high as 6.2 – this is consistent very strongly with causation. The levels of B12 in these disease can be extremely high which would of course put them in the upper quartile of the study. It is important to be aware too that when high B12 levels are investigated most cases of malignancy have been undiagnosed and not metastasised i.e. the high B12 is the first indicator of malignancy.
Hence potentially well people could have been included in the study but in whom the high B12 actually indicated a significant life threatening disease was about to manifest.
The mechanism for high B12 in the case of liver disease is elevation of serum binding cobalamins from hepatocyte damage and reduced clearance of B12 by the liver – this is also the reason why non-malignant liver disease, such as cirrhosis, are also a cause of high B12. In the other solid tumours the mechanism is believed to be both the secretion of binding proteins and also stimulation of white cell overproduction, as below.
Malignant blood diseases including for example chronic myeloid leukaemia, the acute leukaemias and other myeloproliferative diseases all elevate B12. The mechanism being white blood cells (granulocytes) releasing B12 binding proteins.
I have already elucidated the connection between renal disease and high B12 – the suggested mechanism for this is elevation of serum B12 binding proteins due to reduced renal clearance.
A further piece of the jigsaw is those people in whom there is a genetic reason affecting the function of the transcobalmin II bound B12 (holocobalamin) receptor. This specialised receptor is the mechanism by which B12 is taken into the cells in healthy people. Genetic disorders in this receptor lead to high serum B12 levels but actual functional deficiency. The proportion of the population affected by a number of different genetic changes is currently unclear but probably runs at least around 1-2% so is not uncommon7. It stand to reason that these individuals are at higher risk of several diseases due to their functional B12 deficiency despite elevated serum levels.
Is there a mechanism by which high B12 levels might be a cause of harm?
We clearly know that inappropriate supplementation of some nutrients can be harmful e.g. excess iron -due to causing mitochondrial dysfunction -and of course excess of the fat soluble vitamins are harmful. We have a biological understanding as to why these are harmful. To date no one has been able to indicate how excess B12 can cause harm.
It is notable that the Institute of Medicine. Food and Nutrition Board does not set an upper limit for B12 intake because of its low potential for toxicity. It states “no adverse effects have been associated with excess vitamin B12 intake from food and supplements in healthy individuals”8
Findings from intervention trials support this conclusions. For example in one trial 9, vitamin B12 supplementation (in combination with folic acid and vitamin B6) did not cause any serious adverse events when administered at doses of 1.0 mg daily for 5 years.
A caveat – what about cyanocobalamin?
At the risk of appearing to contradict everything I have written there is one area of supplementation that is of potential concern and that is the use of cyanocobalamin. Cyanocobalamin is an artificial compound of B12 created by the reagents used in X-ray crystallography in the 1940’s when the structure of B12 was first established.
This occurred because of B12’s high affinity for cyanide- the treatment for cyanide poisoning (if you are quick enough!) is intravenous hydroxycobalamin.
It has always struck me as an absurdity that this artefactual compound has been used for the next 80 years as a supplement –in the UK NHS this is the only prescribable oral form, in the USA it is commonly used as an injectable form, and of course it is the form found in the majority of purchased supplements.
Cyanocobalamin is normally only found in the body in trace amounts due to ingestion of small amounts of cyanide mainly from smoking but also other environmental toxicity.
The reason for raising cyanocobalamin supplementation as a potential issue is because despite the common claim that the amount of cyanide that is released and needs to be detoxified is miniscule, in fact in renal impairment there is evidence to indicate concern. Given the nature of the population this study was performed on then this is particularly relevant.
A study10 has shown that cyanide metabolism is abnormal in individuals with Chronic Kidney Disease due to decreased clearance. In this study injection of methylcobalamin improved nerve damage that was caused by renal failure – “uraemic nephropathy.”
It was shown that injecting methylcobabalamin increased the serum levels of cyanocobalamin which was then removed from the body by urinary excretion i.e. the body uses methylcobalamin to mop up cyanide by forming cyanocobalamin that can then be excreted. Cyanocobalamin being excreted by the kidney at three times the rate of methylcobalamin.
Clearly if we give cyanocobalamin instead of methyl cobalamin not only are we not providing mechanism for mopping up cyanide but we are also increasing the total cyanide burden on the body.
So overall what can we conclude?
Having indicated that they excluded injectable B12 from the study and that they had no data on oral supplementation then the authors make some exceptionally questionable statements. In the discussion they acknowledge “the underlying mechanism of the association of plasma concentration of Vitamin B12 with mortality is incompletely understood” but then go on to say “The results of this study could also be clinically interpreted in the context of oral vitamin B12 supplementation.”
This is an entirely arbitrary statement and completely unjustified by the study – the paucity of evidence for stating this is indicated by them supporting their statement with two referenced studies of B vitamin supplementation neither of which indicated any issues of mortality. The first paper simply showed no benefit of using Vitamin B supplementation (not just B12) on the progression of coronary heart disease. The second paper suggested a possible increased association of hip fracture with B vitamins but in fact no statistical increase with B12 supplementation and again no comment on mortality.
I think this has been sufficient to “set the hares running” by those with vested interests against any supplementation recommendation. Charitably we could describe this as lazy reporting/journalism although my inclination is to describe it as deliberately biased reporting.
This study, which despite its title is NOT a general population study, does not provide any evidence whatsoever that supplementing B12 has any potential to be dangerous. It fails even on its basic premise that high B12 is a cause of increased mortality. The link between high B12 and mortality is much more clearly explained in reverse i.e. diseases that are more likely to kill you are causes of high B12. All the other data we have indicates no cause for concern with B12 supplementation other than cyanocobalamin in renal impairment, but this is not a formulation I would ever recommend.
It widely acknowledged that B12 deficiency is rife, often hard to determine and definitively a cause of higher risk of a number of serious diseases – dementia, vascular disease, neurological disease, mental health disorders etc. – these are not simply associations as we can see mechanisms by which low B12 is causative. We also know that a failure to intervene in timely manner with B12 deficiency can lead to irreversible neurological damage. Any attempt to scaremonger people away from addressing these risks with appropriate supplementation is clearly to be lamented.
We’d love your comments on this article. It’s easy, just post your questions, comments or feedback below.
Dr Morris qualified as a medical doctor in 1994 and spent six years in hospital medicine – mostly in general adult medicine, but also in paediatrics and Accident and Emergency.
In 2000 David moved into family general practice and was a GP partner for many years. During this time he was also extensively involved in commissioning health care services.
Dr Morris has significant training and experience in complementary therapies such as acupuncture and homeopathy, and ran a primary care based pain clinic for over a decade using acupuncture therapies.
With many thanks to Dr Morris for this blog.
- Association of Plasma Concentration of Vitamin B12 With All-Cause Mortality in the General Population in the Netherlands. Jose L. Flores-Guerrero, MD, et al JAMA Netw Open. 2020;3(1):e1919274. doi:10.1001/jamanetworkopen.2019.19274.
- Urinary Albumin Excretion Is Associated with Renal Functional Abnormalities in a Nondiabetic Population Sara-Joan Pinto-Sietsma et al JASN October 2000, 11 (10) 1882-1888.
- The association between vitamin B12, albuminuria and reduced kidney function: an observational cohort study. McMahon GM et al BMC Nephrol. 2015 Feb 2;16:7. doi: 10.1186/1471-2369-16-7.
- High vitamin B12 levels are not associated with increased mortality risk for ICU patients after adjusting for liver function: a cohort study. Fiona M Callaghan et al ESPEN J. 2014 Apr 1; 9(2): e76–e83.
- Plasma Homocysteine, but Not Folate or Vitamin B-12, Predicts Mortality in Older People in the United Kingdom Alan D. Dangour et al The Journal of Nutrition, Volume 138, Issue 6, June 2008, Pages 1121–1128.
- The pathophysiology of elevated vitamin B12 in clinical practice Andrès et al QJM: An International Journal of Medicine, Volume 106, Issue 6, June 2013, Pages 505–515.
- Cellular Uptake of Cobalamin: Transcobalamin and the TCblR/CD320 Receptor Edward V. Quadros et al Biochimie. 2013 May; 95(5): 1008–1018.
- Institute of Medicine. Food and Nutrition Board. Dietary Reference Intakes: Thiamin, Riboflavin, Niacin, Vitamin B6, Folate, Vitamin B12, Pantothenic Acid, Biotin, and Choline. Washington, DC: National Academy Press, 1998.
- Homocysteine lowering with folic acid and B vitamins in vascular disease. Lonn E et al. N Engl J Med. 2006;354:1567-77.
- Koyama K et al. Abnormal cyanide metabolism in uraemic patients. Nephrol Dial Transplant. 1997;12:1622–8.
- Effect of homocysteine-lowering B vitamin treatment on angiographic progression of coronary artery disease: a Western Norway B Vitamin Intervention Trial (WENBIT) substudy. Løland KH et al Am J Cardiol. 2010 Jun 1;105(11):1577-84.
- Association of High Intakes of Vitamins B6 and B12 From Food and Supplements With Risk of Hip Fracture Among Postmenopausal Women in the Nurses’ Health Study. Meyer HE et al JAMA Netw Open. 2019 May 3;2(5).