No, your supplement stack won’t kill you

Before we start, let me remind everyone that we need to invest in robust gerotherapeutics – that is small molecule drugs - that will substantially extend lifespan. So please support real, hardcore aging research, which ranges from basic to translational. 

Here are three ways to support us: 
https://www.levf.org/donate 
 
Buy and hold VITA tokens to support VitaDAO 
https://www.vitadao.com/

 
Listen to the VitaDAO Aging Science Podcast 

 
 

Meanwhile, there is time to discuss what we can currently do. 

Many people are taking supplements, some of which could indeed extend your lifespan. Emphasis on could. The idea is certainly not far-fetched if you have followed the science. But what happens when you combine multiple promising supplements and substances? Some people like Andrew Steele recently argued on twitter that this is likely dangerous.  

https://twitter.com/statto/status/1803784862563697093

  

My stance is that we do not know enough about combinations, but I consider it unlikely that they would be very harmful and Andrew’s proclamation too certain. 

Should you supplement 40 different compounds? Probably not because many of the compounds just by themselves already lack strong evidence. Would it be harmful to take 5 or 10 different supplements? I doubt it. In fact, most treatments are likely to be additive, albeit with diminishing returns. 

Cohort studies and intuitions about combinations 
Let us start with what I will call “The argument from diet”. Consider this. Our diet is full of bioactive substances and yet there is little evidence for unexpected interactions, where two healthy dietary patterns together are harmful. (To be entirely fair, this is also an area that is understudied and very difficult to tackle.) 

I would like to start with an almost philosophical rebuttal. Lifestyle choices also have a substantial impact on our physiology, so they too should be subject to interaction effects. If harmful interactions are indeed common it should be risky to combine good sleep, a healthy diet, endurance and strength training, with whey powder and the occasional tequila sunrise because this is already an immensely complex mixture of things. Any deviation from the mean pushes you into territory that is not well-covered by observational studies. If less than 1% of the population follows an overall super healthy lifestyle we will not have enough data on them (quantity) and the data quality will be low since they are such an elite subset, introducing additional biases. In the most extreme case everyone can be taken as a unique point in a highly multi-dimensional space (c.f. Curse of Dimensionality). Wouldn’t a healthy lifestyle be quite risky if there is a non-trivial number of dangerous interactions? 

Although the preceding argument is meant as a somewhat clever combination of intuition and reductio ad absurdum, this argument is not entirely crazy. One can certainly argue that being top quartile for some healthy behaviors and dietary patterns is good but being top decile in every possible way is not. We simply have no idea whatsoever about these complex interactions even just considering bioactives plus healthy lifestyle. One can make the reasonable case for moderation from this! (Biohacking and supplement naysayers usually make the unreasonable, lazy argument for moderation. Just heed the 80:20 rules, they will say, followed by sneering and implications of orthorexia.) 

Just to remind you what a healthy diet entails, you could consider it a combination of dozens of interventions! Low sodium, high potassium, high marine n3, low saturated fat, low cholesterol, high magnesium, chlorophyll, low heme iron, low glycemic load, low sugar, high flavonoids, terpenes and lignans, high isothiocyanate, high marine taurine, phytoestrogens, phytosterols, dietary salicylates, ergosterol, fucoidan, ergothioneine, high phylloquinone, complex sulfur compounds, ... the list of potentially beneficial bioactive compounds could go on for pages. How do you define and count the complexity of a healthy lifestyle? Half of these things you could supplement at diet like levels. Are you running a risk by doing so or just replicating the risk of a regular healthy diet? 

One can argue that drugs are a different beast because they hit very specific targets and they hit them hard (at least some approved drugs do), and their detoxification follows specific pathways, either renal or hepatic. But doesn’t exercise put strain on the kidney too just like all the plant bioactives and oxalates? Kidney failure is in fact the most likely cause of death from rhabdomyolysis. I certainly think the detoxification risk argument has some merit, although it is important to point out that it can also apply to dietary patterns.  

Back to Andrew’s argument. Personally, I am worried about liver toxicity from combining multiple hormetics / irritants / unspecific Nrf2 activators. Take, EGCG as a good example. There is quite a bit of evidence for liver toxicity at very high intakes. The same problem can happen with gastrointestinal irritants, because drugs and supplements are provided as highly concentrated pills that can lead to very high, non-physiologic concentrations of the substance. Historically, for example, potassium supplements were associated with deadly bowel perforations. Not because potassium is harmful, but because of locally elevated concentrations. This was largely resolved with delayed release formulations. 

Sometimes mechanistic understanding can help us avoid side effects. 
 
Genetic evidence 
One of the most interesting findings to come out of genetics and large GWAS studies was that most SNPs, at least those affecting quantitative traits, are indeed additive. If you have three SNPs for height you will be on average taller than if you had two. There is no such thing as a catastrophic interaction, where four of the SNPs taken together will cause the opposite. Since longevity is a quantitative trait the same will be true for lifespan. However, it is not clear if drugs behave like common SNPs. If most detectable and common SNPs show very mild phenotypes this could explain why SNPs are additive whereas drugs and supplements are not. Perhaps negative interactions are more common with rare, high-impact variants. Either way, if you take a diverse range of moderately effective supplements that all affect different longevity pathways, it should follow that the likelihoods are in your favor. 
 
“The role of non-additive genetic variation (dominance and epistatic) in complex traits in human populations remains elusive, because it is difficult to estimate and also because theory predicts it to be small relative to additive variance. ...Our results provide new evidence that genetic variance for complex traits is predominantly additive and that sample sizes of many millions of unrelated individuals are needed to estimate epistatic variance with sufficient precision.” 
https://www.cell.com/ajhg/fulltext/S0002-9297(21)00056-2?dgcid=raven_jbs_etoc_email 

“Consistent with the genetic architecture of common diseases, gene expression is predominantly additive, but a minority of transcripts display non-additive effects.” 
https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1003502 

Drug-Drug interactions and clinical evidence 

It is true that undesirable drug-drug interactions are somewhat common but do these drug-drug interactions shorten your lifespan? Do we even have a list of positive drug-drug interactions? I doubt anyone is keeping track of these. So that is certainly a bias. It is also important to point out that these are interactions among a random set of drugs. In the case of longevity supplements and drugs these have already been preselected to 1. be somewhat beneficial and 2. have low risk of interactions in the case of supplements (hence they are GRAS). 

One can make the argument that polypharmacy is harmful but that is not a good outcome to look at. The number of drugs a person takes usually correlates negatively with their health status and this is almost impossible to correct for statistically (1). In clinical practice negative drug interactions are also often accepted because there is no alternative. Taking two or three immunosuppressants might increase your risk of cancer and infection but it is still better than a swift death through organ rejection. 

Rational polypill approaches have in fact been quite successful, e.g. the recent PolyIran cluster randomized study with over 50 000 participants. A fixed combination of a statin, aspirin and two blood pressure lowering medications successfully reduced major cardiovascular events by 39%.  
https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(19)31791-X/abstract 

A meta-analysis of polypill studies, excluding the above one, with over 25 000 patients across 8 studies also found reduced MACE and all-cause mortality. While to which extent different treatments are additive remains to be seen, these results certainly confirm the safety of this approach. 
https://www.sciencedirect.com/science/article/abs/pii/S0033062022000056 

Last time I interviewed Prof. Bruce Neal for the VitaDAO podcast, he was also quite bullish on polypills for blood pressure lowering. 

Although not all studies of polypills meet the gold standard for lowering of all-cause mortality, they at least demonstrate substantial safety of a rational combination approach. 

Mouse data 
We will consider three types of evidence here: Aubrey de Grey’s robust mouse rejuvenation study (RMR), interventions testing program (ITP) single arm and ITP combination arm. My initial intuition developed from a quip by Rich Miller, which I will paraphrase as: “it’s really hard to kill mice, that was a surprising finding from the ITP”. The ITP should inform our thinking about drug and supplement safety. It is literally defined as testing unproven but promising compounds! This primes our intuition that, on average, monotherapies are unlikely to be harmful, no matter what you take. 

However, if mice are robust to monotherapies, they are probably also robust to combos as detailed in the intro (the assumption of additivity with diminishing returns). Do we have evidence to support this? From the ITP, very little but less than none. All combos in the ITP were neutral to somewhat beneficial, e.g. rapamycin + acarbose, rapamycin + metformin. The ITP also tested one extract (green tea) and a mixture without issues (protandim is a “patented blend of 5 herbal ingredients with antioxidant activity”). 

Stephen Spindler tested ramipril and simvastatin finding evidence for synergy and not just additive effects (Spindler et al. 2016). A cocktail of rapamycin, acarbose and phenylbutyrate “delays aging phenotypes” as shown by the lab of Warren Ladiges, although the authors unfortunately did not have the funds to run a full lifespan study. 
https://www.nature.com/articles/s41598-022-11229-1 

Let us look at the results of Aubrey’s RMR as posted recently and home in on the combinations: https://twitter.com/aubreydegrey/status/1807160868289036663

If we look at the RMR update in June 2024. Considering median lifespan, 5 out of 5 combination therapies in female mice are at least as good as control. 3 out of 5 combo therapies in males are at least as good as control. Considering approximated max LS (20% survivorship), this rises to 4 out of 5 in males. If you ask me, a success rate of 8 out of 10 suggests that you loaded the dice in your favour. And let’s not fool ourselves, this is a game of roulette, just like takin Aspirin or eating peanut butter. Life is full of risks, random, stochastic variation. Anaphylaxis or asphyxiation from swallowing a pill the wrong way. It is all a gamble and the rational choice is to make good bets. 

Invertebrate data and conclusions 

My intuition says that invertebrates are not a good model for drug synergies because they have very fragile gene expression networks due to their short lifespan. Be that as it may, we can still look at earlier work published by my colleagues from Singapore who studied combinations. Jan Gruber’s lab looked at 10 pairs of drugs finding that “none of the drug combinations were toxic” (Admasu et al. 2018). However, in many cases they also were not simply additive. Five drug combinations extended lifespan significantly and 5 did not, although numerically all the but one of these led to increased lifespan. 

 
In another lifespan study also using worms the authors found the following among 12 combinations (Phelps et al. 2024). Three of them led to additive benefits, seven to some benefit, one was neutral and one was harmful. 

Based on this we can conclude that combination therapies will likely not be actively harmful, but there is certainly a high risk that they are wasteful since the benefits may not be additive and there may be weak interference between different drugs and supplements. What we really need are robust one-time treatments that lead to sustained rejuvenation. Candidates include e.g. blood dilution, young blood or plasma transfer, stem cell transplants, gene therapies, repgrogramming, senolytics etc. 
 

References and further reading 

(1) Here I wanted to share a brief note on the hopeless level of confounding in observational studies. This example occurred to me based on my personal experience. So I suffer from iatrogenic dry eye disease, but dry eye disease is also a very common age-related condition. When I was staying in a hotel I could not exercise in their gym because it was too bright there due to large windows. Dry eye is a known cause of photophobia and many people are also sensitive to sweat and wind, which you would be exposed to during exercise. Now imagine the following, you observe that people who exercise less die earlier even after adjusting for cholesterol and all the other frou frou. How are you going to adjust for dry eye and all other incipient, slowly progressive age-related diseases that cause invisible disabilities and make people less willing to exercise? You will end up measuring not the benefit of exercise, but some underlying age-related process that causes aging and conditions that make it harder to exercise. 
There will be thousands of subtle reverse causality chains that you cannot adjust for in epidemiology. Thus epidemiology is about as speculative as is work in worms or mice. All of these are hypothesis generating and never definitive! 
Did people die earlier because they exercised less or because they had an invisible disability that made them die earlier and made them exercise less? We don’t know. 
 
Virk, Ghazala S., et al. "The effectiveness of polypill for the prevention of cardiovascular disease: a meta-analysis of randomized controlled trials." Cureus 15.10 (2023). 
 
Al Hennawi, Hussam, et al. "Polypill therapy in cardiovascular disease: A meta-analysis of randomized controlled trials." Current Problems in Cardiology 48.8 (2023): 101735. 

Admasu, Tesfahun Dessale, et al. "Drug synergy slows aging and improves healthspan through IGF and SREBP lipid signaling." Developmental cell 47.1 (2018): 67-79. 

Phelps, Grace B., et al. "Comprehensive evaluation of lifespan-extending molecules in C. elegans." bioRxiv (2024): 2024-06.