Hide table of contents

September 2019 update: the TAME trial has been funded. See comment by avturchin. This post is now outdated.

Edit: Upon suggestion (see comments) I set up a Guesstimate model. There, you can find wider confidence intervals and probability distributions for my final estimates. You may want to combine my estimates with your own priors on general intervention's effectiveness and thereby potentially correct for the high levels of uncertainty in my model. Otherwise do not take this analysis too seriously, since it is not very robust and it has ample margin of improvement. I encourage anyone to make their own adjustments or even make a new analysis based off on this one if interested in improving this work. Read the comments to get an idea of what further modifications could be implemented.

I encourage everyone interested in trying to evaluate interventions in the cause area of aging to read at least the paragraphs "The Idea" and "Future directions in evaluating the cause area of aging", since they contain ideas broadly applicable for evaluations in this area, remaining valid even if the calculation of the cost-effectiveness of this specific intervention could be flawed.



In this post I will try to calculate the expected cost per life saved of the Targeting Aging With Metformin (TAME) trial, in an attempt to improve Turchin’s estimate. I found some of Turchin’s assumptions were unnecessary or unjustified. He did not provide an expected value calculation and he did not apply the necessary discounts. His figure is true only if some of his many assumptions are, and this led to a result that I thought to be many orders of magnitudes off: $0.24 per life saved. In reality, if you scroll down to the Calculation section, you can see that, accounting for the icebreaking effect of the TAME trial on the FDA, I came to a result that is not distant to Turchin’s. This happened because Turchin did not account for the icebreaking effect, compensating the omission of the discounts. Otherwise, If the icebreaking effect is disregarded, the results are different, yet still show high cost-effectiveness. Again, you can find those results in the Calculation section. In the last section I outlined future directions and strategies that could be helpful to evaluate potentially much greater opportunities in the field of aging research.

Writing this analysis was challenging and some parameters remain somewhat subjective. It’s possible I made some mistakes, so I will present the calculation in a way that makes it easier to change values and redo it. If you find mistakes or make other important observations, please comment and I’ll modify the post as soon as possible. In general I tried to proceed in a similar way to how GiveWell’s CEAs [1] are made.

The TAME (Targeting Aging With Metformin) trial is motivated by two reasons:

  1. According to AFAR [2], the TAME trial could have an icebreaking effect on the FDA. It will: 
    [...] pave the way for the Food and Drug Administration (FDA) to consider aging a modifiable condition and an official “indication” for which treatments can be developed and approved. This could lead to increased funding and research, and expanded and better drug development to target aging factors, not just specific age-related diseases and conditions. TAME can open the door to ongoing research and other drug developments that will help us all live healthier, longer.
  2. Testing metformin on a healthy population could prove its beneficial effects (compression of late life morbidity and life extension) on healthy people in a definite way. After that, the official “blessing” of metformin by the FDA would cause physicians to start suggesting it to their patients. 

The idea

The core of Turchin’s idea is right: Let’s say Longevity Escape Velocity [3] will happen at date x. If we extend the lives of people who would have died before date x, making them reach date x, we are then ”saving their lives”. This could be done by raising the life expectancy of a portion of the world using simple interventions such as metformin.

The given in this rationale is that Longevity Escape Velocity will happen at some point, and I think this is very safe to assume. There is no physical law that makes Longevity Escape Velocity impossible [4] and there are examples in nature of negligibly senescent animals. Early basic research about strategies to bring aging under medical control [5] by targeting its core hallmarks is already beginning to translate into commercial application [6].

Variables needed in the calculation

A)  Expected number of years added by metformin: 1 year

Many studies suggest that metformin could postpone age-related pathologies [7]; Gwern picked some relevant ones in this detailed article. The largest study is by Bannister et al. It's particularly relevant because of the number of subjects and because it introduced a population without diabetes to be compared with a diabetic one: 

We identified 78 241 subjects treated with metformin, 12 222 treated with sulphonylurea, and 90 463 matched subjects without diabetes. This resulted in a total, censored follow-up period of 503 384 years.

It found an all cause mortality reduction in diabetics patients treated with metformin compared with a population on non diabetics:

With reference to observed survival in diabetic patients initiated with metformin monotherapy [survival time ratio (STR) = 1.0], adjusted median survival time was 15% lower (STR = 0.85, 95% CI 0.81-0.90) in matched individuals without diabetes.[...]

We can extrapolate a Risk Ratio from this quote:

Unadjusted event rates [...], surprisingly, were lower in those treated with metformin than in their matched controls (14.4 vs. 15.2 per 1000 person-years, respectively; p = 0.054). 

This data translates to around 1 year of more life for diabetics than the non diabetic control. The conversion from Risk Ratio to years added is explained by Gwern in his article.

There are some considerations to make:

  • The mean age of participants who had been given metformin was 61, and it’s possible that starting earlier would make the effect greater.

  • Since diabetics lived more than the control group it’s possible that healthy people would live even more, especially because non diabetic controls had less morbidity. In the study, table 1 shows that more diabetics had already been treated for disorders and had slightly worse parameters than the control group. This could mean that their life expectancy without metformin would have been worse than the one of the control group. On the other hand, at page 7 it is noted that: Due to the association between type 2 diabetes and increased cardiovascular risk, people with type 2 diabetes are more likely to be receiving exercise and lifestyle interventions and close monitoring and control of blood pressure and cholesterol levels. Hypertension and hypercholesterolaemia are risk factors for cardiovascular disease but are generally asymptomatic. Therefore, these conditions may be less well diagnosed in the control group.  
    In the study it is also noted that the control group could have more undiagnosed morbidities. In fact another interesting finding is this: When compared with matched, non-diabetic controls, diabetic patients with high co-morbidity who were treated with metformin had particularly improved survival (Charlson index ≥3: STR=0.67, 0.59–0.77), and this pattern increased with increasing morbidity (Figure 3). Importantly, survival was better with metformin even in those people who had not received cardiac prophylactic medications at baseline, but consistent survival benefits were observed with metformin when used in people with a prior history of each prophylactic treatment subgroup.
    This could be due to less undiagnosed morbidities leading to more treatment or (most probably) due to the fact that the STR, so the difference in death rates, starts to show when diabetics get older, that is the same time when their morbidity increases.

  • The study is a correlative result, although there are theories of why metformin may have beneficial effects in healthy people. Metformin could reduce cardiovascular events and could protect from cancer [7]; it could influence various metabolic pathways linked to aging and mimic caloric restriction

  • Turchin cites a possible life extending effect of metformin of 3 years. This presentation (by Stephen B. Kritchevsky, PhD. Given at the OAIC Annual Meeting April 19, 2016) at first glance seems to support his figure, stating a 30% effect size, that would translate to around 3 years. But this effect size regards studies looking for the reduction of the risk of various disorders such as cardiovascular diseases, dementia and cancer and not all cause mortality, so we can’t use it in this analysis. The effect sizes of metformin regarding various disorders have been shown to be higher than the effect size regarding all cause mortality. Moreover these studies were done on diabetics, not healthy people. 

The first two consideration could make the 1 year figure a bit higher and the third one lower. The last one has no effect on it. So I’ll leave it at 1 year. If you are able to quantify the considerations above in an objective manner please comment.

B) Statistical power of the TAME trial: worst case = 10%, average case = 47%, best case 84%

The objective of the TAME trial will be comparing the risk of various morbidities among metformin's users and non users. It will not check for all cause mortality, so I’ll take Gwern’s estimate of 10% - extrapolated as if all cause mortality was the endpoint of TAME - as a worst case, since checking for all cause mortality leads to a smaller effect size, that influences the power of the trial negatively. On the other hand, the projections in Kritchevsky’s presentation seem to be too optimistic. They assume at worst an effect size of 20%, and a 30% effect size as a best case. I’m highly skeptical of effect sizes greater than 20%, since the studies done till now on metformin’s effect on various morbidities were done on diabetics, who have already an increased risk of the morbidities studied. I’ll take the 20% effect size as a best case. According to the table in the presentation, if the effect size is 20%, the statistical power could vary between 58% and 84%. I'll take 84% as the best case. The average between 10% and 84% is 47%. 

C) Number of people dying of aging in a year near LEV: LEV in 2030: 65M, LEV in 2065: 88M, LEV in 2100s: 98M

I expect LEV could happen somewhere between 2030 and 2100: average 2065. The later LEV happens, the more people will be dying of aging at that time, and so more would be saved by anti-aging interventions. Taking an early date for LEV makes the estimate conservative. The number of people dying of aging in 2030 will be a fraction of the number of people born around 1950. At that time the world population was 2.52 billions and the crude birth rate - the total number of live births every 1,000 people in a year - was 36.876, so approximately 93 millions people were born. Today, aging is responsible for around 70% of all deaths. If this percentage remains the same (in reality it will increase, but I will take a conservative approach), in 2030 approximately 65 millions people will die of aging. If LEV happens around 2065, then, at that time, people born in the 80s will be dying of aging. In the 80s the world population was approximately 4.5 billions and the crude birth rate was 27.863. So the number of people dying of aging in 2065 will be approximately (4,500,000,000/1000)*27.863*0.7 = 88 millions. If LEV happens in 2100, then people born today will be dying of aging at that time. Today’s world population is 7.6 billions and the crude birth rate is 18.5, so the number people dying of aging in a year around 2100 will be (7,600,000,000/1000)*18.5*0.7= 98 millions.

D) Additional share of the world population who would take metformin consistently: worst case: 0.13%, average case: 1.3%, best case: 13%

I expect that such an intervention will be implemented only among the upper-middle income and high income shares of the world population. This amounts to 16% of the world population although this percentage is increasing, so this is a conservative estimate. It also has to be taken into account that only a percentage of this population will take metformin consistently enough to benefit from it. According to the CRN, in 2017, 76% of the US population took dietary supplements in the previous 12 months and 21% of them for healthy aging. It is also reported that 45% of those who do not take dietary supplement might consider taking them if a doctor recommended it.

About the survey:

The survey was conducted between Aug. 24–28 by Ipsos Public Affairs and was funded by CRN. The survey was conducted online in English and included a national sample of 2,001 adults aged 18 and older living in the United States, including 1,528 among those who are considered supplement users. The survey has been conducted annually since 2000. The precision of Ipsos online polls are measured using a credibility interval. In this case, the poll has a credibility interval of plus or minus 2.5 percentage points for all respondents, and plus or minus 2.9 percentage points for supplement users (see Ipsos’ online polling methodology for more info).

According to the study above, a majority of the US population uses supplements, but this is not true for other developed countries, with differences in prevalence of supplements usage up to tenfold.

Taking all this into consideration, the share of people taking supplements for healthy aging in the world should be between 0.16*0.76*0.21/10 = 0.0025 (0.25%) and 0.16*0.76*0.21 = 0.025 (2.5%). Average 1.3%.

This variable is about how many more people do not currently take metformin but will as a result of the TAME trial. In order to improve the sense of scale about how widespread metformin currently is, I want to report a few statistics: According to the American Diabetes Association, in 2015, 9.4% of the US population had diabetes. 96% of them have type 2 diabetes and 76% were diagnosed.  
In the UK, 57% of people with type 2 diabetes are treated with metformin (Bannister et al., 2014, p.1). If metformin usage is similar among diabetics in the US, that means that almost 4% of the US population currently uses metformin. This is in accordance with the high number of prescriptions of this medicine in the US: 83.9 millions in 2015. The number is so high probably because of prescription renewals and prescriptions for other conditions other than diabetes.

I want to combine all these numbers together in three possible scenarios:

Worst case: In this case metformin’s prescription for healthy people remains not very widespread, maybe because of side effects (although some of them can be mitigated by the extended release version), or maybe because metformin has to be taken for multiple years in order for it to have an effect on longevity, or maybe because it will be seen as less “natural” than supplements like multivitamins. In the worst case scenario, the number of healthy people taking metformin is well below the number of ill people taking metformin, and also well below the number of people taking supplements for healthy aging. There could be as much as a tenfold decrease from the 1.3% of people taking supplements for healthy aging in the world, maybe more. So 0.13% of the world population. In the US this would translate to around 0.76*0.21/10 = 0.01, so 1% of the US population, that is still a lot. I chose a correction of one order of magnitude because this percentage could mean that a minority of older people (currently 14.5% of the US population is older than 65, but in other countries like Japan and Italy the percentage is around 30%) without diabetes would get a metformin prescription, maybe in the face of risk factors for cardiovascular conditions. There would be also a minority of young healthy people taking metformin, probably a bigger share than today’s self-experimenters.

There could be a case even worse than this, in which the TAME trial is mostly ignored by physicians and maybe a correction of two orders of magnitudes would be required. I believe this is too unlikely, and it is counterbalanced by a similar rationale for the best case, so I will not include it. The sum of the cases (even cases not included here), each multiplied by its probability, is probably very close to the average case.

Average case: In this case the negative considerations of the worst case are compensated by the positive ones in the best case. Or simply metformin becomes one supplement used for healthy aging like any other today. In this case its usage in the world would be around 1.3%, as calculated earlier.

Best case: In this case metformin becomes really widespread thanks to all the media attention it gets. This is partly believable, because a surge of media attention for aging research and interventions is to be expected in the next decades: This will occur when the basic science and the clinical trials being done today, not only for metformin, will translate into something available for the public. At that point, medicine against aging could be one of the most widely discussed topics in public discourse. Supplement usage in general could also continue to rise in many countries. In this case metformin’s usage could be as widely spread across the world as today’s supplements for healthy aging in the US. So 2.5% of the world population would use metformin. I do not expect more because this already means that 15% (0.76 * 0.21 = 0.15) of the population of upper-middle income and high income countries would take metformin.

There could be an even better case, in which almost all the elder population and many young people take metformin, resulting maybe in more than 30% of the population of developed countries using metformin. But I think this is too improbable and so it’s not included in the calculation.

E) By what extent other therapies will reduce metformin’s marginal effect before Longevity Escape Velocity is reached: 50%

Metformin seems to have multiple effects on multiple metabolic pathways that are aging related, although it remains unclear if its main effect could be on just one of them, and if the pathway influenced the most is known at all. Metformin’s effect is upstream enough in the metabolism-to-damage chain that the maintenance-type therapies being developed now, which affect the downstream changes produced by aging, could all reduce metformin’s marginal effect. And even if metformin somehow reduced only one kind of aging damage (downstream), a more powerful therapy could render metformin less useful. By how much it depends on a variety of factors, among which include:
- How close together in time the first big impact therapies for aging will be available.
- How many big impact therapies are needed in order to reach LEV.
In my intuition, the more abruptly LEV arrives the bigger the marginal effect of metformin could be. If only a single event is sufficient for LEV to arrive, metformin’s effect is full. Otherwise, another therapy could account for a gain of many years of healthy life but not bring LEV, decreasing metformin’s marginal effect by an extent, and more of those kind of therapies could subsequently come. This variable is tough to evaluate and I’m not leaning for any of the two possibilities. My correction is 50%. I understand this estimate really leaves something to be desired, and I do not know if I could do much better using only current evidence.

F) Chance the government will not complete funding the trial: 82%

This discount is necessary if we want to calculate only the marginal effect of donations to TAME. In order to estimate this chance I contacted Nir Barzilai, one prominent proposer of the TAME trial and author of Metformin as a Tool to Target Aging. He confirmed they applied for government funding, although it is still unsure if the government will fund the trial. Considering that around 18% of the applications to NIH result in an award, the chance the government will not fund the trial is 82%.

G) Expected number of years LEV will be advanced by, considering the icebreaking effect of TAME: 3.6

The TAME trial was conceived in 2014 and till till now AFAR was able to raise half of the $75 millions needed (see H). I could project that there is a 50% chance that in the next four years the funding will be completed anyway. Splitting chances evenly among the next four years, each year has a 12.5% chance of being the one in which the trial is funded. Extending it four years further, considering that eight years from now something else could cause the icebreaking effect anyway, there’s a 12.5% chance per year for the next eight years that the icebreaking will happen spontaneously. So the expected number of years LEV will be advanced is, let’s say, no more than:  0.125*8 + 0.125*7 + 0.125*6 + 0.125*5 + 0.125*4 + 0.125*3 + 0.125*2 + 0.125*1 = 4.5.

Will advancing the date of the icebreaking effect by 4.5 years result in an advance of LEV of 4.5 years? Probably not, since the FDA not recognising aging as an indication does not stop aging research, and it‘s not clear if the icebreaking effect is a bottleneck for achieving LEV. Such an effect could, though, increase the budget of aging research and make research in this field more focused on translation and on the hallmarks of aging instead of single diseases, their symptoms, or transversal topics (to get a sense of how the budget is allocated right now you can see it directly on NIH’s website). If the icebreaking effect enables the funding of one or more projects which are bottlenecks to LEV, then LEV's date is advanced. The main probability at play here is if NIH would indeed increase its budget on aging or spend it better. If this does not happen I do not anticipate other actors would step in who wouldn’t otherwise. Since the expert opinion on this is that such a thing will happen (it is the objective of TAME stated by AFAR and Barzilai), I will put this probability at 80%. 4.5*0.8 = 3.6.

Coming up with a good value for this variable is difficult, and the figure is not very robust. It should be noted, though, that in the calculation the number of people dying per year will have the most influence and the size of this variable does not influence the bottom line much.  

H) Dollars necessary to complete the funding of the TAME trial: $37.5M

In order to evaluate this variable I contacted Nir Barzilai and Odette van der Willik, the director of the grant programs at AFAR. They said that till now AFAR was able to raise half of the $75 millions necessary to fund the trial. So the other half remains to be funded. $75M/2 = $37.5M


Accounting for everything: H/(A*B*C*D*E*F + G*C)

Worst case: 37500000/(1*0.10*65000000*0.0013*0.5*0.82 + 3.6*65000000) = 0.16$
Average case: 37500000/(1*0.47*88000000*0.013*0.5*0.82 + 3.6*88000000) = 0.12$
Best case: 37500000/(1*0.84*98000000*0.025*0.5*0.82 + 3.6*98000000) = 0.11$

Only accounting for the icebreaking effect: H/(G*C))

Worst case: 37500000/(3.6*65000000) = 0.16$
Average case: 37500000/(3.6*88000000) = 0.12$
Best case: 37500000/(3.6*98000000) = 0.11$

Without accounting for the icebreaking effect: H/(A*B*C*D*E*F)

Worst case: 37500000/(1*0.10*65000000*0.0013*0.5*0.82) = 10824$
Average case: 37500000/(1*0.47*88000000*0.013*0.5*0.82) = 170$
Best case: 37500000/(1*0.84*98000000*0.025*0.5*0.82) = 44$

Edit: Upon suggestion (see comments) I set up a Guesstimate model (here). You can find wider confidence intervals (and probability distributions) for my final estimates there. You may want to combine my estimates with your own priors on general intervention's effectiveness and thereby potentially correct for the high levels of uncertainty in my model.

This result shows that most of the cost-effectiveness of TAME comes from its icebreaking effect. The variable G is not very robust although it should be noted that the majority of influence on the cost-effectiveness come from the scope of the problem (the variable C) so with even lower values of G, the figure would be small. Overall this funding opportunity is potentially very cost effective, although there could be further corrections I missed. Five cases out of six are at least one order of magnitude better than GiveWell’s recommendations, although this evaluation is probably less robust than GiveWell’s ones. A counterbalance to its lack of robustness (although probably very small) is that this analysis does not account for DALYs averted and the economic gains of a compression of the old age morbidity of a nation (the “longevity dividend”), that could enhance the cost-effectiveness a bit more.

It should be noted that the worst case that does not account for the icebreaking effect possibly acts like a cap. I do not expect the real cost effectiveness to be worse than that, or, if worse, not much worse.

Accounting for the icebreaking effect, in the denominator, I multiplied the expected years LEV should be advanced by TAME with the number of people dying of aging per year. The result of the multiplication remains exactly the same even in the face of a slow spreading of rejuvenating therapies (and so LEV) across the world in many years. If 3.6 years seems too much it has to be considered that even approaching 0, it will generate a result always smaller than the ones not accounting for the icebreaking effect.

The order of magnitude of the results that do not account for the icebreaking effect is highly influenced by the percentage of people who will take metformin consistently (variable D), making it probably the most relevant factor in the calculation together with the number of people dying of aging in a year near LEV (variable B), although this one varies in a smaller interval. 

QALYs saved and the longevity dividend

In this post I did not account for the QALYs saved by metformin without considering LEV. I expect this figure to be probably higher than 1 for an individual taking metformin, since the RRs for many disorders is around 0.80 [7], suggesting metformin compresses the period of morbidity at the end of life. According to Barzilai this would save around 4 trillions of dollars. But this would be a completely different calculation and I do not expect the result to be competitive with other interventions in Effective Altruism if not summed with the ones in this analysis. 

Future directions in evaluating the cause area of aging

Financing the TAME trial may prove to be an effective intervention, but the effect of metformin or other existing simple interventions is negligible if compared to the potential effect of translational basic research being conducted now that directly targets the hallmarks of aging. Organisations working in this area are more difficult to evaluate, due to the uncertainty inherent to basic research. Nonetheless I think the work that some of them are doing is being terribly neglected despite their research being a bottleneck to the goal of curing age related diseases. Evaluating those opportunities could have an impact too high to ignore. My approach will probably be to classify the hallmarks of aging by their current neglectedness, tractability and scope, trying to find all the companies or non profit organisations making research in each one of them and classifying the different organisations by how they approach the problem. This will be a much more difficult task than the present evaluation. Other interesting questions I will try to answer relate to the impact of advocacy and inducing policy changes in this area, potentially having a multiplicative effect on dollars donated.


[1]: I’m still in the process of learning how to do good cost-effectiveness analyses. I worked considering the example of GiveWell’s CEAs in the sense that I tried to find all the discounts applicable, and consider counterfactuals. Having said this I hope I’m not misrepresenting what GiveWell does.

[2]: AFAR is the American Federation for Aging Research and its mission comprises “Identifying and funding a broad range of cutting-edge research most likely to increase knowledge about healthy aging”. TAME has been under reviews through several funding mechanisms and has received planning funding from AFAR (Barzilai et al., 2016, p. 1). Till now this organisation raised half the funding needed for the trial (source: email exchange with Nir Barzilai and Odette van der Willik, the director of the grant programs at AFAR).

[3] Longevity Escape Velocity, or LEV, is reached when life expectancy raises at least one year per year. This concept was first introduced by David Gobel, co-founder of the Methuselah Foundation, and spearheaded by Aubrey de Grey in his book Ending Aging. After the first treatments targeting aging will be available to the public, there will be a point in which societies become effectively non-aging through reaching LEV before the complete elimination of aging happens, maybe much before it. This because of two reasons:

  • Multiple interventions are needed in order to target all the aspects of aging, and at first they will be imperfect but they will increase life expectancy more and more as they improve.
  • Eliminating the aging damages accumulating in a human lifetime is a different challenge than eliminating the damages accumulating in a longer than human lifespan. Therefore being able to eliminate damages responsible for current human death by aging does not guarantee achieving negligible senescence, but additional therapies may be needed. Given the fast rate at which technology improves after the first proof of concepts work, it's very probable that subsequent advances will go faster than the rates of death by aging. Especially because at the point when a second wave of therapies will be needed, much more funding will be available (the idea of eliminating aging will be something really mainstream) and researchers will have much better tools than today's, while the challenge will be probably similar in difficulty and general approach (removing damaging molecules, repairing structures inside the cell, replacing or repairing tissue etc.).

[4]: As Richard Feynman puts it in his 1964 lecture at the Galileo Symposium in Italy titled “What Is and What Should Be the Role of Scientific Culture in Modern Society”:

It is one of the most remarkable things that in all of the biological sciences there is no clue as to the necessity of death. If you say we want to make perpetual motion, we have discovered enough laws as we studied physics to see that it is either absolutely impossible or else the laws are wrong. But there is nothing in biology yet found that indicates the inevitability of death. This suggests to me that it is not at all inevitable, and that it is only a matter of time before the biologists discover what it is that is causing us the trouble and that that terrible universal disease or temporariness of the human’s body will be cured.

[5]: Serious efforts about a comprehensive strategy to bring aging under medical control started in the early 2000s, thanks to Aubrey de Grey’s early proposals (2002) to the biogerontology community and his landmark book Ending Aging (2007), providing a definition of the problem that renders it attackable. At first it encountered some resistance, that, though, was not adopted as the main position:

We need to remember that all hypotheses go through a stage where one or a small number of investigators believe something and others raise doubts. The conventional wisdom is usually correct. But while most radical ideas are in fact wrong, it is a hallmark of the scientific process that it is fair about considering new propositions; every now and then, radical ideas turn out to be true. Indeed, these exceptions are often the most momentous discoveries in science. (MIT Technology Review, 2006). 

With the hindsight of today it is safe to say that de Grey’s ideas proved to be the kind that causes a paradigm shift. “The Hallmarks of Aging” (2013) is a cornerstone in the field that guides today’s research and it contains the same ideas first spearheaded by de Grey, with aging defined as a similar categorisation of damages, with a comprehensive, divide et impera, view of the problem.

Some particularly developed or interesting branches today are stem cell research, in vivo reprogramming using Yamanaka Factors, senescent cells removal, preventing epigenetic errors. All this is possible also thanks to the enlargement and upgrade of the toolbox researchers have at their disposal (very famous example being CRISPR/Cas9).

It is increasingly clear to the scientific community that targeting aging is theoretically superior to treating individual chronic diseases (Matt Kaeberlein et al., 2015) and that “treating one disease at a time, the most you can expect is to exchange one disease for another” as Nir Barzilai points out in his TED Talk.

[6]: The private sector in the area began to flourish in the last three years with prominent examples such as Oisin Biotechnologies, Ichor Therapeutics, Covalent Biosciences (All three spun from SENS research Foundation), Unity Biotechnology, AgeX, InsilicoMedicine. Many of them have tens of millions in investments. An overview of the private sector in this area can be found in Jim Mellon’s book “Juvenescence: Investing in the Age of Longevity” (2017). 

[7]: Here I organise some relevant papers cited in Gwern's analysis, in the paper by Bannister et al. and in Kritchevsky’s presentation, plus others found indipendently. Note that some of them compare metformin with other drugs that could be harmful. Dozens are left out.

All cause mortality and different diseases of aging
1. C.A. Bannister et al. 2014 Can people with type 2 diabetes live longer than those without? A comparison of mortality in people initiated with metformin or sulphonylurea monotherapy and matched, non-diabetic controls.
2. Campbell et al. 2017, Metformin reduces all-cause mortality and diseases of ageing independent of its effect on diabetes control: A systematic review and meta-analysis.
3. Stevens et al 2012, Cancer outcomes and all-cause mortality in adults allocated to metformin: systematic review and collaborative meta-analysis of randomised clinical trials
4. Saenz et al. 2005, Metformin monotherapy for type 2 diabetes mellitus
5. Morgan CLl et al. 2012, What next after metformin? A retrospective evaluation of the outcome of second-line, glucose-lowering therapies in people with type 2 diabetes
6. Pantalone KM, Kattan MW, Yu C et al. 2012, Increase in overall mortality risk in patients with type 2 diabetes receiving glipizide, glyburide or glimepiride monotherapy versus metformin: a retrospective analysis
7. Morgan CLl et al. 2014 Association between first-line monotherapy with sulphonylurea versus metformin and risk of all-cause mortality and cardiovascular events: a retrospective, observational study.

Cardiovascular disease and mortality
1. Selvin et al. 2008, Cardiovascular outcomes in trials of oral diabetes medications: a systematic review.
2. Lamanna et al. 2011, Effect of metformin on cardiovascular events and mortality: a meta-analysis of randomized clinical trials.
3. Kooy A, de Jager J, Lehert P et al. 2009, Long-term effects of metformin on metabolism and microvascular and macrovascular disease in patients with type 2 diabetes mellitus.
4. Roussel R. et al. 2010, Metformin use and mortality among patients with diabetes and atherothrombosis.
5. Aguilar D. et al. 2011 Metformin use and mortality in ambulatory patients with diabetes and heart failure.
4. Roumie CL, Hung AM, Greevy RA et al. 2012, Comparative effectiveness of sulfonylurea and metformin monotherapy on cardiovascular events in type 2 diabetes mellitus: a cohort study.
5. Morgan CLl et al. 2014 Association between first-line monotherapy with sulphonylurea versus metformin and risk of all-cause mortality and cardiovascular events: a retrospective, observational study.
6. Johnson JA et al. 2005 Reduced cardiovascular morbidity and mortality associated with metformin use in subjects with type 2 diabetes

1. Currie CJ, Poole CD, Gale EAM. 2009, The influence of glucose-lowering therapies on cancer risk in type 2 diabetes
2. Stevens et al 2012, Cancer outcomes and all-cause mortality in adults allocated to metformin: systematic review and collaborative meta-analysis of randomised clinical trials.
3. Franciosi M et al. 2013, Metformin therapy and risk of cancer in patients with type 2 diabetes: systematic review.  
4. Jacek Kasznicki, Agnieszka Sliwinska, and Józef Drzewoski, 2014 Metformin in Cancer Prevention and Therapy.
5. Hyun Hee Chung et al. 2013, The Relationship between Metformin and Cancer in Patients with Type 2 Diabetes.

Mechanisms of action
1. Nir Barzilai et al. 2016, Metformin as a Tool to Target Aging.
2. Kirpichnikov D, McFarlane SI, Sowers JR. 2002, Metformin: an update

Sorted by Click to highlight new comments since:

TAME study got needed funding from a private donor:

"After closing the final $40m of its required $75m budget with a donation from a private source, the first drug trial directly targeting aging is set to begin at the end of this year, lead researcher Dr Nir Barzilai has revealed."


Thanks, great info. This post is officially outdated :)

I think that there are other cost-effective interventions in life extension, including research in geroprotectors combinations and brain plastination.

Hi Emanuele,

I saw your request for commentary on Facebook, so here are some off-the-cuff comments (about 1 hour's worth so take with appropriate grains of salt, but summarizing prior thinking):

  • My prior take on metformin was that it seems promising for its space (albeit with mixed evidence, and prior longevity drug development efforts haven't panned out, but the returns would be very high for medical research if true), although overall the space looks less promising than x-risk reduction to me; the following comments will be about details of the analysis where I would currently differ
  • The suggestion of this trial moving forward LEV by 3+ years through an icebreaker effect boosting research looks wildly implausible to me
    • LEV is not mainly bottlenecked on 'research on aging,' e.g. de Grey's proposals require radical advances in generally medically applicable stem cell and genetic engineering technologies that already receive massive funding and are quite challenging; the ability to replace diseased cells with genetically engineered stem cell derived tissues is already a major priority, and curing cancer is a small subset of SENS
    • Much of the expected gain in biomedical technology is not driven by shifts within biology, and advances within a particular medical field are heavily driven by broader improvements (e.g. computers, CRISPR, genome sequencing, PCR, etc); if LEV is far off and heavily dependent on other areas, then developments in other fields will make it comparatively easy for aging research to benefit from 'catch up growth' reducing the expected value of immediate speedup (almost all of which would have washed away if LEV happens in the latter half of the century)
    • In particular, if automating R&D with AI is easier than LEV, and would moot prior biomedical research, then that adds an additional discount factor; I would bet that this happens before LEV through biomedical research
    • Getting approval to treat 'aging' isn't actually particularly helpful relative to approval for 'diseases of aging' since all-cause mortality requires larger trials and we don't have great aging biomarkers; and the NIH has taken steps in that direction regardless
    • Similar stories have been told about other developments and experiments, which haven't had massive icebreaker effects
    • Combined, these effects look like they cost a couple orders of magnitude
  • From my current epistemic state the expected # of years added by metformin looks too high
  • Re the Guesstimate model the statistical power of the trial is tightly tied to effect size; the larger the effect size the fewer people you need to show results; that raises the returns of small trials, but means you have diminishing returns for larger ones (you are spending more money to detect smaller effects so marginal cost-effectiveness goes a lot lower than average cost-effectiveness, reflecting high VOI of testing the more extravagant possibility)
  • Likewise the proportion using metformin conditional on a positive result is also correlated with effect size (which raises average EV, but shifts marginal EV lower proportionate to average EV); also the proportion of users seems too low to me conditional on success

Hey! Thanks for the comment! I really appreciate it. For some reason I'm only seeing it now and by chance. I don't know if I didn't get the notification or if I missed it.

I'm not sure if this is the post I was asking feedback for though. This analysis is from nine months ago, and my views on it changed. On Facebook I was probably referring to this other post I made recently: A general framework for evaluating aging research. Part 1: reasoning with Longevity Escape Velocity. [EDIT: I just saw you made a comment under that post too, so never mind].

Regarding the content of your comment: I agree with most of it. In fact 3.6 years is probably a big overestimation. However, I still think, in general, that bringing LEV forward could be a big contributor to the cost-effectiveness of aging research in general. In my newer post I lay your same arguments about improving technology that may subsume the effect of today's research, making it less cost-effective. This factor also influences variable E in the TAME analysis, which I also probably vastly overestimated. For the very specific case of AI potentially automating R&D I think the timeline is longer than for LEV achieved through biomedical research (I'm taking the view that arises from the probability distribution given by AI researchers), but, as you said, it's not the only technology that would make some of the efforts made now less useful.

Maybe I'm still less "pessimistic" than you in the sense that I think that an ice-breaking effect could enable more research on what are neglected facets of aging for which treatments could be devised much more quickly. The foundational research is not very neglected, while there are wide areas of translational research that could use much more funding and that are necessary to reach the final goal. From the lifespan.io's Rejuvenation Roadmap you should get a preliminary idea.

Your example using the SENS approach is correct: areas like stem cell research and cancer research don't seem to be underfunded. But they are only two pieces of the puzzle. Some others are being much more neglected. That's why SENS itself gives higher priority to the most neglected areas, like mitochondrial dysfunction and crosslinks, which should be also more tractable (an interesting fact is that Aubrey de Grey often emphasises neglectedness, tractability and scope in his conferences, but I haven't heard anyone within EA pointing this out). If stem cells research, cancer and other difficult and highly funded areas were all there is to aging research, it wouldn't look like a very good candidate EA cause. In fact, not only de Grey but many researchers in the area are pursuing projects they believe are very much funding constrained (example: Steve Horvath).

About the comparison with x-risk reduction: Yes, I broadly agree that x-risk reduction looks overall more promising as a cause area. However I think that many x-risk focused interventions have a higher level of uncertainty. It also seems that within Effective Altruism little to no effort has been made to evaluate aging research, while, to me, it looks highly competitive with many of the other focuses of EAs (some specific interventions inside aging research should be very recognisably better). So it should be analysed further, especially because we may be missing out on especially important opportunities.

Regarding the expected number of years added by metformin: I think "one year" is a very conservative number given the evidence I've presented, and you'll often hear researchers estimating more.

Interesting Analysis! Since you already have confidence intervals for a lot of your models factors, using the guesstimate web tool to get a more detailed idea of the uncertainty in the final estimate might be helpful, since some bayesian discounting based on estimate's uncertainty might be a sensible thing to do. (https://www.lesswrong.com/posts/5gQLrJr2yhPzMCcni/the-optimizer-s-curse-and-how-to-beat-it)

It might also make sense to make your ethical assumptions more explicit in the beginning (https://www.givewell.org/how-we-work/our-criteria/cost-effectiveness/comparing-moral-weights), especially since the case against aging seems to be less intuitive than most of givewells interventions.

Here how I would reason about moral weights in this case:

In this case the definition of a "life saved" is pretty different than what normally means. Normally a life saved means 30 to 80 DALYs averted, depending if the intervention is on adults or children. In this case we are talking about potentially thousands of DALYs averted, so a life saved should count more. On the other hand there's also to take into consideration that when saving, for example, children who would have died of malaria, you are also giving them a chance of reaching LEV. It's not a full chance as in the present evaluation, but something probably ranging from 30% to 70%.

Additional consideration: some people may want to consider children more important to save than adults. Introducing age weighting and time discounting could seem reasonable in this case, since even if you save 5000 DALYs you are only saving one person, so you might want to discount DALYs saved later in life. On the other hand there are reasons to disagree with this approach: Saving an old person and guaranteeing him/her to reach LEV means also "saving a library". A vast amount of knowledge and experience, especially future experience would have been otherwise completely destroyed. In fact I am not so sure I would apply time discounting myself for this reason.

Regarding bayesian discounting:

I just read how GiveWell would go about this (https://blog.givewell.org/2011/08/18/why-we-cant-take-expected-value-estimates-literally-even-when-theyre-unbiased/). To account for it I would need a prior distribution (or more than one?). I also have difficulty making the calculation, since Guesstimate doesn't let me calculate the variance of the random variables. I will try with other means... maybe with smaller data sets and proceeding by hand or using online calculators.

I would also like to introduce probability distributions in the whole analysis and turn some arguments made in the explanations of some variables in variables in their own right, and I would like to add some more informations (for example the safety profile and history of metformin and the value of information of the trial) based on feedback I'm receiving. This would mean rewriting many sections though, and this will require time.

For now I put an "Edit" at the beginning in order to warn readers not to take the numbers reached too seriously, but I invited them to delve in some more broadly applicable ideas I presented in the analysis that could be useful for evaluating many interventions in the cause area of aging.

I think, it might be best to just report confidence intervals for your final estimates (guesstimate should give you those). Then everyone can combine your estimates with their own priors on general intervention's effectiveness and thereby potentially correct for the high levels of uncertainty (at least in a crude way by estimating the variance from the confidence intervals).

The variance of X can be defined as E[X^2]-E[X]^2, which should not be hard to implement in Guesstimate. However, i am not sure, whether or not having the variance yields to more accurate updating, than having a confidence interval. Optimally you'd have the full distribution, but i am not sure, whether anyone will actually do the maths to update from there. (But they could get it roughly from your guesstimate model).

I might comment more on some details and the moral assumptions, if i find the time for it soon.

Thank you, I applied your suggestion by modifying the text. I just noticed that Guesstimate gives you the standard deviation. I guess I had to familiarise with the tool.

Thank you for the feedback!

I'm still learning and comments really help me to be more accurate and they steepen my learning curve. I set up a Guesstimate model (https://www.getguesstimate.com/models/10848). I didn't know about this tool, it is really helpful!

Tomorrow I will improve the guesstimate and get back to you with another comment regarding the bayesian discounting you proposed and the moral weights. I also might make other changes to the evaluation together with the ones you suggested, especially considering that Guesstimate lets me toy with probability distributions.

Thanks for this detailed analysis. I think that the main difference in our estimations is the number of adopters, which is 1.3 percent in your average case. In my estimation, it was almost a half of the world population.

This difference highlights the important problem: how to make really good life-extending intervention widely adopted. This question is related not only to metformin, but for any other interventions, including now known interventions such as sport, healthy diet and quitting smoking, which all depends on a person's will.

Taking a pill will require fewer efforts than quitting smoking, and around 70 percent of US adult population is taking some form of supplements. https://www.nutraceuticalsworld.com/contents/view_online-exclusives/2016-10-31/over-170-million-americans-take-dietary-supplements/

However, supplements market depends on expensive advertising, not on real benefits of the supplements.

Metformin isn't a supplement though. It's unlikely it would ever get approved as a supplement or OTC, especially given that it has serious side effects.

That is why I think that we should divide discussion in two lines: One is the potential impact of simple interventions in life extension, which are many, and another is, is it possible that metformin will be such simple intervention.

In case of metformin, there is a tendency to prescribe it to the larger share of the population, as a first line drug of diabetes 2, but I think that its safety should be personalized by some genetic tests and bloodwork for vitamin deficiency.

Around 30 mln people in US or 10 per cent of the population already have diabetes 2 (https://www.healthline.com/health/type-2-diabetes/statistics) and this population share is eligible for metformin prescriptions.

This means that we could get large life expecting benefits replacing prescription drugs not associated with longevity - with longevity associated drugs for the same condition, like metformin for diabetes, lazortan for hypertension, aspirin for blood thining etc.

Curated and popular this week
Relevant opportunities