How long should a mouse live?
Pabis, Kamil, et al. "The impact of short-lived controls on the interpretation of lifespan experiments and progress in geroscience–through the lens of the “900-day rule”." Ageing Research Reviews (2024): 102512.
https://www.biorxiv.org/content/10.1101/2023.10.08.561459v1
https://www.sciencedirect.com/science/article/abs/pii/S1568163724003301
ABSTRCT
Although lifespan extension remains the gold standard for assessing interventions hypothesized to impact the biology of aging, there are important limitations to this approach. Our reanalysis of lifespan studies from multiple sources suggests that the use of short-lived control cohorts tends to exaggerate the relative efficacy of putative longevity interventions. Moreover, due to the high cost and long timeframes of mouse studies, it is rare that a particular longevity intervention will be independently replicated by multiple groups.
To facilitate identification of successful interventions, we propose an alternative approach. The level of confidence we can have in an intervention is proportional to the degree of lifespan extension above the strain- and species-specific upper limit of lifespan, which we can estimate from comparison to historical controls. In the absence of independent replication, a putative mouse longevity intervention should only be considered with high confidence when control lifespans are close to 900 days or if the final lifespan of the treated group is considerably above 900 days. Using this “900-day rule” we identified several candidate interventions from the literature that merit follow-up studies.
Introduction
When evaluating the robustness of a mouse study some of the top criteria are:
lifespan of the controls (higher is better)
bodyweight changes, or lack thereof if we want to identify something that is not acting via caloric restriction
sample size (N>40 is a good starting point)
robustness across genetically diverse strains (e.g. use genetically heterogeneous strains)
robustness across sexes
The key idea advanced in our paper is that many long-lived mouse strains live around 900-days, especially the popular HET3 and B6 strains. Thus 900 days can be used as a historical benchmark when comparing lifespan curves for mouse studies. Additionally, we should strive to extend the lifespan of our controls to around 900 days so that drugs are tested in the longest-lived possible conditions, since it is very likely that life-shortening effects are idiosyncratic and not related to aging.
Put in human terms, malaria can kill you. Alleviating malaria will extend a population’s lifespan. This does not mean that malaria causes aging or that anti-malarials are longevity drugs. Similarly, we believe that many short-lived mice die due to stress, fighting, infection, etc. rather than through aging. Alleviating this premature death has nothing to do with slowing aging.
Follow-up question
The question we want to investigate in this follow-up post to our preprint is whether there is an optimal cutoff for mouse control lifespans that defines a robust study. Robust here means simply a result that can replicated by other investigators. For this analysis, we consider the ITP study as the replication study and all other studies as hypothesis-generating.
This question is quite interesting and was raised on my twitter timeline.
We believe that studies reporting positive results against a long-lived control population (lifespan > cutoff) are more likely to be robust. The question is whether we can test this systematically by varying the cutoff which could allow us to define an unbiased cutoff. In our previous work we proposed that controls should live 900 days (± 50 days). This is based on a biological argument, and several assumptions, rather than a statistical one. Below we will try to make a more quantitative case based on the ratio of false positives to false negatives.
One major caveat of this analysis is that the data quality and the number of published mouse studies is low making it difficult to systematically evaluate different cutoffs. Especially because defining a successful mouse study is somewhat subjective. Therefore we choose a more narrative approach.
Results
If we use a cutoff for control mice at 900 days (± 50 days), we find two true positives (rapamycin, NDGA) and one maybe positive (Aspirin) in the DrugAge dataset. We consider NDGA a true positive because it works rather consistently in male but not female HET3 mice across ITP cohorts (tested four times). Aspirin only worked once and only in males (tested three times) so it is difficult to classify.
If we use a cutoff for control mice at 800 days (± 50 days), we find the same true positives as above. However, in addition we find four false positives (curcumin, green tea, nicotinamide riboside [NR] derivatives, and metformin). Classifying metformin and NR derivatives is difficult because several studies showed no benefit or even life shortening in the case of metformin. If we exclude metformin and NR from our list of false positives we still have two additional false positives in total (curcumin, green tea).
Fisetin and resveratrol deserve their own discussion since they were not in the DrugAge and/or the ITP dataset that we used for our manuscript. Nevertheless, these two examples also show that relaxing the 900d rule to 800d would produce many more false positives that will not replicate in rigorous studies.
In a small study fisetin was reported to extend the median lifespan of male and female C57BL/6 mice from an estimated 745 days to 873 days (Yousefzadeh et al. 2018). Recently after submission of our manuscript the ITP reported that fisetin fails to extend the lifespan of HET3 mice. This would be consistent with the notion that the initial lifespan result was not robust, although could be also attributed to a strain difference since HET3 mice show more limited senescence-related pathology than do B6 mice (Harrison et al. 2023).
Baur et al. (2006) found that resveratrol extends the lifespan of obese mice whose median lifespan was just shy of 770 days. This finding was not replicated in the ITP nor by Pearson et al. (2008) when comparing treated animals with controls that had a median lifespan of around 840 days. Again, this would be consistent with the notion that the initial lifespan result was not robust due to short-lived controls.
Summary
Overall applying relaxed lifespan criteria we find 3-6 false positives (curcumin, green tea, fisetin, NR derivatives, and possibly resveratrol, metformin), 2 true positives (rapamycin, NDGA) and one maybe positive (aspirin). All false positives are eliminated using the 900-day rule with no additional false negatives. If we consider aspirin a “hit” in the ITP, then the 900-day rule would produce 1 false negative.
Notes, limitations and additions
Although the ITP was the first study to show that rapamycin extends mouse lifespan we can nevertheless compare the published rapamycin studies with the ITP result using our methodology “as if we used the non-ITP data to predict the ITP result”. Rapamycin (and arguably NDGA) are the only drugs that increase lifespan in at least one sex in the ITP and in long-lived B6 or hybrid mice outside of the ITP.
For convenience and to exclude potential biases we heavily rely on DrugAge to define “all other studies”. When appropriate we try to reference other relevant studies that may not have been reported in DrugAge, although the discussion of these studies here is not exhaustive.
Somewhat in line with the 900-day rule, although this one is hard to call, is the story of ACE inhibitors. Captopril extends male lifespan in the ITP. Enalapril has non-significant benefits. Similarly, ramipril appears to have small, non-significant benefits in long-lived F1 hybrid mice (Spindler 2016).
In summary, if it works in long-lived mice it is likely robust!
References
Yousefzadeh, Matthew J., et al. "Fisetin is a senotherapeutic that extends health and lifespan." EBioMedicine 36 (2018): 18-28.
Harrison, David E., et al. "Astaxanthin and meclizine extend lifespan in UM-HET3 male mice; fisetin, SG1002 (hydrogen sulfide donor), dimethyl fumarate, mycophenolic acid, and 4-phenylbutyrate do not significantly affect lifespan in either sex at the doses and schedules used." GeroScience (2023): 1-22.
Baur, J. A., Pearson, K. J., Price, N. L., Jamieson, H. A., Lerin, C., Kalra, A., Prabhu, V. V., Allard, J. S., Lopez-Lluch, G., Lewis, K., Pistell, P. J., Poosala, S., Becker, K. G., Boss, O., Gwinn, D., Wang, M., Ramaswamy, S., Fishbein, K. W., Spencer, R. G., … Sinclair, D. A. (2006). Resveratrol improves health and survival of mice on a high-calorie diet. Nature, 444(7117), Article 7117. https://doi.org/10.1038/nature05354
Spindler, Stephen R., Patricia L. Mote, and James M. Flegal. "Combined statin and angiotensin-converting enzyme (ACE) inhibitor treatment increases the lifespan of long-lived F1 male mice." Age 38 (2016): 379-391.