For more than a century, calorie restriction (CR), reducing energy intake without malnutrition, has been one of the most reproducible interventions for extending lifespan in laboratory animals. Yet two influential studies unsettled the field by reporting that, in some inbred mouse strains, CR could shorten rather than extend life. These findings raised the possibility that genetic background might fundamentally reshape the response to CR, complicating both mechanistic interpretation and any eventual translation to humans.
A new Perspective published in Life Metabolism revisits that conclusion and argues that the evidence for large strain-specific life-shortening effects is weaker than it appeared. In this Perspective, Prof. John R. Speakman and Dr. Sharon E. Mitchell suggest that much of the reported heterogeneity may reflect low statistical power and experimental design, rather than robust genotype-specific biology.
Their argument begins with a simple but consequential point: sample size. In the widely cited study by Liao and colleagues, the maximum number of animals used to estimate lifespan response was only five ad libitum-fed and five calorie-restricted mice per strain, and some restricted groups became even smaller after censoring early deaths. For two strains, the effective dataset consisted of the lifespan of only a single mouse. Under those conditions, distinguishing a true strain effect from sampling noise becomes extremely difficult.
To test that idea, the authors pooled the original lifespan data for males and females under ad libitum feeding and CR, and then repeatedly resampled these distributions to generate 41 simulated strains with the same small group sizes as the original experiment (Figure 1). Strikingly, the simulated datasets reproduced the same broad pattern that drove the controversy: some apparent lifespan extension, some no effect, and some lifespan shortening. They also reproduced the weak correlation between male and female responses within strains, suggesting that many dramatic strain differences can arise even when animals are sampled at random from a common population.
The Perspective does not argue that genotype is irrelevant. The real data showed slightly greater variance than the simulations, leaving open the possibility of a modest genuine strain effect. But the central point is that any such effect was likely much smaller than originally implied. Later evidence is consistent with that interpretation: in a 2021 follow-up study using at least 30 mice per group, most of the previously reported life-shortening effects did not replicate, and only one of eight strain-sex comparisons still showed reduced lifespan, with a much weaker magnitude than before.
Beyond revisiting a long-running controversy, the article makes a broader methodological case for the field. Lifespan studies under CR are highly sensitive to sample size, housing conditions, and feeding protocol. Group housing under CR, for example, may introduce stress or aggression-related confounding, while inconsistent designs make true genotype-by-diet interactions harder to resolve. The authors therefore call for more standardized studies, with adequate statistical power and, where feasible, individual housing that allows food intake to be quantified accurately.
Taken together, the message is not that all variation in response to calorie restriction should be dismissed. Rather, claims of dramatic life-shortening effects in specific genetic backgrounds require far stronger evidence than small, noisy datasets can provide. This Perspective argues for a more disciplined interpretation of the CR literature, one that separates plausible biological heterogeneity from artefacts introduced by underpowered experimental design.
DOI
10.1093/lifemeta/loag006