Discussion
The question was when ageing itself should count as a target for a trial or a regulator, who decides what counts as evidence that it has been slowed, by when, and what would make them revisit it. The texts answer its parts unevenly, and the unevenness is the finding.
On the target they come close to agreement, from different starting points. Gems reaches it by refusing the line between ageing and disease; Mattison and others by the monkey studies, which they read as showing that the age-related increase in disease vulnerability in primates is malleable; Scott and others by the value a slowing of ageing would have; the PROSPR page by the gap between lifespan and healthspan; Barzilai by asking that the two be joined in health-adjusted survival. None of these is a decision that ageing may be the indication on which a treatment is approved. They are arguments that it should be aimed at, made by people who study it.
On evidence the texts divide, and the division runs along time. The outcomes people care about, illness, frailty and death, take decades. The measures that can be read within a trial take months, and every text that proposes one also says what it has not yet shown. Horvath writes that his clock may become a useful surrogate marker; Belsky and others that their measure is a first step toward a single-assay measurement of the pace of ageing; Fahy and others that their result must be replicated; Waziry and others that there is no gold standard and that their trial's follow-up ended with the intervention. The demand that Barzilai reports, for hard outcomes in mammals rather than a parade of biomarkers, and his concession that human trials cannot wait for death, together state the dilemma the measurement papers are trying to solve.
On who decides, the texts are thinnest. The research texts leave the decision about evidence to the process of validation, carried out by the people who build and use the measures, under review boards and, in one case, a regulator's permission to run the study. The one government text names the regulator as the body to which expert recommendations on an endpoint are addressed, and it proposes an endpoint of a different kind, intrinsic capacity rather than methylation. It also sets the only period in the texts: three years instead of decades, as a programme's aim.
On what would reopen a decision, the texts are richest. The measurement papers state, before anyone has relied on their measures, the conditions under which those measures could be doubted: that change in the measure track change in healthy life, that results hold in other populations, that participants be followed beyond the trial. The pilot trial names replication. The monkey studies show that two answers to one question can be sent back to the designs that produced them. The economic model names a pending trial and its own calibration. Waziry and others add an assumption for planners, of modest effects, whose failure would itself be a reason to look again.
Contributions
The paper makes three contributions. First, it separates the question of whether ageing can be slowed, which it does not answer, from the question of who decides what would show it, which it reads in the texts. Second, it shows that in these texts the target is better argued than the evidence, and the evidence better argued than its governance: the arguments for aiming at ageing are confident, the measures are hedged, and the holder of the decision is mostly left unnamed. Third, it shows that the conditions for revisiting are written with unusual care in the measurement literature, as the steps by which a marker becomes a surrogate endpoint, and that the government text adds a different endpoint and a period. Setting these side by side is the paper's contribution; which way any reader should weigh them is left to the reader.
Limits
The paper reads nine texts that could be quoted, not the literature. Important documents could not be read from this session, among them two pages of the National Institute on Aging, and six works are known here only from their abstracts. The government text is a programme page, not a guidance or a decision of the regulator, and the paper read nothing the regulator itself has said about ageing as an indication; it cannot say what the regulator has decided. The findings of the trials and animal studies are given as their authors state them and are not assessed, combined or extended to people. The texts are mostly from the United States, and the paper says nothing about how other regulators frame the question. Finally, the reading is this study's interpretation; another reader could find in the same texts a different allocation of the decision.
Further research
Three lines of work follow. The first is to read the regulator's own texts on endpoints for trials that target ageing, and the consensus document on intrinsic capacity that the PROSPR page lists, to see whether the allocation the texts imply is the one the regulator accepts. The second is to follow the conditions the measurement papers wrote for themselves: whether follow-up of trial participants, results in more diverse populations, or evidence about the correspondence between a measure's change and a change in health have since been reported, and by whom. The third is comparative: whether the question of who decides what counts as evidence of slowed ageing is answered differently where the regulator, the research funder and the clinician are differently placed. Each would test the finding of this paper, that the decision about evidence for slowing ageing is, in the texts read here, still largely held by the people who build the measures.