This paper asks a very interesting question – is speed of development during childhood associated with speed of aging in middle age, and in which direction? Some excellent longitudinal data was brought to bear on the question, involving relevant measures across multiple decades. The findings, though weak and somewhat statistically tenuous, of a positive association, are interesting.

I had a number of concerns, though, with the way the paper is conceptualised/interpreted, and how the data are analysed.

• In the introduction, repeatedly saying “X goes hand in hand with Y” instead of just “X is associated with Y”, aside from being annoying, also implies a stronger relation than there is between the variables. For example, it is claimed that “In childhood, poor health goes hand in hand with slow development (Miller et al., 2016).” But the citation is only about stunting as a consequence of severe malnutrition, not a general association between health and developmental speed. There is clear evidence in the other direction too – obesity (a primary example of poor health) is associated with faster, not slower, development in adolescence.

• The authors frame the hypothesis that “fast-developing children will go on to be fast-aging adults” as a prediction of life history theory in evolutionary psychology (LHT-P), and take their support of this hypothesis as support of that theory. This seems highly questionable. A positive association between development speed and aging speed is expected under the most basic idea that there are individual differences in speed through the life course. It could be considered a default hypothesis. LHT-P proposes various things that are not assessed in this study, and its theoretical basis has been harshly critiqued by many (e.g. see special issue in Evolution and Human Behaviour, 2020). Given that this study does not test any key predictions from the theory (even the present authors couch its predictions as maybes: LHT-P “posits that faster development might maximize reproductive success in stressful environments,
  but the tradeoff might be poorer health and faster aging”) and that the theory is regarded by many as fundamentally flawed, it does not seem a good framing for the study.

• Similarly, the authors characterise the opposite hypothesis, that children who have poor health and slow development will continue to have poor health and fast aging as supporting a theory that biological health is programmed antenatally and shapes both child development and adult aging, but again, it’s not at all clear what links the theory with this specific hypothesis. (The cited paper by Barker was unhelpful in this regard.) If the authors want to link their hypotheses to theories, they should do so much more thoroughly and rigorously. But there is no need to try to manufacture theoretical bases for the question. It’s a very interesting question in and of itself.

• I don’t think the Latent Profile Analysis is a sensible approach. It is implausible that the variation in developmental speed is categorical with three categories of individuals. (What could explain such a thing?) Like nearly every other variable, it will be continuously distributed, so why not just analyse it like that (e.g. with a composite variable)? As the authors say, “LPA is a type of mixture model where the observed distribution on several variables is assumed to be comprised of a mixture of two or more latent distributions”. What basis is there for this assumption in this case? It seems to me there is no basis. If there is none, the analysis is inappropriate. It would be good to see the correlations between all measures too, to get an idea of how coherent (or not) these speed variables are.

• Especially strange is that the profiles were then converted back into three continuous variables based on each participant’s probability of being classified into each of the three profiles. Then there is a dimensional approach referred to later, in the supplementary material, which seems like the more logical of all the approaches (and incidentally has the more convincing effects).

• The association between rate of development and the pace of aging measure, as analysed, is tenuous. For the profiles, the 95% confidence intervals around the pace of aging means for each rate of development profile all include 1.00 (i.e. the mean of the whole sample). This is not good evidence for an effect. For the continuous measures in the main text, the associations were tiny and either marginally significant or not at all. Of the individual rate of development measures, adiposity rebound was the only decent predictor, suggesting that the effects likely relate to food intake rather than life history.

• The outcome variable of pace of midlife biological aging seems like a measure of biomarkers of health. How can biomarkers of aging be distinguished from biomarkers of health? Why was the epigenetic clock not used as a measure of aging in study 1? What about facial aging? Use of these kinds of variables would better convince that you are measuring pace of aging specifically.

• In Study 2, age at menarche is taken as the rate of development measure. This illustrates why LHT-P is a poor model of variation in this measure. Age at menarche has decreased greatly in modern times – that clearly doesn’t fit with age of development being sped up by growing up in harsher environments (environments have become less harsh over time), but it does fit with the food intake account.

• “Because obesity is related to both earlier menarche and faster aging, we additionally
  identified girls who were obese in adolescence and controlled for this in our models.” Why not just control for BMI, rather than making an arbitrary cutoff and controlling for the dichotomous variable? I would definitely want to see the results of the former analysis.

• Study 2 was billed as a conceptual replication of these results, but its finding – a link between age at menarche and aging per an epigenetic clock – had already been found thrice before. So fine, but not novel, and if it’s a conceptual replication of Study 1, then it suggests that Study 1 is not novel either (conceptually). And this wasn’t mentioned in the introduction.