Review of Sex differences in the association of photoperiod with hippocampal subfield volumes in older adults: A crosssectional study in the UK Biobank cohort:

By Majrashi et al.,

The authors have submitted a very well written manuscript for consideration, which was a genuine pleasure to read. The logic of the introduction established through proxy species leading to the hypothesis for humans was very nice. I have no issues with the introduction or discussion.

There are a few minor issues with methods - mostly tied to the multiple correction problem, but this is not insurmountable, and I think the authors can correct this quite quickly. The way I see it, you are treating this as a univariate problem (I.e., volume in every hippocampal subfield is independent), which is almost certainly not the case. Individuals with smaller areas CA1 will likely also have reduced volume in CA2/3 and so on. This makes your problem inherently multivariate. I suggest using an approach like PLS regression (with sex and photoperiod as predictors). You can still report the correlations, but unless you correct for the 2(sex) by 6 (ROI) x 2(hemisphere) = p corrected 0.002 comparisons, this isn’t quite right. Note that you should probably either include your covariates of no interest in the model and examine their effects separately, or regress them out a priori.

With a sample this large, it is also hard to evaluate significance without effect sizes - please add partial eta squared values for the ANCOVA interaction and main effects.

Finally, when reporting correlations, please be consistent about reporting either r or r^2.

I sincerely doubt your results will change, the story is very clear, and adding the multivariate modelling to account for the multiple tests should help to propel this story into print where it firmly belongs.

Best wishes,

John A. E. Anderson

The authors have done a commendable job of revising this manuscript and taking into account the comments of the reviewers. Please include output from the PLS analysis requested in the first review (e.g., weights for each hippocampal subregion from the first latent variable). Merely describing the outcome of the analysis is not a full presentation of the results. Ideally for a PLS analysis, one would want to see 1) the proportion of variance the first LV accounted for, 2) whether the LV was significant (via permutation tests) 3) how the variables loaded onto the LV - this could be shown graphically as a bar-chart.