Abstract
“new version of Freesurfer v 6.0” – Not that new, has been out a couple of years, and will definitely not stay new
"in all participants" – a bit unclear, I assume you mean “across” participants

Methods
Were there no other exclusion criteria than quality of the T1?
Were patients with psychiatric or somatic disorders included?
Depression and medication are known to affect hippocampal volumes, were they accounted for?
It looks like the standard Freesurfer pipeline was followed which is published in detail elsewhere, this methods text can be make more compact.
How was the poor segmentation of the subfield assessed, only by visual inspection? Why were not outlying volumes calculated and excluded?
The hippocampus volumes were only adjusted by the TBV, but they should have also been adjusted by age and sex, se for instance the recommendation by Barnes et al doi:10.1016/j.neuroimage.2010.06.025
Please clarify the number of comparisons the p values were corrected for. I count to 10, left and right hemispheres for 5 subfields. Is this correct?
Was the Molecular Layer excluded?
I thought the hippocampal tail was excluded but appears in Table 1.

Results
The Results section is difficult to follow since it includes a lot of the methodological approach, it is unclear is these models have already been described in the methods in which case it should not be repeated in the Results. Otherwise move model details to the Methods.
Figure 1 is redundant, and has been published elsewhere. The 5 subfields investigated here can be visualized in Figure 3.

The authors addressed some of my concerns but there are a few remaining issues.
I don’t agree with the argument that inclusion of all medical condition makes the results more generable, especially since some of the psychiatric condition e.g. subtypes of mood disorder are related with photoperiod. For instance, since the sample is very large and likely dominated by healthy subjects the observed effects are driven by this group. However, no inference can be made regarding the applicability of the findings on specific subgroups e.g. seasonal affective disorder, or other psychiatric conditions.

I also think it is problematic that no attempt was done to control that the findings still stand when outliers are removed. The argument that as long as the FS pipeline did not crash every segmentation is included is not valid. Manual visual inspection of individual hippocampal subfields is very difficult and time consuming. The subfield segmentation is also far from perfect, for instance in our lab we are often seeing 10% volume differences between 2 T1 scans of the same subject. Just because the pipeline did not crash there is no assurance that it worked perfectly, other papers have also applied outline exclusion e.g. Cao et al 2017 in Mol Psychiatry that excluded five standard deviations outliers.

In general, issues that can be addressed in the current data should be addressed and not heaped up on the limitation section which would make interpretation of the results ambiguous.