In this very well written manuscript J Tong et al make the important link between retinal imaging and histology. The paper is timely with ever more data being published on retinal optical coherence tomography (OCT), based on the assumption of a firm correlation with histology. Such correlative imaging/histological data are however lacking. Hence a gap in the literature present study aims to fill.

The main concern this referee has is with regard to extrapolation from cross-sectional data on a relationship between ageing and degeneration of specific cell populations in the retina. There are two levels of extrapolation. First, from cross sectional OCT data to longitudinal OCT data by means of modelling. Second, from localised histological data to whole retina histological data. The latter is correctly discussed by the authors as a limitation of the study: "INL cell densities were also only available along the temporal meridian, resulting in assumptions of similar densities across meridians; this may not be accurate, given meridian-specific differences in GC and photoreceptor density as well as INL thickness." But more glaringly, the histological data stem from different patients than the OCT data. I have great reservations of correlating data from one study cohort with another.

This major concern aside, there is also lack of discussing the existing literature of longitudinal OCT changes in neurodegenerative disease and controls. This is relevant because there are ongoing discussions, for example for the INL, how to interprete data on a group level which is, as correctly pointed out by Tong et al, at the limit of the axial resolution of OCT.

The paper does not follow published OCT quality control and reporting guidelines and Figure 1B shows that this will be an issue. The OCT B-scan shared in this image was taken by the OCT technician with the retina not horizontally aligned in the live window. Hence the asymmetric signal intensity for the OPL/ONL to the right and left of the foveola caused by oblique illumination of the Henle fibres. There is a known effect from this on other segmented retinal layer thickness data. This can be repaired, but will require revision of individual B-scans.

Figure 1C and D suggest a perfect segmentation at the outer part of the volume scan which in practise is not always achieved. Where there individual B-scans cut off at the borders? If yes please report.

Fig 1E requires a legend explaining the colours. I presume they are linked to 1D, but this is very confusing for the perimacular area.

The formula in line 173 requires a reference.

The classification into age groups of 32.4, 46.4 and again 32.4 (line 186) is unusual and I think duplication of the 33.4 year group was a typo.

Abbreviations (eg Rx) in Table 2 require explanation.

Figure 2A suggests that there are ganglion cells in the foveola?
That should not be the case and is likely a downstream effect from lack for OCT QC according to validated guidelines. Most likely this is an artefact from the segmentation algorithm.

Table 1 suggests that there are many more subjects in the INL group than the GCL data group.Yet the error bars in the graphs to the right are smaller for the GCL than for the INL in Figure 2A&B;? How can this be explained?

The error bars for Figure 2C are as expected from the smaller numbers. The altiduinal pattern does however look very odd.

The Bland-Altman plots in Figure 3A&BV; show that there is a problem because the distribution is not random around the zero-line.

Figure 3 is potentially very interesting because it makes the link to histological data. But Figure 1B is idealistic. This does not match the data on the INL shown earlier. Likewise the altitudinal pattern for the GCL in Figure 3A does not match earlier OCT data. Is there any chance to get more complete histological data for the grid data reported?

Figure 5 is largely speculative because OCT does not permit to delinate the individual cell types presented. The major issue is that data from different sources are collated.

Taken together this is a very interesting and timely study which does address a relevant point. The challenge remains to demonstrate these correlations between OCT and histological data experimentally.