Features Associated with Visible Lamina Cribrosa Pores in Individuals of African Ancestry with Glaucoma: Primary Open-Angle African Ancestry Glaucoma Genetics (POAAGG) Study

Round 1

Reviewer 1 Report

The manuscript “Features Associated with Visible Lamina Cribrosa Pores in Individuals of African Ancestry with Glaucoma: Primary Open-Angle African Ancestry Glaucoma Genetics (POAAGG) study” by Jalin A. Jordan et al. provides statistical analyses – both uni- and multi-variate – connecting visibility of Lamina Cribrosa pores (LCP) to various morphological and clinical glaucoma-related factors and characteristics amongst the African Ancestry glaucoma cohort of the POAAGG study.  The paper is generally well written, and the study and its results are interesting and could over time prove to be clinically important.  I would recommend several important areas of improvement before a decision to publish is made:

Major comments:

11)      It would be very valuable to add a figure or two to support the LCP identification/reconciliation processes described in the Methods.  For example, what do LCPs look like in the fundus photos including examples of no pores, <=3 pores, and >3 pores; what were examples of disagreement between the original gradings as well as in the re-gradings; an example of an ungradable photo; etc.?

22)      In Materials and Methods around line 80 it is stated that the data set includes glaucoma cases, suspects, and controls.  However, the analyses, data tables, and associated discussions seem to focus only on the glaucoma cases.  Would it be possible to add analyses for the three subject categories vis-à-vis one another to an opening table?  It was surprising not to encounter this over the course of the manuscript.

33)      How was the threshold of 3 pores decided upon to delineate the “yes” and “no” pores cases?  How do we know if this is a proper or optimal threshold?

44)      As this is not a case-control study and, more importantly, because the prevalence of LCPs is quite high in this study (typically in the range of 50-90+% of cases), the use of odds ratios (both OR as well as aOR) is misleading in numerous cases.  For example, the Deep Cup Depth univariate category has an OR of 6.47 where the relative risk ratio is only 1.77.  The confidence intervals do not even come close to including this 1.77 level.  There are numerous other similar examples as well.  As such, the usage of relative risk ratios in place of OR-based results is strongly preferred for LCP analyses (again, where the prevalence or exposure is very high).

55)      The last sentence of the abstract implies that there is sufficient evidence that physicians apply LCP findings clinically, however it isn’t clear if doing so currently adds diagnostic or clinical value.  Furthermore, additional information such as refinement of the methodology and/or assessments of reproducibility, etc. would be needed before a methodology were to be clinically useful.  Please edit this sentence to soften the statement.

66)      Like the final sentence of the abstract, the last sentence of the manuscript again calls on clinicians to utilize LCP information in the management and treatment of glaucoma patients.  However, it does not seem that appropriate guidelines exist yet, and this (and any similar statements) should be softened to an optimistic but less definitive outlook.

Minor comments:

77)      In the opening sentence of the abstract, “as well” should be “as well as.”  In addition, the nature of the “impact” should be expanded at the very end of the sentence: i.e., “impact on []”.

28)      Why weren’t the 439 eyes where the original graders agreed as “none” regraded?  For example, would any of those cases be put into the “ungradable” category if they had been regraded?  (Since this would only increase the degree of regrading by about 15%, it would seem beneficial to do so.)

99)      BMI is described as “basal metabolic index” (in Table 1).  This should be corrected to “body mass index”.

110)      Were parameters related to eye shape/length evaluated?  For example, it would be interesting to see if axial eye length or refractive measures had any relationship to LCP visibility.  The aspect of interest here would be if magnification was having any impact on the LCP visibility.

111)      While this study was conducted entirely using fundus photos, how do the authors feel about the potential of OCT (optical coherence tomography) and the 3D data it provides to quantify LCPs (and perhaps other aspects of the LC mesh)?  How does pore visibility in OCT compare to that in CFPs?  Please include such aspects in the discussion.

No comments in particular, though a small number of minor typos were detected so a review is recommended (one typo in the abstract is noted in the above comments).

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

I appreciate the opportunity to review this paper. The paper explores the biomarker relevance of a series of phenotypical and demographic factors in association with LCP. The paper in general is very well written and is very intuitive in its narrative. I enjoyed readding the paper. It was unfortunate to see that the GWAS hits did not show any significance. However, the paper does a good job of presenting relevant phenotypical factors.

I believe this paper only requires a minor revision. One part I would suggest some clarification in the DNA sample processing and quality control. This is mentioned in the methods in lines 82-83 but is not developed. Also, I am experienced with fast STRUCTURE, but not familiar with the specific pipeline used. I would suggest expanding on that. You do have a citation, but I would appreciate having a bit more information in the paper.

One thing I believe may be useful is to add more on the distribution of the q0 and q1 of the sample. How do the cases, suspects and controls tally on that metric? What about the African, Afro-Caribbean, and Afro-American groupings. These aspects are not directly related to the setup of the assessment, but they provide context on the characteristics of the sample. This may be interesting as samples were collected from a variety of locations in Philly, I wonder if the sampling may be the direct reason why the SNPs assessments were not successful.

Last, I wonder about the performance of the multivariate models. You used a backward selection model using a very relaxed p<0.2. The assessment was not designed to assess predictability performance, but I think it may be relevant to consider some of that since you are referring to these phenotypical factors as biomarkers.

Just curious why GEEs were used? those are useful for longitudinal or correlated data, you don’t have longitudinal data here. Was the possibility of some collinearity of the factors in the multivariate models the motivation? Were they to address normality concerns?

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The revisions to the manuscript overall are acceptable, but I do renew my concern regarding the use of Odds Ratios (OR) as the basis for much of the statistical analysis.  Since the incidence of pores was often quite high, the reported ORs are likewise artificially high (e.g., the 6.47 value in Table 2, among others).

At the very least, the others should insert a paragraph into the Discussion describing the limitations of and potential inaccuracies in using ORs for this application in which incidence is rather high.  I continue to recommend changing to a different statistic, though.

Author Response

Please see the attachment

Author Response File: Author Response.pdf