Review History
Neck circumference is a highly reliable anthropometric measure in older adults requiring long-term care

Thank you for your revised version and rebuttal. I am pleased to accept your manuscript. There are some small edits that I think are needed (listed below) but I think you can address these in the preparation of the final version of your work. Congratulations!

Lines 62–63: “Among anthropometric indices, measurement of the neck circumference has the less time-consuming” should perhaps be “Among anthropometric indices, measurement of the neck circumference is less time-consuming”.

Table 2: “Age (years) 75≤” should perhaps be “Age (years) ≤75” (compare with the following two rows).

Figure 3: Shouldn’t the y-axis be “Absolute difference in…”? I’d suggest adding “absolute” before “differences” in the caption as well.

[# PeerJ Staff Note - this decision was reviewed and approved by Mike Climstein, a PeerJ Section Editor covering this Section #]


**PeerJ Staff Note:** Although the Academic and Section Editors are happy to accept your article as being scientifically sound, a final check of the manuscript shows that it would benefit from further editing. Therefore, please identify necessary edits and address these while in proof stage.

Thank you for your patience. I’m not the original editor of your manuscript but was invited to take that role recently. I’ve read over our three reviewer reports and as you can see, Reviewer #1 has no changes to suggest. Reviewers #2 and #3 have both provided some helpful comments that I think will assist you in revising your manuscript.

Reviewer #2 makes a good point about the reliability versus the utility of the measure. Your title makes it clear to me that you are interested in the reliability of the measure but I agree that you could more clearly motivate why measurements of neck circumference are important/useful in the first instance (expanding on Lines 57–71). Their suggestion that you could consider the validity of this measure by correlating it to body fat (and muscle mass if available) is something you could consider adding, or you could keep your focus entirely on the reliability aspect.

Reviewer #3 suggests a reference to add. You are welcome to include this if you find it helpful. The topic of how valid measurements are against a gold standard would be interesting, but I suspect that it would fall outside of your intentions in the present manuscript. Their other comments warrant your consideration and response.

I will add some of my own comments. These are often requests for clarification of some details of the analyses (speaking as a biostatistician who often works in anthropometry). I look forward to seeing a revised version of your manuscript with responses to each of the reviewers’ and my points and any changes in your manuscript tracked.

Your use of Bland-Altman for two examiners seems slightly unusual here to me. Bland-Altman is for method comparison rather than reliability analysis per se, and your two examiners are not methods of measurement or, as far as I could tell, of specific interest in themselves. Bias between the two examiners would not seem to be of particular interest to me (unless one was, say, experienced and one a novice, which is not the case from Line 100, although exploring such a difference would require more instances of each type, but this would lead to a situation where Bland-Altman analysis would seem entirely appropriate to me). If there was statistically significant evidence of a bias (e.g., Lines 133–135), what would this mean unless we knew exactly who the two raters were (the sign of the bias cannot be interpreted without this information)? If you mean that these are two randomly selected raters, the bias doesn’t seem as interpretable to me and in that case, it is only the greater (in magnitude) value in the LoA that seems particularly important to me. I wonder what you feel the bias and LoA add on top of the ICC.

I would also have considered using a mixed model for the calculation of the LoA to reflect the randomly selected (representative) raters if this was retained but it was not clear to me that this was done. You may well have another interpretation intended here that I’m not picking up on, so could you clearly explain what you’re aiming to achieve with the Bland-Altman analysis in your manuscript beyond what is already shown through the ICC analysis?

I was initially unclear how the ICC(1,1) calculation (Line 125) was performed as each examiner measured each participant twice, giving four measurements per participant. For this, the underlying mixed model is slightly more complicated than with a single assessor. Could you clearly explain what was done around Line 125? From Lines 147–148, it sounds as if this was done separately for each rater, but again without knowing the raters, I find this difficult to interpret and a single intra-rater ICC could be provided using data from both.

I also wondered if it would be useful to use Spearman-Brown’s prophecy formula to indicate how many measurements would be needed from an examiner to achieve satisfactory reliability (e.g., while a single weight or height measurement ought easily to be sufficient for WHO’s 0.95 reliability criterion, multiple circumference measurements can be needed). Given the results you use, your ICCs are high enough to justify single measurements, but you could do this using the lower 95% CI limit for the ICCs (unless these are also greater than 0.95 or whatever you consider acceptable in this specific context).

Related to the above, 95% CIs are needed for all ICCs.

The presentation of Pearson’s correlation on Line 150 (and again on Line 154) should be removed unless it can be justified. Same for Table 2 and anywhere else it is presented. The use of ICCs (and Bland-Altman’s method) is intended to avoid the problems that arise from using such correlation coefficients.

The MDC on Line 152 needs units (cm). The interpretation on Line 166 isn’t quite correct and should be revised. The focus from the title is on reliability, and while reliability affects MDC, I wonder if the presentation of the MDC could be more clearly motivated for the reader.

Were the two measurements taken in random orders for each participant (i.e., was each examiner first for around 52 participants, or, alternatively, were the measurements counterbalanced)?

I think you can look at ECW/TBW against the difference between examiners, but I wouldn’t call that difference “measurement error” (e.g., Lines 38, 142, and 157) as we have no gold standard. Perhaps just “measurement difference”? You could also argue for using the absolute value of this difference if the question was would edema make measurements harder.

I wonder if reporting ICCs for some subgroups, including perhaps categories defined by ECW/TBW such as Line 115, might be interesting and useful. I appreciate that this becomes challenging with the total sample size, but was (and you could do this for both intra- and inter-rater) reliability higher or lower for men or women, with edema (as you give on Line 153 and Table 2 for inter-rater reliability) or without, younger or older? These groups need to be meaningful to you and I mean these only as examples.

When you report descriptives in Table 1, which rater/measurement occasion are these from for neck circumference?

For Figure 3, I would have thought that the difference in measurements (first measurement or mean of both for each rater) would have been on the y-axis.