Dairy Cow Longevity Is Affected by Dam Parity and Age

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Specific comments on the manuscript

L 69 Are all cows the same breed? Holstein? It should be mentioned, as there might be breed differences on this matter

L 75-76 Convert to 20-130 months ? (to be consistent with units in HL y LPL)

L 77 I assume this refers to First-Calving?

L 87 and L 102 What are the distributions (and link functions) assumed in these GLMM models for the response variables? HL and LPL are generally  right-skewed. Not sure about PLI.  

L 106   It should be "the linear regression coefficient of parity number on the response variable" 

L 115  What is the scale used for Age of Dam ?   years? Rounded to nearest unit? Important to better understand regression coefficients given  in results  

L 120 I assume this comparison was performed on LSM? 

Table 1.  Although I understand these are LSM obtained from GLMM, I still expected PLI be  approx. equal to LPL/HL (as defined earlier), but it is larger?  Why?.  

Table 2.  According to these data for HL and LPL,  Age at First Calving would be around 30 mo? Is this consistent with local AFC? It seems high

Tables 1 and 2. There are 2 things that strike me from these results

1. I would expect a larger difference in the magnitude of standard errors for different parity groups, given the mcuh lower n available for later compared to earlier parities. 

2. It also seems very odd that a difference of magnitude 1 (or even less!) in LSMs for different parities is declared significant (P<0.05) when standard errors are close to 3/4 units or magnitude? This applies to all three response variables. These large SE imply that all confidence intervals are highly overlapped. I wonder how can this happen? In my experience with similar models  and variables (HL, LPL) the magnitude of SE and CI95%, are largely dependent on sample size, and significance is also consistently associated to non overlapping 95%CI.  

I would suggest to have a further look into the model and see how these SE behave when some effects are dropped from the model or when different comparison tests are used.   

L 142 : check R2 032?

L 145: check R2 032?

Discussion:  In general, I find the discussion pertinent and well supported.  Perhaps a genetic perspective is lacking in this discussion, which is mainly devoted to physiological aspects. Especially since longevity is a trait that has been considered in selection indexes already for several decades. 

Conclusions:  The conclusions seem to make sense from a biological perspective, but I have some doubts about the LSM, SE and results from the comparison test, presented in T1 and T2. I recommend the authors thoroughly check these results to rule out any errors in the proposed models.

Author Response

Please see the attachment. 

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This is a nice, well written study addressing a current knowledge gap.  I feel that there needs to be a bit more clarity regarding the ethical approval of this work - currently there is no mention of whether ethical approval was granted which warrants correction.  It is also unclear whether the authors' had permission to use data included in the Dairy Herd Improvement Database for research purposes and further clarity about this would be beneficial. 

Detailed suggestions: 

Abstract

The inclusion of p-values in the abstract is not strictly necessary but if you do decide to keep these here, please report the precise p-values, not "p < 0.05" (down to p = 0.001, below which "p < 0.001" is conventional).  

Introduction

Line 50: Please correct "we hypothesize" to the past tense "we hypothesized"

M+M

General:

Were there any exclusion criteria used? If so, this needs including here.  If not, a brief statement saying this might be worthwhile.  I also think it would be worth including here if any dairy breed was included, or if you only included certain breeds.  Additionally, based on your other criteria I presume that only female animals were included but it would be good to state this so the reader does not have to make assumptions.  

The description of the statistical modelling omits some important information about the GLMM that warrants inclusion:

·       How were assumptions of the models (including data distribution) and goodness of fit verified? What criteria were used to compare models and select the final model?  

·       How were missing data handled? 

·       What link function was used? 

·       What covariance structure was chosen? 

I’m afraid that from your reporting of the statistical analysis I don’t understand why you chose to model the data using both a GLMM and analysis of co-variance.  Please could an explanation be included in the M+M to clarify this - I can see that you have reported the linear effect in the results, but I fail to see the clinical relevance of this, so these results don't really explain why this was done.  I also cannot see how you manipulated categorical data to fit a linear model – this also requires explanation.       

Line 63: Please write out the abbreviations here in full the 1st time they are used in the main body of the text (I appreciate they are written out in the abstract but it bears repeating here for clarity)

Line 68: The end of this sentence suggests that PLI is calculated by dividing a proportion of LPL by HL which I don't think is your intention.  I suggest rephrasing this to read "...calculated as LPL divided by HL expressed as a percentage" (if this is the correct interpretation).  If I am wrong and you did divide a proportion of LPL by HL, what proportion this was needs to be stated.  

Line 72: It is unclear why 5 farms were selected and how these were selected.  Were they the only 5 that met the criteria or were they a subset that was selected for another reason?

Line 76: I suggest you slightly re-order this line to read "no cows culled before calving" for grammatical correctness and clarity.  

Line 79: The final number of observations is not inclusion criteria (unless you set an a priori cap of 12792).  This would be better placed either in the results section, or in the statistical analysis section as information regarding the models.  

Line 84: It is unclear why you chose to include season of birth in your model – did you think this might have an effect? If so, some background information in the introduction would be helpful for context. 

Line 93: How was each season defined (i.e. which months were assigned to each season)

Results

General:

I think the results section would be improved if descriptive statistics included, including population statistics.   

Similarly to the abstract, please report exact p-values in line with APA guidelines.  

Tables 1 and 2: I find the formatting of these tables to be a bit confusing as they are currently laid out with comparisons being made down the columns, rather than along the rows.  I would suggest that you swap the columns and rows around so that your outcomes (HL, LPL, PLI) each have a row and parity has a column each.  I would then also include a column at the end (far right) for p-values. The number in each group can be included as a footnote or in a row at the bottom of the table.     

Example:

Longevity metric	Parity		p-value
	Primiparous	Multiparous
HL
LPL
PLI

 

 

Discussion

General: There is no discussion of the limitations of this study and this needs to be included here.  I also feel that there is some overreach in parts of the discussion – for example, it is suggested that colostrum quality (typically lower in heifers) might be affecting the longevity of the progeny in this study, but your study does not have the data to support this, so it is speculation.  Furthermore, you also discuss the separation of calves from dams potentially minimising maternal behaviour effects on offspring but not every dairy farm does this and of those that do, the timing of cow-calf separation varies substantially - your study does not include this data so these conclusions cannot be drawn.  Having said this, if you include some discussion of study limitations, these concerns will be addressed as part of this.  

Line 157-158: The introductory line that ends … “are discussed as follows” is not necessary, consider removing this. 

Line 213: Consider replacing ‘biological model’ with ‘management’ (or something similar) as ‘biological model’ suggests that cow-calf separation is a natural part of bovine postpartum behaviour, when it is actually performed as a human choice to aid management. 

Author Response

Please see the attachment. 

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you for addressing the reviewer comments so thoroughly.  This manuscript has been much improved by this and I have no further comments.