Critique
• The authors should provide more information about how that diagnosis determined weight loss. Was this by history or through medical records?
• The authors should provide a timeframe for weight loss
• On page 3, a reference as needed for the definition of cancer cachexia
• There are 3 groups (metastatic, locally advanced, resectable but not a candidate for surgery) that have different outlets and should be analyzed separately.
• There are no quality metrics for the CT scans’ reader. There was a single reader, but nothing has been given about the reliability regarding the Least Significant Change as a random error. This is critically important because changes as little as 5% may have clinical relevance. Interclass coefficients are inadequate
• The psoas muscle only was the measure of skeletal muscle which is not validated. The usual L3 cuts include 7 muscle groups.
• In the multi-variant analysis, weight loss was prognostic. I would question the method for propensity matching, which does not substitute for stratification of a randomized clinical trial and is constructed by the authors and hence subject bias
• There are several confounders between the weight loss group and the non-weight loss group, suggesting that there are likely hidden confounders that are likely to bias the result.

The title should be. "Does pretreatment weight loss predict survival in patients with advanced pancreatic cancer?"
How precise were the CT scan skeletal muscle (psoas muscle, SCF, VFA) readings. That is the reproducibility of the CT scan measurements..

Critique
• This study design itself is adequate
• The outcomes include skeletal muscle area and weight change on precision. The authors have added that Dr K H did repeat measurements for reasons of precision; however they do not provide data regarding the precision of his readings. This should be presented as the least significant change and not as ICCs
• Patients were weight frequently. I assume that different scales were used though this may not be true. Since, again the weight is an outcome and there may be variability between scales it would be important to make some mention of the precision. Significant variability and precision may lead to a type II error