To the Editor: Langsted and Nordestgaard, of the Copenhagen General Population Study (CGPS),1 stated that “calculated LDL cholesterol concentrations do not change in response to normal food intake in individuals with and without diabetes after correction for hemodilution” (1). Using unpaired, nonrandomized samples, the authors estimated lipid differences between fasting and nonfasting individuals. Within-individual changes were not measured. Thus, to us, the authors’ conclusions about fasting/nonfasting “changes,” “increases,” or “decreases” were not appropriate. Notably, the maximum difference in the mean triglyceride concentration between fasting and nonfasting individuals with diabetes was 18 mg/dL (0.2 mmol/L). This difference apparently doubled to 35–44 mg/dL (0.4–0.5 mmol/L) (as graphically read) after adjusting for age and sex. We believe that a contrast of this magnitude, apparently arising from covariate adjustments, suggests considerable confounding between fasting and nonfasting individuals. Although Langsted and Nordestgaard did not mention the data in their report, the CGPS had previously reported significant differences between fasting and nonfasting individuals with respect to age, sex, smoking, and so forth (2), underscoring the possible confounding effect of self-selection by participants choosing fasting or nonfasting blood draws. Although the CGPS is a large-scale study, we do not believe that sample size per se affects confounding; it affects only statistical power. Albumin concentrations in the CGPS were significantly lower in nonfasting individuals than in fasting individuals (approximately 4%–5% lower at 1–2 h postprandially, as graphically read, P < 0.05), as were the concentrations of some lipids, e.g., LDL cholesterol (LDL-C) (approximately 9%–20% lower at 1–2 h postprandially, as graphically read, P < 0.01). After adjustment for albumin, the statistical significance in LDL-C differences disappeared. The CGPS used albumin as a proxy for hemodilution; that is, …