The large within- and between-sample variability in breast milk lipid content greatly complicates the collection of representative samples in field studies. The main purpose of this study was to validate the ability of individual daytime samples to predict the 24-h lipid concentration of breast milk. We also studied maternal, child, and other factors (time of day and interval between feeds) associated with the within- and between-mother variability in milk lipid content. Fifty-two primiparous urban Guatemalan women between 1 and 4 mo postpartum were studied. Milk samples were collected during six 2-h intervals from 0600 to 1800, and throughout the night when the child breast-fed. On average, the 24-h pooled milk samples contained 4.2 +/- 0.92% (mean +/- SD) lipids and the best concordance with this value was obtained with samples collected between 0600 and 0800 (concordance correlation coefficient = 0.60, P < 0.05). None of the regression equations to predict the 24-h lipid content of breast milk based on daytime samples reached a sufficiently high predictive power to be recommended for the estimation of individual child intake. Time of day and time elapsed since the last feeding were significant determinants of diurnal variations in milk lipid content, whereas between-mother variability was explained by maternal weight (P = 0.05) and body mass index (P < 0.05). For the collection of milk samples in surveys and pre-post studies, we recommend standardization of time of day and interval between feeds.