Increasing the lactating sow feed intake is generally considered beneficial, but high inter-individual variability of the measurements makes difficult designing the trials on this subject. A meta-analysis of 10 trials (1111 sows, 8 genetics lines, parity 1–8, lactation duration 19–31d) was undertaken, comparing “control” diets (n = 12) to diets supplemented with the same strain of live yeast Saccharomyces cerevisiae boulardii CNCM I-1079 (LY) at 1.0 or 2.0 × 10⁹ cfu/kg (n = 12), to increase the chance to detect feed intake differences. In these trials, individual feed intake of sows (ranging 3.9 to 8.5 kg/d) was measured; trials were performed in commercial farms or close situation, in Europe or North America. Measurements of litter weight at weaning and number of piglets weaned could be retrieved from all 10 trials, but not changes of sow weight or P2 backfat thickness. Data were analyzed by ANOVA (SPSS 19.0) incorporating a random effect of trial, the treatment effect being considered weighted or not for experimental power or measurement variability. Effect of location of the experiment, year (1999 to 2013), lactation length, or duration of pre-treatment with live yeast were incorporated in the model, then excluded because of lack of significance (P > 0.20). Litter size, litter weight at weaning, parity or backfat thickness were included as covariate, and then discarded from final model as neither significance nor correlations were found. Live yeast added to sows diets stimulated appetite overall (Table 1), leading to a significant increase of feed intake ranging between +200 (weighted by number of replicates) and +300 g/d (un-weighted). Feed intake difference higher than the mean standard error of measurement was found in 6 trials out of 10. Nonetheless, factors affecting the magnitude of yeast effect could not be elucidated with only 10 trials in the compilation. Subsequently, the present results deserve being confirmed with new studies including accurate measurements of sows DMI and body condition. Table 1. DMI (kg) per treatment according to the statistical model

Weighted	Control	LY	SEM	P-value
None	5.714	5.997	0.346	0.03
1/Var	5.725	5.985	0.340	<0.01
No. of replicates	5.608	5.805	0.345	0.01