The reliability of energy intake assessment depends on the accuracy of the dietary software in secondary data from female CrossFit athletes

INTRODUCTION: Accurate estimation of energy and macronutrient intake is essential in studies involving athletes, particularly when evaluating dietary adequacy and investigating risks related to insufficient energy intake. Although commercial nutrition applications are widely used due to accessibility, their accuracy compared with validated research tools remains uncertain. This short communication compared nutrient intake estimates generated by the commercial WebDiet (WD) software with those obtained using the research-grade Nutrition Data System for Research (NDSR) in female CrossFit athletes, with emphasis on energy and carbohydrate intake. METHODS: This cross-sectional analysis used secondary data from 47 adult female CrossFit athletes assessed at PROMOVE São Camilo, São Paulo, Brazil. Dietary intake was measured using 2 non-consecutive 24-h food recalls administered by trained nutrition professionals and analyzed using WD and NDSR. Usual intake was estimated with the Multiple Source Method. Paired t-tests, effect sizes, Pearson correlations, intraclass correlation coefficients (ICC), and Bland-Altman analyses were applied to compare methods. RESULTS: WD produced significantly lower estimates of total energy intake (1762.78 ± 503.01 kcal) compared with NDSR (1843.49 ± 504.28 kcal; P = 0.022; d = 0.346). Carbohydrate intake, expressed in g/d and g/kg, was also underestimated by WD (P < 0.001). Protein and fat estimates did not differ significantly between methods. Agreement analyses demonstrated moderate to excellent ICC values for energy, carbohydrate, and protein intake (ICC: 0.86-0.89), while fat showed moderate agreement (ICC = 0.70). Bland-Altman plots indicated systematic underestimation by WD, particularly for energy (bias = 80.7 kcal) and carbohydrates (bias = 22.3 g), with wide limits of agreement suggesting substantial individual variability. CONCLUSION: WD consistently underestimated energy and carbohydrate intake when compared with NDSR, while protein and fat estimates remained more comparable between methods. Agreement analyses confirmed that, although correlations were high, the magnitude of individual differences was meaningful, especially for variables critical to dietary assessment in athletes. These findings highlight that the accuracy of nutrient estimates depends strongly on the software used. Research-grade platforms such as NDSR provide more reliable reference values for evaluating energy intake in studies involving physically active populations, whereas commercial applications may introduce systematic bias with potential implications for nutritional screening, interpretation of dietary adequacy, and decision-making in sports nutrition.