This study explored the effect of breakfast on FFM estimation and found no clinically relevant difference between fasted and non-fasted measurements. In 90% of participants, FFM estimates changed less than 1 kg compared to their fasting value. In four participants, the difference exceeded the pre-set limit of 1 kg, of whom two admitted to having breached the research protocol. The most pronounced mean difference in FFM was found after 3 h; this was statistically significant but not clinically relevant, remaining well below 1 kg difference. We hypothesized that this difference after 3 h was due to the uptake of the breakfast at this time. In the first 2 h after breakfast, most of the food remains in the stomach, which would not affect impedance outcomes [15]. However, body weight increased due to the breakfast mass, influencing the calculation since weight is a variable in the Kyle formula [3].

The results of our study are consistent with findings of Hollander-Kraaijeveld et al. (2020, n = 84), in cystic fibrosis patients. These authors found a mean decrease of 0.2 kg in FFM after eating a non-standardized meal, with a difference of < 1 kg of FM and FFM in 86% of the patients [16]. Although the trends compared to our study are in different directions in these (metabolically) substantially differing subject populations, the average change was minor and not clinically relevant. Furthermore, both studies found opposing individual responses, with participants increasing or decreasing FFM estimation after eating.

Androutsos et al. (2015, n = 55) studied the effect of a high-fat meal or high-carbohydrate meal on impedance and FM estimation [17]. These authors reported that average FM increased most after 2 h, increasing 0.8 kg with a median difference of 4.8%. While the present study found a decrease in FM estimation, both studies have found non-clinically relevant changes (Table 2).

Our results are consistent with the three studies mentioned in the ESPEN guideline [9,10,11]. These all found differences in resistance and impedance outcomes before and after eating. However, none of these differences appeared clinically relevant, i.e. leading to treatment changes. In hindsight, in our opinion, we never had robust evidence to let our patients fast. Thus, all studies in the literature report similar results, while concluding on statistically significant differences in BIA outcomes but without clinically relevant differences between fasted and non-fasted BIA measurements.

In our study, a difference of ≥ 1 kg in FFM was considered clinically relevant, whereas Hollander-Kraaijeveld et al. used ≥ 1.5 kg as a clinically relevant difference. If we also defined 1.5 kg as a clinically relevant difference, we would only have one outlier left, which empowers the argument of measuring in a non-fasting state.

According to the literature, there is a within-day variability of 1–2% of resistance when performing an SF-BIA measurement and a weekly intra-person variability of 2–3.5% [3]. Comparing these values with the CV of 0.42% in FFM between time points found in this study, the variation due to having breakfast is below the within-day and the weekly variability. This finding further supports our opinion that measuring in a fasted state is unnecessary.

Weight increased on average by 0.2 kg 1 h after breakfast and decreased by 0.2 kg after 4 h (Table 2). This increase was due to the weight of the breakfast, while the decrease can be explained by losing water due to urinating before each measurement. There was no statistically significant change in reactance, while resistance and impedance decreased by 6 Ω after 3 h (Table 2). This indicates that this change caused the increased FFM estimation at t3, rather than weight, since the average weight was similar at baseline and at this time point.

This study has limitations, including the homogeneity of our study population, mostly compromising of females aged 24–38 with a healthy BMI. This homogeneity reduces the external validity of this study, but there is no evidence to suggest that there would be other findings in other types of populations. Furthermore, participants could choose between two breakfast meals. Differences in composition between the meals could have affected the BIA measurements differently, although we found no difference between breakfast groups regarding all outcomes. Hollander-Kraaijeveld et al. also did not find clinically relevant differences, and their participants had no limitations regarding nutritional intake [16]. Finally, we cannot be completely certain whether all subjects correctly reported their adherence to the study protocol.

Concerning the strengths, the research was standardized, following the SOP. Moreover, all measurements were performed by the same researcher, and on each study day, the SF-BIA was calibrated using the same device throughout the study.