Estimation of non-shivering thermogenesis and cold-induced nutrient oxidation rates: Impact of method for data selection and analysis.

Background & Aims: Since the discovery of active brown adipose tissue in human adults, non-shivering cold-induced thermogenesis (CIT) has been regarded as a promising tool to combat obesity. However, there is a lack of consensus regarding the method of choice to analyze indirect calorimetry data from a CIT study. We analyzed the impact of methods for data selection and methods for data analysis on measures of cold-induced energy expenditure (EE) and nutrient oxidation rates.

Methods: Forty-four young healthy adults (22.1 ± 2.1 years old, 25.6 ± 5.2 kg/m, 29 women) participated in the study. Resting metabolic rate (RMR), cold-induced thermogenesis (CIT), and cold-induced nutrient oxidation rates were estimated by indirect calorimetry under fasting conditions during 1 h of cold exposure combining air conditioning (19.5-20 °C) and a water perfused cooling vest set at a temperature of 4 °C above the individual shivering threshold. We applied three methods for data selection: (i) time intervals every 5 min (5min-TI), (ii) the most stable 5-min period of every forth part of the cold exposure (5min-SS-4P), and (iii) the most stable 5-min period of every half part of the cold exposure (5min-SS-2P). Lately we applied two methods for data analysis: (i) area under the curve as a percentage of the baseline RMR (AUC) and; (ii) the difference between EE at the end of the cold exposure and baseline RMR (Last-RMR).

Results: Mean overall CIT estimation ranged from 11.6 ± 10.0 to 20.1 ± 17.2 %RMR depending on the methods for data selection and analysis used. Regarding methods for data selection, 5min-SS-2P did not allow to observe physiologically relevant phenomena (e.g. metabolic shift in fuel oxidation; P = 0.547) due to a lack of resolution. The 5min-TI and 5min-SS-4P methods for data selection seemed to be accurate enough to observe physiologically relevant phenomena (all P < 0.014), but not comparable for estimating over-all CIT and cold-induced nutrient oxidation rates (P < 0.01). Regarding methods for data analysis, the AUC seemed to be less affected for data artefacts and to be more representative in participants with a non-stable energy expenditure during cold exposure.

Conclusions: The methods for data selection and analysis can have a profound impact on CIT and cold-induced nutrient oxidation rates estimations, and therefore, it is mandatory to unify it across scientific community to allow inter-study comparisons. Based on our findings, 5min-TI should be considered the method of choice to study dynamics (i.e. changes across time) of CIT and cold-induced nutrient oxidation rates, while 5min-SS-4P and AUC should be the method of choice when computing CIT and cold-induced nutrient oxidation rates as a single value.