A neural mechanism for affective well-being: Subgenual cingulate cortex mediates real-life effects of nonexercise activity on energy

A neural mechanism underlying the real-life effects of nonexercise activity on well-being promises novel intervention targets.


This PDF file includes:
Figs. S1 to S3 Sections S1 to S4 Tables S1 to S5 Figure S1. Research plan Section S1.

Robustness against alternative definitions of physical activities
To test if the definition of physical activities used may have influenced our results, we conducted additional analyses. While the definition of exercise vs. non-exercise activities considers both physiological and psychological processes involved in physical activity (e.g., the (social) context, motive, structure, duration, and energy expenditure; for details see introduction section and (4)), a more parsimonious definition following categories of physical activities derived from energy expenditure estimates that are well-established and broadly used (e.g., to investigate effects of physical activity on somatic health outcomes), can be vigorous vs. light/moderate physical activity, therewith focusing on the physiological aspects of physical activity. Therefore, to test if the latter and alternative definition of physical activities influences our results, we conducted a series of additional analyses using the extant analyses procedures but the definition of vigorous vs. light/moderate physical activity instead of the differentiation non-exercise vs. exercise activity. First, participants' physical activity levels were correlated across the different definitions, see Table S1. Second, in our discovery study, using the definition of vigorous vs. light/moderate physical activity instead of the differentiation non-exercise vs. exercise activity lead to the same results, i.e., a significant effect of light/moderate activity yet a non-significant result for vigorous activity on energetic arousal (see Table S2). Third, control analyses in the replication study further substantiated this finding: light/moderate activity but not vigorous activity predicted energetic arousal (see Table S2).
Together, these additional analyses using the alternative definition of physical activities lead to the same findings as the distinction of exercise vs. non-exercise activities therewith providing further evidence that momentary positive effects of physical activity on energetic arousal in real-life are specifically driven by every-day activities of light/moderate intensity (which are most prevalent in human life; https://sites.google.com/site/compendiumofphysicalactivities/) but not by vigorous activities (oftentimes characterizing physical exercise sessions; https://sites.google.com/site/compendiumofphysicalactivities/. For details on these analyses, refer to Section S1. For details on these analyses, refer to Section S1.

Section S2. Robustness against sleep as a potential confounder
To test whether sleep may confound our findings, we conducted additional analyses. We operationalized sleep using the accelerometer data. Since participants had been thoroughly instructed to wear the devices throughout all waketimes but not during sleep, we quantified sleep duration via the non-wear times of the accelerometers. In particular, we parameterized sleep duration as the daily amount of accelerometer non-wear time during night times, which we defined as the time frame from 22.30 to 07.30 since participants received e-diary prompts each day between 7:30 and 22:30. Put simply, this sleep parameter represents the non-wear time between 22:30 and 7:30 for each night. We entered this parameter into our main analysis as an additional within-subject variable on a day level (sleep minutes per participant and per day), which lead to a three-level multilevel model for this additional analysis: measurements (on level 1) were nested within days (on level 2) and participants (on level 3). However, incorporating sleep duration (i.e., controlling for potential effects of sleep duration within the night on the e-diary ratings of the following day) did not alter our findings (Interaction analyses sgACC*non-exercise activity: Beta coefficient = -0.148; P value = 0.001).

Section S3. Robustness against potential outliers in exercise duration
We researched if potential outliers in exercise duration, i.e., data of participants engaging in relatively high amounts of exercise across the study week influenced our results. Therefore, we first explored the distribution of exercise duration per participant and study week in the discovery vs. the replication study, see Figures S1 and S2 below. Given the distribution of the data, we decided to define outliers as exercise duration > 600 minutes/participant/week, yielding two potential outliers in the replication study. However, excluding these outliers from our main analysis did not change the results (Interaction analyses: sgACC*non-exercise activity: Beta coefficient = -0.150; P value = 0.001).

Figure S2. Distribution of exercise duration in the discovery study
The histogram depicts the distribution of the cumulative exercise duration across the study week (in minutes; x-axis), in which participants of the discovery study (y-axis) engaged in; 43 of a total of 67 participants did not engage in any physical exercise at all across the study week.

Figure S3. Distribution of exercise duration in the replication study
The histogram depicts the distribution of the cumulative exercise duration across the study week (in minutes; x-axis), in which participants of the replication study (y-axis) engaged in; 33 of a total of 83 participants did not engage in any physical exercise at all across the study week.

Section S4. Robustness against exercise intensity as a potential confounder
To test if exercise intensity may have influenced our results, we conducted additional analyses. Therefore, we operationalized exercise intensity in two different ways. First, as the mean Movement Acceleration Intensity (MAI; see methods section for details) across all exercise activities when participants wore the accelerometers. Second, to compensate for missing accelerometer data during exercise sessions when participants could not wear the devices (such as swimming), we manually assigned metabolic equivalent (MET) values to all exercise labels that had been reported by the participants after the study week via an advanced day reconstruction method (see methods section and (42)), using the recognized MET compendium by Ainsworth and colleagues (https://sites.google.com/site/compendiumofphysicalactivities/). However, adding exercise intensity as an additional covariate into our main analysis did not change the results (Interaction analyses sgACC*non-exercise activity using the operationalization of exercise intensity via MAI: Beta coefficient = -0.148; P value = 0.001; Interaction analyses sgACC*non-exercise activity using the operationalization of exercise intensity via MET: Beta coefficient = -0.148; P value = 0.001).    (44)) Beta coefficient = -0.148; P value = 0.001 To test whether the socioeconomic status of the participants may have confounded our results, we conducted additional analyses, entering socioeconomic status as quantified by Lampert et al.'s questionnaire (30), and perceived social status derived from the MacAthur scale ((44); see Methods section for details) into our main analysis (either as single predictors or both predictors together). However, adding the additional covariates did not change our findings.