A Pilot Randomized Control Trial Testing a Smartphone-Delivered Food Attention Retraining Program in Adolescent Girls with Overweight or Obesity

Abstract

Background/objectives: Attention bias (AB) toward food is associated with obesity, but it is unclear if programs designed to reduce AB can impact adolescents' eating behavior. We investigated whether a two-week, smartphone-delivered attention retraining (AR) program (vs a control program) altered food AB in adolescent girls with overweight.

Methods: Participants completed three food-cue visual-probe trainings/day. The AR and control programs directed attention away from food stimuli during 100% and 50% of trainings, respectively. Before and after completion of the programs, girls completed a food-cue visual-probe task while undergoing magnetoencephalography (MEG), and then a laboratory test meal.

Results: Sixty-eight adolescents were randomized; 58 completed post-program visits. There was minimal effect of condition on AB scores (β [95%CI] = -1.9 [-20.8, 16.9]; d = -0.06). There was a small effect of condition on energy intake (EMM_control = 1017 kcal, EMM_AR = 1088 kcal, d = 0.29). Within the AR group, there was slightly blunted initial engagement in brain areas associated with reward response and subsequent increased goal-directed attention and action control.

Conclusions: We found preliminary support for efficacy of an intensive smartphone-delivered AR program to alter neural correlates of attention processing in adolescent girls with overweight or obesity. Studies with larger sample sizes are needed to elucidate if AR trainings disrupt the link between food AB and eating behavior.

Keywords: adolescence; attention bias; attention retraining; obesity; smartphone program.

Figure 1

Consort diagram of participant retention.

Figure 2

Effects of (A) condition and (B) condition by LOC-eating on change in $∆$_{AB score}. Note: LOC-eating; loss-of-control eating. Estimated marginal means (EMM) and 95% confidence intervals presented for the total sample (EMM_AR 1.85 [−8.44, 12.13]; EMM_control 4.27 [−6.25, 14.80]), AR group (EMM _LOC-eating 0.02 [−12.13, 12.18]; EMM _{no LOC-eating} 3.67 [−16.18, 23.52]), and Control group (EMM _LOC-eating 1.97 [−11.02, 14.97], EMM _{no LOC-eating} 6.57 [−11.81, 24.95]). The presented estimated marginal means for the effect of condition are from the models adjusted for LOC-eating and the LOC-eating × condition interaction term.

Figure 3

$∆$_{oscillatory power} in Surface ROIs During Attention Capture (0–250 ms). Note. Estimated marginal means of change (post intervention − pre intervention) in beta band oscillatory power are presented for all a priori identified ROIs. To obtain oscillatory power estimates, we log transformed ratios (pseudo-Z oscillatory power in congruent trials/pseudo-Z oscillatory power in incongruent trials); therefore, estimated marginal means are unitless. The presented estimated marginal means for the effect of condition are from the models adjusted for LOC-eating and the LOC-eating × condition interaction term.

Figure 4

$∆$_{oscillatory power} in Surface ROIs During Attention Deployment (250–500). Note: Estimated marginal means of change (post intervention − pre intervention) in beta band oscillatory power are presented for all a priori identified ROIs. To obtain oscillatory power estimates, we log transformed ratios (pseudo-Z oscillatory power in congruent trials/pseudo-Z oscillatory power in incongruent trials); therefore, estimated marginal means are unitless. The presented estimated marginal means for the effect of condition are from the models adjusted for LOC-eating and the LOC-eating × condition interaction term.