Obesity is one of the leading risk factors of non-communicable diseases worldwide [49]. The prevalence of obesity remains high in developed nations and is increasing worldwide [34]. While some interventions to lose weight have been successful they are rarely successful at allowing people to maintain this weight loss in the long term [9, 23], where maintenance is clinically defined at least 1 year post weight loss [21]. The primary factors that contribute to obesity are unhealthy eating and sedentary behavior [49]. These behaviors are particularly difficult to change in the long-term because they are often enacted habitually [14, 40]. Moreover, weight loss maintenance will not be achieved if an intervention only temporarily interrupts these habits, as old behavioral patterns will resume once the intervention has ceased [13]. Therefore, there is a need to develop weight loss interventions that specifically target and aim to break the unhealthy habits that contribute to the maintenance of obesity.

A habit is an overlearned process that generates automatic responses to contextual cues [13], and is reinforced by repetition [27]. For example, arriving home after work each day and immediately having an unhealthy snack would likely lead to the formation of an unhealthy eating behavior habit such that a person may find themselves approaching the kitchen every time they enter the home regardless of hunger or desire to eat. The role of habit in unhealthy eating behavior and sedentary behavior has been demonstrated previously (for review, see: [12, 46]). Several studies have demonstrated that the more habitual a person’s eating and/or sedentary behavior is, the more likely they are to engage in that behavior [2, 8, 47]. Moreover, engaging in these behaviors leads to weight gain in the long term [32].

The few dietary interventions that have attempted to break habitual unhealthy eating behavior have shown promise in terms of maintenance of change in unhealthy eating or weight loss maintenance [7, 13]. For example, Carels et al. [7] compared the efficacy of a cognitive-based weight loss intervention with a habit-based intervention and found that while both approaches resulted in weight loss, the habit-based intervention resulted in weight-loss maintenance. The habit-based intervention was designed to disrupt unhealthy habits, and to develop new healthy habits. Further, preliminary results indicate that habit-based interventions are also successful at changing sedentary behavior [28]. For example, Matei et al. [28] demonstrated that prompting participants to replace sitting with light-intensity activity, and repeating and monitoring this action, led to a decrease in sedentary behavior and habits, and an increase in walking and activity habits.

In addition to unhealthy habits, poor executive function has been shown to be associated with unhealthy eating behavior [3], physical activity [19], and weight gain [33]. In particular, individuals with obesity have lower levels of executive function than their healthy weight counterparts, even after controlling for medical comorbidities [42]. Executive function refers to a set of higher order cognitive processes that includes planning, inhibiting behavior, cognitive flexibility, working memory and central coherence [5, 31]. Thus, executive function processes are essential to an individual’s ability to modify their behavior (for review, see: [1]). Researchers have demonstrated that improving elements of executive function results in reduction of unhealthy behaviors such as consumption of high caloric foods in healthy weight populations [4]. It might also be possible to change unhealthy habits that prevent weight loss in individuals with obesity by improving executive function. Dual-process theory of self-regulation [43] suggests that automatic, bottom-up processes, such as those that characterise habitual action, can be regulated by executive function to allow for top-down processes, such as goals, to direct behavior [20].

A manualised Cognitive Remediation Therapy (CRT) intervention program for adult obesity has been developed to improve executive function in individuals with obesity, targeting specifically those processes that individuals with obesity have difficulties with [36]. CRT, in particular, promotes reflection on thinking styles, develops metacognition and helps to explore and apply new thinking strategies in everyday life [44]. The primary function of CRT is to improve the thinking process rather than the content [45]. Findings from a recent randomized controlled trial in which adults with obesity underwent CRT for obesity (CRT-O) showed that those in the CRT-O group demonstrated significant improvement in executive function and an 6.6% average weight loss at the 3 months post treatment compared to a no-treatment control group [36, 37].

Given that improving executive function may also have an impact on habitual unhealthy behavior, the aim of the present study was to perform a secondary analysis of the CRT-O trial by Raman et al. [37] in order to determine if CRT-O decreases unhealthy eating and sedentary behavior habits, and if these changes are maintained at follow-up. An additional aim was to determine whether changes in executive function mediate any observed changes in habitual behavior. It is hypothesised that a reduction in both unhealthy eating and sedentary behavior habits will be observed after the intervention period in the CRT-O condition only and that these changes will be maintained at follow-up. Further, it is expected that the reduction in unhealthy habit strength observed in those who received CRT-O will be mediated by improvement in executive function.

There were no differences at baseline between experimental (n = 42) and control (n = 38) conditions on any of the included outcome variables or by age or gender, all ps > .05. Further, there were no differences between those who remained in the study and those who dropped out (n = 17) on any outcome variable, age or gender, all ps > .05. However, participants were more likely to drop-out of the control condition (n = 12) than the experimental condition, n = 5; χ² = 4.615, p = .032.

Wisconsin card sort task

There were main effects of time and condition which were qualified by a significant time by condition interaction, F (2,122) = 26.736, p < .001, Cohen’s d = 1.325. Contrasts examining these effects separately for each time comparison revealed that WCST perseverance errors reduced in the CRT-O condition, MD = − 5.854, SD = 4.216, but not the control group, MD = − 0.226, SD = 4.455, from baseline to post-intervention, F (1,61) = 22.928, p < .001, Cohen’s d = 1.086. Similarly, perseverance errors reduced in the CRT-O condition, MD = − 6.108, SD = 4.532, but not the control group, MD = 1.038, SD = 5.087, from baseline to follow-up, F (1,61) = 34.311, p < .001, Cohen’s d = 1.500, see Fig. 1.

Trail making task

There was a main effect of time which was qualified by a significant time by condition interaction, F(2,116) = 4.788, p = .010, Cohen’s d = 0.574. Contrasts examining these effects separately for each time comparison revealed that TMT performance improved in the CRT-O condition, MD = − 19.673, SD = 35.442, but not in the control group, MD = 0.870, SD = 24.858, from baseline to post-intervention, F (1,58) = 5.842, p = .019, Cohen’s d = 0.637. Similarly, TMT performance improved in the CRT-O condition, MD = − 19.791, SD = 37.945, but not the control group, MD = 0.467, SD = 32.538, from baseline to follow-up, F (1,58) = 4.869, p = .031, Cohen’s d = 0.578, see Fig. 2.

Unhealthy eating habit

There were no significant main effects of time or condition, nor was the interaction between time and condition significant. However, planned contrasts- which analysed these effects separately for baseline to post-intervention and for baseline to follow-up- revealed that habit strength significantly reduced in the CRT-O group, MD = − 1.588, SD = 0.986, compared to the control group, MD = 0.032, SD = 1.021, from baseline to follow-up, F (1,60) = 39.101, p < .001, Cohen’s d = 1.616, but not from baseline to post-intervention, F (1,60) = .083, p = .778, Cohen’s d = 0.063, see Fig. 3.

Sedentary behavior habit

There were main effects of time and condition which were qualified by a significant time by condition interaction, F (2,120) = 30.770, p < .001, Cohen’s d = 1.432. Contrasts examining these effects separately for each time comparison revealed that sedentary habit strength reduced in the CRT-O condition, MD = − 1.412, SD = 0.946, but not in the control group, MD = 0.199, SD = 0.825, from baseline to post-intervention, F (1,60) = 65.326, p < .001, Cohen’s d = 2.086. Similarly, sedentary habit strength reduced in the CRT-O condition, MD = − 1.660, SD = 1.306, but not in the control group, MD = 0.170, SD = 0.591, from baseline to follow-up, F (1,60) = 42.949, p < .001, Cohen’s d = 1.692, see Fig. 4.

Mediation analyses

The indirect effect of condition on change in unhealthy eating habit, through change in WCST perseverance errors¹ was significant, β = 0.396, 95% [CI: 0.119, 0.790]. This indirect effect accounted for 31.56% of variance in the total effect and an additional 8.58% of the variance in unhealthy eating habits. The indirect effect of condition on change in unhealthy eating habit, through change in TMT performance was not significant, β = 0.026, 95% [CI: -0.108, 0.199], and only accounted for 1.62% of the variance in the total effect. The effect of intervention condition on change in habit strength was partially mediated by change in WCST performance, as the direct effect of condition on habit change remained significant once change in WCST was added to the model, β = 0.838, 95% [CI: 451, 1.687]. Overall, the direct and indirect effects accounted for 46.97% of the variance in change in unhealthy eating habits. See Fig. 5 for standardised coefficients between all variables in the model.

The indirect effect of condition on change in sedentary behavior habit, through change in WCST perseverance errors was significant, β = 0.277, 95% [CI: 0.033, 0.626]. This indirect effect accounted for 21.21% of variance in the total effect and an additional 3.91% of the variance in sedentary behavior habits. The indirect effect of condition on change in sedentary behavior habit, through change in TMT performance was not significant, β = − 0.010, 95% [CI: -0.140, 0.193], an only accounted for 0.8% of the variance in the total effect. The effect of intervention condition on change in habit strength was partially mediated by change in TMT performance, as the direct effect of condition on change in habit remained significant once change in TMT was added to the model, β = 1.037, 95% [CI: 0.758, 2.148]. Overall, the direct and indirect effects accounted for 44.95% of the variance in change in sedentary behavior habits. See Fig. 6 for standardised coefficients between all variables in the model.

This is the first study to examine the effects of CRT on habitual behavior, both sedentary activity and unhealthy eating, in individuals with obesity. Following from evidence that CRT improves executive function [26], we hypothesised that this improvement in executive function would provide individuals with the mental flexibility to break unwanted habits, which in turn would account for changes in lifestyle outcomes- specifically weight loss- as documented in a previous study (Raman et al., 2016). From the current analysis we found that CRT-O improved executive function, both WSCT and TMT, compared to the control group, at both post-treatment and three-month follow-up, with large effect sizes. Additionally, those in the CRT-O group also had reduced habitual sedentary behavior at post-treatment and follow-up, but only reduced habitual unhealthy eating at 3 month follow-up, with large effect sizes. Importantly, changes in habitual behavior were mediated by changes in one element of executive function that is, cognitive flexibility measured by performance on the WCST.

These findings are the first to demonstrate that improving executive function has a direct impact on habitual behaviors, building on previous research that has provided support for the effectiveness of habit based interventions in the promotion of weight loss [25]. Specifically, how enhancing self-regulatory skills and automaticity through habit-based intervention can lead to weight loss in adults. While previous research has shown the potential of executive function training to improve health outcomes in non-clinical populations [4], few studies have applied such training to individuals with obesity, or specifically examined the effect on unhealthy habits. Perseverance- as measured by the WCST- appeared to be the most influential element of executive function in changing unhealthy eating habits and sedentary behavior. This outcome is said to reflect cognitive flexibility [6], and has been associated with weight previously where it has been shown that individuals with anorexia nervosa or obesity have more rigid thinking styles and difficulty perceiving alternative ways of coping with problems compared to healthy weight controls [10]. In the current context, it appeared that improving this element of executive function may have allowed participants to break their unhealthy routines.

The TMT is said to measure different elements of executive function compared to the WCST; namely, visual attention and task switching [41]. This may partially explain why improvements on the TMT did not mediate changes in habit strength in the same way that changes in WCST performance did. These measures did not correlate, providing further evidence that these tasks index distinct elements of executive function. It appears that TMT performance is not relevant to interrupting the unhealthy eating and sedentary behavior habits of individuals with obesity. While previous research has demonstrated TMT deficits in individuals with anorexia nervosa compared to healthy weight controls [39], research examining differences between individuals with obesity and healthy weight controls has not consistently found this effect [11]. Thus, future research that examines the role of TMT performance or other measures of task switching and visual attention in individuals with obesity is warranted.

The finding that changes in executive function mediated changes in unhealthy habits provides a mechanistic explanation for how habits are broken (or at least reduced).

Habits are primarily interrupted when the behavior becomes impossible (i.e., unhealthy menu option removed) or the context in which the cue to act changes or the value of the outcome changes [22]. One interpretation of the results is that individuals with greater cognitive flexibility may be able to use these interruptions to adjust their behavior in line with their weight loss goals rather than persisting with an unhealthy choice. Interestingly, habit strength for unhealthy eating habits did not improve from baseline to post-intervention, while sedentary habits did. The greater lag between intervention and effect for this outcome may indicate that unhealthy eating habits are more difficult to break. The results also reflect a qualitative difference between the two behaviors- stopping an unhealthy eating habit versus engaging in more physical activity. Previous research has demonstrated a distinction between health behaviors that require the inhibition or initiation of a response in terms of associated processes [29, 35].

CRT-O appeared to improve specific elements of executive function. Changes in executive function mediated changes in unhealthy eating habits and sedentary behavior habits. These results are important to consider when implementing behavioral weight loss programs to individuals with obesity. Specifically, interventions that include components that train executive function may be particularly useful as improvements in executive function relate to reduced unhealthy eating and sedentary behavior habits, which have flow on effects to weight loss. Furthermore, these results have implications for habit theory suggesting that cognitive flexibility is a key mechanism in the reduction of unwanted habits. Moreover, cognitive flexibility may broadly reflect individual differences in the ability to break habitual behavior.