Erasmus MC heeft ervoor gezorgd dat je Mijn BSL eenvoudig en snel kunt raadplegen. Je kunt je links eenvoudig registreren. Met deze gegevens kun je thuis, of waar ook ter wereld toegang krijgen tot Mijn BSL.
Om ook buiten de locaties van Erasmus MC, thuis bijvoorbeeld, van Mijn BSL gebruik te kunnen maken, moet je jezelf eenmalig registreren. Dit kan alleen vanaf een computer op een van de locaties van Erasmus MC.
Eenmaal geregistreerd kun je thuis of waar ook ter wereld onbeperkt toegang krijgen tot Mijn BSL.
Login
Als u al geregistreerd bent, hoeft u alleen maar in te loggen om onbeperkt toegang te krijgen tot Mijn BSL.
In-person mindfulness-based interventions (MBIs) have been shown to decrease symptoms of anxiety and stress in autistic adults, who often report high levels of these symptoms. Little is known about the effectiveness of remote MBIs for this population, which may be particularly useful given the common barriers autistic adults face in accessing in-person treatment. This study examined the feasibility and effectiveness of an app-based mindfulness intervention for autistic adults.
Method
This randomized controlled trial (RCT) examined whether a 6-week remote intervention, using a customized version of the Healthy Minds Program app, reduced symptoms of anxiety and perceived stress in 89 autistic adults. Participants were randomly assigned to either the mindfulness intervention or a wait-list control (WLC) group. The WLC group received the intervention after the RCT. Self-report measures of anxiety, perceived stress, positive and negative affect, and trait mindfulness were administered at several timepoints.
Results
The mindfulness group showed significant decreases in anxiety symptoms and perceived stress relative to the control group, with medium to large between-groups effect sizes (ηp2 0.07 to 0.14). These benefits, as well as significant decreases in negative affect and increases in trait mindfulness, were replicated when the WLC group subsequently received the intervention, and were retained in both groups 6 weeks after conclusion of the intervention.
Conclusions
Results demonstrate both the feasibility and effectiveness of a remote mindfulness self-guided intervention for reducing perceived stress and anxiety symptoms in autistic adults. Future research can investigate the specific processes of how such an intervention exerts its effects.
Lindsay Bungert is now at The Donald and Barbara Zucker School of Medicine, Hofstra University, Hempstead, NY.
John D. E. Gabrieli and Liron Rozenkrantz are dual senior authors.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Autism is a neurodevelopmental condition characterized by social communication difficulties and repetitive behaviors and/or restricted interests. Approximately 1% of individuals worldwide (Baxter et al., 2015; Zeidan et al., 2022) and 1 in 36 children in the USA (Maenner et al., 2023) are diagnosed with autism. Autistic individuals often have at least one co-occurring mental health condition, of which anxiety is among the most common (Hollocks et al., 2019). It is estimated that between 20 and 65% of autistic adults have one or more anxiety disorders (Fombonne et al., 2020; Jadav & Bal, 2022; Nimmo-Smith et al., 2020; Underwood et al., 2023). Further, self-reported stress is higher in autistic adults than in the general population, and this heightened stress is associated with less independence in daily living and worse subjective quality of life (McQuaid et al., 2022). Thus, strategies to help manage symptoms of anxiety and stress may be particularly important for autistic individuals.
Mindfulness has been described as intentionally paying attention to the present moment without judgement (Creswell, 2017; Kabat-Zinn, 2003). Mindfulness-based interventions (MBIs) are designed to cultivate increased mindfulness and have been recognized to benefit health and well-being. Meta-analyses have shown that MBIs are associated with moderate to large decreases in stress and anxiety in nonclinical populations (Khoury et al., 2015; Querstret et al., 2020). A systematic review of 44 meta-analyses found that MBIs were effective in reducing stress in a variety of populations, and were superior to active controls (e.g., psychoeducation) but not to “specific active controls” (e.g., therapies specifically known to reduce stress and anxiety) (Goldberg et al., 2022). Additionally, an MBI with a nonclinical population showed decreased perceived stress compared to both the no-treatment and active-control groups (Dark-Freudeman et al., 2022). In a meta-analysis, MBIs were found to reduce anxiety symptoms in a population with anxiety disorders, when compared to controls, including both wait-list control (WLC) and active-control groups. Moreover, the MBIs were as effective as cognitive behavioral therapy (CBT) in reducing anxiety symptoms (de Abreu Costa et al., 2019). Further, a randomized controlled trial (RCT) found that a common mindfulness intervention, Mindfulness-Based Stress Reduction, was as effective as medication in the treatment of anxiety disorders (Hoge et al., 2023).
There is evidence for positive impacts of in-person MBIs for autistic adults. Feasibility studies of in-person group interventions reported improvements in positive outlook, with trends toward improvement in quality of life and trait mindfulness (Beck et al., 2020), and decreases in symptoms of anxiety, depression, and perceived stress (Agius et al., 2024). A 6-week mantra-like recitation intervention resulted in improvements in symptoms of anxiety, stress, depression, rumination, and emotion regulation (Corney et al., 2023). Further, a 9-week in-person group-based MBI with autistic adults resulted in reduced symptoms of anxiety and depression, and increased positive affect, with benefits maintained at the 9-week-follow-up (Kiep et al., 2015).
RCTs have also shown benefits of in-person mindfulness interventions for autistic individuals. One study reported significant decreases in symptoms of depression, anxiety, and rumination, and increases in positive affect, in autistic adults who were randomized to an MBI versus a WLC group (Spek et al., 2013). Another RCT of autistic adults examined the effects of an MBI versus an active control (relaxation and social support psychoeducation) on quality of life (Braden et al., 2022). Both groups experienced improvement in quality of life but gains were larger in the MBI group. Another study found that an MBI demonstrated benefits equivalent to CBT in reducing symptoms of anxiety and depression in autistic adults, and these benefits were sustained for 3 months after treatment ended (Sizoo & Kuiper, 2017). There was also decreased rumination and negative affect, and increased positive affect (Sizoo & Kuiper, 2017). Yet another study found that an MBI and an active control (social support/education) both led to decreases in symptoms of trait anxiety and depression in autistic adults (Pagni et al., 2023). The MBI group alone showed improvements in mindfulness traits (specifically in nonjudgment), as well as executive functioning (Pagni et al., 2023). Together, these findings provide evidence that in-person MBIs may decrease anxiety symptoms and improve well-being in autistic individuals, although not always above and beyond that of active control groups. It should be noted that several reviews indicate that results are limited and should be interpreted with caution due to the small number of methodologically sound studies (Benevides et al., 2020; Loftus et al., 2023).
With the spread of digital media, MBIs have become more available and convenient (Plaza et al., 2013). Delivered through online smartphone apps for mindfulness training, remote MBIs enable participants to practice short meditation sessions in the comfort of their chosen location and time of day that best suits them. Such flexibility and accessibility allow for the potential to scale the beneficial effects of traditional MBIs. Similar to in-person MBIs, evidence suggests that digital MBIs are effective in decreasing stress (Kirk et al., 2023) and in reducing anxiety symptoms both in nonclinical populations (Economides et al., 2018) and in adults diagnosed with a variety of anxiety disorders (Boettcher et al., 2014). In one study where participants were randomized to either a 3-week online group-based MBI, a 3-week online group-based CBT condition, or a WLC group, participants receiving the MBI showed significant improvement in symptoms of anxiety, depression, perceived stress, and insomnia, as compared to the WLC group, with no significant differences in outcomes compared to the CBT group (Wang et al., 2023). Further, these findings persisted at the 6-month follow-up. Additionally, several meta-analyses of online RCTs found that MBIs compared to control groups were associated with decreases in symptoms of anxiety and stress (Chen et al., 2023; Reangsing et al., 2023; Spijkerman et al., 2016), further supporting the promise of online mindfulness in improving mental health.
There is some evidence that remote MBIs reduce anxiety symptoms and perceived stress in autistic adults. A feasibility study of a 6-week virtual group-based mindfulness curriculum found that autistic adults reported decreased levels of stress and anxiety symptoms (maintained at 3 months post-intervention) but no change in depression symptoms (Lunsky et al., 2022). Participants also experienced increases in mindfulness and self-compassion, most notably at follow-up. Further, increases in trait mindfulness correlated with decreases in both depression and anxiety symptoms but not stress. This contrasts with the findings of another feasibility study based on the same remote 6-week mindfulness group-based intervention but with a new sample consisting solely of autistic women, reporting no statistically significant gains in outcome measures (Redquest et al., 2022). Additionally, a 5-week online training program for autistic adults, which focused on self-compassion (a component of mindfulness), showed improvements in self-compassion, symptoms of anxiety and depression, positive and negative affect, emotion regulation difficulties, and psychological well-being (Cai et al., 2024). An RCT examined the effectiveness of a mindfulness app in reducing depression symptoms in autistic adults, with a particular interest in investigating the impact of habit training, where participants were instructed to pair their app usage with a daily morning routine (Stecher et al., 2024). The App + Habit group reported greater daily app usage than either the WLC group or the App only group, both during the 8-week intervention and the 8-week period following the intervention when app usage was also assessed. Further, the App + Habit group showed the greatest decrease in depressive symptoms from baseline, at both the 8-week and 6-month follow-ups (Stecher et al., 2024).
The only remote mindfulness-based RCT examining symptoms of anxiety in autistic adults, to our knowledge, was a pilot study comparing an online self-guided MBI program (n = 14 completing the study), an online CBT self-guided intervention, and a WLC group (Gaigg et al., 2020). Individuals in both intervention groups were expected to complete the program within 6 to 8 weeks. Participants in both treatment conditions exhibited significant reductions in anxiety symptoms post-intervention and at the 3-month follow-up as compared to the control group, although these gains dissipated by the 6-month follow-up (Gaigg et al., 2020). No significant improvements were found in depression symptoms, overall mental health and well-being, intolerance to uncertainty, alexithymia, or emotional acceptance, as compared to the control group. Limitations of this study include a small sample size and that groups were not matched at baseline on degree of anxiety symptoms. In fact, the MBI group had lower levels of anxiety symptoms at baseline, suggesting that the effects of the MBI group may have been stronger had this not been the case. Additionally, this study was not entirely remote, as several assessments were conducted in-person. Finally, no objective data were obtained that tracked time spent on the programs.
Thus, emerging evidence demonstrates that in-person MBIs appear beneficial for autistic adults and that remote MBIs may have similar benefits. Online MBIs are a scalable, cost-effective, convenient, and highly accessible tool that could reach many autistic individuals who might not otherwise seek or be able to access treatment. This is particularly important because a common treatment barrier for autistic adults is accessibility to effective and affordable mental health services (Maddox & Gaus, 2018). Additional research is needed to determine the impact of remote mindfulness interventions on autistic adults.
The primary objective of this study was to determine the efficacy of a 6-week smartphone mindfulness intervention in reducing anxiety symptoms and perceived stress in autistic adults who were randomly assigned to either a WLC group (also referred to as “control group”) or an intervention group. We aimed to examine improvements in anxiety symptoms, perceived stress, and other secondary outcome measures, in the intervention group relative to the control group. Next, we aimed to utilize the WLC group as a replication sample, investigating pre-to-post-intervention changes after they received the intervention. We also aimed to assess whether any changes in outcome measures would be sustained at the 6-week follow-up. Finally, we aimed to assess the feasibility of this remote intervention by examining percentage completion of the intervention, as well as assessment completion rates, and qualitative feedback. Based upon existing literature, we hypothesized that the intervention group would show significantly greater decreases in anxiety symptoms and perceived stress, increased positive affect, decreased negative affect, and increased trait mindfulness, as compared to the control group. We also expected to replicate these benefits when the WLC group received the intervention after the RCT phase of the project, and expected these improvements to be maintained at the 6-week follow-up after the intervention was completed.
Method
Participants
Individuals were recruited from January through November 2022 from two databases: (1) the Autism Research Participant Database at the Massachusetts Institute of Technology, which consists of over 150 adults recruited from the community with an existing clinical diagnosis of autism and who have met criteria on the Autism Diagnostic Observation Schedule (Lord et al., 2000); and (2) Simons Foundation Powering Autism Research for Knowledge (SPARK), which consists of over 20,000 adults with a professional autism diagnosis (SPARK Consortium, 2018) and is confirmed to have a high degree of autism diagnosis validity (Fombonne et al., 2022). We did not advertise that the study was specifically targeting anxiety symptoms and perceived stress but instead stated that we were interested in the effects of mindfulness on well-being. Informed written consent was obtained from all individuals before participation in the study.
Eligible participants were ≥ 18 years, reported a professional diagnosis of autism, and had daily access to a smartphone. Exclusion criteria consisted of the following: non-corrected hearing impairment that impacts one’s ability to listen to the app, far below average nonverbal intellectual ability, defined as approximately ≥ 2 SDs below the mean on the Test My Brain, matrix reasoning subtest (see below for details), listening comprehension skills that may compromise one’s ability to understand the app content, defined as below 60% correct on the Listening Comprehension screener (see below for details), and considerable meditation experience (Online Resource).
Of 173 autistic adults who were invited to participate in the study, 117 completed the screening measures. Of those, 91 individuals met eligibility criteria and 89 completed baseline measures and were randomly assigned to either the intervention group, n = 45, or the WLC, n = 44 (Fig. 1). The intervention lasted for 6 weeks. The control group received the same intervention after their 6-week wait. The sample of 89 individuals who completed baseline measures consisted of adults who were primarily assigned female at birth (55.06%), cisgender (85.39%), and White (82.02%), with a mean age of 35.77 years (SD = 10.69). Of the 89 individuals enrolled in our study, at baseline 72% scored in the high range (defined below) on the measure of trait anxiety. However, since we did not administer diagnostic clinical interviews to assess for clinically significant anxiety disorders, these individuals cannot be considered a clinical population in regard to anxiety.
Fig. 1
Consolidated Standards of Reporting Trials (CONSORT) diagram. For WLC: pre-test, after 6-week wait; post-test, at end of intervention; follow-up, 6 weeks after intervention ended. For the intervention group: post-test, at end of intervention; follow-up, 6 weeks after intervention ended
×
Procedure
Participants completed the entire study remotely. The screening surveys were completed on Qualtrics (https://www.qualtrics.com) and all other surveys were completed on Research Electronic Data Capture (REDCap). REDCap is a secure web-based application designed to support data capture for research studies (Harris et al., 2009). Participants were able to complete all measures on their smartphone or computer (laptop, desktop, or iPad). Participants received automated emails and reminders from REDCap instructing them to complete the surveys. To decrease the possibility of missing data, all questions from the standardized outcome measures were “required” and all surveys in the set needed to be completed in one sitting.
Eligible individuals were administered baseline measures and then randomly assigned to either the intervention group or control group through block randomization, a pseudo-randomization process designed to address potential baseline differences between groups for the main outcome measure (anxiety). We pseudo-randomized participants in blocks based on their low, medium, or high baseline trait anxiety scores to ensure similar representation of trait anxiety scores between groups. On a rolling basis, each participant was alternately assigned to the intervention group or the control group. The first block of participants with low, medium, and high trait anxiety scores was assigned to the intervention group, the second block was assigned to the control group, and so on. The cutoffs for the trait anxiety scores from the State-Trait Anxiety Inventory for Adults (STAI) were based on prior studies that typically consider STAI scores to be high at 45 and above (Kindler et al., 2000; Millar et al., 1995) and have categorized scores into “no or low anxiety” (20–37), “moderate anxiety” (38–44), and “high anxiety” (45–80) (Kayikcioglu et al., 2017). Neither researchers nor participants were blinded to group assignment.
Self-reported levels of anxiety, perceived stress, positive and negative emotions, and trait mindfulness were measured at four time points throughout the study: before the start of the intervention (participants were given directions on how to download the app after they filled out the pre-intervention assessments, see below), halfway through the 6-week intervention, at the end of the 6-week intervention, and at the 6-week follow-up. The control group was assessed at two additional time points: before they began their waiting period and halfway through their 6-week waiting period.
Several measures were taken to ensure that participants understood the study procedures and experienced as few difficulties as possible in their participation (Online Resource). Additionally, emotional discomfort that participants attributed to the practices was tracked on a Weekly Check-In form. A protocol was established to contact individuals who expressed a high degree of discomfort and to determine if they should stop participating (Online Resource, Optional Questionnaires).
All participants received a list of mental health resources at the start of the study. Participants were compensated for their time, as follows: $15 for completion of the screening surveys, $20 for each subsequent set of surveys, with the last three timepoints including increasing bonuses of $5, $15, and $25, respectively. Individuals could cease participation at any time and were compensated for the portion of the study they completed.
The intervention consisted of a customized version of the Healthy Minds Program (HMP) smartphone app (Healthy Minds Innovations, 2021) which focuses on four pillars of well-being: awareness, connection, insight, and purpose. It included 6 weeks of alternating didactic “lessons” and guided mindfulness “practices.” Each lesson provided scientific background for the practices and lasted up to 7 min. Practices lasted either 10 or 15 min, as chosen by the participant. A total of 29 lessons and practices constituted the structured curriculum. Other optional practices were included on the app as well. Participants were instructed to listen to at least one activity per day, 5 days per week, for 6 weeks. It was suggested, but not required, that they listen to the activities in order. They were also told that they could relisten to activities as desired and listen to the optional practices as they saw fit. For each mindfulness practice, participants chose between the type (sitting or active), the speaker (a male or female narrator), and the duration (10 or 15 min). (See Online Resource for curriculum content and description of app customization.) Participants were given free access to the app for an unlimited time.
We assessed participants’ app usage through the app dashboard, which tracked the number of minutes participants listened to each lesson and practice activity as well as the options that they chose for each mindfulness practice. For the primary analyses, we included individuals who used the app at least once (n = 79). However, we performed intent-to-treat analyses as well, wherein we included those participants who did not use the app at all (n = 81).
Measures
See Online Resource for reliability details of the measures, as reported in the literature.
Screening Measures
The Test My Brain matrix reasoning subtest (referred to here as TMB) is an online validated and normed proxy for nonverbal intelligence (Germine et al., 2012; Singh et al., 2021). For each item, participants are asked to choose the image that best completes a given pattern. The threshold for exclusion in this study was approximately ≥ 2 SDs below the normed mean (for mobile phone administration). In this study, the McDonald’s omega reliability coefficient for this measure was 0.90. The Listening Comprehension screener consists of 20 multiple-choice questions we created based on language and concepts from the HMP app (Healthy Minds Innovations, 2021). These include questions about scenarios presented auditorily, without accompanying text. Exclusion was defined as achieving below 60% correct. In this study, McDonald’s omega was 0.94 for this measure. For both measures, higher scores indicate better performance.
Measures of Anxiety Symptoms and Perceived Stress
To assess anxiety symptoms, we administered the STAI, a 40-item self-report questionnaire assessing both state anxiety (20 items) and trait anxiety (20 items) (Spielberger et al., 1983). Responses for both subscales are rated on a 4-point Likert scale from not at all to very much so for the state responses and almost never to almost always for the trait responses. The State Anxiety subscale measures how one is currently feeling, whereas the Trait Anxiety subscale assesses how one is feeling in general. Scores on each subscale range from 20 to 80. Higher scores are indicative of a greater level of anxiety. In this study, the reliability estimates as measured by McDonald’s omega were 0.95 for the Trait Anxiety subscale and 0.97 for the State Anxiety subscale. The alpha coefficients were 0.94 for the Trait Anxiety subscale and 0.96 for the State Anxiety subscale. We also measured symptoms of anxiety with the Patient-Reported Outcomes Measurement Information System (PROMIS) Anxiety-Short Form (v.10, 8a). This is an eight-item self-report measuring anxiety over the past week on a 5-point Likert scale from never to always, with higher scores indicating greater anxiety (Cella et al., 2010). We used the raw scores, which range from 8 to 40 and can be converted to T-scores. In this study, McDonald’s omega was 0.94 and the alpha coefficient was 0.92.
To measure perceived stress, we administered the Perceived Stress Scale (PSS), which is a 10-item self-report measure of perceived stress over the past month (Cohen & Williamson, 1988). Response options are based on a 5-point Likert scale from never to very often. Scores range from 0 to 40, with higher scores signifying a greater perception of stress. In this study, McDonald’s omega was 0.93 and the alpha coefficient was 0.92.
Measure of Positive and Negative Affect
The Positive and Negative Affect Schedule (PANAS) is a commonly used measure of positive and negative affect in mindfulness studies (Roemer & Medvedev, 2023). In a review of the impact of mindfulness training on negative affect, 29 of 65 studies used the PANAS (Schumer et al., 2018). The PANAS, Short Form, is a 20-item self-report of 10 positive and 10 negative emotions experienced over the past week, rated on a 5-point Likert scale from very slightly or not at all to extremely (Watson et al., 1988). There are Positive and Negative Affect scales, with scores on each ranging from 10 to 50. Higher scores on the Positive Affect scale indicate a greater degree of positive emotions, whereas higher scores on the Negative Affect scale indicate a greater degree of negative emotions. In this study, the reliability estimates as measured by McDonald’s omega were 0.93 for the Positive Affect scale and 0.93 for the Negative Affect scale. The alpha coefficients were 0.91 for the Positive Affect scale and 0.89 for the Negative Affect scale.
Mindfulness Measures
To measure trait mindfulness, we used the Mindfulness Attention Awareness Scale (MAAS), a 15-item self-report, with questions assessing mindfulness in daily activities, emotions, and sensations (Brown & Ryan, 2003). Items are rated on a 6-point Likert scale from almost always to almost never. Raw scores range from 15 to 90, with higher scores indicating a greater degree of mindfulness. In this study, McDonald’s omega was 0.90 and the alpha coefficient was 0.87. We also measured trait mindfulness with The Five Factor Mindfulness Questionnaire-Short Form (FFMQ-15). This is a 15-item self-report measure of five aspects of trait mindfulness, rated on a 5-point Likert scale from never or very rarely true to very often or always true (Baer et al., 2006). The five facets are as follows: Observing, Describing, Acting with Awareness, Non-Judging of Inner Experience, and Non-Reactivity to Inner Experience. The total raw score ranges from 15 to 75. Higher scores indicate greater levels of trait mindfulness. Research suggests excluding the Observing subscale due to instability in the five-factor structure with that subscale (Baer et al., 2006). Therefore, we omitted the Observing subscale from the total raw score and from exploratory analyses of the facets. In this study, McDonald’s omega for the total score was 0.86 and the McDonald’s omega values for the facet subscales ranged from 0.74 to 0.85. The alpha coefficient for the total score was 0.77 and the alpha coefficients for the facet subscales ranged from 0.72 to 0.81.
See Online Resource for measures of autistic traits as well as the optional questionnaires.
Data Analyses
We conducted statistical analyses using STATA software (Statacorp LLC, 2019) and R (R Core Team, 2022). To determine whether groups were equivalent on demographic variables and baseline characteristics, we conducted chi-squared tests on sex, gender diversity, and race, performed two-tailed t-tests on the Social Responsiveness Scale and the TMB (n = 79 sample), and performed Mann–Whitney U tests on age, TMB (n = 89 sample), Listening Comprehension, the Autism Quotient, and the Comprehensive Autistic Trait Inventory, which were all nonnormally distributed. To determine if the outcome measures differed by group at baseline, we conducted two-tailed t-tests or Mann–Whitney U tests (on nonnormally distributed variables). Further, we conducted chi-squared tests to assess whether there were different post-test and follow-up attrition rates between groups. To assess app utilization, we conducted Mann–Whitney U tests on various measures of utilization, all nonnormally distributed.
We conducted two-way, repeated-measures analyses of variance (ANOVA) to determine the effect of group (intervention and control) and time (pre- and post-intervention) on the outcome measures. Variables which were nonnormally distributed (state anxiety and negative affect) were log transformed to correct the nonnormality. One variable (anxiety in the past week) was square root transformed to correct issues with heterogeneity of variance. Greenhouse–Geisser corrections are reported due to sphericity not being assumed. For post hoc analyses of significant interaction effects (via ANOVA), we conducted one-tailed paired t-tests by group on pre-to-post comparisons, to determine differences over time. In addition, we conducted multiple linear regression analyses to control for one variable that differed between groups at baseline (TMB), to determine whether change scores on outcome measures could still be predicted from group.
The control group received the mindfulness intervention after their 6-week waiting period. To determine the effect of the intervention on the control group, we conducted one-tailed paired t-tests on their pre-to-post outcome measures. We also performed analyses combining the two groups to maximize statistical power and further probe the replicability of the RCT finding. We conducted one-way repeated-measures ANOVAs to analyze change over time in outcome measures across pre, post, and follow-up timepoints. We report Greenhouse–Geisser corrected p-values due to not assuming sphericity. We followed up significant findings with pairwise comparisons for pre-to-post, pre-to-follow-up, and post-to-follow-up. Friedman’s test was used for variables that demonstrated nonnormal distributions. Post hoc tests of nonnormal variables were conducted using Wilcoxon signed-rank tests.
To confirm the repeated measures findings for the primary sample, we performed general linear modeling (GLM) in the form of linear regression on the repeated time series, using panel data. We conducted random effects panel regressions, reporting robust values to account for any heteroskedasticity. Additionally, to determine whether changes in the outcome measures may have been due to increased trait mindfulness, we assessed the relationship between participants’ pre-to-post change in outcome measures and their pre-to-post change in mindfulness scores. We calculated Spearman’s correlations due to nonnormally distributed change scores. We corrected for multiple comparisons of similar measures (anxiety and mindfulness), timepoints (pre, post, follow-up), and multiple correlations, using false discover rate (FDR) corrections. FDR p-values are reported. For effect sizes, we calculated Cohen’s d or Hedges’ g (for sample sizes under 50).
Using G*Power (Faul et al., 2009), we conducted a post hoc power analysis of the between-groups ANOVA of the primary sample, which revealed that we were powered at 80% to detect medium or larger effect sizes of 0.37 or higher (equivalent to F-values of 2.64) at a significance level of p = 0.05. We assumed a test–retest correlation of 0.50. See Online Resource Table S1 for sample size details.
Results
Baseline Characteristics, Attrition, and App Utilization between Groups
A total of 64 out of the 89 individuals (71.91%) scored in the high range on baseline trait anxiety (control group M = 52.73, SD = 13.20, intervention group M = 53.27, SD = 11.31). Groups were equivalent on all demographic variables. Groups differed only on TMB (Table 1) with the control group having significantly higher TMB scores (z = − 2.35, p = 0.02; control group M = 29.18, SD = 4.28, intervention group M = 27.31, SD = 3.79). Groups did not differ significantly on any outcome measures at baseline (Table 1). The same findings regarding comparisons at baseline held for the sample of 79 individuals included in the primary analyses (Online Resource, Table S2). Additionally, we compared rates of attrition and app utilization between the intervention group and the control group (after the control group participated in the intervention). There was no significant difference in attrition at post-test or follow-up (post-test attrition: intervention group, 13.33%, control group, 18.18%, X2(1) = 0.39, p = 0.53; follow-up attrition: intervention group, 20.0%, control group, 25.0%, X2(1) = 0.32, p = 0.57). There were no significant differences between groups in the total percent of the required curriculum completed, the number of days practiced, the total number of minutes of the required curriculum practiced, the total number of minutes of the required lessons practiced, the total number of minutes of the required practices practiced, or the total number of minutes practiced (Online Resource Table S3). Also, see Online Resource Table S3 for the percentages of participants in both groups who completed at least 50%, 75%, and 95% of the required curriculum. Additionally, there were no sex differences in the combined group’s outcome pre-to-post change scores.
Table 1
Demographic and baseline outcome measure comparisons between groups
Demographic variable
(n = 89)
Intervention group
(n = 45)
% or M (SD)
WLC (n = 44)
% or M (SD)
X2(df)
z or t
p
Sex
Assigned male at birth
40.0
50.0
X2(1) = 0.90
0.34
Assigned female at birth
60.0
50.0
Gender diversity
Cisgender
82.22
88.64
X2(1) = 0.73
0.39
Gender diverse
17.78
11.36
Race
White
77.78
86.36
X2(4) = 2.40
0.66
Asian
11.11
4.55
Black, African American, or African
2.22
0.00
Latina, Latino, or Hispanic
2.22
2.27
More than one race
6.67
6.82
Age
34.60 (11.21)
36.97 (10.12)
z = − 1.44
0.15
TMB
27.31 (3.79)
29.18 (4.28)
z = − 2.35
0.02
LC
18.78 (1.24)
18.98 (1.55)
z = − 1.31
0.19
AQ
35.11 (8.23)
35.20 (8.25)
z = − 0.18
0.86
CATI
154.98 (22.97)
162.57 (24.29)
z = − 1.79
0.07
SRS
102.22 (29.95)
104.61 (26.18)
t = -0.40
0.69
Outcome measure
State anxiety
46.42 (12.27)
43.98 (15.91)
z = 1.23
0.22
Trait anxiety
53.27 (11.31)
52.73 (13.20)
t = 0.21
0.84
Anx. past week
23.33 (6.71)
23.20 (7.23)
t = 0.09
0.93
Perceived stress
22.11 (7.32)
21.52 (7.55)
t = 0.37
0.71
Positive affect
27.71 (7.64)
28.86 (9.91)
t = -0.62
0.54
Negative affect
24.31 (8.03)
24.11 (8.83)
z = 0.50
0.62
MAAS
50.18 (13.55)
48.64 (11.84)
t = 0.57
0.57
FFMQ
34.84 (6.13)
33.30 (7.70)
t = 1.05
0.30
z, z statistic for Mann–Whitney U; TMB, Test My Brain matrix reasoning subtest; LC, Listening Comprehension screener; AQ, Autism Quotient; CATI, Comprehensive Autistic Trait Inventory; SRS, Social Responsiveness Scale; State anxiety, State-Trait Anxiety Inventory for Adults- State Anxiety subscale; Trait anxiety, State-Trait Anxiety Inventory for Adults- Trait Anxiety subscale; Anx. past week, anxiety in past week, from Patient-Reported Outcomes Measurement Information System Anxiety-Short Form; Perceived stress, Perceived Stress Scale; Positive affect, Positive and Negative Affect Schedule, Short Form- Positive Affect subscale; Negative affect, Positive and Negative Affect Schedule, Short Form- Negative Affect subscale; MAAS, Mindfulness Attention Awareness Scale; FFMQ, Five Factor Mindfulness Questionnaire-Short Form; Bolded values indicate p < 0.05
Feasibility
Over 72% of adults in both groups completed at least 75% of the required curriculum. Also, 80% of individuals in the intervention group and 75% in the control group completed the follow-up assessments. All 71 participants who responded to the Post-Intervention Feedback Questionnaire reported that the app was very (73%) or moderately (27%) easy to use. Most participants (65%) reported that it was not at all or slightly difficult to find time to use the app. Additionally, 65% of participants reported that the app was very (31%) or moderately (34%) helpful for them. Only 3% of participants reported experiencing significant emotional discomfort from the app-based practices, 16% reported moderate emotional discomfort (Online Resource, Table S4), and none reported any adverse events or indicated that they chose to stop the intervention due to emotional discomfort.
Mindfulness Training Resulted in Significant Improvement for All Outcome Measures
Between subjects, repeated-measures ANOVAs revealed statistically significant group (intervention group, control group) by time (baseline, post-intervention for intervention group, 6 weeks post-baseline for control group) interaction effects for each of the outcome variables, with medium to large effect sizes: state anxiety, F(1,77) = 11.31, p < 0.01, ηp2 = 0.13; trait anxiety, F(1,77) = 7.94, p = 0.01, ηp2 = 0.09; anxiety in the past week, F(1,77) = 12.50, p = < 0.01, ηp2 = 0.14; perceived stress, F(1,77) = 7.30, p = 0.01, ηp2 = 0.09; positive affect, F(1,77) = 6.15, p = 0.02, ηp2 = 0.07; negative affect, F(1,77) = 10.95, p < 0.01, ηp2 = 0.12, MAAS mindfulness, F(1,77) = 5.75, p = 0.02, ηp2 = 0.07, and FFMQ mindfulness, F(1,77) = 7.12, p = 0.02, ηp2 = 0.08. (Main effects are reported in Table S5.) GLM analysis confirmed all of these findings. We conducted random effects regressions with group (intervention, control) and time (baseline and post-intervention for intervention group, and baseline and 6 weeks post-baseline for control group) as predictors for each of the outcome measures. Group-by-time interactions revealed that the intervention group showed significantly greater improvements on all outcome measures compared to the control group (Online Resource, Table S6).
Post hoc one-tailed paired t-tests by group, and accompanying effect sizes, revealed that the intervention group showed significant pre-to-post-test improvement on each of the outcome measures, whereas the WLC group did not improve significantly on any measure (Fig. 2, Table 2). Table 2 also shows the effect size differences between the intervention group and control group, which for most outcome measures are at least half of a standard deviation larger for the intervention group compared to the control group. See Online Resource and Table S7 for exploratory analyses of the FFMQ facets.
Fig. 2
Violin plots of pre-post comparisons by group. n = 79. Intervention, intervention group; Control, WLC. Pre for both groups, baseline scores; Post for intervention group, after end of 6-week intervention; Post for control group, after end of 6-week wait; State anxiety, State-Trait Anxiety Inventory for Adults- State Anxiety subscale; Trait anxiety, State-Trait Anxiety Inventory for Adults- Trait Anxiety subscale; Anx. in past week, anxiety in past week, from Patient-Reported Outcomes Measurement Information System Anxiety-Short Form; Perceived stress, Perceived Stress Scale; Positive affect, Positive and Negative Affect Schedule, Short Form- Positive Affect subscale; Negative affect, Positive and Negative Affect Schedule, Short Form- Negative Affect subscale; MAAS, Mindfulness Attention Awareness Scale; FFMQ, Five Factor Mindfulness Questionnaire-Short Form. *p < 0.05; FDR-corrected for multiple anxiety measures and multiple mindfulness measures. Used the non-transformed data for these variables for visualization purposes
Table 2
Post hoc pre-post comparisons for mixed ANOVAs, primary sample
Measure
Pre
M (SD)
Post
M (SD)
t
pa
g
gdiff
Intervention group
State anxietyb
1.66 (0.12)
1.59 (0.13)
− 3.52
< 0.01
0.55
0.61
Trait anxiety
54.59(10.90)
47.97 (10.39)
− 4.86
< 0.01
0.61
0.47
Anx. past week
23.62 (6.44)
19.89 (5.82)
− 3.88
< 0.01
0.60
0.71
Perceived stress
22.51 (7.52)
18.76 (6.70)
− 3.27
< 0.01
0.52
0.53
Positive affect
27.32 (8.09)
29.38 (7.10)
1.81
0.04
0.27
0.37
Negative affectb
1.37 (0.15)
1.31 (0.13)
− 3.24
< 0.01
0.42
0.60
MAAS
50.62 (14.45)
56.14 (12.37)
2.81
< 0.01
0.40
0.37
FFMQ
34.24 (6.11)
39.03 (6.01)
4.42
< 0.01
0.77
0.61
WLC
State anxietyb
1.61 (0.16)
1.62 (0.15)
0.92
0.84
− 0.06
Trait anxiety
52.45 (13.32)
50.57 (13.79)
− 1.83
0.11
0.14
Anx. past week
23.10 (7.37)
23.95 (8.19)
1.03
0.84
− 0.11
Perceived stress
21.19 (7.47)
21.26 (7.73)
0.08
0.53
− 0.01
Positive affect
28.98 (10.10)
27.98 (9.54)
− 1.73
0.95
− 0.10
Negative affectb
1.35 (0.16)
1.38 (0.16)
1.35
0.91
− 0.18
MAAS
49.33 (11.63)
49.69 (11.57)
0.34
0.37
0.03
FFMQ
33.60 (7.74)
34.93 (8.58)
1.78
0.08
0.16
Intervention group n = 37. WLC n = 42
Pre for both intervention group and WLC, baseline scores; Post for intervention group, after end of 6-week intervention; Post for WLC, after end of 6-week wait; gdiff, the difference in Hedge’s g between the intervention group and WLC; State anxiety, State-Trait Anxiety Inventory for Adults- State Anxiety subscale; Trait anxiety, State-Trait Anxiety Inventory for Adults- Trait Anxiety subscale; Anx. past week, anxiety in past week, from Patient-Reported Outcomes Measurement Information System Anxiety-Short Form; Perc. Stress, Perceived Stress Scale; Positive affect, Positive and Negative Affect Schedule, Short Form- Positive Affect subscale; Negative affect, Positive and Negative Affect Schedule, Short Form- Negative Affect subscale; MAAS, Mindfulness Attention Awareness Scale; FFMQ, Five Factor Mindfulness Questionnaire-Short Form; Bolded values indicate p < 0.05
aFDR-corrected for multiple anxiety measures and multiple mindfulness measures
bLog-transformed
×
The findings remained the same when utilizing the full sample (n = 81) for an intent-to-treat analysis, i.e., adding the two participants who did not engage in any intervention practices (Online Resource, Tables S8 and S9).
Due to group differences in baseline TMB scores, we conducted multiple linear regression analyses to determine whether pre-to-post change scores on the outcome measures could be predicted from group, controlling for TMB. Group, but not TMB, significantly predicted change scores (in the expected direction) for all variables except for trait mindfulness (Online Resource, Table S10). For both mindfulness measures, only TMB was a significant predictor, with higher TMB scores predicting less change in trait mindfulness scores. Examining the TMB correlations with the change scores, by group, revealed no significant associations (see Online Resource, Table S11).
Mindfulness Training Resulted in Improvement on All Measures in the Control Group
After participating in the intervention, the control group replication sample showed significant pre-to-post improvement on all outcome measures (effect sizes ranged from 0.36 to 0.73), corroborating the findings from the intervention group (Table 3). GLM analysis confirmed these findings. We conducted random effects regressions with time (pre, post) as a predictor for each of the outcome measures. Time was a significant predictor for all outcome measures, showing that the control group replication sample improved significantly on all measures from pre-to-post (Online Resource, Table S12).
Table 3
Replication sample pre-to-post-test comparisons (WLC after engaging in intervention)
Measure
Pre
M (SD)
Post
M (SD)
t or z
pa
g or r
State anxiety
45.39 (15.51)
40.19 (12.75)
− 2.83
< 0.01
0.36
Trait anxiety
50.33 (13.47)
45.06 (11.90)
− 4.45
< 0.01
0.41
Anx. past week
24.17 (8.52)
19.92 (6.79)
− 4.75
< 0.01
0.54
Perceived stress
21.03 (8.10)
16.92 (7.01)
− 4.37
< 0.01
0.53
Positive affect
27.58 (9.17)
31.67 (9.17)
3.97
< 0.01
0.44
Negative affectb
25.47 (9.42)
20.00 (6.88)
4.32
< 0.01
0.72
MAAS
48.89 (11.77)
57.42 (11.10)
5.19
< 0.01
0.73
FFMQ
34.56 (8.59)
38.42 (7.89)
4.89
< 0.01
0.46
n = 36
Pre, after 6-week wait; Post, after end of 6-week intervention; State anxiety, State-Trait Anxiety Inventory for Adults- state subscale; Trait anxiety, State-Trait Anxiety Inventory for Adults- Trait Anxiety subscale; Anx. past week, anxiety in past week, from Patient-Reported Outcomes Measurement Information System Anxiety-Short Form; Perc. Stress, Perceived Stress Scale; Positive affect, Positive and Negative Affect Schedule, Short Form- Positive Affect subscale; Negative affect, Positive and Negative Affect Schedule, Short Form- Negative Affect subscale; MAAS, Mindfulness Attention Awareness Scale; FFMQ, Five Factor Mindfulness Questionnaire-Short Form; Bolded values indicate p < 0.05
aFDR-corrected for multiple anxiety measures and multiple mindfulness measures
bz statistic for Wilcoxon signed-rank tests, and r statistic for effect size calculated from Wilcoxon signed-rank tests
Mindfulness Training Resulted in Significant Improvement over Time on Most Measures
Collapsing across groups to estimate the intervention effects on a larger sample size, we found a statistically significant change over time (pre, post, follow-up) for all outcome measures on one-way repeated-measures ANOVAs (Online Resource, Table S13). Post hoc pairwise comparisons revealed statistically significant improvement from pre-to-post and pre-to-follow-up and no significant change from post-to-follow-up, on state anxiety, trait anxiety, anxiety over the past week, perceived stress, negative affect, MAAS, and FFMQ (Table 4). Only positive affect showed no significant improvement over time. GLM analysis confirmed these findings. We conducted random effects regressions with time (pre, post, follow-up) as a predictor for each of the outcome measures. Time was a significant predictor for all outcome measures, except positive affect (Online Resource, Table S14).
Table 4
Post hoc pre, post, follow-up comparisons, within-subjects repeated-measures ANOVAs
Measure
Comparison
t or z
pa
d or r
State anxietyb
Pre-post
4.14
< 0.01
0.51
Pre-follow
3.11
< 0.01
0.38
Post-follow
− 0.94
0.45
− 0.11
Trait anxiety
Pre-post
− 3.09
< 0.01
0.52
Pre-follow
− 2.95
0.01
0.50
Post-follow
0.14
0.89
− 0.03
Anx. past wk.b
Pre-post
5.29
< 0.01
0.65
Pre-follow
4.18
< 0.01
0.51
Post-follow
− 0.29
0.87
− 0.04
Perceived stress
Pre-post
− 3.49
< 0.01
0.58
Pre-follow
− 2.85
0.01
0.48
Post-follow
0.64
0.52
− 0.12
Positive affect
Pre-post
2.27
0.08
0.38
Pre-follow
1.01
0.31
0.18
Post-follow
− 1.25
0.31
− 0.22
Neg. affectb
Pre-post
5.55
< 0.01
0.68
Pre-follow
4.02
< 0.01
0.49
Post-follow
− 1.15
0.25
− 0.14
MAAS
Pre-post
3.39
< 0.01
0.58
Pre-follow
3.37
< 0.01
0.56
Post-follow
− 0.02
0.98
0.00
FFMQ
Pre-post
3.77
< 0.01
0.65
Pre-follow
3.94
< 0.01
0.67
Post-follow
0.17
0.98
0.03
n = 67, groups combined
State anxiety, State-Trait Anxiety Inventory for Adults- State Anxiety subscale; Trait anxiety, State-Trait Anxiety Inventory for Adults- Trait Anxiety subscale; Anx. past wk., anxiety in past week, from Patient-Reported Outcomes Measurement Information System Anxiety-Short Form; Perceived stress, Perceived Stress Scale; Positive affect, Positive and Negative Affect Schedule, Short Form- Positive Affect subscale; Neg. affect, Positive and Negative Affect Schedule, Short Form- Negative Affect subscale; MAAS, Mindfulness Attention Awareness Scale; FFMQ, Five Factor Mindfulness Questionnaire-Short Form; Bolded values indicate p < 0.05
aFDR-corrected for multiple anxiety measures and multiple mindfulness measures, as well as for multiple timepoints
bz statistic for Wilcoxon signed-rank tests, and r statistic for effect size calculated from Wilcoxon signed-rank tests
Additionally, at follow-up, there was a significant decrease in the number of adults with the high trait anxiety designation (50.7%) as compared to before the intervention (71.6%; n = 67, X2(1) = 4.25, p = 0.04).
App usage during the follow-up period was quite minimal. The mean number of total minutes used during the 6-week follow-up period was only 9.98 min (SD 21.76) compared to the 224.31 mean number of total minutes used during the intervention (SD 93.20).
Improvement in Outcome Measures Was Correlated with Increased Trait Mindfulness
To determine whether changes in the outcome measures may have been due to increased mindfulness, we assessed the relationship in the combined group, between participants’ pre-to-post change in outcome measures and their pre-to-post change in mindfulness scores. Mindfulness change scores (i.e., increased trait mindfulness) were significantly correlated with improvements in all outcome change scores (Online Resource, Table S15).
In this study, we conducted the largest RCT to date, to our knowledge, of a remote MBI with autistic adults, examining the impact on anxiety symptoms and perceived stress. Employing a 6-week smartphone-based mindfulness intervention, our primary hypothesis was that practicing mindfulness would decrease symptoms of anxiety and stress, which tend to be high in autistic adults (anxiety: Fombonne et al., 2020; Jadav & Bal, 2022; Nimmo-Smith et al., 2020; Underwood et al., 2023; stress: McQuaid et al., 2022). In the RCT phase of the study, autistic adults who were assigned to the mindfulness intervention, relative to a WLC group, exhibited significant reductions in anxiety symptoms, perceived stress, and negative affect, and significant increases in positive affect, and trait mindfulness. These findings were replicated when the control group engaged in the mindfulness intervention. Moreover, these benefits, with the exception of positive affect, were sustained when participants from both groups were reassessed 6 weeks after the completion of the mindfulness intervention.
The autistic adults in our sample exhibited high rates of anxiety symptoms, according to their trait anxiety scores. Specifically, at baseline, 72% of the autistic adults enrolled in the study met the high trait anxiety designation. This makes it all the more valuable that the mindfulness intervention reduced anxiety symptoms and perceived stress. It is important to note that because we did not administer diagnostic clinical interviews to determine the presence of clinically significant anxiety disorders, the individuals with a high trait anxiety designation cannot be considered a clinical population with regard to anxiety.
The present findings are congruent with previous studies showing that in-person and remote MBIs are often helpful in reducing anxiety symptoms and/or stress in adults in both the general population and those with anxiety disorders or high levels of anxiety (Boettcher et al., 2014; Economides et al., 2018; Hoge et al., 2023; Khoury et al., 2015; Querstret et al., 2020; Wang et al., 2023). Importantly, among the individuals in our study who completed follow-up, the percentage categorized as having high trait anxiety dropped significantly from 72% at baseline to 51% at follow-up. These results are also consistent with research demonstrating the beneficial impact of in-person MBIs for autistic adults (Agius et al., 2024; Kiep et al., 2015; Pagni et al., 2023; Sizoo & Kuiper, 2017; Spek et al., 2013). In fact, for in-person studies, effect sizes of significant changes were primarily in the medium range, with some large effect sizes as well, which generally aligns with the magnitude of the effect sizes in the present study.
The present study extends the findings from the one study to our knowledge, that examined the impact of a remote MBI on symptoms of anxiety in autistic adults (Gaigg et al., 2020). The present study expands on that study by being entirely remote, using an app, which provides added convenience, obtaining objective data regarding time spent on the intervention, controlling for baseline trait anxiety, and using a larger sample size. Results from the present study corroborate their findings that anxiety symptoms decreased at post-test, compared to the control group (Gaigg et al., 2020).
Consistent with many other MBIs (Chen et al., 2023; Lunsky et al., 2022; Pagni et al., 2023; Spijkerman et al., 2016; also, for meta-analysis see Goldberg et al., 2019), participating in this intervention led to significant increases in trait mindfulness, on both the MAAS and the FFMQ. In fact, some of the largest effect sizes in this study were on the mindfulness measures. Additionally, pre-to-post increases in trait mindfulness change scores correlated significantly with improvement in all outcome measure change scores, suggesting that although causality cannot be inferred, enhancement of trait mindfulness may play a role in the improved outcomes.
For a detailed comparison of our findings to in-person and remote MBIs with autistic adults, see Online Resource.
Replicability is important in evaluating the generalizability of a finding. In the present study, the benefits of the MBI demonstrated in the RCT phase of the study were fully replicated when the control group engaged with the MBI. Persistence of benefits is also important in evaluating the value of an intervention. In the present study, benefits on all measures besides positive affect were fully retained over a 6-week period after the MBI, even though voluntary engagement with the app was markedly reduced in that period.
The feasibility of the intervention was demonstrated in several ways. First, over 72% of individuals in both groups completed at least 75% of the mindfulness curriculum. Second, 80% of individuals in the intervention group and 75% of those in the control group completed follow-up assessments. This is comparable to the range reported in a meta-analysis of attrition in smartphone interventions, which reported a mean attrition rate of 24% at short-term follow-up (Linardon & Fuller-Tyszkiewicz, 2020). Third, post-intervention feedback indicated that the majority of participants reported that it was easy to use the app, that it was not at all or only slightly difficult to find time to listen to the app, and that the app was very or moderately helpful for them. Further, no participants reported any adverse events or indicated that they chose to stop the intervention due to emotional distress. These findings suggest that the MBI was well-received and feasible for most participants.
Improved access to mental health treatments is needed for autistic adults because they often face barriers in accessing effective and affordable mental health services (Maddox & Gaus, 2018). This may be partly due to transportation difficulties due to being less likely to drive and to barriers in using public transportation (Kersten et al., 2020; Mazurek et al., 2023; Pfeiffer et al., 2022). Additionally, high unemployment rates among autistic adults (Espelöer et al., 2023; Roux et al., 2017) may result in fewer financial resources to allocate to treatment services as well as transportation to treatment. In a meta-synthesis examining barriers to mindfulness-based in-person therapy, autistic adults expressed concern about time and travel (Hartley et al., 2022). Moreover, in a virtual group-based MBI, autistic participants reported scheduling as one of the main challenges (Redquest et al., 2022), suggesting there may be fewer barriers with individual, self-guided interventions.
Further benefits of remote treatment with a mindfulness app include having options that would not be possible in a live class, such as being able to choose among various narrators and activity session lengths as well as being able to repeat activities and relisten to material. Thus, individuals can tailor the intervention in ways that best suit their needs, which may be particularly helpful for autistic individuals. For example, in our study, we found that many participants chose the active versus sitting version for their mindfulness practices, and a number of participants reported that they very much appreciated having this option.
Limitations and Future Directions
There are several limitations to this study. First, the results may not be generalizable to individuals with backgrounds different from those in our study, such as adults whose intellectual ability is well below the average range (the cutoff we used was approximately ≥ 2 SDs below the normed mean on a proxy for nonverbal IQ). Second, while we made some app customizations, we did not involve autistic collaborators who may have suggested important changes to the study. It is possible that if the intervention had been specifically tailored to autistic participants that the effects may have been even stronger. Third, participants were not blind to their group assignment; therefore, we cannot discount the possibility of expectancy bias playing a role, especially because we did not ask about expectations in advance. However, a number of participants reported specific ways that the intervention was helpful for them. This feedback, combined with the improved outcome scores, suggests that participants made important gains regardless of potential expectancy bias. Relatedly, having an active control group would help determine specific effects of the intervention as opposed to the possibility of “digital placebo” (Torous & Firth, 2016) and would likely show smaller intervention effect sizes since larger intervention effect sizes tend to be found when compared to an inactive versus an active control group (Linardon, 2020).
In the present study, we noted an unexpected finding where a proxy (TMB) nonverbal IQ score (and not group) was a significant predictor of both mindfulness change scores, with higher TMB scores predicting less change in trait mindfulness scores. It is unclear why this was the case, and it would be helpful to explore whether this finding would be replicated in a larger sample. Importantly, however, nonverbal IQ did not correlate significantly with any of the outcome measure change scores, by group.
Future research can explore whether there are certain aspects of an MBI that are more helpful in effecting change, such as whether specific mindfulness practices are particularly beneficial and whether different aspects of an MBI are best for targeting certain outcomes (e.g., anxiety, versus stress, versus affect). In this intervention, there were approximately equal numbers of mindfulness practices (e.g., breath awareness) and didactic lessons (e.g., explaining why mindfulness is helpful). Comparing an active control group that only receives didactic mindfulness lessons to an intervention group that only receives mindfulness practices may be particularly illuminating.
It is of interest that there was such low app usage during the follow-up period, despite the multiple benefits and high scores on app usability. Studies examining such post-intervention usage in nonautistic groups have also reported reduced usage after intervention periods (e.g., Chu et al., 2022; Galante et al., 2021; van Emmerik et al., 2018). However, one study of autistic adults found no meaningful change in self-reported number of days meditating during the 8-week follow-up period that app usage was assessed (Stecher et al., 2024). We did not ask for or necessarily expect continued usage. Nevertheless, nearly all benefits were fully sustained over the 6-week follow-up, consistent with long-term benefits of MBIs on anxiety symptoms, lasting weeks to months in autistic adults (Gaigg et al., 2020; Lunksy et al., 2022; Sizoo & Kuiper, 2017).
The sustained benefits suggest that the participants may have found ways to incorporate mindfulness into their daily lives. In fact, in the HMP app, tips are provided at the end of each practice, suggesting ways in which to bring mindfulness into one’s daily life. It is possible that offering these tips contributed to the sustained outcome effects during follow-up, despite low app usage, perhaps by enhancing habit-forming behaviors which increased mindfulness in daily life. Making mindfulness more of a habit may be an important factor contributing to maintenance of longer-term intervention gains. An app-based MBI to reduce depressive symptoms in autistic adults found that the group which received habit training to increase their use of the app reported significantly fewer depressive symptoms at the 6-month follow-up compared to the group who received the app without the habit training (Stecher et al., 2024). Future research could examine what types of beneficial activities participants may be engaging in during follow-up, whether the benefits of app-based mindfulness last not only weeks and months but even years, and whether something like an annual booster could sustain benefits over multiple years.
It is unclear what accounts for the discrepancies in exploratory findings regarding the FFMQ facets in the current study, as compared to other MBIs with autistic adults (Pagni et al., 2023) and reviews with nonautistic populations which have found that the Non-Judging and Acting with Awareness facets tend to show the strongest correlations with outcomes (Carpenter et al., 2019; Mattes, 2019). Of note is that the Describing facet, the only facet exhibiting significant improvement compared to the control group in the current study, is likely associated with alexithymia, a condition that autistic adults commonly experience, and is defined as difficulty labeling and identifying one’s emotions (Kinnaird et al., 2019). Alexithymia has been associated with anxiety (Hendryx et al., 1991; Palser et al., 2018) and is also linked to differences in interpretations of bodily sensations which may be viewed as frightening or unexpected (Palser et al., 2018). Future studies could determine whether improvements in the Describing scale are associated with improvements in alexithymia, and whether alexithymia mediates the relationship between change in Describing and improved outcome measures.
The weaker findings regarding positive as compared to negative affect in the present study are also of note. Some systematic reviews have found no evidence of in-person MBIs improving positive affect (Goyal et al., 2014) and that mindfulness was found to reduce negative affect but not increase positive affect (Conversano et al., 2020). Further, in a study examining the association between mindfulness and affect over a 3-month period, mindfulness predicted reduced negative affect but did not predict increased positive affect (Jose & Geiserman, 2024). There is evidence suggesting that compassion training specifically is associated with increased positive affect (Förster & Kanske, 2022; Klimecki et al., 2013). Compassion training with autistic adults improved positive affect, among other outcomes (Cai et al., 2024). Although the app used for the present study included elements of compassion training commonly included in MBIs, perhaps emphasis on this component would lead to more robust findings concerning positive affect.
Further, future interventions should implement improvements suggested by autistic collaborators and involve autistic adults in the design or modification of a mindfulness-based app. Recommendations from our participants should be considered in future studies. These include more choices of narrators, with varied pacing options, providing transcripts of all online material, and having more in-app practice reminders. This last suggestion has been found to play an important role in participation and outcome in smartphone interventions, leading to increased adherence (Linardon & Fuller-Tyszkiewicz, 2020) and larger effect sizes (Linardon, 2020).
This remote mindfulness intervention was both feasible and effective, decreasing symptoms of anxiety, perceived stress, and negative affect in autistic adults, as well as increasing trait mindfulness, with improvements maintained at the 6-week follow-up. Online mindfulness interventions may be especially beneficial for autistic adults who might not otherwise seek in-person treatment due to barriers in accessing mental health care. Future research can include active control conditions, investigate the specific processes by which such an intervention is effective, identify characteristics of autistic individuals most or least likely to benefit from such a program, and involve autistic collaborators in the design or modification of a mindfulness-based app.
Acknowledgements
The authors thank the individuals who participated in this research study. We appreciate obtaining access to recruit participants through SPARK research match on SFARI Base. The authors also thank the Hock E. Tan and K. Lisa Yang Center for Autism Research as a funding source for this project (CL, IT, LB). We thank Healthy Minds Innovations, Inc. for use of the Healthy Minds Program app, and for the app customization. The authors also thank Dr. Ethan Scherer who provided statistical guidance.
Declarations
Ethics Approval
This study was approved by the Massachusetts Institute of Technology Committee on the Use of Humans as Experimental Subjects (Institutional Review Board). All procedures in this study were performed in accordance with the ethical standards of the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.
Informed Consent
Informed written consent was obtained from all individuals before participation in the study.
Conflict of Interest
The authors declare no competing interests.
Use of Artificial Intelligence
Artificial intelligence was not used.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
share
DELEN
Deel dit onderdeel of sectie (kopieer de link)
Optie A:
Klik op de rechtermuisknop op de link en selecteer de optie “linkadres kopiëren”
Met BSL Psychologie Totaal blijf je als professional steeds op de hoogte van de nieuwste ontwikkelingen binnen jouw vak. Met het online abonnement heb je toegang tot een groot aantal boeken, protocollen, vaktijdschriften en e-learnings op het gebied van psychologie en psychiatrie. Zo kun je op je gemak en wanneer het jou het beste uitkomt verdiepen in jouw vakgebied.