The aims of study 1 were threefold: (1) to estimate the prevalence of self-reported AExs attributed to meditation and symptom worsening during the HMP, (2) to assess how participants felt about having practiced meditation despite AExs, and (3) to examine potential predictors of AExs. Prior literature indicates that psychological vulnerability factors, such as baseline mental health symptoms and loneliness, are positively associated with the reporting of AExs [11,25]. Accordingly, we expected that psychological resilience factors would predict fewer AExs. Thus, we selected predictors representing psychological vulnerability and resilience from the available measures. Baseline levels on predictors are of particular interest, as most research on the topic is cross-sectional. As such, past reports of AEx frequency may include distressing experiences not directly caused by meditation training but instead reflecting an individual’s preexisting propensities toward challenging emotional states. Thus, predictors included demographic variables, baseline psychological vulnerability factors (baseline depression, anxiety, loneliness, experiential avoidance, and fear of missing out), baseline resilience factors (self-compassion), and meditation practice variables (amount of HMP app usage and perceived barriers to meditation).

This study is a secondary analysis of data from a randomized trial whose sample size was determined by feasibility and powered to detect small-to-moderate between-group differences (d=0.30) on its primary outcome, as described in the original trial protocol [39]. While the current analysis was not specified before data collection, sensitivity power analysis with the pwr R package [52] showed that the available sample (n=315 participants who completed posttest AEx measures) was powered to detect zero-order correlations between hypothesized AEx predictors and AEx occurrence of r≥0.16 at 80% power.

Study 1 explored the prevalence, severity, and predictors of AExs among participants in a digital meditation training program. A total of 27.9% (88/315) of participants reported AExs that they attributed to meditation, and 6.7% (21/315) of participants reported some degree of functional impairment that usually lasted for 1 day or less. The study showed 10.3% attrition (35/351). Some participants may have dropped out because of AExs related to their use of the HMP. Under the most conservative assumption that all participants who dropped out experienced an AEx, the estimated prevalence would increase to 35.3% (124/351). Higher baseline depression, anxiety, loneliness, experiential avoidance, and perceived barriers to meditation were consistently associated with more AExs, whereas self-compassion showed a protective association in 1 model. Demographic characteristics and meditation dosage (eg, number of HMP usage days and meditation practice minutes) did not significantly predict AExs. These findings provided the basis for further confirmatory testing of identified predictors of AExs.

Study 2 drew data from a randomized trial using a 4-week HMP intervention in a distressed sample of US adults. This study aimed to estimate the prevalence of AExs in this sample and to test whether the AEx predictors identified in study 1 would replicate. On the basis of study 1 findings, we hypothesized that higher baseline depression, anxiety, loneliness, and perceived barriers to meditation would be associated with (1) more frequent AExs, (2) greater odds of reporting at least one specific AEx from the MRAES-MBP, and (3) a higher total number of reported specific AExs from the MRAES-MBP. Additionally, we exploratorily examined whether these same predictors were associated with depression and anxiety symptom worsening.

We made one deviation from the preregistration. We originally intended to conduct 2 sets of analyses on specific AExs (hypotheses 2 and 3), one including all reported AExs and another limited to those rated as “very distressing” or “somewhat distressing.” However, the number of participants who rated their experiences as distressing (n=73) was too small to support adequate model estimation. As a result, we do not report the second set of analyses. This deviation limited the study’s ability to identify risk factors for experiencing specific AExs from the MRAES-MBP that were subjectively evaluated as distressing; however, due to an insufficient number of participants reporting them, stable estimation was not possible.

We also conducted several exploratory analyses that were not preregistered. In study 2, participants’ compensation was not tied to their engagement with the HMP. Consequently, 7.8% of the sample (31/396) completed only the introductory guided meditation, which lasted approximately 5 minutes, and 4% (16/396) completed only the psychoeducation lessons without engaging in any guided meditations. This created a quasi-experimental setting that allowed us to compare AEx rates between participants who only received minimal or no exposure to meditation (and therefore were unlikely to have experienced AExs due to meditation) and those who were exposed to meditation.

Participants in study 1, including those who experienced AExs, generally reported being glad to have practiced meditation. Hence, in study 2, we aimed to isolate experiences that were perceived as unequivocally negative. To achieve this, as described in the following section, we removed the words “challenging” and “difficult” from the general AEx occurrence item (as experiences generally viewed as nonharmful may be challenging and difficult, eg, learning to play a musical instrument). We also included follow-up items assessing to what extent participants evaluated specific AExs that they experienced as distressing or positive. In addition, to avoid biasing participants by querying only negative experiences, we also assessed positive experiences that participants may attribute to their meditation practice.

Study 2, a preregistered confirmatory follow-up to study 1, found a lower overall prevalence of AExs that participants attributed to meditation, with 10.10% (40/396) of participants reporting experiencing AExs compared to 27.9% (88/315) in study 1. In this study, attrition was 25.3% (134/530). Assuming most conservatively that all participants who dropped out experienced an AEx, the estimated prevalence would increase to 32.8% (174/530), which is comparable to the estimate in study 1 (124/351, 35.3%). Impairment was reported by 3% (12/396; 21/315, 6.7%, in study 1). This difference could be attributed to emphasizing the distressing nature of these experiences in study 2 (ie, removing “challenging” and “difficult” from the item), lower baseline anxiety in study 2 sample (sample 1: mean 63.97, SD 5.59; sample 2: mean 61.54, SD 6.27), and differences in the populations sampled (ie, undergraduates vs general population adults). Compared with study 1, fewer participants in study 2 reported impairment or clinically significant symptom worsening, with 90% (36/40) of the participants who had distressing experiences indicating that they were still glad to have used HMP and 56% (93/166) evaluating at least one of the experiences from the MRAES-MBP as neutral or positive. Furthermore, 93.7% (373/398) of the participants reported deriving positive experiences from their engagement with HMP, suggesting predominantly positive appraisals relative to the 10.1% AEx rate. In terms of predictors, study 2 confirmed that higher baseline loneliness and perceived barriers to meditation were significantly associated with more AExs. However, contrary to study 1, baseline anxiety and depression were not consistent predictors of AExs. A protective effect of female gender and White race emerged in study 2, whereas demographic variables had no predictive utility in study 1. The prevalence of AExs and symptom worsening did not differ between participants who did and did not complete guided meditations, showing that at least some of the reported AExs stemmed from sources other than meditation. Overall, study 2 corroborated several findings of study 1 while yielding a lower estimate of AEx prevalence and providing a more nuanced picture of AEx genesis and participants’ subjective evaluations of their experiences.

Study 1 included an open-ended question, “What was helpful in responding to/managing challenging, difficult, or distressing meditation-related experiences?” in the postintervention survey. We analyzed participants’ responses using qualitative document analysis [55], a flexible, inductive method for identifying patterns of meaning (themes) from the data. This approach maximizes the preservation of participants’ voices by allowing themes to emerge from the data in a “bottom-up” manner rather than imposing predefined categories. Data analysis was conducted by 3 coders trained in qualitative coding. Through an iterative process of initial coding, data segments were systematically labeled until consensus among coders was achieved. This was followed by axial coding to identify overarching themes and relationships between labels [56]. Once consensus was reached, a narrative analysis of the themes was conducted. Pseudonyms were used to maintain participants’ confidentiality.

Responses varied from a few words to several sentences; the majority were brief. We identified 8 core themes summarizing the strategies that participants used to cope with AExs. Several participants described seeking social support (theme 1) from family, friends, or mental health professionals. For example, Avery reported, “talking to a therapist about my frustrations.” Others reported connecting to a broader sense of humanity through feelings of connection (theme 2), such as warmth, oneness, or an invisible bond with others. For example, Alex found it “helpful to remind myself I was not alone” and that “many people go through similar difficult experiences and they will be there for me.” Parker shared, “I went and wished well on others until I felt better about myself.”

Some participants coped by stepping away from the practice (theme 3), either mentally or physically. Sam shared, “I would take time away from the practice to find something to soothe my mind,” while Casey said they would “get up and complete a different task.” Others responded by accepting the present moment (theme 4), staying present with nonjudgment. Casey reflected on “reminding myself to acknowledge the thought and not judge it but that it’s okay if the mind wanders,” while Kai shared, “trying to ground myself or be present in the moment.” This acceptance often led to re-engaging with the practice (theme 5) through strategies such as refocusing attention, following the instructor’s guidance, or trying a different meditation technique. Casey noted that “slowly bringing [the mind] back to the breath was helpful.” Justice shared: “I used other practices that were taught in the previous week to try and get over that.”

Participants also sought to express and understand their emotions (theme 6) through introspection, crying, or talking with others. Jordan explained this as “[remaining] calm and thinking through why I was feeling the emotions I was feeling.” In the same vein, Dakota engaged in “reflecting on why I was feeling any challenge or distress.” To manage overwhelm (theme 7), participants reported slowing down the pace of the program, relaxing, using breathwork, taking breaks, spending time alone, or using medication. Taylor said they would “take some time to calm down and unwind,” while Ari noted they would “pause the practice and wait until I was ready to meditate a small amount of time later.”

Finally, participants engaged in reflective and reframing techniques (theme 8), such as prioritizing personal needs, adopting a growth mindset, or practicing gratitude and compassion. Quinn described using positive reframing: “Appreciate the difficulties and challenges I have faced and move forward with a willingness to be better.” Similarly, Phoenix used positive self-talk to cope: “primarily I had to remind myself that I am important too.”

The qualitative findings provided context for the quantitative results, offering insight into how participants understood and navigated AExs that emerged during the study. Participants described a wide range of coping strategies, highlighting the diversity of ways individuals relate to and manage challenging experiences. Many drew on skills learned through the HMP or from prior experience. These responses emphasized participants’ psychological resilience and capacity for self-regulation in the face of discomfort. As one participant noted, “while facing these experiences can be difficult, it’s nice to remember that facing them can improve your life overall.”

In this series of studies, we investigated the prevalence and predictors of self-reported AExs and worsening symptoms that occurred during participants’ engagement with the HMP, a self-guided digital MBI, as well as participants’ subjective evaluations of their experiences and coping strategies. The prevalence estimates were largely consistent with prior literature. Higher baseline depression, anxiety, loneliness, experiential avoidance, and perceived barriers to meditation predicted more AExs. Most participants evaluated learning to meditate positively, including those who reported AExs, and described using a variety of coping strategies, including techniques learned through HMP.

The 27.9% (88/315) of participants prospectively evaluated prevalence of self-reported AExs that participants attributed to meditation observed in study 1 aligned with estimates from previous survey studies that used the same measure in a retrospective observational study context (32%) [11,12]. Study 2 found a lower prevalence of 10.1% (40/396). This discrepancy may be partially due to the wording of the item in study 2, which specifically emphasized distressing effects, rather than more broadly including “challenging” or “difficult” experiences that some participants may actually perceive as neutral or positive. For example, learning to regulate attention through meditation practice is likely challenging or difficult for most individuals. However, lower prevalence rates were also observed across other AEx measures, suggesting that participants in study 2 experienced fewer AExs overall, possibly due to lower baseline symptoms. Pooled across the 2 studies, the prevalence of AExs attributed to meditation was 18% (128/711). The proportion of participants endorsing at least one item on the MRAES-MBP [30] was 65.4% (206/315) in study 1 and 41.9% (166/396) in study 2 (pooled 52.3%, 372/711). This pattern is consistent with prior research indicating that checklist-based self-report assessments yield higher AExs prevalence estimates compared to single-item or open-ended questions [6,11,12]. Across both studies, 10 (1.4%) of the 711 participants reported AExs occurring regularly or frequently. Eight (1.1%) participants reported moderate to severe impairment attributed to meditation, and 4 (0.6%) reported impairment lasting longer than 1 week.

In study 1, 10.8% (34/316) and 12% (38/316) of participants and in study 2, 11.8% (49/417) and 10.1% (42/417) experienced a clinically significant ≥3 T-score increase of depression and anxiety symptoms, respectively (pooled 11.3%, 83/733, for depression and 10.9%, 80/733, for anxiety). This estimate is higher than the estimated 3.6% rate of clinically significant symptom worsening among participants of the Mindfulness-Based Stress Reduction program offered by a community health clinic, an estimate calculated from participants’ records collected over 14 years (n=2155) [29]. Baer et al [13] reported symptom deterioration rates (2%‐6%) in an RCT similar to community Mindfulness-Based Stress Reduction. One explanation for the discrepancy is that this study applied a lower threshold for clinically significant symptom worsening. In our samples, the MCID of 3 T-score points [45,46] corresponds to roughly 0.4 SD on PROMIS Depression and 0.5 SD on PROMIS Anxiety. In contrast, Baer et al [13] used a higher cutoff—approximately 1 SD on their depression and anxiety measures. Hirshberg et al [29] defined clinically significant harm as a shift from functional to moderate, or from moderate to severe symptom levels, an approach that generally requires a larger change in symptoms than the MCID used here. To contextualize these findings, the share of the participants who experienced depression and anxiety symptom improvement was 51.9% (164/316) and 45.6% (144/316), respectively, in study 1, and 55.2% (230/417) and 61.4% (256/417), respectively, in study 2 (pooled 53.8%, 394/733 for depression and 54.6%, 400/733 for anxiety).

Across both studies, baseline loneliness consistently predicted AExs, suggesting that individuals who felt more isolated were more vulnerable to difficult experiences. This was further supported by findings from the qualitative study 3, in which some participants described using their actual or perceived social connections as a way to cope with AExs. Those with higher baseline loneliness may have had less social support or sense of connection. Thus, social context may serve as both a risk factor and a protective buffer in the occurrence of and coping with AExs.

In study 1, baseline levels of depression and anxiety were associated with several AEx outcomes; however, this was not replicated in study 2. Baseline experiential avoidance emerged as a risk factor across several AEx measures in study 1; however, it was not assessed in study 2. This finding was supported by the qualitative data, where some participants described coping with challenging experiences by expressing and processing their emotions. Individuals with a stronger tendency to avoid unwanted internal experiences may have been less able or willing to engage in this form of coping, potentially increasing their vulnerability to AExs.

Neither the number of days using the app nor total meditation minutes predicted AExs in Study 1. This contrasts with Britton et al [6], who found evidence that AExs emerge according to the principle of biological gradient, where greater exposure leads to a higher incidence of the experience. However, the Britton et al study involved in-person mindfulness-based cognitive therapy with much higher exposure—an average of over 32 hours of formal and informal practice during the 8-week program and approximately an additional 26 hours of practice until the 3-month follow-up, when AExs were evaluated. In our study, participants averaged 3.76 hours of meditation (SD 1.99) during the 2-week study 1 period, and 1.27 hours of meditation (SD 1.53) during the 4-week study 2 period. Notably, perceived barriers to meditation predicted AExs in both studies, suggesting that subjective anticipated difficulties with the practice, such as internal resistance, frustration, or misunderstanding, may contribute to negative outcomes.

Importantly, in study 2, participants who, as determined by the objective app logs, had minimal (<5 min) or no exposure to guided meditations reported AExs they attributed to using the HMP app at the same rate as those who engaged in guided meditations. This suggests that these experiences were not caused by meditation and therefore cannot be classified as AEfs. When considered alongside the above results that baseline symptoms and participant characteristics seem to be the most robust predictors of AExs, at least among the variables we assessed, our data provide a caution against the practice of inferring causality from self-attribution or observational data. Although the temporality criterion is satisfied in such data—experiences occurred during or following meditation practice—the specificity criterion cannot be evaluated without a comparison condition, as emphasized in the Bradford Hill framework for causal inference [57]. Prior work has similarly found that participants sometimes attribute experiences to MBIs that are better explained by unrelated causes or natural symptom fluctuations [25].

Most participants evaluated their meditation experience and use of HMP positively, including 89.8% (79/88) of those who reported AExs in study 1 and 90% (36/40) in study 2. This estimate is similar to Baer et al [13], who reported that between 85% and 92% of the participants who had AExs rated them as not at all upsetting or only somewhat upsetting. An observational study [11] also found that 88% of the survey participants who had AExs were still glad to have practiced meditation. Future research could further disentangle the immediate valence and appraisal of experiences during meditation training from the interpretive meaning-making that practitioners construct around those experiences.

In sum, the contemplative science field has made substantial strides in recent years in documenting negative experiences that may arise during meditation training [8]. Our study extended this work by examining such experiences in the context of a digital MBI. The overall low rates of reported AExs and impairment, the absence of statistically significant differences between meditating and nonmeditating groups, the high proportion of participants who experienced an AEx yet still endorsed being glad to have meditated, and the resilience reflected in participants’ coping strategies collectively argue against generalized concerns about potential widespread harms associated with the digital MBI evaluated here. That said, a small number of participants may report serious or prolonged AExs during digital meditation training. Such experiences should be monitored in both research and clinical contexts, and their causal connection to the intervention should be systematically evaluated.

This study holds important public health implications because meditation apps are the most widely used category of mental health mobile apps, with millions of active users [15,16]. US Food and Drug Administration guidelines require the collection and reporting of adverse events in sponsored trials of investigational devices; however, because the vast majority of digital meditation interventions are classified as wellness devices and are therefore outside the scope of US Food and Drug Administration–cleared digital therapeutics, the onus remains on the creators to monitor for challenging experiences [58,59]. Across digital mental health apps broadly, participants report AExs such as symptom worsening, distress caused by app features, self-harm or suicidal ideation or intent, and rehospitalization, with frequencies ranging from common to very rare, and differing widely based on the clinical population targeted with the app [20]. Given the frequency with which AExs occur, their screening and reporting should be integrated into digital mindfulness programs and apps. Our data also suggest that meditation programs should inform new practitioners that, as in daily life, distressing experiences may occur during meditation. Having realistic expectations about the kinds of experiences that can arise during meditation will allow individuals to make informed choices about participating in meditation training. In addition, some researchers have conjectured that when novice practitioners encounter challenging experiences, they may be more likely to discontinue practice as a result [12]. If meditation programs more fully inform individuals about the range of potential experiences and that some degree of distress during meditation is not uncommon and should even be expected for most participants [6,13], new practitioners may be better able to work with such challenges.

In the data we reviewed, the majority of reported AExs were brief and short-lived. However, a subset of individuals may experience more intense and longer-lasting AExs that they subjectively attribute to meditation. Developers of meditation programs are encouraged to incorporate screening for vulnerability factors—such as those identified in this and other studies—into their offerings. Individuals who are flagged as high risk for AExs could be more closely monitored through in-app questionnaires and, when clinically indicated, escalated to receive support within the app, be directed to external resources for meditators experiencing distress, or given information on general mental health support (eg, calling an emergency crisis line).

It is clear from prior qualitative research that some AExs attributed to meditation [10,60], particularly intensive meditation practice and/or esoteric practice of the sort not encountered in secular meditation programs or apps, are more severe, long-lasting, and potentially debilitating. Future research on meditation AExs and AEfs would benefit from differentiating between types of meditation practice, in terms of techniques, duration, intensity, and context (eg, app vs in person; with a trained teacher vs without a trained teacher). Indeed, many meditation traditions speak of the potential negative consequences of meditation practiced with the wrong understanding, or in the wrong context, or without an appropriate teacher [61].

The research makes a few contributions to the field. First, we emphasize the importance of distinguishing between AExs—any distressing experiences occurring during an intervention—and AEfs, which refer to harmful experiences that are caused by the MBI and would not otherwise occur. As these data illustrate, participants report meditation-related AExs even in the absence of meditation. Hence, subjective attribution alone cannot be taken as sufficient evidence of causality. Rare events, such as severe or prolonged AExs and their possible relationship to meditation, may be studied on a case-by-case basis, as previously suggested [6]. However, we encourage the researchers to rely on established causal inference methods (eg, RCTs) to evaluate whether more common mild AExs should be classified as AEfs—that is, iatrogenic effects caused by meditation. Beyond RCTs, hybrid designs such as stepped-wedge trials could evaluate AExs in real-world app deployments.

Second, given the prevalence of distressing or difficult experiences in the context of meditation, it is important for future research to actively monitor and report AExs in clinical trials, with particular attention to how they are operationalized. Our study yielded varying AEx prevalence estimates depending on the measures used. Therefore, the way AExs and AEfs are defined and measured plays a critical role in shaping findings. Researchers are encouraged to clearly specify how AExs and AEfs are defined in their studies, which measurement scales are used, the degree to which their study design allows causal inferences to be drawn, and to interpret their findings in light of those methodological choices.

There are several limitations to note. First, although the study advanced prior observational research on AExs in digital MBIs by establishing temporal precedence and using standardized meditation training, it did not include a control group. As a result, we were unable to compare AExs in meditation training to a base rate and test the causal relationship between meditation and AExs. Second, while attrition rates in this study (study 1: 36/351, 10.3% and study 2: 134/530, 25.3%) were at or below estimates for attrition in RCTs of mental health apps generally [62], it is possible that some participants who discontinued their involvement in our studies did so due to AExs related to their use of the HMP. This could lead to an underestimate of AExs based on those who completed postintervention measures. Under the most conservative assumption that all participants who dropped out experienced an AEx, the estimated prevalence would increase to 35.3% (124/351) in study 1 and 32.8% (174/530) in study 2. Third, these data are useful for understanding AExs in the context of digital trainings in which participants received relatively low doses of meditation—an average of 3.76 hours in study 1 and 1.27 hours in study 2. We cannot infer from these data the impacts of digital or in-person meditation training when the dose is 10 times or, in very intensive contexts, 50 times greater. Future trials could explicitly contrast app-based secular practices with more intensive forms of practice (eg, standardized 8-wk MBIs and meditation retreats). Additionally, although using several different operationalizations and measures allowed us to capture a broad range of AExs, the study relied exclusively on self-report measures. Previous studies have reported objective adverse events, such as hospitalization occurring during the trial [63]; however, such severe adverse events are rare. To better evaluate mild AExs of the type reported here, future studies could adapt AEx measures for informant report and combine self-reported and informant-reported AExs to enable more objective assessment. Finally, this study mainly focused on the risk factors associated with AExs, but it may be illuminating for research to more closely examine protective factors.