2.1 Questionnaires

Every questionnaire used in this survey has been previously validated in English and in French. All items and questionnaires in English and French are available in open access (https://osf.io/5u6am/). Questionnaires included in this survey relate to PA habits and the main correlates of PA found in the literature (income, education level, disorder severity and duration, medication use, substance use, and suicide ideations; St-Amour et al. (2021), Vancampfort et al. (2011), Vancampfort et al. (2018)).

2.2 Statistical Analysis

Descriptive statistics (N, %, mean) are used to describe the sociodemographic characteristics, the substance use disorder, insomnia, and PA level, preferences, and barriers in this sample. A Poisson multivariate regression was done to identify the main sociodemographic and clinical factors associated with weekly PA level. A graph presenting the distribution of PA level among participants is available in supplementary material (Figure S1). The age, social level, education level, household income, body mass index, emotion regulation difficulties, BPD symptoms, tobacco and alcohol use disorders, depression, past suicide attempts, psychotropic medication use and self-efficacy were included in the model. These independent variables were selected based on previous papers examining the determinants of PA in adults with mental disorders (Schuch et al., 2017; Vancampfort et al., 2011; Vancampfort et al., 2015). The R function stepAIC was used with an initial full model and combined with a forward and backward selection (Zhang, 2016). The final model has the lowest Akaike information criterion.

The multicollinearity of this model has been tested by calculating the variance inflation factor (VIF). No predictors had VIF values greater than 10 after the stepwise variable selection procedure. Statistical analyses were done with the R software version 4.2 with the libraries ggstaplot, stargazer, MASS, and summarytools (Patil, 2021; Zhang, 2016). Research materials, data and R codes are available in open access on Open Science Framework (https://osf.io/c3gvx/).

3.1 Levels of Physical Activity

In this sample, 122 (65.9%) participants declared being physically active according to the guidelines from the WHO (completing at least 150 minutes of PA weekly; WHO. Regional Office for Europe & United Nations Economic Commission for Europe, 2022). On average, participants complete 165.5 (SD = 282.1) minutes of travel-related PA, 315.2 (SD = 685.7) minutes of work-related PA, and 126.0 (SD = 224.9) minutes of leisure-related PA for a total of 606.7 (SD = 766.2) minutes of PA weekly including all three domains. In this sample, men performed significantly more PA than women (p = 0.002). For detailed results comparing total PA levels according to body mass index, sex, level of education, age, and country, see supplementary material (Figure S2).

3.2 Barriers

Barriers preventing participants from engaging in PA were measured with self-efficacy scales from 0 to 100. For each barrier, participants had to rate their self-efficacy to engage in PA when encountering said barrier. Scores were reversed for clarity, therefore a higher score means a more important barrier. The PA barriers are illustrated in Figure 1. For detailed barriers according to age, education level, sex, body mass index, and country, see supplementary material (Figures S3-S7).

Note: The higher the mean indicator (vertical black line) on the abscissa, the greater the barrier. The different curves represent the distribution of data for each barrier. The barriers in order from the top are: Being tired, Feeling pressure at work, Recovering from an injury, Being in a bad mood, Having personal problems, Having more interesting things to do, Without the support of friends or family, Having other engagement, Feeling unwell, Having friends at home, Feeling anxious, Not reaching previously fixed training goals, Having family problems, During vacations.

3.3 Preferences

The detailed frequency to which each PA was declared as being among one’s preferred are represented in Figure 2. For detailed PA preferences according to body mass index, sex, country, education level, and age, see supplementary material (Figures S8-S12).

The details for preferred modalities and contexts of PA are presented in Figure 3.

3.4 Advice

The details of the preferences regarding PA advice are presented in Figure 4.

Note: For the PA advisor, the scores for each professional were calculated by asking participants to indicate which 3 professionals they prefer receiving advice from with the numbers 1 to 3 (1 = 3rd choice; 2 = 2nd choice; 3 = 1st choice). All professional without number were attributed the number 0. The mean score for each professional (including 0s) were computed, reversed for clarity (making a mean score closer to 1 a higher preference for that professional) and reported here.

3.5 Factors Associated with Physical Activity

Seven univariate outliers have been excluded because they had an exceptional self-reported PA level (i.e., more than 3500 min per week). A bivariate regression correlation table is presented in Figure S13 as supplementary material to show existing correlation in included variables. In this sample, the total level of PA is associated to the age, the social level, the education level, the income, the body mass index, the level of BPD symptoms, having a tobacco use disorder, the level of depression, the number of psychotropic medications used and the self-efficacy to PA. The total score of DERS-18 and ALPHA questionnaire were initially in the full model, but they have been excluded during the variables selection presented in statistical section. For all the details see Table 2.

4.1 Level of Physical Activity

First, we observe a relatively high level of PA in this sample. With a mean of more than 600 minutes weekly, participants are on average more active than the general population (about 340 minutes weekly; Colley et al. (2018)) and individuals with mental disorders (about 270 minutes weekly; Vancampfort et al. (2017)). With a little over 60%, this sample presents a similar proportion of active individuals as the general population (about 60%; Colley et al. (2018)) but a bigger proportion than individuals with other mental disorders (about 45%; Vancampfort et al. (2017)). These data are surprising and might indicate that the mean PA level is skewed by extreme individuals with large PA volume. However, the present data are difficult to put in perspective because of the absence of studies reporting PA level in this population (St‐Amour et al., 2022).

4.2 Barriers

The main barriers to PA reported in this study, as measured by the lowest reported self-efficacy to overcome them, are having friends over, having other engagements, recovering from an injury, having something else more interesting to do and feeling too tired. In comparison, individuals with severe mental disorders or substance use disorder reported lack of motivation, fatigue, having no one to engage in PA with, not having enough energy, would not be able to keep up, and lack of financial resources as barriers to PA (Abrantes et al., 2011; Romain, Longpré-Poirier, et al., 2020). The barriers reported here were then somewhat different from those reported in other populations with mental disorders with the exception of lack of energy/feeling tired. Surprisingly, lacking support or not having anyone to engage in PA with was the least important barrier in adults with BPD but was among the main barriers in populations with other mental disorders (Abrantes et al., 2011; Romain, Longpré-Poirier, et al., 2020). This difference might be explained by the relational nature of the personality disorder. Indeed, one of the main characteristics of BPD is the difficulty in maintaining interpersonal relationships (Euler et al., 2021). These difficulties might lead individuals with BPD to prefer engaging in PA alone and therefore not considering lacking support or not having someone to do PA with as a major barrier to PA. Moreover, individuals with BPD tend to be more impulsive and to prefer quick smaller rewards than delayed larger rewards than healthy individuals and those with other mental disorders (Gunderson et al., 2018). Therefore, they might be tempted to engage in quickly rewarding activities rather than PA. However, more studies are needed to thoroughly understand this difference.

4.3 Preferences

The most frequently reported favourite PAs are somewhat consistent with the findings from populations with other mental disorders (Abrantes et al., 2011; Romain, Longpré-Poirier, et al., 2020). Indeed, walking is the preferred PA regardless of the studied population (Abrantes et al., 2011; Romain, Longpré-Poirier, et al., 2020). The clear preference of walking as a PA might be attributed to it being practical, self-paced and controlled, inexpensive, and not needing a lot of resources (Abrantes et al., 2011).

As mentioned before, participants in this study clearly prefer engaging in PA alone than with other people. This result differs greatly from what is seen in other populations with mental disorder having mostly no preference for doing PA alone or in groups. This difference might also be explained by the interpersonal difficulties experienced by individuals with BPD (Euler et al., 2021).

In this sample, most participants prefer supervised PA compared to unsupervised PA. This result is also surprising and in opposition to the findings observed in other populations. Indeed, in populations with other mental disorders, there is either a clear preference for unsupervised PA or no clear preference regarding supervision (Abrantes et al., 2011; Romain, Longpré-Poirier, et al., 2020). This result is unexpected considering previously observed results and interpersonal difficulties of individuals with BPD. However, this difference might be explained by the perception of the different relationships in play. Individuals with BPD tend to feel closer from those peripheral to their social network and farther from the more central individuals (Beeney et al., 2018). Therefore, they might feel closer to an outside individual like a kinesiologist or physical therapist supervising their PA session but would not want to share these moments with friends or relatives.

5.1 Correlates of Physical Activity

In this study, sociodemographic and clinical variables have some opposite correlation to PA level with what is observed in previous systematic reviews (Vancampfort et al., 2011). Indeed, age, education level, and household income are negatively correlated and social status is positively correlated to PA level. Also, body mass index, BPD symptoms, tobacco use disorder, alcohol use disorder and number of psychotropic medications are positively correlated. Higher levels of depression is associated with lower PA level. Finally, self-efficacy level is also positively linked to PA level which is also observed in a recent meta-analysis (Cabassa et al., 2020). Among these correlations, the most surprising are those with the education level, the household income, the level of BPD symptoms, the tobacco and alcohol use disorder, and the number of psychotropic medications. The association between lower income and education level, and higher PA level in the present sample might be explained by a greater proportion of work-related PA in individuals with lower income and education level (Statistics Canada et al., 2020). Moreover, work-related PA is more strongly correlated with the total PA volume than leisure time PA and most studies reporting PA level only report leisure time PA (Vancampfort et al., 2011). Health-related surprising correlates (body mass index, BPD symptoms, substance use disorders and medication) might be explained by the nature of psychiatric comorbid disorders in our participants (St-Amour et al., 2021; Vancampfort et al., 2011). Following the suggestion of a reviewer, a sensitivity analysis was realized using only leisure time PA as dependent variable, confirming the supposition regarding education level and household income’s surprising relation to PA. Moreover, relations between BMI and alcohol use disorder, and leisure time PA have also been reversed which is more in line with results from the literature. Finally, past suicide attempts became significantly associated with leisure time PA. For full details of these sensitivity analysis, see supplemental material (Table S2).

5.2 Limitations

However, this study suffers from some limitations. First, BPD diagnosis was self-reported making it hard to ensure this sample is composed solely of adults with BPD. This sample is also small considering the observational design adopted. Second, this study could have attracted more active than inactive participants due to its theme (i.e., PA) subjecting it to recruitment bias and increasing PA levels artificially. Using self reported PA measures may subject our results to recall bias and therefore affect the validity of the data. Future studies could use representative national surveys reporting both BPD diagnosis and objectively measured PA. However, to our knowledge, no such survey is available at the moment. Moreover, the unequal recruitment between countries biases the results with an over-representation of Canada and France limiting the generalizability in other countries. The online nature of this survey also poses some limitations. Indeed, online surveys are only accessible to those with access to a computer and sufficient informatics literacy to complete them. In addition, the cross-sectional design of this study prevents us from establishing causality relations between PA and its correlates. Finally, participants may have developed questionnaire fatigue because of the length of the survey, but this was not measured nor analyzed.

5.3 Conclusions

To our knowledge, this is the first study analyzing PA level, preferences, barriers, and correlates in adults with BPD. This information is primordial in developing future studies analyzing PA’s medium- to long-term effect in adults with BPD. With treatment guidelines already suggesting the use of PA (Blum et al., 2008; Linehan, 2014), and indirect evidence indicating potential benefits of PA in alleviating BPD symptoms (Mehren et al., 2020), there is an urgent need for data analyzing the effect of PA on the different components of BPD to test this hypothesis. Pending future results, PA could be used by healthcare professionals (mostly kinesiologists, psychiatrists, and psychologists) to treat or alleviate comorbid disorders in this population. In doing so, they should base their intervention on the present barriers and preferences (i.e., walking, individual intervention, supervised sessions, outdoor, etc.) to ensure greater adherence (Romain, Bernard, et al., 2020). Future studies should also use nationally representative surveys to answer these questions and aim to compare these results with those from individuals with other metal disorders.

6.1 Data Accessibility

Data is available via the Open Science Framework account (https://doi.org/10.17605/OSF.IO/C3GVX/) DOI: 10.17605/OSF.IO/C3GVX

6.2 Author Contributions

• Contributed to conception and design: SS, LC, PB
• Contributed to acquisition of data: SS, LC, DD, GL
• Contributed to analysis and interpretation of data: SS, PB
• Drafted and/or revised the article: SS, LC, JL, DD, GL, PB
• Approved the submitted version for publication: SS, LC, JL, DD, GL, PB

6.3 Conflict of Interest

The authors have no conflict of interest to declare.

6.4 Funding

This study was supported by PB’s Junior-1 research grant from the Fonds de Recherche du Québec - Santé.

6.5 Acknowledgments

We would like to thank every participant for their valuable contribution in realizing this study and everyone who got involved in the promotion of the survey and the recruitment process.

6.6 Preprint

The pre-publication version of this manuscript can be found on MedRxiv at https://doi.org/10.1101/2022.05.24.22275513 (DOI: 10.1101/2022.05.24.22275513)