A Remote Adapted Physical Activity Intervention for Women With Breast Cancer and Severe Depressive or Anxiety Symptoms: Series of N-of-1 Trials With Ecological Momentary Assessment

Study Design

This N-of-1 study followed an A-B-A design (representing the three phases of the study) and lasted 16 weeks. No randomization nor blinding were used due to the nature of the study. Each A phase (2-week each) represents the pre- and post-intervention baseline measures, and the B phase (12-week) represents the intervention phase. For the whole 16 weeks, participants received a daily EMA prompt to report their depressive and anxiety levels. The intervention (B phase) included two to three (un)supervised remote PA sessions per week coupled with motivational and educational text messages.

Recruitment

Participants were recruited from Maisonneuve Rosemont or Santa Cabrini Hospitals (Montreal, Canada). Flyers, outlined the study aims, inclusion criteria and contact information for the research coordinator, were distributed through psycho-oncology consultations and shared to patients on the psycho-oncology unit waitlist and on the oncology center website and social networks. Potential participants were screened by telephone and included if they: 1) were diagnosed with a non-metastatic BC and currently completing treatment or completed treatment at least three months prior; 2) reported a high level of self-reported depressive (Patient Health Questionnaire score ≥ 15) or anxiety symptoms (General Anxiety Disorder’s questionnaire score ≥ 15); 3) were aged 18 to 65 years; 4) were considered inactive (less than 150 minutes per week or Godin’s questionnaire score < 23; Amireault et al., 2015); 5) possessed a smartphone. Participants were excluded if they: 1) reported a psychotic or schizophrenia disorder diagnostic; 2) answered positively to the PA Readiness Questionnaire for Everyone; 3) received a weekly psychological treatment from a clinician; 4) reported a major functional or physical disability; 5) were unable to provide consent. The consent form was sent via e-mail before the initial evaluation. Participants were rewarded $150 CA upon study completion.

Physical Activity Intervention

The intervention combined a remote adapted PA intervention, supervised by a kinesiologist, coupled with weekly text messages for 12 weeks. The characteristics of this intervention were based on the ACSM recommendations, identified behavioral change techniques and PA preferences in interventional context among women with BC (Caudroit, Chevance, et al., 2025). The detailed description of the intervention development and content have been previously published (Caudroit, Lapointe, et al., 2025). This intervention has been found as feasible, acceptable, and associated with high adherence rates (Caudroit, Lapointe, et al., 2025).

Participants had three supervised remote PA sessions from week three to six. From week seven to 14, participants had two supervised remote PA sessions, with the choice between supervised or unsupervised sessions for the third weekly session. The supervised PA sessions included: 1) combination of behavioral change techniques and aerobic exercise with resistance or yoga exercises (participants’ choice); 2) the durations of sessions ranged from 30 (weeks 3-6) to 60 minutes (weeks 7-14) but actual durations were adaptable based on participants’ self-reported fatigue level; 3) the effort intensity was selected by the participant, guided by the proposed intensity of the day provided by the kinesiologist, explained through the rate of perceived exertion, progressively increasing, and varying between 1 and 8 (different for aerobics, resistance, yoga) on a scale of 10 during the 14 weeks. The unsupervised PA sessions could be based on previous supervised sessions or personalized after discussions with the participant (e.g., 2 weekly walks of 30 minutes).

Measures

Online questionnaires and EMA were used to measure our variables of interest during the phases of our study. In line with the SCRIBE, we used validated questionnaires to describe participants demographic characteristics and clinical features (Tate et al., 2016). A research-oriented app was installed (EthicaData) on participants’ smartphones to obtain self-measured daily rating of depressive and anxiety symptoms.

Questionnaires

Before the intervention, online questionnaires collected information on age, educational achievements, marital status, income, number of children, date of diagnosis, treatments, current medications, and history of cancer. According to the SCRIBE recommendations, generalization measures were also performed to increase the external validity of the study. Thus, GAD-7 and PHQ-9 were filled by participants three times during the study (details in Figure 1; Tate et al., 2016). The respective validated thresholds of these scales were used to describe the individual scores (see details in the Supplementary Materials [Bernard, 2025S]). The Patient Health Questionnaire (PHQ-9) is a screening instrument with nine items, developed to assess the severity of depression (Kroenke et al., 2001). For each item, the respondent is asked to rate how often each symptom occurred over the last two weeks, on a Likert scale ranging from 0 “not at all” to 3 “nearly every day”. The sum score (range to 0 to 27) indicates the degree of depression with score of ≥ 15 representing severe levels of depression (Howell et al., 2015). The Generalized Anxiety Disorder (GAD-7) is a one-dimensional instrument created to detect symptoms of generalized anxiety disorders (Spitzer et al., 2006). Core symptoms within the past two weeks were queried with seven items on a four-point Likert scale rated from 0 (not at all) to 3 (nearly every day). The total GAD-7 score can range from 0 to 21 and a score ≥ 15 represents severe anxiety symptoms levels (Howell et al., 2015). PHQ-9 and GAD-7 are both recommended questionnaires by the Canadian Association of Psychosocial Oncology (Howell et al., 2015).

Click to enlarge

Figure 1

Study Design

Daily Assessments

Every day during the Phase B, participants rated the severity of depressive and anxiety symptoms using a 0-100 visual analog slider (0: not at all – 100: as much as possible). Participants rated the severity of depression symptoms with 8 adapted items from the Patient Health Questionnaire (PHQ-9, Kroenke et al., 2001). The severity of anxiety symptoms was measured with 6 adapted items from the GAD-7 (Spitzer et al., 2006). All EMA items are presented in Table 1.

Internal Validity

Single case experimental studies may be subject to rival hypotheses that can explain changes in the dependent variables like maturation, question-behavioral effect, and other external factors (St-Amour et al., 2024). To ensure the internal validity (i.e., the observed change is attributable to the intervention rather than other factors), participants were questioned at the end of Phase B about important event having occurred in their life during the phase they just finished that could have a prolonged (positive or negative) impact on their mental health symptoms.

Statistical Analyses

A piecewise linear regression has been performed for each participant. This analysis is particularly well-suited for time series with Gaussian or Poisson distribution (Lapointe et al., 2023; Wilbert, 2021). We considered the auto-correlation between symptoms data only for models with a Gaussian distribution because it was implemented only for the latter in the scan package (Wilbert, 2021). Through those analyses, we compared the daily depression and anxiety level during and after the intervention to baseline measures. Each piecewise regression was carried out to examine the level effects (i.e., the difference between the mean of symptoms in Phase A with Phase B and A’, divided by the standard deviation of the residuals) and test the slope effects (i.e., the continuous change of symptom levels due to the PA intervention). Two models were systematically compared by including or not the trend. The model with the highest R² was finally selected for each dependent variable.

Transparency and Openness

The study protocol was registered a priori with OSF registries (see Supplementary Materials [Bernard et al., 2022S]). Analyses and graphics have been performed with R 4.3 and ggplot2, and scan packages. Data, open materials, and R scripts are available online (Bernard, 2024S-a).

Participants Characteristics

Between 2022 May and 2023 June, 18 participants (10 during treatment and eight post-treatment) were included. Baseline sociodemographic, health and cancer-related information, and PA adherence sessions rates are in Table 2. Two participants were dropped during this study. Some participants postponed the beginning of their intervention phase due to events out of their control making some A phases longer for some than others. Regarding scores on PHQ-9 and GAD-7, all participants experienced severe psychological distress (score ≥ 15 on, at least, one scale) with 13 patients who had severe depressive disorders, one who had severe anxiety disorders and four who had severe depressive and anxiety disorders. The EMA adherence rates were ranged from 48% to 88%, with eight participants with a rate above 70%.

Intervention Effects

Figures 2 and 3 present the daily mean of depression and anxiety symptoms for all participants and local regressions. Table 3 and 4 present the results of piecewise regressions. The level regression coefficient indicates a mean level change after the beginning of intervention. The slope regression coefficient indicates the daily decrement following the beginning of intervention.

Click to enlarge

Figure 2

Depression Scores in Three Conditions With Regression Lines for Each Phase

Click to enlarge

Figure 3

Anxiety Scores in Three Conditions With Regression Lines for Each Phase

A significant decrease of depressive symptoms during the intervention has been found among 10 participants (id2, id3, id4, id5, id8, id13, id10, id11, id12, id15, see details in Table 2). As presented in Figure 2, different patterns of symptom reduction can be observed. For id4, id5, id8, id12, a negative and significant effect was found for level and slope. In other words, benefits from our intervention were quick and progressive. For id2, id3, id13, the beginning of intervention was associated with a significant higher level of symptoms, however, these symptoms gradually decreased (i.e., negative slope) during the intervention. A significant reduction of daily depressive symptoms was observed only for the level (id10, id15) or slope (id11).

A significant decrease of anxiety symptoms following the introduction of the intervention has been found for nine participants (id1, id2, id3, id4, id5, id10, id11, id12, id13, see details in Table 3). Different patterns of symptom reduction have also been found (see Figure 3). For id12, a negative and significant effect was found for level and slope. For id3, the beginning of intervention was associated with significant a higher level of anxiety symptoms, however, these symptoms gradually decreased during the intervention. A significant reduction of daily anxiety symptoms was observed only for the level or slope for id10, id11, and id1, id2, id4, id5, id13, respectively.

In summary, a significant reduction of depression and anxiety symptoms was found among seven participants (id2, id3, id4, id5, id10, id12, id13). A no significant effect of our intervention on mental health symptoms was found for four participants (id6, id9, id14, id16). Also, id6 showed a significant progressive increase of depressive and anxiety symptoms during the follow-up phase.

Significant slope changes were observed more often than level changes (8/10 for depressive and 7/9 anxiety symptoms), indicating that the impact of PA on mental health was generally gradual. Seven participants with a significant reduction of depressive symptoms during the intervention showed maintained benefits during the follow-up phase (id, 3, id4, id5, id8, id11, id12, id13). Four women experienced maintained benefits in terms of anxiety (id1, id4, id5, id13). The statistical modeling of time series explained 15% to 74%, and 12% to 54% of the variance for depressive and anxiety symptoms, respectively (more details are provided in the Supplementary Materials [(Bernard, 2025S]).

A visual analysis of repeated PHQ-9 and GAD-7 scores (see Figure S.1. and S.2. in the Supplementary Materials [Bernard, 2024S-b]) shows that most of the participants (with complete data) had a score below the clinical cut-off at the end of intervention.

Personal Event During Intervention

Eleven participants were available for the short interview at the end of intervention. The reported personal events are presented in the Supplementary Materials (Bernard, 2025S).