The Research Ethics Committee approved this trial on October 22, 2019 (reference: 069-DEFI/068-CES), and the protocol was registered on the US National Library of Medicine (ClinicalTrials.gov) with the identifier NCT04749732 on February 10, 2021. This is a single-center, prospective, two-arm, single-blinded (to patients) randomized controlled trial (RCT) that enrolled participants between March 2021 and July 2022. Participants provided informed consent. The study protocol is published and available online.

Participants were recruited from the outpatient clinic of the Angiology and Vascular Surgery Department of Centro Hospitalar Universitário de Santo António (CHUSA), Porto, Portugal, between January 2021 and July 2022.

Inclusion criteria for the study were: (1) PAD with IC (Fontaine II or Rutherford 1–3) due to atherosclerotic disease and stable IC for more than 3 months; (2) Ankle Brachial Index (ABI) below 0.9 at rest or below 0.73 after exercise (20% decrease); (3) Age range between 50 and 80 years; (4) MWD in treadmill test between 50 and 600 m.

Exclusion criteria were: (1) asymptomatic PAD; (2) critical ischemia (Fontaine III/IV or Rutherford 4–6); (3) previous lower extremity vascular surgery, angioplasty, or lumbar sympathectomy; (4) any condition other than PAD that limits walking; (5) unstable angina or myocardial infarction diagnosed in the last 6 months; (6) inability to obtain ABI measure due to non-compressible vessels; (7) use of cilostazol and pentoxifylline initiated within 3 months before the investigation; (8) active cancer, renal disease, or liver disease; (9) severe chronic obstructive pulmonary disease (GOLD stage III/IV); (10) severe congestive heart failure (NYHA class III/IV); (11) diagnosis of a psychiatric disease that impairs daily life activities and/or with medical records of decompensation episodes in the last year and/or non-adherence to drug therapy; and (12) cognitive impairment (MMSE ≤ 15 for illiterate patients, 22 for those with 1–11 years of schooling; 27 for >11 years).

Participants were randomly assigned in a 1:1 ratio to receive an exercise prescription with a behavior change intervention supported or not by a smartphone app, using a computer-generated randomization system, with randomly selected block sizes of 5 stratified by age (50 to 65 years old; 66 to 80 years old), and maximal walking distance (50 to 325 m; 326 to 600 m) (flow diagram as Supplementary Material S1). It was impossible to mask participants, the outcome assessor, and the psychologist for group allocation after randomization due to the nature of the interventions. The statistician was masked for group allocation.

All participants received standard PAD treatment, a physical exercise prescription, and a behavior change intervention. Standard treatment for PAD and IC followed the guidelines of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines (AHA/ACC) (2, 3, 19, 20).

The physical exercise prescription consisted of walking sessions, performed for at least 30 min per session, three times a week in the area of residence of the participants. Near-maximal pain during training was the outcome of claudication pain (21). The behavior change intervention was based on Self-Determination Theory (SDT), which focuses on the type and quality of motivation, and the satisfaction of basic psychological needs, as pre-eminent behavioral determinants (22–24). According to this theory, individuals become more autonomous or self-determined on a continuum that ranges from external motivation to internalization of motivation. As this continuum progresses, individuals are more likely to engage in new, long-term behaviors. To achieve this, motives or extrinsic reasons for changing and adhering to a new behavior must be internalized and become as intrinsic as possible. Thus, the behavioral change intervention aimed to facilitate this internalization process to nurture a more autonomous motivation. Also, motivational interviewing principles are considered effective in promoting and encouraging change (11, 25, 26). Patients received two face-to-face behavior change sessions, at baseline (T1) and 3 months (T2), conducted by a health psychologist, for a total of 4 h (120 min per session). Complementary, booster phone calls over the 24-week intervention period were used to support participants and identify barriers to goal completion, based on a personalized and individualized self-management approach to promote long-term adherence to the exercise prescription (27, 28).

Medical history and clinical data were collected from the participants' electronic clinical medical records, and sociodemographic data were collected through a clinical interview.

Depressive symptoms were assessed by the Geriatric Depression Scale-5 (GDS), which includes five dichotomous items (yes/no) with higher results corresponding to more depressive symptoms (29, 30).

Anxiety symptoms were assessed by the Geriatric Anxiety Inventory-Short Form (GAI-SF), which includes five dichotomous items (yes/no), with higher results corresponding to more anxiety symptoms (31, 32).

The modified Gardner-Skinner Treadmill Protocol was used, according to which participants began to walk on the treadmill at 1 km/h with a 0% grade (33). After 2 min, the speed increases to 1.6 km/h at 0% grade. Then the speed is increased by 0.8 km/h every 2 min until reaching 3.2 km/h. After reaching 3.2 km/h, the speed is kept constant, and the grade increases by 2% every 2 min.

The 6-Minute Walk Test (6 MWT) was performed to evaluate the functional capacity of the individual to walk over a total of 6 min on a 100 ft (≈30 m) hallway (34) according to the American Thoracic Society guidelines (35).

The primary outcomes were pain-free walking distance (PFWD), functional walking distance (FWD), maximal walking distance (MWD), and 6 MWD at 3 and 6 months, measured by two tests. The treadmill test (33) and the 6-Minute Walk Test (34, 35). The walking distance measured with a standardized treadmill test is a widely used tool for functional assessment and monitoring of exercise rehabilitation, where PFWD, FWD, and MWD are objective measures of improvement. In turn, the 6 MWT is an individualized test that assesses the submaximal level of functional capacity (35, 36). The patient chooses the intensity of the exercise and can rest during the test, which better reflects the functional level of exercise practiced in daily physical activities. The primary outcomes reflect the effectiveness (beneficial effect) and not the harm (adverse effect) of the intervention.

Secondary outcomes were: (1) health-related quality of life (QoL) assessed by the Vascular Disease-Specific Quality of Life Questionnaire (VascuQoL-6, scores ranging from 6 to 24, higher scores indicate better QoL) (36–38); and by the 12-Item Short Form Health Survey (SF-12, includes two summary component scores, higher scores indicate better physical and mental QoL) (39, 40); and (2) the Walking Impairment Questionnaire (WIQ) was used to assess the daily walking ability of patients in three domains: distance, speed, and climbing stairs (values range from 0 to 100%, and higher scores indicate less impairment in walking abilities/performance) (41).

In most RCTs, especially in clinical settings, it is possible to assess the effectiveness of an intervention using a single primary outcome. However, in many situations, a comprehensive understanding of the effects of an intervention requires analysing multiple outcomes. Indeed, patients' health status cannot be fully evaluated using a single outcome, and distinct outcomes may provide different but equally important information about the effectiveness of the intervention (42). Fortunately, developments in statistical methods have allowed researchers to account for multiple outcomes in RCTs (43), especially those that represent the same outcome at different time points, as well as to use suitable families of distributions. In this study, we expected the WalkingPad intervention to improve patients' MWD, PFWD, FWD, and 6 MWD (42, 43). For this reason, these four outcomes were used to define the effectiveness of the intervention and thus, this study comprises these four primary outcomes.

After screening the clinical medical records of patients attending the outpatient clinic of the Angiology and Vascular Surgery Department, 268 patients were identified. Of these, 119 were included and invited to attend a screening evaluation at the hospital to confirm inclusion criteria (at baseline), and 73 were included in the study, randomized, and allocated to one of the two groups (Figure 1). Of these, 60 participated in the three assessment moments. Patients were randomized into two groups: The Control Group (CG)—35 patients (31 men; mean age 64.0 ± 7.16 years) and the Experimental Group (EG)—38 patients (33 men; mean age 63.3 ± 6.73 years). Table 1 shows the sociodemographic and clinical characterization of the sample and between-group differences at the baseline.

There was a significant main effect of time in primary and secondary outcomes. Results of the mixed-effects models show that the average MWD improved significantly over time for the whole sample (from T1 to T2 and from T2 to T3). The average PFWD and FWD increased over time, but from T2 to T3, the increase was insignificant. The same was true for 6 MWD.

Regarding secondary outcomes, the average QoL disease-related increased significantly over time, from T1 to T2 and from T2 to T3. The average physical and mental QoL increased over time, but from T2 to T3, the increase was not significant. The average WIQ distance increased over time, but from T2 to T3, the increase was only borderline significant. The average WIQ speed increased over time, but from T2 to T3, the increase was not significant. The average WIQ stairs increased from T1 to T2 and T2 to T3 over time. Results are shown in Table 2.

By the end of the study, no significant between-group differences regarding the average of the primary outcomes were found. Results are shown in Table 3. Notice, however, that the sample size of this study only allowed to detection of large effect sizes, which means that increasing the sample size would probably allow uncovering of some medium or small effects.

Therefore, we have proceeded with further modeling analysis in order to understand whether the relationships of the primary outcomes with other variables have evolved differently for the two groups over time. First, we performed an internal analysis of the outcome to investigate how baseline scores evolved over time. Second, we conducted an external analysis to understand how other baseline variables affected the outcome as time went by. To perform the group comparison, several multilevel models with three-way interactions of the form “Variable x Group x Moment” have been evaluated. Interestingly, this modeling analysis only detected between-group differences for the MWD outcome. More specifically, unlike the control group, in the EG the MWD has never lost its association with the baseline scores over time (more than that, time strengthened this relationship). Moreover, patients of the EG with no extreme anxiety showed a higher increase in the MWD by the 6-month follow-up, compared with the CG. We emphasize that the sample size of this study only allowed to detect large effect sizes and therefore, all significant effects found in this study are strong. Details of these two models are given below.

Although from the quantitative point of view, there is no significant between-group difference in the MWD improvement (score discounting the baseline MWD) over time, after controlling for participants' differences at the baseline, it is possible to uncover some important differences. Results are shown in Table 4. More specifically, the predictive power of participants' baseline MWD over time was significantly different for the two groups. In the CG, this predictive power did not vary significantly over time; however, in the EG, it increased and became significantly higher at T2 (estimate difference of 4.89, p = 0.013) and T3 (estimate difference of 4.67, p = 0.024, at T3), when compared to the CG. This means that in the EG, a part of the improvement is explained by participants' baseline MWD, in contrast to the CG, where this causal relationship was not observed. Thus, participants from the EG improved a proportion of their baseline MWD and, as a result, the higher the baseline MWD, the higher the improvement achieved afterward. This between-group difference is particularly prominent in the 6-month follow-up, with an improvement of 5.17 (p = 0.003) in the EG and 0.44 (p = 0.779) in the CG. Thus, the baseline MWD only predicts the MWD improvement at T3 in the EG. Notice that this is a qualitative between-groups difference in the MWD over time.

The predictive power of MWD was intensified over time, meaning that participants from the EG benefited from the WalkingPad app, except those with a weak walking ability (Figure 1). It also indicates that the MWD in the EG, unlike in the CG, exhibited strong stability over time.

Furthermore, anxiety (fixed at baseline) had a significant between-group different effect on MWD (depression also showed a similar result, but it disappeared when controlling for anxiety). Except for patients with extreme anxiety symptoms at the baseline, the EG showed a higher increase in the MWD at T3 (increase difference = 0.91, p = 0.026). For patients with extreme anxiety symptoms at the baseline, the increase from T1 to T3 was not significantly different for the two groups (increase difference = 0.75, p = 0.137) (Figures 2, 3).

The WalkingPad app may be recommended to PAD patients, except for those with weak walking ability and extreme anxiety symptoms. We hypothesized that the app can trigger more anxiety in people who are already anxious. Thus, in clinical practice, the health professional must assess the patient's anxiety to understand whether the physical exercise prescription should involve the app or not.

The 6-month duration is crucial to achieving results in maximal walking distance and QoL, as well as the behavioral change intervention that seems to have been the essential driver for promising results. Thus, this is the first study with a 6-month follow-up that achieved promising results in walking distances and quality of life. The behavior change intervention carried out in person and by telephone during the 6 months was decisive for the effectiveness of the physical exercise program, considering the importance of behavior change for the consistent adoption of a healthy habit. This study also shows that physical exercise programs should be at least 6 months in duration because disease-related QoL increased significantly over time, but overall mental and physical QoL only increased from T1 to T2, suggesting that the effect of WalkingPad program on QoL was progressive and not immediate. Furthermore, this study suggests that a smartphone app may not be useful for all patients. In fact, the app may be recommended for people with medium to high baseline MWD scores, unless they have extreme anxiety symptomatology, and for those who still have good walking skills (high MWD) at baseline, associated with a stronger positive effect over time.

This study has some limitations that must be recognized. The sample was collected in only one hospital, being a unicentric study, and the results cannot be generalized to structured and supervised interventions (SET). Furthermore, the lack of gender heterogeneity in the sample (larger number of male patients) and the single-blind nature of the study, when the ideal was to be a double-blind study, must be recognized as limitations.

The sample size was small. Data were analysed according to the most recent statistical recommendations for longitudinal data, namely, using multilevel (or mixed-effect) models with suitable families of distributions. Several models have been investigated and we only report those (two) that were able to detect significant effects. Models are like microscopes whose augmenting power is defined by the sample size, with larger size providing greater power. We emphasize that the sample size of this study only allowed the detection of large effect sizes (increasing the sample would probably detect more effects), and therefore, all significant effects found in this study are strong. For this reason, we believe that our results will be easily corroborated by future studies, as long as they use the same conditions and the same statistical methodology (R codes used in this project can be provided on formal request).

The study lasted only 6 months, and the durability of the program's effects was unknown. Thus, more research is needed to determine the durability of these findings. An exploratory and usability study must be carried out to understand which application functionalities should be improved to promote the other results, and the same protocol should be applied, including more intense and personalized psychological follow-up for more anxious people. Furthermore, future studies should test the differences between gender in the use of the app and how this use influences adherence to physical exercise and outcomes.