The study was a parallel-group, randomized trial with 52 weeks of intervention and subsequent effectiveness follow-up a year after baseline. All participants provided oral and written informed consent prior to commencing any study procedures. The original protocol has been published [23], while the prespecified statistical analysis plan is available as Multimedia Appendix 1. Reporting is in accordance with the CONSORT statement.

The study was approved by the scientific ethical committee of the Capital Region of Denmark (H-1-2014-074) and registered at clinicaltrials.gov (NCT02341690).

Inclusion criteria were T2D diagnosis, ≥18 years of age, and referral to a municipality health promotion center or hospital in the participating municipality by the individual’s general practitioner. Exclusion criteria were medical contraindications to exercise, for example, chronic complications in the musculoskeletal system, painful osteoarthritis, or heart conditions [25]; declining to participate in an exercise program at the health promotion center or hospital; current participation in other intervention studies; or insufficient Danish language skills. All individuals who met for an appointment at the participating health promotion centers or hospital were screened through medical records and at a screening interview with a health professional at the center or hospital.

The data were collected at the participating health promotion centers (Amager, Vanløse, Østerbro, Vesterbro in the Municipality of Copenhagen, Denmark; and Municipality of Guldborgsund, Denmark) and a hospital (Bornholm, Denmark).

Participants were randomly allocated (2:1) into 1 of 2 arms: (1) standard care + InterWalk app–based IWT (IWT group) or (2) standard care + the standard exercise program (StC group). Following the initial 12-week supervised exercise program, the participants in the IWT group were further randomly allocated (1:1) into (1) IWT, no additional support (IWT_only group) or (2) IWT, with additional support (IWT_support group), that is, participants in the IWT_only group and the IWT_support group underwent similar interventions during the 12-week exercise program, and allocation to either of these was concealed until after the 12-week intermediate assessment.

Participants were randomized using random permuted blocks stratified by sex (2 levels) and center (6 levels). The allocation sequence was generated through a standardized computer program by an independent statistician (RC) and stored on a password-protected computer by an independent researcher (RN) who was not involved in any study procedures. Following the completion of all baseline measurements, the independent researcher was contacted and performed allocation (to StC or IWT_only or IWT_support). The respective group allocation for the initial 12 weeks (StC or IWT) was subsequently returned by email to the health professional who informed the participant about the allocation by telephone call. Information about IWT_only or IWT_support was not disclosed by the independent researcher to the health professionals until week 12. Following the 12-week intermediate assessment, the independent researcher was contacted and the allocation into IWT_only or IWT_support was returned by email to the health professional who informed the participant about the allocation by telephone call. The health professionals carried through the data collection and intervention, and thus, were only blinded to the primary outcome.

Primary and key secondary outcome measurements are described in further detail in the published protocol [23]. Outcome measurements were conducted by trained health professionals at the respective center or hospital.

The primary outcome measure was change in objectively measured MVPA time (minutes/day) from baseline to 52-week follow-up. PA outcomes were assessed using accelerometers (AX3; Axivity) attached to the participants’ thigh and back using an adhesive tape (Fixomull Stretch; BSN Medical Inc.) and worn for 7-10 consecutive days at baseline, 12 weeks, and 52 weeks. Accelerometer setup and download of raw data (100 Hz) and downsampling to 30 Hz were performed using OmGui (version 1.0.0.28) [28]. The data were then exported to ActiGraph raw binary files (gt3x files) and resampled into counts (agd files) using ActiLife (version 6.11.6). The final data reduction and generation of PA outcome variables were done using a custom-built software developed at the University of Southern Denmark [29]. MVPA time was defined according to the Freedson cut point, ≥1952 counts per minute (CPM) [30], using the vertical axis of the accelerometer placed on the back. Participants were included in the analyses of objectively measured PA if they had ≥3 days of ≥22 hours of measurement (ie, allowing for a 2-hour nonwear time [31]). Accelerometer wear time and nonwear time are presented in Multimedia Appendix 2.

Key secondary outcome measures include changes in physical and mental HRQoL, physical fitness (peak oxygen consumption [VO_2peak]), self-rated PA energy expenditure (PAEE), exercise motivation, weight, and waist circumference, which were assessed at baseline, 12 weeks, and 52 weeks. Physical and mental HRQoL were assessed using the 12-item Short-Form Health Survey (SF-12), including 8 subscales [32]. The Physical Component Summary (PCS; score 0-100) was calculated based on the 4 subscales: Physical Functioning, Role Physical, Bodily Pain, and General Health; and the Mental Component Summary (MCS; score 0-100) was calculated based on the 4 subscales: Vitality, Social Functioning, Role Emotional, and Mental Health [32]. VO_2peak (ml O₂/minute) was estimated by regression based on the participants’ weight, height, sex, and the acceleration (G) during the last 30 seconds of fast walking in the 7-minute standardized walking test in the InterWalk app [33]. Self-rated PAEE was assessed using the Recent Physical Activity Questionnaire (RPAQ) [34]. Exercise motivation was assessed using the Relative Autonomy Index (RAI) for the Behavioral Regulation in Exercise Questionnaire-2 (BREQ-2) [35]. Weight (kg) was measured using an electronic weight, and waist circumference (cm) was measured midway between the most distal part of the costae and the most proximal part of the iliac crest.

Exploratory secondary outcomes were changes in objectively measured light-intensity PA (100 to <1952 CPM) time (minutes/day), total PA level (CPM), and steps (numbers/day) [30,36] assessed by accelerometers worn on the participants’ back as described above. Sitting time (minutes/day) was derived from the accelerometer placed on the thigh as described elsewhere [31]. BMI (kg/m²) was calculated based on weight and height. Any adverse events were reported by the participants to the health professionals at the center or hospital.

The minimal important difference was considered to be 10 minutes of MVPA per day. Based on existing experimental evidence, we assumed the SD of the change in MVPA time from baseline to 52-week follow-up to be between 1.2 and 2.3 times the difference between groups [37]; accordingly, SD twice the minimal important difference (20 minutes of MVPA per day) was used in the sample size calculations. To maintain a conservative sample size, considering multiple comparisons in a 3-arm trial, Bonferroni adjustment was used in the power calculation. A total of 190 participants were required to achieve a statistical power (1-β) of 80% with an α level of .017 (0.5/3) using an unpaired t test (2-sided). Allowing for 30% attrition, 272 participants (91 in the StC group and 181 in the IWT group) should be recruited. The intervention settings enabled recruitment until December 15, 2016, and thus the sample size would include 272 participants or truncated at the number of participants included at the end of the recruitment period—whichever was reached first. On December 15, 2016, 214 participants had been included and they constituted the final analysis population.

A statistical analysis plan was developed and published at the Centre for Physical Activity Research website prior to commencing the statistical analyses [38]; see Multimedia Appendix 1. All continuous outcomes were analyzed using repeated-measures mixed linear models with the baseline value of the relevant variable as a covariate, including fixed effect factors for group 1 (2 levels: IWT/StC), group 2 (3 levels: IWT_only, IWT_support, and StC), time (3 levels: baseline, 12-week, 52-week), and group 1 × time interaction; and random effects (patient ID). The analyses were adjusted for the stratifying factors sex (2 levels: male and female) and center (6 levels: Amager, Vanløse, Østerbro, Vesterbro, Guldborgsund, and Bornholm). The following model assumptions were investigated: (1) linearity, (2) normality of residuals, (3) homogeneity of residuals variance, and (4) independence of residual error. Data were analyzed based on the intention-to-treat population (no imputation for missing data) as observed, including all randomized participants with available data at baseline; repeated measures mixed linear models were valid assuming that data were “missing at random.” Results are reported as least squares means (LS means) for each group and the difference between them with 95% CIs. Analyses of categorical outcomes for dichotomous endpoints were performed using the Fisher exact test and reported based on the observed proportions.

To investigate the robustness of the primary analyses, 5 sensitivity analyses were conducted, including 4 analyses of differences in changes in MVPA time from baseline to 52-week follow-up between the IWT group and the StC group: (1) per-protocol analysis, including complete cases of the primary outcome at baseline and 52-week follow-up with ≥70% of the prescribed exercise (3 sessions per week for 52 weeks) completed in the IWT group and no registered IWT sessions in the InterWalk app in the StC group; (2) analysis including all randomized participants with available data at baseline with a conservative single-imputation nonresponder imputation technique (ie, missing data replaced with the baseline observation carried forward); (3) analysis with the subsample of participants not undergoing intervention during a period of InterWalk app malfunctions connected to a major restructuring of the iOS (version 9; available during the period from September 16 to October 21, 2016); and (4) analysis with daytime criteria for inclusion, that is, ≥3 days of ≥14 hours of daytime measurement (6 AM to 22 PM; ie, allowing for 2 hours of nonwear time during this period). In the fifth sensitivity analysis, we used an alternative cut-off at 3000 CPM to investigate changes in forced walking time from baseline to 52-week follow-up between the IWT group and the StC group.

Subgroup analyses were conducted to investigate the effects of IWT versus StC on change in MVPA time (minutes/day) after 52 weeks among subgroups of sex (men/women); T2D duration (≤5 years/>5 years); alcohol consumption (within recommended levels/above recommendations); smoking habits (smoker/nonsmoker); highest level of education (International Standard Classification of Education 2011 [ISCED-2011] levels 0-4/ISCED-2011 levels 5-8); civil status (single, divorced, or widowed/married or cohabiting); baseline SF-12 PCS level (high/low); and baseline SF-12 MCS level (high/low). SF-12 PCS and MCS were divided into high and low using median split. Moreover, differences in median (25th to 75th percentile) duration of IWT (minutes/week) over 52 weeks were investigated among these subgroups.

All reported 95% CIs and P values are 2-sided. Because more than 1 comparison was made, key secondary outcome analyses were controlled for the false discovery rate using the Benjamini–Hochberg step-up procedure [39]. All statistical analyses were performed using Stata/IC 13 (StataCorp).

Between January 2015 and December 2016, 762 individuals were screened for inclusion (Figure 1). Of these, 548 were excluded primarily due to declining to participate (n=209) or physical challenges (n=100). Of the 214 participants enrolled in the study, 74 participants were allocated to the StC group and 140 participants to the IWT group. A total of 130 participants completed the 52-week follow-up, which was finalized in December 2017. Participant demographic and clinical characteristics are presented in Table 1. The mean age of participants was 59.6 (SD 10.6) years and 128/214 (59.8%) were men. They had a mean BMI of 34.6 (SD 6.6) kg/m², a median (25th-75th percentile) T2D duration of 2.6 (0.2-8.6) years (Table 1). At baseline, participants spent a median (25th-75th percentile) of 30.2 (16.9-48.1) minutes/week in MVPA (Table 1).

The IWT group engaged in IWT with a median (25th-75th percentile) frequency of 0.3 (0.1-0.7) sessions/week for 11.4 (4.1-22.3) minutes/week at a median intensity of 0.17 (0.11-0.23) G (Multimedia Appendices 3 and 4). For the subgroup that completed the 7-minute standardized walking test in the InterWalk app at baseline, this corresponded to an average intensity across sessions of 65% (SD 49-78) of their maximal intensity at baseline. According to self-reported IWT adherence, participants in the IWT_support group used the InterWalk app 1-2 times/week or more in most weeks; the primary reason for not using the InterWalk app was illness (Multimedia Appendix 4). For the IWT_support group, 148/188 (78.7%) motivational interviews were completed during follow-up (Multimedia Appendix 5). However, only 27/47 (57%) participants completed all 4 interviews and the adherence to the Danish Diabetes Association walking groups was low as only 5-6/47 (11%-12%) participants participated in this activity. Only a limited number of participants included IWT as a specific part of the goals during follow-up (18/47, 38%).

No change in MVPA time was observed from baseline to 52-week follow-up within the StC and IWT groups (P=.81 and P=.91, respectively; Figure 2) and no between-group difference was observed (–0.8 minutes/day; 95% CI –8.1 to 6.4 minutes/day; P=.82; Table 2). From baseline to 12-week follow-up, MVPA increased by a mean of 3.6 (95% CI 0.2 to 6.9) minutes/day in the IWT group and was unchanged in the StC group. However, no difference was observed between the groups (Multimedia Appendix 6).

Only 5 participants in the IWT group fulfilled the per-protocol criteria and the analysis was thus omitted. The remaining sensitivity analyses supported the primary analysis (Multimedia Appendix 7).

From baseline to 12- and 52-week follow-up, SF-12 PCS score (ie, physical HRQoL) increased by a mean of 4.3 (95% CI 3.1 to 5.6) points and 4.2 (95% CI 3.0 to 5.4) points, respectively, in the IWT group and remained unchanged in the StC group. Thus, the SF-12 PCS score increased by a mean of 3.7 (95% CI 1.2 to 6.1) points and 4.3 (95% CI 1.8 to 6.9) points (Benjamini–Hochberg adjusted P=.007) more in the IWT group compared with the StC group at 12- and 52-week follow-up, respectively (Multimedia Appendix 6 and Table 2). According to post hoc linear regression analysis, a change in MVPA and IWT duration was not associated with a change in SF-12 PCS over 52 weeks (Multimedia Appendix 8).

Waist circumference decreased by a mean of –3.1 (95% CI –4.0 to –2.2) cm from baseline to 52-week follow-up in the IWT group with no change in the StC group (Table 2). Thus, waist circumference apparently decreased by a mean of –2.3 (95% CI –4.1 to –0.4) cm more in the IWT group compared with the StC group (Benjamini-Hochberg adjusted P=.06; Table 2). From baseline to 12-week follow-up, waist circumference decreased by a mean of –2.9 (95% CI –3.7 to –2.0) cm and –3.2 (95% CI –4.5 to –1.9) cm in the IWT and StC group, respectively. No differences were observed between the groups (Multimedia Appendix 6). According to a post hoc linear regression analysis, every 10 minute/day increase in MVPA was associated with a mean decrease of 0.6 (95% CI –1.1 to –0.1) cm in waist circumference (r=–0.21; P=.03), while every 10 minute/week increase in IWT duration was associated with a mean decrease of 0.6 (–1.1 to –0.2) cm in waist circumference (r=–0.30; P=.008; Multimedia Appendix 8).

No differences in the changes in SF-12 MCS (mental HRQoL), VO_2peak, RPAQ self-rated PAEE, BREQ-2 RAI, and weight from baseline to 12- and 52-week follow-up, respectively, were observed between the groups (Multimedia Appendix 6 and Table 2).

Among the exploratory secondary outcomes, no between-group differences were observed from baseline to 12- and 52-week follow-up, except for light-intensity physical activity time and steps that increased from baseline to 12-week follow-up by a mean of 20.1 (95% CI 7.5 to 32.7) minutes/day and 1124 (95% CI 255 to 1992) steps/day, respectively, more in the IWT group compared with the StC group (Multimedia Appendix 6 and Table 2).

From baseline to 52-week follow-up, there were no differences between the IWT_only and IWT_support groups in the changes in primary, key secondary, and exploratory secondary outcomes (Multimedia Appendix 9). Moreover, no subgroup effects of IWT versus StC on the change in MVPA time were observed (P>.1 for interaction; Figure 3), nor were there any differences in mean IWT duration over 52 weeks among the subgroups in the IWT group (Multimedia Appendix 10).

No adverse events or harms were reported to the health professionals.