The study was approved by the Danish National Committee on Health Research Ethics (H-18011067) and the Danish Data Protection Agency (P-2019-512) and registered at ClinicalTrials.gov (NCT03679130). The study adheres to the principles of the Helsinki declaration. Written informed consent was obtained at inclusion.

The development of FitMum was inspired by stakeholders: 27 semistructured interviews with Danish pregnant women, midwives, and obstetricians were performed to explore the feasibility, facilitators, and barriers to PA during pregnancy.

FitMum was a single-site randomized controlled trial (RCT) conducted from 2018-2021 at the Department of Gynecology and Obstetrics at Copenhagen University Hospital–North Zealand, Denmark [32]. A total of 220 healthy, inactive pregnant women with gestational ages of ≤15 weeks and 0 days were included (visit 1). Participants were randomized 1:2:2 into CON, EXE, and MOT groups, respectively (Figure 1). Participants were invited to a test visit at the 29th gestational week (visit 2) and the 35th gestational week (visit 3).

All 3 groups were offered standard maternal care. The EXE group was offered 1-hour group-based supervised exercise training at moderate intensity 3 times per week in a gym and swimming pool. The MOT group was offered 4 individual and 3 group PA motivational counseling face-to-face sessions of 1 to 2 hours duration during pregnancy and 1 weekly, personalized text message to support PA. The motivation technique applied is inspired by motivational interviewing [34], self-determination theory [35], and behavior change techniques [36].

Interventions ran from randomization until delivery. The target PA level for the EXE and MOT groups was at least 30 minutes per day at a moderate intensity as recommended in Denmark to healthy pregnant women [37]. Interventions were converted into online versions during the COVID-19 pandemic restrictions introduced in Denmark on March 11, 2020, and throughout the study period. The EXE group could access the swimming pool for 3 months during this period.

The data collection procedures are illustrated in Figure 2. PA was continuously monitored by the activity tracker from randomization to delivery, by PPAQ at visits 1, 2, and 3, and by DLW at visit 2.

The primary outcome was MVPA (minutes per week) from randomization to visit 2. PA was from inclusion to delivery continuously captured by a wrist-worn commercial activity tracker (Vivosport, Garmin International) [38] with a built-in heart rate monitor and accelerometer. Baseline PA was captured from inclusion to randomization (6 full days). PA with a metabolic equivalent of task (MET) value of ≥3 in bouts of at least 10 consecutive minutes was recorded automatically as MVPA by the activity tracker [38]. Secondary outcomes measured by the activity tracker were PA duration at moderate and vigorous intensities; steps; active time; active kilocalories; floors climbed; and minimum, maximum, resting, and average heart rate from randomization to delivery. At inclusion, the activity tracker was preset with PA notifications turned off and an identical face of the tracker showing only clock and battery level. After randomization, women in the MOT group were encouraged to personalize the tracker with, for example, individual goal settings and PA notifications as part of the intervention. Interaction with the tracker was neither encouraged nor controlled for the EXE and CON groups. Throughout the study period tracker software was automatically updated [38].

PA was digitally self-reported by participants using the PPAQ-DK [28] at visits 1, 2, and 3. The questionnaire assesses PA related to everyday activities during the current trimester (eg, household, occupational, sports, and transportation) [27].

Participants collected 2 baseline urine samples prior to visit 2, drank the DLW dose at the visit, and then collected and stored 5 postdose urine samples at home on days 1, 4, 7, 11, and 14 and later at −80°C. [31,39]. The calculation of total energy expenditure (TEE) was based on the Weir equation [39], and the active energy expenditure (AEE) was calculated by subtracting the basal metabolic rate (BMR) from the TEE. BMR was estimated by an equation appropriate for pregnant women [40]. PA level (PAL) was calculated by dividing TEE by BMR.

PA was transferred via Bluetooth from the activity tracker to the Garmin Connect app (Garmin International) [38] from which Fitabase (Small Steps Labs LLC) obtained the data via the programming interface. PA was monitored through Fitabase, and participants were reminded if the activity trackers were not synchronizing. PA data were downloaded from Fitabase, processed, and cleaned in R software (R Foundation for Statistical Computing).

Statistical analyses were performed according to the statistical analysis plan, which includes a sample size calculation [33] using R. Data are presented as means and standard deviations for symmetric distributions, medians and IQRs for skewed data, and frequencies and percentages for categorical variables. The level of statistical significance was 5% except for the primary hypothesis which consisted of 2 subhypotheses; the type I error for each hypothesis test was a priori set to 2.5% to obtain a family-wise error rate of 5%. Wald-based 95% CI were given for all reported estimates [33]. Intention-to-treat analyses using all randomized participants were performed for the primary outcome. Missing observations in tracker data due to nonwear time were imputed by multiple imputations in 25 data sets using a prespecified seed, preselected baseline variables (body weight, age, PA, educational level, and parity), and the random forest imputation model from the mice R package [41]. A statistician blinded for the intervention performed the imputation and the primary outcome analysis as an analysis of covariance model adjusting for baseline PA. MVPA before and during the COVID-19 pandemic was compared within groups with a linear regression model. Cumulative trajectories were estimated from the imputed data using a generalized additive model with a penalized regression spline with point-wise 95% confidence bands estimated by a bootstrap procedure [42]. For the PPAQ-DK outcome, a constrained linear mixed model was fitted with the observation times as a factor [43]. Both within and between-group effects were reported as estimated differences in means. For the DLW outcome, a one-way analysis of variance was used to compare the 3 group averages. For the DLW outcome, a 1-way analysis of variance was used to compare the 3 group averages. Linear regression was used to model the relationship between attended intervention sessions and attained MVPA in the EXE and MOT groups.

In total, 220 pregnant women were included from October 2018 to October 2020. Of those, 219 were randomly allocated to CON (45/219), EXE (87/219) or MOT (87/219; Figure 1). Maternal baseline characteristics are presented in Table 1.

From randomization to visit 2, 15.1% (33/219) of participants were lost to follow-up (CON: 10/45, 22%; EXE: 10/87, 11%; MOT: 13/87, 15%). The main reason (18/33, 55%) was personal matters (eg, time consumed with participation or family reasons). From randomization to delivery, 18.7% (41/219) of participants were lost to follow-up, and proportions were similar across groups (Figure 1).

Participants randomized to EXE participated in 1.4 (95% CI 1.2 to 1.6) exercise sessions per week from randomization to visit 2, and 1.3 (95% CI 1.1 to 1.5) exercise sessions per week from randomization to delivery. Participants randomized to the MOT group joined 5.2 (95% CI 4.7 to 5.7) counseling sessions during their pregnancy.

All tracker outcomes are presented in Figure 3 and accompanying statistics in Multimedia Appendix 1. PA at a vigorous intensity (minutes per week) was higher in EXE than in both CON and MOT (CON vs EXE: randomization to visit 2: 13, 95% CI 4 to 22; randomization to delivery: 13, 95% CI 4 to 22; MOT vs EXE: randomization to visit 2: 9, 95% CI 1 to 16, randomization to delivery: 9, 95% CI 1 to 17). In addition, the maximum heart rate was 2 (95% CI 0.3 to 3) beats per minute higher in EXE compared with CON from randomization to visit 2. No other tracker outcomes differed between groups.

PPAQ-DK was completed for visits 1, 2, and 3 by 100% (219/219), 83.1% (182/219), and 77.2% (169/219) participants, respectively. Figure 6 shows the PA behaviors categorized by intensity and type. Differences between and within groups are shown in Multimedia Appendix 3 and Multimedia Appendix 4.

Total activity did not change from visit 1 to visit 2 in CON, EXE, or MOT, but PA decreased significantly from visit 1 to visit 3 in all groups (Multimedia Appendix 4). PA at moderate intensity was maintained at the same level over the course of pregnancy in CON, EXE, and MOT. However, participants in MOT increased PA at vigorous intensity from visit 1 to visit 2 and visit 1 to visit 3 (Multimedia Appendix 4). When combined (MVPA), the activity level (MET hours per week) did not change through pregnancy in any of the groups (CON: visit 1-2: –1, P=.90; visit 1-3: –4, P=.36; EXE: visit 1-2: 4, P=.10; visit 1-3: 1, P=.61; MOT: visit 1-2: 2, P=.40; visit 1-3: –5, P=.37; data not shown).

The MET hours per week spent at sports activities increased significantly from visit 1 to visit 2 and visit 1 to visit 3 for both EXE and MOT, while no changes were observed in CON (Multimedia Appendix 4). A comparison between groups revealed that sports was significantly higher in EXE compared with CON and MOT at both visit 2 and visit 3 (Multimedia Appendix 3).

A total of 134 participants (CON: 24/45, EXE: 53/87, MOT: 57/87) completed the DLW test and were included in the analysis. TEE (P=.14), AEE (P=.38), and PAL (P=.66) did not differ between groups (TEE [kcal per day]: CON 2215 [SD 238], EXE 2330 [SD 264], MOT 2331 [SD 260]; AEE [kcal per day]: CON 543 [SD 106], EXE 592 [SD 160], MOT 587 [SD 155]; and PAL [TEE/BMR]: CON 1.33 [SD 0.06], EXE 1.35 [SD 0.11], MOT 1.34 [SD 0.09]; Multimedia Appendix 5).

Adverse events and serious adverse events from inclusion to delivery among all participants did not differ between groups (Multimedia Appendices 6-8).

FitMum aimed to investigate the effects of offering EXE or MOT to generate evidence about how to implement PA in healthy pregnant women’s lives. We hypothesized that both EXE and MOT would increase MVPA in pregnancy compared with CON but that EXE would be more effective than MOT [33]. The study confirmed that EXE was more effective than CON, whereas MOT was not more effective than CON, and EXE and MOT did not differ. The number of adverse events and serious adverse events did not differ between groups.

Several previous RCTs have used strategies like ours to examine how to increase PA in pregnant women and at the same time assessed the PA level by objective methods [13,24,26,44,45]. Seneviratne et al [24] conducted a 16-week stationary biking program in overweight and obese pregnant women and reported improved aerobic fitness compared to controls. When determining PA objectively by accelerometry, Hayman et al [26] found an immediate increase in MVPA after 4 weeks of tailored PA advice and access to a resource library. On the contrary, no increase in PA as determined by accelerometry was found after a combined aerobic and strength exercise program [44], face-to-face individual PA consultations [13], or app-based PA behavior change techniques [45].

Women in EXE were encouraged to participate in 3 hours of EXE per week, but the participants attended on average less than half of the sessions, and throughout their pregnancy, the MVPA level was only a third (54 of 150 minutes per week) of the internationally recommended amount [2]. As expected, MVPA was positively associated with the number of exercise sessions attended. Noticeably, EXE had a higher level of vigorous intensity PA compared with both CON and MOT. This was supported by a higher maximum heart rate among EXE. Exercising at vigorous intensity is in accordance with recent suggestions for healthy pregnant women [46,47]. MOT had a high intervention attendance, but even though MOT contained face-to-face counseling, text messaging, activity tracker use, and behavior change techniques as recommended [13,48,49], we found no effect on MVPA compared with CON. The processes behind this finding are currently being assessed via mixed methods. The cumulative MVPA in EXE was significantly higher compared with MOT in the late part of pregnancy, and the same tendency was seen between CON and EXE. Interestingly, women who received the online EXE intervention due to COVID-19 restrictions joined 45% more exercise sessions compared with those who received the physical intervention.

Combining 3 different methodologies to assess PA using objective (activity tracker and DLW) and subjective (PPAQ-DK) methods provides insight into the complexity of PA. The activity tracker offers 24/7 measures of PA, and due to its convenience the tracker can be worn for a long period of time. However, commercial trackers are not designed for research purposes, and tracker algorithms are unknown. The PPAQ is considered one of the most valid and reliable questionnaires for the assessment of PA in pregnant women [27,50], but the inherent bias of self-reported PA is inevitable. The administration of the PPAQ-DK may have led to a heightened awareness of activity among participants [50], especially for members of the MOT group, who received a thorough review of their PA level at the counseling sessions. This might explain the perceived increase in vigorous intensity PA in MOT as determined by PPAQ-DK. DLW is the reference method for the determination of free-living energy expenditure and has previously been used to estimate PA level in pregnant women [39,51], but this is the first intervention study in pregnant women to include DLW. We found no significant differences between groups in TEE, AEE, or PAL, but this might be due to a lack of power, as TEE and AEE were 50 to 100 kcal per day higher in EXE and MOT compared with CON. On the other hand, active kilocalories recorded by the tracker and total activity obtained from the PPAQ-DK, which are equivalent to AEE from DLW, did not differ between groups. Therefore, the total activity probably did not differ between groups.

FitMum is the first RCT to compare the effectiveness of 2 different PA interventions in pregnant women. Strengths comprise the robust design based on the power of randomization, which leaves the internal validity high, and the comprehensive assessment of PA. The primary outcome was measured by a commercial activity tracker, which measured PA continuously, but no data on the validity of the tracker activity measurements has been published. The activity tracker may increase PA due to its motivational impact [49,52], but it might also not capture all activities. Notably, by default the tracker only reported activities with a MET value of ≥3 in bouts of at least 10 consecutive minutes as MVPA [38], and this might partly explain the relatively low MVPA in this study. An additional limitation was the impact of COVID-19 and the need to convert the physical interventions into online ones.

Findings from this RCT demonstrate that offering EXE is more effective than CON to implement MVPA in healthy pregnant women’s lives. Offering MOT was not more effective than CON; EXE and MOT also did not differ. The MVPA in the intervention groups did not reach the recommended PA level in pregnancy. Changing the intervention to online due to COVID-19 restrictions did not affect MVPA level but increased exercise participation. Based on the most effective intervention on MVPA during pregnancy (EXE) and the increased level of EXE sessions attended in the online setup during the COVID-19 pandemic, it might be beneficial to add home-based, online exercise sessions in future prenatal PA interventions.