Self-monitoring of dietary intake is a cornerstone of behavioral weight loss treatment [1], and past research has demonstrated that the frequency of self-monitoring is positively associated with weight loss [2]. Despite its utility, dietary self-monitoring typically declines over the course of treatment [2,3]. Novel strategies are needed to improve dietary self-monitoring engagement [4]. One strategy involves enriching self-monitoring with other theoretically and empirically supported behavior change techniques such as tailored goals and feedback, action plans, and skills training [5,6].

A second strategy includes building mastery, self-efficacy, and self-regulation—key constructs of behavior change in Carver’s Control Theory [7] and Bandura’s Social Cognitive Theory [8]— before asking participants to engage in dietary self-monitoring. Fostering self-regulatory skills may provide an opportunity for mastery and, in turn, strengthen self-efficacy [9], which has been linked to greater weight reduction [10]. Behavioral weight loss interventions typically involve self-monitoring multiple behaviors or outcomes simultaneously during treatment [2,11,12], which can serve as an efficient strategy for producing behavior change but may detrimentally impede performance on each item or result in greater treatment dropout [13,14]. Failing to develop mastery of a behavior can weaken self-efficacy [15], leading to worse treatment outcomes. In addition, when people begin to self-monitor a behavior less frequently, it is likely that they will also self-monitor a corresponding behavior less frequently, as demonstrated in an analysis of self-monitoring patterns [16].

We propose a novel solution that aims to attenuate the decline in engagement by employing a sequential [17] self-monitoring approach, wherein individuals track only body weight for a period of time and then begin to track diet. Tracking body weight was chosen as it requires minimal effort, provides an opportunity for habit formation (eg, track every morning upon waking), and is efficacious for weight loss [18]. We focused on only self-monitoring of body weight during the first month based on prior research demonstrating that enhanced engagement in the first month of treatment may have long-lasting repercussions [19]. Sequential approaches are based on the premise that mastery is more easily obtained if only 1 new behavior is targeted at a time [20,21]. Recent research finds that both sequential and simultaneous approaches are effective for behavior change [21] and that focusing on a single component instead of a multicomponent intervention produces comparable weight loss [22], suggesting that a simpler approach is sufficient to produce behavior change.

We tested this sequential strategy using a remotely delivered intervention that utilizes a popular commercially available mobile app—MyFitnessPal. Utilizing technology for self-monitoring dietary intake has been shown to produce greater adherence and less-pronounced declines in engagement than traditional paper-based tracking methods [23,24]. As demonstrated in recent reviews [25-27], smartphone apps can produce significant weight loss, although most existing trials that use commercial apps lack fully powered designs or are pilot studies. Interventions without counseling that utilize commercial technology for dietary self-monitoring have produced clinically meaningful weight losses between 2.5 kg and 5.5 kg at 3 months and beyond in recent studies [28-32].

In the current randomized trial, we test a weight loss intervention among adults with overweight or obesity that targets the aforementioned strategies of including empirically supported behavior change techniques, promoting mastery and self-efficacy through self-monitoring of body weight before diet, and utilizing a free, commercially available app (MyFitnessPal). We hypothesize that a sequential approach will produce greater weight loss and self-monitoring engagement at 3 months compared with a traditional simultaneous approach and to an “off-the-shelf” app.

GoalTracker was a 3-arm randomized controlled trial comparing 3 stand-alone weight loss interventions: (1) a Simultaneous self-monitoring arm in which participants simultaneously tracked body weight and dietary intake each day and received additional empirically supported behavior change techniques (see Table 1) via email for the entirety of the intervention, (2) a Sequential arm, consisting of identical intervention components but allowing for mastery of 1 skill (ie, self-monitoring of body weight) before beginning self-monitoring of diet, and (3) an App-Only arm that tracked only diet with no additional behavior change components. Study evaluation visits were held at baseline, 1 month, and 3 months. Self-reported weight was collected at 6 months. All study procedures were approved by the Duke University Institutional Review Board (protocol #: D0822; date of approval: 10/14/16).

Inclusion criteria comprised men and women aged 21 to 65 years with a body mass index (BMI) between 25.0 and 45.0 kg/m² who were interested in losing weight through dietary change. We required participants to have an iPhone or Android smartphone, email address, access to a bathroom scale, and written English fluency. Participants needed to be willing to download the mobile app on their phone and not track diet or body weight using any other modality (eg, other health or weight tracking apps, websites, and paper diaries) for the duration of the intervention. We excluded participants if they were enrolled in another weight loss intervention, had used MyFitnessPal to track diet in the past 6 months, had lost ≥10 lb or used a weight loss medication in the past 6 months, had previous or planned bariatric surgery, or if weight loss would be contraindicated (eg, pregnancy or <12 months postpartum or in need of medical or psychiatric intervention such as for cancer diagnosis, eating disorder, uncontrolled hypertension, diabetes mellitus, cardiovascular event, or congestive heart failure). A total of 2 criteria were amended during the trial recruitment to promote generalizability of findings: the BMI criteria were expanded to include participants in the 40.0 to 45.0 kg/m² range, and the weight change criteria were adjusted to no longer exclude individuals who gained more than 10 lb in the past 6 months. The institutional review board approved both amendments.

Recruitment occurred between April and September 2017 in central North Carolina via a university-affiliated research website and listservs, social media postings (Twitter and Facebook), ClinicalTrials.gov registry, and community advertisements (Craigslist, Nextdoor, and paper flyers). Advertisements provided a description of the study and eligibility criteria. Participants were enrolled on a rolling basis until we met our intended sample size.

We directed interested individuals to a study website with descriptive information and a screening questionnaire that assessed all eligibility criteria, including participants’ height and weight. Study personnel contacted eligible candidates within 3 business days to schedule an in-person baseline visit. During the baseline visit, trained study staff obtained written informed consent, confirmed eligibility, collected anthropometric measurements, and assisted participants in installing and navigating the MyFitnessPal mobile app; participants then completed an online survey.

Using simple random assignment, participants were then randomized by study staff to 1 of 3 treatment arms using Excel’s random number generator to allocate participants equally (1:1:1) across conditions. Randomization was revealed to participants by study personnel; as such, study staff were not blinded to treatment allocation but were blinded to the allocation sequence. Participants then reviewed materials describing their treatment condition and goals (see Intervention Design section below) in writing and with study staff to reduce contamination.

In-person follow-up visits occurred at 1 month and 3 months. At 1 month, study staff provided participants with information on their goals for the remainder of the intervention (see Intervention Design section below for details). We compensated participants with Amazon electronic gift cards (US $12 at baseline, US $6 for each follow-up visit, and US $5 bonus for completing dietary measures). Questionnaires were administered in English via a desktop computer. There was no contact with participants from months 3 to 6, and participants were not asked to self-monitor in MyFitnessPal during this time (though they could still do so if desired). At 6 months, study staff contacted participants via email and text message to collect self-reported body weight. Data collection ended in March 2018.

Participants were randomized to 1 of 3 conditions: (1) Simultaneous, (2) Sequential, or (3) App-Only, as outlined below and in Table 1. The CALO-RE taxonomy is used to describe behavior change techniques [5]. The intervention period lasted 12 weeks.

Sample size was calculated based on power to detect a 3.5 kg difference in weight change at 3 months between the Sequential arm and the App-Only arm (our primary comparison) using 3-month results from previous remotely delivered weight loss interventions for our Sequential arm [43] (results in kilograms were provided upon request by the author) and our App-Only arm [44]. Our power analysis (G*Power 3.1.9.2.) determined that 31 participants per group were needed to achieve 80% power for a 2-sided test with an alpha level of .05. To account for attrition of 10% and to obtain equal-size groups, we aimed to recruit 105 participants (35 per group). In exploratory analyses, we compared weight change between the Sequential arm and the Simultaneous arm, although we were not adequately powered to detect a significant effect.

For the baseline characteristics, we computed descriptive statistics stratified by treatment arm. To determine whether baseline characteristics differed by retention status, we used the Pearson chi-square test for categorical variables, analysis of variance for continuous variables, and Fisher exact tests with small cell counts. All analyses were 2-tailed. Participants who became ineligible during the study period up to 3 months were excluded from the analyses. Investigators remained blinded to outcomes until the completion of the 6-month trial.

We used intent-to-treat analyses to test our primary aim using linear mixed modeling with an unstructured covariance matrix and restricted maximum likelihood estimates to examine changes in weight over time by treatment arm. We did not control for any additional variables, and we assumed missing at random and used SAS 9.4 PROC MIXED (SAS Institute) for these analyses. For 6-month weight values sent via photo, we subtracted 0.172 kg (0.4 lb) to account for participants holding a device on the scale to take the photo. To account for the 6-month self-reported weight data without photos, we used a regression model to adjust for age, gender, and race/ethnicity [45]. Participants who sent a photo of their 6-month weight did not differ on any measured sociodemographic characteristics from those who did not send a photo (data not shown). We used chi-square tests to assess proportion of participants achieving ≥3% and ≥5% weight loss; we assumed noncompleters did not achieve this clinical threshold.

Given non-normally distributed intervention engagement data, we reported medians and interquartile ranges (IQR). To examine differences between treatment arms, we used Wilcoxon Mann-Whitney U tests (if 2 arms) and the Kruskal-Wallis tests (if 3 arms). We used Spearman rank correlation coefficients (r_s ) to examine the relation between self-monitoring engagement and change in weight. We also assessed for contamination by exploring whether participants self-monitored when they were not expected to do so.

Of the 670 individuals who completed the online screen for eligibility, 58.3% (391) were ineligible, whereas 23.7% (159) were invited to attend the baseline visit. We enrolled 105 participants and randomized them equally to 3 treatment arms (n=35, for each; see Figure 1 for Consolidated Standards of Reporting Trials [CONSORT] diagram). During the trial, 5 participants became ineligible (3 because of pregnancy, 1 because of cancer diagnosis, 1 because of previously undisclosed eating disorder). Of the remaining 100 participants, 84.0% (84 of 100) completed the 1-month visit and 76.0% (76 of 100) completed the 3-month visit. Between 3 and 6 months, one additional participant became ineligible because of pregnancy. At 6 months, 78% (77 of 99) of participants self-reported their weight. Participant retention did not differ significantly between arms at any time point (1 month: P=.84; 3 months: P=.23; 6 months: P=.32). We had no missing self-monitoring engagement data.

Table 4 illustrates the baseline characteristics of GoalTracker participants. At baseline, participants had a mean age (SD) of 42.7 (11.7) years and BMI of 31.9 (4.5) kg/m² and were predominantly female (84/100) and employed (78/100). One-third (33/100) were racial or ethnic minorities, most were married or living with a partner (64/100), and the majority had at least a college education (83/100). The majority (56/100) did not track diet in the month before baseline, although most had experience tracking body weight (87.0%). Completers at 3 months differed from noncompleters in race or ethnic minority status (P=.03), with 16% (11/67) of non-Hispanic white participants and 39% (13/33) of racial or ethnic minority participants missing the visit.

Figure 2 displays weight change over time by treatment arm. Weight change was significant over time for all arms (see Table 5). In our primary analysis, the Sequential arm did not significantly differ from the App-Only arm in weight change at 1 month (P=.06), 3 months (P=.78), or 6 months (P=.72). In exploratory analyses, the Sequential arm did not differ from the Simultaneous arm in weight change at 1 month (P=.36), 3 months (P=.92), or 6 months (P=.45).

The proportion of participants achieving at least 3% weight loss at 3 months was similar between arms (Sequential: 44%, 15/34 vs App-Only: 29%, 10/34, P=.21; exploratory analysis: Sequential vs Simultaneous 41%, 13/32, P=.77). Likewise, weight loss of at least 5% at 3 months occurred in 21% of Sequential participants (7/34) and 15% of App-Only participants (5/34), which was not significantly different (P=.52). In exploratory analyses, the proportion of participants with ≥5% weight loss did not significantly differ between the Sequential arm and the Simultaneous arm (31% of participants, 10/32, P=.32).

As expected, the Sequential arm tracked weight significantly more days than the App-Only arm (who was not asked to track weight) over the 12-week intervention; median (IQR) 70%, (28%-90%) vs 1% (0%-8%), respectively (P<.001). In exploratory analyses, the frequency of days participants self-monitored weight did not differ between the Simultaneous arm (73%; 25%-90%) and the Sequential arm over 12 weeks (P=.92).

As expected, the frequency with which the Sequential arm and the App-Only arm tracked diet in weeks 1 to 4 was significantly different (see Table 6). Between weeks 5 and 12 (once the Sequential arm began tracking diet), there was no longer a significant difference between the Sequential and App-Only arms; 27% (4%-80%) vs 21% (0%-62%), respectively (P=.54). In exploratory analyses, there were no significant differences in the frequency of days of self-monitoring diet between the Simultaneous and App-Only arms over the intervention period; 77% (27%-96%) vs 42% (17%-75%), P=.10. Table 6 displays additional self-monitoring and action plan completion outcomes (see Figures 3 and 4 for weekly data).

A low-intensity intervention utilizing a commercial app for self-monitoring resulted in comparable weight loss at 3 months, with no variability between the Sequential arm and the “off-the-shelf” App-Only arm. Nevertheless, loss of 3% to 5% of initial weight has been linked to improved health outcomes [46,47], suggesting that GoalTracker is an efficacious intervention for clinically meaningful weight loss.

The addition of evidence-based features such as weekly action plans, behavioral lessons, and tailored feedback did not substantially impact outcomes over and above the core intervention (ie, self-monitoring and in-app feedback) during our 12-week treatment, which parallels findings from several digital health weight loss trials [30,48] but not others [6]. Although most commercial weight loss apps do not include many evidence-based features [49], we suspect that weight loss might still occur with the inclusion of goals, daily self-monitoring, and a daily reminder to track. However, we suspect that had the trial been of a longer duration, the benefit of these enhanced features (eg, weekly lessons) would have become apparent in the findings; indeed, by 6 months, trends suggest continued weight loss at 6 months for the Simultaneous arm and relatively less weight regain in the Sequential arm compared with the App-Only arm.

Given that the GoalTracker trial compared 3 multicomponent interventions, we are unable to isolate the effect of self-monitoring diet and weight, along with each of the additional intervention components. Using a factorial design in consort with the multiphase optimization strategy (MOST) [50] would allow researchers to investigate the unique impact of each intervention component and then build and test an optimized intervention.

We found that self-monitoring engagement was high and that greater frequency of self-monitoring was related to greater weight loss. Contrary to our hypothesis, the Sequential arm did not demonstrate significantly greater engagement in self-monitoring dietary intake than the App-Only arm. In fact, once the Sequential arm was instructed to begin tracking diet after the first month, they tracked only 27% of days. It is possible that the Sequential arm participants’ minimal weight loss in the first month may have negatively impacted their future engagement or that being asked to track diet after a 1-month period seemed like an additional burden that many were unwilling to begin (indeed, almost half or 47% of Sequential participants never or rarely [<20% of days] tracked diet in weeks 5 to 12 [data not shown]). Although the trial was not powered to detect differences between the Simultaneous and Sequential arms, it appears that the Simultaneous arm outperforms the Sequential arm in frequency of tracking diet. This finding suggests that concurrently tracking diet and weight may have reinforced use of the app and activated the overarching goal of losing weight [51], thus leading to high engagement in both entities.

Future studies could consider framing the initial period before self-monitoring diet as a time during which to build self-regulatory skills rather than focus on weight loss, as has been done previously [9]. We selected weight tracking as the precursor to diet tracking for the Sequential arm because we wanted a behavior that could target the theoretical constructs of mastery—considered the best way to strengthen self-efficacy—and self-regulation. Self-weighing is a well-accepted strategy [52] in which mastery can be achieved [16], and self-regulatory capacity can be strengthened [53]. It is possible that providing more rationale for using a sequential approach would have encouraged participants to engage in diet tracking once asked to do so.

Notably, GoalTracker is the first weight loss trial to compare a sequential self-monitoring approach with a traditional approach that asks participants to track multiple components simultaneously. Previous work has compared a simultaneous approach with either a sequential or single component approach in other contexts, with mixed results [21,54]; no examination has focused on self-monitoring or digital approaches for weight loss.

The App-Only arm in GoalTracker performed better than expected. Intervention participants in Laing et al’s trial used the same MyFitnessPal app for diet tracking and lost 0.27 kg at 3 months and 0.03 kg at 6 months and had poor intervention engagement [44]. Possible explanations for the difference in weight loss between Laing et al’s trial and GoalTracker include GoalTracker’s use of specific goals to track diet daily and to lose 5% of initial weight by a specified end date and usage of phone-based reminder notifications.

In comparison with other randomized trials of commercial [28,29,32,55-58] or researcher- designed [24,59,60] apps for self-monitoring of diet, GoalTracker’s Simultaneous arm tended to have greater adherence to diet tracking, whereas the Sequential arm had lower adherence. Given that most weight loss trials of commercial apps are pilot studies and/or were not powered to detect an effect in weight change between treatment arms [26,28,29,38,56,61], more fully powered studies are needed that examine the efficacy of commercial apps for weight loss.

GoalTracker’s Simultaneous arm had a comparable or higher proportion of participants achieving 5% weight loss compared with other weight loss interventions that used mobile apps for self-monitoring dietary intake (range: 26%-35%) [28,59,62,63] but lower rates than some interventions including counseling (range: 42%-44%) [29,60,61].

Strengths of this trial include the collection of objective self-monitoring data for all participants via an API, use of a popular commercially available smartphone app, and ability to isolate the effect of a sequential versus simultaneous self-monitoring approach. In addition, this trial mimicked real-world weight loss experience (ie, no run-in period, prebaseline visit, or orientation session); consequently, it is possible that removal of these treatment barriers allowed for inclusion of participants with lower motivation and readiness to change. This design may have greater external validity but may make it harder to detect an effect between arms. Another strength was that the trial had little contamination between arms, which has been a problem in past app-based trials where up to 50% of participants in no-treatment control arms were found to have used commercial apps during the study period [44,55,63].

As this study was powered on superiority rather than equivalency, we cannot definitively assert that the treatment arms produce comparable weight loss. In addition, we collected self-reported weight at 6-months because of logistical reasons; however, we were encouraged to find that no additional attrition occurred between 3 and 6 months, despite no contact occurring during that period and no incentive given to provide a weight value. As is common in behavioral interventions, we provided minimal financial compensation to offset costs of attending study visits. Although we acknowledge that financial compensation can serve as an incentive for some to participate—and thus, may result in response bias on self-report measures—we expect this is unlikely, given that compensation was appropriately low. In addition, neither study staff nor participants were blinded to treatment arm, and we required participants to have access to a bathroom scale, although this mimics the real-world population who would track weight. Finally, this study did not include a pure control arm without an intervention, which may have led to an underestimation of treatment effects, as could the possibility of data not actually missing at random.

This study adds to the limited literature of randomized trials that assess the efficacy of commercially available mobile apps for weight loss [20,26]. In the GoalTracker trial, all 3 versions of the intervention produced weight loss and had high self-monitoring engagement, with no significant impact of additional features nor differential findings between a sequential versus simultaneous approach to self-monitoring. These results suggest that regardless of the order in which diet is tracked, using tailored weight and calorie goals and a commercial app can produce clinically significant weight loss in one-third of individuals. Stand-alone digital health treatments may be a viable option for those looking for a lower intensity approach who are willing and able to track.