A large body of evidence on the maintenance of weight loss and/or behavior change following the end of initial interventions resoundingly shows a relapse effect, characterized by weight regain and/or behavioral decline back toward baseline levels [1-3]. This has led to a concerted focus on weight loss maintenance interventions, defined in the current literature as “extended contact” interventions—the intervention that continues after initial weight loss intervention, typically at a comparatively lower intensity than the initial intensive phase of intervention. Extended contact interventions have been consistently shown to somewhat mitigate the relapse effect [4-6]. What is not known is whether extended contact helps mitigate weight regain in the longer-term or whether it simply delays weight regain until after the extended intervention contact ceases.

Moreover, 2 previous studies evaluating text message–delivered, extended contact interventions for weight loss maintenance have shown that 6 months after extended contact ceases, participants’ body weight has remained, on average, significantly reduced compared with baseline [7,8]. However, another 2 studies that compared an extended contact intervention with a control condition after 2 months [9] or 6 months [10] of no contact found no between-group maintenance effects, although notably neither of these interventions achieved significant intervention effects at the end of text message–delivered extended contact. Research is needed to understand whether the improvements gained through extended contact behavioral interventions can be sustained following the end of such interventions, particularly in comparison with a control condition.

The “Get Healthy, Stay Healthy” (GHSH) intervention was an extended contact program delivered via text messages for 6 months following completion of an initial 6-month community-wide lifestyle telephone coaching program called “Get Healthy Service” (GHS) in Australia [11]. The GHSH intervention was evaluated in a randomized controlled trial (RCT) compared with normal practice following GHS (no ongoing intervention contact). Anthropometric (weight and waist circumference) and behavioral (physical activity and dietary) indicators were assessed at baseline (following completion of GHS), 6 months (following completion of GHSH), and 12 months (no-contact maintenance follow-up). We have previously reported that the GHSH intervention was feasible to deliver using semiautomated Web-based technology and was highly acceptable to participants [12]. Changes in body weight and physical activity (but not dietary outcomes) between baseline and 6 months were significantly better for the GHSH intervention group compared with the control group [12].

This paper aims to evaluate maintenance beyond the period of the GHSH extended contact intervention in multiple ways. First, we used the traditional method of assessing and comparing changes from baseline contemporaneously after some period of noncontact (ie, at 12 months, which is after 6 months of noncontact for the intervention group and after 12 months of noncontact for the control group). This method establishes whether an intervention has a lasting effect, after allowing some time for intervention recipients to relapse. Second, we used a novel approach to directly assess and compare the degree of changes over the first 6 months of noncontact between the intervention group (between 6 and 12 months) and the control group (between baseline and 6 months). This method adds to the previous by determining whether the degree of changes after extended-contact intervention are any different to changes that naturally occur over the same amount of noncontact time without extended contact. Finally, both of these maintenance perspectives consider only changes at the group level. We, therefore, also examined individual-level changes (personal increase/decrease/no change) over the first 6 months of noncontact, to explore to what extent nonsubstantial changes in group averages reflect all or most participants making no changes or are due to large increases by some participants being offset by large decreases by other participants.

A detailed description of this RCT is published elsewhere [11]. Eligible consenting participants were randomized in a 1:1 ratio to the GHSH intervention and control groups, via a randomization website, by a research assistant with no involvement in participant recruitment. Randomization was across 2 strata (≥ or < the median of 3 kg weight loss during GHS). Recruitment began in August 2012 and 12-month follow-up data were collected until August 2014. Ethical clearance was received from the Human Research Ethics Committee at The University of Sydney (Protocol number: 03-2011/13523).

The GHS is available to adults (≥18 years and older) residing in New South Wales, Australia and is available for free via self-referral and health professional referral. Participants completing GHS between August 2012 and February 2013 were eligible to join the GHSH trial if they had no intention of re-enrolling in GHS coaching, were not involved in other GHS evaluation substudies, and owned a mobile phone. All eligible clients completing the initial contact intervention (GHS) within the recruitment time frame were invited to participate in GHSH during their final coaching call. Interested participants were mailed an information sheet and consent form and then contacted via telephone to establish their eligibility and willingness to participate. Verbal consent to participate was audio recorded, and participants returned a signed consent form via reply-paid post.

The GHSH-extended contact intervention was delivered via individually tailored text messages. Tailoring data were collected during an initial and an interim telephone call (around 12 weeks), during which participants worked with a trained coach to set a 12-week weight goal (weight maintenance or further weight loss) and two 12-week goals for physical activity and/or dietary behavior change, with targets consistent with national guidelines [13,14]. For each behavioral goal (diet and/or physical activity), participants were asked to identify rewards for reaching their goal, expected benefits, preparatory behaviors for goal attainment, barriers and solutions, and a person who could support them to reach their goals. Participants selected their desired number of text messages (from 3-13 per fortnight), timing of texts (eg, 6 am), and type of texts. Overall, 4 types of texts targeted different behavior change strategies, each with different permitted frequencies: prompts to self-monitor weight (once per fortnight), goal checks for behavioral goals (from once per fortnight to once per week for each goal), real-time behavioral prompts (from none to 4 per fortnight for each goal), and goal resets for weight and behavioral goals (1 in week 6 and 1 in week 18). At 12 weeks, participants received a second telephone call from their coach to update their tailoring goals and preferences.

To minimize trial attrition, control participants were posted brief written feedback of results following each assessment. The control group received no other contact.

Details of the data collection are reported elsewhere [11]. Briefly, data were collected at baseline, 6 months, and 12 months. Most outcomes were collected by computer-assisted telephone interviews (CATI), conducted by a research assistant, who was initially blinded to group allocation (information collected in the interviews limited this blinding at 6 and 12 months). The outcomes and measures were the same as those collected in the initial intervention, with the addition of an objective monitor of physical activity and a nutrition assessment tool, the Fat and Fiber Behavior Questionnaire (FFBQ) [15].

As previously reported [11], the sample size had been chosen a priori to provide 90% or more power to detect the following expected differences between groups in primary outcomes with 5% 2-tailed significance: 2 sessions per week of self-reported MVPA, 1 daily serving each of fruit and vegetables, 2 kg body weight, and a 4 cm waist circumference. The study was not powered a priori for questions concerning within-groups changes. For the FFBQ indices, fruit intake, takeaways, and sweetened drinks only, power was adequate (≥80%) to detect differences between groups meeting the minimum differences of interest (MDI). The MDIs were set at 1 kg weight, 1 cm waist circumference, 30 min or 0.5 sessions/week physical activity, 0.5 servings per day of fruit and vegetables, 0.5 takeaway meals per week, 0.25 cups per day of sweetened drinks, and 0.2 units on the FFBQ indices [12].

Maintenance is considered in three ways. First, whether anthropometric and behavioral outcomes are comparatively better after noncontact (at 12 months) for those who received the GHSH extended contact intervention than for those who had not (controls), second, comparing changes during the first 6 months of noncontact within the GHSH intervention group (ie, between 6 and 12 months) with changes in the control group (ie, between baseline and 6 months), and finally, considering behavioral maintenance at the individual-level during the first 6 months of noncontact in the intervention group.

Statistical analyses were performed using SPSS Statistics version 22 (IBM, USA) and STATA version 13 (StataCorp LP, USA). Significance was set at P<.05, 2-tailed. Changes within groups were assessed using paired t tests. All differences between groups were estimated adjusting for the same potential confounders as per the main outcome evaluation of the GHSH intervention [12]. Differences between the groups in their changes (for all outcomes and all time frames) were assessed using separate linear regression models adjusting for baseline values of the outcome and confounders. When assessing change over the period from baseline to 12 months, baseline values were taken as the beginning of the GHSH evaluation. When assessing changes over the first 6 months of noncontact, baseline values were taken as the beginning of the GHSH evaluation for controls or at 6 months for the intervention group. Group differences in daily values of accelerometer MVPA (log-transformed) were assessed using generalized estimating equation models, which accounted for repeated measures (1-7 days per participant per assessment), adjusted for confounders, and controlled comparisons for baseline values via the method outlined by Twisk [21]. Marginal means of the appropriate back-transformed expression were used to obtain the contrasts in minutes per week. The absence of substantial changes in group averages can be suggestive of maintenance; however, this can be achieved by large worsening in some participants being offset by others’ large improvements and does not necessarily show whether individuals maintained their personal outcomes. Accordingly, we further describe how many of the intervention participants maintained their outcomes during noncontact (ie, individual-level maintenance).

Participants who remained in the study at 12 months had an average (mean [SD]) age of 53.4 (SD 12.3) years and baseline BMI of 29.2 (SD 5.9) kg/m² (Table 1). Approximately two-thirds of the participants were female. Retention over the 12-month trial was high overall (92.5%, 211/228), but slightly lower in the intervention group (86.8%, 99/114) than in the control group (98.2%, 112/114; P=.002). Those lost to follow-up at 12 months (n=17) had significantly heavier baseline BMI, were more likely to smoke at baseline, and reported consuming fewer vegetables and more sweetened drinks at baseline than those who participated in the 12-month follow-up CATI (n=211; see Table 1).

Figure 1 shows changes from baseline to 12 months in study outcomes in the control group (12 months of noncontact after the initial GHS intervention) and in the GHSH intervention group (following 6 months of GHSH extended contact then 6 months of no contact). Results are described in units but plotted relative to a substantial decline or worsening (ie, MDI), to indicate whether the absence of a significant change was more consistent with maintenance (ie, a substantial worsening is unlikely) or an inconclusive result (ie, the error is too large to draw a conclusion). Over 12 months, the GHSH intervention group made significant improvements to waist circumference and the fiber index score, and the control group significantly improved their waist circumference and fat index score. The only significant declines between baseline and 12 months were self-reported walking, which declined in the intervention group, and self-reported vigorous activity, which declined in controls (Figure 1). Self-reported moderate physical activity and most of the dietary outcomes were maintained in both groups.

Although both groups had displayed a large degree of behavioral maintenance at 12 months, extended contact was still associated with a significant advantage over control treatment for body weight (P=.04) and accelerometer-assessed MVPA (P=.01), with differences between groups averaging approximately 1.3 kg and 25 min per week (Table 2). Only small and nonsignificant differences between groups were seen with the other outcomes; however, CI included meaningful differences in self-report physical activity and vegetable intake.

Figure 2 shows the changes over the first 6 months of noncontact within each group. Statistically significant changes over the first 6 months of noncontact occurred only in the control group, not in the intervention group following extended contact. All of these changes were worsening of outcomes rather than improvements (Figure 2). Intervention changes (all nonsignificant) during this time frame were suggestive of maintenance for waist circumference, moderate activity, and all of the dietary outcomes, but margins of error precluded definitive conclusions concerning the other outcomes.

When compared with the control group changes over their first 6 months of noncontact, those receiving intervention fared significantly better than those receiving usual care in terms of accelerometer-measured MVPA and self-reported moderate activity (Table 2). Other differences between the groups were all nonsignificant. All were small (except for walking) and a substantial effect of extended contact was unlikely for weight, FFBQ indices, takeaways, sweetened drinks, and fruit intake (based on the CI). However, CI included potentially meaningful differences in waist circumference, walking, vigorous physical activity, and vegetable intake.

In addition to the maintenance seen in the group averages, many intervention participants had maintained their behavioral outcomes during 6 months of noncontact, whereas a minority of participants had worsened in their outcomes during that period. Figure 3 shows the percentage of intervention participants who had maintained their outcomes during noncontact. Nearly all outcomes were maintained (ie, no worsening of ≥ the MDI) by the majority of participants. The proportion of maintenance was lowest for weight (55.5% [56/101]) and highest for sweetened drinks (91.2% [93/102]), with a substantial proportion of maintenance coming in the form of further improvement for most outcomes. Notably, weight regain was still reasonably common (44.5% [45/101]), despite this maintenance intervention and despite the overall intervention effects for weight outcomes.