This study was a randomized controlled trial with a 12-month intervention and 12-month follow-up period. The participants were allocated to 1 of the 3 treatment arms: eHealth, eHealth+group, or eHealth+individual treatment in a 1:1:1 ratio. All treatment arms received a web-administrated eHealth program, and additionally, the eHealth+group or eHealth+individual arms received 3 remotely facilitated group or individual meetings.

The eHealth program used in this study, Healthy Weight Coaching (HWC), is a 1-year treatment program based on the ACT framework, aiming to increase psychological flexibility and teach skills related to weight management [36]. The program consists of web-administrated weekly exercises integrating ACT with behavioral weight management themes (eating behavior, physical activity, sleep, and stress; Table S1 in Multimedia Appendix 1). The HWC program is readily available in Finnish public health care, and this study was designed to investigate whether minimal enhanced support in a group or individual format could improve attrition and response to the treatment.

The version of the program used in this study adopted a softer, more weight-neutral coaching style than the HWC protocol, with prioritization on value-based actions, mindfulness, and health-promoting changes instead of weight loss itself. The enhanced group and individual meetings in eHealth+group and eHealth+individual arms adhered closely to the ACT framework, incorporated key lifestyle change components of the program, and aimed to deepen the participants’ understanding of the program’s themes.

The participants were each assigned a personal coach who provided periodic, tailored, one-on-one written feedback and facilitated any additional meetings. All the 3 coaches in this study had master’s degrees in nutrition and further education in psychology. They met biweekly to ensure adherence to this study’s protocol. The coaches worked across the 3 treatment arms, being randomly assigned to one-third of each.

Participants were Helsinki city employees with class 1 and class 2 obesity [37]. Recruitment was conducted by the staff of the occupational health care and through recruitment advertisements on the city employees’ communication platforms. Enrollment was carried out by research assistants from February to March 2021. Inclusion criteria required participants to have a BMI between 30 to 40 kg/m², the ability to participate in group or individual meetings, access to a computer with internet services, and fluency in the Finnish language. Exclusion criteria included participation in other weight loss programs, pregnancy or lactation within the past 6 months, weight changes exceeding 5 kg within the past 3 months, major medical conditions affecting safety or readiness for weight loss, and the use of weight loss medication.

The sample size of 23 participants per treatment arm was determined by a statistician (JM) with priori power analysis using a statistical power of 0.80 with 2-sided test and 0.05 α, modeling the mean weight change and SD in previous weight loss studies (mean 3.8, SD 4.6 kg) [38]. With an expected attrition rate of 40%, the target number of participants was set at 38 per treatment arm.

Following recruitment, participants underwent a health screening on a phone by a medical doctor covering their health status and medications. Those who passed the screening were then referred for laboratory assessments and body composition measurements.

In March 2021, participants who provided written consent and completed baseline measurements were randomized into 1 of the 3 treatment arms (see figure in the Results section). Randomization was carried out by a statistician (JM) using stratified random allocation with SPSS (IBM Corp). Participants were stratified by sex (male or female), age groups (20-35, 35-50, and 50-65 years), and BMI categories (≥30 to 35 or ≥35 to 40 kg/m²).

The intervention period lasted for 12 months, from March 2021 to March 2022, after which a 12-month follow-up period was conducted until March 2023 without further support. Participant assessments were conducted at baseline, midintervention (6 months), postintervention (12 months), and follow-up (24 months) at Occupational Health Helsinki. The following measurements were taken:

Of the baseline characteristics (Table 1), metabolic syndrome was calculated from the laboratory and anthropometric measurements following the International Diabetes Foundation’s 2006 definition [39]. Smoking was self-reported in a baseline survey. The professional role was determined using Statistics Finland’s Classification of Occupations 2010 index [40]. Program usage was obtained from the log file. Medication changes were tracked every 3 months as part of the eHealth program.

Data analysis was conducted using IBM SPSS Statistics (version 28.0.0.0; IBM Corp). Baseline differences were analyzed by chi-square for the categorical variables and ANOVA for the continuous variables. Differences in mean values between the measurement points and the treatment arms were analyzed using the generalized estimating equations (GEEs) method [41]. The GEE method includes all the available data points and uses estimated means in the analysis, applying the intention-to-treat (ITT) analysis concept to decrease bias caused by dropouts. The article only reports results from the ITT analysis, while completer analyses are available in Multimedia Appendix 1.

Considering GEEs, we first used a model examining the interaction between the measurement points and treatment arms, with post hoc analysis comparing the effects of the eHealth+group and the eHealth+individual arms to the eHealth arm. For further analysis, we used a single-arm model, concentrating on the changes observed across the 4 measurement points in the entire study sample. After the primary analysis, we added baseline weight, sex, age, and applicable medication or medication change (eg, blood pressure medication in systolic and diastolic blood pressure analysis) to the single-arm model to explain the results. The changes in medication are described in Table S3 in Multimedia Appendix 1. Weight, BMI, waist circumference, HbA_1c, cholesterol, HDL, low-density lipoprotein, systolic blood pressure, diastolic blood pressure, and pulse rate were normally distributed. Glucose, triglycerides, alanine transaminase, hs-CRP, and muscle percentage had a positively skewed distribution, and thus, gamma distribution with log link function was used in the GEEs analysis. Visceral fat and fat percentage had a negatively skewed distribution, and thus, inverted variables were used in the analysis with gamma distribution and log link function.

Crosstabs with Pearson chi-square was used when analyzing the group level differences of categorical weight variable (participants were divided into 6 categories based on weight loss percentage: <0, 0-2.9, 3-4.9, 5-5.9, 5-9.9, or ≥10), and change in metabolic syndrome prevalence between the measurement points. Categorical weight loss and metabolic syndrome prevalence are reported for the entire baseline sample, with the last observation carried forward. Changes in muscle and fat percentage, as well as metabolic syndrome incidence, are expressed as percentage points (%-point), describing the arithmetic difference between the 2 percentages.

Spearman correlation coefficient was used to analyze correlations between the weight change percentages (nonparametric). Dancey and Reidy’s definition was used when determining the strength of the correlation [42].

Cohen d effect size was calculated using the square root of the average variance of measures. In the analysis comparing the treatment arms, we used an effect size calculator developed by the statistician (JM) that adjusts for baseline differences. For the single-arm analysis, we used the University of Colorado Springs Effect Size Calculator [43].

Figures were drawn with GraphPad Prism (GraphPad Software).

This study’s protocol was approved by the Ethics Committee of the Helsinki and Uusimaa Hospital District (April 29, 2020, HUS/922/2020) and conducted per the ethical standards of the Declaration of Helsinki. Informed consent was obtained from all individuals included in this study. To ensure privacy and confidentiality, study data were pseudonymized and securely processed, preventing any association with individual participants. Participants received the intervention and individual health data but received no additional compensation. Neither the participants nor the data assessors were blinded to the treatment arms.

Of the 143 people assessed for eligibility, 111 were randomized into research groups (Figure 1). Most participants were female (n=92, 83%) and worked as experts (n=46, 41%) or specialized experts (n=34, 31%; Table 1). More than half were either former or current smokers (n=59, 53%). Of the 111 participants, 73 (66%) had metabolic syndrome, 23 (21%) had mental health problems, 21 (19%) had osteoarthritis, 7 (16%) had type 2 diabetes, and 15 (14%) had sleep apnea. The mean age was 50 (SD 8.9) years, and the mean BMI was 34.2 (SD 2.9) kg/m². Baseline characteristics did not differ between the groups except for triglycerides (P=.006), which showed statistically significant elevation in the eHealth arm compared to the other two treatment arms.

The retention rate was 95% at the 6-month mark, 86% at the 12-month mark, and 61% at the 24-month mark of the intervention (Figure 1). The participants completed, on average, 58.6% (SD 32.7%) of the eHealth program. There were no differences in retention or progression between the treatment arms.

We did not observe any significant between-group differences in changes in the primary outcomes of weight and categorized weight (P=.76 and P=.99, respectively; Table 2, and Tables S4 and S5 in Multimedia Appendix 1). However, we did discover an interaction between the time points and treatment arms for HDL (P=.03) and triglycerides (P=.002).

Considering HDL, further analysis showed a greater increase in the eHealth arm compared to the eHealth+individual arm between 0 and 6 months (P=.09, d=0.35) and 0-12 months (P=.03, d=0.28). The interaction between the time points and treatment arms was explained by sex (P=.002), cholesterol medication use (P<.001), and changes in cholesterol medication (P<.001). When participants who were using cholesterol medication at baseline or had changes in their medication were excluded, the interaction between the treatment arm and time point was insignificant (P=.17). This suggests that the difference in HDL between the treatment arms was likely due to variations in medication use.

For triglycerides, the eHealth arm demonstrated a greater decrease than the eHealth+individual arm at all time points relative to baseline (0-6 months P=.03, d=0.41; 0-12 months P<.001, d=0.53; 0-24 months P=.002, d=0.47; Table 2). However, the eHealth arm had higher triglyceride levels at baseline compared to the other groups (Table 1).

To study the overall effect of the intervention, the 3 groups were considered as a single eHealth group and combined for further analysis, as reported below.

This study examined whether adding minimal individual or group-based face-to-face support to a 12-month coach-assisted eHealth intervention enhances outcomes by strengthening connections with coaches or leveraging peer support. Contrary to our hypothesis, no significant differences in primary outcomes were found between the groups, indicating that 3 live sessions—whether individual or group-based—did not provide additional benefits over the eHealth protocol alone. Interestingly, the eHealth-only group showed greater improvements in HDL cholesterol and triglycerides compared to the individual session group. However, post hoc analyses attributed these differences to cholesterol medication use and higher baseline triglyceride levels rather than the intervention itself. Thus, incorporating infrequent live sessions into the eHealth protocol does not appear to offer additional advantages.

In the entire study population, the intervention resulted in a mean weight loss of 1.5% from baseline to 12-month postintervention and 0.7% from baseline to 24-month follow-up. When considering individual weight change, the results are polarized: the received support promoted weight loss for some participants, while others did not respond. The weight changes were notably correlated between different measurement points, suggesting the individual weight change to be rather consistent.

Throughout the intervention and follow-up, there was a progressive decrease in mean waist circumference, along with improvements in the body composition variables at the 12-month mark.

Additionally, there was a slight enhancement in HDL cholesterol levels at postintervention and follow-up, suggesting positive effects on lipid metabolism. Although we observed a decrease in the fasting glucose levels during the intervention and follow-up, similar results were not observed in HbA_1c, a marker measuring the average blood glucose levels over the past few months. This indicates that the observed improvements in glucose levels might not be consistent or lasting enough to affect overall glycemic control.

Unexpectedly, the hs-CRP increased significantly by the 24-month follow-up, suggesting elevated inflammation not attributable to outliers but potentially reflecting mild infections or systemic inflammation [44]. We do not attribute the increase in the hs-CRP to heightened metabolic disturbances, as we noted a slight, though statistically insignificant, improvement in the prevalence of metabolic syndrome.

There are several potential explanations for the lack of intergroup differences. It is possible that only 3 sessions in the hybrid conditions do not yield a sufficient difference compared to the coach-assisted eHealth intervention, and by increasing the number of sessions in the hybrid intervention, differences between the groups could be observed. In Teeriniemi et al [29], participants with a BMI over 30 kg/m² lost an average of 4.4% of their weight in 12 months when they received a hybrid intervention combining eHealth with 8 group meetings based on cognitive behavioral therapy, compared to just 1.6% weight loss in the eHealth-only arm. Therefore, a more intensively enhanced hybrid approach may be needed to further improve somatic outcomes.

Notably, the intervention took place during the COVID-19 era, leading to all group and individual meetings being conducted remotely. As seen in previous research on therapeutic relationships [45], this shift could have hindered the alliance between the participants and the coach. Particularly, the unplanned move to remote group meetings in the eHealth+group treatment arm may have affected the group’s cohesion and perceived peer support [46], thereby impacting effectiveness and attenuating differences between the treatment arms.

Overall, it is likely that differences in weight change between the 3 study arms are attenuated by the lack of power resulting from smaller-than-expected changes in weight, with high variability. The total weight loss is modest compared to the 2019 meta-analysis by Lawlor [12], reporting a mean absolute weight change of 5.5 kg in third-wave cognitive behavior therapies. Furthermore, the weight loss outcomes observed in our study fall short in comparison to the results reported from previous implementations of the HWC eHealth program, which achieved a 4.6% weight loss during the 12-month intervention [47]. Additionally, a “review of reviews” by Kupila et al [23] reported mean differences in weight loss results ranging from –0.12 (95% CI –0.64 to 0.41) kg in eHealth interventions compared with face-to-face care to –4.32 (–5.08 to –3.57) kg in eHealth interventions compared with no care. However, our results align with the 12-month weight loss of –2.08 kg reported by Mueller et al [33] for another ACT-based hybrid intervention (eHealth+phone support) conducted during the COVID-19 pandemic.

Several factors could contribute to the lower weight loss observed in our study. First, our study used a more weight-neutral coaching style than the one in Kupila et al [47], which may have impacted its effectiveness in promoting weight loss compared to the earlier version. It needs to be noted that a newer version of the HWC is already in clinical use. Therefore, the results of our study do not demonstrate the current effectiveness of the HWC.

Second, our study involved a different population characterized by lower BMI compared to the study of Kupila et al [47]. In their study, the greatest weight loss was observed among individuals with a baseline BMI ≥40 kg/m², a subgroup that was not included in our study. Yet, in Kupila et al [47], the mean weight loss for those with a baseline BMI <40 kg/m² was 3.2%—more than the weight change observed here.

Third, the influence of the COVID-19 pandemic coinciding with the 2021-2022 intervention cannot be overlooked, as research suggests that the pandemic led to disruptions in lifestyle behaviors that could affect weight management. The mean weight gain during the pandemic was higher [48] than average [49], indicating difficulty in weight maintenance. These effects are evident in the Finnish population: between 2017 and 2023, the prevalence of obesity increased 3%-points in men and 4%-points in women [50]. In contrast, the prevalence of obesity had remained unchanged between 2012 and 2017 [51]. In addition, an increase in food consumption and sedentary activity was observed during the pandemic [52]—changes that were potentially intensified in this study population, most of whom worked in expert positions and shifted to remote work during the pandemic.

Lastly, our study involved a different population characterized by significant occupational burden from mentally consuming expert work. This occupational burden may have posed challenges to weight management despite participation in the intervention, especially with the combination of the eHealth program and remote knowledge work. This is per the earlier results showing that early weight loss predicts 1-year weight loss in web-based programs [53] and shows that the early weight loss could even predict long-term weight maintenance.

In the secondary analysis with the overall sample, long-term improvements in waist circumference and lipid metabolism are consistent with findings from previous nondiet studies [54-57]. Additionally, the increase in muscle mass contrasts with previous findings of skeletal muscle mass loss in weight loss studies [58]. These changes are likely attributed to our intervention’s focus on guiding participants toward making healthy dietary choices and increasing physical activity without imposing calorie restrictions. However, our findings of no improvement in long-term glucose metabolism contrast with previous studies showing that lifestyle weight loss interventions reduce the risk of elevated fasting glucose levels and HbA_1c [59-61], suggesting that a more prominent weight loss may be necessary to achieve sustained improvements in glycemic control.

This study presents both strengths and limitations. A key strength is the long follow-up period, which allowed for the investigation of the treatment’s long-term effects—rare in typical ACT-based interventions. Additionally, the comprehensive array of outcome measures, including body composition and laboratory tests alongside traditional metrics such as weight and waist circumference, offers a holistic understanding of the intervention’s effects. This study also ensured accuracy and reliability by conducting all measurements and laboratory tests on-site despite the onset of COVID-19. Additionally, the emergence of new weight loss medications was addressed by screening and excluding participants using weight loss medications, requesting that participants refrain from starting new weight loss medications during the intervention, regularly collecting information on any changes in medication, and excluding from the analysis those who began using them during the follow-up period.

However, some limitations impact especially the interpretation of the overall study effects. First, the participant group was homogeneous—all employed and primarily middle-aged women in expert professions—making generalization to other populations challenging. Second, the lack of blinding and participants’ awareness of their treatment arms before randomization could have affected motivation, particularly for those not assigned to their preferred group. Third—affecting interpretation of the overall results—the absence of an inactive control group prevents comparison against a baseline, making it difficult to isolate the treatment effects from external influences or placebo effects and thus establish causality. These limitations, arising from the integration of this study within routine occupational health care services and the need for ethical service provision to willing participants, highlight important considerations for future research aiming to balance real-world applicability with rigorous study design.

Our study found no differences in weight or other somatic health variables between the eHealth arm and intervention combining eHealth with minimal group or individual enhancement. Moving forward, further research is needed to determine if there is a threshold where face-to-face meetings provide additional benefits in hybrid interventions. Moreover, there is a need to explore for whom and under what conditions eHealth and hybrid models may be most effective.