Closing the Gap in Behavioral Weight Loss Therapy: Prospective Analysis of Clinical Real World Data of a Four-Year Health Insurance-Financed Program

1. Introduction

Obesity defined as a body mass index (BMI) of >30 kg/m² leads to a higher risk of metabolic (e.g., diabetes mellitus type 2, arterial hypertension) and psychological (e.g., depression) diseases, and higher risks for incapacity for work and early retirement [1,2]. In persons with obesity, without treatment, usually with increasing age further weight gain occurs, regardless of baseline weight [3]. In contrast, even a moderate weight reduction can improve comorbidities and cardiovascular risk factors [4]. Other important goals of obesity treatment are an increase in quality of life, prevention of disability and loss of productive years [5].

National [6] and international [5] guidelines recommend an interdisciplinary behavioral treatment as the basis for obesity treatment. Usually it is a combination of nutritional, exercise and behavioral therapy [6] focused on problem identification and problem solving. This includes, e.g., self-monitoring, goal-settings and stress management. By individualizing components therapy adherence can be increased [7]. Behavioral weight loss treatment can be the basis for further treatment steps, such as bariatric surgery, but also, if successful, a relevant therapeutic intervention with important clinical improvements for patients [8].

However, for most persons in need, treatment options are rare due to a lack of financing for corresponding programs through the health care systems, and in industrialized countries obesity and low income are clearly associated [9]. Our evaluated behavioral weight loss therapy program offered a long-term behavioral weight loss treatment without a cost sharing by the patient. In contrast to many other behavioral weight loss programs [8,10,11], it intentionally avoided the use of meal replacement shakes, in order to provide a program for patients who are not suitable for full meal replacements for health, personal, monetary or other reasons. In addition, the German healthcare system currently does not fund the use of meal replacement products. For this reason, only programs without meal replacement products would be eligible for applying for nationwide health insurance coverage.

Here we report clinical real world data of an interdisciplinary four-year, fully health insurance funded Leipzig Obesity Management program, which due to its real world nature has an additional value to already available data from purely research-funded programs [12]. Data of the first segment of treatment was already published [13]. We now evaluated the efficacy of the complete four years of the program for weight change, and changes of metabolic and psychological parameters. In the case of premature termination, all available data up to the point of termination were taken into account.

2. Materials and Methods

2.1. Patients

Inclusion criteria for participation were a BMI ≥35 kg/m², age ≥18 years and an insurance contract with the local health insurance company AOK PLUS. In Germany, behavioral weight loss therapy is not generally covered by health insurances. Our program was to our knowledge the first four-years-program in Germany, which secured full coverage through a contract between the hospital and the health insurance company.

Exclusion criteria were severe mental disorders, substance abuse and other conditions, which would impair the regular participation in the therapy program, or secondary causes of obesity like endocrine disorders.

Reported data is observational real world data, all patients with data were part of the program and there was no untreated control group.

2.2. Interventions and Assessments

At baseline assessment (t₀), a blood draw was performed and routine medical and metabolic laboratory parameters were determined (Supplementary Information, Table A1). Medical history and anthropometric parameters were assessed by a physician in a standardized way. The EUROHIS-QOL 8 item index [14] and the patient health questionnaire (PHQ-9) [15] were used to assess quality of life and depressive symptoms, respectively. If the psychological assessment by questionnaires suggested clinically relevant problems in comparison to population norms, an in-person session with a clinical psychologist was scheduled. Furthermore, a nutritional history was assessed by a dietitian.

Following baseline assessments the interdisciplinary board meeting, consisting of a physician from the department of internal medicine, a dietitian, a psychologist and a case manager, decided about each patient’s treatment plan. If the patient had a primary indication for bariatric surgery (based on the national German guidelines [16]) and the patient agreed, bariatric surgery was offered. From August 2014 until September 2016, 411 patients performed baseline assessments, until the end of May 2020 all patients had either completed the whole Leipzig Obesity Management program or dropped out. Of the 411 baseline patients, 29 (7%) had a primary indication for bariatric surgery and were transferred to the department of surgery. One patient was included a second time into the program after dropping out after first inclusion during the first program segment. The second time was excluded from further evaluation. All other patients (n=381, 93%) started the behavioral weight loss program (Figure 1) and were included into courses of n=11-18 patients. The whole program consisted of four successive 1-year courses (“segments”), with the first segment having the highest intensity (i.e., most therapy hours), and following segments having less scheduled classes.

There were three options for similar behavioral treatment programs during the first segment. Patients were grouped regarding their physical ability for participation in the exercise sessions and the daytime when they could best participate (Supplementary Information, Table A2).

Participants absolved one of the three program options, which lasted approximately one year. Each course consisted of dietary counseling, behavioral therapy and sports courses and lasted approximately twelve months. The whole program consisted of four successive 1-year courses (“segments”), with the first segment having the highest intensity (i.e., most therapy hours), and following segments having less scheduled classes. During the first segment of the program, there were up to six individual diet therapy sessions, 6-12 nutritional group therapy sessions, 10-12 behavioral group therapy sessions and 40-48 group exercise sessions. The appointments took place in the obesity clinic of the University Hospital Leipzig/Germany, some of the exercise courses were conducted by associated regional partners (health sports clubs) [17,18,19].

Exercise sessions consisted of 45-60 min. of aerobic (e.g., Nordic walking, swimming) and anaerobic (e.g., weight lifting) exercises of light to medium intensity, adapted to the physical abilities of participants. Dietary counselling group sessions were 90 min., individual sessions 30 min. long. Treatment focused on increasing the intake of complex carbohydrates, a reduction in the intake of short-chained carbohydrates and saturated fatty acids, taking into account individual needs. Behavioral therapy group sessions lasted 90 min. and focused on goal-setting developing sustainable implementation strategies and problem solving. The sessions of all components (behavioral, diet and exercise therapy) took place within the year in a coordinated sequence so that the topics complement each other (Supplementary Information, Table A2).

At the end of the first segment of the behavioral weight loss program, an assessment comparable to the baseline assessment took place (t₁). Thereafter, the interdisciplinary board meeting reviewed each patient`s progression to determine the further progression of therapy. If the patient wished to obtain bariatric surgery, a physician from the department of bariatric surgery was consulted. After completing the first program segment, bariatric surgery could be performed at any time of the program following the aforementioned algorithm.

In the second to fourth segment, frequency of appointments decreased (Supplementary Information Table A2). After completion of each segment lasting approximately one year, a segment assessment (t₂-t₄) identical to baseline assessment was performed. If necessary, patients could assign to additional appointments with the dietitian and/or psychologist throughout the whole time of treatment.

The aim was to achieve a participation rate of at least 80%. Patients with irregular attendance without giving reasons were excluded prematurely. Patients who did not reach the 80% target for health reasons were allowed to continue participating in the program. Patients who ended the program prematurely were defined as dropouts. All patients included into the analysis of the Leipzig Obesity Management program terminated the program (after completing t₄ or prematurely) until May 31th 2020. The first therapy segment was started by 381 patients (n=272, 71.4% female). Mean ±standard error of the mean (SEM) age was 43.3 ±0.7 years, mean BMI 44.3 ±0.4 kg/m² and mean body weight 127.3 ±1.3 kg (Table 1). At first presentation, 118 (31%) patients had type 2 diabetes mellitus and 233 (61%) arterial hypertension. In average costs per patient ranged between 2,000 € for the first segment of therapy and 700 € for each of the second to fourth segment and were fully covered by the health insurance company.

2.3. Assessments of Laboratory Parameters

The laboratory parameters were determined in the Institute for Laboratory Medicine, Clinical Chemistry and Molecular Diagnostics at the Leipzig University Hospital. The calculation of the homeostasis model assessment of insulin resistance (HOMA-IR) was used to determine insulin resistance. Therefore, fasting glucose (mmol/l) was multiplied by fasting insulin (pmol/l) and divided by 156.2625 [20].

Patients were grouped as “patients with diabetes” if patients reported a previously established diagnosis of type 2 diabetes, were taking diabetes medication, had a hemoglobin A1c (HbA1c) ≥6.5% (48 mmol/mol) and/or a fasting plasma glucose ≥7 mmol/l. Patients with changes of diabetes medication throughout the course of the program were excluded from the analysis of glucose parameters (n=24). For assessment of HOMA-IR, additionally all patients with insulin treatment were excluded (n=21). Concordantly, for the assessment of parameters of lipid metabolism, patients with changes in lipid lowering medication through the time course of the study were excluded from the analysis (n=15).

2.4. Statistical Analysis of Medical Data

All patients who started the first year of therapy (n=381) were included in the evaluation. People who quit the program prematurely or get a bariatric surgery during the course of the program were included in the evaluation with the available data within the behavioral therapy phase. Controlling for patient’s age at baseline and correcting for multiple testing due to the different time set points according to Bonferroni [21], mixed model analyses were used to evaluate changes of parameters of interest of repeated measurements from t₀ to t₄ for all assessed parameters.

To model BMI changes over time, all available measurements were used in a sequence of linear mixed models that treated time as a continuous variable and patient as a random term. The first model in the sequence only contained time linearly, the second added its square, the third its cube and the fourth added time to the fourth power. The optimal model was selected by minimizing the Bayesian information criterion. Gender did not show a significant difference in the linear model, which is why it was not included [22].

Data are presented as mean ±standard error of the mean (SEM). For analysis, Excel 2016 (Microsoft Corporation version 16.0; Redmond, WA, USA), SPSS (IBM® version 25.0; Chicago, IL, USA), STATA (version 16.0) and R (The R Foundation for Statistical Computing, Version 3.5.1, Vienna, Austria) were used. All analyses used a two-tailed α <0.05 as level of significance.

2.5. Assessment and Statistical Analysis of Healthcare Costs

Average treatment and program costs were evaluated for the program group over a period of time starting three years before program and ending with the fourth year of treatment. Treatment costs included inpatient, outpatient and drug costs. There were also patients who did not incur any further healthcare costs in addition to the costs of the therapy program. These patients were also included in the evaluation. Because of dropouts and different durations of participation in the program, we standardized the costs outcomes to costs per capita and 30 days. We analyzed the changes of costs per time period and tested for significance with the Mann-Whitney U test correcting for multiple testing according to Bonferroni-Holm to maintain a global significance level of 5% [23].

3. Results

Of 411 eligible patients who fulfilled the in- and exclusion criteria and wanted to participate, 381 participants started the Leipzig Obesity Management program. During the whole time of treatment, 73 (19%) participants underwent bariatric surgery and 234 (61%) quit the program prematurely. Self-reported reasons for dropping out were personal and familial (28%) or health reasons (16%) (Supplementary Information, Table A3).

The whole four years of the behavioral weight loss program were completed by 77 (20%) participants (Figure 1). The average time between t₀ and t₄ was four years and seven months.

Between t₀ and the last available measurement before completion (t₄) or drop out, respectively, including measurements between the segment assessments, 61% of patients lost weight, mean weight loss was 2.9 ±0.6 kg (2.3 ±0.4%), p<.001 (Figure 2). Taking into account all available data at the two comparative time points irrespective of therapy adherence, mean body weight change during the first year from t₀ to t₁ (n=243) was -5 ±0.6 kg (p<.001), from t₀ to t₂ (n=126) -5.6 ±1.1 kg, from t₀ to t₃ (n=94) -4.9 ±1.2 kg and from t₀ to t₄ (n=77) -4.7 ±1.4 kg.

3.1. Mixed Model Analysis (Starting at t₀ with n=381)

All patients who started the behavioral weight loss program (n=381) were included in the mixed model analysis regarding all measured parameters. For all assessed parameters, mean values from t₀ to t₄ and tests for significant time effects using mixed model analysis are depicted in Table 1; results are listed for the whole sample, anthropometric data also separately for women and men. Available patient measurements with data at different assessments for each parameter are shown in Supplementary Information, Table A4.

Average ±standard error of the mean. Derived from the mixed model analysis, significance levels depicted through asterisks on the respective mean value refer to tests of differences between baseline (t₀) and the respective time point, controlling for age and multiple testing according to Bonferroni [19]. Distinct p-values are shown for tests of significant time/treatment effects within the mixed model analysis. ALAT, alanine aminotransferase; ASAT, aspartate aminotransferase; EUROHIS-QOL, EUROHIS-QOL 8 item index [12]; GGT, gamma glutamyltransferase; HbA1c, hemoglobin A1c, HDL, high density lipoprotein; LDL, low density lipoprotein; PHQ-9, Patient Health Questionnaire [13]. ^§Patients with type 2 diabetes without changes in diabetes medication during the course of the program; n=24 were thus excluded from this analysis: in 17 patients glucose lowering medication was intensified, in 7 patients reduced; ^†patients with type 2 diabetes without changes in diabetes medication during the course of the program and without insulin therapy; n=45 were thus excluded from this analysis; ‡patients without changes in lipid lowering medication, during the course of the program; n=15 were thus excluded from this analysis: in 10 patients lipid lowering medication was intensified, in 5 patients reduced. ***p<.001, **p<.01, *p<.05.

Compared to t₀, most health parameters showed statistically significant decreases at t₄. Specifically, the BMI was reduced from 44.3 ±0.4 kg/m² at t₀ to 43.4 ±0.5 kg/m² at t₄ (p<.01), with a significant treatment effect over the whole time course (p<.001) (Table 1). HOMA-IR decreased from 6.0 ±0.3 at t0 to 5.0 ±0.3 at t₄ (p<.05; whole time course p<.01). The HbA1c between t₀ and t₄ did not change statistically significant for the whole cohort (p=.1), but the mixed-model analysis over the whole time course showed a statistically significant effect (p<.001). Analyses of subgroups of diabetic and patients without diabetes revealed same tendencies as the whole group (Table 1, Figure 3A). For the whole sample, low density lipoprotein (LDL) cholesterol decreased from 3.3 ±0.1 mmol/l at t₀ to 3.2 ±0.1 mmol/l at t₄ (p<.05; whole time course p<.05). The quality of life-score (EUROHIS-QOL) improved from 3.3 ±0.0 points at t₀ to 3.6 ±0.1 points at t₄ (p<.001; whole time course p<.001; Figure 3B) One of the EUROHIS-QOL items, unrelated to health “How satisfied are you with the conditions of your living place?”, did not change, unlike all other items (Supplementary Information, Table A5 and A6). The depressive symptoms score (PHQ-9) decreased from 3.2 ±0.0 points at t₀ to 3.6 ±0.1 points at t₄ by (p<.001; whole time course p<.001). A total of 217 persons completed the PHQ-9-score for depressive symptoms at least twice, 43 (20%) people had a reduction of 5 points or less, 17 (8%) had an increase of 5 points or more.

For analysis of BMI changes, the model containing terms up to the cubic terms was optimal. All patients with at least two weight measurements were included, starting at t₀ with n=268. BMI reached its lowest value approximately two years after beginning of the program. Subsequently, BMI increased moderately, but remained well below the baseline value (Figure 4).

3.2. Completer Analysis of Variance of All Measured Parameters (n=77)

For the analysis of variance of all measured parameters, all patients who completed t₄ were included (n=77). Mean weight loss for this group was 4.6 ±1.4 kg (-3.4 ±1.1%, both t₀ vs. t₄ and whole time course p<.001, Supplementary Information, Table A7). At the last measurement (t₄), 28 (36%) persons had a weight loss of 5% or more and 16 (21%) of 10% or more compared to the initial weight. Further, HbA1c decreased by 0.1 ±0.1%-points (p<.05; whole time course p<.001) and HOMA-IR by 0.9 ±0.3 (p<.01; whole time course not statistically significant with p=.16, Table A8). For the whole sample, LDL cholesterol showed a trend for a decrease with -0.1 ±0.1 mmol/l (p=.07; whole time course p=.18). The quality of life-score (EUROHIS-QOL) improved by 0.2 ±0.1 points (p<.01; whole time course p<.05) and the depressive symptoms score (PHQ-9) did not change statistically significantly (p=.45; whole time course p=.19).

3.3. Exploratory Analysis of Healthcare Cost Changes Over Time (Starting Three Years Before Program)

When comparing the average costs for outpatient care before and during treatment, the trend increase did not reach a significant level (p=.12). Average costs for drugs also numerically changed up and down per year over time and increased continuously from the second year of the program. When comparing the average costs for drugs before and during treatment, the trend increase did not reach a significant level (p=.24). When comparing the average costs for hospital care before and during treatment, the trend increase did not reach a significant level (p=.51). In summary, all healthcare costs were numerically higher before and after the program, but none of the observed changes were statistically significant (Figure 5).

4. Discussion

Weight loss through behavioral weight loss programs leads to a clinically relevant improvement of obesity associated comorbidities [8]. However, therapy adherence and coverage of costs of such programs by healthcare providers is essential for success. Often, the duration of programs fully financed by health insurances is too short or coverage is not available at all, or patients quit programs prematurely yielding only partial effects. Reasons for limited motivation to participate include unrealistic weight loss expectations and a non-individualized therapy [24,25]. This may lead to an initial weight loss followed by more or less rapid weight regain, resulting in little benefits for patients [26]. Further reasons for unsatisfying therapy effects may be a single focus on nutritional therapy or physical activity [27], but not a combination of both interventions, accompanied by cognitive behavioral training as recommended by international guidelines [5,28].

Often, behavioral weight loss programs include a phase of formula diet at the beginning of the treatment [8,10,11]. In our Leipzig Obesity Management program, we did not include a formula phase to be able to offer the program also to people who are not able to conduct a meal replacement diet due to personal preference, medical reasons or the high costs of such therapeutic options. In addition, the German healthcare system does not fund the use of meal replacement products.

During the first intensive year of the multimodal treatment program, patients who absolved the first year lost 5 kg of their initial bodyweight. The second to fourth year aimed to maintain weight loss achieved in the first year. This goal was achieved considering all patients from the beginning up to the last measured weight. There was no statistically significant weight regain after t₁, and mean bodyweight at the end of the program was significantly lower than at the beginning of the program.

Up to date, only few long-term behavioral weight loss intervention studies without meal replacement and with an intervention lasting longer than 1 year are available. In the US American “Practise-based Opportunities for Weight Reduction” (POWER) study, participants were randomized in two different intervention and one control group. One of the intervention groups had a remote support via telephone or internet, the other intervention group had an in-person weight loss support. The intervention groups had a 6 months intensive therapy with 1.5 years follow-up program, which had more frequent diet counselling visits or generally contact to clarify questions and increase motivation (in-person group: 30 group sessions, 10 individual sessions, 21 phone contacts; remote support group 33 phone contacts) than our program during the same time period. At the end of observation, a weight loss of 5.2% or 5.1 kg was achieved in the in-person treatment group, 5.0% or 4.6 kg in the remote support group, and 1.1% or 0.8 kg of the control group [29]. The more intense follow up program may explain higher weight loss outcomes than in our study. Furthermore, the mean baseline BMI of the intervention groups were 36.0 kg/m² in the remote support group and 36.8 kg/m² in the in-person treatment group, which also differed clearly from the population of our study with an average BMI of 44.3 kg/m².

As stated in the obesity guidelines of the American Diabetes Association of 2022, a weight reduction of 3 to 5% can relevantly improve health parameters as glycemic control and other cardiovascular risk factors [30]. In our cohort, all assessed parameters of glucose metabolism improved significantly in both groups, participants with and without diagnosis of type 2 diabetes at baseline. As expected, also in patients without diabetes, insulin resistance was common, mean HOMA-IR at baseline was 5.3 ±0.2. According to data from the American Diabetes Prevention Program, risk of diabetes in this group is extremely high [31].

In both, persons with and without diabetes, a reduction of the LDL cholesterol concentrations leads to a risk reduction for atherosclerotic cardiovascular disease [32,33]. The current analysis showed a significant decrease of LDL cholesterol during the time of treatment. Even a diet induced change of 0.1 mmol/l, as achieved by the participants of our program, has a positive effect on the risk for cardiovascular disease [34].

The C-reactive protein (CRP) is associated with inflammatory processes in the human body. In our study, concentrations decreased with bodyweight reduction. It is known that life style changes, especially exercise, have positive effects on CRP [35]. Another influencing factor appears to be nutritional behavior. Many components of a healthy and balanced diet seem to have a positive effect on inflammatory body processes [36]. All of these factors could have contributed the CRP decrease seen in our cohort.

In our study, scores for quality of life and depressive symptoms improved significantly. However, because the study lacked a placebo control group, nonspecific treatment expectation affecting mood cannot fully be ruled out (Hawthorne effect) [37]. Nevertheless, the EUROHIS-QOL item on the life situation did not change, which could indicate that effects on PHQ-9 and EUROHIS-QOL were driven by the beneficial medical effects of the behavioral weight loss program. In contrast, in the other not health related EUROHIS-QOL item about the financial situation, we saw a significant improvement, probably because of an improved capability of our patients to being employed. However, this is speculative because socioeconomic data were not assessed at measurements t₁-t₄. For PHQ-9, a change of five or more points is considered clinically relevant [38]. In our cohort, 20% of the patients who completed the questionnaire on at least twice had an improvement of five or more points from baseline to the last available measurement, whereas only 8% of the participants a deterioration of this magnitude.

Drop-out rates during the four years of our program were between 36% (first treatment year) and 11% (fourth year), the mean drop-out rates per treatment year was 23%. Additionally, 19% of patients starting the program were transferred to surgical obesity treatment during the program and did not complete all four years of the behavioral weight loss treatment (Figure 1). Neri et al. (2024) investigate reasons for dropouts in adults with overweight and obesity absolving a cognitive behavioral treatment [39]. They reported a dropout-rate between 5 and 62% in the studies included in their review. The focus was primarily on the reasons for dropping out. We only recorded these in categorized form. However, health and familial reasons appear to be a common reason. Looking at the dropout-rates of the studies in relation to the duration of the intervention, the duration seems to have an impact on the dropout-rate [39]. To our knowledge, no other evaluation of a behavioral treatment program under real world conditions with an intervention period of four years has not yet been published. Thus, a comparison of the dropout-rate is severely limited and our results fit into the broadly diversified field of dropout-rates.

Although the evaluation of cost efficacy of reported behavioral weight loss programs is difficult due to inconsistent methodologies of analyses, behavioral weight loss interventions seem to be cost-effective [40], with a non-linear relationship between health care costs and BMI-change [41]. Without intervention, in persons with obesity further weight gain occurs with increasing age regardless of the initial weight [3]. Consequently, considering previous literature [42], we hope that the program costs will be compensated on the long run by reduced health care costs for obesity associated comorbidities. However, this verification requires a longer observation period.

Some limitations of our study have to be discussed. First, our study was not a randomized clinical trial with an untreated group, but an uncontrolled observation of real world data. Thus, we cannot exclude unspecific treatment effects. However, our aim was to evaluate a behavioral weight loss program which was fully covered by a health insurance company and open to all patients insured with that company under real world conditions. We wanted to investigate if our program, which clearly differs in patient selection and available resources from controlled study conditions, which are usually far away from real world situations, can achieve meaningful benefits for participants. Second, health care costs could only be collected for fully completed therapy years. From our point of view, the cost analysis nevertheless provides a good overview for an initial assessment of cost efficiency. Third, our evaluation only covers the effects on body weight and health parameters during the treatment period. For evaluating maintenance of change, a follow up assessment in the years after the end of the program should be conducted. A study on this issue is already ongoing.

5. Conclusions

The current evaluation shows that the four-year Leipzig Obesity program, which is fully funded by a health insurance company, is a promising treatment approach for people with obesity class 2 and 3. A comparison with other studies showed that the intensity of follow-up care is decisive for long-term success of therapy. Health effects were maintained until the end of the program, although there was a slight vanishing trend of achieved medical benefits after the first intensive year of therapy. With regard to health care costs, there was no significant change during the time of evaluation. Here, more accurate studies with longer observation periods are necessary for a better assessment. However, the most outstanding feature of this evaluation is the real world data. In contrast to randomized, controlled studies under research conditions, our real world data allows an unbiased view of the achievable clinical effectiveness of the Leipzig Obesity Management program and is thus an important contribution to the development of treatment strategies for obesity.

6. Patents