SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

Intensive obesity treatment is mandated by federal health care reform but is costly. A partial subsidy for obesity treatment could lower the cost of treatment, without reducing its efficacy. This study sought to test whether a partial subsidy for obesity treatment would be feasible, as compared to a fully subsidized intervention. The study was a pilot randomized trial. Participants (n = 50) were primary care patients with obesity and at least one comorbid condition (diabetes, hypertension, dyslipidemia, or obstructive sleep apnea). Each participant received eight weight loss counseling visits as well as portion-controlled foods for weight loss. Participants were randomized to full subsidy or partial subsidy (2 vs. 1 meal per day provided). The primary outcome was weight change after 4 months. Secondary outcomes included changes in blood pressure, waist circumference, and health-related quality of life. Participants in the full and partial subsidy groups lost 5.9 and 5.3 kg, equivalent to 5.3% and 5.1% of initial weight, respectively (P = 0.71). Changes in secondary outcomes were similar in the two groups. A partial subsidy was feasible and induced a clinically similar amount of weight loss, compared to a full subsidy. Large-scale testing of economic incentives for weight control is merited given the federal mandate to offer weight loss counseling to obese patients.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

The recent passage of federal health care reform includes a provision that services receiving an “A” or “B” recommendation from the US Preventive Services Task Force will be covered without cost sharing (1). One such service is intensive weight loss counseling, defined by the Task Force as a minimum of twice monthly counseling for a minimum of 3 months (2). A total of 34% of the US population is obese (3), including 39% of patients seen in primary care (4). Thus, the cost of providing high-intensity counseling to all obese persons is potentially overwhelming. The treatment provided in the Diabetes Prevention Program (DPP), a highly effective lifestyle intervention for weight loss (5,6), costs ∼$1,200 per person for the first year of treatment (although substantially less if delivered in a group setting) (7,8).

If weight management services are to be provided to a large number of individuals, the need to provide treatment must be balanced against the potential costs. For example, the cost of treatment to health care payers (e.g., Medicare) should be limited. Also, higher cost treatments should be targeted to appropriately selected individuals to maximize their effectiveness. A partial subsidy has been modeled as a cost-effective way to deliver obesity treatment (9). Previous studies have demonstrated the effect of economic incentives for weight loss including deposit contracts, partial reimbursement of treatment costs, food provision, and direct cash incentives for weight loss (10,11,12,13,14,15,16,17). However, additional data are needed to determine how best to use the principles of economics to improve the management of obesity (18).

This pilot randomized trial was designed to test the feasibility and initial efficacy of a partial (∼40% of total treatment costs) versus a full in-kind subsidy during a short-term (4 months) weight loss program. The subsidy was directed to the use of portion-controlled foods for weight loss, a modality that health care payers normally do not cover (19). We hypothesized that partial subsidy of the program would be feasible, as compared to full subsidy, and would produce clinically similar weight loss—i.e., that the partial subsidy would deliver equal clinical benefit at lower cost. Standard economic theory suggests that individuals who have demonstrated a willingness to pay for a service are likely to value the service more highly.

Methods and Procedures

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

Participants

Participants (n = 50) were recruited from primary care practices at the University of Colorado (Denver, CO). Participants had to be 18 years of age or older, have a BMI of 30–49.9 kg/m2, and have a diagnosis code for one of the following: (i) impaired fasting glucose or type 2 diabetes; (ii) elevated blood pressure, including hypertension; (iii) dyslipidemia (high low-density lipoprotein cholesterol, low high-density lipoprotein cholesterol, or hypertriglyceridemia); or (iv) obstructive sleep apnea. A list of patients was generated from the electronic database maintained at the University of Colorado Denver (20), and recruitment letters were then mailed to potentially eligible individuals. Persons that called to express interest in the study were screened by telephone to ensure that they did not have medical contraindications to weight loss and did not have a weight change of ≥5% within the past 6 months. Other exclusion criteria included severe mental illness (schizophrenia or bipolar disorder), untreated major depression, and medications associated with large weight fluctuation (e.g., glucocorticoids, second generation antipsychotics, and weight loss medications). The study was given expedited approval by the Colorado Multiple Institutional Review Board.

Screening and randomization

Individuals found to be eligible after the phone screen attended two in-person screening visits prior to beginning weight loss. At the first visit, participants gave informed consent, completed questionnaires, and underwent noninvasive physical measurements—height, weight, waist circumference, and blood pressure. (Blood tests (e.g., glucose, lipids) were not performed due to the study's small budget and because the beneficial effect of a 5–10% loss of initial weight has previously been well characterized (21,22,23).) At a second screening visit, participants underwent an assessment of weight loss readiness, using the Weight and Lifestyle Inventory (24). Individuals who satisfactorily completed both screening visits were randomized to a treatment condition (full subsidy or partial subsidy) and began weight loss counseling. Randomization was conducted by a biostatistician at a separate location.

Portion-controlled foods

Participants were randomized to receive all of the food necessary to follow a partial meal replacement diet (“full subsidy”) or to receive half the amount of food (“partial subsidy”) needed. The study compared in-kind subsidy of food costs, with portion-controlled foods distributed by the study team. The study completely covered the costs of weight loss counseling. No money was distributed to participants for food purchases or for any other purpose.

A structured meal plan with meal replacements and portion-controlled foods was used to replace two out of three meals per day. This eating plan has been shown to produce additional weight loss beyond that of isocaloric diets composed of conventional foods (25,26,27). Weight loss foods were donated by Health Management Resources Corporation (Boston, MA). These included meal replacement shakes (160–170 calories, 16 g protein) and seven types of entrees (200–250 calories, 10–15 g of protein). Participants randomized to the partial subsidy condition were advised to purchase a second meal of portion-controlled foods on their own, and suggestions were provided for store-bought products. All participants were advised to use meal replacement shakes and bars as well as portion-controlled entrees for two meals per day, along with snacks of fresh fruits and vegetables. They were advised to consume the third meal on their own. Guidelines were provided for the third meal (i.e., 3–4 ounce serving of protein, ½ to 1 cup of carbohydrate, 1–1.5 cups of steamed vegetables).

Counseling intervention

Each participant was provided with eight weight loss counseling visits during the study period. At each visit, participants were weighed. The counselor then collected and reviewed food records, discussed the educational content of the current session, and set goals prior to the next visit. Visits lasted 20–30 min each. Visits were led by a nonphysician provider, given the time demands of primary care practice and data suggesting that PCPs perceive their own effectiveness to be limited (28). A trained lay provider was used, based on the lower cost (29). Written materials used for counseling were adapted from the DPP—these handouts have been made public for dissemination (http:www.bsc.gwu.edudpplifestyledpp_part.html). Prior to starting the intervention, the Principal Investigator (A.G.T.) and counselor (S.F.) spent one half day reviewing the etiology and treatment of obesity and an additional half day role playing the modified DPP handouts, alternating as counselor and participant. Additionally, the Principal Investigator observed the first several sessions with study participants and then intermittently thereafter.

Outcomes

The primary outcome of the study was weight change. Our goal was to achieve an average weight loss of ∼7% of initial weight, similar to the weight losses in the DPP. Secondary outcomes included changes in BMI, blood pressure, waist circumference, and health-related quality of life. Weight was measured on a digital scale (Befour PS 6600; Befour, Sackville, WI). Height was measured using a wall mounted stadiometer. Blood pressure was measured with an automated cuff (Omron HEM 780; Omron, Kyoto, Japan). Three blood pressure measurements were taken, with a 1-min rest between measurements, and averaged. Waist circumference was measured using a Gulick II tape measure, with successive measurements taken until two consecutive measurements were within 1 cm of each other. These two measurements were then averaged. Health-related quality of life was measured with the Medical Outcomes Study, Short Form 12 and with the Impact of Weight on Quality of Life (30,31).

Statistical analysis

Weight change data were analyzed with independent sample t-tests, as were changes in BMI, blood pressure, and waist circumference. For participants who dropped out prior to completing the study, an empirically derived weight regain of 0.3 kg per month was used to impute an end-of-study weight (32). For participants that dropped out, weight gain was imputed until baseline weight was reached or until the end of the intervention period. A secondary analysis changed examined weight losses only among study completers.

Linear regression was used to check for associations between baseline variables and the primary outcome, weight change. Only prerandomization variables were tested for their association with weight change (33). Variables associated with weight change (P < 0.2) were entered into a multivariable regression. Post hoc analyses examined whether intra-study markers of adherence were associated with weight change. Measures of adherence were tested only among participants who completed the study, as imputation of values for participants who dropped out could have biased these analyses.

Sample size

It was not feasible to power the study as a noninferiority trial. Such a study would have required several hundred participants per arm. However, with 25 participants per arm, we had 81% power to detect a difference in weight loss of 3 kg between the two groups, assuming a two-sided α of 0.05 and a pooled s.d. of 3.7 kg after 3 months of treatment (34). With n = 37 completers, we had 82% power to detect a difference in weight loss of 3.5 kg between the two groups.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

Participant flow

Participant flow is outlined in Figure 1. A total of 50 patients were randomly assigned to one of the two groups. One individual developed vasculitis after randomization, and a second participant was removed from the study after being diagnosed with breast cancer. These two individuals were considered censored for medical reasons and were not included in final analyses. Thus, the total sample size for the primary outcome was 48 (n = 24 per group).

image

Figure 1. Flow of study participants. *Other exclusions: declined randomization (n = 10); recent weight fluctuation (n = 8); prior weight loss surgery (n = 5); BMI <30 kg/m2 or ≥50 kg/m2 (n = 5); missed screening visits (n = 2); no comorbid conditions (n = 1); not receiving care at University of Colorado (n = 2); other, e.g., lack of transportation (n = 4).

Download figure to PowerPoint

Baseline characteristics

Baseline characteristics of the study participants are shown in Table 1. Half of study participants were men, and 25% were either African American or Latino. Participants had an average BMI of 38.1 kg/m2 and were taking an average of 5.4 prescription medications. Over one-third of participants had diabetes, and the majority had hypertension and dyslipidemia. Those in the partial subsidy condition were more often female, had a BMI 1.8 units higher, and a lower prevalence of diabetes and sleep apnea. None of the differences between the two groups in baseline characteristics were statistically significant.

Table 1.  Baseline characteristics of study participants
inline image

Primary outcome

Study outcomes are shown in Table 2. Participants in the full subsidy and partial subsidy groups lost 5.9 and 5.3 kg, equal to 5.3% and 5.1% of initial weight (P = 0.71 for difference). However, the 95% confidence interval for the between-group difference in weight change was −2.7 kg to 3.9 kg. Thus, we cannot rule out a difference in weight change in this range. Changes in BMI paralleled weight losses.

Table 2.  Outcomes
inline image

A completers-only analysis of the full and partial subsidy groups (n = 17 and n = 20, respectively) showed losses of 7.4 and 6.3 kg (P = 0.55 for comparison), equal to 6.6% and 6.0% of initial weight.

Secondary outcomes

No significant differences were seen between the groups for changes in blood pressure or waist circumference. Both groups sustained improvements in these parameters. Regarding changes in health-related quality of life, overall decreases were seen in the Medical Outcomes Study, Short Form 12, while improvements were seen in the Impact of Weight on Quality of Life. Between-group differences for changes in health-related quality of life were not significant.

Regression analysis

Univariable association of baseline variables with weight change is shown in Table 3. Only age and ethnicity were associated with weight change in univariable analysis (P < 0.20) and thus were entered into a multivariable regression. In multivariable analysis, the total variance in weight change explained by these two variables was 15.9%.

Table 3.  Baseline predictors and intra-study correlates of weight change
inline image

Post hoc analyses

Post hoc analyses were conducted among completers (N = 37) to assess whether postrandomization markers of adherence correlated with weight change. These correlations are also shown in Table 3. Three measures of adherence were tested—number of visits attended, amount of portion-controlled foods used, and number of days that food records were kept (29,35,36). Of these three measures, two (portion-controlled food use and record keeping) were significantly associated with weight change. All three measures of adherence correlated significantly with each other.

Attrition

Attrition was higher in the full subsidy group (29%) than in the partial subsidy group (17%), but the difference was not statistically significant (P = 0.30).

Adverse events

A total of 18 adverse events occurred among 15 participants. Two of these were serious adverse events leading to hospitalization. One of these was a participant who developed cholecystitis and required gall bladder removal. The other serious adverse event, vasculitis, was unrelated to study participation but led to censoring of the participant for study outcomes.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

In this pilot study, participants that paid for half of their food in an obesity treatment program lost a similar amount of weight as did participants that received all of their food for free. Participants, all of whom had obesity and at least one weight-related medical condition, lost 5.3% and 5.1% of initial weight (full or partial subsidy, respectively). The study fell short of the goal weight loss of 7% of initial weight. However, participants had only eight visits (instead of 16, as in the DPP), and weight losses reached the minimum level of clinical significance (5%) in both groups. Weight losses were clinically similar in the two groups. Because the study could not feasibly be powered as a noninferiority trial, a statistically significant difference in weight loss between the groups cannot be ruled out. As expected, participants that were more adherent during the trial lost a greater amount of weight (36). Both groups sustained improvements in blood pressure and in waist circumference which would be clinically significant, if maintained. Changes in health-related quality of life were similar in the two groups. Attrition was higher in the full subsidy group, although this was not statistically significant. It is possible that individuals in the full subsidy group valued the treatment less than those individuals who were asked to pay part of the cost of treatment and thus, would have been more likely to drop out.

These results suggest that a partial subsidy could be an effective method to lower the cost of obesity treatment, without reducing its efficacy. Although formal cost-effectiveness analysis is beyond the scope of this study, we estimated that the per-person cost of treatment was $552 for the partial subsidy group and $944 for the full subsidy group. Thus, the combination of intensive weight loss counseling plus an adjunctive therapy (portion-controlled foods) can be provided for ∼$550. (Of these amounts, the cost of the counseling (∼$160) represents the minority, whereas the retail cost of the portion-controlled foods ($392 in the partial subsidy group, $784 in the full subsidy group) represents the majority of costs.) The total cost of treatment in the partial subsidy condition is less than the cost of attending Weight Watchers for 1 year (∼$640) (37) and is similar to the cost of group weight loss interventions that are beginning to be reimbursed by health care payers (7,38). Participants in the two groups used a similar number of portion-controlled food products, indicating that individuals provided with a partial subsidy purchased these foods on their own. If these results could be replicated in larger trials, the implications would be clear: more individuals can undergo treatment at a lower cost. Such a finding would be highly relevant, given the mandate of health care reform to provide intensive obesity treatment to more people (1).

Our results are consistent with those from previous studies suggesting that economic incentives are a useful adjunct to traditional lifestyle modification for weight loss. Jeffery et al. found that monetary contracts were effective at inducing weight loss (12,13). Butsch and colleagues showed that patients participating in a university-based weight loss program achieved better attendance and greater weight losses when offered 50% reimbursement for the cost of the program, as compared to no reimbursement (10,11). Volpp et al. showed that two types of incentives, a lottery or a deposit contract with matching payments, both were more effective than daily weigh-ins alone (16). A second study by the same group of investigators found that a deposit contract led to greater weight loss at 32 weeks (39). (A follow-up visit 36 weeks after the intervention revealed significant weight regain, although no treatment or incentives were provided during the follow-up period.) Finkelstein et al. tested the effect of direct cash incentives per percentage of initial weight lost. They showed that a larger incentive led to greater weight losses during the first 3 months and that weight losses were similar after 6 months, when incentives were equalized across groups (15).

The results of the current trial also are consistent with a study by Jeffery, Wing, and colleagues, showing that food provision is a useful tool for weight loss in the short term (14,17,40). (The effect of food provision was attenuated, however, after 30 months (14,17,40).) A separate study by Wing, Jeffery, and colleagues reported that the use of structured meal plans induced slightly, but not significantly, more weight loss than did the provision of food (17). The results of the latter study by Wing and Jeffery suggest that the specificity of the diet prescription is the key ingredient, rather than the provision of food. However, six randomized trials included in a meta-analysis by Heymsfield et al. (17) plus at least two additional randomized trials (41,42), have reported that use of meal replacements leads to weight losses 2.5–3 kg greater than diets comprised of conventional foods. Therefore, the inclusion of portion-controlled foods in weight loss interventions appears to be an important component of treatment. The current study did not have a true control group to test whether a recommendation to use meal replacements would lead to similar weight losses, as compared to actual provision of food.

The present study has several important limitations. First, as noted above, the trial could not feasibly be powered as a noninferiority study. This would have required a sample size of several hundred participants per arm. Second, the outcomes assessor (S.F.) could not feasibly be blinded to treatment group assignment. Third, no blood testing was done, so the effect on glycemic control, lipids, and other markers of weight-related cardiovascular risk is unknown. Fourth, the results are short-term, and it is unclear to what extent weight losses will be maintained. Finally, study participants were generally middle class and with private insurance. Thus, it is unclear whether the use of partial subsidy can be generalized to low income populations (43). Strengths of the study include the use of a lower cost counselor, the recruitment of a patient population that was representative of primary care practice (50% men; 25% African American or Latino), and the achievement of a clinically significant weight loss.

In summary, partial in-kind subsidy of a short-term, intensive weight loss intervention, consisting of counseling and portion-controlled foods, was feasible and induced weight losses clinically similar to those achieved via a full subsidy. Participants in both groups lost over 5% of initial weight. Future studies should test the effect of cash reimbursement and/or in-kind subsidy for weight management, with a long-term goal of integrating economics into the clinical management of obesity (18). In the context of federal health care reform and an increased emphasis on prevention, evidence-based and financially sustainable provision of obesity treatment will be needed.

Acknowledgment

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

The authors thank Dr Libby, PhD, for her review of the manuscript, and Dr Ogden, PhD, for statistical consultation. We thank Drs Kelly, MD, Penaloza, MD, and Pace, MD, for their assistance in patient recruitment. We also thank Health Management Resources Corporation (Boston, MA) for their generous support of the study and we thank the Division of General Internal Medicine at the University of Colorado Denver for internal funding. The study funders had no role in the design of the trial or in the writing of this manuscript.

Disclosure

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES

J.O.H. serves on the advisory board for Nutrisystem. The other authors declared no conflict of interest.

REFERENCES

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods and Procedures
  5. Results
  6. Discussion
  7. Acknowledgment
  8. Disclosure
  9. REFERENCES