Skip Navigation U.S. Department of Health and Human Services
Agency for Healthcare Research Quality
Archive print banner
Use of Behavioral Therapies for Treatment of Medical Disorders

This information is for reference purposes only. It was current when produced and may now be outdated. Archive material is no longer maintained, and some links may not work. Persons with disabilities having difficulty accessing this information should contact us at: Let us know the nature of the problem, the Web address of what you want, and your contact information.

Please go to for current information.

Part 1. Impact on Management of Patients with Diabetes Mellitus (continued)



The analytic question of whether behavioral therapies for individuals with diabetes mellitus improve glycemic control or other health outcome measures requires definition of several terms.

Diabetes. The criteria developed by the American Diabetes Association (Diabetes Care 2003; 26:S33-50) were used to determine a diagnosis of diabetes mellitus. Basically, these are: symptoms of diabetes and a casual plasma glucose >200 mg/dl, OR fasting plasma glucose >126 mg/dl, OR 2-hour plasma glucose >200 mg/dl during an oral glucose tolerance test (OGTT).

Behavioral therapy or intervention. Behavioral therapy or intervention was defined to include four categories of interventions:

  1. Cognitive behavioral therapy—using one or more interventions derived from the cognitive behavioral model of behavior change, including techniques such as cognitive restructuring, motivational interviewing, positive reinforcement, contingency management, problem-solving, goal setting, self-monitoring, skills training, and modeling.
  2. Relaxation-based interventions—interventions whose primary focus is on teaching patients to relax including progressive relaxation training, electromyographic, or thermal biofeedback.
  3. Behavioral diet/exercise interventions—interventions whose goal was to influence health outcomes through behavioral changes in diet or exercise using techniques such as caloric monitoring, portion control, exercise regimens, and individualized dietary prescriptions.
  4. Blood glucose awareness training—interventions whose goal was to teach patients how to interpret physical symptoms, moods, feelings, and external cues to estimate blood glucose level.

Behavioral therapies or interventions would not include: educational programs that provide didactic education or information to the patient on how to manage diabetes with minimal (e.g., 10 minutes or less) or no interactive behavioral training, traditional multifamily therapy; education only; traditional group therapy; peer group counseling; and traditional family-oriented support.

Health outcomes. Three categories of health outcomes were included:

  • Glycemic control as measured by glycosylated hemoglobin (either as hemoglobin A1C (HbA1C), hemoglobin A1 (HbA1) or glycosylated hemoglobin (GHb) or glucose measurements (e.g., fasting blood glucose or area under the serum glucose curve).
  • Diabetes related health events, such as foot infection, amputation, or diabetic ketoacidosis.
  • Control of risk factors that can enhance the potential for poor health outcomes in diabetic patients, including obesity, hypertension, and hyperlipidemia.

Among studies that included health outcomes in at least one of the above categories, we also reviewed data provided on subjective outcomes, including health-related quality of life, adjustment to disease, self-efficacy, stress/hassles, distress, and mood.

Search Strategy

There were two basic search strategies developed for the systematic literature review. The first of these combined the MeSH term "diabetes mellitus" with a behavioral therapy concept (implemented using MeSH terms "'behavioral disciplines and activities'/or cognitive therapy"). The second search strategy focused on patient education using MeSH terms "diabetes mellitus" and "patient education." Both searches employed a standard search strategy for randomized control trials. The strategies were conducted in MEDLINE®, PsychINFO, and Web of Science (1966 through June 2003) and were limited to articles pertaining to humans and published in the English language. The exact texts of the search strategies are provided in Appendix B.

Supplemental searches were conducted in Web of Science and the National Guideline Clearinghouse™. References lists of relevant systematic reviews and meta-analyses were also checked.

Additional articles were included at the suggestion of peer reviewers and as a result of ongoing secondary searches of the literature such as articles cited in other recent systematic reviews and meta-analyses. A recent systematic review of randomized controlled trials on the effectiveness of self-management training in type 2 diabetes by Norris, Engelgau, and Narayan (2001) was particularly useful.

Literature Screening

Abstracts and the full-text versions of articles identified in the MEDLINE® and other searches were screened by the investigators against six exclusion criteria:

  • Study subjects are not diabetic or hyperglycemic.
  • Majority of study subjects are not adults.
  • Study design is not a randomized controlled trial.
  • No medical outcome is reported.
  • No behavioral intervention is reported.
  • "Other" reason (e.g., editorial, review article).

Overall, there were 736 potentially relevant articles reviewed for this study. After an initial screening of their titles and abstracts, 209 (28 percent) were reviewed in their full-text versions. Of these, 61 (29 percent) met our inclusion criteria, and 148 (71 percent) were excluded.

Data Abstraction

For each of the 61 included articles, basic study parameters were abstracted into an evidence table. These included:

  • Study identification (authors, publication year).
  • Inclusion and exclusion criteria (for the study being abstracted).
  • Description of study design.
  • Description of patient population (number in each study group, number of drop-outs, baseline measures such as
  • HbA1C).
  • Description of interventions, treatment duration; outcomes or results (below).
  • Quality assessment (below).

The "Outcomes/Results" column of the table reported three basic categories of findings:

  • Metabolic control (e.g., HbA1C).
  • Significant health events (e.g., hospitalization, physician or emergency department visits).
  • Measures of risk (e.g., specific measure of weight, body mass index, blood pressure).

The last column of the evidence table indicates the presence or absence of specific criteria used to assess each article's internal and external validity. Criteria used to determine internal validity are:

  • Randomization.
  • Clear description of the randomization method.
  • Concealment of allocation (e.g., through the use of sealed envelopes).
  • Details of the study's blinding method (patient, investigators, outcome assessors).
  • The number of withdrawals in each study group.

Factors affecting external validity are:

  • The presence or absence of clear description of the patient population.
  • Description of the intervention(s) that are detailed enough to reproduce.
  • Codification of intervention in manual.
  • Description of provider training.
  • Patient assessment for a Diagnostic and Statistical Manual for Mental Disorders (DSM) diagnosis.

Study biases, study limitations, and other comments are noted at the bottom of the last column of the evidence table.

A psychology graduate student and a research assistant with a bachelor's degree in psychology completed the initial data abstraction for each included article. Each data abstraction was overread by a physician and a psychologist.

In addition to the evidence table, we summarized the results of the reviewed studies in two ways. First, we tabulated the proportion of studies that indicated a statistically positive effect in any of the primary clinical outcomes of interest. Since statistically significant results would only be expected in 5 percent of studies by chance alone, this provides a general-purpose, albeit crude, assessment of the presence of a treatment effect in a pool of diverse studies. Second, we focused on studies in which the outcome measure was glycemic control in terms of GHb, HbA1, or HbA1c and for which mean and variance could be estimated. For these studies, we calculated effect sizes (means and confidence intervals) using the Comprehensive Meta Analysis software (Englewood, NJ).

Return to Contents
Proceed to Next Section

The information on this page is archived and provided for reference purposes only.


AHRQ Advancing Excellence in Health Care