Michael P. Pignone, M.D., M.P.H.a; Alice Ammerman, Dr.P.H., R.D.b; Louise Fernandez, R.D., P.A.-C., M.P.H.b; C. Tracy Orleans, Ph.D.c; Nola Pender, Ph.D., R.N., F.A.A.N.d; Steven Woolf, M.D., M.P.H.e; Kathleen N. Lohr, Ph.D.f; Sonya Sutton, B.S.P.H.g
The authors of this article are responsible for its contents, including any clinical or treatment recommendations. No statement in this article should be construed as an official position from the Agency for Healthcare Research and Quality, or the U.S. Department of Health and Human Services.
Address correspondence to: Michael P. Pignone, M.D., M.P.H., Division of General Internal Medicine, UNC School of Medicine, 5039 Old Clinic Building, CB No. 7110 UNC Hospitals, Chapel Hill, NC 27599-7110; E-mail: pignone@med.unc.edu.
Select for copyright and reprint information. The USPSTF recommendations based on this review are online.
The summaries of the evidence briefly present evidence of effectiveness for preventive health services used in primary care clinical settings, including screening tests, counseling, and chemoprevention. They summarize the more detailed Systematic Evidence Reviews, which are used by the U.S. Preventive Services Task Force (USPSTF) to make recommendations.
Epidemiology
Methods
Results
Discussion
Acknowledgments
References
Notes
Diseases associated with unhealthy dietary behavior rank among the leading causes of illness and death in the United States.1,2 Major diseases in which diet plays a role include coronary heart disease, some types of cancer, stroke, hypertension, obesity, osteoporosis, and non-insulin-dependent diabetes mellitus.1 All of these diseases are major causes of morbidity and mortality in this country.3 Although diet is associated with multiple health outcomes, the ability of counseling to change dietary patterns and improve health is unclear. In this report, counseling is defined as a cooperative mode of interaction between the patient and primary care physician or related healthcare staff to assist patients in adopting behaviors associated with improved health outcomes.4
To address the question of whether counseling can improve dietary patterns, we performed an extensive systematic evidence review on behalf of the U.S. Preventive Services Task Force (USPSTF).1 This larger report comprehensively updated the chapter on dietary counseling from the second edition of the Guide to Clinical Preventive Services,5 and it is available from the Agency for Healthcare Research and Quality (AHRQ) at http://www.preventiveservices.ahrq.gov.
In 1996, the USPSTF recommended counseling adults and children older than 2 years of age to limit intakes of saturated fat and cholesterol, to maintain caloric balance in diets, and to emphasize foods that are high in fiber.5 An updated recommendation,6 dealing specifically with the question of dietary counseling, accompanies this summary of the evidence and is also available at http://www.preventiveservices.ahrq.gov.
We searched the MEDLINE® database for randomized controlled trials (RCTs) published between 1966 and December 2001 that examined the effectiveness of counseling in changing dietary behavior. Search terms are provided in Appendix Table 1. We supplemented our searches by reviewing the bibliographies of included articles and querying experts in the field during an extensive peer review process.
We included only studies that had been conducted with patients similar to those found in primary care practices and that had measured dietary behavior change. We excluded studies that specifically recruited patients with previously diagnosed chronic illnesses (e.g., heart disease, diabetes, renal failure) or that required special diets (e.g., prenatal interventions); however, we did include studies that enrolled patients with known risk factors for chronic diseases (e.g., elevated cholesterol, hypertension, obesity, family history of heart disease). Studies that enrolled only overweight or obese patients for the purpose of weight management were not included; a forthcoming USPSTF report on screening for obesity will examine these articles.7
All included articles used a randomized controlled study design. Because our main outcome of interest was dietary change, we excluded studies that reported only biochemical markers (e.g., serum vitamin A level) or anthropomorphic measures (e.g., weight, proportion of body fat) with no direct measure of dietary behavior. We also excluded studies in which the diet was externally controlled (i.e., provided in a residential institution or distributed by researchers). Trials had to be of at least 3 months' duration and have a minimum retention rate of 50 percent for inclusion.
Senior investigators reviewed titles and abstracts to identify which full manuscripts to review and made the final decisions about inclusion or exclusion. Other team members then reviewed individual articles and abstracted selected information into evidence tables. When multiple articles described the same study, we used the most complete article as the main source of data and used the other articles for supplemental information. Team members discussed disagreements with reviewers and made final decisions by consensus.
We used net change in consumption, defined as change in the intervention group from baseline to followup minus the change in the control group from baseline to followup, as the main outcome. We reported unadjusted outcomes from the article when they were presented. In some cases when necessary data were not presented in the article, we were able to calculate them from other information that was presented.
To facilitate comparison of effectiveness of counseling on dietary change across studies that used a variety of different outcome measures, two investigators independently classified the magnitude of dietary change in each study as "small," "medium," or "large." The study team resolved disagreements by consensus. We developed a definition of small, medium, and large changes based on the distribution of findings from the studies and the limited information available about the relationship between dietary change and health outcomes.
For saturated fat, we defined small as an absolute net difference between intervention and control groups of 0 to 1.2 percentage points, medium as a difference of 1.3 to 3.0 percentage points, and large as a difference of greater than 3.0 percentage points. When studies reported only change in proportion of calories from total fat, we classified large as a difference of greater than 10 percentage points, medium as a difference of 5.1 to 9.9 percentage points and small as a difference of less than or equal to 5 percentage points. We classified effect sizes based on the difference in the number of servings of fruit and vegetables per day consumed by the intervention and control groups. We defined small as a difference of less than 0.3 servings per day, medium as a difference of 0.4 to 0.9 servings per day, and large as a difference of greater than or equal to 1.0 serving per day. For fiber we defined a small effect size as a net difference of less than 2.0 g per day of fiber, medium as 2.0 to 4.0 g per day, and large as greater than 4.0 g per day.
If studies did not provide data on our main outcomes of interest, we used the relative change in the outcome reported (e.g., grams of fat consumed, dietary risk scores) to guide our definition of magnitude of change. The relative change was defined as the net change divided by the baseline value in the control group. A relative change of 25 percent or greater was considered large, 10 percent to 24 percent medium, and less than 10 percent small.
We examined the effect of different intervention characteristics, including intensity, the risk status of the patient populations studied, the study setting, and the use of well-proven counseling elements, on the magnitude of change in dietary behavior achieved. We considered trials that examined multiple nutrients as separate studies for these analyses. Because of concern about double-counting studies, we repeated the analyses with each study's effect counted only once (once using the largest effect and again using the smallest effect) and found similar results. Because of heterogeneity in the outcomes, we did not attempt meta-analysis.
Two senior reviewers independently rated the intensity of the dietary intervention as "low," "medium," or "high" based on the number and length of counseling contacts. Interventions with only one contact of 30 minutes or less were considered low intensity, those with six or more contacts of 30 minutes or more each were considered high intensity, and all others were considered medium intensity.
Each study's intervention "setting" was classified as:
Finally, we examined the studies to determine whether they included as part of their intervention any of seven counseling elements (using a dietary assessment, enlisting family involvement, providing social support, using group counseling, emphasizing food interaction, encouraging goal setting, and using advice appropriate to the patient group being studied) that have been effective in previous research on dietary behavior change.8
Using the techniques established by the USPSTF Methods group, we rated the quality of each article as good or fair, based on criteria affecting internal validity.9 All studies that would be considered poor quality were excluded before the final review stage.
This evidence report was funded through a contract to the RTI-University of North Carolina Evidence-based Practice Center from AHRQ. Staff of the funding agency contributed to the study design, reviewed draft and final manuscripts, and made editing suggestions.
We identified a total of 129 abstracts for review from our literature searches. After review of the 129 abstracts, we identified 74 articles examining the effect of counseling on dietary behavior. After full article review, we excluded an additional 49 articles from our analysis because they did not meet our eligibility criteria. Reasons for exclusion are provided in Appendix Table 2.
We retained 21 studies reported in 25 articles that met our eligibility criteria.10-34 Across this body of literature, 17 studies addressed changes in consumption of dietary fat, 10 studies addressed changes in consumption of fruits and vegetables, and 7 studies addressed changes in consumption of dietary fiber, for a total of 34 intervention "arms." Eleven studies addressed changes in one dietary element and 10 addressed changes in 2 or 3 elements. Four studies included interventions for other behavioral risk factors for chronic disease, such as offering smoking cessation or encouraging increased physical activity.21,23,25,30 All included studies were considered to be of good quality, based on randomized design, high retention rates, and use of appropriate outcome measures.
Eight studies were performed in primary care settings. In seven of the eight studies, primary care providers performed the dietary counseling,10,11,20,27,28,30,34 and in the remaining study, nutrition counseling was performed through referral within the clinic.23 Five studies used self-help materials and/or interactive health communications (e.g., telephone messages, computer-generated mailings) to deliver counseling.12,14,15,22,31 Eight studies were performed in special research clinics,13,16,21,24,25,29,32,33 with counseling performed in most cases by a nutritionist or other specially trained counselor.
Nearly all the studies provided information on the dietary assessment tool used to assess outcomes and, in some cases, to guide counseling. Of the 21 studies, 12 used some version of a validated food frequency questionnaire, 2 used single- or multi-day diet recall, 2 used food diaries, and 4 used other specific instruments. One study did not report how assessment was performed.24 The full systematic evidence review,1 available online (http://www.preventiveservices.ahrq.gov), gives more information about the specific assessment instruments and their accuracy and reliability.
Table 1 describes the 17 studies that examined the effect of counseling on intake of dietary fat. Nine studies reported specifically on change in the percentage of calories from saturated fat.13-16,24-26,28,29 The remaining eight studies used other measures of fat intake, including grams of saturated or total fat consumed or study-specific outcome scales.10-12,20-23,30 Studies that measured only total fat intake focused much of their interventions on reducing saturated fat intake and hence are retained in this analysis.
Six studies focusing on the effect of counseling on reducing patients' consumption of saturated fat achieved a large effect (>3 percentage point reduction),12,13,16,24,25,29 five achieved a medium effect (1.3 to 3.0 percentage point absolute reduction),14,20,21,23,30 and six had only a small effect (less than 1.3 percentage points).10,11,15,22,27,28 For the nine studies reporting change in percentage of calories from saturated fat, net reductions ranged from 0.9 to 5.3 percentage points.
We identified 10 studies that examined the effect of counseling on fruit and vegetable intake (Table 2).12-15,21,22,28,31-33 Most of the studies (6 of 10) did not define which foods (e.g., potatoes or legumes) were considered fruits or vegetables or what constituted a serving.11,1214,15,21,33 Among these 10 studies, three demonstrated that dietary counseling produced small to no increases (<0.3 servings per day) in fruit and vegetable consumption,12,21,28 five demonstrated medium increases ranging from 0.3 to 0.8 servings per day,13,15,22,31,33 and two demonstrated large effects, increasing fruit and vegetable consumption by 1.4 and 3.2 servings per day.14,32
Seven studies examined the effect of counseling on fiber intake (Table 3).10,11,14,15,23,28,34 Five studies showed small increases in the amount of additional fiber consumed (range, 0.3 g to 1.6 g per day).10,11,15,23,28 One study reported differences in daily fiber intake between intervention and control groups of 2.7 g for men and 6.0 g for women at 1-year followup,34 and another found a net change of 3 g.14
Next, we examined the characteristics of the available trials that could possibly explain the differences in effectiveness that we found. Explanatory factors included the intensity of the intervention, the risk status of the patient, the setting for delivery of the intervention, and the use of specific counseling elements that had previously been shown to be effective in producing behavior change. The findings presented combine interventions for the intake of all nutrients (fat, fruit and vegetable, fiber) together, as there were too few studies of counseling about fruit and vegetable or fiber intake alone to make comparisons among intervention characteristics.
As depicted in Table 4, studies using higher intensity interventions produced larger effect sizes than studies using lower intensity interventions. Among nine study arms classified as high intensity, five (55 percent) produced large changes in dietary behavior, three (33 percent) produced medium changes, and one (11 percent) produced only a small change. Of the 18 medium-intensity study arms, one (6 percent) produced a large effect, 10 (55 percent) produced medium effects, and 7 (39 percent) produced small effects. Of the 7 low-intensity study arms, 1 (14 percent) produced a large effect, 1 (14 percent) produced a medium effect, and 5 (71 percent) produced small effects. Higher intensity studies enrolled either patients at risk for chronic disease or selected motivated patients at average risk who may not be representative of the usual patients in primary care practices. They also used well-trained counselors (most often dietitians or nutritionists) to provide counseling.
Twenty-one study arms were conducted using unselected patients, and 13 were conducted using patients with identified risk factors for chronic disease. After stratifying by intervention intensity, we could find no clear relationship between the risk status of the patients and the effect size achieved.
Studies conducted in special research clinics were more likely to produce larger effects than studies performed in other settings, in large part because the interventions in these clinics were of higher intensity. In addition, most involved counseling by trained personnel (usually dietitians or nutritionists) who were focused mainly on counseling about diet. Primary-care-based interventions produced small or medium effects; more intensive studies produced larger effects. Studies using interactive health communications had effects that were larger than those with direct primary care counseling but smaller than those found in research-clinic based studies.
Several components of counseling are thought to be associated with improved behavioral outcomes: using a dietary assessment, enlisting family involvement, providing social support, using group counseling, emphasizing food interaction (such as taste testing, cooking), encouraging goal setting, and using advice appropriate to the patient group being studied.1 We examined each study to determine how many of these elements were included in their interventions. Many interventions were not described in sufficient detail to determine with certainty the absence or presence of these study components. The total number of identified components ranged from 0 to 7, with a median of 2.
As shown in Table 5, studies employing a greater number of components had larger effect sizes. Of 6 study arms employing three or more components, 4 (67 percent) produced large effects and 2 (33 percent) produced medium effects; among 24 study arms employing 1 to 2 components, 4 (17 percent) produced large effects, 11 (46 percent) produced medium effects, and 9 (37 percent) produced small effects. Among 4 study arms reporting no components, all produced small effects.10,11 We did not identify a sufficient number of studies to determine whether any single component was associated with an independent effect on the magnitude of change in dietary behavior.
Researchers have used a wide range of interventions to examine the effect of behavioral counseling on dietary patterns among predominantly healthy adult patients. Among the studies we identified, low-intensity interventions in unselected primary care adult patients produced small or medium changes in self-reported dietary outcomes. Medium- to high-intensity interventions generally produced medium or large changes in dietary behavior, but these studies were generally conducted either in adult patients with known risk factors for chronic disease or performed in special research clinics with highly motivated or selected patients. These interventions also generally used highly trained providers who focused on dietary behavioral change. The specific health effects of these dietary behavior changes are not clear, but epidemiological data suggest that the moderate or large differences in dietary behavior are likely to be associated with lower rates of cardiovascular disease and possibly some forms of cancer.1
Among the factors affecting the response to dietary counseling, the intensity of the intervention was strongly associated with the magnitude of dietary change: medium- to high-intensity interventions produced larger changes than low-intensity interventions. Interventions conducted in special, study-specific research clinics were generally more effective than those performed in primary care clinics, but the effect of study setting was highly correlated with intensity. Interventions using self-help materials and interactive communications (computer-tailored mailings, telephone counseling) along with brief provider advice produced medium changes and appeared to be relatively feasible for use in primary care practices that have system support for their delivery. Interventions using greater numbers of well-proven counseling elements also were more likely to produce large or medium effect sizes than those reporting use of few or no components.
Our systematic review has several limitations. First, because we are extracting information from published studies, we are missing several pieces of important data that were not reported regularly. Second, identifying the appropriate measure of dietary change is difficult. Our main outcome measure, self-reported change in dietary behavior, relies on individual self-report, usually from validated food frequency questionnaires that have limited ability to measure small changes in dietary intake accurately and precisely. In addition, patients receiving dietary interventions may be more likely to report positive changes in dietary behavior than control patients, which could also lead to an overestimation of actual benefit. Although the use of biomarkers is often recommended as a more objective means of measurement, it is unclear whether available biochemical markers accurately reflect actual change in diet, may be influenced by medication use and smoking, and may not be any better correlated with health outcomes than patient self-report.
Because we also have little direct evidence about the effect of dietary changes on the risk for important health outcomes,1 we cannot determine with certainty whether the small changes in dietary behavior seen in the lower-intensity trials will translate into changes in the incidence of chronic disease.
The lack of standard outcome measures for each nutrient makes synthesis of the available evidence, including meta-analysis, difficult to perform and interpret. To provide some means of comparison, we rated study outcomes as small, medium, and large, but these definitions were not developed a priori and only partially reflect the limited body of data that links dietary change with specific health outcomes. We did not formally assess for publication bias; smaller trials with negative results may not have been published, which could lead to an overly optimistic impression of the effect of counseling. Finally, we did not have sufficient information to determine the relationship between the cost of dietary interventions and the effect achieved.
Future research should address promising leads already highlighted in this paper and identify novel means to deliver dietary advice in effective and efficient ways. Broadly speaking, research can be pursued along several dimensions. First, research is warranted as to whether dietary assessment leads to more effective counseling and subsequent behavior change when compared with general dietary advice not preceded by an assessment. Better assessment tools for measuring dietary change, including better validated biochemical markers and novel means of documenting dietary consumption, such as hand-held computer diaries, will be useful to address concerns about measurement bias. The interaction between clinical interventions and broader public health, environmental, legislative, and economic interventions to change dietary behavior requires further study as well.
In addition, more in-depth examinations of the effectiveness of specific components and intensities of dietary counseling are needed. Studies with longer followup periods and linkages to actual health outcomes will also be important. The paucity of studies evaluating referral to health professionals outside the primary care setting for either one-on-one or group counseling is striking. Studies of dietary interventions delivered by special research clinics are common, but they are not representative of the resources typically available to primary care providers.
Better epidemiologic studies and randomized trials assessing the clinical as well as population-level benefits of small dietary changes would help clarify the effectiveness of brief counseling interventions. Studies examining the effectiveness of interventions to change consumption of other foods, food patterns, or nutrients, including fish, the Mediterranean diet, legumes, sodium, and calcium or dairy products are warranted, as they each appear to have important relationships to health outcomes.1 Finally, cost-effectiveness studies comparing interventions through different health communication channels and at varying levels of intensity are needed to determine the most feasible approaches. This information, along with data concerning the health benefits of incremental dietary change, will help determine the relative value of dietary counseling compared with other clinical preventive interventions.
This study was developed by the RTI-UNC Evidence-based Practice Center under contract to the Agency for Healthcare Research and Quality (Contract No. 290-97-0011), Rockville, MD. We acknowledge the assistance of David Atkins, M.D., M.P.H., Chief Medical Officer of the AHRQ Center for Practice and Technology Assessment and Jean Slutsky, P.A., M.S.P.H., the Task Order Officer, for their advice and counsel throughout this project. Finally, we thank our RTI-UNC EPC colleagues Russell Harris, M.D., M.P.H., Co-Director of the RTI-UNC Clinical Prevention Center (University of North Carolina), Linda Lux, M.P.A. and Loraine Monroe of RTI for their assistance in this project and production of this article.