Research Paper

Phase II Study: Computer-Assisted Cholesterol Reduction

Albert Jerome, Ph.D.

SleepKey
New! help for insomnia
QuitKey
quitting smoking
DietMate
diet & exercise
NicotineFreeKids
links for teens
Health Authority
tobacco, cholesterol,
hypertension
TobaccoWars
news

Introduction

The purpose of this Phase II research study was to develop and evaluate a dietary and exercise intervention for high blood cholesterol implemented via a hand-held microcomputer.

Methods

Subjects. One hundred eight subjects (43 men and 65 women) were recruited for the study through newspaper ads and announcements at local worksites. Primary selection criteria were as follows: 1) LDL > 120 or total cholesterol > 200; 2) body mass index < 43; 3) no previous history of medication to control cholesterol; 4) currently exercising < 3 times per week; 5) no known cardiac, pulmonary or respiratory disease.

Ninety-three subjects (86.1%) were Caucasian, 10 (9.3%) were Black; 4 (3.7%) were Hispanic; and 1 (.9%) was Asian. Subjects ranged in age from 26 to 60 (M= 44.3 years, SD=8.8). Mean level of education was 15.9 years (SD=2.2). At pretreatment, subjects had a mean weight of 187.1 pounds (SD=33.4) and body mass index (BMI; weight in kilograms divided by height in meters squared) of 29.7 (SD=4.8). Mean cholesterol levels were 237.7 mg/dL (SD=24.7) for total cholesterol, 153.2 mg/dL (SD=20.6) for low density lipoproteins (LDL), 53.0 mg/dL (SD=16.6) for high-density lipoproteins (HDL), and 163.0 mg/dL (SD=91.0) for triglycerides. Seventy-six percent of subjects were "overweight" as defined by a BMI of 25 or more for women and 28 or more for men (a la Wood, Stefanick, Williams, & Haskell, 1991); 24% were of normal weight. The vast majority of subjects (n=92; 85.2%) had made at least one previous attempt to lose weight and 52 subjects (48%) had lost 10 or more pounds on at least one previous diet.

Measures.

Demographics, cholesterol history, weight loss history, and medical screening. During the pretreatment assessment, subjects completed a background questionnaire that included basic demographics, history of cholesterol levels and treatment for cholesterol, weight and dieting history, as well as current medical conditions and medications that might preclude participation.

Cholesterol. Full lipid panels were assessed at pretreatment and at 3-month and 6-month assessments by an independent laboratory using enzymatic procedures (ACA Analyzer, Du Pont Company).

Weight and related measures. Weight; blood pressure; waist, biceps, and hips circumference; and resting pulse were measured at each assessment point using standard procedures. Height was assessed at pretreatment only. Body mass index was computed by dividing weight (in kilograms) by height (in meters) squared.

Motivation and confidence. Motivation to reduce cholesterol, confidence in achieving cholesterol reduction goals, and confidence in maintaining cholesterol reduction were assessed on 0 - 10 self-report scales during each assessment.

Compliance questionnaire. Compliance with diet, exercise, and record keeping were rated on 1 - 7 self-report scales at 3-month and 6-month assessments.

Body-Satisfaction Questionnaire. An abbreviated form of the Body Satisfaction Questionnaire (developed by the Eating Disorders Clinic at West Virginia University) containing nine items was used. Ratings on 1 to 6 self-report scales (where 1=extremely dissatisfied and 6=extremely satisfied) were obtained for overall body satisfaction and satisfaction with eight specific body parts (shoulders, chest, arms, waist, stomach, buttocks, hips, thighs).

Satisfaction measures. Ratings of overall satisfaction, likelihood of recommending to a friend, likelihood of purchasing the program, and likelihood of using the program consistently for at least 12 months were rated on 1 -7 self-report Likert scales at 3-month and 6-month assessments. During the 3-month assessment a structured interview was also conducted to assess what was liked or disliked about each program and to solicit ideas for program modifications.

Treatment Conditions. Subjects were randomly assigned to one of two treatment conditions: 1) Standard Behavioral Self-Management or the Computerized Cholesterol Reduction Program.

Standard Behavioral Self-Management Program (SBSM). The standard behavioral self-management treatment included the Eater's Choice book (Goor & Goor, 1995), food logs, and an exercise guide. Eater's Choice is a 633 page book containing: 1) educational information on the relationship between diet and blood cholesterol; 2) instructions for following a Step One diet; 3) guidelines for meal planning; 4) instructions for keeping food logs; 5) and a variety of low-fat menus.

Subjects also received one of two exercise programs, based on degree of risk for heart disease which was assessed at pretreatment using American College of Sports Medicine (ACSM) guidelines (ACSM, 1993). Subjects not deemed to be at increased risk for heart disease received The Exercise Book (Stanford Center for Research in Disease Prevention [SCRDP], 1986), a 52-page book that encourages sedentary individuals to take the fist steps toward beginning to exercise. It gives choices of activity, instructions for starting, and hints for exercising safely and with enjoyment. Subjects deemed to be at "increased risk" received the Walking Kit (SCRDP, 1994 ). The Walking Kit provides step by step instructions for beginning and maintaining a walking program.

Computerized Cholesterol Reduction Program (CCRP). Subjects in the CCRP condition received the CCRP computer, a 15-page program guide that provided instructions for using the computer, and an NIH pamphlet containing educational material on high blood cholesterol (So you have High Cholesterol, US Department of Health and Human Services [DHHS], 1993). In addition, subjects deemed to be at increased risk received The Walking Kit to guide their exercise. Subjects not deemed to be at increased risk followed the standard exercise program implemented in the CCRP (described below).

Hardware. The computer used in this study measured approximately 6 1/2" by 3 3/4" by 1" thick (with cover closed), weighed just under 1 lb., and operated on 4 AA batteries. The computer had seven operating buttons plus an on/off button and volume control, and was encased in a hard plastic shell with a cover that folded back to become a stand. The LCD display measured 3 3/4" by 1 7/8" and had two sections. The bottom section of the LCD was dot matrix type and could display graphics or text (8 lines by 32 characters per line). The top section of the LCD had 4 icons for prompting meals, exercise, weighing, and drinking water and displays for volume setting, date, and time. The computer had an 8-bit 80C88 microprocessor with 256 kilobytes ROM and a 4-bit 75305 microprocessor with 32 kilobytes RAM. In addition to the computer, a program manual, a gourmet cookbook and an instructional video were provided.

Program Operation. The computerized treatment was highly prescriptive and interactive. Behavioral principles such as goal setting, self-monitoring, stimulus control, feedback, and shaping were integrated at all levels of the program to foster changes in eating and exercise habits and to maximize user compliance.

When subjects first began the program, the computer provided a tutorial of key functions and then prompted them to enter height, weight, age, sex, activity level, weight loss goals, usual mealtimes, and preferred exercises. On the basis of this information, a set of menus consistent with Step-One dietary guidelines was individualized to a caloric level that projected gradual, steady weight loss. The menus were nutritionally balanced, meeting the recommendations of the Dietary Guidelines for Americans (U.S. Department of Agriculture & U.S. Department of Health and Human Services, 1990), as well as those of nine other major health organizations that publish nutritional guidelines.

Subjects input their weight daily and the computer made periodic changes to the menus to adjust calories up or down as needed to maintain weight loss within a safe range (approximately .5 to 1.5 lb.. weekly). The computer used visual (icons and text) and auditory prompts to remind subjects to eat, exercise, and drink water, thereby encouraging the development of stimulus control over these behaviors. In response to cues, users picked the appropriate action by highlighting an item from a list of actions and pressing a button. The computer then provided a series of screens that allowed them to make appropriate selections and record their behaviors. Following selection of a meal, the caloric content and percent calories from fat were displayed, and subjects were given the option of recording their meal or first changing item quantities or modifying the menu by using a substitution list to replace specific items with nutritionally equivalent foods.

Following a one-week baseline period, exercise targets were provided for a wide range of exercises selected by the user. Subjects were prompted by an icon and a text message to exercise three times per week and the amount of exercise per session was very gradually increased up to a moderate expenditure of 300 - 400 calories.

The program included an extensive feedback system that provided messages at several different times: immediately following the entry of certain behaviors, as a summary at the end of the day, as a reminder or "to do" list in the morning, and as a summary at the end of each week. The feedback system tracked important variables such as total caloric intake, percent calories from fat, number of exercise sessions, and total calories expended during exercise. From this historical data, variables were classified according to their discreet patterns. For example, the first missed exercise session following compliance with exercise would be classified as a "slip." Similarly, a variety of other patterns were defined on the basis of the frequency and timing of the behaviors recorded. Each classification of behavior for a given variable was associated with a set of feedback messages that varied in frequency, phrasing, and tone according to the anticipated needs of the user..

An additional source of daily feedback was provided by a summary screen--accessible at any time--that documented progress toward meeting daily goals for caloric intake, percent calories from fat, and exercise. Longer-term feedback was provided by a weekly summary and by charts and graphs that showed history of weight loss, calories expended through exercise, and caloric intake.

General Procedures

Recruitment and screening. Subjects were recruited via newspaper ads and notices placed at local worksites. Initial screening was conducted by telephone and interested subjects who appeared to meet the screening criteria were scheduled for an introductory meeting at our lab.

During the meeting: 1) volunteers were shown a brief educational video discussing the impact of blood cholesterol on risk for heart disease and the basics of dietary treatment for high cholesterol; 2) the design of the study was described and subject participation detailed; and 3) physician consent forms were handed out. Subjects were told to get the form signed, return them to us, and have their physician authorize blood tests.

Physician consent to participate. All subjects were required to have written consent from a physician to participate in the study. Volunteers were provided with consent forms, instructions, and a project description to give to their physician. Physicians were asked to sign the form and contact Access of Reston to authorize three cholesterol blood tests. Forms were returned to PICS via mail or hand-delivered by participants.

Blood tests. Blood tests were conducted at pretreatment, 3-month and 6-month follow-up. Blood was drawn by Access of Reston (Reston, Virginia) and full lipid panels tests conducted by Friends Medical Laboratory (Catonsville, Maryland). Results were faxed to PICS. Copies of results were provided to subjects or their physicians upon verbal request.

Assignment to treatment conditions. Volunteers were randomly assigned to one of the two treatment conditions and scheduled for a pretreatment assessment if they met selection criteria and had obtained their physician's signature on the consent form.

Pretreatment assessment. During the pretreatment assessment: 1) informed consent was obtained; 2) height, weight, circumference measures, and body fat were measured; 3) questionnaires assessing demographics, cholesterol history, weight loss history, medical history, motivation, and confidence were administered; 4) treatment materials were distributed; and 5) brief instructions were provided for getting started with treatment.

Follow-up assessments. Follow-up assessments were conducted 3- and 6-months after the pretreatment assessment. For each follow-up, subjects were contacted, asked to return to Access of Reston to get their blood drawn for a cholesterol test, and scheduled for individual follow-up visits at our lab. Assessments were conducted by research associates following a standard protocol. During follow-up assessments: 1) weight and circumference measures were recorded; 2) questionnaires assessing compliance with treatments, motivation, and confidence were administered; and 3) a structured interview exploring likes, dislikes, and suggestions for improvement was administered.

Results

Treatment groups. Fifty-three subjects were assigned to the Eater's Choice condition (EC) and 55 were assigned to the DietMate condition (DM). Treatment groups did not differ significantly on mean pretreatment weight, LDL, HDL, or triglycerides. The proportion of overweight subjects did not differ by group.

Follow-up period. The first follow-up evaluation was conducted a mean of 10.7 weeks (SD=5.8) following the pretreatment evaluation and the second assessment occurred a mean of 26.9 weeks (SD =5.9) following the pretreatment evaluation. The length of time between the initial visit and the second visit did not differ between the groups; however; the length of time between the first and third visits was significantly longer for subjects in the DM group (M=27.1, SD=3.2 vs. M=25.6, SD=1.8; t(68.4)=2.8, p<.01).

Attrition rate. Eighty-nine subjects (82%) completed all follow-up evaluations and blood tests (45 in the EC group and 44 in the DM group). Subjects who completed all evaluations and blood tests had more years of education than subjects who missed one or more appointments (16.1 years vs. 14.9 years, t(106)=2.15, p<.05), but did not differ significantly on age, sex, cholesterol levels, triglycerides, weight, or BMI at pretreatment.

For analyses of treatment outcome, the 89 subjects who completed all of the assessments are included. LDL readings could not be obtained for three subjects at various visits, therefore analyses involving LDL levels are based on fewer than 89 subjects as noted below in the pertinent sections.

DietMate program failures. Ten of 44 subjects (23%) in the DietMate condition who completed treatment reported hardware or software problems that were sufficiently severe enough to interfere with their treatment progress (e.g., complete program "crash," loss of historical data, display going blank or becoming unreadable). Computer problems were rectified as soon as they were reported by fixing or replacing the defective unit. In the analyses that follow, we took the most conservative approach and included subjects with computer problems with all other subjects.

Outcome measures. Changes in cholesterol, weight and other physical measures were analyzed using separate mixed model Analyses of Variance (ANOVAs) for 3-month and 6-month follow-ups. For these analyses, the between subjects factor was treatment group (DM vs. EC) and the time factor was assessment point (pretreatment vs. 3-month follow-up or pretreatment vs. 6-month follow-up). Means for cholesterol-related variables are listed in Table 1; means for other physical data are presented in Table 3.

Cholesterol. At the 3-month follow-up, there were significant effects of time for total cholesterol, LDL, and HDL ,(F(1, 84) = 55.7, p<.001, F(1, 84) = 245.1, p<.001, and F(1, 84) = 19.5,p <.05), indicating that both groups significantly reduced their total and LDL cholesterol and significantly increased their HDL cholesterol. Similarly, at the 6-month follow-up, there were significant effects of time for total cholesterol and LDL (F(1, 85) = 6.9, p=.01 and F(1, 85) = 5.8, p<.05), but changes in HDL were not significant. There were no significant group or group by treatment interactions at either follow

Table 1.

Means and Standard Deviations for Cholesterol and Related Variables at Pretreatment and Across Visits.

Note: Except for LDL, sample sizes are 44 for DM and 45 for EC. Sample sizes for LDL: pretreatment DM=43, EC=44; 3-month follow-up DM=43, EC=43; at 6-month follow-up DM=43, EC=44 .

The magnitude and distribution of changes in LDL cholesterol can be seen more clearly by examining Table 2 below.

Table 2.

Frequency Distribution of Percentage Decrease in LDL Cholesterol by Group Across Visits

Note: Numbers refer to number of subjects; numbers in parentheses are group percentages.

n=43 for DM at 3-months and 6-months; n=43 for EC at 3-months and n=44 at 6-months.

Weight loss and changes in Body Mass Index. Significant effects of time were seen for weight and body mass index at both the 3-month (F(1, 84) = 76.8, p<.001 and (F(1, 84) = 76.3, p<.001) and 6-month assessments (F(1, 85) = 6.9, p=.01 and (F(1, 87) = 35.2, p<.001). Group and group by time effects were not significant. These results indicate that subjects in both groups evidenced significant reductions in weight and BMI, but the magnitude of changes did not differ between groups.

Means and standard deviations are presented in Table 3.

Table 3.

Means and Standard Deviations for Weight and Body Mass Index Across Visits.

Note: For all measurements n= 44 for DM group and n=45 for EC group.

Changes in weight and BMI were analyzed further by comparing subjects of normal weight with overweight subjects (based on pretreatment BMI). An independent samples t-test revealed that mean weight reduction from the initial visit to the second visit was significantly greater for overweight subjects than for normal weight subjects (M=6.9 lb., SD=6.8 vs. M=3.9 lb., SD=4.2; t(55.3) = 2.47, p<.05); however, this difference was no longer significant at the third visit. Repeating the mixed model ANOVAs on weight and BMI using only overweight subjects revealed the same pattern of results as the original analyses: A significant effect of time, but no significant effect of group or group by time interaction. The changes in weight among overweight subjects are shown in Table 4 below.

Table 4.

Frequency Distribution of Weight Loss by Group Across Visits for Overweight Subjects

Note: Numbers refer to number of subjects; group percentages are in parentheses.

Sample sizes are 33 for DM and 35 for EC at 3- and 6-months, respectively.

Changes in other physical measures. Subjects in both groups improved on all physical measures between their initial visit and the 3-month follow-up (i.e., there was a significant effect for time for all variables in the analyses; see Table 5 for results of tests of significance, see Table 6 for means and standard deviations). For both groups, these improvements remained significant at the 6-month follow-up, except for changes in systolic and diastolic blood pressure which were no longer significant (see Table 5). Regarding group differences, diastolic blood pressure dropped to a greater degree for subjects in the EC condition at the second visit than for subjects in the DM condition (see Tables 5 and 6), but this difference was not significantly different at 6-month follow-up. No other main effects for group or interaction effects for group by time were significant.

Table 5.

F Statistics and Associated Probabilities (p) for Type I Error for Physical Measurements Across Visits

Note: Degrees of freedom are 1, 85 for comparisons at 3-month and 1, 86 for comparisons at 6-months.

Table 6.

Means and Standard Deviations for Physical Data Across Visits.

Body satisfaction. Ratings of overall satisfaction with body appearance and average satisfaction with eight body parts (shoulders, chest, arms, waist, stomach, buttocks, hips and thighs) were analyzed with mixed model ANOVAs. Results revealed a significant effect of time for overall satisfaction and satisfaction with body parts at both follow-up periods F(1, 84) = 21.91, p<.001 and F(1, 87) = 18.64, p<.001 at 3-month follow-up; and F(1, 84) = 27.8, p<.001) and F(1, 87) = 30.5, p<.001 at 6-month follow-up. Consistent with findings for other outcome variables, group and group by treatment effects were not significant, thereby indicating that the magnitude of changes did not differ as a function of treatment group.

Compliance. Compliance with monitoring food intake was assessed by reviewing retrospective self-reports of the number of meals recorded, on average, each week during treatment. For each week, subjects were rated as being "compliant" if they reported recorded, on average, at least one meal per day in their computer or diary. From these data a "percent compliance" score was calculated by dividing the number of compliant weeks by the number of treatment weeks for each subject. Results showed that subjects in the DM condition were compliant for a significantly greater percentage of total treatment weeks (68%, SD = 34) than subjects in the EC condition (48%, SD = 35), t(49)=2.5, p<.05). A frequency distribution reflecting the extent of compliance among subjects across groups is presented in Table 7. Consistent with the comparison of means, a chi-square analysis on categories in this table revealed a significant advantage for the DM condition, c2(3, N=89) = 8.67, p<.05).

Table 7 .

Frequency Distribution of Extent of Compliance Across First 30 Weeks.

Note: Numbers refer to number of subjects; numbers in parentheses are

group percentages. Sample sizes are 44 for DM group and 45 for EC group.

Bivariate Pearson Correlation Coefficients were calculated between each of the outcome measures discussed above and the percentage of weeks subjects were compliant (see Table 8 below). In general, there was a significant association between compliance and measures of body size and body satisfaction. No significant relationships between compliance and other outcome measures emerged.

In order to further explore the impact of compliance on treatment outcome, we identified a subgroup of the 89 subjects that were compliant (recorded, on average, at least one meal per day in their computer or diary) for at least four weeks during the initial treatment period. Using this definition, 74 of 89 subjects were identified as being compliant. The proportion of subjects classified as compliant did not differ by group.

Analyses of outcome variables were repeated using this subgroup of subjects. Results showed a significant treatment by time interaction for LDL cholesterol at the 3-month follow-up (M=144.4 mg/dL, SD=19.8 for DM subjects; M=149.0, SD=18.3 for EC subjects; (F(1, 70)= 5.4, p<.05), indicating that DM subjects had significantly greater decreases in LDL cholesterol during the first 3-months of the intervention. A significant time effect also emerged for reduction in triglycerides for all subjects at 3-month follow-up (M=166.7, SD=95.7 at pretreatment, M=156.3, SD=88.7 at 3-months), F(1, 70)= 4.6, p<.05. The significance of the results changed little for outcome measures at the six-month follow-up, except that the reduction in LDL for all subjects was no longer significant at p<.05 (p=.056). The pattern of significance for the main and interaction effects for all other comparisons was the same as the results obtained by using all subjects (see above).

Table 8 .

Pearson Correlation Coefficients for Variables Related to Percent Compliance.

Note: Negative correlations for body satisfaction variables reflect improved satisfaction with greater compliance; for all other variables, positive coefficients indicate improvement in health risk indicators with increased compliance. *p<.05, **p<.01, ***p<.001

Exercise. Changes in exercise from pretreatment to each follow-up assessment were analyzed by comparing the number of exercise sessions per week and the average duration of each session reported during the four-week period preceding each visit. Mixed model ANOVAs revealed significant main effects of time for the number and duration of exercise sessions at both the 3-month (F(1, 76)= 33.4, p<.001 and (F(1, 76)= 10.1, p<.01) and 6-month assessments (F(1, 80)= 14.5, p<.001 and F(1, 80)= 4.5, p<.05). Collapsed across groups, mean number of sessions per week were 1.8 (SD=2.3), 3.7 (SD=2.3), and 3.1, and overall mean duration of sessions (in minutes) were 28.1 (SD=28.3), 38.0 (SD=24.5), and 34.2 (SD=27). Tests for the effects of group and the time by group interaction were not significant. Thus, subjects in both groups made positive changes in exercise during the first three months of treatment and maintained these changes during the follow-up period.

Ratings of program satisfaction. Four measures of program satisfaction were assessed at each follow-up: 1) overall satisfaction with the program; 2) likelihood of purchasing the program; 3) likelihood of using the program consistently for at least 12 months; and 4) likelihood of recommending the program to a friend. All variables were measured on a 1-10 scale (where 1 is "extremely dissatisfied" and 10 is "extremely satisfied"). As can be seen by examining Table 9, all of the averaged ratings were relatively high, indicating general satisfaction with both of the programs. Group comparisons using t-tests were made on these four satisfaction variables at each of the two visits to yield eight between-group comparisons. At the 3-month follow-up, subjects in the DM condition reported that they would be more likely to use the program for 12 months than subjects in the EC condition (M=7.6 for DM; M=6.3 for EC; t(95)=2.31<.05). No other analysis was statistically significant.

Table 9 .

Mean Satisfaction Ratings for All Subjects

Note: All scales are measured from 1 to 10 where 1 is unfavorable and 10 is favorable.

Summary and Conclusions

Primary results of the study were that subjects in both treatment groups evidenced statistically significant reductions in total cholesterol, LDL, weight, BMI, and various circumference measures at 3-month and 6-month follow-ups. Reductions in blood pressure were apparent at 3-months but not at 6-months. Among the entire sample of subjects who completed both follow-ups, there were no significant differences between groups on the magnitude of change on any of these variables. However, when subjects who were not at least minimally compliant with their treatment were removed from the analyses, subjects in the DM condition had significantly greater decreases in LDL than subjects in the EC condition at the 3-month visit.

Analyses of compliance (defined as number of weeks that subjects continued to record food intake) indicated that subjects in the DM condition continued recording food intake longer than subjects in the EC condition. Among all subjects who completed both follow-ups, length of compliance was signficantly correlated with changes in weight, BMI, and various body circumference measures.

Analyses of changes in exercise were consistent with findings for other outcome variables: At both the 3-month and 6-month assessments mean number of exercise sessions per week and mean duration of exercise sessions were greater than pretreatment levels for subjects in both treatment conditions.

A final notable finding was generally high levels of satisfaction, reported at both follow-up points, which did not differ between groups.

Overall, these findings support the effectiveness of both computerized and self-help methods for weight loss and cholesterol reduction. Some support for the hypothesized advantage of the CCRP over paper and pencil methods was also demonstrated. First, subjects in the DM condition continued monitoring food intake longer than subjects in the EC condition, and among subjects who monitored food intake for at least four weeks, DM subjects evidenced significantly greater reductions in LDL at the 3-month follow-up.

And second, although not statistically significant, all but one of the observed differences on primary outcome variables were in the anticipated direction (favoring the DM group).

These results suggest that one of the primary advantages of the CCRP is that it facilitates self-monitoring. Compliance with self-monitoring was shown in this study to be significantly correlated with changes in weight and body circumference, and has been shown in other studies to be a significant predictor of successful weight loss.

References

American College of Sports Medicine. (1993). Resource manual for guidelines for exercise testing and prescription. (2nd ed.). Philadelphia, PA: Lea & Febiger.

Goor, R., & Goor, N. (1995). Eater's choice: A food lover's guide to lower cholesterol. (4th ed.). Boston: Houghton Mifflin Company.

Stanford Center for Research in Disease Prevention. (1986). The exercise book (2nd ed.). Stanford, CA: Author.

Stanford Center for Research in Disease Prevention. (1994). The walking kit (3rd ed.) Stanford, CA: Author.

United States Department of Agriculture & United States Department of Health and Human Services. (1990). Dietary Guidelines for Americans (3rd ed.; Home and Garden Bulletin No. 232). Washington, D.C.: U.S. Government Printing Office.

United States Department of Health and Human Services. (1993). So you have high blood cholesterol (NIH publication no. 93-2922). Washington, D.C.: U.S. Government Printing Office.

Wood, P. D., Stefanick, M. L., Williams, P. T., & Haskell, W. L. (1991). The effects on plasma lipoproteins of a prudent weight-reducing diet, with or without exercise in overweight men and women. The New England Journal of Medicine, 319 (18), 1173-1179.