Diabetes Disease Stage Predicts Weight Loss ... - Wiley Online Library

Diabetes Disease Stage Predicts Weight Loss Outcomes with Long-Term Appetite Suppressants Mehmood A. Khan,* John V. St. Peter,* Gail A. Breen,† Guilford G. Hartley,‡ and John T. Vessey§

Abstract KHAN, MEHMOOD A., JOHN V. ST. PETER, GAIL A. BREEN, GUILFORD G. HARTLEY, AND JOHN T. VESSEY. Diabetes disease stage predicts weight loss outcomes with long-term appetite suppressants. Obes Res. 2000:8:43– 48. Objectives: Characterize degree of weight loss with stage of diabetes and describe its effect on cardiovascular disease risk factors in obese patients with and without diabetes. Research Methods and Procedures: Retrospective cohort analysis from patients participating in a long-term weight management protocol using diet, exercise, behavioral modification, and appetite-suppressant therapy. Patient groups, with (n ⫽ 19) and without diabetes (n ⫽ 19) were matched for age, gender, and weight before weight loss therapy. The effect of 12 months of therapy on weight, blood pressure, glycemic control, lipid profile, and medication requirements were tested. Additionally, patients were grouped or staged based upon therapy required for control of diabetes at the beginning of weight loss intervention. Analysis of covariance described relationships between diabetes disease stage and weight loss at 12 months. Results: Nondiabetic patients had greater mean reduction in BMI than the diabetic group (7.98 kg/m2 vs. 4.77 kg/m2, p ⬍ 0.01). A significant linear trend (p ⬍ 0.001) for decreasing weight loss with stage of diabetes was observed. Blood pressure, lipid profile, and glycemia improved significantly. The average daily glyburide-equivalent dose decreased from 9.4 to 3.0 mg (p ⬍ 0.01). Discussion: Patients with diabetes lost less weight than

Submitted for publication February 24, 1999. Accepted for publication in final form July 18, 1999. From the *Hennepin County Medical Center Divisions of Endocrinology and ‡Internal Medicine and §University of Minnesota Division of Epidemiology, Minneapolis, MN 55415 and the †Philadelphia College of Pharmacy, Philadelphia, PA 19104-4495. Address correspondence to Mehmood A. Khan, MD, Hennepin County Medical Center, Division of Endocrinology, #865B, 701 Park Avenue South, Minneapolis, MN 55415. E-mail: [email protected] Copyright © 2000 NAASO.

similarly obese patients without diabetes. Regardless of differential weight loss between groups, cardiovascular disease risk factors improved. Hypoglycemic medication requirements decreased with weight loss therapy. A predictive relationship may exist between diabetes disease stage before weight loss therapy and future weight loss potential. Key words: diabetes, obesity, weight loss, appetite depressants, outcome

Introduction An expert panel from the National Heart, Lung and Blood Institute of the National Institutes of Health has recently released evidenced-based treatment recommendations for obesity because of its increasingly significant impact on public health in the United States (1). The prevalence, in the United States, of overweight (body mass index [BMI] 25 to 29.9 kg/m2) and obese (BMI ⱖ 30 kg/m2) individuals is increasing across age, gender, and ethnicity (2– 4). From 1980 to 1994, it is estimated that the percentage of those considered obese increased from 14.5% to 22.5%. Obesity has a substantial impact on mortality (4). There is an increased prevalence and risk of developing comorbid conditions associated with obesity, including diabetes mellitus, hyperlipidemia, hypertension, and coronary artery disease (4). Developing rational approaches for the management of excess weight is important given current population trends and societal costs related to obesity (5). Clinically, the presentation of the insulin-resistant syndrome seen in patients with Type 2 diabetes may include increased central adiposity, increased blood pressure, increased fasting triglyceride and very low density lipoprotein concentrations, increased small dense low density lipoprotein particles, and decreased high density lipoprotein. Improvements in blood pressure and lipid profile can occur in concert with weight loss. Understanding the response of individuals to weight loss methods and the metabolic effects of weight reduction in those with diabetes may provide OBESITY RESEARCH Vol. 8 No. 1 Jan. 2000

43

Diabetes Diseases Stage Predicts Weight Loss, Khan et al.

insight into appropriate interventions that lead to sustained weight loss. Additionally, identification of metabolic markers that predict degree of weight loss could be clinically useful in targeting individuals for therapy and establishing achievable goals. Weight reduction in obese individuals with and without diabetes, through diet and behavior modification, is generally correlated with improved insulin sensitivity, glucose tolerance, blood pressure control, and lipid profile (6 – 8). The majority of studies, addressing the issue of weight loss in patients with diabetes and obesity have demonstrated that relatively small weight loss improves comorbid conditions (9 –11). Others have suggested a moderate weight loss is necessary (12). In either case, these studies were of relatively short duration (8,13). Comparable weight reductions in obese populations with and without diabetes have been noted when employing diet modification or diet in conjunction with behavioral and exercise modification (14). However, a preponderance of previous studies, employing nonpharmacological weight management strategies, note that patients with type 2 diabetes lose less weight than patients without diabetes (13,15,16). Reasons for these discrepant results may be related to differences in weight-reduction programs or to physiological differences between those with and without diabetes (17–19). An additional reason may be related to the progressive nature of type 2 diabetes. Clinically, as diabetes progresses, therapy must be altered to match these advancing stages. Thus, therapy may be a valid marker of the pathophysiological stage of diabetes. We investigated the magnitude of weight lost by individuals with and without type 2 diabetes as well as relative to type of diabetes therapy, a surrogate marker of disease stage (20).

Research Methods and Procedures Data for this analysis was derived from a 4-year, openlabel trial of d,l-fenfluramine/phentermine (d,l-fenfluramine 20 mg by mouth TID and phentermine 30 mg by mouth daily) in the treatment of obesity (21). In addition to pharmacotherapy, attendance at monthly group support sessions was required. Dietary intervention, exercise, and behavior modification were prescribed. Nutrition intervention in groups of up to 20 subjects were led by a registered dietitian and met weekly for 6 weeks during which a formal didactic program of nutrition education was presented. This was followed by monthly meetings for reinforcement of the didactic content, until the end of the 10th month. Meetings were conducted every 2 months thereafter for the duration of participation. Each subject was prescribed with an individualized calorie restriction diet aimed to supply 25 kcal/kg ideal body weight per day for women and 30 kcal/kg per day for men. Individual follow-up appointments were available for participants with the registered dietitian to discuss topics they were uncom44

OBESITY RESEARCH Vol. 8 No. 1 Jan. 2000

fortable discussing within groups. A multivitamin and mineral supplement was recommended. An exercise program was recommended to each participant aimed at increasing by 300 kcal the amount of energy expended per day on each of 3 days each week. All participants (n ⫽ 199) were at least 30% over estimated ideal body weight and were required to maintain regular medical follow-up visits with the prescribing physician for the duration of their involvement in the weight loss program (21). For purposes of this analysis, each obese diabetic study participant was retrospectively matched to the first identified, nondiabetic participant, which was of the same gender, age (⫾10%), and BMI (⫾10%) immediately before weight loss therapy. Outcome measures at baseline and after completing 1 year of therapy were compiled. These included; weight (kg), BMI (kg/m2), systolic blood pressure (SBP, mmHg), diastolic blood pressure (DBP, mmHg), fasting plasma glucose (glucose, mg/dL), hemoglobin A1C (HbA1C, %), and hypoglycemic agent dose. Additionally, fasting plasma concentrations of total cholesterol, low density lipoprotein cholesterol (LDL, mg/dL), high density lipoprotein cholesterol (HDL, mg/dL), and triglyceride (Trig, mg/ dL) were collected. Comparisons of baseline and 12-month measurements (Table 1) were made using paired Student’s t test when appropriate. Oral hypoglycemic dose requirements before and after weight loss were compared using the Wilcoxon signed-rank test. The diabetic group was ultimately categorized into two subgroups based upon type of therapy, diet, or drug. Analysis of covariance was employed to describe changes in BMI among the three cohorts (no diabetes, diabetic diet-controlled, and diabetic drug-controlled) while statistically controlling for any differences in baseline BMI. Post hoc pairwise comparisons of treatment group changes in BMI based upon model marginal mean estimates and a polynomial trend analysis were completed. Statistical significance was assumed at p ⬍ 0.05.

Results A total of 38 patients (19 with type 2 diabetes and 19 without diabetes) were identified for inclusion in this analysis. Baseline and 12-month follow-up characteristics appear in Table 1. No significant differences between the two groups are apparent at baseline except fasting plasma glucose and HbA1C. At 12 months, when compared to their respective pretreatment values, both groups experienced significant changes (see Table 1). An analysis of covariance was conducted to compare the change in BMI between the two groups (diabetes vs. no diabetes). For this analysis the difference between initial and final BMI was the dependent variable, and the group (diabetes vs. no diabetes) was the independent variable, with initial BMI as the covariate. The distribution of the


Table 1. Demography and outcomes in obese patients without and with diabetes before and after weight loss No diabetes

Age (years) Weight (kg) BMI (kg/m2) Glucose (mM)¶ HbA1C (%) SBP (mmHg) DBP (mmHg) Cholesterol (mM)** LDL (mM)** HDL (mM)** Triglyceride (mM)†† Glyburide eq (mg/day)

Diabetes

Baseline

12 month

Baseline

12 month

48.1 ⫾ 8.8 112.0 ⫾ 16.5 41.3 ⫾ 4.5 5.60 ⫾ 0.77 4.8 ⫾ 0.6 141.4 ⫾ 4.8 84.0 ⫾ 8.7 5.40 ⫾ 1.07 3.58 ⫾ 0.75 1.05 ⫾ 0.20 1.64 ⫾ 1.07

88.8 ⫾ 15.7* 32.7 ⫾ 4.3* 4.98 ⫾ 0.47* 5.1 ⫾ 0.4 132.1 ⫾ 12.3* 79.2 ⫾ 6.2† 4.60 ⫾ 1.06* 2.84 ⫾ 0.74* 1.18 ⫾ 0.26* 1.28 ⫾ 0.72†

47.9 ⫾ 9.4 120.7 ⫾ 18.5 42.5 ⫾ 4.8 11.29 ⫾ 3.56 8.4 ⫾ 0.5 133.6 ⫾ 10.3 83.2 ⫾ 9.0 5.67 ⫾ 1.07 3.86 ⫾ 0.90 0.94 ⫾ 0.19 1.88 ⫾ 0.53 9.4 ⫾ 7.6

103.8 ⫾ 17.3‡ 36.6 ⫾ 4.7‡ 7.49 ⫾ 4.06‡ 6.4 ⫾ 1.6‡ 135.8 ⫾ 14.8 81.5 ⫾ 7.0 5.08 ⫾ 0.71§ 3.28 ⫾ 0.62‡ 1.17 ⫾ 0.23‡ 1.39 ⫾ 0.53‡ 3.0 ⫾ 3.3‡

Data are mean ⫾ SD. * p ⬍ 0.01 vs. non-diabetic baseline. † p ⬍ 0.04 vs. non-diabetic baseline. ‡ p ⬍ 0.01 vs. diabetic baseline. § p ⬍ 0.03 vs. diabetic baseline. Conversions: ¶ glucose ⫻ 18.01 ⫽ mg/dL; ** cholesterol ⫻ 38.67 ⫽ mg/dL; †† triglyceride ⫻ 88.57 ⫽ mg/dL.

dependent variable was positively skewed, therefore, the natural log of this variable was used for the actual analysis. The results of this analysis demonstrate that the group without diabetes had a larger mean reduction in BMI than the group with diabetes (7.98 kg/m2 vs. 4.77 kg/m2, p ⬍ 0.01). The group with diabetes was further categorized by diabetes therapy before beginning obesity treatment (i.e., dietor drug-controlled) yielding three groups; no diabetes (14 female, 5 male), diabetes receiving diet therapy (4 female, 3 male), and diabetes receiving hypoglycemic drug therapy (10 female, 2 male). Fasting plasma glucose, HbA1C, and years of diagnosed diabetes at the start of weight loss intervention for these three groups are listed in Table 2. Using analysis of covariance and controlling for initial BMI and sex, pairwise comparisons demonstrated significantly different change in BMI between the group without diabetes and the group with diabetes and receiving drug therapy (p ⬍ 0.002). Although the change in BMI in the diet-treated group was statistically indistinguishable from the other two, a significant linear trend (p ⬍ 0.004) for decreasing change in BMI was identified (Figure 1).This significant decreasing trend in weight lost is also inversely correlated to the baseline HbA1C (r ⫽ 0.50, p ⬍ 0.05). However, when the diabetes therapy group and HbA1C are analyzed together as covariates, only the type of diabetes therapy remains significant (r ⫽ 0.57, p ⬍ 0.05).

Twelve of the 19 diabetes patients were receiving medication to control hyperglycemia, 10 were receiving glyburide or glipizide. To compare oral hypoglycemic requirements, a glyburide equivalent dose was estimated (glyburide:glipizide ratio of 1⬊2) (22,23). Nine of 10 patients had a reduction in their dose of medication from baseline. The average glyburide-equivalent dose decreased over 12 months from 9.4 to 3.0 mg (p ⬍ 0.01). Figure 2 demonstrates individual patient oral hypoglycemic equivalent dose requirements over time.

Discussion This analysis compared weight lost by obese patients with and without type 2 diabetes, enrolled in a weight loss program, which incorporated appetite suppressant use in addition to structured diet, exercise prescription, and behavioral modification group therapy. As noted by others, with sustained weight loss there are improvements in cardiovascular risk factors in patients with or without diabetes (6 – 8). The improvement in glucose tolerance secondary to weight reduction during our program (pharmacological plus diet, exercise, and behavioral modification) has clinical implications in patients with diabetes. The trend for decreasing hypoglycemic medication requirements should be anticipated for diabetes patients that are enrolled in weight reduction programs, especially those employing pharmacoOBESITY RESEARCH Vol. 8 No. 1 Jan. 2000

45


Table 2. Initial average glucose control in non-diabetic and diet- or hypoglycemic agent-controlled diabetic groups

N Fasting glucose (mM)* HbA1C (%)* Years of diabetes†

No diabetes

Diet-controlled

Drug-controlled

19 5.60 ⫾ 0.77 4.8 ⫾ 0.6

7 8.31 ⫾ 2.19 7.3 ⫾ 0.9 0.34 ⫾ 0.56

12 13.05 ⫾ 3.01 9.1 ⫾ 2.2 3.98 ⫾ 3.09‡

Data are mean ⫾ SD. * Fasting glucose and HbA1C values for all groups are different from each other (p ⱕ 0.02). Conversion; glucose ⫻ 18.01 ⫽ mg/dL. † Years of diabetes estimated from date of earliest notation of diagnosis to enrollment in weight management program. †‡ Drug-controlled significantly different from diet-controlled groups (p ⱕ 0.01).

logical therapies. Clinicians must be aware of this finding and closely monitor glycemic control and symptoms of hypoglycemia. Weight reduction in patients with diabetes clearly has beneficial effects. Nonetheless, the relative benefit of any weight loss intervention must be balanced with potential associated risks. In this case, long-term use of these medications was associated with an increased risk of cardiac valvulopathy (24,25). In spite of statistically correcting for any differences in starting weight, the diabetic group in our study did not lose as much weight as a matched comparison group without

diabetes. These findings differ from those of a 1-year study conducted by Guare and colleagues who found comparable weight loss between groups with and without diabetes (14). However, disparate weight losses are noted by other investigators who also employed no pharmacotherapeutic intervention (13,15–17). The etiology of this finding is still unclear. Henry and colleagues described greater weight loss by those without diabetes, but the data were confounded by heavier starting weight in that group. Greater absolute weight loss has been noted with greater starting weight (26). Beginning with similar starting weights and using an eating

Figure 1. Relationship between change in model-adjusted BMI (DeltaBMI; kg/m2), after 1 year of weight loss therapy and hemoglobin A1C (HbA1C) before weight loss intervention. Groupings for analysis are based upon therapy for diabetes before beginning weight loss therapy; obese diabetics controlled with diet therapy alone (DM Diet; n ⫽ 7), obese diabetics treated with drug therapy (DM Rx; n ⫽ 12), and age-, gender-matched, similarly obese non-diabetics (NonDM; n ⫽ 19). All HbA1C are significantly different from each other (p ⬍ 0.02). There is a significant linear trend in DeltaBMI across the three groups (p ⬍ 0.001). Only DeltaBMI for NonDM and DM Rx are significantly different (p ⬍ 0.01). Data are mean ⫾ SE.

46



Figure 2. Change in daily glyburide-equivalent dose after 1 year of participation in a weight management program. The average equivalent-dose significantly decreased from 9.4 to 3.0 mg daily (p ⬍ 0.01).

behavior scale, Wing et al. attribute observed differences in weight lost to the ability of those without diabetes to significantly reduce nutrient intake as compared to a matched, diabetic group (16). In a separate report on weight loss in a large number of obese patients with type 2 diabetes, Wing and colleagues note that patients treated with diet only tended to lose more weight than those treated with medication and that sex and initial percentage overweight were the only significant predictors of weight lost (27). In our analysis of covariance using three groups (no diabetes, diet-, or drug-treated diabetes), type of treatment or “stage” remained a significant predictor of change in BMI after adjusting for starting BMI and sex. The mechanisms of differential weight loss between those without diabetes and those at various stages of diabetes are unknown. Possible reasons include altered regulation of energy balance in diabetes, such as dissimilar appetite regulation or glycosuria associated with increasing hyperglycemia (16,28,29). Certainly hypoglycemia is known to stimulate appetite and may result in attenuation of weight lost in patients receiving insulin or insulin secretagogues. However, our data and those of others also demonstrate a trend for less weight lost in patients with diabetes treated by diet alone (27). Clearly in this group, hypoglycemia is not a plausible mechanism. Lastly, antidiabetic therapy could have a direct effect that resists weight loss. Others have “staged” with respect to diabetes therapy but not noted significant relationships between weight loss outcomes and pretreatment glycemic control or type of diabetes therapy (8,16,27,30). However, these investigators did not describe their findings in diabetes-affected cohorts with

respect to non-diabetic cohorts. By analyzing information from age-, gender-, and weight-matched study groups with and without diabetes, we note a significant trend between pretreatment disease stage, according to type of diabetes therapy, and weight loss outcome after 12 months of weight loss therapy. If verified, this relationship may be useful for explaining variability in weight loss between study populations and may provide a tool for clinicians in determining reasonable weight loss goals for patients with diabetes. Clinically, diabetes is known to progress through multiple metabolic stages with increasing hyperglycemia. We describe an inverse relationship between diabetes disease stage, based upon type of diabetes treatment, and subsequent degree of response to therapy for obesity. The mechanisms responsible for these observations warrant further study.

Acknowledgments The effort of our research assistant Annie Singh in the compilation of the data is recognized. References 1. Anonymous. Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults—The evidence report. National Institutes of Health [published erratum appears in Obes Res. 1998;6:464]. Obes Res. 1998;(Suppl 2):51S–209S. 2. Taubes G. As obesity rates rise, experts struggle to explain why. Science. 1998;280:1367– 8. 3. Wickelgren I. Obesity: how big a problem? Science. 1998; 280:1364 –7. OBESITY RESEARCH Vol. 8 No. 1 Jan. 2000

47


4. Anonymous. Executive summary of the clinical guidelines on the identification, evaluation, and treatment of overweight and obesity in adults. Arch Intern Med. 1998;158:1855– 67. 5. Quesenberry CPJ, Caan B, Jacobson A. Obesity, health services use, and health care costs among members of a health maintenance organization. Arch Intern Med. 1998;158:466 – 72. 6. Henry RR, Wallace P, Olefsky JM. Effects of weight loss on mechanisms of hyperglycemia in obese non-insulin-dependent diabetes mellitus. Diabetes. 1986;35:990 – 8. 7. Uusitupa MI, Laakso M, Sarlund H, Majander H, Takala J, Penttila I. Effects of a very-low-calorie diet on metabolic control and cardiovascular risk factors in the treatment of obese non-insulin-dependent diabetics. Am J Clin Nutr. 1990; 51:768 –73. 8. Wing RR, Koeske R, Epstein LH, Nowalk MP, Gooding W, Becker D. Long-term effects of modest weight loss in type II diabetic patients. Arch Intern Med. 1987;147:1749 –53. 9. Rabkin SW, Boyko E, Wilson A, Streja DA. A randomized clinical trial comparing behavior modification and individual counseling in the nutritional therapy of non-insulin-dependent diabetes mellitus: comparison of the effect on blood sugar, body weight, and serum lipids. Diabetes Care. 1983;6:50 – 6. 10. Willey KA, Molyneaux LM, Yue DK. Obese patients with type 2 diabetes poorly controlled by insulin and metformin: effects of adjunctive dexfenfluramine therapy on glycaemic control. Diabetic Med. 1994;11:701– 4. 11. Capstick F, Brooks BA, Burns CM, et al. Very low calorie diet (VLCD): a useful alternative in the treatment of the obese NIDDM patient. Diabetes Res Clin Pract. 1997;36:105–11. 12. Stewart GO, Stein GR, Davis TM, Findlater P. Dexfenfluramine in type II diabetes: effect on weight and diabetes control. Med J Aust. 1993;158:167–9. 13. Henry RR, Wiest-Kent TA, Scheaffer L, Kolterman OG, Olefsky JM. Metabolic consequences of very-low-calorie diet therapy in obese non-insulin-dependent diabetic and nondiabetic subjects. Diabetes. 1986;35:155– 64. 14. Guare JC, Wing RR, Grant A. Comparison of obese NIDDM and nondiabetic women: short- and long-term weight loss. Obes Res. 1995;3:329 –35. 15. Jackson RA, Moloney M, Lowy C, et al. Differences between metabolic responses to fasting in obese diabetic and obese nondiabetic subjects. Diabetes. 1971;20:214 –27. 16. Wing RR, Marcus MD, Epstein LH, Salata R. Type II diabetic subjects lose less weight than their overweight nondiabetic spouses. Diabetes Care. 1987;10:563– 6. 17. Schutz Y, Golay A, Felber JP, Jequier E. Decreased glucose-induced thermogenesis after weight loss in obese subjects: a predisposing factor for relapse of obesity? Am J Clin Nutr. 1984;39:380 –7.

48


18. Lean ME, Powrie JK, Anderson AS, Garthwaite PH. Obesity, weight loss and prognosis in type 2 diabetes. Diabetic Med. 1990;7:228 –33. 19. Ravussin E, Bogardus C, Schwartz RS, et al. Thermic effect of infused glucose and insulin in man: decreased response with increased insulin resistance in obesity and noninsulindependent diabetes mellitus. J Clin Invest. 1983;72:893–902. 20. Saltiel AR, Olefsky JM. Thiazolidinediones in the treatment of insulin resistance and type II diabetes. Diabetes. 1996;45: 1661–9. 21. Hartley GG, Nicol S, Halstenson C, Khan M, Pheley A, St. Peter JV. Long-term results from phentermine, fenfluramine, diet, behavior modification, and exercise for treatment of obesity. Obes Res. 1997;5:58S (abstr.). 22. Kilo C, Meenan A, Bloomgarden Z. Glyburide versus glipizide in the treatment of patients with non-insulin-dependent diabetes mellitus. Clin Ther. 1992;14:801–12. 23. Alexis G, Henault R, Sparr HB. Conversion from glipizide to glyburide: a prospective cost-impact survey. Clin Ther. 1992;14:409 –17. 24. Khan MA, Herzog CA, St. Peter JV, et al. The prevalence of cardiac valvular insufficiency assessed by transthoracic echocardiography in obese patients treated with appetite-suppressant drugs. N Engl J Med. 1998;339:713– 8. 25. Khan MA, St. Peter JV, Herzog CA, Hartley GG. Appetite suppressant-related valvulopathy in obese diabetic and nondiabetic individuals. Diabetes. 1998;47:A311 (abstr.). 26. Murray DC. Treatment of overweight: I. Relationship between initial weight and weight change during behavior therapy of overweight individuals: analysis of data from previous studies. Psychol Rep. 1975;37:243– 8. 27. Wing RR, Shoemaker M, Marcus MD, McDermott M, Gooding W. Variables associated with weight loss and improvements in glycemic control in type II diabetic patients in behavioral weight control programs. Int J Obes. 1990;14:495– 503. 28. Feldman JM, Lebovitz FL. Tests for glucosuria: an analysis of factors that cause misleading results. Diabetes. 1973;22: 115–21. 29. Golay A, Swislocki AL, Chen YD, Reaven GM. Relationships between plasma-free fatty acid concentration, endogenous glucose production, and fasting hyperglycemia in normal and non-insulin-dependent diabetic individuals. Metabolism. 1987;36:692– 6. 30. Goldner MG, Knatterud GL, Prout TE. Effects of hypoglycemic agents on vascular complications in patients with adult-onset diabetes. 3. Clinical implications of UGDP results. JAMA. 1971;218:1400 –10.