! Experiments (strongest evidence) ! Observational data

Transcription

1 HEALTH CARE COSTS ASSOCIATED WITH ELEVATED BODY MASS INDEX AND THE COST-EFFECTIVENESS OF WEIGHT LOSS Henry Glick Adam Tsai University of Pennsylvania THE EPISTEMOLOGY OF HEALTH (How do we determine what is healthy and what is not?)! Associations and Causes - Yellow fingers and lung cancer - Coffee and lung cancer - Smoking and lung cancer - Birth order and Downs syndrome - Hormone replacement therapy and coronary artery disease - Autologous bone marrow transplantation and breast cancer survival HOW DO WE ESTABLISH CAUSALITY?! Experiments (strongest evidence)! Observational data - Coherence with existing information (biological plausibility) - Consistency of association - Time sequence - Specificity of association - Reversibility of association - Strength of association * Quantitative strength * Dose-response relationship * Study design! Are there any randomized trials that demonstrate the beneficial impact of diet and exercise on health outcomes such as death, disease, and disability? - Poor control of diabetes and microvascular and macrovascular events - HIV and AIDS 1 2

2 GENERAL STRATEGY FOR IDENTIFYING THE DISEASE BURDEN/COSTS ASSOCIATED WITH ELEVATED BODY MASS INDEX! Grab a bunch of people with elevated body mass indices (BMI) and a bunch of people with average/ ideal BMIs! Assess the difference in the disease burden / health care cost between the two groups! Attribute the difference to elevated BMI WHAT CAN GO WRONG?! The two groups may not differ in BMI alone, but also may differ in other factors that affect disease burden/cost - Age - Gender - Socioeconomic status - Genetics - Medical conditions, obesity-related/unrelated - Other unmeasured/unmeasurable factors HOW DO WE ADDRESS THESE ISSUES?! To account for other differences between the groups: - Use statistical techniques that attempt to control for the differences - Often not possible to fully account for differences between groups in observational studies! To sort out causal chain - Collect data that allow one to differentiate between the alternative, potential causes HAVE WE SORTED OUT THE CAUSAL CHAIN?! Do we know the independent effects of weight, physical activity, and fitness?! Clinical judgment: What are health risks for an obese person who vigorously exercises an hour a day, 5 days a week, and -- where necessary -- maintains normal blood pressure, lipid, and blood sugar levels by taking niacin, statins, and metformin?! Could our aesthetic judgments about overweight/ obesity be affecting our scientific judgments?! BMI might not be the cause of the disease/cost, but might be another expression of a common cause of both BMI and disease/cost (e.g., yellow fingers may not cause lung cancer, yellow fingers and lung cancer may both be caused by smoking) 3 4

3 BAD NEWS?! The Swedish Obese Subjects (SOS) study invited participants to receive bariatric surgery (gastric banding, variable gastric banding, vertical banded gastroplasty, or gastric bypass)! It constructed a comparison group by computerized matching, taking into account 18 variables related to morbidity and mortality - Participants in this group were treated according to existing routines for obesity management at different primary health care centers! The study -- whose follow-up is ongoing -- periodically reports results: Sjostrom et al. NEJM. 2004; 351: Agren, et al. Obesity Research. 2002;10: Narbro, et al. Arch Intern Med. 2002;162: RESULTS, SOS 6-YEAR FOLLOW-UP Variable Weight loss Surgery Usual Care P-value 2 years (%) < years (%) < Cumulative hospital days, 6 years Total < Non-surgery Cumulative inpatient costs ($), 6 years Total <0.001 Non-surgery Annual Pharmacy costs (SEK) Total NS Years ?NS?! What conclusions (if any) should we draw about whether or not weight loss reduces healthcare costs? 5 6

4 TYPES OF STUDIES FOR QUANTIFYING THE COSTS OF ELEVATED BMI / MODIFYING BMI! Cost-of-illness - Prevalence-based cost-of-illness studies (PBCOI) * Cohort studies * Attributable risk/cost studies - Incidence-based cost-of illness studies! Treatment of obesity studies - Randomized trials - Decision analyses/simulation models COST-OF-ILLNESS STUDIES! Prevalence- vs. Incidence-Based - Prevalence-Based (PBCOI) * Measure the value of resources used or lost due to a condition during a specified period of time (e.g., a year), regardless of the time of onset of the condition - If population/ disease is in steady state, correctly weights different durations of disease - If not in steady state -- for example if the prevalence is increasing -- can misweight some costs - Incidence-Based * Measure the value of resources used or lost due to a condition from its onset until cure or death 7 8

5 PBCOI: COHORT STUDIES! Identify two groups of study participants, those who have been exposed to the risk factor under study (e.g., elevated BMI) and those who have not been exposed - Two main types of published cohort studies: * Nationally representative samples * Less representative samples! Estimate health care costs of study participants! Predict (with varying degrees of technical sophistication) cost as a function of BMI - The cost difference associated with BMI is taken to represent the cost difference due to the exposure NATIONALLY REPRESENTATIVE COHORT STUDIES! Use data from nationally representative samples of individual study participants, e.g., - Medical Expenditure Panel Survey (MEPS): self-report confirmed through physician offices, hospitals, and insurance coverage - National Health Interview Survey (NHIS) - Healthcare for Communities household telephone survey (HCC) - National Health and Nutrition Examination Survey (NHANES) - CDC Behavioral Risk Factor Surveillance System (BRFSS)! Obtain estimates of cost and BMI for each study participant - Cost: obtain direct measures or obtain measures of resource use and multiply by estimates of unit costs! Predict cost as a function of BMI 9 10

6 LESS REPRESENTATIVE COHORT STUDIES! Use less representative sample of individual study participants, e.g., - HMO s - Employers! Obtain estimates of cost and BMI - Administrative data on costs (e.g., HMO or employer insurance records) - Direct measurement of BMI and risk factors or self-report of BMI and risk factors * Health Risk Appraisals * Health Surveys! Predict cost as a function of BMI PBCOI: ATTRIBUTABLE RISK/COST STUDIES! Obtain national estimates of the one-year cost of diseases associated with BMI (from various sources)! Obtain estimates of the population-attributable risk% (PAR%) or obesity-attributable etiologic fractions (i.e., proportion of disease/cost that is due to obesity) - e.g. suppose 40% of the population is nonoverweight/obese and 5% have coronary heart disease (CHD); suppose 60% of the population is overweight/obese and 20% have CHD, then the PAR% equals: (0.6 x ( )) / ((0.6 x 0.2) + (0.4 x 0.05) Fract OW x (Risk diff) / Expected risk = 0.09 / ( ) = 0.64 i.e., 9% of the 14% total is "due" to obesity - More formally: P(RR - 1) / (1 + P(RR - 1)) where P equals prevalence of obesity and RR equals the relative risk for disease (e.g., 0.2/0.05)! Multiply cost x PAR% to obtain the cost-of-illness 11 12

7 PBCOI: ATTRIBUTABLE RISK/COST STUDIES (cont.)! Issues with PAR% - Cause vs. association - Can PAR%s be transferred between studies, or are they specific to the prevalence of the risk factor, the prevalence of disease, other factors? * Prevalence of overweight differs widely * Prevalence of disease differs within ethnic groups at the same BMI SUMMARY OF COST STUDIES BY TYPE OF STUDY * Study Nat rep Cohort Nonrep Cohort Attrib Risk/Cos t Other 1. Finkelstein x 2. Sturm x 3. Heitoff x 4. Detournay x + 5. Andreyeva x 6. Kuriyama x + 7. Thorpe x 8. Wang (2002) x 9. Goetzel x 10. Quesenberry x 11. Pronk x 12. Thompson (2001) x 13. Burton x 14. Raebel x 15. Robbins x 16. Tucker x 17. Wang (2004) x 18. Oster x 19. Wolf (1998) x 20. Colditz x 21. Wolf (1996) x 22. Birmingham x Swinburn x Allison x 25. Wang (2002) x 26. Thompson (1999) x 13 14

8 27. Daviglus x 28. Gorsky x + * Results of MEDLINE search for English-language articles since 1996 that directly estimate (rather than review) the costs of obesity + Non-US COST-EFFECTIVENESS STUDIES! Even if we can prove that elevated BMI is associated with excess costs, does not necessarily mean that we can save money by having people reduce their BMI! Should therefore evaluate the value for the cost of weight loss programs! Cost-effectiveness studies compare the difference in costs and difference in outcomes between patients who "lose weight" with those who do not. Observed differences in costs and outcomes (assumed to be) caused by observed weight loss COST-EFFECTIVENESS STUDIES (II)! Two main types of studies: trial-based evaluations vs decision analyses/simulation models! Trial-based: Directly observe weight loss, costs, and outcomes - Strength: Direct observation - Weakness: Usually short term in duration! Decision analyses/simulation models: Usually directly observe weight loss only; use epidemiologic models for relationships between weight loss and outcome and weight loss and cost - Strength: Can model life-time costs - Weakness: Assumption- rather than data-driven )C / )E 15 16

9 COST-EFFECTIVENESS STUDIES (III)! Studied Interventions include: - Lifestyle intervention (multidisciplinary) * University-based SUMMARY OF COST-EFFECTIVENESS STUDIES. BY TYPE OF STUDY * Study DPP Nationally Rep Trial x Nonrepres Trial Decision Analysis * Commercial weight loss programs Wylie-Rosett x * Managed care programs - Weight loss medications - Bariatric surgery! Outcomes used for these studies vary - lb/kg/bmi units lost - Case of diabetes mellitus prevented - Quality-adjusted life years saved! How should we compare studies with these different outcomes? Foxcroft Martin x Christou Agren Agren Spielman Oster Craig, Fang Lamotte DPP (Herman) DPP (Eddy) * Partial results of Medline search Nonrandomized K Non-US x x K x K x K x x x x x x 17 18

10 STUDY RESULTS. PREVALENCE OF OVERWEIGHT/OBESITY * Study Data Source Overweight % Obese Hedley NHANES Sturm HCC Thompson ('01) HMO Wang ('02) NHIS Finkelstein NHIS NHANES 1994 NHANES NHANES Oster HMO * Overweight commonly defined as BMI >25 and <30; obesity commonly defined as BMI >30 NHANES Overweight defined as BMI >25 and <29; obesity defined as BMI >29 ONE-YEAR PER-PERSON COST/EXPENDITURE/CHARGE ESTIMATES Study Over-weight Obese Pooled Nationally Representative Cohort Heithoff Sturm Finkelstein Kuriyama (K) Less Representative Cohort Wang (04) Quesenberry Thompson (01) Bungum Attributable Risk/Cost Oster NA 345 NA Wolf (1998) NA 1137 NA Colditz NA 1050 NA * Generally pooled by assuming that the prevalence of overweight is 34.7% and the prevalence of obesity is 30.5% (Hedley), except for Kuriyama et al (Japan), where prevalences are 26% and 2.8%, respectively K Non-US 19 20

11 ONE-YEAR NATIONAL COST ESTIMATES * COST-EFFECTIVENESS OF TREATMENT OF OBESITY $ (billions) Study Overweight Obese Total Nationally representative cohort Heithoff Sturm Finkelstein Kuriyama (K) Less representative cohort Wang (04) Quesenberry Thompson (01)* Bungum Attributable risk/cost Oster * NA 21.3 NA Wolf (1999) NA 70.1 NA Colditz NA 64.7 NA Average (N) 11.6 (8) 38.5 (11) 45.0 (8) Study Intervention Outcome measure Result Randomized controlled trials, nationally representative DPP DPP Randomized Wylie- Rosett Lifestyle vs placebo DM Prevented $24,400 QALY $51,600 Metformin weakly dominaed by lifestyle Computerized lifestyle Lb lost $6 - $18 Foxcroft Orlistat QALY $76,667 Cohort Studies Martin Medical vs surgical $/LB lost >$250 both arms, NS * Overweight = Cost per overweight person * 70,500,000; obesity = cost per obese person * 61, 610,

12 COST-EFFECTIVENESS OF TREATMENT OF OBESITY Study Decision Analyses DPP (Herman) DPP (Eddy) Craig, Fang Intervention Lifestyle vs placebo Lifestyle vs. placebo Bariatric surgery Outcome measure Result QALY 1100 QALY QALY $5,000-35,600 Lamotte Orlisata QALY $4,415-25,500 Spielman Commercial WL Cost/lb lost Oster Unspecified % weight loss $2 - $26 Dominates SUMMARY OF FINDINGS (I)! Prevalence - Overweight: 34% - 37% - Obesity: 19% - 34%! Conclusion: Self-reported weight and height seems to be consistent across a number of studies! Might there be a consistent reporting bias?! What would be the impact of such a bias on the results of cohort studies? - Under-reporting of the prevalence of obesity would underweight the impact of obesity costs * If mis-reporting generally leads to the classification of higher cost overweight/obese people as being normal weight, the average cost among normal weight people would be increased (thus attributing some of the costs of overweight/obesity to normal weight) - Impact on mean cost among overweight/obese people would depend on whether the misreporting was uniformly distributed among overweight/obese people 23 24

13 SUMMARY OF FINDINGS (II)! Annual Cost Per Person Cohort Attrib Risk/Cost SOURCES OF DIFFERENCE BETWEEN COHORT STUDIES AND PBCOI (II)! Data used in attributable risk/cost studies potentially prone to double counting Overweight: -$104 - $273 NA Obesity: $262 - $767 $345 - $1137! Conclusions - Overweight/obesity associated with substantial costs - Cohort studies appear to yield lower estimates of the cost of disease than do attributable risk/cost studies SOURCES OF DIFFERENCE BETWEEN COHORT STUDIES AND PBCOI (I)! Data used in cohort studies usually come from one or a small number of studies, often with consistent data, whereas data used in attributable risk/cost studies often come from many studies, often with inconsistent data - e.g., the population attributable fraction may be specific to a particular population with a particular prevalence of overweight/obesity and of disease. It may not be transferable to other populations with other prevalences - e.g., Most of these studies included national cost estimates for some or all of the following diseases: * Coronary heart disease (CHD) * Hypertension * Hypercholesterolemia * Gall bladder * Stroke * DVT/Pulmonary embolism * Type II diabetes * Osteoarthritis * Breast cancer * Colon cancer * Endometrial cancer * Obstructive sleep apnea * Congestive heart failure * Gastroesophageal reflux disease - However, the cost of diabetes can include CHD, hypertension, and hypercholesteremia; the costs of hypertension can include CHD and stroke, and the cost of hypercholesterolemia can include CHD. Thus the cost of CHD might be counted four times in these studies * In the short-term, what do you think would happen to the PAR% if there was a sudden increase in the number of overweight/obese people? 25 26

14 SUMMARY OF FINDINGS (III)! Some studies suggests that costs of overweight/obesity greater for overweight/obese women than for overweight/obese men (data not shown) SUMMARY OF FINDINGS (IV)! J-Shaped Curve for the association between BMI and costs (i.e., higher costs for underweight and obese) - Occurs even though men are more likely to be overweight, and there is at most a tendency for more women than men to be obese Source: Heithoff, 1997! Strong or weak evidence reported in 7 of the cohort studies (Heithoff, Goetzel, Quesenberry, Thompson, Burton, Wang '02, Kuriyama)! Ideal BMIs vary between 21 and 27.5! What conclusions (if any) about weight loss should we draw from the J-shaped curve? 27 28

15 SUMMARY OF FINDINGS (V)! Overweight/obesity, Chronic Disease, and Cost - Most of the studies that controlled for chronic medical conditions associated with obesity found that there was still an "independent" effect of weight after accounting for these - Some studies controlled for variables that may be in the causal pathway of overweight/obesity (e.g., poor nutritional habits, hyperlipidemia, blood glucose, depression, etc.) - What conclusions (if any) should we draw about whether or not overweight / obesity alone increases health care costs?! Overweight/obesity, Other Risk Factors, and Cost - Some studies accounted for confounding, while others did not SUMMARY OF FINDINGS (VI)! Cost of treating obesity - Few studies - Heterogeneous outcomes - The three studies that directly compared lifestyle intervention with medications found lifestyle to be more cost-effective (what does insurance cover?) - Only three published studies estimated lifetime benefits (vs. within-trial benefits) - The two long-term decision analyses of the DPP (Herman et al, Eddy et al) arrived at very different conclusions about the cost-effectiveness of treating obesity (why?) See Table, end of handouts 29 30

16 TECHNICAL FINDINGS! Cost data (in cohort studies) highly skewed - Where reported, the median much smaller than the mean or mean much smaller than the S.D. (given costs can t be negative, can infer observations with very high costs) - Makes appropriate statistical analysis difficult, and many of the cost estimates used in the papers may be flawed! Cost data inconsistent - Different studies use different measures: * Costs * Expenditures * Charges - Known to differ consistently in health care (most likely increase in the listed order) - From a social perspective, costs are more important than expenditures or charges SO WHAT?! You perform a cost-of-illness study, and report that overweight/obesity is responsible for between $60 and $375 in health care costs per person per year in the U.S. (i.e., between $30 and $70+ billion per year)! What are these numbers supposed to be? - Actual measures of cost? - Warning signs? - Projections of what they will be?! What incentives do investigators have to report higher or lower results?! How do these numbers help us make decisions? - About treatment? - About Prevention?! What numbers would be better? 31 32

17 CONCLUSIONS (I) Overweight/obesity is a major public health problem that costs society billions of dollars, and we should be doing everything we can to combat it.! Reasons to agree with this statement - Consistent association between BMI, health, and cost - Biological plausibility - Quantitatively strong association - Dose response - Correct time sequence CONCLUSIONS (II)! Reasons to question this statement - The bulk of the evidence we have for this statement come from studies like the ones that supported our public health decisions about smoking and lung cancer and about hormone replacement therapy and CAD - We haven t performed a single trial that demonstrates the impact of weight loss on final health outcomes such as death, disease, and disability * When we had the chance (randomizing 5,000 patients and following them for 7 years), why did we decide to study weight loss in diabetics rather than weight loss in the general public? - We haven t disaggregated the independent effects of weight, physical activity, and fitness - Even if overweight/obesity is a health/economic problem, it is not clear we have cost-effective interventions for its treatment / prevention CONCLUSIONS (III)! At the moment, there are better and worse opinions about statements of this kind, but definitive evidence is lacking 33 34

18 REFERENCES NATIONALLY REPRESENTATIVE COHORT STUDIES US 1. Finkelstein EA, Fiebelkorn IC, Wang G. National medical spending attributable to overweight and obesity: how much, and who's paying? Health Affairs. 2003;22: Sturm R. The effects of obesity, smoking, and drinking on medical problems and costs. Obesity outranks both smoking and drinking in its deleterious effects on health and health costs. Health Affairs. 2002;21: Heithoff KA. Cuffel BJ. Kennedy S. Peters J. The Association Between Body Mass and Health Care Expenditures. Clinical Therapeutics. 1997;19: Andreyeva T, Sturm R. Ringel JS. Moderate and severe obesity have large differences in health care costs. Obesity Research. 2004;12: Daviglus ML, Liu K, Yan LL, et al. Relation of body mass index in young adulthood to Medicare expenditures in older age. JAMA. 2004;292: Thorpe KE, Florence CS, Howard DH, Joski P. The rising prevalence of treatd disease: effects on private health insurance spending. Health Affairs. 2005;W5: Wang G, Zheng ZJ, Health G, Macera C, Pratt M, Buchner D. Economic burden of cardiovascular disease associated with excess body weight in U.S. adults. American Journal of Preventive Medicine. 2002;23:1-6. Non-US 8. Detournay B. Fagnani F. Phillippo M. Pribil C. Charles MA. Sermet C. Basdevant A. Eschwege E. Obesity morbidity and health care costs in France: an analysis of the Medical Care Household Survey. International Journal of Obesity & Related Metabolic Disorders. 2000;24: Kuriyama S, Tsuji I, Ohkubo T, Anzai Y, Takahashi K, Watanabe Y, Nishino Y, Hisamichi S. Medical care expenditure associated with body mass index in Japan: the Ohsaki Study. Int J Obes Relat Metab Disord Aug;26(8): LESS REPRESENTATIVE COHORT STUDIES 10. Goetzel RZ. Anderson DR. Whitmer RW. Ozminkowski RJ. Dunn RL. Wasserman J. The relationship between modifiable health risks and health care expenditures. An analysis of the multi employer HERO health risk and cost database. The Health Enhancement Research Organization (HERO) Research Committee. Journal of Occupational & Environmental Medicine. 1998;40: Quesenberry CP. Caan B. Jacobson A. Obesity, Health Services Use, and Health Care Costs Among Members of a Health Maintenance Organization. Arch Intern Med. 1998;158: Pronk NP. Goodman MJ. O'Connor PJ. Martinson BC. Relationship Between Modifiable Health Risks and Short term Health Care Charges. JAMA. Volume 282(23), 15 December 1999, pp Thompson D. Brown JB. Nichols GA. Elmer PJ. Oster G. Body mass index and future healthcare costs: a retrospective cohort study. Obesity Research. 2001;9: Burton WN. Chen CY. Schultz AB. Edington DW. The economic costs associated with body mass index in a workplace. Journal of Occupational & Environmental Medicine. 1998;40: Bungum T, Satterwhite M, Jackson AW, Morrow JR. The relationship of body mass index, medical costs, and job absenteeism. American 35 36

19 Journal of Health Behavior. 2003;27: Raebel MA, Malone DC, Conner DA, Xu S, Porter JA, Lanty FA. Health services use and health care costs of obese and non-obese individuals. Archives of Internal Medicine. 2004;164: Robbins AS, Chao SY, Russ CR, Fonseca VP. Costs of excess body weight among active duty personnel, U.S. Air Force, Military Medicine. 2002;167: Tucker LA, Clegg AG. Differences in health care costs and utilization among adults with selected lifestyle-related risk factors. American Journal of Health Promotion. 2002;16: Wang F, McDonald T, Champaigne LJ, Edington DW. Relationship of body mass index and physical activity to health care costs among employees. Journal of Occupational and Environmental Medicine. 2004;46: PREVALENCE-BASED COST OF ILLNESS United States 20. Oster G. Edelsberg J. O'Sullivan AK. Thompson D. The clinical and economic burden of obesity in a managed care setting. American Journal of Managed Care. 2000;6: Wolf AM. Colditz GA. Current estimates of the economic cost of obesity in the United States. Obesity Research. 1998;6: Colditz GA. Economic costs of obesity and inactivity. Medicine & Science in Sports & Exercise. 1999;31:S Wolf AM. Colditz GA. Social and economic effects of body weight in the United States. American Journal of Clinical Nutrition. 1996;63:466S 469S. Non-U.S. 24. Birmingham CL. Muller JL. Palepu A. Spinelli JJ. Anis AH. The cost of obesity in Canada. CMAJ (Canadian Medical Association Journal). 1999;160: Swinburn B. Ashton T. Gillespie J. Cox B. Menon A. Simmons D. Birkbeck J. Health care costs of obesity in New Zealand. International Journal of Obesity & Related Metabolic Disorders. 1997;21: INCIDENCE-BASED COST OF ILLNESS 26. Thompson D, Edelsberg J, Colditz G, Bird Am Oster G. Lifetime health and economic consequences of obesity. Archives of Internal Medicine. 1999;159: Gorsky, Pamuk E, Williamson DF, Shaffer PA, Koplan JP. The 25-year health care costs of women who remain overweight after 40 years of age. Am J Prev Med Sep-Oct;12(5): COST-EFFECTIVENESS OF TREATMENT 28. Herman WH, Hoerger TJ, Brandle M, et al. The cost-effectiveness of lifestyle modification or metformin in preventing type 2 diabetes in adults with impaired glucose tolerance. Annals of Internal Medicine 2005;142: Eddy DM, Schlessinger L, Kahn R. Clinical outcomes and cost-effectiveness of strategies for managing people at high risk for diabetes. Annals of Internal Medicine 2005;143: Spielman AB. Kanders B. Kienholz M. Blackburn GL. The cost of losing: an analysis of commercial weight-loss programs in a metropolitan area. Journal of the American College of Nutrition 1992;11; Wylie-Rosett J. Swengionis C. Ginsberg M, et al. Computerized 37 38

20 weight loss intervention optimizes staff time: the clinical and cost results of a controlled clinical trial conducted in a managed care setting. Journal of the American Dietetic Association 2001;101: Foxcroft DR. Milne R. Orlistat for the treatment of obesity: rapid review and cost-effectiveness model. Obesity Reviews 2000:1; Oster G. Thompson D. Edelsberg J. Bird AP. Colditz GA. Lifetime health and economic benefits of weight loss among obese persons. American Journal of Public Health 1999:89; Craig BM. Tseng DS. Cost-effectiveness of gastric bypass for severe obesity. American Journal of Medicine 2002;113: Fang J. The cost-effectiveness of bariatric surgery. Am J Health Behav 2003;27(4): Lamotte M, Annemans L, Lefever A, Nechelput M, Masure J. A health economic model to assess the long-term effects and cost-effectiveness of orlistat in obese type 2 diabetic patients. Diabetes Care 2002;25: Medicine 2002;162: Agren G. Narbro K. Jonsson E. Naslund I. Sjostrom L. Peltonen M. Cost of inpatient care over 7 years among surgically and conventionally treated obese patients. Obesity Research 2002;10: Sjostrom L, Lindroos AK, Peltonen M et al. Lifestyle, diabetes, and carciovascular risk factors 10 years after bariatric surgery. New England Journal of Medicine. 2004;351: Martin LF, Tan TL, Horn JR, Bixler EO, Kauffman GL, Becker DA, Hunter SM. Comparison of the costs associated with medical and surgical treatment of obesity. Surgery 1995;118(4): Hedley AA, Ogden CL, Johnson CL, Carroll MD, Curtin LR, Flegal KM. Prevalence of overweight and obesity among U.S. children, adolescents, and adults, JAMA 2004;291: Christou NV, Sampalis JS, LIberman M, Look D, Auger S, McLean APH, MacLean LD. Surgery decreases long-term mortality, morbidity, and health care use in morbidly obese patients. Annals of Surgery 2004;240(3): SPECIAL INTEREST 38. Allison DB, Zannolli R, Narayan KMV. The Direct Health Care Costs of Obesity in the United States. American Journal of Public Health. 1999;89: Narbro K. Agren G. Jonsson E. Naslund I. Sjostrom L. Peltonen M. Pharmaceutical costs in obese individuals: comparison with a randomly selected population sample and long-term changes after conventional and surgical treatment: the SOS intervention study. Archives of Internal 39 40

21 Demonstration that P H (R H - R L ) / {(P H R H ) + (P L R L )} = P(RR - 1) / {1 + P(RR - 1)} Demonstration that P H (R H - R L ) / {(P H R H ) + (P L R L )} = P(RR - 1) / {1 + P(RR - 1)} 1. Begin with: P H (R H - R L ) / {(P H R H ) + (P L R L )} = P(RR - 1) / {1 + P(RR - 1)} 2. Substitute 1-P H for P L on left-hand side: P H (R H - R L ) / {(P H R H ) + (P L R L )} = 3. Multiply out: P H (R H - R L ) / {(P H R H ) + ((1-P H ) R L )} P H (R H - R L ) / {(P H R H ) + ((1-P H ) R L )} = {(P H R H ) - (P H R L )} / {(P H R H ) + R L - (P H R L ) 4. Divide numerator and denomiator by R L ; if the result equals R H / R L, substitute RR: = (P H RR - P H ) / (1 + P H RR - P H ) 5. Simplify elements containing P H : 1. Begin with risk difference as a fraction of total risk: {P H ([R L RR]-R L )} / {(P H RR R L ) +([1-P H ] R L )} 2. Simplify RL in the numerator {P H R L (RR-1)} / {(P H RR R L ) +([1-P H ] R L )} 3. Divide numerator and denominator by RL {P H (RR-1)} / {(P H RR) +([1-P H ])} 4. Rearrange denominator {P H (RR-1)} / {1 + (P H RR) - P H } 5. Simplify elements in denominator containing P H : {P H (RR-1)} / {1 + P H (RR - 1)} QED = {P H (RR - 1)) / {1 + P H (RR - 1)} QED 41 42

22 RESULTS GIVEN CONTROL FOR COMORBIDITIES Unadj Gluc/ diab Study Adj cost BP Lipids Depres CHD Other Daviglus 1010 x x Abnorm EKG 690 Pronk 1134 x x 479 Raebel 699 x x x 456 Wang ' x 533 Wang ' x x Arthritis, back 654 pain Quesen-berry 953 x x x x x Muscular pain -645 Average cost, unadjusted: 1070 Average cost, adjusted: