PAPER Pharmacotherapy for Obesity: a Quantitative Analysis of Four Decades of Published Randomized Clinical Trials

Home , Benzphetamine

International Journal of Obesity (2002) 26, 262–273 ß 2002 Nature Publishing Group All rights reserved 0307–0565/02 $25.00 www.nature.com/ijo PAPER Pharmacotherapy for obesity: a quantitative analysis of four decades of published randomized clinical trials

CK Haddock1*, WSC Poston1, PL Dill1, JP Foreyt2 and M Ericsson3

1University of Missouri-Kansas City and Mid America Heart Institute, St Luke’s Hospital, Kansas City, Missouri, USA; 2Baylor College of Medicine, USA; and 3University of Umea˚, Sweden and University of Missouri-Kansas City, Missouri, USA

AIM: This article provides the first comprehensive meta-analysis of randomized clinical trials of medications for obesity. METHOD: Based on stringent inclusionary criteria, a total of 108 studies were included in the final database. Outcomes are presented for comparisons of single and combination drugs to placebo and for comparisons of medications to one another. RESULT: Overall, the medications studied produced medium effect sizes. Four drugs produced large effect sizes (ie d > 0.80; amphetamine, benzphetamine, fenfluramine and sibutramine). The placebo-subtracted weight losses for single drugs vs placebo included in the meta-analysis never exceeded 4.0 kg. No drug, or class of drugs, demonstrated clear superiority as an obesity medication. Effects of methodological factors are also presented along with suggestions for future research. International Journal of Obesity (2002) 26, 262 – 273. DOI: 10.1038=sj=ijo=0801889

Keywords: pharmacotherapy; meta-analysis; clinical trials

Introduction Task Force on the Prevention and Treatment of Obesity.8 No comprehensive meta-analytic studies of obesity pharma- Goldstein and Potvin7 conducted a detailed review of 20 cotherapy have been published to date, but three meta- long-term (ie those lasting 6 months or longer) phenter- analyses have been published on single drugs. Lijesen et al1 mine, mazindol, fenfluramine, dexfenfluramine and fluox- meta-analyzed eight controlled and 16 uncontrolled trials of etine studies from 1967 to 1993. While they presented a human chorionic gonadotropin and concluded that it was comprehensive review of clinical trials on these drugs and not effective for the treatment of obesity. Four placebo- concluded that extended treatment was beneficial for those controlled, double-blind studies of sibutramine efficacy for patients unable to lose weight without pharmacotherapy, reducing visceral fat were meta-analyzed by Van Gaal et al,2 they did not provide a quantitative synthesis of these drugs’ who concluded that it was effective in reducing waist cir- efficacy. The National Task Force on the Prevention and cumference and visceral fat when compared to placebo. Treatment of Obesity8 also provided a comprehensive Greenway3 examined phenylpropanolamine (PPA) trials review of the safety and efficacy of FDA-approved and since 1973, including trials that were reviewed by previous selected non-approved anti-obesity medications, ie (bold investigators4,5 and concluded that the average weight loss drugs are those that were identified as FDA-approved) in excess of placebo (kg=week) at the end of studies had amphetamine=dexamphetamine, benzphetamine hydro- decreased since 1985. chloride, dexfenfluramine, diethylpropion, fenfluramine, Two multi-drug reviews that have been cited as meta- fluoxetine, mazindol, methamphetamine hydrochloride, analytic confirmation of obesity pharmacotherapy’s efficacy6 phendimetrazine tartrate, phentermine (hydrochloride were authored by Goldstein and Potvin7 and the National and resin), phenlypropanolamine, the combination of phentermine – fenfluramine, sibutramine and sertraline, but they also did not quantitively synthesize outcomes of the drug therapies. The Task Force reviewed English-language *Correspondence: CK Haddock, Health Research Group, University of trials that evaluated drug safety and efficacy that lasted for a Missouri-Kansas City, 4825 Troost, Room 124, Kansas City, MO 64110, minimum of 24 weeks and concluded that pharmacotherapy USA. for obesity, when combined with appropriate behavioral E-mail: [email protected] Received 5 January 2001; revised 14 August 2001; approaches, helped many obese patients lose weight or accepted 1 October 2001 maintain their weight loss.8 Obesity medication meta-analysis CK Haddock et al 263 The purpose of this study is to provide a comprehensive mocriptine, buspar, cimetidine, fluvoxamine, human chor- meta-analytic review of anti-obesity agents, both prescrip- ionic gonadotropin, human growth hormone, leptin, tion and over-the-counter (OTC), and drugs that are=were naloxone=naltrexone or synthyroid. We also did not include FDA-approved and are=were used off-label for obesity.8 The dietary supplements, which are defined by the Dietary Sup- specific aims are to evaluate the clinical efficacy of obesity plement Health and Education Act of 19949 as products medications and determine whether methodological factors intended to supplement the diet that contain one or more (eg length of treatment, year of publication) are system- of the following ingredients: (1) a vitamin; (2) a mineral; (3) atically related to treatment outcome. In addition, based a herb or other botanical; (4) an amino acid; (5) a dietary on the results of this meta-analysis, suggestions for future substance for use to supplement the diet by increasing the research are discussed. total dietary intake; or (6) a concentrate, metabolite, constituent, extract, or combination of any of the previously described ingredients. Examples of substances in this cate- Methods gory include 5-hydroxytryptophan (5-HTP), ma huang (ephe- Literature search drine), guarana (caffeine), chitosan, chromium (picolinate Inclusion criteria. For this meta-analysis, we evaluated and nicotinate), dehydroepiandrosterone (DHEA), garcinia only anti-obesity agents that are=were FDA-approved for cambogia=hydroxycitric acid, pyruvate and St John’s Wort the treatment of obesity, both prescription and OTC, and (hypericin). drugs that are FDA-approved and are used off-label for Studies which met the following stringent criteria were obesity (see Table 1). We compiled this list by examining included in the review: (1) the data were contained in those that were highlighted by the National Task Force on published reports in peer-reviewed journals; (2) only the Prevention and Treatment of Obesity8 as being currently human studies were included; (3) an English version of the approved for the treatment of obesity in the United States, study was available; (4) a direct comparison between an those that were used off-label, and including recently obesity drug therapy designed to produce weight loss and approved drugs that were not reviewed by the Task Force.8 another treatment modality or a control group of obese In addition, we based our selection of drugs on extensive individuals was provided; (5) participants were assigned consultation with several experts in the field of obesity randomly to treatment groups and the randomization research. scheme was not broken during assignment (ie some partici- We retained fenfluramine and dexfenfluramine in the pants assigned randomly, some haphazardly); (6) groups review because they were widely studied and used in clinical were distinguishable on relevant parameters (eg drug type, settings even though they were removed from the market in use of lifestyle intervention); (7) the study provided suffi- 1997. We did not include experimental obesity agents such cient outcome data to compute an effect size based on as acarbose, beta-adrenoreceptor agonist (BRL 26830A), bro- weight loss (see effect size definition below); (8) the study

Table 1 Anti-obesity agents

Generic name Example trade name DEA schedule Number of studiesa Years of publications (range)

Amphetamine (dexamphetamine)b Biphetamine II 6 1969 – 1974 Benzocaine Slim Mint, Trocaine OTC 1 1999 Benzphetamineb Didrex III 3 1960 – 1987 Dexfenﬂuramineb,c Redux IV 18 1989 – 1998 Diethylpropionb Tenuate, Tenuate Dospan IV 13 1965 – 1983 Fenﬂuramineb,c Pondimin IV 15 1966 – 1992 Fluoxetined Prozac None 11 1987 – 1995 Mazindolb Sanorex, Mazanor IV 28 1969 – 1982 Methamphetamineb Desoxyn II 1 1960 Orlistate Xenical None 6 1995 – 1999 Phendimetrazineb Bontril, Plegine, Prelu-2, X-Trozine III — f — f Phentermine (HCL and resin)b Adipex-P, Fastin, Oby trim, Ionamin IV 9 1969 – 1992 Phenylpropanolamine (PPA)b Dexatrim, Acutrim OTC 9 1975 – 1999 Sertralined Zoloft None — f — f Sibutraminee Meridia IV 5 1991 – 1998

Note: FDA, Food and Drug Administration; DEA, Drug Enforcement Agency; OTC, over the counter. aNumber of studies in our database that address each drug. Because some studies address more than one drug, the sum of these numbers is greater than the total number of studies included in the database. bDrugs highlighted by the National Task Force on the Prevention and Treatment of Obesity (1996). cRemoved from the market in 1997. dFDA-approved drugs that have been used off-label for obesity. eApproved by the FDA after the 1996 report by the National Task Force on the Prevention and Treatment of Obesity. fNo studies of these drugs met our inclusionary criteria.

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 264 was published on or before December 1999 (to provide a DXti À DXci di ¼ point to begin coding and data analysis). Unfortunately, we DXsi were not able to code a large number of studies that address where DXti is the mean weight loss in the treatment group drug treatments for obesity. Uncodable studies typically did D of the ith study, Xci is the mean of the control or not present data in a manner where group outcomes could D alternative treatment in the ith study, and Xsi is the be precisely distinguished (eg cross-over studies where data pooled standard deviation of change for the two groups. were only presented at the conclusion of the study) or did When sufficient data were not reported to directly com- not present sufficient data to compute an effect size (typi- pute d, standard alternative methods of deriving the cally these studies presented no data on outcome variability effect size were used if possible.13 This effect size controls nor information where outcome variance could be esti- for both placebo effects and lifestyle treatments in esti- mated). Finally, there were some studies where drugs were mating the effect of the drugs. Because effect sizes based used for weight maintenance following obesity treatment, on change scores tend to be large compared to those but were not used as part of a primary treatment to produce based on post-test mean differences,14 direct comparisons weight loss.10 – 12 We located a small number of maintenance to reviews using effect sizes based on post-test scores articles and, although codeable, they were not included in should be avoided. the analyses. The statistical procedures for combining effect sizes in this Studies were located by computer searches of databases review weight each study by an inverse function of its sample (eg Medline, PsychInfo), reviewing tables of content= variance — as the sampling variance becomes small, the reference sections of journals, abstracts, previous reviews, weight becomes large (Shadish and Haddock,15 formula past empirical studies, relevant book chapters, and recent 18 – 7). When a single study provided more than one rele- issues of journals which regularly publish obesity research vant effect size for an analysis, all within-study effect sizes (eg American Journal of Clinical Nutrition, International Journal were aggregated to avoid statistical dependency. Consistent of Obesity and Related Metabolic Disorders, Journal of the Amer- with conventions in meta-analysis, results described only as ican Medical Association, Journal of Consulting and Clinical insignificant where adequate data were not presented to Psychology; and Obesity Research). In addition, a number of compute an effect size were conservatively coded as zero. individuals who regularly publish in the obesity literature Hedges’ correction for small sample bias (Hedges,16 Formula were asked to provide personal lists of obesity studies that 4) was applied to all effect sizes. address pharmacotherapy. Based on inclusionary criteria and In addition to providing d as a measure of effect size, raw the search procedures, a total of 108 randomized clinical differences between treatment group weight losses are also trials (published in 103 articles) were located. The character- provided for each study. Although meta-analytic methods istics of these studies are outlined in the results section have been developed for statistically combining raw data in below. their original metric, many studies did not provide sufficient data on the variance of weight changes with which to compute appropriate weights.15 Thus, these unweighted Coding of studies raw differences between treatment groups are provided for A Pharmacotherapy for Obesity: a Meta-Analysis of Control Trials descriptive purposes only and should be interpreted with Coding Manual containing the operational definitions of the caution. variables used in this review was developed (available via email in MS Word for Windows format upon request from the first author). Reliability of coding was maintained by providing intensive training of the project assistants, includ- Statistical analyses ing approximately 20 h of didactic and coding practice. Each Standard statistical procedures for meta-analysis were used coder was required to reach perfect agreement with sample for data analysis.17,18 Outcomes were considered a post-test if studies coded by the principal investigator (CKH) prior to they (1) occurred at the end of an intact treatment package, coding other studies. Finally, another project research assis- (2) were the only outcome provided by the authors, or (3) tant independently verified all coding. Because the majority were designated as a post-test by the study authors. For many of codes used in this review required little judgement (eg studies, coding follow-up or long-term (ie after the initial average weight of subjects, drug name), consistent coding post-test outcomes) effect sizes proved challenging. was easily achieved. When parameters varied during the Although several studies continued to monitor the weights course of a study (eg drug dose), an average of that parameter of patients past initial or post-test assessments of drug was coded for the meta-analysis. effects, many significantly altered the original study design and thus precluded our ability to present directly interpre- Coding and combining effect sizes table treatment effect sizes. Thus, these studies essentially The standardized mean difference, d, based on change scores present a new research trial based on previously treated (reduction in weight), used as the measure of effect size in patients. Only clinical trials that maintained the basic this review, is defined as: structure of their research design and allowed for ‘clean’

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 265 examination of long-term treatment effects were coded as 10.0) 5.0) 7.3)

11.5) follow-ups. – – – – 7.4) 4.2) 2.0)

7 3.3) 5.3) 6.0) 5.1) – – – – – – – 0.80 0.20 1.6 0.10 0.10 7 7 7 7 7 Results Characteristics of studies Appendix 1 contains a complete list of studies included in 2.0) 3.3 (1.7 11.3) 3.82 ( 10.5) 3.00 ( 2.4) 3.3 (0.20 – – – –

4.3) 0.89 (0.30 the meta-analysis. Of the 108 clinical trials included, 102 2.6) 2.7 (2.1 3.2) 2.41 ( 9.3) 2.7 ( 6.1) 2.08 (0.30 5.2) 3.6 (0.6 3.3) 3.5 (2.4 – – – – – – – 1.3 0.4 0.4 1.5 were primarily concerned with pharmacologically induced 7 7 7 7 weight loss. In the remaining six studies weight loss was a secondary outcome, with factors such as macronutrient intake serving as the primary endpoint of concern. However,

6.1) 1.9 (0.60 these six studies used medication designed to promote 5.9) 2.5 (2.4 7.3) 0.73 ( 21.3) 5.1 ( 13.1) 3.5 ( 11.6) 2.41 (1.2 9.3) 0.78 ( 10.1) 3.03 (0.4 10.3) 5.02 (3.0 8.8) 2.8 (1.5 7.3) 1.8 (0.8 – – – – – – – – – – – weight loss and reported sufﬁcient data to be included in ssary information. Weeks of treatment are based on the meta-analysis. Publication dates of the studies ranged placebo effect sizes. Data presented for kg lost for drug

– from 1960 to 1999, with 9.3% in the 1960s, 42.6% in the ect sizes (computed for drugs with less than three studies). 1970s, 12.0% in the 1980s and 36.1% published in the 1990s. Average age of the subjects included in the clinical trials was 40.7 y, although actual ages ranged from 5 y (Stewart et al,94 Appendix 1) to 77.0 y. 60) 81.8 4.10 (1.4 20.0) 100 5.3 (4.0 150) 66.7 4.03 (1.6 360) 100 7.1 (4.0 – – day using lifestyle kg lost with kg lost Drug 15) 50 5.2 (4.5 130) 92.9 8.9 (3.7 75) 100 6.5 (1.9 120) 64.3 5.06 (2.5 30) 83.3 6.3 (3.6 75) 85.7 3.03 (0.90 – –

= 3) 84.1 5.8 (2.2 – – – – – –

– Single drug vs placebo: post-treatment outcomes

cient data to compute drug Table 2 presents design characteristics of studies providing fi Dosage drug – placebo comparisons. Weeks of treatment varied greatly by drug, with more recently introduced medications employing longer treatment periods (eg PPA and benzphe- 91.7) 15 (15 100) 37.4 (30 100) 2.4 (1 88.3) 302.9 (190 100) 27.5 (15 100) 72.4 (57 100) 80.1 (39 – – – – – – 100) 75 (75 100) 14.0 (10.0 – 86.9) 123.3 (100 100) 57.4 (32.5 tamine with an average of 7.4 and 8.9 weeks, respectively, – – – – and sibutramine and orlistat averaging 14.5 and 47.5 weeks of treatment). The majority of patients were female, with the proportion of females ranging from 57.6 to 88.5%. Consis- tent with published guidelines, most studies used some form 85.7) 77.7 (63.6 100) 80.7 (41.2 100) 72.1 (9.1 100) 84.9 (63.6 83.7) 73.6 (46.5 100) 85.0 (72.2 100) 83.1 (56.3 100) 86.0 (70 – – – – – – – – 100) 88.3 (70 of lifestyle management program even though these guide- 95) 57.6 (0 100) 64.4 (0 – – – lines were published long after many studies were in print.19 This partially reflects the multidisciplinary nature of obesity treatment and the recognition that successful drug interventions must address eating and activity behaviors in obese 30) 78.8 (71.9 32) 58.3 (0 340) 72.2 (51.2 74) 86.5 (70.6 36) 88.2 (76.5 49) 85.8 (70.2 268) 83.0 (52.9 29) 88.5 (70 68) 71.5 (11.1 30) 84.4 (59.4 136) 64.1 (0 – – – – – – – – – – – patients. With the exception of benzocaine, patients receiving drug therapy, whether or not it was combined with lifestyle modification, experienced greater average weight loss than

657) 164.5 (46 patients in the placebo groups. Unweighted weight loss – 32) 21 (12 33) 22 (13 76) 29.4 (12 52) 26.8 (15 295) 44.1 (5 32) 18 (4 58) 21.2 (6 50) 17.7 (8 138) 55.7 (9 36) 22.4 (10 – – – – differences ranged from 7 0.80 to 3.82 kg, with most drugs – – – – – – demonstrating modest weight losses relative to placebo. It is interesting to note that five drugs (benzocaine, dexfenfluramine, diethylpropion hydrochloride, fenfluramine and 17) 21 (12 60) 55.2 (7 16) 23 (14 76) 236.9 (46.7 mazindol) had studies in which no weight loss or weight – – 56) 46.6 (5 52) 21.2 (5 18) 20 (5 20) 23.5 (8 24) 32 (15 14) 23.5 (8 26) 27.3 (15 – – – – – – – – – gain occurred relative to the placebo groups, as noted by the placebo: post-treatment outcomes Weeks Number at post-test Female (%) Percentage negative values in the drug-placebo value ranges. vs 7.4 (2 17.6 (6 13.2 (2 Figure 1 presents the effect sizes and 95% confidence intervals for drug – placebo comparisons. Confidence intervals for the effect sizes are presented only for those drugs

Single drug with three or more studies in the meta-analysis database. uramine (14) 30.0 (4

ﬂ These results represent the post-test outcome of studies : unless otherwise indicated, numbers in parentheses are the range of the distribution of values. Aggregate data based on studies that presented nece uramine (14) 9.7 (4

ﬂ without consideration of possible design differences such hydrochloride (9) (hydrochloride and resin) amine (PPA) (7) conditions, placebo conditions, and drug minus placebo are unweighted. Negative numbers for kg changes indicate weight gain. Drugs not included in this table did not have studies that met the study inclusionary criteria that also provided suf any treatment, because studies rarely presented data separately for those weeks involving medication use. An asterisk (*) indicates unweighted eff Table 2 Amphetamine (2) 16 (16 Benzocaine (1) 14 7 10 100 100 96 100 0.80 1.60 Benzphetamine (3) 8.9 (1.6 Dexfen Diethylpropion Fen Mazindol (22) 11.0 (2 Fluoxetine (11) 27.5 (6.0 Orlistat (6)Phentermine (6) 47.5 (16 Phenylpropanol- Drug numberof studies) of any treatment Drug Placebo DrugSibutramine (4)Note 14.5 (8 Placeboas (mg) study treatment length, drug drug with placebodose etc. placebo (kg) Outcomes suggest that the

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 266

Figure 1 Effect sizes and 95% conﬁdence intervals of drug — placebo comparisons. Note: high=low lines were not constructed for amphetamine and benzocaine due to insufﬁcient studies ( < 3). Horizontal line at an effect size of zero represents no treatment effect.

various drugs generally produced comparable weight losses. each drug type, the fact that each effect size is based on a However, it could be argued that amphetamine, benzpheta- group of patients rather than a single subject’s outcome and, mine, fenfluramine and sibutramine produced the largest because a correlation coefficient represents an effect size,21 it mean effect sizes among the drugs studied, each of which was a priori decided that a correlation coefficient of 0.30 were in the ‘large’ range of effect size magnitude (ie > 0.80) as would be judged to be a potentially important indicator of is typically defined in meta-analysis.20 Furthermore, fenflur- association. However, traditional inferential tests also are amine and sibutramine produced significantly better weight reported. The correlations between effect size and treatment losses (based on the effect size 95% confidence intervals) length and publication year are presented in Table 3. than five of the other drugs studied, ie benzocaine, dexfen- There were no significant associations between effect size fluramine, fluoxetine, mazindol and orlistat. Nevertheless, and treatment length for any of the seven drugs examined, both drugs overlapped with amphetamine, benzphetamine, although the magnitude of the correlation for phentermine diethylpropion, phentermine and PPA, suggesting that there was large, suggesting that treatment length did influence were no statistically significant differences between these phentermine’s effect size. Table 3 also presents correlations drugs’ effect sizes. between the amount of weight (kg) lost in both treatment and placebo groups for the seven drugs. Overall, longer treatment was associated with greater weight loss in both Effect of length of treatment drug and placebo groups. Treatment length and amount of Not surprisingly, there was a strong correlation between the weight loss (kg) were significantly correlated only for dexfen- year of publication of a study and the length of the study’s fluramine (both drug and placebo) and the placebo groups in treatment (r ¼ 0.436, P < 0.001) among studies providing diethylpropion hydrochloride studies. Patients receiving single drug vs placebo outcomes. However, there was no diethylpropion hydrochloride or orlistat (or those participat- overall relationship between treatment length and effect ing in diethylpropion hydrochloride or orlistat study placebo size (r ¼ 0.016, P ¼ 0.875), suggesting that many drugs may groups) also demonstrated greater weight loss with increased have their greatest impact on weight early in treatment. treatment time, but the correlations were not statistically Seven drugs were judged to have a sufficient number of significant. Patients receiving placebo in both fenfluramine studies and variance in treatment length to be examined in and mazindol studies experienced less weight loss with within-drug analyses (ie dexfenfluramine, diethylpropion increasing treatment time. Thus, with a few exceptions, hydrochloride, fenfluramine, mazindol, orlistat, phenter- increasing treatment length was associated with greater mine and PPA). Given the small number of studies within weight loss among patients receiving drug or placebo.

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 267 Table 3 Relationships among treatment length and year of publication on post-treatment single drug vs placebo outcomes

Length of treatment Year of publication

kg lost in kg lost in kg lost in kg lost in Drug Effect size drug group placebo group Effect size drug group placebo group

All studies 0.016 (0.875) 0.430 ( < 0.001) 0.395 ( < 0.001) 0.055 (0.592) 0.182 (0.080) 0.165 (0.115) Dexfenﬂuramine 7 0.080 (0.787) 0.578 ( 7 0.030) 0.613 (0.020) 7 0.576 (0.031) 7 0.543 (0.036) 7 0.367 (0.179) Diethylpropion hydrochloride 7 0.084 (0.830) 0.555 (0.121) 0.797 (0.010) 0.385 (0.306) 0.049 (0.900) 7 0.303 (0.427) Fenﬂuramine 7 0.057 (0.846) 7 0.066 (0.838) 7 0.366 (0.242) 0.216 (0.459) 0.442 (0.151) 0.253 (0.427) Mazindol 0.254 (0.254) 7 0.139 (0.547) 7 0.408 (0.066) 7 0.093 (0.680) 7 0.061 (0.793) 0.250 (0.275) Orlistat 0.146 (0.782) 0.456 (0.441) 0.809 (0.097) 0.017 (0.975) 0.623 (0.262) 0.879 (0.050) Phentermine 0.685 (0.202) 0.035 (0.956) 7 0.420 (0.482) 0.141 (0.790) 7 0.346 (0.502) 7 0.115 (0.828) Phenylpropanolamine (PPA) 7 0.132 (0.778) 0.210 (0.651) 0.291 (0.527) 7 0.091 (0.845) 0.171 (0.713) 0.169 (0.717)

Note: values in cells represent correlation coefﬁcient and its associated P-value (in parentheses). Only drugs with a sufﬁcient number of primary studies were included in this table.

Effect of publication year Single drug vs placebo: follow-up outcomes Overall, there was no relationship between year of publica- Nineteen studies provided data where (1) the research design tion and effect size, suggesting that the placebo-controlled remained intact and (2) data were presented for a follow-up outcomes of drug studies are not changing over time assessment period after the formal study was complete. (r ¼ 0.055, P ¼ 0.592). Most of the seven selected drugs Lengths of the follow-ups ranged from 1 to 136 weeks demonstrated no or positive, but statistically insignificant, (mean ¼ 17.3; median 6.0). The relationship between relationships between publication year and effect size except length of follow-up and effect size was not significant for dexfenfluramine. The correlation between dexfenflura- (P ¼ 0.066), although the magnitude of the correlation was mine’s effect size and year of publication was strong and moderate (r ¼ 7 0.430). Table 4 presents follow-up effect negative, suggesting that effect size in dexfenfluramine stu- sizes from these randomized clinical trials. dies decreased over time. This was also demonstrated in the All studies discontinued drug treatment during the negative correlation between publication year and the follow-up period. In addition, most studies did not provide amount of weight (kg) lost in patients taking dexfenflura- booster sessions during this time. Thus, the follow-up peri- mine, suggesting that later dexfenfluramine studies that met ods for the majority of studies examined represent post- our inclusion criteria demonstrated less weight loss than treatment observation periods without any extension of earlier studies. While the same trend was noted in patients the treatments, not reflecting the current long-term treat- receiving placebos in dexfenfluramine studies, the correla- ment paradigm in obesity management. With the exception tion was not statistically significant. of amphetamine and mazindol, all drugs with follow-up

Table 4 Single drug vs placebo: follow-up outcomes

Providing booster Number at follow-up sessions (%) Drug Drug (number continued kg lost kg lost Drug 7 Effect size of studies) Drug Placebo Drug Placebo (%) with drug with placebo placebo (kg) (95% CI)

Amphetamine (2) 19.5 18.5 00 0 7 3.45 7 2.5 7 0.95 0.225* (19 – 20) (16 – 21) Dexfenﬂuramine (9) 33.6 33.0 11 11 0 6.04 4.99 1.05 0.401 (5 – 116) (5 – 108) ( 7 1.5 – 11.7) (1.9 – 8.5) ( 7 3.6 – 3.5) (0.238 – 0.564) Mazindol (4) 18.3 17.8 00 0 7 0.28 7 0.85 0.58 0.674 (10 – 23) (10 – 24) ( 7 6.1 – 2.3) ( 7 5.5 – 1.1) ( 7 0.60 – 1.8) (0.329 – 1.019) Phentermine (2) 16.3 18.0 0 0 0 8.23a 5.8a 2.43a 0.810* (8 – 25) (9 – 27) Sibutramine (2) 31.3 32.5 50 50 0 4.87 2.5 2.37 1.05* (15 – 48) (15 – 50) (3.3 – 6.4) (1.7 – 3.3) (1.63 – 3.1)

Note: unless otherwise indicated, numbers in parentheses are the range of the distribution of values. Aggregate data based on studies that presented necessary information. An asterisk indicates unweighted effect sizes (computed for drugs with less than three studies). Drugs not included in this table did not have studies that met the study inclusionary criteria that also provided sufﬁcient data to compute drug 7 placebo effect sizes. Data presented for kg lost for drug conditions, placebo conditions, and drug minus placebo are unweighted and represent the follow-up weight of participants minus their baseline weight. Negative numbers for kg changes indicate weight gain. Numbers in parentheses represent the number of studies contributing data to a particular row; studies may contribute data to more than one row. aWeight losses for individual groups only reported in one of the two phentermine studies.

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 268 periods demonstrated sustained weight loss. Phentermine exception of diethylpropion vs PPA – caffeine and mazindol and sibutramine maintained fairly large placebo subtracted vs PPA – caffeine. No general patterns of effectiveness weight losses (ie 2.43 and 2.37 kg, respectively) and had the emerged based on drug pharmacology, eg PPA – caffeine, an largest effect sizes, ranging from 0.810 to 1.05. Dexfenflur- OTC preparation, was consistently less effective than pre- amine had a smaller placebo-subtracted weight loss and a scription drugs it was compared with, eg mazindol and more modest effect size, but results are based on a larger diethylpropion. In contrast, ephedrine – caffeine, another number of studies than the effect sizes for phentermine and OTC, performed moderately better than dexfenfluramine. sibutramine. Given the small number of studies providing Most drug – drug comparison trials were relatively short follow-up assessments, the effects of moderators of outcome duration, ranging from 2 to 15 weeks. Overall, mazindol were not explored. was compared to other drugs more often than any of the others that met our study inclusion criteria. Mazindol tended to demonstrate greater weight losses than the drugs Combination drug vs placebo: post-treatment outcomes it was compared with, with effect sizes in the medium range Six studies provided comparisons of combination drugs vs in most cases. placebo and met study inclusionary criteria. None of the combination drug trials included assessments that met our definition of a follow-up. Outcomes for the combination Discussion drugs are presented in Table 5. We meta-analyzed published, randomized, controlled trials Length of treatment for combination drug trials ranged of obesity pharmacotherapies identified by the National Task from 6 to 32 weeks and the majority of patients were women. Force on the Prevention and Treatment of Obesity8 as being All patients in these trials received some form of lifestyle currently approved, those that were used off-label, recently modification (ie modification of diet and=or physical activ- approved drugs that were not reviewed by the Task Force,8 ity). All drug combinations produced placebo-subtracted and recently removed drugs that were previously approved weight loss, ranging from 0.30 kg for PPA – benzocaine to by the FDA for obesity management. Overall, the studied 9.60 kg for fenfluramine – phentermine. Effect sizes for phen- drugs produced medium effect sizes with only four drugs termine – fenfluramine (1.48), PPA – caffeine (2.58), and having effect sizes greater than 0.80 (amphetamine, benz- methamphetamine – phenobarbatol (5.04) were all very phetamine, fenfluramine and sibutramine) and only one large. It also should be noted that, with one exception exceeding 0.90 (sibutramine). As noted in Table 2, the (PPA – benzocaine), all combination drugs produced larger absolute placebo-subtracted weight losses associated with effect sizes than any single drug except fenfluramine and studies of single drugs included in the meta-analysis never sibutramine (see Figure 1). However, these effect sizes are exceeded 4.0 kg. Thus, the incremental benefit of obesity based on a much smaller number of studies (n ¼ 1 for most) drug treatments, in addition to lifestyle interventions, and some of the combinations are now viewed as having appears to be modest. It is interesting that there was no significant adverse health impacts (eg fenfluramine – drug, or class of drugs, that demonstrated clear superiority. phentermine, methamphetamine – phenobarbatol). As can be seen in Figure 1, many of the 95% confidence intervals for the studied drugs overlapped, indicating that the differences in effect sizes among many of the drugs were Drug – drug comparisons not statistically significant. For example, sibutramine, the There were few drug – drug comparisons to examine (ie 18 in drug with the largest effect size, had overlapping confidence total). Table 6 summarizes the data, including effect sizes, for intervals with benzphetamine, diethylpropion, fenflura- all drug – drug comparisons. Most effect sizes for the drug – mine, phentermine and PPA, and the effect size for amphe- drug comparisons were in the medium range with the tamine exceeded its lower boundary. The only clearly

Table 5 Combination drug vs placebo: post-treatment outcomes

Drug combination components Number at Percentage post-test Female (%) using kg lost First drug Weeks of lifestyle kg lost with Effect (number of studies) Second drug Third drug any treatment Drug Placebo Drug Placebo treatment with drug placebo size

PPA (2) Caffeine 6.0 27 25.5 83.6 73.1 100 2.4 1.4 2.58 (6.0 – 6.0) (26 – 28) (23 – 28) (78.6 – 88.5) (67.9 – 78.3) (2.1 – 2.7) (0.9 – 1.9) PPA (1) Benzocaine 14 10 10 100 100 100 1.9 1.6 0.136 Fenﬂuramine (1) Phentermine 32 58 54 75.8 72.9 100 14.2 4.6 1.481 Methamphetamine (1) Phenobarbatol 20 7 5 ——100 2.4 0.5 5.043 Amphetamine (1) Laevoamphetamine Meth-aqual-one 8 20 20 ——100 3.3 0.2 0.873

Note: PPA, phenylpropanolamine. None of the effect sizes are presented with conﬁdence intervals due to the small number of studies providing tests of combination drug treatments. Blank cells represent factors that were not presented in the primary study.

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 269 Table 6 Drug – drug comparisons: post-treatment outcomes

Number at Percentage post-test Female (%) using First drug Weeks of lifestyle kg lost kg lost (number of studies) Second drug any treatment Drug 1 Drug 2 Drug 1 Drug 2 treatment with drug 1 with drug 2 Effect size

Amphetamine (1) Amphetamine – synthroid 12 53 48 96.2 97.9 100 5.2 6.7 7 0.519 Benzocaine (1) Benzocaine – PPA 14 7 10 100 100 100 0.80 1.9 7 0.511 Diethylpropion (1) PPA – caffeine 8 23 25 95.7 96 100 3.6 3.1 0.872 Dexfenfluramine (1) Ephedrine – caffeine 15 43 38 83 78 100 6.9 8.3 7 0.292 Fenfluramine (1) Metformin 8 10 10 100 100 100 7.7 5.2 0.772 Fluoxetine (1) Benzphetamine 8 32 33 88 80 100 3.7 3.2 0.209 Mazindol (9) All other drugs 9 18.9 19.7 85.2 87.9 78 5.47 4.3 0.257 (2 – 16) (5 – 31) (4 – 32) (59.4 – 100) (71.9 – 100) (2.9 – 9.0) (1.6 – 8.1) (0.043 – 0.471) (2) Amphetamine 12 17.5 18 59.4a 71.9a 50 4.9 3.75 0.096 (8 – 16) (5 – 30) (4 – 32) (3.4 – 6.4) (1.6 – 5.9) (3) Diethylpropion 11.7 17.3 19.3 96.2a 90.9a 100 6.7 5.1 0.310 (11 – 12) (2.9 – 9.0) (2.0 – 8.0) ( 7 0.073 – 0.693) (3) Phenmetrazine 5.3 18.3 18 91.1 91.9 66.7 5.1 4.2 0.034 (2 – 8) (80 – 100) (85.7 – 100) (3.5 – 6.7) (2.9 – 5.7) (0.342 – 0.410) (1) Phentermine 2 15 17 100 100 100 6.7 5.5 0.114 (1) PPA – caffeine 6 28 27 82.1 88.9 100 4.1 3.7 0.870 Phentermine (2) All Other Drugs 7 33 32 82 84.7 100 7.1 5.95 0.446 (2 – 14) (16 – 50) (15 – 49) (64 – 100) (69.4 – 100) (6.1 – 8.3) (5.6 – 6.3) (1) Phenmetrazine 2 16 15 64 69.4 100 6.1 5.6 0.050 (1) Diethylpropion 12 50 49 64 69.4 100 8.3 6.3 0.574 Sibutramine (1) Dexfenfluramine 12 112 112 90.2 93 100 4.5 3.2 0.347

Notes: PPA, phenylpropanolamine. There were nine total studies comparing Mazindol to other drugs. Individual comparisons of Mazindol to a particular drugaddto 10 because one study contributed to more than one group of these comparisons. aOnly one study reported data for this factor. ineffective drug was benzocaine, with a negative effect size digm emphasizes that obesity is a chronic disease requiring of 7 0.35 and a placebo-subtracted weight loss of 7 0.80 kg. sustained treatment.22 Thus, having drug-free follow-ups in However, the outcome of benzocaine was based on one obesity treatment should be akin to expecting sustained anti- randomized clinical trial. hypertensive effects after drug discontinuation in hyperten- Another surprising result was that treatment length and sion patients, ie there is no more reason to expect that year of publication did not influence effect size over all drugs obesity medications will have significant prolonged effects studied. While longer treatments were associated with than would be expected with any other chronic disease greater weight loss, this was true for both drug and placebo pharmacotherapy. Given this new treatment paradigm, obe- groups and the relationships were roughly equivalent (ie sity pharmacotherapy studies may need to shift from the r ¼ 0.430 and 0.395, respectively), suggesting that this idea of using drug-free follow-up periods to studying long- effect was independent of drug treatments. When examining term continuous or intermittent drug administration. Never- individual drugs, it was striking to note that several demon- theless, some drugs continued to provide important weight strated negative relationships between treatment length and loss maintenance during drug free follow-up periods. For effect size (eg dexfenfluramine, diethylpropion, fenfluramine example, both phentermine and sibutramine demonstrated and PPA), suggesting that their effectiveness decreased with large effect sizes and modest placebo-subtracted weight increasing treatment time, although these correlations were loss (ie 2.43 and 2.37 kg, respectively) even though small and not statistically significant. While there was no pharmacotherapy had been discontinued during the overall relationship between publication year and effect size, follow-up period and PPA provided sustained weight loss it is interesting to note that there was a statistically signifi- with continued administration. cant negative correlation between publication year and With the exception of PPA – benzocaine, drug combina- effect size for dexfenfluramine, indicating that treatment tions appeared to be the most potent weight loss agents, effectiveness (and weight loss in drug treatment groups) producing large effect sizes ranging from 0.873 to 5.043 and decreased over the last decade of published studies. Mazindol placebo-subtracted weight losses ranged from 1.0 to 9.6 kg. and PPA also had negative, but small and statistically insig- Unfortunately, several of the constituent drugs in these nificant associations between effect size and publication combinations have demonstrated substantial negative side- year. effects and have been removed from the market or are not Few drug studies provided follow-up outcome data and available (eg amphetamine, fenfluramine and methamphe- those that did discontinued pharmacotherapy during this tamine) or have been or are under consideration for removal time. This is notable given that the current treatment para- (eg PPA, ephedra).23 – 27

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 270 Prior to conducting this meta-analysis, several moderator dock and a minority scientist development grant from the variables were hypothesized to affect the outcomes of the American Heart Association, awarded to Dr Poston. clinical trials but were not included in this review. For instance, the effects of drug side effects and attrition due References to side effects was a primary interest of the meta-analysis 1 Lijesen GK, Theeuwen I, Assendelft WF, Van Der Wal G. The team. However, few studies provided useable data on either effect of human chorionic gonadotropin (HCG) in the treatment patterns of or attrition due to side effects. Further, many of obesity by means of the Simeons therapy: a criteria-based meta-analysis. Br J Clin Pharmac 1995; 40: 237 – 243. studies provided only scant description of controls used to 2 Van Gaal LF, Wauters MA, Peiffer FW, De Leeuw IH. Sibutramine ensure the internal validity of the trial or codeable descrip- and fat distribution: is there a role for pharmacotherapy in tions of lifestyle components. As with all meta-analyses, we abdominal=visceral fat reduction? Int J Obes Relat Metab Disord were limited to coding factors that are presented in primary 1998; 22(Suppl): S38 – S40. 3 Greenway FL. Clinical studies with phenylpropanolamine: a studies. Therefore, standards for reporting important para- meta-analysis. Am J Clin Nutr 1992; 55: 203S – 205S. meters in obesity drug clinical trials are needed to increase 4 Scoville BA. Review of amphetamine like drugs by the Federal the utility of future reviews of this literature. Drug Administration: clinical data and value judgments. In: G Bray (ed). Obesity in perspective, DHEW publication no. (NIH) The results of this meta-analysis have several important 75 – 708. US Government Printing Office: Washington, DC; implications for obesity pharmacotherapy. First, increasing 1973. length of drug treatment does not lead to more weight loss; 5 Weintraub M. Phenylpropanolamine as an anorexiant agent thus, longer treatments appear to promote weight mainte- weight control: a review of published and unpublished studies. In: JD Morgan, DV Kagan, JS Brody (eds). Phenylpropanolamine: nance, but further weight loss beyond the typical plateau at risks, benefits, and controversies. Praeger: New York; 1985. pp 53 – 6 months is unlikely. In addition, the amount of weight lost 79. above and beyond that achieved in placebo treatments, most 6 Committee on Nutrition, Massachusetts Medical Society. Obesity of which included some form of lifestyle management (ie treatment using drug therapy. White paper. The Massachusetts Medical Society: Waltham, MA; 1998. diet, exercise, or both) is typically modest (ie usually greater 7 Goldstein DJ, Potvin JH. Long-term weight loss: the effect of than 2 kg) and never exceeded 4 kg in the single drug vs pharmacological agents. Am J Clin Nutr 1994; 60: 647 – 657. placebo comparisons. Some may argue that this is a very 8 National Task Force on the Prevention and Treatment of Obesity. small incremental improvement given the costs and risks Long-term pharmacotherapy in the management of obesity. JAMA 1996; 276: 1907 – 1915. associated with drug therapies; however, this amount of 9 Dietary Supplement Health and Education Act of 1994 (DSHEA). weight is important (ie nearly or exceeding 1 BMI unit) Public Law 103 – 417: 103rd Congress, 2nd Session Senate, and drug treatments often can be more accessible and Report 103 – 410 Washington, DC; pp, 1 – 49. easier to use than structured lifestyle modification programs 10 Apfelbaum M, Vague P, Ziegler O, Hanotin C, Thomas F, Leute- negger E. Long-term maintenance of weight loss after a very- that typically are based in obesity treatment centers. In low-calorie diet: a randomized blinded trial of the efficacy and addition, obesity medications can result in important reduc- tolerability of sibutramine. Am J Med 1999; 106:179– 184. tions in overall medication use and net costs associated with 11 Wadden TA, Bartlett SJ, Foster GD, Greenstein RA, Wingate BJ, obesity-related comorbidites.27 Stunkard AJ, Letizia KA. Sertraline and relapse prevention training following treatment by very-low-calorie diet: a controlled Finally, more recent studies that did not meet our inclu- clinical trial. Obes Res 1995; 3: 549 – 557. 10 – 12 sion criteria and were not included in this meta-analysis 12 James WPT, Astrup A, Finer N, Hilsted J, Kopelman P, Rossner S, suggest that pharmacotherapy may be particularly helpful in Saris WHM, Van Gaal LF. Effects of sibutramine on weight promoting long-term weight maintenance. For example, maintenance after weight loss: a randomized trial. Lancet 2000; 356: 2119 – 2125. 75% of patients taking sibutramine following a 4-week very 13 Rosenthal R. Parametric measures of effect size. In: H Cooper, LV low calorie diet (VLCD) maintained 100% of their initial Hedges (eds). The handbook of research synthesis. Russell Sage: weight loss 1 y after completing the VLCD, compared with New York; 1994. pp 231 – 244. only 42% of placebo-treated patients.10 Similarly, James et 14 Lipsey MW, Wilson DB. The efficacy of psychological, educa- 12 tional, and behavioral treatment: confirmation from meta-ana- al found that patients receiving sibutramine for an 18- lysis. Am Psychol 1993; 48: 1181 – 1209. month maintenance phase, following a 6-month 600 kcal 15 Shadish, WR, Haddock, CK. Combining estimates of effect size. diet þ sibutramine weight loss segment, maintained a greater In: H Cooper, LV Hedges (eds). The handbook of research synthesis. amount of their initial weight loss when compared to Russell Sage: New York; 1994. pp 261 – 281. 16 Hedges LV. Estimation of effect size from a series of independent patients randomized to receive a placebo. Thus, 43% of the experiments. Psychol Bull 1982; 92:490– 499. drug-treated patients maintained 80% of their initial weight 17 Hedges LV, Olkin I. Statistical methods for meta-analysis. Academic loss as compared to only 16% of the placebo-treated patients. Press: San Diego, CA; 1985. Future studies should examine the effectiveness of drug 18 Cooper H, Hedges LV. The handbook of research synthesis (Part VI). Russell Sage: New York; 1994. therapies as long-term weight maintenance agents. 19 Cohen J. A power primer. Psychol Bull 1992; 112: 155 – 159. 20 National Institutes of Health (NIH), National Heart, Lung, and Blood Institute (NHLBI). Clinical guidelines on the identification, evaluation, and treatment of overweight and obesity: the evidence Acknowledgements report. US Government Press: Washington DC; 1998. This paper was supported by a faculty research grant from 21 Ozer DJ. Correlation and the coefficient of determination. the University of Missouri-Kansas City awarded to Dr Had- Psychol Bull 1985; 97: 307 – 315.

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 271 22 Kopelman PG. Obesity as a medical problem. Nature 2000; 404: 15 Brightwell DR, Naylor CS. Effects of a combined behavioral and 635 – 643. pharmacologic program on weight loss. Int J Obes Relat Metab 23 Kernan WN, Viscoli CM, Brass LM, Broderick JP, Brott T, Feld- Disord 1979; 3: 141 – 148. mann E, Mogenstern LB, Wilterdink JL, Horwitz RI. Phenylpro- 16 Brodbin P, O’Connor CA. A double-blind clinical trial of an nanolamine and the risk of hemorrhagic stroke. New Engl J Med appetite depressant, fenfluramine, in general practice. Practi- 2000; 343: 1826 – 1832. tioner 1967; 198: 707 – 710. 24 Kolanowski J. A risk – benefit assessment of anti-obesity drugs. 17 Brun LD, Bielmann P, Gagne C, Moorjani S, Nadeau A, Lupien, Drug Safety 1999; 20:119– 131. PJ. Effects of fenfluramine in hypertriglyceridemic obese sub- 25 Scheen AJ, Lefebvre PJ. Pharmacological treatment of obesity: jects. Int J Obes Relat Metab Disord 1988; 12:423– 431. present status. Int J Obes Relat Metab Disord 1999; 23:47– 53. 18 Campbell CJ, Bhalla IP, Steel JM, Duncan LJP. A controlled trial 26 Haller CA, Benowitz NL. Adverse cardiovascular and central of phentermine in obese diabetic patients. Practitioner 1977; 218: nervous system events associated with dietary supplements 851 – 855. containing ephedra alkaloids. New Engl J Med 2000; 343: 19 Connolly VM, Gallagher A, Kesson CM. A study of fluoxetine in 1833 – 1838. obese elderly patients with type 2 diabetes. Diabetic Med 1995; 27 Greenway FL, Ryan DH, Bray GA, Rood JC, Tucker EW, Smith SR. 12: 416 – 418. Pharmaceutical cost savings of treating obesity with weight loss 20 *Conte A. Evaluation of Sanorex-a new appetite suppressant. medications. Obes Res 1999; 7: 523 – 531. J Obes Bariat Med 1973; 2: 104 – 107. 21 *Conte A. Evaluation of Sanorex-a new appetite suppressant. J Obes Bariat Med 1973; 2: 104 – 107. 22 *Conte A. Evaluation of Sanorex-a new appetite suppressant. Appendix 1: Studies included in the meta analysis J Obes Bariat Med 1973; 2: 104 – 107. References 23 Crommelin RM. Nonamphetamine, anorectic medication for 1 Abramson R, Garg M, Cioffari A, Rotman PA. An evaluation of obese diabetic patients: controlled and open investigations of behavioral techniques reinforced with an anoretic drug in a mazindol. Clin Med 1974; 81:20– 24. double-blind weight loss study. J Clin Psychiatry 1980; 41: 24 Dahms WT, Molitch ME, Bray GA, Greenway FL, Atkinson RL, 234 – 237. Hamilton K. Treatment of obesity: cost – benefit assessment of 2 Alger S, Larson K, Boyce VL, Seagle H, Fontvielle A, Ferraro R, behavioral therapy, placebo, and two anorectic drugs. Am J Clin Rising R, Ravussin E. Effect of phenylpropanolamine on energy Nutr 1978; 31: 774 – 778. expenditure and weight loss in overweight women. Am J Clin 25 Davidson MH, Hauptman J, DiGirolamo M, Foreyt JP, Halsted Nutr 1993; 57: 120 – 126. CH, Heber D, Heimburger DC, Lucas CP, Robbins DC, Chung J, 3 Allen GS. A double-blind clinical trial of diethylpropion hydro- Heymsfield SB. Weight control and risk factor reduction in obese chloride, mazindol, and placebo in the treatment of exogenous subjects treated for 2 y with orlistat: a randomized controlled obesity. Curr Ther Res 1997; 22: 678 – 685. trial. JAMA 1999; 281: 235 – 242. 4 Altschuler S, Conte A, Sebok M, Marlin R, Winick C. Three 26 DeFelice EA, Chaykin LB, Cohen A. Double-blind clinical eva- controlled trials of weight loss with phenylpropanolamine. Int luation of mazindol, dextroamphetamine and placebo in treat- J Obes Relat Metab Disord 1982; 6: 549 – 556. ment of exogenous obesity. Curr Ther Res 1973; 15: 358 – 366. 5 Altschuler S, Frazer DL. Double-blind clinical evaluation of the 27 DeFelice E, Bronstein S, Cohen A. Double-blind comparison of anorectic activity of phenylpropanolamine hydrochloride drops placebo and 42 – 548, a new appetite suppressant, in obese and placebo drops in the treatment of exogenous obesity. Curr volunteers. Curr Ther Res 1969; 11: 256 – 262. Ther Res 1986; 40: 211 – 217. 28 Drent ML, Larsson I, William-Olsson T, Quaade F, Czubayko F, 6 Atkinson RL, Greenway FL, Bray GA, Dahms WT, Molitch ME, von Bergmann K, Strobel W, Sjostrom L, van der Veen EA. Hamilton K, Rodin J. Treatment of obesity: comparison of Orlistat (RO 18-0647), a lipase inhibitor, in the treatment of physician and nonphysician therapists using placebo and anor- human obesity: a multiple dose study. Int J Obes Relat Metab ectic drugs in a double-blind trial. Int J Obes Relat Metab Disord Disord 1995; 19: 221 – 226. 1977; 1: 113 – 120. 29 Elliott BW. A collaborative investigation of fenfluramine: anor- 7 Bacon GE, Lowery GH. A clinical trial of fenfluramine in obese exigenic with sedative properties. Curr Ther Res 1970; 12: 502 – children. Curr Ther Res 1967; 9:626– 630. 515. 8 Baird IM, Howard AN. A double-blind trial of mazindol using a 30 Elmaleh, MK, Miller, J. Controlled clinical evaluation of a new very low calorie formula diet. Int J Obes Relat Metab Disord 1977; anorectic agent in obese adults. Pa Med 1974; 77:46– 50. 1: 271 – 278. 31 Enzi G, Baritussio A, Marchiori E, Crepaldi G. Short-term and 9 Bandisode MS, Boshell BR. Double blind clinical evaluation of long-term clinical evaluation of a non-amphetaminic anorex- mazindol (42 – 548) in obese diabetics. Curr Ther Res 1975; 18: iant (mazindol) in the treatment of obesity. J Int Med Res 1976; 4: 816 – 824. 305 – 318. 10 Bolding OT. Diethylpropion hydrochloride: an effective appetite 32 Enzi G, Crepaldi G, Inelmen EM, Bruni R, Baggio B. Efficacy and suppressant. Curr Ther Res 1974; 16:40– 48. safety of dexfenfluramine in obese patients: a multi-center 11 Bradley MH, Raines J. The effects of phenylpropanolamine study. Clin Neuropharmac 1988; 11(Suppl): S173 – S178. hydrochloride in overweight patients with controlled stable 33 Ferguson JM, Feighner JP. Fluoxetine induced weight loss in hypertension. Curr Ther Res 1989; 46:74– 84. overweight non-depressed humans. Int J Obes Relat Metab Disord 12 Bray GA, Ryan DH, Gordon D, Heidingsfelder S, Cerise F, Wilson 1987; 11: 163 – 170. K. A double-blind randomized placebo-controlled trial of sibu- 34 Finer N, Finer S, Naoumova RP. Prolonged use of a very low tramine. Obes Res 1996; 4: 263 – 270. calorie diet (Cambridge diet) in massively obese patients attend- 13 Breum L, Astrup A, Andersen T, Lammert O, Nielsen E, Garby L, ing an obesity clinic: safety, efficacy, and additional benefit from Quaade F. The effect of long-term dexfenfluramine treatment on dexfenfluramine. Int J Obes Relat Metab Disord 1989; 13:91– 93. 24-hour energy expenditure in man: a double-blind placebo 35 Galloway DB, Logie AW, Petrie JC. Prolonged action fenflura- controlled study. Int J Obes Relat Metab Disord 1990; 14: 613 – mine in nondiabetic patients with refractory obesity. Postgrad 621. Med J 1975; 51:155– 157. 14 Breum L, Pedersen JK, Ahlstrom F, Frimodt-Moller J. Comparison of an ephedrine=caffeine combination and dexfenfluramine in the treatment of obesity: a double-blind multicentre trial in general practice. IntJObesRelatMetabDisord1994; 18:99– 103. *Represents three independent studies.

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 272 36 Goldrick RB, Hevnstein N, Whyte HM. Effects of caloric restriction 57 {Kornhaber A. Obesity-depression: clinical evaluation with a and fenfluramine on weight loss and personality profiles of pati- new anorexigenic agent. Psychosomatics 1973; 14: 162 – 167. ents with long-standing obesity. Austr NZ J Med 1973; 3:131– 141. 58 Kutnowski M, Daubresse J, Friedman H, Kolanowski J, Krzen- 37 Goldstein DJ, Rampey AH, Potvin JH, Fludzinski LA. Fluoxetine towski G, Scheen A, van Gaal L. Fluoxetine therapy in obese in obese patients with noninsulin-dependent diabetes mellitus. diabetic and glucose intolerant patients. Int J Obes Relat Metab Clin Res 1992; 40: 240A (abstract). Disord 1992; 16(Suppl): S63 – S66. 38 {Holman SL, Goldstein DJ, Enas GG. Pattern analysis method 59 Lafreniere F, Lambert J, Rasio E, Serri O. Effect of dexfenflur- for assessing successful weight reduction. Int J Obes Relat Metab amine treatment on body weight and postprandial thermogen- Disord 1994; 18: 281 – 285. esis in obese patients: a double-blind placebo-controlled study. 39 Goldstein DJ, Rampey AH Jr, Enas GG, Potvin JH, Fludzinski LA, Int J Obes Relat Metab Disord 1993; 17:25– 30. Levine LR. Fluoxetine: a randomized clinical trial in the 60 Langlois KJ, Forbes JA, Bell GW, Grant GF Jr. A double-blind treatment of obesity. Int J Obes Relat Metab Disord 1994; 18: clinical evaluation of the safety and efficacy of phentermine 129 – 135. hydrochloride (Fastin) in the treatment of exogenous obesity. 40 Gray DS, Fujioka K, Devine W, Bray GA. A randomized double- Curr Ther Res 1974; 16: 289 – 296. blind clinical trial of fluoxetine in obese diabetics. Int J Obes 61 Lawson AAH, Roscoe P, Strong JA, Gibson A, Peattie P. Compar- Relat Metab Disord 1992; 16:S67– S72. ison of fenfluramine and metformin in the treatment of obesity. 41 {Greenway F, Herber D, Raum W, Morales S. Double-blind, Lancet 1970; i: 437 – 441. randomized, placebo-controlled clinical trials with non-pre- 62 Levine LR, Rosenblatt S, Bosomworth J. Use of a serotonin re- scription medications for the treatment of obesity. Obes Res uptake inhibitor, fluoxetine, in the treatment of obesity. Int J 1999; 7: 370 – 378. Obes Relat Metab Disord 1987; 11: 185 – 190. 42 {Greenway F, Herber D, Raum W, Morales S. Double-blind, 63 Levine LR, Enas GG, Thompson WL, Byyny RL, Dauer, AD, Kirby randomized, placebo-controlled clinical trials with non-pre- RW, Kreindler TG, Levy B, Lucas CP, Mcllwain HH, Nelson EB. scription medications for the treatment of obesity. Obes Res Use of fluoxetine, a selective serotonin-uptake inhibitor, in the 1999; 7: 370 – 378. treatment of obesity: a dose – response study. Int J Obes Relat 43 Guy-Grand B, Apfelbaum M, Crepaldi G, Gries A, Lefebvre P, Metab Disord 1989; 13: 635 – 645. Turner P. International trial of long-term dexfenfluramine in 64 Lucas CP, Sandage BW. Treatment of obese patients with dexfen- obesity. Lancet 1989; 2: 1142 – 1144. fluramine: a multicenter, placebo-controlled study. Am J Ther 44 {Pfohl M, Luft D, Blomberg I, Schmulling R-M. Long-term 1995; 2: 962 – 967. changes of body weight and cardiovascular risk factors after 65 Marbury TC, Angelo JE, Gulley RM, Krosnick A, Sugimoto DH, weight reduction with group therapy and dexfenfluramine. Int Zellner SR. A placebo-controlled dose-response study of dexfen- J Obes Relat Metab Disord 1994; 18: 391 – 395. fluramine in the treatment of obese patients. Curr Ther Res 1996; 45 Hanotin C, Thomas F, Jones SP, Leutenegger E, Drouin P. Efficacy 57:663– 674. and tolerability of sibutramine in obese patients: a dose-ranging 66 Mathus-Vliegen EMH, van de Voorde K, Kok AME, Res AMA. study. Int J Obes Relat Metab Disord 1998; 22:32– 38. Dexfenfluramine in the treatment of severe obesity: a placebo- 46 Hanotin C, Thomas F, Jones SP, Leutenegger E, Drouin P. A controlled investigation of the effects on weight loss, cardiovas- comparison of sibutramine and dexfenfluramine in the treat- cular risk factors, food intake, and eating behavior. J Intern Med ment of obesity. Obes Res 1998; 6: 285 – 291. 1992; 232: 119 – 127. 47 Heber KR. Double-blind trial of mazindol in overweight 67 Mathus-Vliegen LMH, Res AMA. Dexfenfluramine influences patients. Med J Aust 1975; 2: 566 – 567. dietary compliance and eating behavior, but dietary instruction 48 Hill JO, Hauptman J, Anderson JW, Fujioka K, O’Neil PM, Smith may overrule its effect on food selection in obese subjects. JAm DK, Zavoral JH, Aronne LJ. Orlistat, a lipase-inhibitor, for weight Diet Assoc 1993; 93: 1163 – 1165. maintenance after conventional dieting: a 1-y study. Am J Clin 68 Mathus-Vliegen EMH. Prolonged surveillance of dexfenflura- Nutr 1999; 69: 1108 – 1116. mine in severe obesity. Neth J Med 1993; 43: 246 – 253. 49 Hoebel BG, Krauss IK, Cooper J, Willard D. Body weight 69 McKay RHG. Long-term use of diethylpropion in obesity. Curr decreased in humans by phenylpropanolamine taken before Med Res Opin 1973; 1: 489 – 493. meals. J Obes Bariat Med 1975; 4: 200 – 206. 70 McQuarrie HG. Clinical assessment of the use of an anorectic 50 Holdaway IM, Wallace E, Westbrooke L, Gamble G. Effect of drug in a total weight reduction program. Curr Ther Res 1975; 17: dexfenfluramine on body weight, blood pressure, insulin resis- 437 – 443. tance, and serum cholesterol in obese individuals. Int J Obes 71 Miach PJ, Thomson W, Doyle AE, Louis WJ. Double-blind cross- Relat Metab Disord 1995; 19: 749 – 751. over evaluation of mazindol in the treatment of obese hyper- 51 Hollander PA, Elbein SC, Hirsch IB, Kelley D, McGill J, Taylor T, tensive patients. Med J Aust 1976; 2: 378 – 380. Weiss SR, Crockett SE, Kaplan RA, Comstock J, Lucas CP, Lodewick 72 Munro JF, Seaton DA, Duncan LJP. Treatment of refractory PA, Canovatchel W, Chung J, Hauptman J. Role of orlistat in the obesity with fenfluramine. Br Med J 1966; 2: 624 – 625. treatment of obese patients with type 2 diabetes: a 1-year rando- 73 Murphy JE, Donald JF, Molla AL, Crowder D. A comparison of mized double-blind study. Diabetes Care 1998; 21:1288– 1294. mazindol (Teronac) with diethylpropion in the treatment of 52 Hooper ACB. Comparison of fenfluramine (with ad libitum food exogenous obesity. J Int Med Res 1975; 3: 202 – 206. intake) with 1000 calorie diet in obesity. J Irish Med Assoc 1972; 74 Noble RE. A controlled study of a weight reduction regimen. 65:35– 37. Curr Ther Res 1971; 13: 685 – 691. 53 Johnson WG, Hughes JR. Mazindol: its efficacy and mode 75 Nolan GR. Use of an anorexic drug in a total weight reduction of action in generating weight loss. Addict Behav 1979; 4: 237 – program in private practice. Curr Ther Res 1975; 18: 332 – 337. 244. 76 O’Connor HT, Richman RM, Steinbeck KS, Caterson ID. Dexfen- 54 Kaplan NM, Jose A. Thyroid as an adjuvant to amphetamine fluramine treatment of obesity: a double blind trial with post therapy of obesity: a controlled double-blind study. Am J Med Sci trial follow-up. Int J Obes Relat Metab Disord 1995; 19: 181 – 189. 1970; 260: 105 – 111. 77 O’Kane M, Wiles PG, Wales JK. Fluoxetine in the treatment of 55 Kolanowski J, Younis LT, Vanbutsele R, Detry JM. Effect of obese type 2 diabetic patients. Diabetic Med 1994; 11: 105 – 110. dexfenfluramine treatment on body weight, blood pressure Eur and noradrenergic activity in obese hypertensive patients. {Represents data from a single study published in multiple J Clin Pharmac 1992; 42: 599 – 606. 56 {Kornhaber A. Obesity-depression: clinical evaluation with a articles. new anorexigenic agent. Psychosomatics 1973; 14: 162 – 167. {Represents two independent studies.

International Journal of Obesity Obesity medication meta-analysis CK Haddock et al 273 78 Oster HL, Medlar RE. A clinical pharmacologic study of benz- 99 Thorpe PC, Isaac PF, Rodgers J. A controlled trial of mazindol 1 1 1 phetamine (Didrex ), a new appetite suppressant. Arizona Med (Sanorex , Teronao ) in the management of the obese rheu- 1960; 17: 398 – 404. matic patients. Curr Ther Res 1975; 17:149– 155. 79 Persson I, Andersen U, Deckert T. Treatment of obesity with 100 Truant AP, Olon LP, Cobb S. Phentermine resin as an adjunct in fenfluramine. Eur J Clin Pharmacol 1973; 6:93– 97. medical weight reduction: a controlled, randomized, double- 80 Petrie JC, Bewsher PD, Mowat JA, Stowers, JM. Metabolic effects blind prospective study. Curr Ther Res 1972; 14: 726 – 738. of fenfluramine-a double-blind study. Postgrad Med J 1975; 51: 101 Valle-Jones JC, Brodie NH, O’Hara H, O’Hara J, McGhie RL. A 139 – 144. comparative study of phentermine and diethylpropion in the 81 Pijl H, Koppeschaar HPF, Willekens FLA, de Kamp IO, treatment of obese patients in general practice. Pharmatherapeu- Veldhuis HD, Meinders AE. Effect of serotonin re-uptake tica 1983; 3: 300 – 304. inhibition by fluoxetine on body weight and spontaneous 102 Van Gaal LF, Broom JI, Enzi G, Toplak H. Efficacy and tolerability food choice in obesity. Int J Obes Relat Metab Disord 1991; 15: of orlistat in the treatment of obesity: a 6-month study dose- 237 – 242. ranging study. Eur J Clin Pharmac 1998; 54:125– 132. 82 Recasens MA, Barenys M, Sola R, Blanch S, Masana L, Salas- 103 Vernace BJ. Controlled comparative investigation of mazindol, Salvado J. Effect of dexfenfluramine on energy expenditure in d-amphetamine and placebo. J Obes Bariat Med 1974; 3: 124 – obese patients on a very-low-calorie-diet. Int J Obes Relat Metab 129. Disord 1995; 9: 162 – 168. 104 Visser M, Seidell JC, Koppeschaar PF, Smits P. The effect of 83 Sainani GS, Fulambarkar AM, Khurana BK. A double blind trial fluoxetine on body weight, body composition and visceral fat of fenfluramine in the treatment of obesity. Br J Clin Pract 1973; accumulation. Int J Obes Relat Metab Disord 1993; 17:247– 253. 27: 136 – 138. 105 Waal-Manning HJ, Simpson FO. Fenfluramine in obese patients 84 Schteingart DE. Effectiveness of phenylpropanolamine in the on various antihypertensive drugs: Double-blind controlled management of moderate obesity. Int J Obes Relat Metab Disord trial. Lancet 1969; 2: 1392 – 1395. 1992; 16: 487 – 493. 106 Walker BR, Ballard IM, Gold JA. A multicentre study comparing 85 Schwartz LN. A non-amphetamine anorectic agent: preclinical mazindol and placebo in obese patients. J Int Med Res 1977; 5: background and a double-blind clinical trial. J Int Med Res 1975; 85 – 90. 3: 328 – 332. 107 Wallace AG. AN 448 Sandoz (mazindol) in the treatment of 86 Seagle HM, Bessesen DH, Hill JO. Effects of sibutramine on obesity. Med J Aust 1976; 1: 343 – 345. resting metabolic rate and weight loss in overweight women. 108 Weintraub M. Long-term weight control: the National Heart, Obes Res 1998; 6: 115 – 121. Lung, and Blood Institute funded multimodal intervention 87 Sebok M. A double-blinded, placebo-controlled, clinical study of study. Clin Pharmac Ther 1992; 51: 581 – 585. the efficacy of a phenylpropanolamine=caffeine combination 109 {Weintraub M, Sundaresan PR, Madan M, Schuster B, Balder A, product as an aid to weight loss in adults. Curr Ther Res 1984; Lasagna L, Cox C. Long-term weight control study I (weeks 0 to 28: 701 – 708. 34): the enhancement of behavior modification, caloric restric- 88 Sedgwick JP. Mazindol in the treatment of obesity. Practitioner tion, and exercise by fenfluramine plus phentermine versus 1975; 214: 418 – 420. placebo. Clin Pharmacol Ther 1992; 51: 586 – 594. 89 Silverstone JT, Solomon T. The long-term management of obe- 110 {Weintraub M, Sundaresan PR, Schuster B, Ginsberg G, Madan sity in general practice. Br J Clin Pract 1965; 19: 395 – 398. M, Balder A, Stein EC, Byrne L. Long-term weight control study 90 Simkin BWL. A controlled clinical trial of benzphetamine II (weeks 34 – 104): an open-label study of continuous fenflur- 1 (Didrex ) in the management of obesity. Curr Ther Res 1960; amine plus phentermine versus targeted intermittent medica- 2:33– 38. tion as adjuncts to behavior modification, caloric restriction, 91 Simkin B, Wallace L. Some quantitative observations on a and exercise. Clin Pharmac Ther 1992; 51: 595 – 601. methampetamine – phenobarbital anorexic compound in obese 111 {Weintraub M, Sundaresan PR, Schuster B, Moscucci M, Stein outpatients. Am J Med Sci 1960; 239: 533 – 538. EC. Long-term weight control study III (weeks 104 to 156): an 92 Sirtori C, Hurwitz A, Azarnoff DL. Hyperinsulinemia secondary open-label study of dose adjustment of fenfluramine and phen- to chronic administration of mazindol and d-amphetamine. Am termine. Clin Pharmac Ther 1992; 51: 602 – 607. J Med Sci 1971; 261: 341 – 349. 112 {Weintraub M, Sundaresan PR, Schuster B, Averbuch M, Stein 93 Sjostrom L, Rissanen A, Andersen T, Boldrin M, Golay A, Kop- EC, Cox C. Long-term weight control study IV (weeks 156 to peschaar HPF, Krempf M. Randomised placebo-controlled trial 190): the second double-blind phase. Clin Pharmac Ther 1992; of orlistat for weight loss and prevention of weight regain in 51: 608 – 614. obese patients. Lancet 1998; 352: 167 – 172. 113 {Weintraub M, Sundaresan PR, Schuster B, Averbuch M, Stein 94 Sonka J, Limanova Z, Zbirkova A, Kratochvil O. Effects of diet, EC, Byrne L. Long-term weight control study V (weeks 190 to exercise, and anorexigenic drugs on serum thryoid hormones. 210): follow-up of participants after cessation of medicine. Clin Endokrinologie 1980; 76:351– 356. Pharmac Ther 1992; 51: 615 – 618. 95 Sproule BC. Double-blind trial of anorectic agents. Med J Aust 114 {Weintraub M, Sundaresan PR, Cox C. Long-term weight con- 1969; 1: 394 – 395. trol study VI: individual participant response patterns. Clin 96 Stewart DA, Bailey JD, Patell H. Tenuate dospan as an appetite Pharmac Ther 1992; 51: 619 – 633. suppressant in the treatment of obese children. Appl Ther 1970; 115 {Weintraub M, Sundaresan PR, Schuster B. Long-term weight 12:34– 36. control study VI (weeks 0 to 210): serum lipid changes. Clin 97 Stewart GO, Stein GR, Davis TME, Findlater P. Dexfenfluramine Pharmac Ther 1992; 51: 634 – 641. in type II diabetes: effect on weight and diabetes control. Med J 116 {Weintraub M. Long-term weight control study: conclusions. Austr 1993; 158: 167 – 169. Clin Pharmac Ther 1992; 51: 642 – 646. 98 Swinburn BA, Carmichael HE, Wilson MR. Dexfenfluramine 117 Weintraub M, Rubio A, Golik A, Byrne L, Scheinbaum ML. as an adjunct to a reduced-fat, ad libitum diet: effects on Sibutramine in weight control: a dose-ranging, efficacy study. body composition, nutrient intake, and cardiovascular risk Clin Pharamc Ther 1991; 50: 330 – 337. factors. Int J Obes Relat Metab Disord 1996; 20: 1033 – 118 Wise PJ. Clinical experience with a new dosage form of phen- 1040. termine hydrochloride. J Obes Bariat Med 1975; 4: 102 – 105.

International Journal of Obesity