Effect of Policosanol on Lipid Levels Among Patients with Hypercholesterolemia Or Combined Hyperlipidemia a Randomized Controlled Trial

ORIGINAL CONTRIBUTION

Effect of Policosanol on Lipid Levels Among Patients With Hypercholesterolemia or Combined Hyperlipidemia A Randomized Controlled Trial

Heiner K. Berthold, MD, PhD Context Policosanol is a natural substance derived from sugar cane that is adver- Susanne Unverdorben, MD tised for its lipid-lowering effects as a nonprescription drug. More than 80 placebo- Ralf Degenhardt, PhD controlled or comparative trials, performed mostly by a single research institute, suggest that policosanol at doses of 5 to 40 mg/d has lipoprotein-lowering effects comparable Michael Bulitta, Dipl-Stat with statins. Ioanna Gouni-Berthold, MD Objectives To determine the lipoprotein-lowering effects of Cuban sugar cane– derived policosanol and to establish, if effective, dose-dependency up to 80 mg/d in OLICOSANOL IS A MIXTURE OF patients with hypercholesterolemia or combined hyperlipidemia. long-chain primary alcohols iso- Design, Setting, and Participants A multicenter (lipid outpatient clinics and gen- lated from sugar cane wax. eral practitioners in Germany), randomized, double-blind, placebo-controlled, parallel- Cuban sugar cane policosanol is group trial conducted from September 29, 2000, to May 10, 2001, of patients with hy- Psold in more than 40 countries mainly percholesterolemia or combined hyperlipidemia having baseline low-density lipoprotein because of its supposed lipid-lowering cholesterol (LDL-C) levels of at least 150 mg/dL (Ն3.88 mmol/L) and either no or 1 effects.1 Other effects attributed to poli- cardiovascular risk factor other than known coronary heart disease, or baseline LDL-C cosanol are antithrombotic, plaque sta- levels of between 150 and 189 mg/dL (3.88-4.89 mmol/L) and 2 or more risk factors. bilizing, and antioxidant actions. Numer- Interventions Open-label 6-week placebo and diet run-in phase followed by a double- ous policosanol products from a variety blind 12-week treatment phase after randomization to 5 groups: 10, 20, 40, or 80 of sources (sugar cane, wheat germ, rice mg/d of policosanol or placebo. bran, beeswax) are available over-the- Main Outcome Measure The percentage change of LDL-C, with changes in other counter and on the Internet in several lipoproteins as secondary outcome measures. countries. Advertising emphasizes pre- Results A total of 143 patients were randomized to 5 equal groups and were ana- dominantly its putative lipid-lowering lyzed on an intention-to-treat basis. In none of the 5 treatment groups did LDL-C effects. levels decrease more than 10% from baseline. No statistically significant difference Virtually all of the published scien- between policosanol and placebo was observed. A nonparametric test analyzing dose- tific literature supporting the benefi- dependency yielded nonsignificant results. In none of the secondary outcome mea- cial effects of policosanol on lipids has sures, namely total cholesterol, high-density lipoprotein cholesterol (HDL-C), very low- density lipoprotein cholesterol, triglycerides, lipoprotein(a), and ratio of total or LDL-C been authored by a single research to HDL-C, were there any significant effects of policosanol. Policosanol was tolerated group from Cuba. In 2002, we re- well without serious adverse events. viewed the existing literature and have stressed the need for independent con- Conclusion In patients with hypercholesterolemia or combined hyperlipidemia, the 1 sugar cane–derived policosanol in usual and high doses does not demonstrate a re- firmation of these results. A recent duction in lipid levels beyond placebo. meta-analysis of natural therapies for hyperlipidemia concluded that polico- Trial Registration clinicaltrials.gov Identifier: NCT00288483 sanol has lipid-lowering properties JAMA. 2006;295:2262-2269 www.jama.com more effective than plant sterols.2 In Author Affiliations: Department of Clinical Pharma- Medicine II, University of Cologne, Cologne (Dr Gouni- summary, the study shows that the low- cology, Institute for Clinical Research, Center for Car- Berthold), Germany. density lipoprotein cholesterol (LDL-C) diovascular Diseases, Rotenburg an der Fulda (Drs Ber- Corresponding Author: Heiner K. Berthold, MD, PhD, thold, Unverdorben, and Degenhardt), Clinical Drug Commission of the German Medical Associa- lowering effects of sugar cane–derived Research Management Pharmaberatung GmbH, tion, Herbert-Lewin-Platz 1, 10623 Berlin, Germany policosanol are similar to the effects of Rheinbach (Mr Bulitta), and Department of Internal ([email protected]).

Downloaded From: https://jamanetwork.com/ on 09/23/2021 HYPERCHOLESTEROLEMIA OR COMBINED HYPERLIPIDEMIA

statins but are achieved without ad- a double-blind manner. The study medi- Patients verse effects. cation was taken once daily, un- Women and men aged 18 to 80 years Until 2004, trials outside of Cuba chewed, with the evening meal. Polico- with known or newly detected iso- studying its effectiveness are lacking sanol was provided by Dalmer lated hypercholesterolemia or com- with the exception of 2 studies.3,4 All Laboratories (La Habana, Cuba) and bined hyperlipidemia attending a par- trials of the lipid-lowering effects have film-coated tablets were manufactured ticipating clinical trial center were used policosanol manufactured by 1 by Madaus AG (Cologne, Germany). eligible for enrollment. To enter the company in Cuba and an astonishing The placebo tablets contained lactose run-in phase, the participants had to consistency exists between the Cuban monohydrate and microcrystalline cel- have an LDL-C level of at least 150 results of individual studies. There- lulose in equal amounts. mg/dL (Ն3.88 mmol/L) at visit 1. Ran- fore, independent verification from Blood was drawn after an overnight fast domized patients were those patients studies outside Cuba on the Cuban poli- for lipid and lipoprotein analysis and whose mean LDL-C level on visits 2 and cosanol, studies of Cuban and other safety parameters at screening visit 1 (–6 3 (5 and 6 weeks after visit 1) was at policosanols in non-Hispanic popula- weeks), visit 2 and visit 3 (–2 or –1 weeks, least 150 mg/dL (Ն3.88 mmol/L). tions, and studies lasting at least 12 respectively), visit 4 (day 0), and visits Inclusion criteria included the fol- weeks are necessary. The Cuban find- 5 and 6 (6 or 12 weeks, respectively). lowing characteristics: baseline LDL-C ings have been challenged by a clini- Dietary counseling was administered at levels of at least 150 mg/dL (Ն3.88 cal trial from the Netherlands that visits 1 and 4. Patients had to adhere to mmol/L) as described above and either showed that a standard dose (20 mg/d) the step I diet of the US National Cho- no or 1 cardiovascular risk factor other wheat germ–derived policosanol is in- lesterol Education Program (The Third than known coronary heart disease effective in lowering total cholesterol Report of the National Cholesterol Edu- (CHD), or baseline LDL-C levels of be- and LDL-C.5 cation Program Expert Panel on Detec- tween 150 and 189 mg/dL (3.88-4.89 Our trial was performed to corrobo- tion, Evaluation, and Treatment of High mmol/L) for patients with 2 or more rate the reported lipid-lowering ef- Blood Cholesterol in Adults [Adult Treat- risk factors. The following risk factors fects of policosanol. The primary ob- ment Panel III]).6 A 3-day food diary was were considered6-9: age (men Ͼ45 years jective was to show a dose-dependent performed before visits 4 and 6. and women Ͼ55 years or postmeno- LDL-C level decrease in comparison The primary outcome measure of our pausal), known CHD, uncontrolled hy- with placebo in patients with hyper- trial was to show the dose-dependent pertension with systolic blood pres- cholesterolemia or combined hyperlip- LDL-C level decrease with policosa- sure of more than 140 mm Hg, HDL-C idemia. To comply with the require- nol compared with placebo. Second- levels of less than 35 mg/dL (Ͻ0.91 ment to meet the International ary outcome measures were to show mmol/L) for the mean of the values at Conference on Harmonization of Tech- dose-dependent total cholesterol, very visits 2 and 3, current cigarette smok- nical Requirements for Registration of low-density lipoprotein cholesterol ing of more than 10 cigarettes per day, Pharmaceuticals for Human Use guide- (VLDL-C), lipoprotein(a), and triglyc- obesity (body mass index [BMI, calcu- line, “Ethnic factors in the acceptabil- eride level decrease, the decrease of the lated as weight in kilograms divided by ity of foreign clinical data,” we tested ratios of total or LDL-C to high- height in meters squared] of Ͼ30), and the hypothesis of whether the lipid- density lipoprotein cholesterol (HDL- family history of premature CHD. lowering effects of policosanol are re- C), and HDL-C increase by policosa- Major exclusion criteria included producible in patients by using sugar nol vs placebo, as well as to show the myocardial infarction, percutaneous cane–derived policosanol produced in tolerability of policosanol vs placebo. transluminal coronary angioplasty or Cuba. The trial was performed according to coronary artery bypass grafting less than the Declaration of Helsinki (current ver- 1 year before trial inclusion, unstable METHODS sion), taking into account the current angina pectoris, uncorrected hypothy- Study Design and Protocol version of German drug legislation roidism, diabetes mellitus, other endo- Our trial was a multicenter (lipid out- (Arzneimittelgesetz), and it was in ac- crine or metabolic diseases, acute patient clinics and general practition- cordance with the principles of Good inflammatory diseases, severe gastro- ers in Germany), placebo-controlled, Clinical Practice. The study protocol intestinal diseases, triglyceride levels of randomized, double-blind trial with 5 in- was approved by the ethics committee more than 500 mg/dL (Ͼ5.65 mmol/L) dependent parallel treatment groups and at each of the 14 participating centers at visits 2 or 3, use of systemic corti- was conducted from September 29, across Germany, and written in- costeroids, anticoagulants, or lipid- 2000, to May 10, 2001. After an open- formed consent was obtained from all lowering drugs, treatment within the diet and placebo run-in phase (no blind- participants. The principal investiga- previous 6 weeks with any medication ing intended) of 6 weeks, patients were tor (H.K.B.) and the statistician (M.B.) that is known to affect lipoprotein lev- randomized to receive either 10, 20, 40, had complete access to the primary els, severe kidney or liver disease, or or 80 mg/d of policosanol or placebo in data. other severe diseases.

Downloaded From: https://jamanetwork.com/ on 09/23/2021 HYPERCHOLESTEROLEMIA OR COMBINED HYPERLIPIDEMIA

Laboratory Analysis took more than 120% of the pre- Reaction Terminology (WHO-ART). The LDL-C, VLDL-C, and HDL-C lev- scribed dose. Adherence was checked The frequencies and incidence rates els were determined by lipoprotein elec- by pill count at visit 4 (day 0) for the were calculated on a per patient base trophoresis and subsequent quantita- placebo run-in phase and at visit 5 (6 and analyzed according to the method tive cholesterol staining.10 Total weeks) and visit 6 (12 weeks) for the by O’Neill.14 cholesterol and triglyceride levels were treatment phase. Patients found not to Sample size estimation was based on determined enzymatically (cholesterol have adhered to the clinical trial regi- the percentage change of LDL-C level oxidase-␳-aminophenazone or glycerol men at visits 5 or 6 were not included from baseline, a level of significance of phosphate oxidase-␳-aminophenazone in the per-protocol analysis. PϽ.025 (1-sided alternative), and a methods, respectively; Boehringer The primary efficacy parameter, per- power of 80%. It was expected that the Mannheim, Mannheim, Germany). Total centage change of LDL-C from base- active treatment would result in an mass of lipoprotein(a) was determined line, was calculated as the percentage LDL-C level decrease of at least 10% in serum by nephelometry (Beckmann change at visit 5 and 6 from pooled base- (with an SD of 11%) compared with pla- Array, Fullerton, Calif), a method that line (mean value of visits 2, 3, and 4). A cebo. Thus, an effect size (Cohen d) of depends on the measurement of the pooled baseline was used to reduce the 0.91 (10/11) was expected. To achieve intensity of scattered laser light emanat- day-to-day variability of lipid and lipo- the calculated power, 20 patients had ing from an illuminated volume of a protein levels measured during the run-in to be enrolled in each treatment group. dilute suspension of small particles. Bio- phase. To assess every randomized pa- Predefined subgroups were ana- chemical and hematological safety tient with regard to the primary param- lyzed in terms of different responses in parameters were determined by stan- eter in the ITT analysis, an end point the primary outcome parameter, per- dard laboratory methods. analysis using the changes from pooled centage change in LDL-C. The sub- All analyses were performed in a cen- baseline to the last visit was performed groups included comparisons be- tral laboratory at the University of (for completers, visit 6 and for random- tween men and women; patients Freiburg im Breisgau, Germany. This ized dropouts, visit 5 was taken as last younger or older than 65 years; base- laboratory applied Good Laboratory visit, respectively). Descriptive results line LDL-C levels of less than 165 Practice standard and was certified ac- were also calculated for the changes from mg/dL (Ͻ4.27 mmol/L), between 165 cording to International Organization pooled baseline to visit 5 (6 weeks) and and 195 mg/dL (4.27-5.04 mmol/L), or for Standardization (ISO-9001) (http:// visit 6 (12 weeks). more than 195 mg/dL (Ͼ5.04 mmol/ www.iso.org), which defines quality The a priori statistical working hy- L); BMI of less than or greater than 28; management systems. pothesis was that the highest dose is number of risk factors; smoker status; more effective in lowering LDL-C lev- and medical care setting (lipid outpa- Statistical Analysis els than the next lower dose and so on, tient clinic vs general practitioner). Randomization was performed by us- the lowest dose being more effective Diet protocols were evaluated using ing a 5 permutation block according to than placebo. Multiple testing proce- Prodi version 4.5 software (Wissen- a list generated by means of the PROC dure was performed by using the trend schaftliche Verlagsgesellschaft, Stutt- PLAN procedure of the statistical pack- test according to Jonckheere and Terp- gart, Germany). The following vari- age SAS version 6.0 (SAS Institute, Cary, stra (JT test).11,12 One-sided level of ables were compared between visits 4 NC). All randomized patients for whom significance was stipulated at P=.025. and 6 with paired statistics: total calo- at least 1 lipoprotein value was mea- The treatment effects between polico- ries, percentage of calories from fat sured during treatment were evalu- sanol and placebo were estimated by (total, saturated, unsaturated), pro- ated in the intention-to-treat (ITT) means of the nonparametric Hodges- tein, carbohydrates, alcohol, intake of group in terms of efficacy irrespective Lehmann estimators with 95% confi- total cholesterol, and fiber intake. of any protocol violations. All patients dence intervals.13 Descriptive and ex- of the ITT group who exhibited no ma- ploratory methods were used for the RESULTS jor protocol violations and sufficient ad- analysis of secondary variables. The sta- The flow of participants through the trial herence were assessed for efficacy in the tistical package SAS version 6.0 was is shown in the FIGURE. A total of 143 per-protocol group. Before the code was used for general calculations. Sample patients were randomized into the broken, a blind review was conducted size was calculated by using the soft- double-blind treatment phase. Nine of to decide which protocol violations ware “N” version 2.2 (IDV-Data Analy- the patients terminated the trial prema- could be characterized as major. Safety sis and Study Planning, Gauting, Ger- turely. The reasons included not meet- and tolerability were assessed for all pa- many). The JT test was calculated using ing inclusion criteria or meeting exclu- tients who had taken at least 1 dose of StatXact-4 for Windows (Cytel Soft- sion criteria in 4 patients, decision of pa- the trial medication, including placebo. ware Corp, Cambridge, Mass). Ad- tient in 3 cases (withdrew), and adverse Patients were classified as nonadher- verse events were coded according to events in 2 patients. Five more patients ent if they failed to take at least 80% or World Health Organization–Adverse had to be excluded from per-protocol

Downloaded From: https://jamanetwork.com/ on 09/23/2021 HYPERCHOLESTEROLEMIA OR COMBINED HYPERLIPIDEMIA

analysis due to the following reasons: er- lations; therefore, 129 patients were ton pump inhibitors (8%), nonsteroi- roneously receiving interchanged trial analyzed per protocol. dal anti-inflammatory drugs (3%), hor- medication (n=2), completion of the Baseline characteristics of the random- mone therapy (6%), benzodiazepines trial but exclusion criteria were discov- ized patients are shown in TABLE 1.Of (4%), selective serotonin reuptake in- ␤ ered afterward (n=1), itraconazole treat- the 143 randomized patients, 56 (39%) hibitors (6%), and 2 agonists (4%). ment became necessary during the trial were men and 87 (61%) were women. Adherence was close to 100% in all (n=1), and insufficient adherence to trial Mean (SD) age was 56 (12) years (range, treatment groups, as defined by the per- medication(pillcount)(n=1).Thus,per- 23-78) and BMI was 27.2 (3.6). All pa- centage of trial medication taken dur- protocol analysis was performed in 129 tients were white and definition of race ing treatment phase verified through pill patients. was by patient self-identification. A total count. of 26 patients (18%) were smokers and Patient Characteristics 4 (3%) were ex-smokers. Forty-two pa- Effects on Plasma Lipids A total of 216 patients were enrolled in tients (29%) had a family history of pre- and Lipoproteins 14 trial centers, of which 143 patients mature CHD. All demographic data were The results of the lipid and lipopro- were randomized and entered in the equally distributed among the groups. tein measurements are shown in ITT analysis (Figure). Of the random- Concomitant medications were also TABLE 2. Only data from the ITT analy- ized patients, 114 (79.7%) were treated equally distributed among the groups sis (as the more conservative statisti- with policosanol and 29 (20.3%) re- and consisted of ␤-blockers (15%), an- cal method) are presented because there ceived placebo, reflecting the antici- giotensin-converting enzyme inhibi- were no differences between ITT and pated allocation. A total of 134 of the tors (10%), angiotensin II type 1 recep- per-protocol analysis. randomized patients (93.7%) com- tor antagonists (5%), diuretics (8%), The mean (SD) LDL-C level of the pleted the trial. Fourteen randomized calcium antagonists (4%), other antihy- total group of randomized patients at patients (9.8%) were excluded from the pertensive medications (8%), thyroid baseline was 187 (36) mg/dL (4.84 [0.93] ITT analysis due to major protocol vio- hormones (13%), aspirin (10%), pro- mmol/L). There were no differences

Figure. Flow of Participants Through the Trial

216 Assessed for Eligibility

1 Withdrew Informed Consent Before Enrollment in Placebo Run-in Phase

215 Enrolled in Placebo Run-in Phase

72 Excluded∗ 54 Did Not Meet Inclusion/Met Exclusion Criteria 12 Refused to Participate 10 Decision by Investigator 2 Adverse Events During Placebo Run-in Phase

143 Randomized

29 Assigned to Receive Placebo 28 Assigned to Receive 10-mg 27 Assigned to Receive 20-mg 27 Assigned to Receive 40-mg 32 Assigned to Receive 80-mg Policosanol Policosanol Policosanol Policosanol

27 Completed Study 26 Completed Study 26 Completed Study 25 Completed Study 30 Completed Study 1 Did Not Meet Inclusion/Met 1 Did Not Meet Inclusion/Met 1 Adverse Event 1 Did Not Meet Inclusion/Met 1 Did Not Meet Inclusion/Met Exclusion Criteria Exclusion Criteria Exclusion Criteria Exclusion Criteria 1 Adverse Event 1 Withdrew 1 Withdrew 1 Withdrew

29 Included in Intention-to-Treat 28 Included in Intention-to-Treat 27 Included in Intention-to-Treat 27 Included in Intention-to-Treat 32 Included in Intention-to-Treat Analysis Analysis Analysis Analysis Analysis

25 Included in Per-Protocol 26 Included in Per-Protocol 25 Included in Per-Protocol 24 Included in Per-Protocol 29 Included in Per-Protocol Analysis Analysis Analysis Analysis Analysis 2 Excluded 1 Excluded (Insufficient 1 Excluded (Did Not Meet 1 Excluded (Wrong Medication) 1 Required Itraconazole Adherence) Inclusion/Met Exclusion 1 Received Wrong Medication Criteria)

*Number of patients excluded do not total 72 because there were multiple citations in 6 cases.

Downloaded From: https://jamanetwork.com/ on 09/23/2021 HYPERCHOLESTEROLEMIA OR COMBINED HYPERLIPIDEMIA

Table 1. Baseline Characteristics of the Randomized Patients* Policosanol

Placebo 10 mg 20 mg 40 mg 80 mg P Characteristics (n = 29) (n = 28) (n = 27) (n = 27) (n = 32) Value† Sex, No. (%) Male 16 (55) 9 (32) 7 (26) 11 (41) 13 (41) .22 Female 13 (45) 19 (68) 20 (74) 16 (59) 19 (59) Age, y 53.8 (13.8) 58.7 (11.5) 57.0 (10.2) 54.9 (9.9) 55.8 (11.8) .55 Height, cm 170 (8) 168 (8) 166 (7) 171 (8) 169 (8) .23 Weight, kg 78 (13) 76 (12) 76 (11) 78 (12) 80 (13) .67 BMI 27.1 (3.9) 26.8 (3.3) 27.5 (3.6) 26.8 (3.7) 27.7 (3.5) .80 % of patients‡ Overweight 69 68 78 67 84 .02 Obese 28 14 37 15 25 Ͻ.001 Smoker status, No. (%) Smoker 4 (14) 3 (11) 8 (30) 4 (15) 7 (22) .37 Ex-smoker 1 (3) 1 (4) 1 (4) 0 (0) 1 (3) .91 Family history of premature 10 (35) 8 (29) 5 (19) 9 (33) 10 (31) .71 cardiovascular disease, No. (%) Abbreviation: BMI, body mass index, calculated as weight in kilograms divided by height in meters squared. *Data are expressed as mean (SD) unless otherwise specified. See “Methods” section for definitions of characteristics. †For sex, percentage of patients who are overweight or obese, smoker status, and family history of premature cardiovascular disease, P values were calculated by 2-sided ␹2 test; and for age, height, weight, and BMI, P values were calculated by analysis of variance. ‡Defined as BMI of at least 25 for overweight and at least 30 for obese.

Table 2. Results of Lipoprotein Measurements* Mean (SD) [% Change From Baseline {SD}]

Lipids, mg/dL Ratio

Total Total LDL-C HDL-C Cholesterol VLDL-C Triglycerides Lipoprotein(a) Cholesterol/HDL-C LDL-C/HDL-C Placebo Baseline 181 (36) 50 (16) 276 (39) 45 (18) 176 (65) 37 (46) 6.0 (1.9) 4.0 (1.5) 6 wk 182 (42) 53 (16) 277 (41) 43 (17) 166 (62) 39 (47) 5.7 (1.9) 3.8 (1.6) [–0.8 {12}] [2.0 {8}] [–1.0 {26}] [–1.9 {15}] [–8.1 {48}] [1.1 {5.0}] [–0.3 {1.1}] [–0.2 {1.0}] 12 wk 166 (45) 49 (16) 257 (51) 42 (21) 157 (75) 24 (52) 5.7 (2.3) 3.8 (2.0) [–9.2 {13}] [–1.0 {10}] [–20 {34}] [–3.1 {18}] [–21 {59}] [3.3 {9.4}] [–0.4 {1.2}] [–0.3 {0.9}] 10-mg policosanol Baseline 200 (55) 52 (12) 293 (65) 41 (14) 154 (49) 33 (31) 6.1 (2.1) 4.2 (1.7) 6 wk 203 (63) 51 (13) 293 (66) 39 (13) 152 (55) 35 (34) 6.1 (2.1) 4.2 (1.8) [1.0 {8}] [–0.3 {8}] [–0.8 {17}] [–2.6 {13}] [–2.3 {49}] [2.0 {5.5}] [–0.1 {1.0}] [–0.02 {0.8}] 12 wk 183 (47) 53 (15) 276 (47) 40 (13) 147 (59) 35 (32) 5.8 (2.7) 3.9 (2.4) [–5.7 {11}] [2.5 {13}] [–11 {23}] [–1.7 {12}] [–7.4 {45}] [1.9 {5.2}] [–0.4 {1.4}] [–0.3 {1.3}] 20-mg policosanol Baseline 185 (25) 55 (15) 279 (30) 39 (12) 156 (58) 38 (45) 5.4 (1.3) 3.6 (1.0) 6 wk 184 (35) 63 (22) 288 (41) 41 (17) 151 (64) 43 (51) 5.1 (1.9) 3.3 (1.6) [–0.5 {13}] [7.9 {13}] [9.4 {23}] [2.1 {15}] [–4.7 {50}] [3.4 {7.9}] [–0.3 {1.3}] [–0.3 {1.2}] 12 wk 175 (31) 56 (13) 271 (32) 41 (12) 149 (55) 43 (50) 5.2 (1.7) 3.4 (1.3) [–5.1 {12}] [0.6 {13}] [–7.5 {21}] [1.4 {12}] [–6.9 {61}] [4.2 {7.9}] [–0.8 {1.1}] [–0.2 {0.9}] 40-mg policosanol Baseline 181 (19) 50 (11) 277 (26) 47 (16) 195 (75) 40 (42) 6.0 (1.5) 3.9 (1.1) 6 wk 184 (32) 48 (15) 273 (39) 47 (18) 209 (86) 39 (44) 6.3 (1.7) 4.2 (1.3) [2.3 {17}] [–0.8 {10}] [3.0 {39}] [–0.2 {14}] [11 {67}] [–0.6 {8.1}] [0.2 {1.1}] [0.2 {0.9}] 12 wk 176 (29) 52 (18) 270 (36) 42 (22) 177 (108) 43 (46) 5.6 (1.6) 3.7 (1.3) [–2.0 {16}] [3.7 {16}] [–6.1 {30}] [–5.5 {13}] [–20 {84}] [2.1 {7.0}] [–0.5 {1.3}] [–0.3 {1.1}] 80-mg policosanol Baseline 186 (44) 50 (15) 284 (51) 48 (21) 196 (84) 35 (31) 6.4 (2.3) 4.2 (1.7) 6 wk 180 (43) 53 (19) 281 (53) 48 (24) 203 (105) 38 (32) 5.8 (2.1) 3.8 (1.6) [–4.0 {17}] [3.5 {16}] [–5.3 {28}] [–0.3 {17}] [2.9 {78}] [2.5 {6.5}] [–0.6 {1.6}] [–0.5 {1.3}] 12 wk 173 (53) 54 (22) 274 (52) 48 (24) 210 (140) 36 (31) 5.9 (3.0) 3.8 (2.6) [–9.0 {12}] [4.6 {14}] [–13 {21}] [–0.7 {17}] [10 {90}] [0.5 {6.8}] [–0.6 {1.7}] [–0.5 {1.5}] Abbreviations: HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; VLDL-C, very low-density lipoprotein cholesterol. SI conversions: To convert HDL-C, LDL-C, VLDL-C, and total cholesterol to mmol/L, multiply by 0.0259; and triglycerides to mmol/L, multiply by 0.0113. *Baseline values were mean levels of 3 independent visits during the run-in phase at –2, –1, and 0 weeks.

Downloaded From: https://jamanetwork.com/ on 09/23/2021 HYPERCHOLESTEROLEMIA OR COMBINED HYPERLIPIDEMIA

between the groups at baseline. There Table 3. Relative Change, Hodges-Lehmann Estimators, and Parametric Mean Differences was a slight decrease in some of the 5 for the Relative Change in LDL-C Levels vs Placebo treatment groups after 6 and 12 weeks Comparison vs Placebo % Change (95% CI) compared with baseline. However, all Relative Change Hodges-Lehmann Mean Difference Calculated these reductions were below 10% and Mean % (SD)* Estimator† by Parametric Test‡ therefore of limited clinical relevance. A Placebo –8.1 (13.8) statistically significant dose-dependent Policosanol, mg/d trend of LDL-C level decrease could not 10 –5.2 (11.2) –2.7 (–10.3 to 4.2) –3.0 (–9.8 to 3.9) be demonstrated (P=.30, 1-sided JT test). 20 –5.1 (12.3) –3.0 (–10.2 to 5.0) –3.1 (–10.2 to 4.1) The corresponding JT test for the per- 40 –2.0 (16.2) –4.8 (–13.4 to 2.8) –6.1 (–14.4 to 2.1) protocol analysis was P=.39. 80 –8.3 (12.4) –0.8 (–7.2 to 6.6) 0.2 (–6.7 to 7.1) ABLE Abbreviations: CI, confidence interval; LDL-C, low-density lipoprotein cholesterol. T 3 shows the relative change, *The relative change in LDL-C between pooled baseline and the last visit (for completers visit 6 at 12 weeks and for Hodges-Lehmann estimators, and the dropouts visit 5 at 6 weeks). †The Hodges-Lehmann estimator is the median of all differences (N) calculated between 2 groups (N = NPl ϫ NTx, where mean differences calculated by para- NPl indicates the number of patients in the placebo group and NTx indicates the number in the respective treatment metric tests for the pair-wise treat- group being compared with placebo). ‡Mean differences were obtained by subtracting the differences of the 2 groups using the results in the intention-to- ment differences regarding the percent- treat analysis (for completers, the results of visit 6 [12 weeks] and for randomized dropouts, the results of visit 5 [6 age change in LDL-C levels for the 4 weeks, which is the last observation carried forward]). active treatment groups vs placebo. Hodges-Lehmann estimators, which are Safety and Adverse Events mine, if effective, the dose-dependency based on ranks, are more robust esti- All 215 patients enrolled received at least of effects and optimal doses. Two of the mators of true differences than para- 1 dose of trial medication and were in- 5 treatment groups were treated with metric estimators. Both approaches cluded in the safety evaluation (172 pa- usual doses (10 and 20 mg/d), while 2 yielded similar results. tients received policosanol and 43 re- groups received high doses (40 and 80 The results of the secondary out- ceived placebo). The safety profile was mg/d) of policosanol. The study revealed come measures are also given in Table 2. excellent. In the double-blind phase, 45 no significant differences between the Mean (SD) baseline levels were 51 (14) patients experienced at least 1 adverse groups in all lipoprotein fractions exam- mg/dL (1.32 [0.36] mmol/L) for HDL-C, event. The proportion between the ined, namely LDL-C, HDL-C, total cho- 282 (42) mg/dL (7.29 [1.09] mmol/L) for groups was not statistically significant lesterol, VLDL-C, triglycerides, and lipo- total cholesterol, 44 (16) mg/dL (1.14 (28% of patients received placebo, 21% protein(a). A nonparametric test (JT test) [0.41] mmol/L) for VLDL-C, 175 (66) received 10 mg, 37% received 20 mg, used to discover a possible dose- mg/dL (1.98 [0.75] mmol/L) for triglyc- 37% received 40 mg, and 34% received dependency as well as comparison of erides, 37 (39) mg/dL (1.32 [1.39] 80 mg). The most commonly reported active treatment with placebo was not µmol/L) for lipoprotein(a), 6.0 (1.8) for adverse events were bronchitis (13%), significant. The results are in contrast to the ratio of total cholesterol to HDL-C, back pain (11%), pharyngitis (9%), up- a large number of published random- and 4.0 (1.4) for the ratio of LDL-C to per respiratory tract infection (7%), and ized controlled trials. HDL-C. There were no significant dif- gastritis (5%). There were no differ- Recently, a systematic review of the ferences between the groups at baseline ences in the frequencies between active literature up to June 2003 of random- or in the changes from baseline at 6 or treatment and placebo or between the ized controlled trials using policosanol 12 weeks between the groups. dosage groups. No serious adverse events was published.2 Concerning the pri- There were no differences in the ef- were found in all treatment groups. mary end point of reduction of LDL-C, fects on the primary outcome variable, Policosanol treatment resulted in no this meta-analysis found 29 eligible stud- percentage change in LDL-C, between relevant changes in weight, vital signs ies with 1528 patients receiving treat- predefined subgroups as defined in the (blood pressure, pulse rate), or rou- ment compared with 1406 patients re- Methods section (data not shown). tine chemistry and hematological labo- ceiving placebo. Twenty-eight of the 29 ratory parameters (data not shown). trials came from Cuba and 1 trial came Diet Protocols from Argentina.15 Weighted estimates of Food records (3 consecutive days at the COMMENT percentage change in LDL-C were end of the run-in phase and at the end Our rigorously controlled trial found no –23.7% for policosanol (mean dose of of the treatment phase) were evalu- effect of Cuban sugar cane–derived poli- policosanol, 12 mg/d; range, 5-40 mg/d) ated using computerized nutrient cosanol on lipoprotein levels in patients vs –0.11% for placebo (PϽ.001). This analysis. No significant differences be- with hypercholesterolemia or com- LDL-C reduction is well in the order of tween the 2 time points (paired statis- bined hyperlipidemia beyond placebo. magnitude that is achieved with com- tics) or between treatment groups Our goal was to corroborate the effec- monly used doses of statins.16 In fact, (analysis of variance) were found (data tiveness of policosanol in lowering lipo- some trials, again all but 1 from Cuba, not shown). protein concentrations and to deter- directly compared the effects of polico-

Downloaded From: https://jamanetwork.com/ on 09/23/2021 HYPERCHOLESTEROLEMIA OR COMBINED HYPERLIPIDEMIA

sanol with statins3,17-21 and found poli- which is in contrast with findings of a ishes cardiovascular morbidity. Be- cosanol equally effective or better. Even variety of animal studies from Cuba.28,29 cause CHD remains the leading cause of a dose of policosanol as low as 2 mg/d The putative mechanism of action of death in industrialized countries and was found to lower LDL-C signifi- policosanol has not been clarified, al- some patients cannot be treated with stat- cantly by more than 15%.22 It has been though suppression of the 3-hydroxy- ins, effective pharmacological lipid- suggested that the lipid-lowering ef- 3-methylglutaryl coenzyme A reduc- lowering alternatives are greatly de- fects of policosanol are dose-depen- tase activity has been postulated.30 sired. The main target of treatment dent from 2 to 40 mg/d.23 There are several possible limitations remains effective lowering of LDL-C lev- The reasons for the discrepant re- to our study. Our study design is gener- els,31 especially because a recent meta- sults between our trial and previous ally accepted for studies investigating regression analysis showed that LDL-C studies are unclear. As nearly all pre- lipid-lowering agents and our trial was level decrease, and not the mechanism vious studies have been performed in a multicenter trial. Lipid and lipopro- through which it is achieved, is impor- Latin America (1 positive study was tein measurements were obtained on 3 tant for reducing CHD risk.32 Our data performed in Russia),4 it cannot be ex- different days at baseline (pooled base- clearly show that policosanol has no cluded that either ethnic or nutri- line) to get reliable values. The measure- lipid-lowering effects beyond placebo. tional factors contribute to the impres- ments were performed in a central, cer- The policosanol-induced changes in sively potent lipid-lowering effects of tified laboratory. The trial consisted of LDL-C, all less than 10%, cannot be con- policosanol reported. Although eth- a placebo and diet run-in phase with in- sidered clinically relevant. When the nic and nutritional differences be- structions to adhere to a National Cho- Adult Treatment Panel III indicates that tween European white and Latin Ameri- lesterol Education Program step I diet,6 drug therapy is considered for LDL-C can patients might play a role, the initiated 4 to 6 weeks before baseline val- lowering, it is reasonable to use doses magnitude of differences reported (no ues were drawn. Diet had no influence achieving a reduction in risk for major effect vs a consistent 25% LDL-C re- on the results, as confirmed by compari- coronary events of 30 to 40%.31 The large duction) makes this explanation un- son of food diaries at the end of the run-in intervention trials with statins have likely. Furthermore, other lipid- phase and the treatment phase. The treat- shown, however, that a risk reduction of lowering drugs, such as statins24 and ment duration was 12 weeks. It might be this extent requires an approximate 30% ezetimibe25 have been shown to have argued that lipid-lowering effects take reduction in LDL-C concentrations.31 no ethnic-specific effects. The same longer to develop, although currently no A considerable health food store and holds true for responses to low-fat diet.26 known pharmacological intervention Internet market has extended the de- Almost all studies were supported by would support this notion. In the meta- velopment of nonprescription polico- 1 sponsor, Dalmer Laboratories, a com- analysis of Chen et al,2 the average treat- sanol, and worldwide sales are con- mercial enterprise founded by the Cen- ment duration of the 29 studies was 30 stantly increasing. A simple search in the ter of Natural Products, National Cen- weeks and a mean LDL-C level de- Internet using the term policosanol gives ter for Scientific Research, La Habana, crease of 23.7% was found. Extracting the about 640 000 items, with discount of- Cuba, to market policosanol. Our group1 subset of the 12 studies that had treat- fers and bulk prices, for a substance that and other studies5,23 have therefore sug- ment durations of 12 weeks or less, the is marketed as “potent, natural and free gested that independent studies should LDL-C lowering effect in these trials was of side effects.” A widely held belief be performed to prove the efficacy of poli- still 24%. Furthermore, the published among users is that policosanol is effec- cosanol. Until now, independently per- positive policosanol studies found lipid- tively lowering cholesterol. A national formed studies are scarce. Our trial is the lowering effects already after 4 to 6 weeks survey documented that alternative first study to investigate sugar cane– of treatment.1 Moreover, the putative medicine use, such as herbal drugs, is derived policosanol independently from lipid-lowering effect of policosanol has increasing and is attributable primarily the aforementioned Cuban research been reported to plateau after 10 weeks to an increase in the proportion of the group but still using Cuban policosa- of administration.29 We conclude there- population seeking alternative thera- nol. A randomized placebo-controlled fore that trial duration cannot be made pies.33 As justifiable as it may be to take trial from the Netherlands of 58 pa- responsible for the observed absence of patients’ wishes for herbal medicines into tients found no effects when using wheat policosanol effects on lipids. We used account, these drugs have to be scruti- germ policosanol (20 mg/d) on lipopro- Cuban sugar cane–derived policosanol nized regarding their effectiveness, ap- teins in patients with normal to mildly that was provided as bulk from Dalmer plying the same scientific criteria as stan- increased cholesterol.5 Very few other Laboratories; therefore, a difference in the dard drug therapies. studies outside Cuba exist but are not source of policosanol as a reason for the It is known from more than 80 trials convincing in quality. An animal study discrepant results is highly unlikely. performed so far that policosanol has performed in Canada in hamsters In general, it has been clearly demon- an excellent safety profile. Also, in our showed that sugar cane and rice wax strated that cholesterol-lowering therapy study, the drug was tolerated well and policosanol have no effects on lipids,27 reduces the risk of CHD and dimin- no serious adverse events occurred. The

Downloaded From: https://jamanetwork.com/ on 09/23/2021 HYPERCHOLESTEROLEMIA OR COMBINED HYPERLIPIDEMIA

frequency of adverse effects was com- gau (n=10), I. Naudts, MD, Dudenhofen (n=30), M. 15. Mirkin A, Mas R, Martinto M, et al. Efficacy and Ritter, MD, Ibbenbu¨ ren (n=4), I. Urlea-Scho¨ n, MD, tolerability of policosanol in hypercholesterolemic post- parable between placebo and the ac- Siegen (n=21), B. Hinkelmann, MD, Kuppenheim menopausal women. Int J Clin Pharmacol Res. 2001; tive treatment groups but more de- (n=15), S. Mantz, MD, Offenbach (n=10), A. Schmidt, 21:31-41. MD, Offenbach (n=10), P. Franz, MD, Berlin (n=5), 16. Jones P, Kafonek S, Laurora I, Hunninghake D. tailed information on the safety profile R. Van der Meij, MD, Hagen-Hohenlimberg (n=10). Comparative dose efficacy study of atorvastatin ver- of policosanol is warranted, especially Acknowledgment: We thank Michael Scha¨ fer, MD, sus simvastatin, pravastatin, lovastatin, and fluvas- since it can be obtained as an over-the- Uwe Phillip Strauss, MD, and Reinhold Sieck, PhD, all tatin in patients with hypercholesterolemia (the employed by Madaus AG, who helped in planning the CURVES study). Am J Cardiol. 1998;81:582-587. counter drug. trial. We are indebted to Karin Thiele, Dipl-Phil, Anke 17. Castano G, Mas R, Arruzazabala ML, et al. Ef- In conclusion, policosanol is a well- Bru¨ hl, PTA, and Ursula Scha¨ fer, MD, all freelance col- fects of policosanol and pravastatin on lipid profile, laborators of the contract research organization, for platelet aggregation and endothelemia in older hy- tolerated substance derived from natu- monitoring the trial, and Claudia Krapf, employed by percholesterolemic patients. Int J Clin Pharmacol Res. ral sources. Our results suggest that it the principal investigator’s research institution, for tech- 1999;19:105-116. nical assistance. Additional thanks to W. Stro¨ sser, MD, 18. Castano G, Menendez R, Mas R, et al. Effects of is devoid of clinically relevant lipopro- consultant for Madaus AG, for helpful discussions. All policosanol and lovastatin on lipid profile and lipid per- tein-lowering properties in white of the named persons received compensation from the oxidation in patients with dyslipidemia associated with funding sponsor with the exception of Mrs Krapf, who type 2 diabetes mellitus. Int J Clin Pharmacol Res. 2002; patients. Still, more independent stud- received no compensation. 22:89-99. ies are required to counterbalance the 19. Castano G, Mas R, Fernandez L, et al. Compari- vast body of available positive trials. REFERENCES son of the efficacy and tolerability of policosanol with atorvastatin in elderly patients with type II Although policosanol has been used for 1. Gouni-Berthold I, Berthold HK. Policosanol: clini- hypercholesterolaemia. Drugs Aging. 2003;20:153- more than a decade in clinical trials, cal pharmacology and therapeutic significance of a new 163. lipid-lowering agent. Am Heart J. 2002;143:356-365. 20. Castano G, Mas R, Fernandez L, Gamez R, Illnait 2. there are still no data on patient- Chen JT, Wesley R, Shamburek RD, Pucino F, Csako J. Effects of policosanol and lovastatin in patients with G. Meta-analysis of natural therapies for hyperlipid- intermittent claudication: a double-blind compara- related outcomes, such as cardiovas- emia: plant sterols and stanols versus policosanol. tive pilot study. Angiology. 2003;54:25-38. Pharmacotherapy. 2005;25:171-183. cular morbidity and mortality. More- 21. Crespo N, Illnait J, Mas R, Fernandez L, Fernan- 3. Prat H, Roman O, Pino E. Comparative effects of dez J, Castano G. Comparative study of the efficacy over, independent information should policosanol and two HMG-CoA reductase inhibitors and tolerability of policosanol and lovastatin in pa- be given to consumers who might take on type II hypercholesterolemia. Rev Med Chil. 1999; tients with hypercholesterolemia and noninsulin de- 127:286-294. policosanol to improve their cardio- pendent diabetes mellitus. Int J Clin Pharmacol Res. 4. Nikitin IuP, Slepchenko NV, Gratsianskii NA, et al. 1999;19:117-127. vascular risk profile. Results of the multicenter controlled study of the hy- 22. Batista J, Stusser R, Saez F, Perez B. Effect of poli- polipidemic drug polycosanol in Russia. Ter Arkh. 2000; Author Contributions: Dr Berthold had full access to 72:7-10. cosanol on hyperlipidemia and coronary heart dis- all of the data in the study and takes responsibility for 5. Lin Y, Rudrum M, van der Wielen RP, et al. Wheat ease in middle-aged patients: a 14-month pilot study. the integrity of the data and the accuracy of the data germ policosanol failed to lower plasma cholesterol Int J Clin Pharmacol Ther. 1996;34:134-137. analysis. in subjects with normal to mildly elevated cholesterol 23. Janikula M. Policosanol: a new treatment for car- Study concept and design: Berthold, Bulitta. concentrations. Metabolism. 2004;53:1309-1314. diovascular disease? Altern Med Rev. 2002;7:203-217. Acquisition of data: Berthold, Unverdorben, Degenhardt, 6. Expert Panel on Detection, Evaluation, and Treat- 24. Tan CE, Loh LM, Tai ES. Do Singapore patients Bulitta. ment of High Blood Cholesterol in Adults. Executive require lower doses of statins? the SGH Lipid Clinic Analysis and interpretation of data: Berthold, Bulitta, Summary of the Third Report of the National Cho- experience. Singapore Med J. 2003;44:635-638. Gouni-Berthold. lesterol Education Program (NCEP) Expert Panel on 25. Pearson TA, Denke MA, McBride PE, Battisti WP, Drafting of the manuscript: Berthold, Gouni-Berthold. Detection, Evaluation, and Treatment of High Blood Brady WE, Palmisano J. A community-based, random- Critical revision of the manuscript for important in- Cholesterol in Adults (Adult Treatment Panel III). JAMA. ized trial of ezetimibe added to statin therapy to at- tellectual content: Unverdorben, Degenhardt, Bulitta. 2001;285:2486-2497. tain NCEP ATP III goals for LDL cholesterol in hyper- Statistical analysis: Berthold, Bulitta. 7. Smith SC Jr, Blair SN, Bonow RO, et al. AHA/ACC cholesterolemic patients: the Ezetimibe Add-on to Statin Obtained funding: Berthold, Bulitta. scientific statement: AHA/ACC guidelines for pre- for Effectiveness (EASE) trial. Mayo Clin Proc. 2005; Administrative, technical, or material support: venting heart attack and death in patients with ath- 80:587-595. Unverdorben, Degenhardt, Gouni-Berthold. erosclerotic cardiovascular disease: 2001 update: a 26. Howard BV, Hannah JS, Heiser CC, Jablonski KA. Study supervision: Berthold. statement for healthcare professionals from the Ameri- Effects of sex and ethnicity on responses to a low-fat Financial Disclosures: None reported. can Heart Association and the American College of diet: a study of African Americans and whites. Am J Funding/Support: This study was sponsored by Madaus Cardiology. Circulation. 2001;104:1577-1579. Clin Nutr. 1995;62:488S-492S. AG, which also provided funding of the contract re- 8. Prevention of coronary heart disease in clinical prac- 27. Wang YW, Jones PJ, Pischel I, Fairow C. Effects search organization involved. Neither Madaus AG nor tice: recommendations of the Second Joint Task Force of policosanols and phytosterols on lipid levels and cho- any of their subsidiary companies manufactures or dis- of European and other societies on coronary lesterol biosynthesis in hamsters. Lipids. 2003;38:165- tributes any lipid-lowering drugs, including statins. prevention. Eur Heart J. 1998;19:1434-1503. 170. Role of the Sponsor: Employees of Madaus AG con- 9. Assmann G, Cullen P, Schulte H. Simple scoring 28. McCarty MF. An ezetimibe-policosanol combi- tributed to the design of the study. The study was scheme for calculating the risk of acute coronary events nation has the potential to be an OTC agent that could conducted by support from the contract research or- based on the 10-year follow-up of the Prospective Car- dramatically lower LDL cholesterol without side effects. ganization, Clinical Research Management Pharmabera- diovascular Munster (PROCAM) study. Circulation. Med Hypotheses. 2005;64:636-645. tung GmbH. The sponsor had no influence on the in- 2002;105:310-315. 29. Varady KA, Wang Y, Jones PJ. Role of policosa- terpretation of the data or the writing of the manuscript. 10. Aufenanger J, Haux P, Kattermann R. Improved nols in the prevention and treatment of cardiovascu- Independent Statistical Analysis: An independent sta- method for enzymic determination of cholesterol in lar disease. Nutr Rev. 2003;61:376-383. tistical analysis was conducted by Walter Lehma- lipoproteins separated by electrophoresis on thin layer 30. Menendez R, Amor AM, Rodeiro I, et al. Polico- cher, PhD, from the Institute for Medical Statistics, In- agarose gels. J Clin Chem Clin Biochem. 1989;27:807- sanol modulates HMG-CoA reductase activity in cul- formatics, and Epidemiology, University of Cologne, 813. tured fibroblasts. Arch Med Res. 2001;32:8-12. Cologne, Germany. Dr Lehmacher reviewed a draft 11. Terpstra TJ. The asymptotic normality and con- 31. Grundy SM, Cleeman JI, Merz CN, et al. Impli- of the manuscript and had complete access to the origi- sistency of Kendall’s test against trend, when ties are cations of recent clinical trials for the National Cho- nal data. Using the same nonparametric test of Jonck- present in one ranking. Indagationes Mathematicae. lesterol Education Program Adult Treatment Panel III heere and Terpstra ( JT test) as the authors, Dr Lehma- 1952;14:327-333. guidelines. Circulation. 2004;110:227-239. cher arrived at the same results. No remuneration was 12. Jonckheere AR. A distribution-free k-sample test 32. Robinson JG, Smith B, Maheshwari N, Schrott H. paid for his services. against ordered alternatives. Biometrika. 1954;41:133- Pleiotropic effects of statins: benefit beyond choles- Study Group: Germany: H.K. Berthold, MD, PhD, Ro- 145. terol reduction? a meta-regression analysis. J Am Coll tenburg (n=30 patients), E. Bo¨ nninghoff, MD, Bec- 13. Hodges JL, Lehmann EL. Estimates of location based Cardiol. 2005;46:1855-1862. kum (n=40), R. Erbel, MD, Frankfurt (n=8), E. on rank tests. Ann Math Statist. 1963;34:598-611. 33. Eisenberg DM, Davis RB, Ettner SL, et al. Trends Ja¨ hnchen, MD, Bad Krotzingen (n=9), U. Julius, MD, 14. O’Neill RT. Statistical analysis of adverse event data in alternative medicine use in the United States, Dresden (n=13), W. Ko¨ ster, MD, Freiburg im Breis- from clinical trials. Drug Inf J. 1987;21:9-30. 1990-1997. JAMA. 1998;280:1569-1575.

Downloaded From: https://jamanetwork.com/ on 09/23/2021