Effect of Simvastatin/Ezetimibe 10/10 mg Versus Simvastatin 40 mg on Serum Vitamin D Levels

Abstract

Backround:

Low levels of 25-hydroxyvitamin D (25(OH)VitD) have been recognized as an emerging cardiovascular disease (CVD) risk factor. Statins are reported to increase 25(OH)VitD concentration. Animal studies suggest that ezetimibe is a moderate inhibitor of intestinal 25(OH)VitD absorption, but its effect in humans is unknown.

Aim:

To investigate whether combined treatment with simvastatin/ezetimibe 10/10 mg would increase 25(OH)VitD levels compared to simvastatin 40 mg monotherapy in patients with primary hypercholesterolemia.

Methods:

In a Prospective Randomized Open-label Blinded End point study, 50 patients with primary hypercholesterolemia received either simvastatin/ezetimibe 10/10 mg (n = 25) or simvastatin 40 mg (n = 25) daily for 3 months. The primary end point was between-group difference in the change of serum 25(OH)VitD levels.

Results:

Simvastatin/ezetimibe 10/10 mg was associated with a 36.7% increase in 25(OH)VitD serum levels (from 6.8 to 9.3 ng/mL, P = .000), while simvastatin 40 mg was associated with a 79.1% increase (from 6.7 to 12.0 ng/mL, P = .008). The increase in 25(OH)VitD levels in the simvastatin 40 mg group was significantly greater compared to that in the simvastatin/ezetimibe 10/10 mg group (P = .04). Both groups exhibited similar reductions in low-density lipoprotein cholesterol (LDL-C) levels.

Conclusion:

For similar LDL-C lowering simvastatin 40 mg is associated with greater increase in 25(OH)VitD compared to simvastatin/ezetimibe 10/10 mg. Whether this difference is relevant in terms of CVD risk reduction is unknown.

Keywords

vitamin D cardiovascular disease simvastatin ezetimibe

Introduction

An emerging amount of evidence suggests that low serum levels of vitamin D (VitD) may be associated with increased risk of cardiovascular disease (CVD).^{1
–3}

Statins may increase serum VitD levels in the context of their pleiotropic effects. Specifically, lovastatin,⁴ simvastatin,⁵ atorvastatin,⁶ and rosuvastatin^{7
–9} may be associated with increases in 25-hydroxyvitamin D (25(OH)VitD) levels, whereas a recent randomized controlled trial found no effect of simvastatin on VitD serum levels.¹⁰

Ezetimibe, a cholesterol absorption inhibitor, is mainly used in combination with statins. Its effect on VitD metabolism has not been studied in humans. Animal experiments indicate that ezetimibe may be a moderate inhibitor of the intestinal 25(OH)VitD absorption.¹¹

Simvastatin 40 mg and simvastatin/ezetimibe 10/10 mg result in low-density lipoprotein cholesterol (LDL-C) reductions of approximately the same magnitude,^12,13 but whether they differ in their effects on serum 25(OH)VitD concentration is unknown. In this context, we aimed to compare the effect of simvastatin 40 mg versus simvastatin/ezetimibe 10/10 mg on serum 25(OH)VitD levels in patients with primary hypercholesterolemia. The primary end point was the between-group difference in the change of 25(OH)VitD levels following 3 months of treatment.

Methods

Patients

Consecutive patients with primary hypercholesterolemia (n = 50) attending the Outpatient Lipid and Obesity Clinic of the University Hospital of Ioannina, Ioannina, Greece participated in the present study. Inclusion criteria were LDL-C levels above those recommended by the National Cholesterol Education Program Adult Treatment Panel III based on each patient risk factors following a 3-month period of lifestyle changes.¹⁴ Exclusion criteria were known CVD, symptomatic carotid artery disease, peripheral arterial disease, abdominal aortic aneurysm, diabetes, triglycerides (TGs) >500 mg/dL (5.65 mmol/L), renal disease (serum creatinine levels >1.6 mg/dL; 141.4 μmol/L), hypothyroidism (thyroid stimulating hormone >5 IU/mL), liver disease (alanine aminotranferase and/or aspartate aminotranferase levels >3-fold upper limit of normal in 2 consecutive measurements), neoplasia as well as clinical and laboratory evidence of an inflammatory or infectious condition. Patients with hypertension were included in the study if they were on stable medication for at least 3 months and their blood pressure was adequately controlled (no change in their treatment was allowed during the study). Patients currently taking lipid-lowering drugs or having stopped them less than 4 weeks before study entry were excluded.

This study had a Prospective Randomized Open-label Blinded End point design. The patients were randomly allocated to receive open-label simvastatin 40 mg or simvastatin/ezetimibe 10/10 mg for 12 weeks. Randomization was performed by means of a computer-generated sequence of random numbers.

All participants were of Greek origin, had similar dietary habits with usual calcium content, and comparable amounts of sun exposure, since none was institutionalized or homebound or had a special dress code. In order to exclude the confounding effect of the seasonal variation of serum 25(OH)VitD levels, we collected all specimens from early autumn to late winter. During this period, sunshine in Greece is of the same duration and therefore we can largely exclude the confounding sunlight effect on 25(OH)VitD levels. Additionally, none of the participants consumed drugs that could interfere with VitD metabolism (ie, osteoporosis treatment) or VitD supplements. Compliance with study medication was assessed at week 12 by tablet counts; patients were considered compliant if they took 80% to 100% of the prescribed number of tablets.

All participants gave their written informed consent before any clinical or laboratory evaluations were performed. The study protocol was approved by the ethics committee of the University Hospital of Ioannina and adheres to the principles of the declaration of Helsinki.

Biochemical Parameters

Blood samples were obtained after a 14-hour overnight fast and were blindly assessed with regard to treatment allocation. All laboratory measurements were performed at the Laboratory of Biochemistry of the University Hospital of Ioannina. Concentrations of fasting plasma glucose, total cholesterol, TGs, and high-density lipoprotein cholesterol (HDL-C) were determined enzymatically on the Olympus AU 600 clinical chemistry analyzer (Olympus Diagnostica, Hamburg, Germany). The HDL-C level was determined using a direct assay (Olympus Diagnostica). The LDL-C was calculated with the Friedewald formula. Apolipoproteins (apo) A-I and apoB were measured using a Behring Nephelometer BN100 and with reagents (antibodies and calibrators) from Dade Behring Holding GmbH (Liederbach, Germany). The apoA-I and apoB assays were calibrated according to the International Federation of Clinical Chemistry standards. Serum creatinine, liver enzymes, and muscle enzymes as well as thyroid function tests were measured using conventional methods.

Serum 25(OH)VitD was determined quantitatively by an enzyme immunoassay method using the reagents from DRG Instruments GmbH kit (Germany), in the Child Health Research Laboratory of University of Ioannina. The sensitivity of the method was 1.28 ng/mL (3.2 nmol/L) and the intra- and interassay variation was 7% for each at the level of 28.8 and 33.6 ng/mL (72 and 84 nmol/L), respectively.

Statistical Analysis

We used G*Power 3.0.10 to calculate the sample size. Based on our previous study,⁹ we estimated that simvastatin would result in an at least 50% increase in 25(OH)VitD, while we hypothesised that the addition of ezetimibe would ameliorate the statin-associated increase in serum VitD levels. Power analysis revealed that a sample size of 25 patients per group would give an 80% power to detect a 15% difference between groups at an α level of 0.05.

Preliminary analyses were performed to ensure no violation of the assumptions of normality and linearity. The Kolmogorov-Smirnoff test was used to evaluate whether each variable followed a Gaussian distribution and logarithmic transformations were accordingly performed. Data are presented as mean ± standard deviation and median (range) for parametric and nonparametric data, respectively. The differences in study parameters between baseline and posttreatment values were evaluated by paired samples t test (or Wilcoxon rank test for non-Gaussian variables). Analysis of covariance, adjusted for baseline values, was used for the comparisons between treatment groups. Significance was defined as P < .05. Analyses were performed using the SPSS 18.0 statistical package for Windows (SPSS Inc, 1989-2004, Chicago, Illinois).

Results

A total of 50 patients (23 men and 27 women, aged 56 ± 10 years) were enrolled and randomized to receive either simvastatin/ezetimibe 10/10 mg (n = 25) or simvastatin 40 mg (n = 25). None of the participants dropped out, while compliance was >80% in all participants. The baseline clinical and laboratory characteristics of study participants are listed in Table 1. Of note, both groups presented with low baseline 25(OH)VitD levels at 6.8 and 6.7 ng/mL, respectively. No significant difference in baseline parameters was found between the 2 groups.

Table 1.

Baseline Characteristics of Study Participants.^a

Characteristics	Simvastatin/Ezetimibe 10/10 mg	Simvastatin 40 mg	P
N (males/females)	25 (12/13)	25 (11/14)	NS
Age, years	54 ± 12	58 ± 8	NS
Current smokers, %	41	38	NS
Body weight, kg	81 ± 10	78 ± 6	NS
BMI, kg/m²	29 ± 3	28 ± 3	NS
WC, cm	103 ± 8	102 ± 7	NS
SBP, mm Hg	123 ± 11	128 ± 12	NS
DBP, mm Hg	78 ± 7	80 ± 5	NS
TC, mg/dL	253 ± 53	266 ± 38	NS
LDL-C, mg/dL	176 ± 48	177 ± 32	NS
TGs, mg/dL	109 (58-194)	104 (73-210)	NS
HDL-C, mg/dL	59 ± 11	62 ± 12	NS
Apo-AI, mg/dL	150 ± 24	158 ± 25	NS
ApoB, mg/dL	108 ± 27	117 ± 23	NS
Glucose, mg/dL	96 ±12	99 ± 13	NS
25(OH)VitD, ng/mL	6.8 (0.2-16)	6.7 (0.5-24)	NS

Abbreviations: BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; WC, waist circumference; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; TGs, triglycerides; Apo, apolipoprotein; 25(OH)Vit D, 25-hydroxyvitamin D; NS, not significant.

^aTo convert the values of triglycerides to mmol/L multiply by 0.01129. To convert the values of cholesterol to mmol/L multiply by 0.02586. To convert the values of glucose to mmol/L multiply by 0.05551. To convert the values of 25(OH)Vit D to nmol/L multiply by 2.5.

In the simvastatin/ezetimibe 10/10 mg group 25(OH)VitD serum levels increased by 36.7% (from 6.8 to 9.3 ng/mL, P = .000), while in the simvastatin 40 mg group a 79.1% increase (from 6.7 to 12.0 ng/mL, P = .008) was noticed (Table 2). The increase in 25(OH)VitD levels was significantly greater in the simvastatin 40 mg group compared to simvastatin/ezetimibe 10/10 mg group (P = .04).

Table 2.

Serum Metabolic Parameters at Baseline and After 3 Months of Treatment.

	Baseline	3 Months	P vs Baseline	Change, %
TC, mg/dL
Simva/eze 10/10 mg	253 ± 53	169 ± 29	.000	−33.2
Simva 40 mg	266 ± 38	175 ± 27	.000	−34.2
TGs, mg/dL
Simva/eze 10/10 mg	109 (58-194)	92 (58-166)	.01	−15.6
Simva 40 mg	104 (73-210)	94 (61-160)	.01	−9.6
HDL-C, mg/dL
Simva/eze 10/10 mg	59 ± 11	60 ± 11	NS	+1.6
Simva 40 mg	62 ± 12	64 ± 12	NS	+1.6
LDL-C, mg/dL
Simva/eze 10/10 mg	177 ± 32	92 ± 19	.000	−48.0
Simva 40 mg	176 ± 48	97 ± 28	.000	−44.8
ApoAI, mg/dL
Simva/eze 10/10 mg	150 ± 24	152 ± 21	NS	+1.3
Simva 40 mg	158 ± 25	164 ± 24	NS	+3.7
ApoB, mg/dL
Simva/eze 10/10 mg	117 ± 23	69 ± 13	.000	−41.0
Simva 40 mg	108 ± 27	66 ± 17	.000	−38.8
Glucose, mg/dL
Simva/eze 10/10 mg	96 ±12	95 ± 12	NS	−1.0
Simva 40 mg	99 ± 13	105 ± 15	NS	+6.0
25(OH)VitD (ng/mL)
Simva/eze 10/10 mg	6.8 (1.0-16.0)	9.3 (2.1-21.6)	.000	+36.7^a
Simva 40 mg	6.7 (2.8-24.0)	12 (3.8-30.8)	.008	+79.1

Abbreviations: Simva/eze, simvastatin/ezetimibe; Simva, simvastatin; TC, total cholesterol; TGs, triglycerides; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol; Apo, apolipoprotein; 25(OH)Vit D, 25-hydroxyvitamin D.

^a P = .04 versus Simva 40 mg group.

Lipid changes were similar between the 2 groups as summarized in Table 2.

Discussion

We showed that for the same degree of LDL-C lowering simvastatin 40 mg was associated with a more than double increase in 25(OH)VitD levels (79.1%) compared to simvastatin/ezetimibe 10/10 mg (36.7%) in patients with primary hypercholesterolemia.

The effect of several statins on VitD metabolism has previously been investigated. In particular, lovastatin,⁴ atorvastatin,⁶ and rosuvastatin^7,8 appear to increase 25(OH)VitD levels. Of note, we have shown that rosuvastatin 40 mg as monotherapy and rosuvastatin 10 mg plus fenofibrate 200 mg or omega-3 fatty acids 2 g were associated with significant and similar increases in the 25(OH)VitD levels (+53%, +64%, and +61%, respectively).⁹ On the other hand, no significant change in 25(OH)VitD levels has been observed with fluvastatin treatment.⁸ Data are inconclusive for simvastatin. A small study showed that treatment with 10 and 20 mg raised plasma levels of 25(OH)VitD and 1,25(OH)₂VitD (the active metabolite) in a dose-dependent ratio.⁵ On the contrary, a recent randomized controlled trial failed to show any effect of simvastatin 40 mg treatment compared to placebo for 1 year on 25(OH)VitD levels in healthy postmenopausal women with osteopenia but without known hyperlipidemia.¹⁰ The authors comment that the observed increase in 25(OH)VitD levels after statin treatment found in previous uncontrolled studies is due to unmeasured changes in indices affecting 25(OH)VitD levels. These may include the coincidence of lifestyle changes in relation to the administration of statin drugs (ie, increased exercise and sun exposure, consumption of food items rich in VitD or supplements).¹⁰ In this study, treatment with simvastatin 40 mg for 3 months resulted in a 79.1% increase in serum 25(OH)VitD levels. Of note, most of our patients were VitD deficient at baseline. The reasons are unknown but may be related to limited sun exposure during autumn and winter. On the contrary, in Rejnmark’s et al study only 3 patients were VitD deficient (25(OH)VitD <10 ng/mL).¹⁰ It can be assumed that the lower the 25(OH)VitD baseline levels, the greater the observed increase following statin treatment. The high incidence of VitD deficiency even in countries with long duration of sunlight may be attributed to genetical differences. These genetical differences may also affect the effect of statins on VitD status.

Several potential mechanisms have been proposed to explain the observed increase in 25(OH)VitD with statin therapy. Since 25(OH)VitD is catabolized in liver and intestine by cytochrome P450 3A4 (CYP3A4),¹⁵ which also metabolizes many statins (along with CYP3A5), the competition in this common catabolic pathway could be the cause for the increased 25(OH)VitD levels under statin treatment. However, increases have also been seen with rosuvastatin which does not use CYP3A4. Another mechanism is that when 3-hydroxy-3 methylglutaryl coenzyme A reductase is inhibited by statins, 7-dehydrocholesterol levels, the common precursor of cholesterol and 25(OH)VitD, may increase.¹⁶ As a result, there is an abundance of cutaneous substrate, which after exposure to sunlight undergoes photolysis giving rise to synthesis of previtamin D₃ and consequently to increased production of 25(OH)VitD.

The more than double 25(OH)VitD increase with simvastatin 40 mg compared to simvastatin/ezetimibe 10/10 mg found in this study could be attributed either to a dose-dependent effect of simvastatin on raising 25(OH)VitD level or to an amelioration of VitD intestinal absorption by ezetimibe or both. It has not been clarified whether there is a dose-dependent effect of statins on 25(OH)VitD levels or what effect could ezetimibe monotherapy have on serum 25(OH)VitD levels in humans.

The lipid-lowering effect of ezetimibe is mediated through a specific inhibition of Niemann-Pick C1 Like 1 (NPC1L1) cholesterol transporter, which was recently shown to also be involved in 25(OH)VitD intestinal absorption.¹¹ Moreover, in vitro experiments indicated that ezetimibe inhibited 25(OH)VitD uptake in human colon carcinoma (Caco-2) cells and human embryonic kidney cells.¹¹ Also, in vivo uptake experiments demonstrated that ezetimibe decreased both medium and distal intestinal fragment 25(OH)VitD contents, but these differences were not significant.¹¹ Similarly, a previous study in rats found that VitD absorption was lower in the ezetimibe group, although the difference remained nonsignificant.¹⁷ Subsequently, it has been assumed that the involvement of NPC1L1 in 25(OH)VitD transportation in vivo may be moderate.¹¹ Notably, the contribution of intestinal absorption to serum VitD levels, when no supplements are taken, is relatively small since 80% to 90% of VitD derives from endogenous production in the skin.¹ In line, a recent study reported a negative but nonsignificant effect on bone mineral density in patients receiving ezetimibe for 1 year.¹⁸

The increase in 25(OH)VitD levels may represent a novel pleiotropic effect of statins. Ezetimibe has been found to either enhance or abrogate the pleiotropic effects of statins,¹⁹ but its effect on 25(OH)VitD levels is largely unstudied. Low levels of 25(OH)VitD have been recognized as an independent cardiovascular²⁰ and cerebrovascular disease²¹ risk factor. We have previously shown that in patients with metabolic syndrome, low 25(OH)VitD was indirectly associated with higher levels of small-dense LDL-C but not with lipoprotein-associated phospholipase A2 or high-sensitivity C-reactive protein.²² The VitD deficiency may be negatively associated with survival,²³ while supplementation may decrease mortality.²⁴

As VitD deficiency is very common, even in countries with increased sunlight, the importance of investigating the effect of lipid-lowering treatments on 25(OH)VitD levels in various populations increases.

Study Limitations and Strengths

There are certain limitations in this study. We included no group receiving monotherapy with ezetimibe, as statins are first-line lipid-lowering drugs. Also, we did not include a 10 mg simvastatin group as current guidelines suggest an at least 40% decrease in LDL-C, which can only be achieved with 40 mg of simvastatin. Additional limitations include the open-label design and the relatively short period of follow-up (3 months). The number of patients in each treatment group is rather small. However, this was an adequately powered study based on a priori power calculations.

Also, this is a clinically relevant study, since we tested the efficacy of equivalent treatments in terms of LDL-C lowering used in everyday clinical practice on 25(OH)VitD levels. Additionally, all comparisons were adjusted for baseline levels and end points were blindly assessed.

Conclusions

For similar, LDL-C lowering simvastatin 40 mg is associated with a more than double increase in serum 25(OH)VitD levels compared to simvastatin/ezetimibe 10/10 mg. Whether this difference is relevant in terms of CVD risk reduction is unknown.

Footnotes

Authors’ Note

The authors Evangelos N. Liberopoulos and Stefania E. Makariou contributed equally to this work.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Makariou

Liberopoulos

Elisaf

Challa

. Novel roles of vitamin D in disease: what is new in 2011? Eur J Intern Med. 2011;22(4):355–362.

Nadir

Szwejkowski

Witham

. Vitamin D and cardiovascular prevention. Cardiovasc Ther. 2010;28(4):e5–12.

Florentin

Elisaf

Mikhailidis

Liberopoulos

. Vitamin D and metabolic syndrome: is there a link? Curr Pharm Des. 2011;16(30):3417–3434.

Wilczek

Sobra

Justova

. [Iatropathogenic effect of Mevacor on vitamin D metabolism]. Cas Lek Cesk. 1989;128(40):1254–1256.

Wilczek

Sobra

Ceska

. [Monitoring plasma levels of vitamin D metabolites in simvastatin (Zocor) therapy in patients with familial hypercholesterolemia]. Cas Lek Cesk. 1994;133(23):727–729.

Perez-Castrillon

Vega

Abad

. Effects of atorvastatin on vitamin D levels in patients with acute ischemic heart disease. Am J Cardiol. 2007;99(7):903–905.

Yavuz

Ertugrul

Cil

. Increased levels of 25 hydroxyvitamin D and 1,25-dihydroxyvitamin D after rosuvastatin treatment: a novel pleiotropic effect of statins. Cardiovasc Drugs Ther. 2009;23(4):295–299.

Ertugrul

Yavuz

Cil

. STATIN-D Study: comparison of the influences of rosuvastatin and fluvastatin treatment on the levels of 25 hydroxyvitamin D. Cardiovasc Ther. 2011;29:146–152.

Makariou

Liberopoulos

Agouridis

Challa

Elisaf

. Effect of rosuvastatin monotherapy and in combination with fenofibrate or omega-3 fatty acids on serum vitamin D levels. J Cardiovasc Pharmacol Ther. 2012;17(4):382–386.

10.

Rejnmark

Vestergaard

Heickendorff

Mosekilde

. Simvastatin does not affect vitamin d status, but low vitamin d levels are associated with dyslipidemia: results from a randomised, controlled trial. Int J Endocrinol. 2010;2010:957174.

11.

Reboul

Goncalves

Comera

. Vitamin D intestinal absorption is not a simple passive diffusion: evidences for involvement of cholesterol transporters. Mol Nutr Food Res. 2011;55(5):691–702.

12.

Pearson

Ballantyne

Sisk

. Comparison of effects of ezetimibe/simvastatin versus simvastatin versus atorvastatin in reducing C-reactive protein and low-density lipoprotein cholesterol levels. Am J Cardiol. 2007;99(12):1706–1713.

13.

Bays

Ose

Fraser

. A multicenter, randomized, double-blind, placebo-controlled, factorial design study to evaluate the lipid-altering efficacy and safety profile of the ezetimibe/simvastatin tablet compared with ezetimibe and simvastatin monotherapy in patients with primary hypercholesterolemia. Clin Ther. 2004;26(11):1758–1773.

14.

Grundy

Cleeman

Merz

. Implications of recent clinical trials for the National Cholesterol Education Program Adult Treatment Panel III guidelines. Circulation. 2004;110(2):227–239.

15.

Hashizume

Shuhart

. Intestinal and hepatic CYP3A4 catalyze hydroxylation of 1alpha,25-dihydroxyvitamin D(3): implications for drug-induced osteomalacia. Mol Pharmacol. 2006;69(1):56–65.

16.

Guryev

Carvalho

Usanov

Gilep

Estabrook

. A pathway for the metabolism of vitamin D3: unique hydroxylated metabolites formed during catalysis with cytochrome P450scc (CYP11A1). Proc Natl Acad Sci U S A. 2003;100(25):14754–14759.

17.

van Heek

Farley

Compton

. Ezetimibe potently inhibits cholesterol absorption but does not affect acute hepatic or intestinal cholesterol synthesis in rats. Br J Pharmacol. 2003;138(8):1459–1464.

18.

Sertbas

Ersoy

Ayter

Gultekin Tirtil

Kucukkaya

. Ezetimibe effect on bone mineral density and markers of bone formation and resorption. J Investig Med. 2010;58(2):295–297.

19.

Kalogirou

Tsimihodimos

Elisaf

. Pleiotropic effects of ezetimibe: do they really exist? Eur J Pharmacol. 2010;633(1-3):62–70.

20.

Kendrick

Targher

Smits

Chonchol

25-Hydroxyvitamin D deficiency is independently associated with cardiovascular disease in the Third National Health and Nutrition Examination Survey. Atherosclerosis. 2009;205(1):255–260.

21.

Makariou

Michel

Tzoufi

Challa

Milionis

HJ.

Vitamin D and stroke: promise for prevention and better outcome [published online June 22, 2012]. Curr Vasc Pharmacol.

22.

Makariou

Liberopoulos

Florentin

. The relationship of vitamin D with non-traditional risk factors for cardiovascular disease in subjects with metabolic syndrome. Arch Med Sci. 2012;8(3):437–443.

23.

Melamed

Michos

Post

Astor

25-hydroxyvitamin D levels and the risk of mortality in the general population. Arch Intern Med. 2008;168(15):1629–1637.

24.

Autier

Gandini

. Vitamin D supplementation and total mortality: a meta-analysis of randomized controlled trials. Arch Intern Med. 2007;167(16):1730–1737.