Changes in Myocardial PerfusionAbnormalities by Positron EmissionTomography After Long-term, IntenseRisk Factor ModificationK. Lance Gould, MD; Dean Ornish, MD; Larry Scherwitz, PhD; Shirley Brown, MD; R. Patterson Edens, PhD;Mary Jane Hess, RN; Nizar Mullani; Leonard Bolomey; Frank Dobbs, PhD; William T. Armstrong, MD;Terri Merritt, MS; Thomas Ports, MD; Stephen Sparler, MA; James Billings, PhD

Objective.\p=m-\Toquantify changes in size and severity of myocardial perfusionabnormalities by positron emission tomography (PET) in patients with coronary ar-

tery disease after 5 years of risk factor modification.Design.\p=m-\Randomizedcontrolled trial.Setting.\p=m-\Outpatientcommunity setting.Intervention.\p=m-\Randomizationof patients to risk factor modification consisting

of very low-fat vegetarian diet, mild to moderate exercise, stress management, andgroup support (experimental group, n=20) or to usual care by their own physicians,consisting principally of antianginal therapy (control group, n=15).

Main Outcome Measures.\p=m-\Quantitativecoronary arteriography and PET atbaseline and 5 years after randomization. Automated, objective measures of sizeand severity of perfusion abnormalities on rest-dipyridamole PET images and ofstenosis severity on arteriograms were made by computer algorithms.

Results.\p=m-\Sizeand severity of perfusion abnormalities on dipyridamole PET im-ages decreased (improved) after risk factor modification in the experimental groupcompared with an increase (worsening) of size and severity in controls. The per-centage of left ventricle perfusion abnormalities outside 2.5 SDs of those of normalpersons (based on 20 disease-free individuals) on the dipyridamole PET image ofnormalized counts worsened in controls (mean \m=+-\SE,+10.3%\m=+-\5.6%)and improvedin the experimental group (mean \m=+-\SE,\m=-\5.1%\m=+-\4.8%)(P=.02); the percentage of leftventricle with activity less than 60% of the maximum activity on the dipyridamole PETimage of normalized counts worsened in controls (+13.5%\m=+-\3.8%) and improvedin the experimental group(\m=-\4.2%\m=+-\3.8%)(P=.002); and the myocardial quadrant onthe PET image with the lowest average activity expressed as a percentage of maxi-mum activity worsened in controls (\m=-\8.8%\m=+-\2.3%)and improved in the experimentalgroup (+4.9%\m=+-\3.3%)(P=.001 ). The size and severity of perfusion abnormalities on

resting PET images were also significantly improved in the experimental group as

compared with controls. The relative magnitude of changes in size and severity ofPET perfusion abnormalities was comparable to or greater than the magnitude ofchanges in percent diameter stenosis, absolute stenosis lumen area, or stenosisflow reserve documented by quantitative coronary arteriography.

Conclusions.\p=m-\Modestregression of coronary artery stenoses after risk factormodification is associated with decreased size and severity of perfusion abnor-malities on rest-dipyridamole PET images. Progression or regression of coronaryartery disease can be followed noninvasively by dipyridamole PET reflecting theintegrated flow capacity of the entire coronary arterial circulation.

(JAMA. 1995;274:894-901)

From the Department of Medicine (Drs Gould, Edens,and Dobbs and Messrs Mullani and Bolomey) and thePositron Diagnostic and Research Center (Ms Hess) ofthe University of Texas Medical School at Houston; thePreventive Medicine Research Institute, Sausalito, Calif(Drs Ornish, Scherwitz, and Billings, Ms Merritt, and Mr

Sparler); California Pacific Medical Center, San Francisco(Drs Ornish and Armstrong); and the University of Cali-fornia at San Francisco (Drs Ornish, Brown, and Ports).

Reprint requests to Division of Cardiology, Room4.258MSB, 6431 Fannin, University of Texas MedicalSchool at Houston, Houston, TX 77030 (Dr Gould).

LIPID-LOWERING trials in patientswith coronary atherosclerosis have dem¬onstrated no progression or partial re¬

gression of coronary artery stenoses bycoronary arteriography compared withprogression of stenosis severity in con¬trols treated with standard therapy.1"24However, these regression trials shareseveral common Mmitations, including thefollowing: the degree of regression was

anatomically modest, being 5% to 10%diameter stenosis units;percentdiameterstenosis is poorly related to flow capacityof coronary arteries or coronary flow re¬

serve4·25"32; progression or regression ofcoronary artery stenoses may be associ¬ated with complex shape changes or re¬

modeling,4 in which the integrated hemo-dynamic effects ofpercent narrowing, ab¬solute arterial lumen area, and length arenot accounted for by any single geometricdimension such as percent stenosis4,2M2;and quantifying single focal stenoses on

coronary arteriograms does not accountfor multiple stenoses, diffuse atheroscle¬rosis, or the associated vasomotor abnor¬malities frequently present33"35 and there¬fore does not reflect the perfusion capac¬ity of the entire integrated coronaryarterial circulation affected by athero¬sclerosis.

Although several lipid-lowering trialshave demonstrated only modest reduc¬tion in percent diameter narrowing intreated vs control groups, these studieshave reported a proportionately greaterdecrease in cardiac events in treatedgroups.* In view of current widespreadinterest in the noninvasive managementof coronary atherosclerosis by choles¬terol-lowering drugs, diet, exercise, andbehavioral interventions, the potentialimpact of these trials on clinical practiceis substantial, but data on changes in

*References 2, 5, 11, 14, 15, 17-19, 22, 24.

Downloaded From: http://jama.jamanetwork.com/ on 07/27/2012

myocardial perfusion abnormalities cor¬

responding to the observed artériograph¬ie changes are lacking. Accordingly, wemeasured the size and severity of myo¬cardial perfusion abnormalities by posi¬tron emission tomography (PET) at restand after dipyridamole stress at baselineand at an average follow-up of 5 years inpatients with coronary artery disease(CAD) randomized to the control groupor to the experimental group undergoingintense risk factor modification.METHODSStudy Patients

Patients selected for participation inthe risk modification program were menand women, 41 to 70 years old, who hadCAD documented by arteriography, hadno recent myocardial infarction, werenot taking lipid-lowering drugs, had leftventricular ejection fraction greater than25%, and resided in the greater SanFrancisco area of California. Individu¬als in each group gave informed consentand were randomized after we obtainedbaseline quantitative coronary arteri¬ography but before other baseline mea¬surements.1,4

Modification of risk factors has beendescribed previously1,4 and consisted ofa low-cholesterol (<5 mg/d), low-fat(<10% of total energy intake) vegetar¬ian diet with 15% protein and 75% com¬

plex carbohydrate augmented with vi¬tamin B12.1,4 Patients stopped smoking,practiced stress management techniquesfor 1 hour daily, and participated in mildto moderate aerobic exercise 3 hoursper week. Adherence to the programwas quantified by a score reflecting a

3-day diet diary, an exercise and stressmanagement diary, and confirmation ofsmoking cessation by plasma cotinineconcentration. A total score of 1.0 indi¬cated 100% adherence to the recom¬mended lifestyle changes, whereas ascore of zero indicated no adherence. Ascore greater than 1.0 indicated that pa¬tients did more than was recommended.

Quantitative Coronary ArteriographyInitial and follow-up coronary arte¬

riograms were performed using the stan¬dard percutaneous femoral approach.Detailed records of the view angles, x-

ray exposures, image intensifier, x-raytube, patient distances, and referencecatheter dimensions were maintained.These same characteristics were usedin follow-up arteriograms to reproduceviews and exposures as closely as pos¬sible. Angiograms were analyzed si¬multaneously in pairs by a technicianunaware of clinical data or group as¬

signment using automated border rec¬

ognition and stenosis analysis techniquesto avoid the potential bias, imprecision,

and uncertainties of visual interpreta¬tion. The primary stenosis dimensionsmeasured by the automated artér¬iographie program previously de¬scribed1·4·25"29 include proximal diameterand cross-sectional lumen area, minimaldiameter and area, distal diameter andarea, the exit angle and exit shape ef¬fects, and calculated measures of sever¬

ity, including percent diameter steno¬sis, percent area narrowing, integratedlength-area effects, and stenosis flowreserve. This program has been vali¬dated in three separate experimentalstudies,26"28 applied in humans,1·4·33·36"40 andused routinely in approximately 5000clinical arteriograms in our laboratory.The 95% confidence interval for steno¬sis flow reserve is ±0.66, with a repro-ducibility ofprimary dimensions of ±3%to ±5%.1·4'25"29PET Imaging

As previously described,36"43 PET im¬aging of myocardial perfusion was per¬formed at rest and after administrationof dipyridamole. Fluoroscopy was usedto mark the cardiac borders for patientpositioning. The PET imaging was per¬formed using the University of Texas-designed cesium fluoride, multislicetomograph with a reconstructed reso¬lution of 12 mm full width at half maxi¬mum (FWHM) in plane and 14 mmFWHM axially. Transmission imageswere obtained to correct for photon at¬tenuation using the segmented atten¬uation correction method.44 Emissionimages were obtained following intra¬venous injection of 18 mCi of cyclotron-produced nitrogen 13 (13N) ammonia36"43or after 40 to 50 mCi of generator-pro¬duced rubidium 82, depending on avail¬ability of a generator (Bristol-Myers-Squibb, Princeton, NJ). The radionuclideused at baseline was also used in thefinal follow-up PET study.

At 5 to 10 minutes after the first doseof rubidium 82 or at 40 minutes afteradministration of the first dose of am¬monia to allow for decay of the firstradionuclide dose, dipyridamole (0.142mg/kg per minute) was infused intra¬venously for 4 minutes. Two minutesafter the infusion was complete, 25% ofpredetermined maximal handgrip washeld by the patient with one hand for 4minutes. Two minutes after the start ofthe handgrip, a second dose of the sameamount of the same radionuclide was

injected intravenously, with the hand¬grip continued for 2 minutes. The PETimaging was then repeated by the same

protocol as for the resting study.Transmission scans contained 100 mil¬

lion to 150 million counts. Emission scansof the whole heart contained 20 millionto 40 million counts for 15 to 20 mCi of

intravenous 13N ammonia and 15 millionto 25 million counts for 40 to 50 mCi ofrubidium 82.

Baseline and final dipyridamole im¬ages were displayed together for directside-by-side comparison and were visu¬ally interpreted by three independentreaders blinded to patient identification,quantitative data, treatment group, andall clinical information. The image pairswere randomized and were indepen¬dently read by the three blinded read¬ers as worse, better, unchanged, or show¬ing mixed changes for different direc¬tional changes in the four quadrants andapex as described below. The visualchange from baseline to final PET im¬age was scored as no change, zero, or

improved on a scale of +1, correspond¬ing to definite but mild improvement, to+3 for maximal improvement, or -1 fordefinite but mild worsening to -3 formaximal worsening. A ±1 change cor¬

responded to a one-color difference (eg,red vs yellow, representing a step incount density of 15% of maximum) be¬tween baseline and final PET scans; a±2 change corresponded to a two-colordifference (eg, red vs green); and a ±3change corresponded to a three-colordifference (eg, red vs blue) for a colorscale as shown in Figure 1. Any changeof +1 or greater was considered im¬proved. Any change of -1 or less wasconsidered worse. Any study with a

change in one part of the heart of +1 or

greater and a change in another part of-1 or less was considered to be a mixedchange.

All measurements ofseverity and sizeof perfusion defects were carried outby computer analysis as previously de¬scribed16,25'38"40 without operator inter¬pretation, visual drawing of borders,visual location of defect, or any otheroperator interaction, background sub¬traction, image enhancement, or otherimage manipulation. A three-dimen¬sional restructuring algorithm generatedtrue short- and long-axis views fromPET transaxial cardiac images andthree-dimensional views of the left ven¬tricle as viewed from the right (septal),anterior, left lateral, and inferior views,reflecting relative regional activity dis¬tribution at rest and after dipyridamolestress. Polar maps and three-dimensionalviews are divided into fixed sections con¬

sisting of septal, anterior, lateral, andinferior quadrants and an apical seg¬ment of the polar or three-dimensionaldisplay.16·25,38"40 The baseline PET scanwas compared with the final PET scan

by automated analysis of differences insize and severity ofperfusion abnormali¬ties, also without operator judgment,selection of areas of interest, or imagemanipulation.

Downloaded From: http://jama.jamanetwork.com/ on 07/27/2012

Figure 1.—Top left, Orientation of three-dimensional positron emission tomog¬raphy (PET) images. Top right, Example of PET after dipyridamole at baseline(upper row) and after 5 years as a control patient (lower row) in right (septal),anterior, left lateral, and inferior views. Relative radionuclide uptake is shownin a graded-color scale ranging from maximum activity (100%) in white down¬ward in 5% increments corresponding to the stepped color scale through red,yellow, green, blue, and black as minimum activity, shown in the color bars. Thefinal study (lower row) compared with baseline (upper row) shows worseningthroughout the heart, particularly interiorly, corresponding to progression ofthree-vessel disease by quantitative coronary arteriography. Bottom right, Ex¬ample of PET after dipyridamole at baseline (upper row) and after 5 years oflifestyle change (lower row) with the same views and color scale. The inferiorperfusion defect becomes less severe with better, more uniform perfusion in thedefect and in the rest of the heart corresponding to regression of three-vesseldisease by quantitative coronary arteriography.

The end points were the severity andsize of perfusion defects on PET imagesat rest and after dipyridamole stress,defined as follows: The end point lowestquadrant average was the average num¬ber of normalized counts for the quad¬rant having the lowest average activity,in which an anterior, septal, lateral, andinferior quadrant surround a centralapex area. The mean value for any givenquadrant with the lowest or minimumactivity was the quadrant that containedthe most severe perfusion defect. Thislowest quadrant average was deter¬mined for the PET image at rest afterdipyridamole stress. This end pointquantified the relative severity of theperfusion abnormality. For example, avalue of 65% would indicate that themean count value for the quadrant withthe lowest counts, and therefore con¬

taining the perfusion defect, would be65% of the normal maximum of 100%.

The end point percentage outside 2.5SDs was the size of the perfusion defectdetermined as the percentage of the car¬diac image outside 2.5 SDs of that ofnormal individuals (based on 20 disease-free persons) for the PET image at restor after dipyridamole stress. Because2.5 SDs includes 97.6% of the normaldistribution, there was only a 2.4%chance that normal values outside 2.5

SDs would be observed.The end point percentage with a ratio

less than 0.6 was a measure of combinedseverity and size ofperfusion abnormali¬ties determined as percentage of myo¬cardium with activity of less than 60% ofmaximum activity (100%) on the PETimage. This measurement gives the sizeof the defect characterized by the se¬

verity threshold of less than 60% of thenormal maximum of 100%, and there¬fore reflects the combined intensity andsize of the defect on the PET image. Avalue of less than 0.6 or less than 60% ofmaximum on the PET image is approxi¬mately 3 SDs below the normal mean of80% ±7% of maximum activity. Because3 SDs contain 99.7% of the normal dis¬tribution, there was less than a 0.3%chance that normal values would be ob¬served below 60% of maximum activity.Statistical Analysis

Changes in perfusion abnormalitiesfrom the baseline to the final PET were

compared between experimental andcontrol groups by unpaired, one-tailed ttest for continuous variables.45"48 Dataare reported as mean±l SEM unlessotherwise noted, reflecting the variabil¬ity of means in each group and calcu¬lated as the SD divided by the squareroot of n, where is the sample size of

the group. A one-tailed t test was jus¬tified on the grounds that improved life¬style changes would not be expected tomake perfusion abnormalities worse, as

specified at the beginning of the study.Two-tailed t tests are also reported asmore rigorous tests of significance.Analysis of variance was carried out us¬

ing the Bonferroni post hoc correctionalgorithm4548in Statviewsoftware (Aba¬cus Concepts Ine, Berkeley, Calif). Fordiscrete variables such as number or

percentage ofsubjects showing changesgreater than 1 SD from baseline values,significance of differences betweengroups was determined by Fisher's ex¬act test (one tailed).45"48

Sample size was determined in initialtrial design by power calculations fromestimated changes expected and vari¬ability ofquantitative artériographie endpoints, the primary end points as pre¬viously reported for 1-year follow-up.1·4In view of previous reports on changesin artériographie stenosis severity inthis1·4 and other trials,1"24 we report ar¬

tériographie changes only briefly in com¬

parison to PET results.At entry into the study, four patients

had one or more completely occludedcoronary arteries with extensive collat¬erals to viable myocardium. These pa¬tients demonstrated severe perfusion ab-

Downloaded From: http://jama.jamanetwork.com/ on 07/27/2012

Table 1.—Changes in Risk Factors*

Experimental Group Control GroupRisk Factor 5-y Follow-up Change Baseline 5-y Follow-up Change Pi

Systolic blood pressure, mm Hg 135±15 128±14t -7±13 137±27 127±13 -11 ±21 .69Dlastolic blood pressure, mm Hg 82±9 77±8± -5±9

Weight, kg 91±15 83±12§ -8±6 76±18 78±19 +2±5 <.001Cholesterol, mmol/L (mg/dL) 5.80±1.35

(225 ±53)4.50±0.85(175±33)§

-1.30±0.95(-50 ±37)

6.45±0.95(250 ±37)

5.95±0.85(230 ±33)

-0.50±1.10(-20±43)

.04

LDL cholesterol, mmol/L (mg/dL) 3.70±1.30(143±50)

2.65±0.90(103±35)§

-1.05±1.03(-41 ±40)

4.35±0.78(168±29)

3.80±0.90(148±35)

-0.50±1.00(-19±38)

.13

HDL cholesterol, mmol/L (mg/dL) 1.05±0.30(40±12)

0.90±0.30(35±11)

-0.15±0.35(-5±13)

1.35±0.55(53 ±22)

1.30±0.40(50±16)

-0.03±0.40(-1±15)

.26

Triglycérides, mmol/L (mg/dL) 2.58±1.34(228±119)

2.78±1.20(247±107)

+0.22±0.98(+19±87)

2.62±2.84(232 ±251)

2.20±0.76(194±67)

-0.42±2.82(-38 ±250)

.35

Adherencell 0.67±0.37 1.14±0.24§ +0.55±0.29 0.60±0.29

*LDL indicates low-density llpoprotein; and HDL, high-density lipoprotein. Numbers Indicate mean±1 SD unless otherwise indicated.fP for change, experimental group vs control group.tFor baseline vs experimental period, P=.03.§For baseline vs experimental period, P<.001.|| See text for explanation.

normalities and myocardial steal afterdipyridamole stress by PET. Myocar¬dial steal by PET after dipyridamolestress is diagnostic for substantial coro¬

nary collateralization as compared withcoronary arteriography.41 Areas of myo¬cardium showing myocardial steal wouldnot be expected to show decreased sizeor severity of perfusion abnormalitiescorresponding to stenosis regression be¬cause the artery is occluded and usuallydoes not reopen with lipid lowering. Myo¬cardial steal and severity of perfusiondefect after dipyridamole stress increasewith time as resting collateral supplyimproves to myocardium distal to theocclusion.49 Accordingly, image datawere analyzed both with and withoutthe four patients with myocardial steal.

RESULTSOf the 48 patients enrolled, five (25%)

of 20 patients in the control group andeight (29%) of 28 patients in the experi¬mental group dropped out or were un¬available for follow-up study, includingone control patient with bypass surgery,leaving 35 patients for analysis. Fifteenpatients (12 men and three women) werein the control group and 20 patients (allmen) were in the experimental group.Mean age of the experimental group was57±6 years and of the control group62 ±8 years.

Time from baseline control PET to thefinal PET study was 5.0±0.2 years(mean±l SEM) and was comparable forcontrol and experimental groups (5.1±0.2years vs 4.9±0.2 years, respectively).Baseline characteristics ofthe control andexperimental study groups were compa¬rable, as previously reported.1

Differences between the control andexperimental groups in the change frombaseline to average values during thefollow-up period for blood pressure,weight, serum cholesterol level, and ad¬herence to the exercise, diet, and stress

management program are shown inTable 1. At final follow-up there was nodifference in frequency, duration, or se¬

verity of angina pectoris between ex¬

perimental and control groups, mostlikely due to revascularization proce¬dures performed in the control group.Twenty patients (57%) were studied with13N ammonia at baseline and final PETstudies, and 15 patients (43%) were stud¬ied with rubidium 82 at baseline andfinal PET studies, with the same pro¬portion in the experimental and controlgroups. No systematic differences in thechanges from baseline to final study wereobserved between these two perfusionradiotracers in the experimental vs thecontrol group.

Figure 1, top right, shows the perfu¬sion images in a control patient afterdipyridamole at baseline (upper row) andat final follow-up 5 years later (lowerrow). In this control patient, the finalstudy in comparison to baseline shows amarked decrease of activity in the in¬ferior myocardium and a definite butless severe decrease in lateral, septal,anterior, and apical myocardium, reflect¬ing decreased perfusion capacity corre¬

sponding to progression of three-vesseldisease documented by quantitativecoronary arteriography. From a patientin the experimental group (bottom right)the final study in comparison to baselineshows increased activity, particularlyin the inferior myocardium but alsoin lateral, septal, anterior, and apicalmyocardium, reflecting increased flowcapacity and regression of three-vesseldisease by quantitative coronary arte¬riography. These examples illustratecharacteristic changes that occurred inmost patients.

Figure 2 shows the change in severityof myocardial perfusion abnormalitiesafter dipyridamole stress measured asthe lowest quadrant activity expressedas a percentage of maximum activity in

the heart. The quadrant with the lowestactivity contains the most severe per¬fusion abnormality of the PET imageafter dipyridamole stress. In the controlgroup, this lowest activity decreased fur¬ther, indicating a more severe or wors¬

ening abnormality on the final study com¬

pared with the baseline study. In theexperimental group, the lowest quad¬rant activity increased, ie, became lesssevere, indicating improvement in theperfusion abnormality on the final studycompared with the baseline study. Thedifferences in changes between controland treated groups were significant(P=.001 for a one-tailed t test; P=.002for a two-tailed t test; and a power of95%).

Figure 3 shows the change in size ofmyocardial perfusion abnormalities af¬ter dipyridamole stress expressed as

percentage of the left ventricle (LV) out¬side 2.5 SDs of normal individuals. Inthe control group, the size of the myo¬cardial perfusion defects increased, in¬dicating a larger perfusion abnormality.In the experimental group, the size ofthe perfusion abnormalities decreased,indicating improvement. The differencein the changes between control andtreated groups was significant (P=.02for a one-tailed t test; P=.05 for a two-tailed t test; and a power of 70%).

Figure 4 shows the change in com¬bined size and severity end point ex¬

pressed as percentage of the LV withactivity less than 60% of maximum ac¬

tivity. In control patients, the percent¬age of the LV below 60% of maximumactivity increased, indicating worseningof the perfusion abnormality. In the ex¬

perimental group, the percentage of theLV below 60% of maximum activity be¬came smaller, indicating improvementin the perfusion abnormality. The dif¬ference in the changes between the con¬trol and experimental groups was sig¬nificant (P=.002 for a one-tailed t test;

Downloaded From: http://jama.jamanetwork.com/ on 07/27/2012

) µ ,_

2d Í1ms;ß o

« Sto c o

go >oO

P=.001

Control

+8+6-+4+2

0-2-4-6

10J -8.8±2.3

+4.9±3.3

Experimental

Figure 2.—Changes in severity of myocardial per¬fusion abnormalities by positron emission tomogra¬phy after dipyridamole stress. Severity is measuredas the lowest quadrant (as shown in Figure 1) av¬erage activity expressed as percentage of maxi¬mum activity. A decrease in activity, shown by aminus sign, indicates worsening; an increase in ac¬

tivity, shown by a plus sign, indicates improvement.The values shown are mean±SEM, with n=16 forthe experimental and n=15 for the control group.

P=.004 for a two-tailed t test; and powerof 95%).

Table 2 shows the changes in threePET end points, including the four pa¬tients who, at entry, had occluded col-lateralized coronary arteries and myo¬cardial steal after dipyridamole stress.The values for the experimental group(n=20) and the values compared withthose for the controls (n=15) in Table 2are somewhat different than in Figures2 through 4. The differences in changesbetween the control group and the ex¬

perimental group (Table 2) remain sta¬tistically significant, although they werenot as large due to unimproved size andseverity of perfusion defects associatedwith myocardial steal in the four pa¬tients with occluded arteries at studyentry. For Table 2, values are forone-tailed t test, with the same resultswith two-tailed t testing.

Figure 5 shows the percentage of pa¬tients with significant changes, eitherworse or better, in size and/or severityof myocardial perfusion abnormalitiesafter dipyridamole stress. Significantchange in each of the three PET mea¬surements of size and/or severity of per¬fusion abnormalities was defined as a

change in each PET measurement onthe final PET study that was greaterthan 1 SD of each PET measurement onthe baseline studies of the control andexperimental groups combined. The per¬centage of patients showing a thresholdchange defined in this way was aver¬

aged for each of the three PET mea¬

surements, and the resulting percent¬age of patients is shown in Figure 5.This analysis of patients based on athreshold change shows that 45% of con¬trols had worsening defects, 50% showedno change, and 5% showed improvement.By comparison, most of the patients in

+10.3±5.6 P=.02QCO ,, +8

_i +4'

g 0> ^_ a

'o - « m

ExperimentalControl

-5.114.8

Figure 3.—Changes in size of myocardial perfusionabnormalities by positron emission tomography af¬ter dipyridamole stress. Size is measured as thepercentage of the left ventricle (LV) outside 2.5 SDsof the normal range. An increase in size, shown bya plus sign, indicates worsening; and a decrease insize, indicated by a minus sign, indicates improve¬ment. The values shown are meanlSEM, withn=16 for the experimental and n=15 for the controlgroup.

the experimental group showed im¬provement or no change. The differencein these changes between the controland experimental groups was significant(P=.03 by Fisher's exact test).

As a secondary end point, PET im¬ages were also visually interpreted in¬dependent of the automated quantita¬tive analysis. Among three blinded read¬ers there was agreement in 33 of 35patients on whether the baseline to finaldipyridamole images showed visualworsening, improvement, or mixedchanges. In the two patients requiringconsensus readings, all three readersagreed that these two patients showedmixed changes.

Visual interpretation of worsening or

improvement did not agree with auto¬mated measurements indicating wors¬

ening or improvement in three patientswho had mixed changes that the auto¬mated method determined as a net totalchange of improvement in one patientand worsening in two, a quantitativeresult not possible by visual interpre¬tation. In 31 of the 32 remaining pa¬tients, the automated measurementsconcurred with visual interpretation ofworsening or improvement. In the onecase of disagreement between auto¬mated and visual analysis, the visualinterpretation was borderline improve¬ment, whereas the automated measure¬

ments showed borderline worsening.Analysis of variance with the Bonfer-

roni post hoc correction algorithm didnot change these conclusions; valuesfor differences between the control andexperimental groups for changes in sizeand severity of perfusion abnormalitiesby PET after dipyridamole were as fol¬lows: for lowest quadrant averagecounts, P=.006; for percentage outside2.5 SD limits, P=.04; for percentage with

ïi

lì

+ 12

*, +8S2>m-> +4-

0

+13.513.8

to -4

P=.002

ExperimentalControl

-4.213.8

Figure 4.—Changes in combined size and severityof myocardial perfusion abnormalities by positronemission tomography after dipyridamole stress.The combined measure of size and severity is ex¬

pressed as the percentage of the left ventricle (LV)with activity less than 60% of maximum, which is 3SDs below the normal mean of 80%. An increase insize and severity is shown by a plus sign indicatingworsening. A decrease in size and severity is shownby a minus sign indicating improvement. The valuesshown are meamSEM, with n=16 for the experi¬mental and n=15 for the control group.

activity less than 60% of maximum ac¬

tivity on the dipyridamole image, P=.01.Myocardial perfusion abnormalities by

PET at resting conditions prior to di¬pyridamole stress also improvedbut notto the same degree as after dipyrida¬mole because resting myocardial perfu¬sion is less affected by coronary arterystenoses than is maximum perfusion ca¬

pacity after pharmacologie arteriolar va-sodilation.2532 The changes in size andseverity of perfusion abnormalities on

resting PET images were improved orbetter in the experimental group as com¬

pared with controls, with values asfollows: for the myocardial quadrant withthe lowest average activity, P=.02 byone-tailed t test and P=.04 by two-tailedt test; for percentage of LV activity withless than 60% of the maximum activityon the resting PET image, P=.01 byone-tailed t test and P=.02 by two-tailedt test; for percentage of LV outside 2.5SDs ofnormalized counts, P=.06 by one-tailed t test and P=.ll by two-tailed ttest.

Figure 6 compares the changes inthese three primary PET measurementswith the changes in minimum lumen di¬ameter, percent diameter stenosis, andstenosis flow reserve documented byquantitative coronary artériographieanalysis. Since the units of these sixartériographie and PET measurementsare different, the absolute difference inchanges between the control and ex¬

perimental groups was expressed as a

percentage of the mean of each mea¬surement at baseline for all patientswithout the minus or plus sign for di¬rection ofchange. For quantitative coro¬

nary arteriography, the difference inchanges between the control and ex-

Downloaded From: http://jama.jamanetwork.com/ on 07/27/2012

Table 2.—Changes in Perfusion Abnormalities by Positron Emission Tomography (PET)*PET End Point Control Group Experimental Group

Lowest quadrant averagecounts, % of maximum -8.8±2.3 +1.6±3.1 .008

% of LV outside 2.5 SDs_+10.3±5.6_-1.9±4.2_.04% of LV with activity

<60% of maximum +13.5±3.8 -1.4±3.7 .004

*Final PET-baseline PET. Includes patients with total coronary artery occlusions at entry. LV indicates leftventricle. Numbers are mean±SD unless otherwise indicated.

perimental groups ranged from 18% to31% of the mean baseline value. ForPET the difference in changes betweencontrol and experimental groups rangedfrom 22% to 96% of the mean baselinevalue.

COMMENTThis study demonstrates that the size

and severity ofmyocardial perfusion ab¬normalities documented by PET at restand after dipyridamole stress decreaseor improve in patients undergoing in¬tense risk factor modification in com¬

parison to an increase or worsening inperfusion abnormalities in patientstreated with standard therapy.

Although it is used as a measure ofchanging stenosis severity in lipid-low-ering trials, percent diameter narrow¬

ing or absolute lumen size documentedby coronary arteriography does not ac¬count for complex shape changes of ste¬noses.4 Percent stenosis is also poorlyrelated to flow capacity or coronary flowreserve2532 and fails to account for dif¬fuse disease present in most patientswith localized coronary artery narrow¬

ing.33"35 The degree of improvement inpercent stenosis in regression trials isquite modest, ranging up to 5% diam¬eter stenosis units for all stenoses andup to 10% of diameter stenosis for moresevere stenoses.1"24·60·51 These modestanatomic changes reported in recent cho¬lesterol-lowering trials raise questionsabout their functional significance. Inthis study, the substantial improvementin perfusion abnormalities measured byPET images documents the functionalsignificance of relatively small changesin the atherosclerotic coronary arterialcirculation associated with this modestextent of artériographie regression inlong-term cholesterol-lowering trials.

The improvement in size and severityof myocardial perfusion abnormalitiesat resting conditions and after dipyri¬damole stress in comparison to worsen¬

ing in controls most likely involves twomechanisms. The first is partial anatomicregression of localized coronary arterystenosis as demonstrated by quantita¬tive coronary arteriography, which im¬proves maximum flow capacity and de¬creases size and severity of perfusionabnormalities after dipyridamole stress.A second mechanism for decreased size

and severity of perfusion abnormalitiesby dipyridamole-PET may be improvedendothelial-mediated coronary arteryand arteriolar vasomotor function oc¬

curring within weeks to months aftervigorous cholesterol lowering but be¬fore anatomic regression occurs.16

Both coronary atherosclerosis andhypercholesterolemia impair endotheli¬al-mediated vasodilation of epicardialcoronary conduit arteries.52"68 Dietary fatrestriction, lipid reduction by drugs, orboth restore this endothelial-mediatedepicardial artery vasodilation in experi¬mental animals53·56·59·62 before anatomicregression of CAD is seen.69 Changes inthe coronary microcirculation may playa role in the observed improvement inperfusion defects. Atherosclerosis ofproximal, conduit, and epicardial coro¬

nary arteries impairs endothelial-medi¬ated vasodilation of the distal microcir¬culation.54·55,59,62,64 Therefore, the coronaryflow response induced by the direct ef¬fect of dipyridamole on arteriolar vaso¬dilation may be augmented by an im¬proved flow-mediated, endothelial-de-pendent, further arteriolar vasodilationin response to the initial increased flowinduced by the effect of dipyridamole.

Since coronary artery stenoses affectmaximum flow capacity much more thanresting flow, the improvement in rest¬ing perfusion abnormalities suggests dif¬fuse improvement in vasomotor func¬tion throughout the coronary arterialand arteriolar circulation in the experi¬mental group compared with controls.

These results also provide a mecha¬nism explaining the substantial decreasein angina pectoris reported to occur earlyin lipid-lowering trials.123 Because myo¬cardial perfusion reflects the integratedeffects of single or multiple stenoses,diffuse atherosclerosis, and vasomotordysfunction on coronary flow, quantita¬tive PET perfusion imaging providesinformation on severity of CAD beyonda single dimension of a single localizedcoronary artery narrowing as measuredon an arteriogram. Furthermore, be¬cause perfusion is related to lumen ra¬dius raised to the fourth power,2529 smallchanges in artériographie lumen diam¬eter that are difficult to measure on thearteriogram produce proportionatelygreater changes in perfusion that are

apparent on PET scan.

Control 55% 45%

99%

P=.03

1% Experimental

— -1-1100 80 60 40 20 20 40 60

Better or No Change, % \ Worse, %

Figure 5.—Percentage of patients with significantchanges greater than a threshold change, as de¬fined below, In size and severity of myocardial per¬fusion abnormalities by positron emission tomogra¬phy (PET) after dipyridamole stress. Significantchange in each of the three PET measurements ofsize and severity of perfusion abnormalities wasdefined as a change in each PET measurement onthe final PET study that was greater than 1 SD ofeach of the PET measurements on the baselinestudies of the control (n=15) and experimental(n=16) groups combined. Values of each of thethree PET end points for each patient were aver¬aged separately for the experimental and controlgroups.

Using end point measurements of ste¬nosis severity, artériographie regressiontrials have compared average change instenosis severity between control andexperimental groups or have comparedthe percentage of patients showing a

change in severity greater than somethreshold value, reflecting inherent vari¬ability of the measurement technique.In this study we analyzed the PET datausing both ofthese approaches and foundconcurring results.

Lipid-lowering trials have also char¬acteristically categorized patients or ste¬noses as showing regression or no changeas opposed to progression in severity.The reason is that progression is asso¬ciated with high risk of future coronaryevents, such as death, myocardial in¬farction, bypass surgery, or balloon an-

gioplasty, whereas partial regression or

stability of coronary artery stenoses isassociated with low risk of coronaryevents.2·15·50·70 Accordingly, in our analy¬sis based on threshold change in thePET images (Figure 5), we categorizedpatients as showing improvement or no

change vs progression in size and se¬

verity of perfusion abnormalities due tothese observed prognostic implications.

In view ofseveral previous trials dem¬onstrating modest artériographie re¬

gression of single, localized coronary ar¬

tery stenoses, the current study reportsunique, long-term follow-up data on thefunctional and mechanistic correlates ofthese artériographie changes seen aftera variety of lipid-lowering interventions.The greater changes in size and sever¬

ity of perfusion abnormalities obtainednoninvasively in this study may be in¬terpreted as being functional conse¬

quences of or correlates with the mod-

Downloaded From: http://jama.jamanetwork.com/ on 07/27/2012

Figure 6.—Comparison of changes by positron emission tomography (PET) and quantitative coronary ar¬teriography (QCA). The PET measures were myocardial quadrant with the lowest average activity (LowQuad), percentage of left ventricle (LV) outside 2 SDs of normals, and percentage of LV activity less than60% of maximal activity. Measurements by QCA were minimum absolute lumen diameter (Min Dia); per¬cent diameter stenosis (% DS) and stenosis flow reserve (SFR) derived from the integrated effects of ab¬solute lumen diameter, % DS, and cumulative length effects based on fluid dynamic equations, all as pre¬viously validated and reported.27 The absolute difference in changes between the control and experimentalgroups (without the minus or plus sign for direction of change) was expressed as a percentage of the meanof each measurement at baseline for all patients.

est anatomie changes in stenosis sever¬

ity measured by invasive arteriography.The principal limitation of the current

study is the relatively small number ofsubjects. However, differences in PETimages between control and experimen¬tal groups were statistically significant.During the 5-year follow-up period ofthis study, five (25%) of 20 control pa¬tients and eight (29%) of 28 patients inthe experimental group dropped out ofthe study or were unavailable for follow-up, or data were lost by computer mal¬function. The dropout rate averaged 5%per year, comparable to that in othertrials, and was comparable in controland experimental groups without iden¬tifiable bias in the results. While some

unrecognized bias from dropouts, smallsample size, or patient selection mightlimit the generalizability of the effectsof the lifestyle intervention, it wouldnot be expected to affect results of sizeand severity of perfusion abnormalitiesby PET. Noninvasive functional mea¬sures of disease severity were compa¬rable to or, in some respects, better thaninvasive coronary arteriography per¬formed in the same patients.

The total cost of diagnosis and follow-up by noninvasive PET at $2200 perstudy for all component costs is compa¬rable to typical total costs of $2500 forsingle-photon emission computed tomog¬raphy (SPECT) stress perfusion imag¬ing and is less than $8000 to $10 000 forall costs ofcoronary arteriography.71 Thisstudy focused on changes in myocardialperfusion using the most quantitative,accurate imaging currently available,which PET provides due to attenuationcorrection and uniform, depth-indepen¬dent high resolution. Pharmacologiestress is also readily standardized de¬spite changing maximal exercise work-

load due to changing anginal thresholdsthat alter the vasodilatory stimulus atfollow-up studies compared with base¬line. Whether changes in myocardial per¬fusion can be reliably followed up bySPECT, exercise stress, or both is un¬clear. Eight recent studies of a total of4064 cases suggest that attenuation ar¬tifacts limit the accuracy of SPECT72and reduce its reported sensitivity andspecificity to 86% and 54%, respectively,compared with 95% and 95% for PETbased on 855 cases in nine reports.71 Inview of technical limitations, further ran¬domized studies would be necessary todocument the validity of SPECT for thisapplication in comparison with coronaryarteriography.

In conclusion, the modest regressionof coronary artery stenosis in patientswith coronary atherosclerosis after in¬tense risk factor modification is associ¬ated with significantly decreased sizeand severity of perfusion abnormalitiesby rest-dipyridamole PET perfusion im¬aging as compared with worsening ofperfusion in control patients treated withstandard antianginal therapy at 5-yearfollow-up. Progression or regression ofCAD can be followed up noninvasivelyby objective automated quantitation or

qualitative visual interpretation of per¬fusion abnormalities by rest-dipyrida¬mole PET perfusion imaging.

This study was supported in part by grants R01HL 26882, HL 26885, HL 42554, and HL 28356 fromthe National Institutes ofHealth, Bethesda, Md; bythe Houston Endowment Foundation, the EnronCorporation, Gerald D. Hines Interests, Henry J.Kaiser Family Foundation, Fetzer Institute, Con¬tinental Airlines, Nathan Cummings Foundation,Bucksbaum Foundation, Pritzker Foundation,Gross Foundation, and Moldaw Foundation; and asa joint collaborative project with the ClaytonFoundation for Research, Houston, Tex.

The authors are indebted to Dahlia Garza, MD,

and Roberto Roberti, MD, of Beth Israel PETCenter, New York, NY, for participating in blindedreadings of PET images.

References1. Ornish DM, Brown SE, Scherwitz LW, et al. Canlifestyle changes reverse atherosclerosis? Lancet.1990;336:129-133.2. Brown G, Albers JJ, Fisher LD, et al. Regres-sion of coronary artery disease as a result of in-tensive lipid-lowering therapy in men with highlevels of apolipoprotein B. N Engl J Med. 1990;323:1289-1298.3. Kane JP, Malloy MJ, Ports TA, Phillips NR,Diehl JC, Havel RJ. Regression of coronary ath-erosclerosis during treatment of familial hypercho-lesterolemia with combined drug regimens. JAMA.1990;264:3007-3012.4. Gould KL, Ornish D, Kirkeeide R, et al. Im-proved stenosis geometry by quantitative coronaryarteriography after vigorous risk factor modifica-tion. Am J Cardiol. 1992;69:845-853.5. Watts GF, Lewis B, Brunt JN, et al. Effects on

coronary artery disease of lipid-lowering diet, ordiet plus cholestyramine, in the St Thomas' Ath-erosclerosis Regression Study (STARS). Lancet.1992;339:563-569.6. Schuler G, Hambrecht R, Schlierf G, et al. Myo-cardial perfusion and regression of coronary arterydisease in patients on a regimen of intensive physi-cal exercise and low fat diet. J Am Coll Cardiol.1992;19:34-42.7. Schuler G, Hambrecht R, Schlierf G, et al. Regu-lar physical exercise and low-fat diet: effects on

progression of coronary artery disease. Circula-tion. 1992;86:1-11.8. Buchwald H, Varco RL, Matts JP, et al. Effectof partial ileal bypass surgery on mortality andmorbidity from coronary heart disease in patientswith hypercholesterolemia. N Engl J Med. 1990;323:946-955.9. Blankenhorn DH, Nessim SA, Johnson RL, San-marco ME, Azen SP, Cashin-Hemphill L. Benefi-cial effects of combined colestipol-niacin therapy on

coronary atherosclerosis and coronary venous by-pass grafts. JAMA. 1987;257:3233-3240.10. Cashin-Hemphill L, MackWJ, Pogoda JM, San-marco ME, Azen SP, Blankenhorn DH. Beneficialeffects of colestipol-niacin on coronary atheroscle-rosis: a 4-year follow-up. JAMA. 1990;264:3013-3017.11. Haskell WL, Alderman EL, Fair JM, et al. Ef-fects of intensive multiple risk factor reduction on

coronary atherosclerosis and clinical cardiac eventsin men and women with coronary artery disease.Circulation. 1994;89:975-990.12. Expert Panel on Detection, Evaluation, andTreatment of High Blood Cholesterol in Adults(Adult Treatment Panel II). Second report of theExpert Panel on Detection, Evaluation, and Treat-ment of High Blood Cholesterol in Adults (AdultTreatment Panel II), National Cholesterol Educa-tion Program. Circulation. 1994;89:1329-1445.13. Blankenhorn DH. Angiographic trials testingthe efficacy ofcholesterol lowering in reducing pro-gression or inducing regression of coronary ath-erosclerosis. Coron Artery Dis. 1991;2:875-879.14. Singh RB, Rastogi SS, Verna R, et al. Ran-domised controlled trial of cardioprotective diet inpatients with recent acute myocardial infarction.BMJ. 1992;304:1015-1019.15. Buchwald H, Matts JP, Fitch LL, et al. Changesin sequential coronary arteriograms and subsequentcoronary events. JAMA. 1992;268:1429-1433.16. Gould KL, Martucci JP, Goldberg DI, et al.Short-term cholesterol lowering decreases size andseverity ofperfusion abnormalities by positron emis-sion tomography after dipyridamole in patients withcoronary artery disease\p=m-\apotential noninvasivemarker of healing coronary endothelium. Circula-tion. 1994;89:1530-1538.17. Scandinavian Simvastatin Survival StudyGroup. Randomised trial of cholesterol lowering in4444 patients with coronary heart disease. Lancet.1994;344:1383-1389.

Downloaded From: http://jama.jamanetwork.com/ on 07/27/2012

18. deLogeril M, Renaud S, Mamelle N, et al. Medi-terranean alpha-linolenic acid\p=m-\rich diet in second-ary prevention of coronary heart disease. Lancet.1994;343:1454-1459.19. Furberg CD, Adams HP, Applegate WB, et al.Effect of lovastatin on early carotid atherosclerosisand cardiovascular events. Circulation. 1994;90:1679-1687.20. Waters D, Higginson L, Gladstone P, et al.Effects of monotherapy with an HMG-CoA reduc-tase inhibitor on the progression of coronary ath-erosclerosis as assessed by serial quantitative ar-

teriography. Circulation. 1994;89:959-968.21. Blankenhorn DH, Azen S, Kramsch DM, et al,and the MARS Research Group. Coronary angi-ographc changes with lovastatin therapy. Ann In-tern Med. 1993;119:969-976.22. Pitt B, Mancini J, Ellis SG, Rosman HS, Mc-Govern ME. Pravastin limitation of atherosclerosisin the coronary arteries (PLAC I). J Am Coll Car-diol. 1994;23:131A. Abstract.23. MAAS Investigators. Effect of simvastatin on

coronary atheroma. Lancet. 1994;344:633-638.24. Zhao XQ, Grown BG, Hillger L, et al. Effectsof intensive lipid-lowering therapy on the coronaryarteries of asymptomatic subjects with elevatedapolipoprotein B. Circulation. 1993;88:2744-2753.25. Gould KL. Coronary Artery Stenoses: A Text\x=req-\book of Coronary Pathophysiology, QuantitativeCoronary Arteriography, PETPerfusion Imagingand Reversal of Coronary Artery Disease. NewYork, NY: Elsevier Science Publishing Co Inc; 1991.26. Gould KL, Kelley KO. Experimental validationof quantitative coronary arteriography for deter-mining pressure-flow characteristics of coronarystenoses. Circulation. 1982;66:930-937.27. Kirkeeide RL, Gould KL, Parsel L. Assess-ment of coronary stenoses by myocardial imagingduring coronary vasodilation, VII: validation of coro-

nary flow reserve as a single integrated measure ofstenosis severity accounting for all its geometricdimensions. J Am Coll Cardiol. 1986;7:103-113.28. Gould KL, Kirkeeide RL, Buchi M. Coronaryflow reserve as a physiologic measure of stenosisseverity, part I: relative and absolute coronary flowreserve during changing aortic pressure and car-

diac workload; part II: determination from arte-riographic stenosis dimensions under standardizedconditions. J Am Coll Cardiol. 1990;15:459-474.29. Gould KL. Identifying and measuring severityof coronary artery stenosis\p=m-\quantitativecoronaryarteriography and positron emission tomography.Circulation. 1988;78:237-245.30. Marcus ML, Harrison DG, White CW, McPher-son DD, Wilson RF, Kerber KE. Assessing thephysiologic significance of coronary obstructions inpatients: importance of diffuse undetected athero-sclerosis. Prog Cardiovasc Dis. 1988;31:39-56.31. White CW, Wright CB, Doty DB, et al. Doesvisual interpretation of the coronary arteriogrampredict the physiologic importance of a coronarystenosis? N Engl J Med. 1984;310:819-824.32. Marcus ML, Skorton DJ, Johnson MR, CollinsSM, Harrison DG, Kerber RE. Visual estimates ofpercent diameter coronary stenosis: 'a battered goldstandard.' J Am Coll Cardiol. 1988;11:882-885.33. Seiler C, Kirkeeide RL, Gould KL. Basic struc-ture-function relations of the epicardial coronaryvascular tree\p=m-\thebasis of quantitative coronaryarteriography for diffuse coronary artery disease.Circulation. 1992;85:1987-2003.34. Seiler C, Kirkeeide RL, Gould KL. Measure-ment from arteriograms of regional myocardial bedsize distal to any point in the coronary arterial tree

for assessing anatomic area at risk. J Am CollCardiol. 1993;21:783-797.35. Hodgson JM, Reddy KG, Suneja R, Nair RN,Lesnefsky EJ, Sheehan HW. Intracoronary ultra-sound imaging: correlation of plaque morphologywith angiography, clinical syndrome and proceduralresults in patients undergoing coronary angioplasty.J Am Coll Cardiol. 1993;21:35-44.36. Gould KL, Goldstein RA, Mullani N, et al. Non\x=req-\invasive assessment of coronary stenoses by myo-cardial imaging during pharmacologic coronary va-

sodilation, VIII: feasibility of 3D cardiac positronimaging without a cyclotron using generator pro-duced Rb-82. J Am Coll Cardiol. 1986;7:775-792.37. Demer LL, Gould KL, Goldstein RA, KirkeeideRL. Diagnosis of coronary artery disease by posi-tron emission tomography. Circulation. 1989;79:825-835.38. Gould KL. PET perfusion imaging and nuclearcardiology. J Nucl Med. 1991;32:579-606.39. Gould KL. Clinical cardiac positron emissiontomography. Circulation. 1991;84(suppl):I-22\p=m-\I-36.40. Hicks K, Ganti G, Mullani N, Gould KL. Au-tomated quantitation of 3D cardiac PET for routineclinical use. J Nucl Med. 1989;30:1787-179741. Demer LL, Gould KL, Goldstein RA, KirkeeideRL. Noninvasive assessment of coronary collater-als in man by PET perfusion imaging. J Nucl Med.1990;31:259-270.42. Mullani NA, Goldstein RA, Gould KL, FisherDJ, Marani SK, O'Brien HA. Myocardial perfusionwith rubidium-82,I: measurement ofextraction frac-tion and flow with external detectors. J Nucl Med.1983;24:898-906.43. Goldstein RA, Mullani NA, Fisher D, Marani S,Gould KL, O'Brien HA. Myocardial perfusion withrubidium-82, II: the effects of metabolic and phar-macologic interventions. J Nucl Med. 1983;24:907\x=req-\915.44. Xu EZ, Mullani NA, Gould KL, Anderson WL.A segmented attenuation correction for PET.J Nucl Med. 1991;32:161-165.45. Snedecor GW, Cochran WG. Statistical Meth-ods. 8th ed. Ames: Iowa State University Press;1989.46. Glantz SA. Primer ofBiostatistics. New York,NY: McGraw-Hill International Book Co; 1987.47. Spatz C, Johnston JD. Basic Statistics. 3rd ed.Monterey, Calif: Brooks/Cole Publishing Co; 1984.48. Stat View. Berkeley, Calif: Abacus ConceptsInc; 1992.49. Demer L, Gould KL, Kirkeeide R. Assessingstenosis severity: coronary flow reserve, collateralfunction, quantitative coronary arteriography, posi-tron imaging, and digital subtraction angiography.Prog Cardiovasc Dis. 1988;30:307-322.50. Brown BG, Zhao XQ, Sacco DE, Albers JJ.Lipid lowering and plaque regression: new insightsinto prevention of plaque disruption and clinicalevents in coronary artery disease. Circulation. 1993;87:1781-1791.51. Fuster V, Badimon L, Badimon JJ, ChesebroJH. The pathogenesis of coronary artery diseaseand the acute coronary syndromes. N Engl J Med.1992;326:242-250.52. Cohen RA, Zitnay KM, Haudenschild CC, Cun-ningham LD. Loss of selective endothelial cell va-

soactive functions caused by hypercholesterolemiain pig coronary arteries. Circ Res. 1988;63:903-910.53. Shimokawa H, Vanhoutte PM. Dietary cod liveroil improves endothelium-dependent responses inhypercholesterolemic and atherosclerotic porcinecoronary arteries. Circulation. 1988;78:1421-1430.54. Selke FW, Armstrong ML, Harrison DG. En-

dothelium-dependent vascular relaxation is abnor-mal in the coronary microcirculation of atheroscle-rotic primates. Circulation. 1990;81:1586-1593.55. Chilian WM, Dellsperger KC, Layne SM, et al.Effects of atherosclerosis on the coronary micro-circulation. Am J Physiol. 1990;258(2, pt 2):H529\x=req-\H539.56. Tomita T, Ezaki M, Miwa M, Nakamura K,Inoue Y. Rapid and reversible inhibition by low\x=req-\density lipoprotein of the endothelium-dependentrelaxation to hemostatic substances in porcine coro-

nary arteries. Circ Res. 1990;66:18-27.57. Takahashi M, Yui Y, Yasumoto H, et al. Lipo-proteins are inhibitors of endothelium-dependentrelaxation of rabbit aorta. Am J Physiol. 1990;258:H1-H8.58. Andrews HE, Bruckdorfer KR, Dunn RC, Ja-cob M. Low-density lipoproteins inhibit endothe-lium-dependent relaxation in rabbit aorta. Nature.1987;327:237-239.59. Harrison DG, Armstrong ML, Freiman PC, He-istad DD. Restoration of endothelium-dependentrelaxation by dietary treatment of atherosclerosis.J Clin Invest. 1987;80:1801-1811.60. Verbeuren TJ, Jordaens FH, Zonnekeyn LL,Van Hove CE, Coene MC, Herman AG. Effect ofhypercholesterolemia on vascular reactivity in therabbit. Circ Res. 1986;58:552-564.61. Tanner FC, Noll G, Boulanger CM, LuescherTF. Oxidized low density lipoproteins inhibit re-laxations ofporcine coronary arteries. Circulation.1991;83:2012-2020.62. Kuo L, Davis MJ, Cannon S, Chilian WM. Patho-physiological conseqences ofatherosclerosis extendinto the coronary microcirculation. Circ Res. 1992;70:465-476.63. Gordon JB, Ganz P, Nabel EG, et al. Athero-sclerosis influences the vasomotor response of epi-cardial coronary arteries to exercise. J Clin Invest.1989;83:1946-1952.64. Zeiher AM, Drexler H, Wollschlager H, Just H.Endothelial dysfunction of the coronary microvas-culature is associated with impaired coronary bloodflow regulation in patients with early atheroscle-rosis. Circulation. 1991;84:1984-1992.65. Zeiher AM, Drexler H, Wollschlager H, Just H.Modulation of coronary vasomotor tone. Circula-tion. 1991;83:391-401.66. Ludmer PL, Selwyn AP, Shook TL, et al. Para-doxical vasoconstriction induced by acetylcholinein atherosclerotic coronary arteries. N Engl J Med.1986;315:1046-1051.67. Hodgson JM, Marshall JJ. Direct vasoconstric-tion and endothelium-dependent vasodilation. Cir-culation. 1989;79:1043-1051.68. Uren NG, Marraccini P, Gistri R, deSilva R,Camici PG. Altered coronary vasodilator reserveand metabolism in myocardium subtended by nor-mal arteries in patients with coronary artery dis-ease. J Am Coll Cardiol. 1993;22:650-658.69. Benzuly KH, Padgett RC, Kaul S, Piegors DJ,Armstrong ML, Heistad DD. Functional improve-ment precedes structural regression of atheroscle-rosis. Circulation. 1994;89:1810-1818.70. Waters D, Craven TE, Lesperance J. Prognos-tic significance of progression of coronary athero-sclerosis. Circulation. 1993;87:1067-1075.71. Gould KL. Reversal of coronary atherosclero-sis: clinical promise as the basis for the non-inva-sive management of coronary artery disease. Cir-culation. 1994;90:1558-1571.72. Wackers FJ. Artifacts in planar and SPECTmyocardial perfusion imaging. Am J Card Imag-ing. 1992;6:42-58.

Downloaded From: http://jama.jamanetwork.com/ on 07/27/2012