- Email: [email protected]
Design, rationale, and baseline characteristics of the Prospective Pravastatin Pooling (PPP) project—A combined analysis of three large-scale randomized trials: Long-Term Intervention With Pravastatin in Ischemic Disease (LIPID), Cholesterol and Recurrent Events (CARE), and West of Scotland Coronary Prevention Study (WOSCOPS)
Design, Rationale, and Baseline Characteristics of the Prospective Pravastatin Pooling (PPP) Projeck A Combined Analysis of Three Large-Scale Randomized Trials: Long-Term Intervention With Pravastatin in lschemic Disease (LIPID), Cholesterol and Recurrent Events (CARE), and West of Scotland Coronary Prevention Study (WOSCOPS) The PPP Project Investigators* The Prospective Pravastatin Pooling (PPP) project is a pooled evaluation of 3 large, placebo-controlled, randomized trials of cholesterol-lowering treatment with pravastatin. It is designed to more reliably evaluate the effect of treatment on coronary and all-cause mortaliand on total coronary artery disease (CAD) events ror specific ulations of interest, including women and the el cc: e y. The trials-long-Term Intervention With Pravastatin in lschemic Disease trial, the Cholesterol and Recurrent Events trial, and the West of Scotland Coronary Prevention Study-each have common design features, including drug, dose, and duration. The project pros tively defines the objectives, end ints, and ana r ‘c plans in a protocol developed Efo re results are known of any individual trial. More than 2,000 (or 10%) of the participants in the pooled data set are women, 1,841 are aged 270 years at trial entry, and >6,000 have a total cholesterol e5.5
mmol/L (2 13 mg/dl). The mean low-densi lipoprotein cholesterol level is 4.2 mmol/L (162 mg/ a, ). The mean blood pressure level is 134/81 mm Hg and 20% are current smokers. Half of the PPP participants have had a prior myocardial infarction. More than 7% have a history of diabetes and 26% have a histo of hypertension. PPP is projected to have data on aTo ut 1,100 CAD deaths, 500 non-CAD deaths, and >l,OOO cancers by study completion. Based on a pooled population of 19,768 patients, with approximately 1,600 deaths and 100,000 patient-years of follow-up, PPP should have oad power to examine the effects of treatment on tota Bmortality, CAD mortality, and cancer incidence, and to determine effects on total CAD events in important subgroups including the elderly, women, diabetics, and those with lower serum cholesterol levels (~5.5 mmol/L (2 13 mg/dll). (Am J Cardiol 1995;76:899-905)
he ProspectivePravastatinPooling (PPP)project was T initiated in 1992to addressthe effectsof pravastatin on all-cause and coronary mortality, on total coronary
vention With Pravastatin in Ischemic Disease (LIPID) study,17the Cholesterol and Recurrent Events (CARE) study,l8 and the West of Scotland Coronary Prevention Study (WOSCOPS).19Collectively, these studies will involve 19,768patients, and by the time of their completion will provide outcome data from nearly 100,000person-years of follow-up. The purpose of this study is to describe the objectives of the project, the trials involved, the baseline characteristicsof the study populations, the definitions of outcome events, and the power to test the study hypotheses.
events in special populations, and to confirm long-term safety. At that time, trials of cholesterol-lowering therapy, and meta-analysesof thesetrials, had clearly demonstrated reductions in coronary events,1-9but the question of the effects of cholesterol lowering on overall mortality remained unresolved for all patient groups.c9 With the advent of the hydroxymethylglutaryl coenzyme A reductaseinhibitors, it has becomepossible to test the effects of larger cholesterol reductions on coronary events and overall mortality.‘0-19 PPP is a prospective METHODS Objectives: The primary goal of the PPPproject is to collaboration involving 3 ongoing, long-term, large-scale, monotherapy trials of pravastatin: the Long-Term Inter- pool the data from the 3 large randomized studies of pravastatin to addressscientific questionsfor which there may be inadequatepower within the individual studies: Manuscript received May 19, 1995; revised manuscript received in particular, the effect of pravastatin on total mortality and accepted July 24, 1995. Address for reprints: Curt Furberg, MD, Department of Public and on specific non-CAD events as well as the effect on Health Sciences, Bowman Gray School of Medicine, Winston-Salem, CAD morbidity within important subgroups. The proNorth Carolina 27157-l 063 ject will also be able to explore the consistencyof effects across the constituent trials and help establish how to *A complete listing of PPP investigators and their affiliations IS given rn the Appendix. generalize the findings from the trials, as well as provide PREVENTIVE
CARDIOLOGY/PROSPECTIVE
PRAVASTATIN
POOLING
PROJECT
899
Three pravastatin trials: The prevention trials that form the basisfor the Pravastatin Pooling PPP-LIPID, CARE, and WOSCOPS Acronym -are described in Table I. These 3 studies have been selected for incluLIPID CARE woscoPs sion in the PPPproject becauseof their Cholesterol and Long-Term West of Scotland Name of trial common design features, including Intervention with Coronary Prevention Recurrent Events drug, dose, and duration, and the closePravastatin in itudy nessof their completion dates.Imporlschemic Disease Australia and USA & Canada Scotland location tantly, they have each been selected New Zealand without prior knowledge of the out9,000 6,000 4,000 Target sample size come and they should therefore pro9,014 6,595 Observed sample size 4,159 vide an unbiasedestimateof the effects 5+ 5 5 Mean follow-up [yr) May 31, 1995 of pravastatin. Furthermdre, each of February 1997 Target termination February 12, 1996 46,200 20,975 32,140 Patient-years the researchquestions to be addressed Fatal 81 nonfatal Fatal CAD Fatal CAD & nonfatal Primary end point by PPP and a detailed analysis plan myacardial infarction myocardial infarction have been carefully defined in adv87 80 1 centralized Sites ance, without any knowledge of trial + 40 satellite + 60 trial centers Major eligibility outcome by treatment. In this way, recriteria sultsof PPPwill provide much stronger Myocardial infarction Myocardial infarction No evidence of Clinical evidence than simply undertaking a 3-20 months before or unstable anaina myocardial infarction retrospective meta-analysisat the conpectoris 3-36 months ’ randomization before screening clusion of the 3 trials. Men and women Men and women Men only Gender Each of these studies is a random31-75 21-75 45-64 Age at entry (yr) ized, double-blind, placebo-controlled, 4.0-7.0 mmol/L e6.2 mmol/L Not applicable Total cholesterol clinical trial of 25 years’ duration using [155-270 mg/dl) (~240 mg/dl) a dose of 40 mg of pravastatin once Not applicable >4.0 mmol/L on 3.0-4.5 mmol/L LDL cholesterol (115-l 74 mg/dl) 2 occasions; daily, and each has been described in b4.5 mmol/L detail previously.*‘-19LIPID is a secat least once ondary prevention trial in 9,014 men but e6.0 mmol/L and women with a history of myocarat least once (155-l 74 mg/dl) dial infarction or hospitalization for unstable angina. CARE is a secondary CAD = coronary artery disease; LDL= lowdensiv lipoprotein cholesterol. prevention trial in 4,159 men and women with a history of myocardial reliable information on various outcomes attributed to infarction and normal or only mildly elevatedcholesterol levels. The WOSCOPSis a primary prevention trial in cholesterol reduction. Specific primary obiectives of PPP:Specific objectives 6,595 men who have had no documented myocardial are: (1) to determine the effect of pravastatin therapy on infarction. PPPoutcomes: ALL-CAUSE MORTALITY: Approximateall-causemortality both for the secondaryprevention trials (LIPID and CARE) and for the 3 trials combined; ly half of the deaths in WOSCOPS are expected to be and (2) to determine the effect of pravastatin therapy on due to CAD versus X to %of the deaths in CARE and cause-specificmortality: CAD mortality, and other car- LIPID. Becausethe proportion of all deathsdue to CAD diovascular and noncardiovascular mortalities. The spe- will differ between the secondary and primary prevencific secondary objectives of PPP are: (1) to determine tion trials, the percent reduction in all-cause mortality the effect of pravastatin therapy on total coronary inci- for a given effect on CAD will also differ. Consequentdence (fatal CAD and nonfatal myocardial infarction), ly, the primary PPP analysis of all-cause mortality will and particularly in prespecified subgroups of special comprise both the secondarypopulations in CARE and interest (primary vs secondary prevention, the elderly, LIPID and, separately,an analysis of all 3 trials includmen vs women, diabetics, current smokers,those taking ing WOSCOPS. CARDIOVASCULAR DISEASE MORTALITY: Three C&eantihypertensive treatment, and specific categories of baseline lipid levels including those with total choles- gories of fatal secondary cardiovascular outcomes will terol ~5.5 rnmol/L); and (2) to determine the effect of be studied: (1) fatal CAD (i.e., fatal myocardial infarcpravastatin therapy on total (fatal plus nonfatal) non- tion, sudden death or fatal CAD, separatelyor in comcoronary events(specifically cancer,trauma, and stroke). bination); (2) other fatal cardiovascular disease (i.e., Additional outcomes of interest will be explored in stroke, aortic aneurysm, and iliofemoral disease, sepadescriptive analysesof the effects of treatment on other rately and in combination); and (3) any fatal cardiovasischemic events, the need for invasive coronary proce- cular disease(categories 1 + 2). This will be examined dures,andhospitalization for prcdefinedcategoriesasout- for all patients, those with established CAD, and those lined later. Combinedeffectson cause-specificmortalities with prior myocardial infarction. NONCARDIOVASCULAR DISEASE MORTALITY: Mortali(including cancer and trauma) will also be used to estity will be described separatelyfor total and cause-spemate effectson total mortality for other populations. TABLE I Characteristics
and Eligibility (PPP) Project
900
THE AMERICAN
JOURNAL
Criteria
of the Trials in the Prospective
OF CARDIOLOGY@
VOL. 76
NOVEMBER
1, 1995
79 * 11
1.8 * 0.7 (156 * 61) 128 * 18
1 .o * 0.2 (38 zt 8)
3.6 zt 0.4 (139 * 15)
5.4 * 0.4 (208 zt 17)
Men (n = 3,583)
70 * 11
78 * 11
1.7 f 0.7 (154 zt 61) 134 f 20
1.2 f 0.3 (45 * 1 1)
3.6 * 0.4 (140 * 14)
5.6 zt 0.4 (215 * 15)
Women (n = 576)
67 i 11
78 zt 11
1.8 * 0.7 (156 * 61) 129 * 18
1 .o * 0.2 (40 f 7)
3.6 * 0.4 (139 f 15)
5.4 f 0.4 (209 * 17)
Total (n = 4,159)
10 67 23 27 f 4
69 zt 11
81 i 11
1.8 f 0.8 (158 zt 74) 133 f 19
0.9 f 0.2 (36 i: 8)
3.8 * 0.7 (149 * 26)
5.6 zt 0.7 (216 i 28)
Men (n = 7,503)
10 46 44 27 * 5
71 * 11
80 * 11
1.7 * 0.8 (155 * 70) 138 * 20
1.1 * 0.3 (43 * 1 1)
4.0 f 0.7 (156 zt 26)
5.9 * 0.7 (229 zt 27)
Women (n = I,51 1)
10 63 17 27 * 4
89 zt 11
81 zt 11
1.8 f 0.8 (158 * 73) 134 f 19
1 .o * 0.2 (37 * 9)
3.9 * 0.7 (150 ziz27)
5.6 ZJZ 0.7 (218 f 28)
Total (n = 9,014)
35 39 26 26 zt 3
73 * 9
84i
10
1.8 +z0.8 (163 * 69) 136 * 17
1.2 * 0.3 (47 f 10)
5.0 f 0.5 (192 * 17)
7.0 st 0.6 (272 i 23)
Men Only (n = 6,595)
20 56 24 27 * 4
70 f 11
81 * 11
1.8 zt 0.8 (160 * 70) 133 5 18
1 .o f 0.3 (40 f 10)
4.2 * 0.8 (163 * 31)
6.1 * 1.0 (235 * 37)
Men (n = 17,681)
13 46 41 27 iz 5
70 f I 1
79 f 11
1.7 * 0.8 (155 i 67) 137 f 20
1.1 f 0.3 (43 * 11)
3.9 e 0.6 (151 * 24)
5.8 * 0.6 (225 i 25)
Women (n = 2,087)
20 55 25 27 * 4
70 f 11
81 * 11
1.8 * 0.8 (159 f 70) 134 * 19
1.1 f 0.3 (41 * 10)
4.2 i 0.8 (162 +z31)
6.1 i 0.9 (234 i 38)
Total (n = 19,768)
PPP
zt 11
16 61 23 28 * 4
woscoPs
67
20 47 33 28 i 6
LIPID
16 63 21 28 iz 4
50 7 26 of population).
70 13 47 as percent
48 7 24 are given
0 1 11 which
64 9 39 disease,
59 10 51 of other
65 8 34 history
100 14 29 and
100 20 37 status
100 13 27
CARE
TABLE II Baseline Characteristics and Risk Factor Levels of Study Population
Risk Factor Total cholesterol mmol/L (w/4 LDL cholesterol mmol/L h/4 HDL cholesterol mmol/L (w/4 Triglyceride mmol/L h/4 Systolic blood pressure (mm Hd Diastolic blood pressure (mm Hgl Heart rate (beats/min) Cigarette smoker Current Former Never Body mass index (kg/mz) History Prior myocardial infarction Diabetes mellitus Hypertension All values ore expressed OS mean * SD (except for cigarette smoker HDL = highdensity lipoprotein; other abbreviations ~1s in Table I.
TABLE III
Estimates of Statistical
Power Among
Outcome
Subgroup
Total mortality
All LIPID + CARE All Prior myocordial infarction A)) All All All Men Women
CAD
mortality
Non-CAD mortality Cancer (fatal + nonfatal) Trauma (fatal + nonfatal) Totol CAD events (Fatal CAD + nonfatal myocardial infarction)
Age iv1 <55 55-64 265 270 Prior myocardial infarction Current smoker Hypertension Diabetes mellitus Cholesterol ~5.5 LDL-C ~3.5 HDLC cl .O Triglyceride >2.5
PPP Populations Number of Patients
Projected Events*
19,768 13,173 19,768 9,91 1
1,600 1,400 1,100 650
0.78 0.77 0.92 0.73
(1213 (13)Z
19,768 19,768 19,768 19,768 17,681 2,087
500 1,150 600 2,900 2,600 300
0.52 0.88 0.60 >0.99 >0.99 0.82
(-2018 (-20)s (-20)s (27)tl (27) (27)
6,635 8,286 4,843 1,841 9,911
600 1,100 1,200 600 2,200
0.98 aO.99 >0.99 0.98 >0.99
(27) (27) (27) (27) (27)
850
>0.99 >0.99 0.91 BO.99 0.99 >0.99 >0.99
(27) (27) (27) (27) (27) (27) (27)
3,871 5,209 1,427 6,198 4,53 1 7,254 3,275
1,100
350 1,100 700 1,100
600
Power to Detect % Reduction (%)t
(18)
(18)
‘Projected events for each category are bared on (1) the original assumptions of each protocol, that each trial contin(2) estimated event rates for the different categories, and (3) th e assumption ues until its planned completion. tPower to detect o difference using 2-sided 5% level test. tDifference in total mortality (12% to 13%) bused on on assumption of an 18% difference in coronary artery disease (CAD) mortality and no effect on non-CAD mortality. 5Possible increases in non-CAD deaths of ~20% have been suggested in published studies and ore shown for illustrative purpaes only. IiDifference in total CAD events of 27% is based on on assumption of on 18% difference in CAD mortality and 50% greater effect for fatal + nonfatal CAD events combined. C = cholesterol; other abbreviations os in Tables I and II. Lipid values ore meowed in mmol/L.
angioplasty or other nonsurgical procedures) or surgical revascularization (e.g., coronary artery bypassgraft surgery) procedures, total episodes of congestiveheart failure or unstable angina (i.e., fatal or hospitalizations for nonfatal), major hepatic disease,hospitalization for psychiatric illness, and occurrenceof myopathy.PPPwill also be able to examine for other potential adverseeffectsthat are suspectedto be drug-related. Effects of treatment on population subgroups (effect modification): The
effect of pravastatin on total coronary incidence will be investigated in specified subgroupsof the population. The following subgroups will be included in an analysisof a modification of treatment effect by risk factors: Age (~55, 55 to 64,265, and 270 years); gender; baseline lipid levels (total cholesterol ~5.5, 5.5 to 6.4, and 26.5 mrnol/L [<213, 213 to 249, and 2250 mg/dl]; low-density lipoprotein cholesterol <3.5, 3.5 to 4.4, 24.5 mmol/L
[<135,
135to 174,and 2175 mg/dl]; high-density lipoprotein cholesterol c 1.0, 2 1.0 mmol/L [c38.7, 238.7 mg/dl]; and triglycerides ~1.5, 1.5 to 2.4, and 22.5 mmol/L [<133, 133 to 219, and 2220 mg/dl]); smoking (classifiedby current smokers, former [at time of randomization], and never); hypertension (patients _._.^ receiving antihypertensive drug _ cific noncardiovasculardiseases.In particular, categories therapy); prior myocardial infarction; and prior estabof noncardiovascular fatal outcomesto be described are lished CAD and diabetesmellitus. The relative effectsof (1) cancer mortality: deaths resulting from cancer over- pravastatinaccording to age,lipid levels, and blood presall and listed by body system and tumor type; (2) trau- surewill also be explored using theseas continuous varimatic death (accidental death and suicides); and (3) oth- ables, and the constancy of the relative risk reduction er noncardiovascular deaths, including cirrhosis and acrossa compositerisk scorederived from thesevariables non-cancer pulmonary disease. Because of the small will be examined.The effectswithin these subgroupson number in eachcategory,a simple tabulation will be pro- cause-specificand total mortality will also be reported. Evaluation and adjudication of the PPP outcomes: The vided for each study and for the combined population. detailed definitions and methods for classification of TOTAL CORONARY INCIDENCE: The effect Of praVaSti+ tin on the total incidence of any fatal CAD plus nonfatal major events of interest and outcomesin the prevention myocardial infarction will be investigated.Each of the 3 studies have been agreedupon by the PPPsteering comtrials has good power to detect a modesttreatmenteffect mittee and have been incorporated in the full PPP proon the combined incidence of fatal and nonfatal CAD for tocol (copy available upon request). These definitions its own study population. The combinedPPPanalysiswill correspondwith those usedin the 3 trials, facilitating the addressthe questionof whether the relative risk reduction direct use of individual trial data. PPP analysis plan: All primary researchquestions for in fatal CAD is different from that in nonfatal CAD, as suggestedin previous meta-analyses,1~2~20 and will exam- PPPand the relevant trials to be included for each quesine for possible effect modification by other risk factors tion are specified in advance in the protocol. The intention-to-treat model of analysis will be used. An ancil(see subgroupsmentioned later). OTHER OUTCOMES OF INTEREST: PPP will also inves- lary analysis estimating treatment effect for compliers tigate the effect of pravastatin on the following out- will be consideredby making appropriate adjustmentof comes: total (i.e., fatal plus nonfatal) cancer,total trau- the intent-to-treat analyses, rather than performing an matic or violent events, and total stroke, each of which analysis by treatment received.21 Primary analyses will be presented unadjusted for are specific secondary objectives. In addition, descriptions will be provided on the outcomesof cardiac cathe- other covariates. Analyses adjusted for other covariates terization, nonsurgical revascularization (e.g., coronary may be undertaken as ancillary analyses. All analyses 902
THE AMERICAN JOURNAL OF CARDIOLOGY”
VOL. 76
NOVEMBER 1, 1995
will be stratified by trial. Primary analyseswill be based on time-to-event data using the stratified log rank test. Analyses adjusting for patient covariates will use a stratified Cox regression approach,provided the proportional hazardsassumption is not violated. If this assumption appearsinvalid, an analysis allowing for a delayed treatment effect will be performed using a time-partitioned Cox regression analysis.22Time-to-event mortality plots will be used for the main analyses.In addition, the total number of eventsin different categorieswill be presented (e.g., the number of cancers and trauma events). When appropriate, analysis of 2 X 2 tables will be based on the stratified Mantel-Haenszel test.23Relative risks will be estimated to determine the percent risk reduction associatedwith pravastatin treatment.Absolute risks will be estimated to determine the number of events averted per 1,000 treated patients. Since detailed individual data will be available from eachtrial, the analysis is a stratified pooling of the 3 studies. Heterogeneity of cause-specific treatment effects acrossthe study populations will be examined, although it is recognized that the power to detect such interactions is low. A model basedon a weighted combination of the effectsseenon each specific causeof death will be developed, enabling estimation of an “expected” net mortality reduction for any given risk population. Basedon the proportion of different types of deaths in each trial, this will enable more robust study-specific reductions in total mortality to be estimated.24 Subgroups have been prespecified based on either biologic rationale or scientific importance. The characteristics previously noted (under effect modification) may be expandedbefore any of the individual trial results are known, in the light of new data emerging from other studies. Tests for “treatment by subgroup” interaction will be undertakento help in the interpretation of results. Examination of effect modification will include a multivariate model incorporating all variables (as previously listed under effect modification) and their treatmentinteraction terms. Results for each subgroup will be presentedas p values and confidence intervals, adjustedfor covariates if necessary,with an appropriate discussion indicating that the results do not take into account multiple comparisons and hence the need for some adjustment or caution in interpreting the results. A simple data set comprising all baseline characteristics and major outcome variables will be created from each trial. These data will enable the time-to-event outcomes as well as individual patient covariates to be examined. The stratified pooled analysesfor PPPwill be undertaken independently of the sponsor by the members of the PPP steering committee. It is proposed that this analysis occur as soon as the last of the 3 trials, presumably LIPID, has been completed. Project organization and publication policy: A governing body, the steering committee, was established in October 1992. It contains representativesfrom all 3 trials, together with an epidemiologist responsible for assisting the investigators with the data analysis and a PPP chairman, both independent of the 3 trials. Three working groups, with representation from each constituent trial, were created to organize and guide the
design and analysis of the project, objectives, and publications, and outcomes and end point definitions. All analysesand assessmentof the results will be conducted independently of the sponsor and its representatives. Publication of theseresults will be made in the name of the PPPproject group and only after the principal results of the 3 individual trials are published.
RESULTS The protocols and case report forms were reviewed and comparedfor the 3 studies participating in the PPP project. Table I describes the participating studies and key entry and enrollment information. Table II presents the baseline characteristicsfor the 3 individual trials and the combined PPP population. The mean age of the 19,768participants at baseline is 58 years, with women being, on average,4.5 years older than men. This overall gender difference in age is primarily a function of the trials contributing the data. The primary prevention trial, WOSCOPS,is composedentirely of men, and it has the youngest study population. Risk factor distributions are presented in Table II. More than 2,000 (or 10%) of the participants in the pooled data set are women; half of the PPP participants have had a prior myocardial infarction, >7% have a history of diabetes,and 26% have a history of hypertension and antihypertensive medication. It is noted that while the WOSCOPSpopulation (free of evidence of rnyocardial infarction by definition) had significantly higher lipid levels and a significantly greater proportion of current smokers, it had the lowest prevalences of diabetes and hypertension. Power calculations: One criterion in the selection of researchquestionsto be examined with PPP is the need for adequatepower to detectplausible and important differences in outcomes. Table III illustrates the approximate power PPPhas to detect such differencesbasedon a projected number of events within each category.This projectednumber provides only a rough guide to the final number of eventsthat will occur, and is basedon the original assumptionsof each protocol, the estimated event rates for the different categories,and the assumptionthat each trial continues until its planned completion. Based on original protocol assumptions,a reduction in coronary mortality that might be expected is 18%, if an average reduction in total cholesterol of 18% was obtained and there was 1% reduction in coronary mortality for each 1% reduction in cholesterol achieved.2*6 The power of the combined analysis to detect such an effect is high (92%) and fair (73%) for the subgroup of patients with myocardial infarction. If one assumesno beneficial or adverseeffect on non-CAD mortality, this would translate to a reduction in total mortality of about 12% for the 3 trials, with reasonable power (78%) to detect such an effect. However, these assumptions are conservativeand a considerably larger reduction in CAD mortality, such as 1.6% for every 1% reduction in cholesterol, as observed in the Scandinavian Simvastatin Survival Study (4s) trial group12is also possible. This would correspond to a reduction in total mortality of =20%, and the power of PPP to detect such an effect would be very high (>99%).
PREVENTIVE CARDIOLOGY/PROSPECTIVE
PRAVASTATIN POOLING PROJECT
903
It will also be important that PPP exclude possible adverseeffects on non-CAD deaths or serious nonfatal outcomes,such as a 20% increasein theseevents.However, with approximately 500 non-CAD deaths projected, PPP will only have about a 1 in 2 chance (52%) of detecting a 20% increase in non-CAD deaths. For this reason, fatal plus nonfatal potential adverseeffects will also be examined: for cancers with >l,lOO projected events,PPPwould have adequatepower to detect a 20% increase, although it would only have a 60% power to detect a 20% increase in trauma. For subgroup analyses,PPP will examine the effects of pravastatin on the combined outcome of CAD events (fatal CAD + nonfatal myocardial infarction). For this outcome,a larger effect of 27% reduction in events(50% more than for fatal CAD) has been assumed.PPP will have very good power to detect such an effect for almost all subgroupslisted; even among women (with the least projected number of events), PPP will still have >80% power.
increased power to examine the effects of treatment on total mortality and on total CAD eventsfor many of the clinically important subgroupsconsideredfor treatment. The size of the PPP population of 19,768followed for >5 years, with nearly 100,000patient-years, 1,600 fatal events, and twice as many nonfatal events expected, offers a unique opportunity to explore these issues. In contrast, for example, 4s yielded =23,000 patient-years of experience and 438 fatal events. The similarity of design of the individual studies,the use of a single therapeutic agent used for 5 to 6 years, and the prospectively defined objectives of the project, along with predefinedend point definitions, are strengths of the project, allowing pooling of individual data and interpretation of the results with little likelihood of bias. Power estimatesindicate a high likelihood that the PPP project will detect beneficial effectsof pravastatin on the sizesanticipated from all available evidenceon all-cause and on coronary mortality in the total population, and generally on total CAD in the subgroupsof special interest.
DISCUSSION Epidemiologic and clinical trials researchhave established the causalrelation of high serum cholesterol, particularly low-density lipoprotein cholesterol, to the risk of CAD and its reversibility.1-9,20,25-27 However, there has been some concern that reducing cholesterol may also carry some risk,~8,2~28and until recently, trials of cholesterol lowering, by diet or drug therapy,have failed to provide unequivocal evidence that such strategiesprolong life. The recently completed 4S12involving 4,444 patients has now produced striking evidence of such a survival benefit over 5 to 6 years among hypercholesterolemic patients (aged 35 to 70 years) (total cholesterol 5.5 to 8.0 mmolL [212 to 310 mg/dl]) with preexisting CAD. These results are likely to lead to broad changes in physician practice, particularly for patients similar to those studied in 4s. DesDite this imoortant result, 4s alone still did not have sufficient power to addressthe question of the effects on overall mortality in women, or on major non-CAD events, and the issue of how widely theseresults should be generalizedacross age,gender,lipid levels, absolutelevels of risk for CAD, choice of lipid-lowering drug, and so on, remains to be determined. Four placebo-controlled trials of pravastatin, designed principally to assessthe effectsof pravastatin on changes in atherosclerosis, have demonstrated modest effects on atherosclerosis,but also collectively demonstrated significant reductions in cardiovascular clinical events.13-15,29,30 A planned meta-analysisof these studies demonstrateda 62% (95% confidence interval 38 to 80) reduction in risk of total myocardial infarction and a 53% (95% confidence interval 29 to 71) reduction in the combined outcomes of nonfatal myocardial infarction or death from any cause.29+30 However, even in combination, thesestudieswere not of sufficient size to detect statistically significant effectson CAD mortality or total mortality. The PPP project will be able to address important questions that are beyond the scope of the 3 individual contributing studies. Combining the data sets provides I
904
THE AMERICAN JOURNAL OF CARDIOLOGY@
I
VOL. 76
Acknowledgment: Bristol-Myers Squibb Pharmaceutical ResearchInstitute, Princeton, New Jersey,the sponsor of all 3 trials, is also supporting and helping with the coordination of the ProspectivePravastatinPooling project. We gratefully acknowledge the secretarial assistance of Edel Williams and Dianne Black in the preparation of the manuscript.
I. Holme I. An analysis of randomized trials evaluating the effect of cholesterol reduction on total mortality and coronary heart disease incidence. Circularion 1990;82:1916-1924. 2. Pete R, Yusuf S, Collins R. Cholesterol-lowering trial results in their epidemological context. Circularion 1985;72(suppl III):III-451. 3. Rossouw JE, Lewis B, Riikind BM. The value of lowering cholesterol after my@ cardial infarction. N Engl J Med 1990;323: 1112-l 119. 4. Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Pmvention Trial I. Reduction in incidence of coronary heart disease. JAMA 1984: 251:351-364. 5. Frick MH, Elo 0, Haapa K, Heinonen OP, Heinsalmi P, He10 P, Huthmen JK, Kaitaniemi P, Koskinen P, Manninen V, MBenpti H, Miilkiinen M, Mitt$ri M, Norola S, Pastemack A, Pikkarainen I. Rome M, Sjiiblom T, NikkilP EA. Helsinki Heart Study: primary prevention trial with gemfibrozil in middle-aged men with dyslipidemia. Safety of treatment, changes in risk factors, and incidence of corenary heart disease. N Engl JMed 1987;317:1237-1245. 6. Keech AC. Does cholesterol lowering reduce total mortality? Postgrad Med J 1992;68:870-871. 7. Davey-Smith G, Pekkanen J. Should there be a moratorium on the use of cholesterol lowering drugs? Br Med J 1992;304:431&434. 8. Muldoan MF, Manuck SB, Matthews KA. Lowering cholesterol concentrations and mortality: a quantitative review of primary prevention trials. Br Med J 1990; 301:309-314. 9. Collins R, Keech A, Pete R, Sleight P (ISIS), Kjekshus J, Wilhelmsen L (4s). MacMahon S, Shaw J, Simes J (LIPID), Braunwald E, Boring J, Hennekens C, Pfeffer M, Sacks F (CARE, Women’s Health Study, Physicians’ Health Study, SAVE), Probstfield J, Yusuf S (Post-CABG Study), Downs JR, Gotto A (AFCAPSI TEXCAPS), Cobbe S, Ford I, Shepherd J (WOSCOPS). Cholesterol and total mcrtality: need for larger trials [letter]. Br Med J 1992;304:1689. 10. Illingsworth DR. Clinical implications of new drugs for lowering plasma cho‘ester01 concentrations. Drugs 1991;41:151-160. I 1. McGovern ME, Mellies MJ. Long-term experience with pravastatin in clinical research trials. Clin Ther 1993;15:57+54. 12. Scandinavian Simvastatin Survival Study Group. Raodomised trial of cbolestern1 lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4s). Lancef 1994;344:1383-1389. 13. Pitt B, Man&i JGB, Ellis SG, Rosman HS, McGovern ME, for the PLAC I Investigators. Pravastatin Limitation of Atherosclerosis in the Coronary Arteries (PLAC I): reduction in atherosclerosis progression and clinical events. J Am Coil
cardi0’ prrss). 14. Crow(inJR, Byington RP, Bond MG, Espeland MA, Craven TE, Sprinkle JW, McGovern ME, Furberg CD. Pravastatin, Lipids and Atherosclerosis in the Carotid NOVEMBER 1, 1995
Arteries (PLAC 11).Am J Cardiol 1995;75:455459. 15. Jukema IW, Btuscbke AVG, van Boven Ad J, Reiber JHC, Bal ET, Zwinderman AH, Jansen H, Boerma GJM, van Rappard FM, Lie KI, on behalf of the REGRESS study group. Effects of lipid lowering by pravastatin on progression and regression of coronary artery disease in symptomatic men with normal to mcderately elevated serum cholesterol levels. Tbe Regression Growth Evaluation Statin Study (REGRESS). Circulation 1995;91:‘2528-2540. 16. Cholesterol Treatment Trialists’ (CTT) Collaboration. Protocol for a prospec tive collaborative overview of all current and planned randomized trials of cholesterol treatment regimens. Am J Cardiol 1995;75:113&1134. 17. The Lipid Study Group. Design features and baseline characteristics of the LIPID (Long-Term Intervention with Pravastatin in Ischemic Disease) study: a randomized trial in patients with previous myocardial infarction and/or unstable angina pectoris. Am J Cardiol 1995;76:474-479. 18. Sacks FM, Rouleau JL, Moye LA, Pfeffer MA, Wamica JW, Arnold JM, Nash DT, Brown LE. Sestier R, Rutherford J, David BR, Hawkins M, Braunwald E for the CARE Investigators. Baseline Characteristics in the Cholesterol and Recurrent Events (CARE) trial of secondary prevention in patients with average semm cholesterol. Am J Cardiol 1995;75:621423. 19. The WOSCOPS Study Group. Baseline characteristics and screening expcrience in the West of Scotland Coronary Prevention Study. Am J Cardiol 1995; 76~485-491. 20. Law MR. Wald MI. By how much and how quickly does reduction in semm cholesterol concentration lower risk of ischaemic heart disease? Br Med J 1994: 308~367-372. 2 1. Newcombe RC. Explanatory and pragmatic estimates of the treatment effect when deviations from allocated treatment occur. Stat Med 1988;8:1179-I 187. 22. Grambsch PM, Themeau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994;s 1:5 15-526. 23. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;22:71%748. 24. Pocock S, Geller N, Tsiatis A. The analysis of multiple end points in clinical trials. Biometrics 1987;43:487-498, 25. Anderson KM, Cast& WP, Levy D. Cholesterol and mortality: 30 years of follow-up from the Framingham study. JAMA 1987;257:2174-2180. 26. Martin HJ. Hulley SB, Browner WS, Kuller LH, Wentwortb D. Serum cholesterol, blood pressure and mortality; implications from a cohort of 361,662 men. Lancer 1986;2:933-936. 27. Davey Smith G, Song F, Sheldon TA. Cholesterol lowering and mortality: the imponance of considering initial level of risk. Er Med J 1993;306:1367-1373. 28. Law MR, Thompson SG, Wald NJ. Assessing possible hazards of reducing serum cholesterol. Br Med J 1994;308:373-379. 29. Furberg CD, Byington RP, Grouse JR, Espland MA. Pravastatin. lipids, and major coronary events. Am J Cardiol 1994:73:1133-l 154.
PREVENTIVE
30. Byington RP, Jukema JW, Salonen IT, Pitt B, Bruschke AV, Hoen H, Furberg CD, Man&i GBJ. Reduction in cardiovascular events during pravastatin therapy: pooled analysis of clinical events of the pravastatin atherosclerosis intervention program. Circulation (in press).
APPENDIX Prospective Pravastatin Pooling (PPP) Project Investigators: (I ) = Subcommittee Chairperson; (2) = End Points Definitions & Outcomes; (3) = Design and Analysis; (4) = Objectives & Publications. Curt Furberg(4) PPP Project Chairman: Department of Public Health Science, Bowman Gray School of Medicine, Winston-Salem, North Carolina; Robert P. Byington(4) PPP Project Coordinator: Depanment of Public Health Science, Bowman Gray School of Medicine, Winston-Salem, North Carolina: Jennifer L. Baker(4): NHMRC Clinical Trials Ctr., University of Sydney, Sydney, Australia; Stuart M. Cobbe(4,l) Department of Cardiology, University of Glasgow, Glasgow, Scotland; Barry Davis(4): School of Public Health, University of Texas, Houston, Texas; Ian Ford(4): Director, Databases Unit, University of Glasgow, Glasgow, Scotland; Paul Glasziou(4): Department of Social & Preventive Medicine, University of Queensland, Herston Qld, Australia; C. Morton Hawkins(4): Director, Coordinating Center for Clinical Trials, University of Texas, Houston, Texas: Anthony C. Keech(4): NHMRC Clinical Trials Ctr., University of Sydney, Sydney, Australia; Stephen MacMahon(4): Clinical Trial Research Unit, University of Auckland, Auckland, New Zealand: Lemuel Moy6(4): School of Public Health, University of Texas, Houston, Texas; Chris Packard(4): Department of Pathologic Biochemistry, University of Glasgow, Glasgow, Scotland: Marc A. Pfeffer(4): Brigham & Women’s Hospital, Harvard Medical School, Boston, Massachusetts; Frank M. Sacks(4,l): Brigham & Women’s Hospital, Hzvard Medical School, Boston, Massachusetts; James Shepherd(4): Professor of Pathological Bicchemisuy, University of Glasgow, Glasgow, Scotland; John Simes(4.1): Director, NHMRC Clinical Trial Cu., University of Sydney, Sydney, Australia; Andrew Tonkin(4): Department of Cardiology, Austin Hospital Heidelburg, Victoria, Australia; Bristol-Myers Squibb Co., Pharmaceutical Research Institute,+ Box 4oo0, Princeton, New Jersey 08543; Cardiovascular Clinical: Margot Mellies(4), Mark McGovern(4). and Savian Nicholas(4); Biometrics & Data Management: Sharon Anderson(4). Jong-Soon Park(4), Murray Bamhart(4), and Joseph Fortunate(4).
TBtistol-Myers Squibb [BMS], the sponsor of all 3 trials, IS also supportlng the PPP colloboratlon and helping with the coordination of the prolect. However, o/l analy ses and assessment of results WI// be conducted Independently of the sponsor and Its representotlves
CARDIOLOGY/PROSPECTIVE
PRAVASTATIN
POOLING
PROJECT
905
mmol/L (2 13 mg/dl). The mean low-densi lipoprotein cholesterol level is 4.2 mmol/L (162 mg/ a, ). The mean blood pressure level is 134/81 mm Hg and 20% are current smokers. Half of the PPP participants have had a prior myocardial infarction. More than 7% have a history of diabetes and 26% have a histo of hypertension. PPP is projected to have data on aTo ut 1,100 CAD deaths, 500 non-CAD deaths, and >l,OOO cancers by study completion. Based on a pooled population of 19,768 patients, with approximately 1,600 deaths and 100,000 patient-years of follow-up, PPP should have oad power to examine the effects of treatment on tota Bmortality, CAD mortality, and cancer incidence, and to determine effects on total CAD events in important subgroups including the elderly, women, diabetics, and those with lower serum cholesterol levels (~5.5 mmol/L (2 13 mg/dll). (Am J Cardiol 1995;76:899-905)
he ProspectivePravastatinPooling (PPP)project was T initiated in 1992to addressthe effectsof pravastatin on all-cause and coronary mortality, on total coronary
vention With Pravastatin in Ischemic Disease (LIPID) study,17the Cholesterol and Recurrent Events (CARE) study,l8 and the West of Scotland Coronary Prevention Study (WOSCOPS).19Collectively, these studies will involve 19,768patients, and by the time of their completion will provide outcome data from nearly 100,000person-years of follow-up. The purpose of this study is to describe the objectives of the project, the trials involved, the baseline characteristicsof the study populations, the definitions of outcome events, and the power to test the study hypotheses.
events in special populations, and to confirm long-term safety. At that time, trials of cholesterol-lowering therapy, and meta-analysesof thesetrials, had clearly demonstrated reductions in coronary events,1-9but the question of the effects of cholesterol lowering on overall mortality remained unresolved for all patient groups.c9 With the advent of the hydroxymethylglutaryl coenzyme A reductaseinhibitors, it has becomepossible to test the effects of larger cholesterol reductions on coronary events and overall mortality.‘0-19 PPP is a prospective METHODS Objectives: The primary goal of the PPPproject is to collaboration involving 3 ongoing, long-term, large-scale, monotherapy trials of pravastatin: the Long-Term Inter- pool the data from the 3 large randomized studies of pravastatin to addressscientific questionsfor which there may be inadequatepower within the individual studies: Manuscript received May 19, 1995; revised manuscript received in particular, the effect of pravastatin on total mortality and accepted July 24, 1995. Address for reprints: Curt Furberg, MD, Department of Public and on specific non-CAD events as well as the effect on Health Sciences, Bowman Gray School of Medicine, Winston-Salem, CAD morbidity within important subgroups. The proNorth Carolina 27157-l 063 ject will also be able to explore the consistencyof effects across the constituent trials and help establish how to *A complete listing of PPP investigators and their affiliations IS given rn the Appendix. generalize the findings from the trials, as well as provide PREVENTIVE
CARDIOLOGY/PROSPECTIVE
PRAVASTATIN
POOLING
PROJECT
899
Three pravastatin trials: The prevention trials that form the basisfor the Pravastatin Pooling PPP-LIPID, CARE, and WOSCOPS Acronym -are described in Table I. These 3 studies have been selected for incluLIPID CARE woscoPs sion in the PPPproject becauseof their Cholesterol and Long-Term West of Scotland Name of trial common design features, including Intervention with Coronary Prevention Recurrent Events drug, dose, and duration, and the closePravastatin in itudy nessof their completion dates.Imporlschemic Disease Australia and USA & Canada Scotland location tantly, they have each been selected New Zealand without prior knowledge of the out9,000 6,000 4,000 Target sample size come and they should therefore pro9,014 6,595 Observed sample size 4,159 vide an unbiasedestimateof the effects 5+ 5 5 Mean follow-up [yr) May 31, 1995 of pravastatin. Furthermdre, each of February 1997 Target termination February 12, 1996 46,200 20,975 32,140 Patient-years the researchquestions to be addressed Fatal 81 nonfatal Fatal CAD Fatal CAD & nonfatal Primary end point by PPP and a detailed analysis plan myacardial infarction myocardial infarction have been carefully defined in adv87 80 1 centralized Sites ance, without any knowledge of trial + 40 satellite + 60 trial centers Major eligibility outcome by treatment. In this way, recriteria sultsof PPPwill provide much stronger Myocardial infarction Myocardial infarction No evidence of Clinical evidence than simply undertaking a 3-20 months before or unstable anaina myocardial infarction retrospective meta-analysisat the conpectoris 3-36 months ’ randomization before screening clusion of the 3 trials. Men and women Men and women Men only Gender Each of these studies is a random31-75 21-75 45-64 Age at entry (yr) ized, double-blind, placebo-controlled, 4.0-7.0 mmol/L e6.2 mmol/L Not applicable Total cholesterol clinical trial of 25 years’ duration using [155-270 mg/dl) (~240 mg/dl) a dose of 40 mg of pravastatin once Not applicable >4.0 mmol/L on 3.0-4.5 mmol/L LDL cholesterol (115-l 74 mg/dl) 2 occasions; daily, and each has been described in b4.5 mmol/L detail previously.*‘-19LIPID is a secat least once ondary prevention trial in 9,014 men but e6.0 mmol/L and women with a history of myocarat least once (155-l 74 mg/dl) dial infarction or hospitalization for unstable angina. CARE is a secondary CAD = coronary artery disease; LDL= lowdensiv lipoprotein cholesterol. prevention trial in 4,159 men and women with a history of myocardial reliable information on various outcomes attributed to infarction and normal or only mildly elevatedcholesterol levels. The WOSCOPSis a primary prevention trial in cholesterol reduction. Specific primary obiectives of PPP:Specific objectives 6,595 men who have had no documented myocardial are: (1) to determine the effect of pravastatin therapy on infarction. PPPoutcomes: ALL-CAUSE MORTALITY: Approximateall-causemortality both for the secondaryprevention trials (LIPID and CARE) and for the 3 trials combined; ly half of the deaths in WOSCOPS are expected to be and (2) to determine the effect of pravastatin therapy on due to CAD versus X to %of the deaths in CARE and cause-specificmortality: CAD mortality, and other car- LIPID. Becausethe proportion of all deathsdue to CAD diovascular and noncardiovascular mortalities. The spe- will differ between the secondary and primary prevencific secondary objectives of PPP are: (1) to determine tion trials, the percent reduction in all-cause mortality the effect of pravastatin therapy on total coronary inci- for a given effect on CAD will also differ. Consequentdence (fatal CAD and nonfatal myocardial infarction), ly, the primary PPP analysis of all-cause mortality will and particularly in prespecified subgroups of special comprise both the secondarypopulations in CARE and interest (primary vs secondary prevention, the elderly, LIPID and, separately,an analysis of all 3 trials includmen vs women, diabetics, current smokers,those taking ing WOSCOPS. CARDIOVASCULAR DISEASE MORTALITY: Three C&eantihypertensive treatment, and specific categories of baseline lipid levels including those with total choles- gories of fatal secondary cardiovascular outcomes will terol ~5.5 rnmol/L); and (2) to determine the effect of be studied: (1) fatal CAD (i.e., fatal myocardial infarcpravastatin therapy on total (fatal plus nonfatal) non- tion, sudden death or fatal CAD, separatelyor in comcoronary events(specifically cancer,trauma, and stroke). bination); (2) other fatal cardiovascular disease (i.e., Additional outcomes of interest will be explored in stroke, aortic aneurysm, and iliofemoral disease, sepadescriptive analysesof the effects of treatment on other rately and in combination); and (3) any fatal cardiovasischemic events, the need for invasive coronary proce- cular disease(categories 1 + 2). This will be examined dures,andhospitalization for prcdefinedcategoriesasout- for all patients, those with established CAD, and those lined later. Combinedeffectson cause-specificmortalities with prior myocardial infarction. NONCARDIOVASCULAR DISEASE MORTALITY: Mortali(including cancer and trauma) will also be used to estity will be described separatelyfor total and cause-spemate effectson total mortality for other populations. TABLE I Characteristics
and Eligibility (PPP) Project
900
THE AMERICAN
JOURNAL
Criteria
of the Trials in the Prospective
OF CARDIOLOGY@
VOL. 76
NOVEMBER
1, 1995
79 * 11
1.8 * 0.7 (156 * 61) 128 * 18
1 .o * 0.2 (38 zt 8)
3.6 zt 0.4 (139 * 15)
5.4 * 0.4 (208 zt 17)
Men (n = 3,583)
70 * 11
78 * 11
1.7 f 0.7 (154 zt 61) 134 f 20
1.2 f 0.3 (45 * 1 1)
3.6 * 0.4 (140 * 14)
5.6 zt 0.4 (215 * 15)
Women (n = 576)
67 i 11
78 zt 11
1.8 * 0.7 (156 * 61) 129 * 18
1 .o * 0.2 (40 f 7)
3.6 * 0.4 (139 f 15)
5.4 f 0.4 (209 * 17)
Total (n = 4,159)
10 67 23 27 f 4
69 zt 11
81 i 11
1.8 f 0.8 (158 zt 74) 133 f 19
0.9 f 0.2 (36 i: 8)
3.8 * 0.7 (149 * 26)
5.6 zt 0.7 (216 i 28)
Men (n = 7,503)
10 46 44 27 * 5
71 * 11
80 * 11
1.7 * 0.8 (155 * 70) 138 * 20
1.1 * 0.3 (43 * 1 1)
4.0 f 0.7 (156 zt 26)
5.9 * 0.7 (229 zt 27)
Women (n = I,51 1)
10 63 17 27 * 4
89 zt 11
81 zt 11
1.8 f 0.8 (158 * 73) 134 f 19
1 .o * 0.2 (37 * 9)
3.9 * 0.7 (150 ziz27)
5.6 ZJZ 0.7 (218 f 28)
Total (n = 9,014)
35 39 26 26 zt 3
73 * 9
84i
10
1.8 +z0.8 (163 * 69) 136 * 17
1.2 * 0.3 (47 f 10)
5.0 f 0.5 (192 * 17)
7.0 st 0.6 (272 i 23)
Men Only (n = 6,595)
20 56 24 27 * 4
70 f 11
81 * 11
1.8 zt 0.8 (160 * 70) 133 5 18
1 .o f 0.3 (40 f 10)
4.2 * 0.8 (163 * 31)
6.1 * 1.0 (235 * 37)
Men (n = 17,681)
13 46 41 27 iz 5
70 f I 1
79 f 11
1.7 * 0.8 (155 i 67) 137 f 20
1.1 f 0.3 (43 * 11)
3.9 e 0.6 (151 * 24)
5.8 * 0.6 (225 i 25)
Women (n = 2,087)
20 55 25 27 * 4
70 f 11
81 * 11
1.8 * 0.8 (159 f 70) 134 * 19
1.1 f 0.3 (41 * 10)
4.2 i 0.8 (162 +z31)
6.1 i 0.9 (234 i 38)
Total (n = 19,768)
PPP
zt 11
16 61 23 28 * 4
woscoPs
67
20 47 33 28 i 6
LIPID
16 63 21 28 iz 4
50 7 26 of population).
70 13 47 as percent
48 7 24 are given
0 1 11 which
64 9 39 disease,
59 10 51 of other
65 8 34 history
100 14 29 and
100 20 37 status
100 13 27
CARE
TABLE II Baseline Characteristics and Risk Factor Levels of Study Population
Risk Factor Total cholesterol mmol/L (w/4 LDL cholesterol mmol/L h/4 HDL cholesterol mmol/L (w/4 Triglyceride mmol/L h/4 Systolic blood pressure (mm Hd Diastolic blood pressure (mm Hgl Heart rate (beats/min) Cigarette smoker Current Former Never Body mass index (kg/mz) History Prior myocardial infarction Diabetes mellitus Hypertension All values ore expressed OS mean * SD (except for cigarette smoker HDL = highdensity lipoprotein; other abbreviations ~1s in Table I.
TABLE III
Estimates of Statistical
Power Among
Outcome
Subgroup
Total mortality
All LIPID + CARE All Prior myocordial infarction A)) All All All Men Women
CAD
mortality
Non-CAD mortality Cancer (fatal + nonfatal) Trauma (fatal + nonfatal) Totol CAD events (Fatal CAD + nonfatal myocardial infarction)
Age iv1 <55 55-64 265 270 Prior myocardial infarction Current smoker Hypertension Diabetes mellitus Cholesterol ~5.5 LDL-C ~3.5 HDLC cl .O Triglyceride >2.5
PPP Populations Number of Patients
Projected Events*
19,768 13,173 19,768 9,91 1
1,600 1,400 1,100 650
0.78 0.77 0.92 0.73
(1213 (13)Z
19,768 19,768 19,768 19,768 17,681 2,087
500 1,150 600 2,900 2,600 300
0.52 0.88 0.60 >0.99 >0.99 0.82
(-2018 (-20)s (-20)s (27)tl (27) (27)
6,635 8,286 4,843 1,841 9,911
600 1,100 1,200 600 2,200
0.98 aO.99 >0.99 0.98 >0.99
(27) (27) (27) (27) (27)
850
>0.99 >0.99 0.91 BO.99 0.99 >0.99 >0.99
(27) (27) (27) (27) (27) (27) (27)
3,871 5,209 1,427 6,198 4,53 1 7,254 3,275
1,100
350 1,100 700 1,100
600
Power to Detect % Reduction (%)t
(18)
(18)
‘Projected events for each category are bared on (1) the original assumptions of each protocol, that each trial contin(2) estimated event rates for the different categories, and (3) th e assumption ues until its planned completion. tPower to detect o difference using 2-sided 5% level test. tDifference in total mortality (12% to 13%) bused on on assumption of an 18% difference in coronary artery disease (CAD) mortality and no effect on non-CAD mortality. 5Possible increases in non-CAD deaths of ~20% have been suggested in published studies and ore shown for illustrative purpaes only. IiDifference in total CAD events of 27% is based on on assumption of on 18% difference in CAD mortality and 50% greater effect for fatal + nonfatal CAD events combined. C = cholesterol; other abbreviations os in Tables I and II. Lipid values ore meowed in mmol/L.
angioplasty or other nonsurgical procedures) or surgical revascularization (e.g., coronary artery bypassgraft surgery) procedures, total episodes of congestiveheart failure or unstable angina (i.e., fatal or hospitalizations for nonfatal), major hepatic disease,hospitalization for psychiatric illness, and occurrenceof myopathy.PPPwill also be able to examine for other potential adverseeffectsthat are suspectedto be drug-related. Effects of treatment on population subgroups (effect modification): The
effect of pravastatin on total coronary incidence will be investigated in specified subgroupsof the population. The following subgroups will be included in an analysisof a modification of treatment effect by risk factors: Age (~55, 55 to 64,265, and 270 years); gender; baseline lipid levels (total cholesterol ~5.5, 5.5 to 6.4, and 26.5 mrnol/L [<213, 213 to 249, and 2250 mg/dl]; low-density lipoprotein cholesterol <3.5, 3.5 to 4.4, 24.5 mmol/L
[<135,
135to 174,and 2175 mg/dl]; high-density lipoprotein cholesterol c 1.0, 2 1.0 mmol/L [c38.7, 238.7 mg/dl]; and triglycerides ~1.5, 1.5 to 2.4, and 22.5 mmol/L [<133, 133 to 219, and 2220 mg/dl]); smoking (classifiedby current smokers, former [at time of randomization], and never); hypertension (patients _._.^ receiving antihypertensive drug _ cific noncardiovasculardiseases.In particular, categories therapy); prior myocardial infarction; and prior estabof noncardiovascular fatal outcomesto be described are lished CAD and diabetesmellitus. The relative effectsof (1) cancer mortality: deaths resulting from cancer over- pravastatinaccording to age,lipid levels, and blood presall and listed by body system and tumor type; (2) trau- surewill also be explored using theseas continuous varimatic death (accidental death and suicides); and (3) oth- ables, and the constancy of the relative risk reduction er noncardiovascular deaths, including cirrhosis and acrossa compositerisk scorederived from thesevariables non-cancer pulmonary disease. Because of the small will be examined.The effectswithin these subgroupson number in eachcategory,a simple tabulation will be pro- cause-specificand total mortality will also be reported. Evaluation and adjudication of the PPP outcomes: The vided for each study and for the combined population. detailed definitions and methods for classification of TOTAL CORONARY INCIDENCE: The effect Of praVaSti+ tin on the total incidence of any fatal CAD plus nonfatal major events of interest and outcomesin the prevention myocardial infarction will be investigated.Each of the 3 studies have been agreedupon by the PPPsteering comtrials has good power to detect a modesttreatmenteffect mittee and have been incorporated in the full PPP proon the combined incidence of fatal and nonfatal CAD for tocol (copy available upon request). These definitions its own study population. The combinedPPPanalysiswill correspondwith those usedin the 3 trials, facilitating the addressthe questionof whether the relative risk reduction direct use of individual trial data. PPP analysis plan: All primary researchquestions for in fatal CAD is different from that in nonfatal CAD, as suggestedin previous meta-analyses,1~2~20 and will exam- PPPand the relevant trials to be included for each quesine for possible effect modification by other risk factors tion are specified in advance in the protocol. The intention-to-treat model of analysis will be used. An ancil(see subgroupsmentioned later). OTHER OUTCOMES OF INTEREST: PPP will also inves- lary analysis estimating treatment effect for compliers tigate the effect of pravastatin on the following out- will be consideredby making appropriate adjustmentof comes: total (i.e., fatal plus nonfatal) cancer,total trau- the intent-to-treat analyses, rather than performing an matic or violent events, and total stroke, each of which analysis by treatment received.21 Primary analyses will be presented unadjusted for are specific secondary objectives. In addition, descriptions will be provided on the outcomesof cardiac cathe- other covariates. Analyses adjusted for other covariates terization, nonsurgical revascularization (e.g., coronary may be undertaken as ancillary analyses. All analyses 902
THE AMERICAN JOURNAL OF CARDIOLOGY”
VOL. 76
NOVEMBER 1, 1995
will be stratified by trial. Primary analyseswill be based on time-to-event data using the stratified log rank test. Analyses adjusting for patient covariates will use a stratified Cox regression approach,provided the proportional hazardsassumption is not violated. If this assumption appearsinvalid, an analysis allowing for a delayed treatment effect will be performed using a time-partitioned Cox regression analysis.22Time-to-event mortality plots will be used for the main analyses.In addition, the total number of eventsin different categorieswill be presented (e.g., the number of cancers and trauma events). When appropriate, analysis of 2 X 2 tables will be based on the stratified Mantel-Haenszel test.23Relative risks will be estimated to determine the percent risk reduction associatedwith pravastatin treatment.Absolute risks will be estimated to determine the number of events averted per 1,000 treated patients. Since detailed individual data will be available from eachtrial, the analysis is a stratified pooling of the 3 studies. Heterogeneity of cause-specific treatment effects acrossthe study populations will be examined, although it is recognized that the power to detect such interactions is low. A model basedon a weighted combination of the effectsseenon each specific causeof death will be developed, enabling estimation of an “expected” net mortality reduction for any given risk population. Basedon the proportion of different types of deaths in each trial, this will enable more robust study-specific reductions in total mortality to be estimated.24 Subgroups have been prespecified based on either biologic rationale or scientific importance. The characteristics previously noted (under effect modification) may be expandedbefore any of the individual trial results are known, in the light of new data emerging from other studies. Tests for “treatment by subgroup” interaction will be undertakento help in the interpretation of results. Examination of effect modification will include a multivariate model incorporating all variables (as previously listed under effect modification) and their treatmentinteraction terms. Results for each subgroup will be presentedas p values and confidence intervals, adjustedfor covariates if necessary,with an appropriate discussion indicating that the results do not take into account multiple comparisons and hence the need for some adjustment or caution in interpreting the results. A simple data set comprising all baseline characteristics and major outcome variables will be created from each trial. These data will enable the time-to-event outcomes as well as individual patient covariates to be examined. The stratified pooled analysesfor PPPwill be undertaken independently of the sponsor by the members of the PPP steering committee. It is proposed that this analysis occur as soon as the last of the 3 trials, presumably LIPID, has been completed. Project organization and publication policy: A governing body, the steering committee, was established in October 1992. It contains representativesfrom all 3 trials, together with an epidemiologist responsible for assisting the investigators with the data analysis and a PPP chairman, both independent of the 3 trials. Three working groups, with representation from each constituent trial, were created to organize and guide the
design and analysis of the project, objectives, and publications, and outcomes and end point definitions. All analysesand assessmentof the results will be conducted independently of the sponsor and its representatives. Publication of theseresults will be made in the name of the PPPproject group and only after the principal results of the 3 individual trials are published.
RESULTS The protocols and case report forms were reviewed and comparedfor the 3 studies participating in the PPP project. Table I describes the participating studies and key entry and enrollment information. Table II presents the baseline characteristicsfor the 3 individual trials and the combined PPP population. The mean age of the 19,768participants at baseline is 58 years, with women being, on average,4.5 years older than men. This overall gender difference in age is primarily a function of the trials contributing the data. The primary prevention trial, WOSCOPS,is composedentirely of men, and it has the youngest study population. Risk factor distributions are presented in Table II. More than 2,000 (or 10%) of the participants in the pooled data set are women; half of the PPP participants have had a prior myocardial infarction, >7% have a history of diabetes,and 26% have a history of hypertension and antihypertensive medication. It is noted that while the WOSCOPSpopulation (free of evidence of rnyocardial infarction by definition) had significantly higher lipid levels and a significantly greater proportion of current smokers, it had the lowest prevalences of diabetes and hypertension. Power calculations: One criterion in the selection of researchquestionsto be examined with PPP is the need for adequatepower to detectplausible and important differences in outcomes. Table III illustrates the approximate power PPPhas to detect such differencesbasedon a projected number of events within each category.This projectednumber provides only a rough guide to the final number of eventsthat will occur, and is basedon the original assumptionsof each protocol, the estimated event rates for the different categories,and the assumptionthat each trial continues until its planned completion. Based on original protocol assumptions,a reduction in coronary mortality that might be expected is 18%, if an average reduction in total cholesterol of 18% was obtained and there was 1% reduction in coronary mortality for each 1% reduction in cholesterol achieved.2*6 The power of the combined analysis to detect such an effect is high (92%) and fair (73%) for the subgroup of patients with myocardial infarction. If one assumesno beneficial or adverseeffect on non-CAD mortality, this would translate to a reduction in total mortality of about 12% for the 3 trials, with reasonable power (78%) to detect such an effect. However, these assumptions are conservativeand a considerably larger reduction in CAD mortality, such as 1.6% for every 1% reduction in cholesterol, as observed in the Scandinavian Simvastatin Survival Study (4s) trial group12is also possible. This would correspond to a reduction in total mortality of =20%, and the power of PPP to detect such an effect would be very high (>99%).
PREVENTIVE CARDIOLOGY/PROSPECTIVE
PRAVASTATIN POOLING PROJECT
903
It will also be important that PPP exclude possible adverseeffects on non-CAD deaths or serious nonfatal outcomes,such as a 20% increasein theseevents.However, with approximately 500 non-CAD deaths projected, PPP will only have about a 1 in 2 chance (52%) of detecting a 20% increase in non-CAD deaths. For this reason, fatal plus nonfatal potential adverseeffects will also be examined: for cancers with >l,lOO projected events,PPPwould have adequatepower to detect a 20% increase, although it would only have a 60% power to detect a 20% increase in trauma. For subgroup analyses,PPP will examine the effects of pravastatin on the combined outcome of CAD events (fatal CAD + nonfatal myocardial infarction). For this outcome,a larger effect of 27% reduction in events(50% more than for fatal CAD) has been assumed.PPP will have very good power to detect such an effect for almost all subgroupslisted; even among women (with the least projected number of events), PPP will still have >80% power.
increased power to examine the effects of treatment on total mortality and on total CAD eventsfor many of the clinically important subgroupsconsideredfor treatment. The size of the PPP population of 19,768followed for >5 years, with nearly 100,000patient-years, 1,600 fatal events, and twice as many nonfatal events expected, offers a unique opportunity to explore these issues. In contrast, for example, 4s yielded =23,000 patient-years of experience and 438 fatal events. The similarity of design of the individual studies,the use of a single therapeutic agent used for 5 to 6 years, and the prospectively defined objectives of the project, along with predefinedend point definitions, are strengths of the project, allowing pooling of individual data and interpretation of the results with little likelihood of bias. Power estimatesindicate a high likelihood that the PPP project will detect beneficial effectsof pravastatin on the sizesanticipated from all available evidenceon all-cause and on coronary mortality in the total population, and generally on total CAD in the subgroupsof special interest.
DISCUSSION Epidemiologic and clinical trials researchhave established the causalrelation of high serum cholesterol, particularly low-density lipoprotein cholesterol, to the risk of CAD and its reversibility.1-9,20,25-27 However, there has been some concern that reducing cholesterol may also carry some risk,~8,2~28and until recently, trials of cholesterol lowering, by diet or drug therapy,have failed to provide unequivocal evidence that such strategiesprolong life. The recently completed 4S12involving 4,444 patients has now produced striking evidence of such a survival benefit over 5 to 6 years among hypercholesterolemic patients (aged 35 to 70 years) (total cholesterol 5.5 to 8.0 mmolL [212 to 310 mg/dl]) with preexisting CAD. These results are likely to lead to broad changes in physician practice, particularly for patients similar to those studied in 4s. DesDite this imoortant result, 4s alone still did not have sufficient power to addressthe question of the effects on overall mortality in women, or on major non-CAD events, and the issue of how widely theseresults should be generalizedacross age,gender,lipid levels, absolutelevels of risk for CAD, choice of lipid-lowering drug, and so on, remains to be determined. Four placebo-controlled trials of pravastatin, designed principally to assessthe effectsof pravastatin on changes in atherosclerosis, have demonstrated modest effects on atherosclerosis,but also collectively demonstrated significant reductions in cardiovascular clinical events.13-15,29,30 A planned meta-analysisof these studies demonstrateda 62% (95% confidence interval 38 to 80) reduction in risk of total myocardial infarction and a 53% (95% confidence interval 29 to 71) reduction in the combined outcomes of nonfatal myocardial infarction or death from any cause.29+30 However, even in combination, thesestudieswere not of sufficient size to detect statistically significant effectson CAD mortality or total mortality. The PPP project will be able to address important questions that are beyond the scope of the 3 individual contributing studies. Combining the data sets provides I
904
THE AMERICAN JOURNAL OF CARDIOLOGY@
I
VOL. 76
Acknowledgment: Bristol-Myers Squibb Pharmaceutical ResearchInstitute, Princeton, New Jersey,the sponsor of all 3 trials, is also supporting and helping with the coordination of the ProspectivePravastatinPooling project. We gratefully acknowledge the secretarial assistance of Edel Williams and Dianne Black in the preparation of the manuscript.
I. Holme I. An analysis of randomized trials evaluating the effect of cholesterol reduction on total mortality and coronary heart disease incidence. Circularion 1990;82:1916-1924. 2. Pete R, Yusuf S, Collins R. Cholesterol-lowering trial results in their epidemological context. Circularion 1985;72(suppl III):III-451. 3. Rossouw JE, Lewis B, Riikind BM. The value of lowering cholesterol after my@ cardial infarction. N Engl J Med 1990;323: 1112-l 119. 4. Lipid Research Clinics Program. The Lipid Research Clinics Coronary Primary Pmvention Trial I. Reduction in incidence of coronary heart disease. JAMA 1984: 251:351-364. 5. Frick MH, Elo 0, Haapa K, Heinonen OP, Heinsalmi P, He10 P, Huthmen JK, Kaitaniemi P, Koskinen P, Manninen V, MBenpti H, Miilkiinen M, Mitt$ri M, Norola S, Pastemack A, Pikkarainen I. Rome M, Sjiiblom T, NikkilP EA. Helsinki Heart Study: primary prevention trial with gemfibrozil in middle-aged men with dyslipidemia. Safety of treatment, changes in risk factors, and incidence of corenary heart disease. N Engl JMed 1987;317:1237-1245. 6. Keech AC. Does cholesterol lowering reduce total mortality? Postgrad Med J 1992;68:870-871. 7. Davey-Smith G, Pekkanen J. Should there be a moratorium on the use of cholesterol lowering drugs? Br Med J 1992;304:431&434. 8. Muldoan MF, Manuck SB, Matthews KA. Lowering cholesterol concentrations and mortality: a quantitative review of primary prevention trials. Br Med J 1990; 301:309-314. 9. Collins R, Keech A, Pete R, Sleight P (ISIS), Kjekshus J, Wilhelmsen L (4s). MacMahon S, Shaw J, Simes J (LIPID), Braunwald E, Boring J, Hennekens C, Pfeffer M, Sacks F (CARE, Women’s Health Study, Physicians’ Health Study, SAVE), Probstfield J, Yusuf S (Post-CABG Study), Downs JR, Gotto A (AFCAPSI TEXCAPS), Cobbe S, Ford I, Shepherd J (WOSCOPS). Cholesterol and total mcrtality: need for larger trials [letter]. Br Med J 1992;304:1689. 10. Illingsworth DR. Clinical implications of new drugs for lowering plasma cho‘ester01 concentrations. Drugs 1991;41:151-160. I 1. McGovern ME, Mellies MJ. Long-term experience with pravastatin in clinical research trials. Clin Ther 1993;15:57+54. 12. Scandinavian Simvastatin Survival Study Group. Raodomised trial of cbolestern1 lowering in 4444 patients with coronary heart disease: the Scandinavian Simvastatin Survival Study (4s). Lancef 1994;344:1383-1389. 13. Pitt B, Man&i JGB, Ellis SG, Rosman HS, McGovern ME, for the PLAC I Investigators. Pravastatin Limitation of Atherosclerosis in the Coronary Arteries (PLAC I): reduction in atherosclerosis progression and clinical events. J Am Coil
cardi0’ prrss). 14. Crow(inJR, Byington RP, Bond MG, Espeland MA, Craven TE, Sprinkle JW, McGovern ME, Furberg CD. Pravastatin, Lipids and Atherosclerosis in the Carotid NOVEMBER 1, 1995
Arteries (PLAC 11).Am J Cardiol 1995;75:455459. 15. Jukema IW, Btuscbke AVG, van Boven Ad J, Reiber JHC, Bal ET, Zwinderman AH, Jansen H, Boerma GJM, van Rappard FM, Lie KI, on behalf of the REGRESS study group. Effects of lipid lowering by pravastatin on progression and regression of coronary artery disease in symptomatic men with normal to mcderately elevated serum cholesterol levels. Tbe Regression Growth Evaluation Statin Study (REGRESS). Circulation 1995;91:‘2528-2540. 16. Cholesterol Treatment Trialists’ (CTT) Collaboration. Protocol for a prospec tive collaborative overview of all current and planned randomized trials of cholesterol treatment regimens. Am J Cardiol 1995;75:113&1134. 17. The Lipid Study Group. Design features and baseline characteristics of the LIPID (Long-Term Intervention with Pravastatin in Ischemic Disease) study: a randomized trial in patients with previous myocardial infarction and/or unstable angina pectoris. Am J Cardiol 1995;76:474-479. 18. Sacks FM, Rouleau JL, Moye LA, Pfeffer MA, Wamica JW, Arnold JM, Nash DT, Brown LE. Sestier R, Rutherford J, David BR, Hawkins M, Braunwald E for the CARE Investigators. Baseline Characteristics in the Cholesterol and Recurrent Events (CARE) trial of secondary prevention in patients with average semm cholesterol. Am J Cardiol 1995;75:621423. 19. The WOSCOPS Study Group. Baseline characteristics and screening expcrience in the West of Scotland Coronary Prevention Study. Am J Cardiol 1995; 76~485-491. 20. Law MR. Wald MI. By how much and how quickly does reduction in semm cholesterol concentration lower risk of ischaemic heart disease? Br Med J 1994: 308~367-372. 2 1. Newcombe RC. Explanatory and pragmatic estimates of the treatment effect when deviations from allocated treatment occur. Stat Med 1988;8:1179-I 187. 22. Grambsch PM, Themeau TM. Proportional hazards tests and diagnostics based on weighted residuals. Biometrika 1994;s 1:5 15-526. 23. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 1959;22:71%748. 24. Pocock S, Geller N, Tsiatis A. The analysis of multiple end points in clinical trials. Biometrics 1987;43:487-498, 25. Anderson KM, Cast& WP, Levy D. Cholesterol and mortality: 30 years of follow-up from the Framingham study. JAMA 1987;257:2174-2180. 26. Martin HJ. Hulley SB, Browner WS, Kuller LH, Wentwortb D. Serum cholesterol, blood pressure and mortality; implications from a cohort of 361,662 men. Lancer 1986;2:933-936. 27. Davey Smith G, Song F, Sheldon TA. Cholesterol lowering and mortality: the imponance of considering initial level of risk. Er Med J 1993;306:1367-1373. 28. Law MR, Thompson SG, Wald NJ. Assessing possible hazards of reducing serum cholesterol. Br Med J 1994;308:373-379. 29. Furberg CD, Byington RP, Grouse JR, Espland MA. Pravastatin. lipids, and major coronary events. Am J Cardiol 1994:73:1133-l 154.
PREVENTIVE
30. Byington RP, Jukema JW, Salonen IT, Pitt B, Bruschke AV, Hoen H, Furberg CD, Man&i GBJ. Reduction in cardiovascular events during pravastatin therapy: pooled analysis of clinical events of the pravastatin atherosclerosis intervention program. Circulation (in press).
APPENDIX Prospective Pravastatin Pooling (PPP) Project Investigators: (I ) = Subcommittee Chairperson; (2) = End Points Definitions & Outcomes; (3) = Design and Analysis; (4) = Objectives & Publications. Curt Furberg(4) PPP Project Chairman: Department of Public Health Science, Bowman Gray School of Medicine, Winston-Salem, North Carolina; Robert P. Byington(4) PPP Project Coordinator: Depanment of Public Health Science, Bowman Gray School of Medicine, Winston-Salem, North Carolina: Jennifer L. Baker(4): NHMRC Clinical Trials Ctr., University of Sydney, Sydney, Australia; Stuart M. Cobbe(4,l) Department of Cardiology, University of Glasgow, Glasgow, Scotland; Barry Davis(4): School of Public Health, University of Texas, Houston, Texas; Ian Ford(4): Director, Databases Unit, University of Glasgow, Glasgow, Scotland; Paul Glasziou(4): Department of Social & Preventive Medicine, University of Queensland, Herston Qld, Australia; C. Morton Hawkins(4): Director, Coordinating Center for Clinical Trials, University of Texas, Houston, Texas: Anthony C. Keech(4): NHMRC Clinical Trials Ctr., University of Sydney, Sydney, Australia; Stephen MacMahon(4): Clinical Trial Research Unit, University of Auckland, Auckland, New Zealand: Lemuel Moy6(4): School of Public Health, University of Texas, Houston, Texas; Chris Packard(4): Department of Pathologic Biochemistry, University of Glasgow, Glasgow, Scotland: Marc A. Pfeffer(4): Brigham & Women’s Hospital, Harvard Medical School, Boston, Massachusetts; Frank M. Sacks(4,l): Brigham & Women’s Hospital, Hzvard Medical School, Boston, Massachusetts; James Shepherd(4): Professor of Pathological Bicchemisuy, University of Glasgow, Glasgow, Scotland; John Simes(4.1): Director, NHMRC Clinical Trial Cu., University of Sydney, Sydney, Australia; Andrew Tonkin(4): Department of Cardiology, Austin Hospital Heidelburg, Victoria, Australia; Bristol-Myers Squibb Co., Pharmaceutical Research Institute,+ Box 4oo0, Princeton, New Jersey 08543; Cardiovascular Clinical: Margot Mellies(4), Mark McGovern(4). and Savian Nicholas(4); Biometrics & Data Management: Sharon Anderson(4). Jong-Soon Park(4), Murray Bamhart(4), and Joseph Fortunate(4).
TBtistol-Myers Squibb [BMS], the sponsor of all 3 trials, IS also supportlng the PPP colloboratlon and helping with the coordination of the prolect. However, o/l analy ses and assessment of results WI// be conducted Independently of the sponsor and Its representotlves
CARDIOLOGY/PROSPECTIVE
PRAVASTATIN
POOLING
PROJECT
905