Clinical and Experimental Pharmacology and Physiology

(2002)

29

,

951–955

BRIEF REVIEW

CLINICAL TRIALS AND TRIBULATIONS

Gordon T McInnes

University of Glasgow, Department of Medicine and Therapeutics, Western Infirmary, Glasgow, UK

SUMMARY

1. Pharmacologists should be involved in all stages of drugdevelopment. Often neglected is the final step, the clinical trialsand other studies that determine clinical utility. The presentarticle illustrates how pharmacoepidemiology can facilitateevaluation of the clinical potential of different drugs used totreat hypertension.

2. The evidence base for the drug treatment of hypertensionis very strong. Large-scale outcome trials, largely based ondiuretics, indicate that stroke events are prevented to the extentexpected from blood pressure reduction, but there appears tobe a shortfall in the prevention of coronary heart diseaseevents.

3. On theoretical grounds, newer agents may be expected tohave benefits in coronary heart disease prevention beyondblood pressure reduction. Recent trials with angiotensin-converting enzyme inhibitors and calcium channel blockerssuggest no advantage over conventional drugs, but short-comings in these studies mean that each is uninformative.

4. Observational studies based on pharmacoepidemiologicalprinciples offer an alternative approach to evaluating out-comes in treated hypertensives.

5. Evidence from the Glasgow Blood Pressure Clinicdatabase suggest that there are outcome differences betweenantihypertensive agents. Angiotensin-converting enzymeinhibitor treatment is associated with a mortality advantage,whereas calcium channel blocker therapy is associated with apoorer prognosis. Preliminary findings from a primary caredatabase support these observations.

6. Long-term follow up of a well-documented high-riskclinical population may allow detection of outcome differencesnot apparent in relatively short-term clinical trials.

7. Appropriate interpretation of observational datanecessitates an understanding of the strengths and limitationsof observational data. Clinical pharmacologists have a critical

role in design and evaluation of pharmacoepidemiologystudies.

Key words: angiotensin-converting enzyme inhibitors,calcium channel blockers, clinical pharmacology, clinicaltrials, coronary heart disease, experimental pharmacology,hypertension, observational studies, pharmacoepidemiology,stroke.

INTRODUCTION

Pharmacology (experimental and clinical) is critical to all aspectsof drug evaluation. In the long and expensive evaluation of a newmolecule, the discipline is heavily involved in the early stages, butis often much less prominent in the final step that determines theeventual clinical utility of the therapy. Thus, clinical pharma-cologists are involved infrequently in clinical trial design andinterpretation.

Large-scale, randomized clinical trials are considered to providethe most reliable evidence of efficacy. However, if trials havedeficiencies in design, conduct or interpretation, findings can beinconclusive. The present article discusses some of the tribulationsthat have plagued major outcome trials in hypertension. In the faceof these difficulties, other approaches, including observationalstudies based on the principles of pharmacoepidemiology, meritconsideration.

CLINICAL TRIALS

The evidence base in support of the drug treatment of hypertensionis among the strongest in medicine. Prospective, randomizedclinical trials in 50 000 individuals have demonstrated conclusivelythat pharmacological reduction of blood pressure reduces the riskof cardiovascular events and all-cause mortality.

1

The magnitude of reduction in stroke events in the trials isexactly that predicted from long-term epidemiological studies forthe differences in systolic and diastolic blood pressure achieved.

1

However, reduction in coronary heart disease events, the predomin-ant complication of hypertension in Western populations, whilesignificant, is rather less than that expected. The shortfall incoronary heart disease prevention may be explained by thedeleterious metabolic effects of diuretics, which formed the basisof therapy in almost all trials.

1

The first challengers to diuretics were the beta-blockers. In the1980s, there were great expectations that these drugs would be

Correspondence: Dr Gordon T McInnes, University Department ofMedicine and Therapeutics, Western Infirmary, Glasgow, G11 6NT, UK.Email: gordon.t.mcinnes@clinmed.gla.ac.uk

Presented at the 35th Annual Scientific Meeting of the AustralasianSociety of Clinical and Experimental Pharmacologists and Toxicologists,Dunedin, New Zealand, 3–5 December 2001.

Received 1 February 2002; revision 7 May 2002; accepted 9 May 2002.

952

GT McInnes

superior to diuretics, particularly because, in patients with coronaryheart disease, beta-blockers reduce the risk of reinfarction ordeath.

2

This hypothesis was tested in a series of large-scale control-led trials. The results were inconclusive.

1

Although these compari-sons

provided

no

support

for

the

concern

that

the metaboliceffects of diuretics may have prevented a reduction in coronaryheart disease in hypertensive patients, this possibility could

not

beexcluded.

In

addition,

beta-blockers

have

the

potential

fordetrimental

metabolic

effects

and

further

work

was needed.Newer antihypertensive agents should avoid these unwanted

actions and could have advantages in cardioprotection. Largeoutcome trials with these drugs (angiotensin-converting enzyme(ACE) inhibitors and calcium channel blockers (CCB)) in hyper-tension were slow to appear. The first results were not publisheduntil 20 years after their introduction.

3–8

Although these trials werenecessary, the results were equivocal. None detected a significantdifference in coronary heart disease events between therapy basedon newer drugs and that based on conventional agents. However,the precision of comparisons in individual trials was weak, withwide 95% confidence intervals for differences (Table 1). Becauseclinically useful differences between therapies in coronary heartdisease events could not be excluded in any of the trials, indi-vidually the trials were not informative.

The Heart Outcomes Prevention Evaluation (HOPE) study

5

merits particular mention. Although not a comparison of a newdrug against conventional therapy in hypertensive patients, it hasbeen interpreted as showing an advantage of ACE inhibitionbeyond blood pressure control. That active therapy (ramipril) wasbetter than inactive therapy (placebo) in reducing heart attacks ishardly surprising, but the magnitude of the reduction was greaterthan that predicted from epidemiological data for the small differ-ence in blood pressure observed. However, the reduction in cardio-vascular events per mmHg difference in blood pressure was nogreater than that seen in other high-risk populations, such asdiabetes, treated with other forms of antihypertensive therapy.

9,10

Inthe absence of a positive control group, treated with another agentproviding equivalent blood pressure control, no definitive conclu-sions can be reached.

TRIBULATIONS OF CLINICAL TRIALS

The recent trials of newer drugs against older drugs have failed usbecause they were too small, the event rate was low, almost 50% ofdeaths were not cardiovascular and strokes were almost as commonas coronary heart disease events. The pattern of events in recenttrials (relatively low proportion of cardiovascular deaths and highrate of stroke events) is not anticipated from epidemiological dataor from trials comparing active therapy against placebo wherecardiovascular deaths and coronary heart disease predominate.

1

Inearly trials,

1

the ratio of coronary heart disease to stroke events was3 : 2, whereas in the recent trials

3,4,6,7

the ratio is 1 : 1. Therefore,the newer drugs appear to have been tested in an environment verydifferent from that in which older drugs were evaluated and thehypothesis of a shortfall in coronary heart disease prevention wasgenerated. Insufficient power and the large proportion of non-vascular mortality make it difficult to detect difference betweentreatments for the outcome of interest, coronary heart disease.Clinical trials provide short-term answers to long-term problemsand are, in effect, surrogates for real life, where treatment is oftengiven for a lifetime.

OBSERVATIONAL STUDIES

In the face of uncertainty from the trials, it is reasonable to examineother forms of evidence. Observational studies are generallyfrowned upon because the introduction of chance bias mayconfound interpretation. However, recently published meta-analyses

11,12

have indicated that well-conducted studies of this typecan provide estimates of treatment effects that are qualitatively andquantitatively similar to those from randomized controlled trials.

The Glasgow Blood Pressure clinic provides a convenientenvironment to observe the long-term effects of antihypertensive

Table 1

Relative risk of myocardial infarction and 95% confidenceintervals in comparisons of newer drugs and conventional drugs (diureticsand beta-blockers)

Trial Drug Relative risk 95% CI

CAPPP

4

ACE inhibitor 0.96 0.77–1.19STOP-2

3

ACE inhibitor 0.90 0.72–1.13CCB 1.18 0.96–1.47

NORDIL

7

CCB 1.16 0.94–1.44INSIGHT

6

CCB 1.09 0.76–1.58ALLHAT

8

AB 1.03 0.90–1.17

Where the 95% confidence interval (CI) includes unity, relative risk notsignificantly different from 1.0 (i.e. no difference between treatments).

ACE, angiotensin-converting enzyme; CCB, calcium channel blocker;AB,

�

-adrenoceptor antagonist; CAPPP, Captopril Prevention Project;STOP-2, Swedish Trial in Old Patients with Hypertension II; NORDIL,Nordic Diltiazem Study; INSIGHT, International Nifedipine GITS study:Intervention as a Goal in Hypertension Treatment; ALLHAT, Antihyper-tension and Lipid-Lowering Treatment to Prevent Heart Attack Trial.

Fig. 1

Kaplan–Maier survival curves for coronary heart disease mortalityin patients who had received angiotensin-converting enzyme (ACE)inhibitors but never calcium channel blockers (CCB; ––––), CCB but neverACE inhibitors (-·-·-·-) and neither (-----). A value of 1.0 represents 100%survival without coronary heart disease death and 0.8 represents 80%survival over time on the horizontal axis.

Clinical trials and tribulations

953

drug therapy. The clinic was established in 1968 and now has acomputerized database of almost 11 000 patients. This providesstandardized records of clinical findings, drug treatment andinvestigations. Patients are record-linked with the RegistrarGeneral (Scotland) and the West of Scotland Cancer Registry; thisprovides information on dates and cause of death and fatal and non-fatal cancer. The database has several attributes: (i) prolongedfollow up; (ii) a large number of subjects; (iii) structured clinicalrecords; (iv) accurate data on outcome; and (v) a high mortalityrate.

The potential value of this resource is illustrated in the findingsof a recent analysis of over 5000 patients prescribed antihyper-tensive medication at the Clinic from 1 January 1980 to 31December 1995. This population was selected to examine theinfluence of newer antihypertensive drugs on outcome. Bothdihydropyridine CCB and ACE inhibitors were introduced intoclinical practice in 1980. Of the patients included, 2295 receivedCCB, 1599 received ACE inhibitors, 3679 received beta-blockersand 3297 received diuretics. As expected, most patients receivedmore than one drug during the 16 year observation period.

The first objective was to address the question of whether CCBpredispose to cancer, as had been suggested in other observationalstudies.

13,14

No such association was found, but an unexpectedobservation was that incident (fatal and non-fatal) and fatal cancerswere reduced in patients who had received ACE inhibitors.

15

Survival

curves

indicated

that

benefit

was

apparent

only

after3–4 years of therapy, providing plausibility for this finding.

By the end of 1995, over 850 patients had died and all-causemortality was significantly lower in patients who had received ACEinhibitors compared with the rest. In observational studies, apossible explanation for such a finding is bias, which may arise if,by chance, ACE inhibitors had been prescribed for low-riskpatients. No such bias was identified and, in an analysis using theCox proportional hazard model

16

to adjust for confounding factors,the survival advantage in ACE inhibitor-treated subjects wasconfirmed. All-cause, cardiovascular and coronary heart diseasemortality was reduced significantly in patients who had receivedACE inhibitors compared with those who had never received ACEinhibitors.

TRIBULATIONS OF OBSERVATIONAL STUDIES

Observational studies are not without tribulations. As well as thepotential for confounding influences that may introduce bias, therecruitment of patients into the Glasgow Blood Pressure Clinicstudy cohort over time makes interpretation difficult. The definitionof time to event, the primary end-point, is a critical problem.

If time to event is related to the start of treatment with the drugof interest (e.g. ACE inhibitor), there is bias in favour of drugs usedroutinely as first-line therapy (diuretics and beta-blockers) becauseACE inhibitor treatment is started later (when the subject is older)and the survival period is shorter. In contrast, if the time ofcommencement is defined as the start of any antihypertensivetherapy, bias is in favour of the drug of interest, which is usuallyused as second- or third-line therapy, because the patient mustsurvive to receive that agent. The drug of interest may be con-sidered as a time-dependent variable, allowing for these biases, butthe interpretation of such data is complex and untested.

The tribulations associated with observational studies can bereduced by limiting the analysis to that which is possible. Thus, thecomparison should be between drugs with similar patterns of usage(i.e. second-line agents). In the context of the Glasgow BloodPressure Clinic database, this means comparison of patients treatedwith ACE inhibitors or CCB, drugs that were used conventionallyas add-on therapy in patients already receiving diuretics and/orbeta-blockers.

COMPARISON OF ACE INHIBITORS AND CCB

In the Glasgow Blood Pressure Clinic study, over 90% of CCB-treated patients received dihydropyridines (more than 50% long-acting preparations, nifedipine gastrointestinal transport system(GITS) or amlodipine). The comparisons were between patientswho had received an ACE inhibitor but never a CCB (

n

= 680), aCCB but never an ACE inhibitor (

n

= 1416), an ACE inhibitor andalso a CCB (

n

= 879) and those who had received neither an ACEinhibitor nor a CCB (

n

= 2232). After adjustment for patientcharacteristics (age, gender, epoch, smoking, blood pressure,creatinine, urea, antidiabetic, lipid-lowering and antiplateletmedication) all-cause mortality was reduced significantly inpatients who had received an ACE inhibitor (but never a CCB) andwas significantly increased in patients who were prescribed a CCB(but never an ACE inhibitor) compared with those who hadreceived neither. Patients who had received both classes had anintermediate risk of death. Qualitatively similar differences, but ofgreater magnitude, were seen for all vascular mortality andcoronary heart disease mortality.

Differences between treatment groups were particularly extremefor coronary heart disease mortality, where survival curves separ-ated early, unlike the findings for cancer, and continued to separateover 10 years of follow up (Fig. 1). This suggests that relativelyshort-term prospective randomized trials may underestimate long-term differences between treatments.

The high coronary heart disease mortality in CCB-treatedpatients may be explained if these drugs were given preferentiallyto patients with coronary heart disease and who may be expectedto have high coronary heart disease mortality. The CCB are used totreat symptomatic coronary heart disease, but treatment with themost frequently used anti-anginal drugs, beta-blockers, was notassociated with excess coronary heart disease mortality, suggestingthat the results were not confounded by indication.

Short-acting CCB (such as nifedipine capsules) have beenreported to increase coronary heart disease risk

17

and were usedwidely during the early period of observation. This could haveinfluenced the results. However, when coronary heart diseasemortality was examined in 4 year epochs between 1980 and 1995,the excess risk associated with CCB treatment did not decline withtime. Thus, increased coronary heart disease mortality seems tohave persisted even after the widespread introduction of long-acting dihydropyridines.

These findings may have substantial health implications. Basedon 857 deaths over 34 519 patient years, it can be estimated thatuse of ACE inhibitors may spare 35 deaths per 1000 patients treatedfor 5 years. In contrast, use of CCB may lead to the loss of 45 livesthat would otherwise be spared (Fig. 2). These calculations offer nosupport for the widespread use of CCB and the relative underuse ofACE inhibitors in hypertension.

954

GT McInnes

GENERALIZABILITY OF FINDINGS

The findings from the Glasgow Blood Pressure Clinic relate tomortality data in patients referred to a secondary/tertiary referralcentre. To assess whether these findings can be extrapolated tomorbidity data in patients managed in primary care, we haveutilized information from the UK General Practice ResearchDatabase.

18

The General Practice Research Database contains standardizedmorbidity and treatment data on several million patients. In apreliminary analysis, 117 965 hypertensive patients treatedbetween 1988 and 1997 were identified. Of these, 105 965 hadreceived treatment. To avoid potential confounding by indication,only those with no evidence of cardiovascular disease (

n

= 80 364)were included. Among these patients, 14 409 had subsequentcardiovascular events.

Compared with those who had received CCB but never ACEinhibitors (

n

= 17 981), patients who had received ACE inhibitorsbut never CCB (

n

= 15 004) had a significantly lower hazard ratiofor coronary heart disease morbidity. Findings were similar inthose patients who received ACE inhibitors (

n

= 11 249) or CCB(

n

= 12 494) as first-line therapy. For cerebrovascular events, thetrends were again significantly in favour of ACE inhibitor-treatedpatients.

Thus, the findings from the General Practice Research Databasesupport those from the Glasgow Blood Pressure Clinic. However,the preliminary General Practice Research Database analysis isadjusted, using Cox proportional hazard model only for age, genderand year of entry. Further adjustment is necessary to confirm theconclusions. In addition, in the comparison of ACE inhibitors andCCB as first-line therapy, approximately 25% of each group sub-sequently received a drug from the other class. Nevertheless, thegeneralizability of the findings is supported.

CONCLUSIONS

The evidence base for the drug treatment of hypertension is strong,but trials seeking evidence of outcome differences between activeagents have been uninformative. If differences are to be detected,future trials must have more coronary heart disease events andlonger duration. Meanwhile, observational data merit consider-ation. Large numbers of well-documented fatal events andprolonged follow up in the clinic setting may allow demonstrationof mortality differences not apparent in short-term prospectivetrials.

The appropriate interpretation of observational data necessitatesunderstanding of both the strengths and weaknesses of pharmaco-epidemiological studies. Clinical pharmacologists have a crucial

Fig. 2

Lives spared or lives lost per 1000 patients treated for 5 years. Data for angiotensin-converting enzyme (ACE) inhibitor-treated patients (

�

) andcalcium channel blocker-treated patients ( ) relative to patients who received neither of these drug classes. The horizontal line is the line of identity. Numbersin columns above the line represent estimated lives spared and numbers in columns below the line represent estimated excess deaths per 1000 patients treatedfor 5 years (e.g. for all-cause mortality, 35 lives are spared if patients are prescribed ACE inhibitors and 45 lives are lost if patients given CCB; switchingfrom CCB to ACE inhibitor is predicted to spare 80 lives). Data is based on 857 deaths (564 vascular and 349 coronary artery disease) over 34 519 patientyears. CHD, coronary heart disease; ACEI, ACE inhibitor.

Clinical trials and tribulations

955

role in this evaluation. Such considerations emphasize theimportance of pharmacology across the range of drug evaluation.

ACKNOWLEDGEMENTS

The

observational

analyses

described

in

this

paper

weresupported

by

an

unconditional

grant

from

Merck.

The

author

isindebted to his coinvestigators, Professor David Hole, ProfessorTony

Lever,

Dr

Peter

Meredith,

Dr

Ross

Morton,

Dr

LilianMurray and Professor John Reid (Departments of Medicine andTherapeutics,

and

Public

Health,

University

of

Glasgow,Glasgow, UK).

REFERENCES

1. Collins R, MacMahon S. Blood pressure, antihypertensive drugtreatment and the risks of stroke and of coronary heart disease.

Br.Med. Bull.

1994;

50

: 272–98.2. Yusuf S, Peto R, Lewis J

et al. Beta blockade during and aftermyocardial infarction: An overview of the randomised trials. Prog.Cardiovasc. Dis. 1985; 27: 335–71.

3. Hansson L, Lindholm LH, Ekbom T et al. Randomised trial of old andnew antihypertensive drugs in elderly patients: Cardiovascularmortality and morbidity the Swedish Trial in Old Patients withHypertension-2 Study. Lancet 1999; 354: 1751–6.

4. Hansson L, Lindholm LH, Niskanen L et al. Effect of angiotensin-converting-enzyme inhibition compared with conventional therapy oncardiovascular morbidity and mortality in hypertension: The CaptoprilPrevention Project (CAPPP) randomized trial. Lancet 1999; 353:611–16.

5. HOPE (Heart Outcomes Prevention Evaluation) Study Investigators.Effects of an angiotensin-converting-enzyme inhibitor, ramipril, oncardiovascular events in high-risk patients. N. Engl. J. Med. 2000; 342:145–53.

6. Brown M, Palmer C, Castaigne A et al. Morbidity and mortality inpatients randomised to double-blind treatment with a long-actingcalcium channel blocker or diuretic in the International Nifedipine

GITS study: Intervention as a Goal in Hypertension Treatment(INSIGHT). Lancet 2000; 356: 366–77.

7. Hansson L, Helmer T, Lund-Johannsen P et al. Randomised trial ofefforts of calcium antagonists compared with diuretics and β-blockerson cardiovascular morbidity and mortality in hypertension: The NordicDiltiazem (NORDIL) Study. Lancet 2000; 356: 359–65.

8. The ALLHAT Officers and Coordinators for the ALLHAT Collabor-ative Research Group. Major cardiovascular events on hypertensivepatients randomized to doxazosin vs chlorthalidone. The Antihyper-tensive and Lipid-Lowering Treatment to Prevent Heart Attack Trial(ALLHAT). JAMA 2000; 283: 1967–75.

9. Hansson L, Zanchetti A, Carruthers SG et al. Effects of intensiveblood pressure lowering and low-dose aspirin in patients with hyper-tension: Principal results of the Hypertension Optimal Treatment(HOT) randomised trial. Lancet 1998; 351: 1755–62.

10. Tuomilehto J, Rastenyte D, Birkenhager WH et al. Effects of calciumchannel blockers in older patients with diabetes and systolic hyper-tension. N. Engl. J. Med. 1999; 340: 677–84.

11. Benson K, Hartz AJ. A comparison of observational studies andrandomized controlled trials. N. Engl. J. Med. 2000; 342: 1878–86.

12. Concato J, Shah N, Horwitz RI. Randomized, controlled trials,observational studies, and the hierarchy of research designs. N. Engl.J. Med. 2000; 342: 1887–92.

13. Pahor M, Guralnik JM, Ferucci L et al. Calcium-channel blockade andincidence of cancer in aged populations. Lancet 1996; 348: 493–7.

14. Pahor M, Guralnik JM, Salive ME, Corti M-C, Cartbonin P, Havlik RJ.Do calcium channel blockers increase the risk of cancer? Am. J.Hypertens. 1996; 9: 695–9.

15. Lever AF, Hole DJ, Gillis CR et al. Do inhibitors of angiotensin-I-converting enzyme protect against risk of cancer? Lancet 1998; 352:179–84.

16. Cox DR. Regression models and life tables. J. R. Stat. Soc. B 1972;34: 187–220.

17. Psaty BM, Heckbert SR, Koepsell TD et al. The risk of myocardialinfarction associated with anti-hypertensive drug therapies. JAMA1995; 274: 620–5.

18. Lawson DH, Sherman V, Hollowell J for the Scientific and EthicalAdvisory Group. The General Practice Research Database. Q. J. Med.1998; 91: 445–52.