Effect of Antihypertensive Treatment in Patients Having Already Suffered From Stroke
Gathering the Evidence
Background and Purpose Drug treatment of high blood pressure has been shown to reduce the associated cardiovascular risk. Stroke represents the type of event more strongly linked with high blood pressure, responsible for a high rate of death or invalidity, and with the highest proportion of events that can be avoided by treatment. Hypertensive patients with a history of cerebrovascular accident are at particularly high risk of recurrence. Specific trials of blood pressure lowering drugs in stroke survivors showed inconclusive results in the past.
Methods We performed a meta-analysis using all available randomized controlled clinical trials assessing the effect of blood pressure lowering drugs on clinical outcomes (recurrence of stroke, coronary events, cause-specific, and overall mortality) in patients with prior stroke or transient ischemic attack.
Results We identified 9 trials, including a total of 6752 patients: 2 trials included 551 hypertensive stroke survivors; 6 trials of hypertensive patients included a small proportion of stroke survivors (536 patients); 1 trial included stroke survivors, whether hypertensive or not (5665 patients). The recurrence of stroke, fatal and nonfatal, was significantly reduced in active groups compared with control groups consistently across the different sources of data (relative risk of 0.72, 95% confidence interval: 0.61 to 0.85). There was no evidence that this intervention induced serious adverse effect.
Conclusions Blood pressure lowering drug interventions reduced the risk of stroke recurrence in stroke survivors. Available data did not allow to verify whether such benefit depends on initial blood pressure level. More data are needed before considering antihypertensive therapy in normotensive patients at high cerebrovascular risk.
For hypertensive individuals, antihypertensive drug treatment has been shown effective and safe in general, from the results of randomized clinical trials and their overviews.1,2 The most apparent benefit of this treatment is the relative risk reduction of stroke incidence, estimated between 30 and 50%. Some uncertainties remain concerning the effect of treatment in some specifically defined subgroups of patients and make the definition of the treatment target population3 not as accurate as desired. The present paper focuses on the particular subgroup of hypertensive patients who have already suffered from stroke or transient ischemic attacks and are consequently at a high risk for future stroke. Two randomized clinical trials (the Carter trial4 and the Hypertension-Stroke Cooperative Study Group (HSCSG) trial5), whose results were published more than 20 years ago, were specifically designed to address this issue. In a recent overview6 the results of these 2 trials were considered separately from other antihypertensive drug trials: “ … the results were heterogeneous for recurrent stroke, which is the most expected outcome in this case. Although no result was significant, the data did not reject the hypothesis of a beneficial effect on stroke recurrence and congestive heart failure incidence under treatment… For obvious ethical reasons, it is probable that such trials will not be repeated in the future, and our knowledge in this domain will remain incomplete.” Because of publication delays, a major relevant trial was omitted from this assessment: the report of the Post-stroke Antihypertensive Treatment Study7 (PATS), a large-scale study of antihypertensive drug treatment inpatients having already suffered from stroke or transient ischemic attack and included irrespective of their blood pressure level, was published a few months previously.
In addition, some of the clinical trials of antihypertensive drug treatment performed in less-specific populations have included a small number of patients who had already suffered from stroke or transient ischemic attack. Data concerning these patients specifically are available through the INDANA (INdividual Data ANalysis of Antihypertensive drug interventions) database.8
Our aim here is to synthesize and weigh the different sources of evidence that are available today, to assess whether blood pressure lowering drugs are effective in hypertensive people who have already suffered a cerebrovascular event.
Subjects and Methods
The trials have been identified through electronic medical database searches, survey of specialized and general medical journals, survey of specialized congress proceedings, reading of overviews, and direct contacts with investigators involved in such trials.* In agreement with the intention-to-treat principle, we have included all data available from randomized controlled trials that included survivors of stroke or transient ischemic attack and that have assessed the effect of antihypertensive drug treatment in terms of morbidity and mortality.
The outcomes analyzed here are those adopted by the Steering Committee of the INDANA project.8 The numbers considered in each outcome are the numbers of patients who suffered: (1) fatal strokes; (2) fatal and nonfatal strokes, excluding transient ischemic attacks; (3) fatal coronary events (including sudden death, defined as death occurring within 24 hours after any symptom); (4) fatal and nonfatal main coronary events (using patients’ history as the criterion for main coronary heart disease in the Hypertension Detection and Follow-up Program [HDFP]1); (5) cardiovascular mortality (including death from pulmonary thrombo-embolism); (6) main cardiovascular events (combining 2, 4, and 5, excluding nonmajor cardiovascular events such as angina pectoris, intermittent claudication, or nonfatal congestive heart failure); and (7) total mortality. For each outcome, patients are taken into account only once, whatever the number of events they experienced.
The outcome of primary interest in this analysis is stroke recurrence, both because the patients are at particularly high risk (see below) and because this outcome has high salience for such patients.
To document the prognostic significance of a prior cerebrovascular event, the increased risk attributable to a prior stroke or transient ischemic attack was explored with the INDANA database. A multiple logistic regression model was used, with stroke recurrence as the dependent variable. The independent covariates included in the model were trial, blood pressure, smoking status, age, sex, history of myocardial infarction, and history of stroke. Serum total cholesterol was not associated with the risk of stroke.
The relationship between risks in the control and treated groups was examined to check which kind of model (multiplicative or additive) was the more appropriate. If the multiplicative model had a better fit, we planned to express the treatment effect as relative risk, which has a direct meaning, the odds ratio being a poor estimate of the relative risk at the level of risk we consider. The combined relative risk has been estimated from summarized data as in a classic meta-analysis, through the EasyMA software.9 To take into account the differences in length of follow-up between trials, the number of patient-years by trial and by group was used as the denominator, rather than the number of patients. This number was computed exactly for the trials included in the INDANA database. It is also precisely given in the publication for the PATS trial.7 For the Carter4 and the Hypertension-Stroke Cooperative Study Group5 trials, it has been imputed from the mean follow-up and the number of patients in each group.
The small number of stroke patients in some subgroups of INDANA trials (namely the Coope trial, the MRFIT, and the STOP) led us to pool their data, to avoid the dispersion of treatment effect estimate, and address the issue of no outcomes in the treatment groups of some trials.
We have distinguished 3 groups according to the type of sources of data in this meta-analysis. The Carter and HSCSG trials were pooled in the “previous meta-analysis” group, the INDANA subgroup data have been pooled in the second group, and the PATS data constitute the third group. Therefore, results for each category are presented separately, and combined if there was no heterogeneity among groups.
To assess a possible difference in treatment effect according to a prior cerebrovascular event, an interaction term has been tested with the available individual data, ie, with the INDANA database.
The robustness of the results has been tested by estimating the number of additional trials of the same size and the same risk level as the largest included, but with no treatment effect, that would be needed to give a nonsignificant overall result.
The HSCSG trial5 included hypertensive patients from 10 US centers (systolic blood pressure between 140 and 220 mm Hg, and diastolic blood pressure between 90 and 115 mm Hg) below 80 years of age, with a history of stroke or transient ischemic attack in the previous year. Four percent of the patients presented with a transient ischemic attack history only; 59% were men. The active treatment was a fixed combination of deserpidine and methyclothiazide; the control group received a placebo in a double-blind fashion. The Carter trial included hypertensive patients (systolic blood pressure of 160 mm Hg or higher and diastolic blood pressure of 110 mm Hg or more) below 75 years of age, with previous ischemic stroke; 57% were men. The active treatment was titrated to a blood pressure control target; the control group was untreated. In these 2 trials, 551 patients in total were included, representing 1388 patient-years, ie, 10% of the data analyzed here.
Among the trials for which data are available through the INDANA database, 6 had included patients with a history of stroke or transient ischemic attack, in Europe or the United States. These trials are the European Working Party on High Blood Pressure in the Elderly Trial (EWPHE),10 the Coope trial,11 the HDFP,12 the Multiple Risk Factor Intervention Trial (MRFIT),13 the Systolic Hypertension in the Elderly Program (SHEP),14 and the Swedish Trial in Old Patients with Hypertension (STOP-Hypertension).15 Other trials from INDANA had excluded such patients. One of these trials (MRFIT13) assessed a multifactorial intervention on cardiovascular risk factor, and we decided to include only the data of the nonsmoker subgroup. The rationale for this decision is given elsewhere.8 The proportion of patients included in these trials with a history of stroke lies between 0.2 and 5% (Table 1⇓). This represents a population of 536 patients, and 2158 patients-years, ie, 15% of the data analyzed here. The proportion of men was 49%, the average age was 65 years, the average blood pressure at entry was 174/97 mm Hg. For the trials in which the proportion of transient ischemic attack is known, it was around 15 to 20%. All these trials assessed an intervention based on stepwise antihypertensive drug treatment, compared with placebo,10,14,15 no intervention,11 or usual care.12,13
The PATS trial included 5665 Chinese patients, having 10 397 patients-years of follow-up, ie, 74% of the data analyzed here, with 72% men, and 60 years of age on average. The proportion of patients included after a transient ischemic attack was 11.8%. There was no inclusion criterion for blood pressure; the average blood pressure at baseline was of 154/93 mm Hg.
The prognostic significance of a prior stroke or transient ischemic attack has been explored among the 5 trials from the INDANA database for which information is available concerning stroke in the follow-up. The adjusted estimated odds ratio for recurrent stroke was 2.72 (95% confidence interval [CI]: 1.86 to 3.98), which represents the same increase in risk of subsequent stroke as an increase of 58 mm Hg of systolic blood pressure, or an increase of 21 years of age.
Plotting for each trial the risk of the control group against the risk of the treated group (data not shown) suggested that a multiplicative model describes the treatment effect better than an additive one. Therefore, relative risk has been adopted as the treatment effect index.
For the main outcome, ie, all strokes, the data from the 2 earlier trials combined with those from INDANA suggested that treatment reduced the risk of stroke in relative terms by 29% (95% CI, 5 to 47%; P=.03) (Table 2⇓). Estimates limited to 1 of these 2 groups did not separately reach a statistically significant level. The point estimate was very close to the corresponding result from the PATS trial (Fig 1⇓) (which alone was highly statistically significant, P=.0013), and the overall combined risk reduction was 28% (95% CI, 15 to 39%; P=.00009). The interaction between treatment effect on stroke and a prior history of stroke or transient ischemic attack, explored in the INDANA data, was not statistically significant (Table 3⇓), despite a difference that could be clinically meaningful. This finding could be due either to a no actual difference or to a lack of power. For stroke death, only the combination of all available data reached a statistically significant level with a relative risk reduction of 29% (95% CI, 6 to 47%; P=.019). There was no evidence of heterogeneity of results among trials.
For the other outcomes, there was no evidence of heterogeneity among trial results. The trend toward benefit was observed for all other outcomes except main coronary events and was statistically significant for combined cardiovascular events, total or fatal only.
We summarized all the available evidence from randomized controlled trials on the effect of blood pressure lowering drugs in survivors of stroke or transient ischemic attack. The results show a significant benefit on stroke recurrence, with approximately 30% prevented by treatment. It is noteworthy that stroke history is, along with age, the most powerful risk factor for stroke in hypertensive patients.
This work has several limitations.
Limitation 1: This is a meta-analysis partially based on trial subgroups, not planned a priori in the conduct of the trials.
Limitation 2: The inclusion criteria comprise transient ischemic attacks as well as strokes. The latter are quite difficult to accurately diagnose in daily practice, and their definition may vary greatly from one trial to another. However, transient ischemic attacks represent a small proportion of patients, and their inclusion is unlikely to alter markedly the treatment estimate. A change in treatment effect estimate would require that this effect be different between patients with transient ischemic attack and patients with definite stroke. Such a hypothesis deserves to be explored, which was not possible with the available data. However, it could be done in further meta-analyses using individual patient data when results from ongoing trials are available.
Limitation 3: From the patient’s point of view, it would have been more relevant to focus on severe stroke, ie, either fatal or disabling, as the main outcome, because a proportion of strokes are without any significant consequence for the quality of life of patients. It is not possible to obtain such results from the currently available data. Ongoing and future trials should take this into account.
Limitation 4: This meta-analysis does not rely on a system that ensures all existing information has been identified. Such a system is not yet available, but may be in the future, thanks to the efforts of organizations such as the Cochrane Collaboration,16 and to a change in policy of the registration agencies, which may ensure public access to trial registries once specific drugs are marketed.
Limitation 5: The data analyzed are from very heterogeneous sources. This heterogeneity concerns the type of treatment administered, as well as patient selection procedures or other factors that might influence the treatment effect. We have tried to take this fact into account in presenting the results according to their source, in addition to the combined overall results. The fact that all sources remarkably agree concerning the main outcome may be partially due to chance.
Limitation 6: The PATS trial represents three-fourths of the data. The overall results are then essentially representative of the PATS results.
The PATS trial deserves specific attention. This is the first powerful trial of antihypertensive drug treatment in patients having suffered from stroke. The initial report does not suggest any important methodological flaw, and the results strongly indicate that the assessed intervention was effective in reducing the incidence of stroke recurrence. The fact that patients were included irrespective of their blood pressure does not constitute per se a defect. It is important to check that initial blood pressure does not influence the treatment effect, but we could logically expect that this effect is at least as important in hypertensive as in normotensive patients. If this were the case, the overall effect could be diluted by this inclusion procedure, compared with including hypertensive patients only.
If we omit the data from the PATS trial, the results suggest that antihypertensive drug treatment reduces stroke recurrence incidence, with a weak level of evidence because of a lack of power and the post hoc nature of the analysis. However, the PATS trial results are impressively in agreement with such an effect, especially regarding the size of the estimate of the treatment effect. With regard to the overall results on stroke, the probability that antihypertensive drugs have no effect is computed at less than 1/5000. The analysis of the robustness of the results shows that the level of evidence is very strong. We estimated that 6 trials of the same size and the same risk level as the PATS trial, but with no treatment effect, would be needed to give an overall result for stroke recurrence as not significant.
Before such results were available, treating hypertension in survivors of stroke was debatable. Indeed, stroke usually means that brain arteries are damaged, and high blood pressure may be needed to ensure sufficient blood supply through such arteries. According to this pathophysiological model, a decrease in blood pressure may be harmful. The alternative is to consider that high blood pressure is still a causal risk factor in stroke survivors; thus blood pressure lowering drugs should provide benefit by decreasing the associated risk. Obviously, our results support the latter therapeutic model.
It seems sensible to examine these results in the context of what we know about hypertensive people in general, most of whom have not suffered a stroke. The estimate from all trials together is more beneficial than that we obtained here, our point estimate being outside the 95% CI given by Collins et al1 (Table 4⇓). However, it has been observed that when trial results were examined separately according to age (older being defined with a cut-off around 65 years), the treatment effect in relative terms appears (nonsignificantly) lower in older than in younger patients for stroke, the opposite pattern being observed with a statistical significance for coronary events (data not shown).17 The risk reduction we observed is closer to that previously estimated in older hypertensive patients. However, the age of the population we considered is on average close to 65 years, and therefore includes both “younger” and “older” patients.
Other reasons may explain a lower estimate of relative reduction attributable to treatment, such as the quality of blood pressure management, or the type of drugs used. It could also be due to a partly deleterious treatment effect (following the first pathophysiological model mentioned above), which may counterbalance the beneficial effect such as the J-curve phenomenon suggested for coronary events.18 Whether previous history of stroke interacts with the blood pressure lowering drug effect should be explored on individual patient data. Such an interaction does not appear significant in the INDANA database, but the number of patients with such a history is small.
The analysis of data from trials in hypertensive patients suggests that relative risk reduction attributable to treatment is the same whatever the pretreatment blood pressure1. Added to this finding, the PATS trial results raise a very important issue: is antihypertensive treatment effective in normotensive people? There is no suggestion here for treating everyone with normal blood pressure, but rather to target the treatment toward patients with high cardiovascular risk, especially high stroke risk, irrespective of the blood pressure level. This hypothesis should be validated before being applied in daily medical practice. The ongoing PROGRESS trial,19 which assesses the effect of perindopril plus indapamide versus placebo in either normotensive or controlled hypertensive stroke survivors, should contribute to answering this important question.
Available evidence on the effect of antihypertensive drug treatment in hypertensive stroke patients shows a substantial reduction in the risk of stroke recurrence, without significant adverse effect. This evidence is in line with the known treatment effect in hypertension. Trials must now assess the effect of blood pressure lowering drugs in normotensive patients at high risk of stroke.
The constitution of the INDANA database was possible thanks to grants from the “Association pour la Promotion de la Recherche et de l’Evaluation en Thérapeutique,” the “Société Française d’Hypertension Artérielle,” the “Fondation pour la Recherche Médicale,” and the “Hospices Civils de Lyon.” This paper corresponds to the work granted by the French Ministery of Foreign Affairs and the “Association Cardiologique du Rhône.”
- Received March 18, 1997.
- Revision received August 18, 1997.
- Accepted September 19, 1997.
- Copyright © 1997 by American Heart Association
Carter AB. Hypotensive therapy in stroke survivors. Lancet. 1970;i:485–489.
Gueyffier F, Boutitie F, Boissel JP, Coope J, Cutler J, Ekbom T, Fagard R, Friedman L, Perry HM, Pocock S, Prineas R, Schron E. INDANA: a meta-analysis on individual patient data in hypertension. Protocol and preliminary results. Thérapie. 1995;50:353–362.
Easy MA. Reference Manual. Version 1996b. Department of Clinical Pharmacology, Lyon, 1996.
Amery A, Birkenhäger W, Brixko P, Bulpitt C, Clément D, Deruyttere M, De Schaepdryver A, Dollery C, Fagard R, Forette F, Forte J, Hamdy R, Henry JF, Joossens JV, Leonetti G, Lund-Johansen P, O’Malley K, Petrie J, Strasser T, Tuomilheto J, Williams B. Mortality and morbidity results from the European Working Party on High Blood Pressure in the Elderly Trial. Lancet. 1985;i:1349–1354.
Coope J, Warrender TS. Randomised trial of treatment of hypertension in elderly patients in primary care. BMJ. 1986;293:1145–1151.
Cochrane Collaboration (Sackett DL, Oxman AD, eds). Cochrane Collaboration Handbook. Oxford: The Cochrane Collaboration, 1995.
Lièvre M, Leizorovicz A. Treatment of high blood pressure in patients aged over 60 years: lessons from randomized clinical trials. Cardiology in the elderly. 1995;3:217–222.
PROGRESS Management Committee. Blood pressure lowering for the secondary prevention of stroke: rationale and design for PROGRESS. J Hypertens. 1996;14:S41–S46.