Control of blood pressure and risk of first acute myocardial infarction: Skaraborg hypertension projectBMJ 1994; 308 doi: https://doi.org/10.1136/bmj.308.6930.681 (Published 12 March 1994) Cite this as: BMJ 1994;308:681
- U Lindblad, Hentorp Health Care Centre, Grona vagen 40A, S-541 51 Skovde, Sweden.a,
- L Ramstam,
- L Ryden,
- J Ranstam,
- S-O Isacsson,
- G Berglund
- a Department of Community Health Sciences, Lund University, Malmo, Sweden
- Department of Medicine, Lund University, Malmo, Sweden
- Department of Cardiology, Karolinska Hospital, Stockholm, Sweden
- Correspondence to: Dr
- Accepted 8 November 1993
Objective: To analyse the relation between treated blood pressure and concomitant risk factor and morbidity from acute myocardial infarction.
Design: Prospective longitudinal study. Treated blood pressures and other variables were used to predict acute myocardial infarction.
Setting: Primary health care in Skaraborg, Sweden.
Subjects: 1121 men and 1453 women aged 40-69 years at registration at outpatient clinics, 1977-81, with no evidence of previous myocardial infarction were followed up for an average of 7.4 years. Subjects were undergoing treatment with drugs to lower blood pressure or had blood pressure that exceeded the systolic or diastolic limits, or both, for diagnosis (>170/>105 mm Hg (patients aged 40-60 years) and >180/>110 mm Hg (older than 60 years)) on three different occasions, or both.
Main outcome measures: First validated event of fatal or non-fatal acute myocardial infarction. Results - In men but not in women there was a negative relation between treated diastolic blood pressure and risk of acute myocardial infarction. Left ventricular hypertrophy and smoking were contributory risk factors in both sexes, as was serum cholesterol concentration in men. In men with normal electrocardiograms (n=345) risk increased with increasing diastolic blood pressure (P=0.047), whereas the opposite was found in men with electrocardiograms suggesting ischaemia or hypertrophy, or both (n=499, P=0.009). In those with a reading of 95-99 mm Hg the relative risk was 0.30 (P=0.034); at >=100 mm Hg it was 0.37 (P=0.027). No similar relations were seen in women or for systolic blood pressure.
Conclusion: It may be hazardous to lower diastolic20blood pressure below 95 mm Hg in hypertensive men with possible ischaemia or hypertrophy, or both. Electrocardiographic findings should be considered when treatment goals are decided for men with hypertension.
As the gradient of diastolic blood pressure is a major determinant of coronary perfusion there may be a point beyond which reduction in diastolic blood pressure is harmful
In this study hypertensive men with ischaemic/hypertrophic electrocardiograms had increased risk for a first myocardial infarction the lower the treated diastolic blood pressure
The opposite was true in men with normal electrocardiograms
No similar relations were found in women or for systolic blood pressure
Important contributing risks for acute myocardial infarction came from smoking and left ventricular hypertrophy in both sexes and from the serum cholesterol concentration in men
The risk for acute myocardial infarction among people with hypertension is related to blood pressure, other concomitant risk factors, and the extent of established arteriosclerotic disease.*RF 1-9* The prognosis is generally regarded to be influenced by the effectiveness of control of blood pressure combined with the impact on other risk factors.1011 Though the risk of myocardial infarction will decrease the lower the blood pressure, there may be a point beyond which reduction is dangerous (the J curve phenomenon).12 Further knowledge about this is of utmost importance for how individual treatment goals are set in clinical practice.
Results from clinical trials cannot be directly extrapolated to patients in routine care because of selected bases for recruitment of participants. Therefore it is important to know the relative impact of various factors on the risk of complications in unselected groups of patients. Such information can be used for the development of practice guidelines for how to treat hypertension.
Our aim was to analyse risk factors for myocardial infarction, including blood pressure during ongoing treatment, that may influence the prognosis in a large cohort of men and women with hypertension treated in primary health care.
Subjects and methods Skaraborg hypertension project
The Skaraborg hypertension project implemented guidelines for the detection, management, treatment, and follow up of hypertensive patients aged 40-69 years at particular nurse based outpatient clinics in the primary health care of six municipalities in Skaraborg county.13 It targeted all subjects with hypertension among the 46 234 (at 1 January 1977) residents who were aged 40-69 years.
A patient was eligible for inclusion and given the diagnosis of hypertension if he or she was being treated for hypertension or had a blood pressure reading that fulfilled certain diagnostic criteria, or both. For people aged 40-60 the limits used for diagnosis were >170/>105 mm Hg and in older subjects >180/>110 mm Hg. Systolic or diastolic limits, or both, had to be exceeded on three different occasions. Treatment aimed at normalising blood pressure to <160/<95 (40-60 years) and <170/<105 (>60 years).
Patients were recruited in two ways. Firstly, general practitioners and specialists referred patients who were already being treated. Secondly, during a two year period the blood pressure was measured routinely in all primary care patients aged 40-69 years who attended any outpatient clinic, irrespective of the reason for visit.
The diagnostic protocol at baseline included a structured medical history obtained by specially trained nurses according to predefined criteria for acute myocardial infarction, angina pectoris, stroke, intermittent claudication, heart failure, diabetes mellitus, and smoking habits.1415 A 12 lead electrocardiogram was recorded and classified according to the Minnesota code.15 Laboratory tests included measurement of total serum cholesterol concentration with a direct enzymatic method,16 measurement of serum creatinine concentration, and a test of albuminuria. Supine (5 minutes' rest) systolic and diastolic (phase V) right brachial arterial pressure were recorded to the nearest 2 mm Hg by specially trained nurses. Height (to the the nearest 1 cm) and weight (to the nearest 0.1 kg) were also recorded. A detailed description of the method has been reported elsewhere.17 Except for electrocardiograms, the baseline procedure was repeated once a year up to 1981.
Drug prescription followed a stepped care schedule17 starting with a β blocker (initially propranolol and from 1979 metoprolol) or a thiazide diuretic (bendrofluazide). If needed, treatment was continued by the combination of the two. Hydralazine could be added as a third step.
During the five years up to December 1981, 3240 patients (1428 men and 1812 women) were included. They are 7% of the targeted population. Most (2367) were referred to the project with ongoing treatment, and hypertension was detected in 873 previously untreated patients. Of the initial recruits, 1196 men and 1495 women completed their first annual check up before January 1982, when formal follow up of the cohort ended. The present study includes a follow up of morbidity among those patients from their first annual check up to December 1987.
Table I gives the characteristics of the study population. At the start of the follow up 825 men (69%) and 894 women (60%) were taking a β blocker alone or in combination. Corresponding figures for treatment with a thiazide diuretic were 638 (53%) and 929 (62%), respectively.
The systolic and diastolic blood pressures recorded at the first annual check up were used as independent variables in univariate and multivariate analyses. They were grouped into five categories of systolic (<150, 150-159, 160-169, 170-179, >=180 mm Hg) and diastolic blood pressure (<85, 85-89, 90-94, 95-99, >=100 mm Hg) but were also used as continuous variables. Categories were chosen based on measurements often referred to in clinical practice. Besides the linear variables, a quadratic term was added to investigate the existence of a non-linear relation between achieved diastolic blood pressure and risk. The concentrations of serum cholesterol and body mass index achieved on this occasion were also used for the analyses. Patients were categorised as current smokers or non-smokers according to the baseline interview. Prevalent diabetes mellitus was based on interviews at baseline and at the first annual check up, while left ventricular hypertrophy was determined from the coded electrocardiograms at baseline.
The Minnesota coded electrocardiograms were categorised as follows: a Q wave suggesting myocardial infarction - 1.1 or 1.2; left ventricular hypertrophy - 3.1 or 3.3 and 4.1-4.3 and/or 3.1 and 3.3 and 5.1-5.3; ST depression - 4.1-4.3; negative T waves - 5.1-5.3; high voltage - 3.1 or 3.3. Patients with a Minnesota code of zero in all positions were defined as having a normal electrocardiogram. Those who had either ST depression or negative T wave or high voltage or any combination of the three were classified as having ischaemic-hypertrophic electrocardiograms.
Serum creatinine concentrations and albuminuria were determined at the start of follow up. The absence of angina pectoris, heart failure, intermittent claudication, and previous stroke was based on combined information from the recordings at baseline and at the first annual check up. Body mass index was calculated as weight (kg)/(height (m))2.
The end point of this study was the first acute myocardial infarction during follow up; fatal and non-fatal cases were analysed together.18 Two sources of information were used: the computerised inpatient register covering all three hospitals in the county and the national mortality register. All registered files that included International Classification of Diseases codes 410-411 were extracted from the hospitals. By using predefined diagnostic criteria,18 the inpatient diagnoses were validated against their hospital records and all events not fulfilling the diagnostic criteria were excluded. For fatal cases the underlying cause of death according to the national mortality register was checked and corrected according to hospital records when available. After the validation, an end point register was created for each subject including data on any first event of a definite acute myocardial infarction during the complete follow up.
In all analyses patients with any evidence of a previous myocardial infarction (medical history, a Q wave on the baseline electrocardiogram, or an event in the end point register during the year between baseline and first annual check up) were excluded.
All analyses were stratified by sex. Cumulative incidence (risk) was calculated with the Kaplan-Meier method.19 The multivariate analyses with Cox's proportional hazards model were based on blood pressure as a continuous and a categorised variable.20 Significance was in all analyses presumed when P<0.05. All tests were two sided.
Risk of acute myocardial infarction in relation to blood pressure
After the exclusion of subjects with previous myocardial infarction, 1121 (94%) men and 1453 (97%) women remained. Table II gives descriptive data according to category of blood pressure. There were only minor differences between the categories in duration of follow up, age, treatment, body mass index, serum cholesterol concentration, and percentage of smokers, diabetic patients, or left ventricular hypertrophy.
Table III shows the percentage risks adjusted for differences in the duration of follow up at seven years. There was an increasing risk of myocardial infarction with decreasing diastolic blood pressure for men. A similar trend was observed for increasing systolic blood pressure in women.
Table IV presents adjusted estimates of relative risk for the various categories of blood pressure. The lowest categories were used as reference. For two of the four higher categories of diastolic blood pressure in men, the 95% confidence interval did not include 1.0, indicating a significantly lower risk than in the reference group. No other relative risk estimate excluded 1.0.
Achieved diastolic blood pressure, entered as a continuous linear term in the Cox's proportional hazards model, did not reach significance, for men or women (table V). The same was true when a quadratic term was added. Instead, left ventricular hypertrophy was the strongest predictor for myocardial infarction followed by age in both men and women. Serum cholesterol concentration (men) and smoking (both sexes) were also significant risk factors. Substituting systolic blood pressure for diastolic changed the results only marginally.
Impact of electrocardiographic findings in men
The higher risk for men of myocardial infarction incurred by a low diastolic blood pressure was further explored by separating men by electrocardiographic findings at baseline. Two subgroups of patients were created, those with normal electrocardiograms (n=345) and those with ischaemic/hypertrophic electrocardiograms (n=499), and analysed separately by the model used for the total group of hypertensive patients. Table VI presents characteristics of the two groups. There were only marginal differences between the blood pressure strata.
The relative risks associated with each category of diastolic blood pressure did not differ significantly in men with normal electrocardiograms (table VIII). With diastolic blood pressure entered in the Cox's model as a continuous variable there was, however, a significant positive relation between diastolic blood pressure and risk (P=0.047). An added quadratic term did not reach significance.
Among the ischaemic/hypertrophic patients, those with a diastolic blood pressure >=95 mm Hg had a significantly lower adjusted relative risk of developing myocardial infarction than those belonging to the group with the lowest blood pressure (table VIII). There was also a significant negative relation with diastolic blood pressure used as a continuous variable (P=0.009). An added quadratic term did not reach significance.
No significant relation between achieved diastolic blood pressure and risk of acute myocardial infarction was seen in the 277 men who had neither normal nor ischaemic/hypertrophic electrocardiograms. Neither could any similar relation be shown for women or for achieved systolic blood pressure in men and women.
The negative relation between achieved diastolic blood pressure and the risk of myocardial infarction in men was seen in those with electrocardiographic changes indicating left ventricular hypertrophy or myocardial ischaemia, or both.21 These signs indicate a less favourable prognosis.1,22 The number of men with such electrocardiographic changes exceed the number who had a normal electrocardiogram. Most events occurred in the group with electrocardiographic abnormalities. This indicates that results of clinical trials, excluding subjects with evidence of overt ischaemic heart disease,4,6,23,24 cannot easily be applied to an unselected population of hypertensive patients.
Left ventricular hypertrophy, which is commonly caused by hypertension,25,26 increases oxygen demand in the myocardium. Narrowing of the coronary arteries caused by coronary atherosclerosis or hypertrophy (induced by hypertension) decreases coronary flow reserve.*RF 27-30* In the coronary circulation oxygen extraction is nearly maximal at rest. In the presence of left ventricular hypertrophy the lower end of the auto-regulation may be shifted to the right. Coronary flow occurs during diastole. If the diastolic blood pressure gradient drops below a critical value myocardial ischaemia may develop, increasing the risk of infarction.*RF 27-30*
Two early studies questioned the goal of the lower the better for the treatment of hypertension.*RF 31-32* Intense debate followed*RF 29,33- 36* when Cruickshank in 1987 showed a J shaped relation between diastolic blood pressure and risk of myocardial infarction in people being treated for hypertension who had established ischaemic heart disease.8 Other studies have confirmed his results among hypertensive subjects*RF 37-41* and in the general population.42 A further study has suggested that change in pressure may need to be considered besides the absolute level of diastolic blood pressure.43
In men with normal electrocardiograms at entry the relative risk of a myocardial infarction increased significantly with increasing diastolic blood pressure. This suggests that men in that subgroup would benefit from treatment to lower diastolic blood pressure, at least down to 85-90 mm Hg. Early detection and treatment of hypertension might prevent electrocardiographic changes and thus improve prognosis.
Other risk factors for myocardial infarction in men, besides the diastolic blood pressure and electrocardiographic abnormalities, were serum cholesterol concentration and smoking. In hypertensive men who already have signs of impaired coronary blood flow it might be more important to eliminate smoking and to lower blood cholesterol concentration than to induce a radical decrease of blood pressure.
Although serum cholesterol concentration was higher in women, it was not as strong a predictor for myocardial infarction as in men. Women were, however, just as vulnerable to smoking and left ventricular hypertrophy as men. Prevention of acute myocardial infarction in hypertensive women should perhaps concentrate on a reasonable reduction of blood pressure and stopping smoking.
Echocardiographically determined left ventricular hypertrophy has a substantially higher sensitivity*RF 44-45* than that electrocardiographically determined,44,46 and echocardiography is proposed to be included as a part of clinical management in some hypertensive subjects.30 The specificity of the electrocardiogram is high,44 however, and in the view of our results it may still be an important tool in the evaluation of hypertensive patients.
Blood pressure measurements at the first annual check up were used as a measure of treatment effect. Patients were continuously recruited during 1977-80. Thus, the individual number of check ups in each patient may vary from one to four. In order not to induce bias by the use of averaged blood pressures over the years we decided to base the study on data from the first annual check up. As a further justification there were only marginal improvements in control of blood pressure after the first year.47
Our results build on observational data. Thus it cannot be completely excluded that the results may in part be caused by selection bias. Since there were only marginal differences between different categories of blood pressure with respect to multiple risk and morbidity at entry, this is, however, very unlikely. The consistency with previous studies also strongly suggests that the results are true reflections of different achieved blood pressures.
The present study includes patients with known hypertension in a well defined catchment area that was carefully evaluated for the presence of high blood pressure. The material should be representative of a general population of hypertensive subjects. Strict criteria were applied for the diagnosis, follow up, and treatment, simplifying a scientific follow up and offering opportunities to draw more representative conclusions of the impact of blood pressure control than those derived from traditional clinical trials of antihypertensive treatment.
The prognosis for the hypertensive cohort has recently been compared with the population from which they were recruited. The risk of myocardial infarction48 and total mortality49 did not differ from that of the population, and the risk of stroke was lower than expected.49 Thus, the main impression is that the prognosis for this cohort was better than expected and on average the intervention should have been beneficial. The results of the present study are consistent with these findings. The relative risks between different segments of the distribution of treated blood pressure showed small differences. The distribution of blood pressure was concentrated around a mean for which risks are low. In other words, this should be regarded as a well treated population. Furthermore, the blood pressures for diagnosis and treatment of hypertension were higher than current recommendations.50
In conclusion, we suggest that a low achieved diastolic blood pressure in men with abnormal electrocardiographic findings may be hazardous. Future trials of antihypertensive agents should focus on including a wide variety of risks including left ventricular hypertrophy.
This study was supported by grants from the Swedish Heart and Lung Foundation, Lund University, National Corporation of Pharmacies' fund for research in social pharmacy and health economics, and the Skaraborg County Council.