Original Article | Published:

Exercise blood pressure and the risk of future hypertension

Journal of Human Hypertension volume 26, pages 691695 (2012) | Download Citation

Abstract

The aim of this prospective cohort study was to identify which blood pressure measurement during exercise is the best predictor of future hypertension. Further we aimed to create a risk chart to facilitate the evaluation of blood pressure reaction during exercise testing. A number (n=1047) of exercise tests by bicycle ergometry, performed in 1996 and 1997 were analysed. In 2007–2008, 606 patients without hypertension at the time of the exercise test were sent a questionnaire aimed to identify current hypertension. The response rate was 58% (n=352). During the 10–12 years between exercise test and questionnaire, 23% developed hypertension. The strongest predictors of future hypertension were systolic blood pressure (SBP) before exercise (odds ratios (OR) 1.63 (1.31–2.01) for 10 mm Hg difference) in combination with the increase of SBP over time during exercise testing (OR 1.12 (1.01–1.24) steeper increase for every 1 mm Hg min−1). A high SBP before exercise and a steep rise in SBP over time represented a higher risk of developing hypertension. A risk chart based on SBP before exercise, increase of SBP over time and body mass index was created. SBP before exercise, maximal SBP during exercise and SBP at 100 W were significant single predictors of future hypertension and the prediction by maximal SBP was improved by adjusting for time/power at which SBP max was reached during exercise testing. Recovery ratio (maximal SBP/SBP 4 min after exercise) was not predictive of future hypertension.

Introduction

The prevalence of hypertension is estimated to be one billion individuals world wide, and hypertension causes 7.1 million deaths per year.1 It is well known that a vast majority of patients with hypertension are not well treated according to guidelines or are unaware of their high blood pressure.2 During recent decades, several different risk charts for cardiovascular morbidity/mortality have been developed to facilitate identification of individuals at high cardiovascular risk.3, 4 In most risk charts, hypertension is used to assess cardiovascular risk.

One way to identify individuals at high risk of developing future hypertension5, 6, 7, 8, 9 and cardiovascular disease10, 11 is to evaluate the blood pressure reaction during exercise testing. Bicycle ergometry testing is commonly performed to diagnose ischemic heart disease but this test can also reveal the general condition of the cardiovascular system. There is, however, a lack of consensus regarding how to interpret and handle individuals with high blood pressure values during exercise testing. Systolic blood pressure (SBP) of 200,10, 11, 12 2107, 13 and 230 mm Hg14 have been considered to be sufficient cutoff values. Other investigators have chosen to divide the blood pressure response in quartiles6 and tertiles15 to predict the risk of future hypertension. Another way to find a suitable predictor for future hypertension and cardiovascular disease is to evaluate the blood pressure level at recovery.8, 16 The diversity in predictive values could be explained by differences in exercise test methods, the definition of the hypertensive reaction and study populations. Several investigations have been performed on healthy young or middle-aged males.9, 10

The aim of this study was to identify which blood pressure measurement during exercise testing best could predict future hypertension in a clinically relevant population without diagnosed hypertension. Further, we aimed to create a feasible risk chart regarding blood pressure reaction during exercise that could predict the risk of future hypertension in the clinical setting.

Patients and methods

The study population consisted of all (n=1047) patients who performed an exercise test at the Department of Clinical Physiology at Sahlgrenska University hospital/Ostra, Goteborg/Sweden between May 1996 and December 1997. The reasons for referral varied and both outpatient and hospital-admitted individuals were consecutively included in this study. The main reasons for referral were unexplained chest pain or investigation of suspected coronary heart disease. The first exercise test during the time period was used for each individual. Medical information on the referral and the exercise test protocol was used to identify previous hypertension, other cardiovascular diseases, current medication and information regarding smoking. We excluded 190 patients who had known hypertension or were taking any blood pressure lowering medication. Patients who had moved abroad or for whom it was impossible to find the referral paper were excluded (N=93). Further, 158 patients had died during follow-up. In 2007 and 2008, 606 patients received a letter with information regarding the study and a request to participate. An informed consent was signed and a questionnaire was presented. The patients were asked whether they had been diagnosed with hypertension since the exercise test. All medication at the time of follow-up was recorded. Further, the questionnaire contained questions regarding smoking and other cardiovascular diagnoses (CVD). The time of follow-up was 10–12 years. Individuals who answered that they had been diagnosed with hypertension and/or individuals treated with blood pressure lowering medication were considered to be hypertensive at follow-up.

Ethical considerations

The Regional Ethical Review Board in Goteborg/Sweden approved the study and it was performed in accordance with the World Medical Association declaration of Helsinki.

Exercise bicycle ergometry test

Blood pressure before exercise was measured manually in the supine position after 2–5 min rest just before initiation of the exercise test with a manometer and a standard cuff (BOSO, Jungingen, Germany). Information regarding medication, smoking habits, weight and height were documented. The test was performed as graded symptom-limited exercise testing with bicycle ergometry according to the Bruce protocol.17 The patient was instructed to pedal at a frequency of 60–80 revolutions min−1. The exercise test started at different workloads (20–50 W) and the increase in workload (10 or 20 W min−1) was applied depending on age, sex and medical history. SBP was measured manually by auscultatory detection of Korotkoff-sound every second minute during the exercise, immediately after exercise and 4 min after exercise in the supine position. Blood pressure registration was performed by the attending nurse at the ward. Diastolic blood pressure was not measured during exercise. The patients exercised until the development of symptoms necessitated termination of the test, or until the patient had made a normal test in relation to the expected performance according to fixed tables based on sex, age and weight.18

All data from the exercise test were stored in a database. From this database we extracted information regarding blood pressure measurements, heart rate, workload and time as well as body mass index (BMI) and smoking status.

Among the variables measured during exercise, the following were considered to be possible predictors of hypertension:

  • SBP before exercise(mm Hg).

  • SBP at 100 W (mm Hg).

  • Maximum value of SBP (SBP max) during exercise (mm Hg).

  • Time at which SBP max was attained for the first time (min).

  • Increase of workload until SBP max is reached for the first time (W).

  • Rate of increase of SBP per time (mm Hg min−1).

  • Rate of increase of SBP per workload (mm Hg W−1).

  • Recovery ratio: maximal SBP/SBP 4 min after exercise.

Statistical methods

Differences by incident hypertension were identified by χ2 and the t-test for categorical and continuous variables, respectively. We used multiple logistic regression to model the dependence of the probability for hypertension as a function of risk factors and covariates. For model comparison we restricted ourselves to the subset of observations with complete information on all variables (255 observations, 56 events). To compare different models not necessarily nested within each other, we used the Akaike information criterion, which is smallest for a model providing a good fit with a small number of parameters.

Goodness of fit was assessed using the Hosmer–Lemeshow test and all models included here had P-values >0.1. In addition, the area under the ROC curve was calculated.

All models were adjusted for age, sex, BMI and smoking. Smoking was categorized as ever (current or former) versus never. The best model was retested in the unrestricted data set (284 observations, 61 events). Variables were included in a stepwise manner by adding the most significant of the remaining variables; multicollinearity among predicting variables in a model was checked using the variance inflation factor in a corresponding linear regression model and the correlation among estimated parameters. None of the models reported showed variance inflation factor >1.5 for any of the dependent variables or a correlation coefficient >0.36.

Odds ratios (OR) and 95% confidence intervals were defined as follows:

  • For BMI with respect to 1 kg m−2 unit increase.

  • For all BP values (including differences) with respect to 10 mm Hg increase.

  • For power with respect to 10 W increase.

  • For time with respect to 1 min increase.

  • For rates with respect to 1 mm Hg min−1 or 1 mm Hg 10 W−1.

  • For recovery ratio with respect to 10% increase.

Following Manolio et al.7 we dichotomized maximal SBP during exercise as 210 mm Hg (men) and 190 mm Hg (women) versus SBP < 210 and < 190 mm Hg, respectively. All analysis was performed using the SAS system (v.9.2, SAS Institute Inc., Cary, NC, USA) and results were considered to be significant at the 5% level (two-sided test).

Results

General

Of the 606 patients who received a questionnaire 352 (58%) responded. Because of incomplete information in the questionnaire, 47 patients were excluded. At baseline, 11 patients had a resting blood pressure before exercise of more or equal to 180 mm Hg and were excluded because of suspected untreated hypertension. This rendered us with a study group of 294 patients. Nine of the included subjects were diagnosed with CVD (myocardial infarction>, angina pectoris or stroke) at the time of the exercise test. Of the evaluated patients 244 were outpatients and 35 were hospital admitted. A total of 15 patients had unknown referral status. Of 294 participants, 67 individuals (23%) reported hypertension at the 10–12 year follow-up in 2007–8. Table 1 shows the baseline properties of participants grouped by incident hypertension. Individuals who were hypertensive at follow-up had a higher BMI and were more often found to be smokers. There were no gender differences with respect to BMI or smoking status except that the female participants were on average 3.1±10.8 years (P=0.02) older than the men (data not shown).

Table 1: Characteristics of the study population (N=294)

SBP during exercise

Figure 1 shows the average value of SBP during exercise as a function of time by incident hypertension, separately for men and women. A similar plot is obtained when the average SBP is plotted versus workload (data not shown). Mean SBP was consistently higher at each point in time, or value of power, among the patients who developed hypertension compared with the patients who stayed normotensive. The dotted line indicates the limiting values of 210 (men) and 190 mm Hg (women) as proposed by Manolio et al.7 Subjects who developed hypertension at follow-up passed the limit (210 for men and 190 mmHg for women) after roughly 10 min of exercise. Subjects who were normotensive at follow-up also reached the limit but on average 2–3 min later and shortly before the end of exercise.

Figure 1
Figure 1

SBP during exercise (mean value ±1 s.e.) by incident hypertension as a function of time, separately for women and men. The horizontal line indicates the limiting values of 190 and 210 mm Hg for women and men, respectively.

Multiple logistic regression

The results of multiple logistic regression of incident hypertension on different explanatory variables are shown in Table 2. All models were adjusted for age, sex, BMI and smoking status. The results given in Table 2 were not affected by increase in workload or its starting value. SBP before exercise was the strongest single predictor of hypertension and model 1 had a smaller value for Akaike information criterion than model 2 and 3. SBP at 100 W was comparable to maximal SBP reached during exercise. However, when maximal SBP was adjusted for total power increase or time until maximal SBP was reached the prediction was better than for SBP before exercise alone. Individuals with high maximal SBP (men 210 and women 190 mm Hg) had 2.98 (confidence interval 1.61–5.51) times higher risk for developing hypertension than those below cutoff. When adjusted for total time until maximal SBP was reached, the OR for future hypertension increased to 3.83 (confidence interval 1.97–7.46) for participants with high maximal SBP during exercise compared with those in the lower category of SBP. The best model was obtained when SBP before exercise and rate of SBP increase over time were included in the same model (Akaike information criterion=263, area under the ROC curve=0.75, cf. Figure 2). The mutual correlation between parameters for SBP before exercise and rate of SBP increase over time was small (r=0.13). The rate of SBP increase over workload was a slightly weaker predictor of future hypertension than the rate of SBP increase over time. Similar results were obtained when the model was tested in the complete data file (data not shown). The recovery ratio as given by the quotient of maximal SBP during exercise over SBP 4 min after exercise was not associated with future hypertension.

Table 2: Risk of future hypertension in relation to possible predictors during exercise testing
Figure 2
Figure 2

The ROC curve for the best model to predict future hypertension, based on SBP before exercise and rate of blood pressure increase over time (model 6 in Table 2).

Predictive probability chart

To predict the absolute risk for hypertension (P) we refit the best model in the largest data set, including only the predictors, which were significantly associated with hypertension, logP/(1−P)=−10.3.+0.088 BMI (kg m−2)+0.044 SBP0 (mm Hg)+0.103 rate (mm Hg min−1).

Denoting the linear predictor by y, the absolute probability for 10-year incidence of hypertension is given by P=exp(y)/(1−exp(y)). Figure 3 shows the predicted probabilities for hypertension by selected categories of SBP before exercise, rate of SBP increase over time and BMI.

Figure 3
Figure 3

Predicted probability for 10-year incidence of hypertension (%). Chart based on SBP before exercise (SBP0), rate of blood pressure increase over time (mm Hg min−1) and BMI.

There was no gender difference in the effect of SBP before exercise and increase of BP on incident hypertension. When both BMI and SBP before exercise where included in the model, age was no longer predictive of incident hypertension. However, BMI (OR=1.09 (1.01–1.19)) was a significant risk factor for hypertension while smoking (OR =1.83 (0.94–3.59)) had a minor influence on the risk of future hypertension.

Discussion

In this study we found that a combination of explanatory variables measured during exercise testing gave the best prediction of future hypertension. The strongest predictor was the combination of blood pressure before exercise and the rate of blood pressure increase over time. SBP before exercise, at 100 W and maximal SBP alone were all predictive of future hypertension. However, maximal SBP adjusted for time or workload was an even better predictor of later hypertension. Further we constructed a risk chart based on our main findings that enables us to implement the results in the clinical risk stratification. To our knowledge this is a novel approach to evaluate blood pressure reaction during exercise.

Blood pressure level before exercise was the strongest single predictor. It is well established that individuals in the high normal state of blood pressure are at higher risk of future definite hypertension.19 Further Miyai et al.6 identified that blood pressure reaction during exercise could improve the prediction of future hypertension in male individuals in the high normal range. In a Norwegian study of healthy middle-aged men several single predictors for blood pressure level and blood pressure reaction during exercise testing were found. Beside strong correlation between baseline blood pressure, both at rest and during exercise, smoking was associated with a higher blood pressure during exercise at follow-up.20 Anticipation may also influence blood pressure before exercise. Everson et al.21 found a 3.8-increased risk of developing hypertension in men who had a strong anticipatory blood pressure rise before exercise testing. Blood pressure increase before exercise reflects an increase in cardiac output with no compensatory decrease in vascular resistance.22 However, because all individuals with high normal blood pressure or anticipatory blood pressure increase do not develop hypertension it is important to identify additional measurements during the exercise test that could improve the prediction of hypertension.

The best model contains blood pressure before exercise and rate of blood pressure rise versus time. In this case, the mutual correlation between parameters for blood pressure before cycling and rate of increase over time was small, suggesting independent pathways to hypertension for both parameters. A difference in blood pressure level of 10 mm Hg between two patients before the test with the same rate of increase in blood pressure resulted in a 62% relative risk increase to develop hypertension. Equivalently, if two subjects had the same blood pressure level before, the one with a 1 mm Hg min−1 steeper rise had a 15% higher risk of developing hypertension in the next 10 years. Interestingly, it has previously been shown that there is an association between sharp blood pressure rise during an exercise test and the risk for stroke and myocardial infarction.23, 24 Even though blood pressure before exercise is associated with future hypertension the prediction was significantly improved when considering the rate of blood pressure rise during exercise.

Maximal SBP was also associated with the development of hypertension. In the CARDIA study by Manolio et al., young adults with a maximal SBP during exercise equal to, or exceeding, 210 in men and 190 mm Hg in women were 1.7 times more likely to develop hypertension 5 years later.7 When we applied the same model in our study we found an even higher OR for developing hypertension, 2.98 (1.61–5.51), which could be explained by several factors; for example, older subjects and longer follow-up period in our study compared with the CARDIA study. This OR increased even further when adjusted for the time at which the blood pressure limits were reached (OR 3.83 (1.97–7.46)). An individual who reaches the maximal blood pressure later/on a higher workload is hence not at an equally high risk of developing hypertension as the person who reaches a maximal blood pressure earlier/on lower workloads.

Recovery ratio was not found to be an explaining variable in our study. This is in contrast to several other investigators who have found a relation between recovery blood pressure level and hypertension as well as myocardial infarction and cardiovascular death.8, 9, 16, 25, 26 We had several missing values for SBP 4 min after exercise that can make the comparison with other variables difficult.

In a clinical setting during exercise testing there is a need for an instrument that could quantify the risk of future hypertension with respect to the blood pressure reaction. The risk chart we constructed considers SBP before exercise testing, rate of blood pressure increase over time and body mass index. Age, gender and smoking did however not significantly contribute to the model. This may be because of the facts that the mean age in our study was over 50 years of age and that the study was rather small. Furthermore, there was no significant difference regarding age between the hypertensive and normotensive population.

Limitations

This study has some limitations. The diagnosis of hypertension at follow-up was based on answers from a questionnaire, which obviously represents some uncertainty. The response rate of the questionnaire was not 100%, which of course affects the quality of the results. However, the response rate is in accordance with epidemiological studies performed during the last decade.27 Time to follow-up varied from 10 to 12 years. The chart we created, as a risk assessment tool for estimating the risk of future hypertension, could be seen as an example of how to implement our results. It is in line with the SCORE-chart for overall cardiovascular risk.4 However, the SCORE chart is based upon 200 000 subjects. Our study is performed with a small population and the chart could be of clinical use when the results are confirmed in a larger population.

Summary and conclusion

The strongest relative risk for future hypertension was found when combining SBP before the exercise test with the increase in blood pressure in relation to time and workload. To make these results more comprehensible and useful in the clinical setting we created a risk chart that can estimate the overall risk for patients to develop hypertension later in life. In clinical practice numerous patients are referred for exercise testing for different reasons. Hopefully this model will help identify patients during the exercise test that are in need of follow-up with respect to blood pressure.

References

  1. 1.

    , , , , . Selected major risk factors and global and regional burden of disease. Lancet 2002; 360(9343): 1347–1360.

  2. 2.

    , , , , , et al. High remaining risk in poorly treated hypertension: the ‘rule of halves’ still exists. J Hypertens 2002; 20(10): 2081–2088.

  3. 3.

    , , , . Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. JAMA 2001; 286(2): 180–187.

  4. 4.

    , , , , , et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J 2003; 24(11): 987–1003.

  5. 5.

    , . Blood pressure response to exercise as a predictor of hypertension. Arch Intern Med 1986; 146(10): 2053–2055.

  6. 6.

    , , , , , . Exercise BP response in subjects with high-normal BP: exaggerated blood pressure response to exercise and risk of future hypertension in subjects with high-normal blood pressure. J Am Coll Cardiol 2000; 36(5): 1626–1631.

  7. 7.

    , , , , , . Exercise blood pressure response and 5-year risk of elevated blood pressure in a cohort of young adults: The CARDIA study. Am J Hypertens 1994; 7(3): 234–241.

  8. 8.

    , , , , , et al. Blood pressure response during treadmill testing as a risk factor for new-onset hypertension. The Framingham Heart Study. Circulation 1999; 99(14): 1831–1836.

  9. 9.

    , , , , , . Blood pressure response after two-step exercise as a powerful predictor of hypertension: the Osaka Health Survey. J Hypertens 2002; 20(8): 1507–1512.

  10. 10.

    , , , , , . Exercise blood pressure predicts cardiovascular death and myocardial infarction. Blood Press Monit 1997; 2(3): 147–153.

  11. 11.

    , , , , , . Supine and exercise systolic blood pressure predict cardiovascular death in middle-aged men. J Hypertens 2001; 19(8): 1343–1348.

  12. 12.

    , , , . Follow-up of normotensive men with exaggerated blood pressure response to exercise. Am Heart J 1983; 106(2): 316–320.

  13. 13.

    , , , . Is there a relationship between exercise systolic blood pressure response and left ventricular mass? The Framingham Heart Study. Ann Intern Med 1992; 116(3): 203–210.

  14. 14.

    , , , . Exaggerated pressure response to exercise in men at risk for systemic hypertension. Am J Cardiol 1990; 66(7): 731–736.

  15. 15.

    , , , . Correction of the hypertensive response in the treadmill testing by the work performance improves the prediction of hypertension by ambulatory blood pressure monitoring and incidence of cardiac abnormalities by echocardiography: results of an eight year follow-up study. Int J Cardiol 2010; 141(3): 243–249.

  16. 16.

    , , , , , . Systolic blood pressure during recovery from exercise and the risk of acute myocardial infarction in middle-aged men. Hypertension 2004; 44(6): 820–825.

  17. 17.

    , . Exercise stress testing in evaluation of patients with ischemic heart disease. Prog Cardiovasc Dis 1969; 11(5): 371–390.

  18. 18.

    , , , . Reference values for exercise tests with continuous increase in load. Clin Physiol 1985; 5(2): 161–172.

  19. 19.

    , , , , , et al. Hemodynamic patterns of age-related changes in blood pressure. The Framingham Heart Study. Circulation 1997; 96(1): 308–315.

  20. 20.

    , , , , , . Predictors of 7-year changes in exercise blood pressure: effects of smoking, physical fitness and pulmonary function. J Hypertens 1997; 15: 245–249.

  21. 21.

    , , , . Anticipatory blood pressure response to exercise predicts future high blood pressure in middle-aged men. Hypertension 1996; 27(5): 1059–1064.

  22. 22.

    , , , . Integrated mechanisms of cardiovascular response and control during exercise in the normal human. Prog Cardiovasc Dis 1976; 18(6): 421–444.

  23. 23.

    , , , , , . Systolic blood pressure response to exercise stress test and risk of stroke. Stroke 2001; 32(9): 2036–2041.

  24. 24.

    , , , , , . Systolic blood pressure response to exercise testing is related to the risk of acute myocardial infarction in middle-aged men. Eur J Cardiovasc Prev Rehabil 2006; 13(3): 421–428.

  25. 25.

    , , , , , . The prognostic value of post-exercise blood pressure reduction in patients with hypertensive response during exercise stress test. Int J Cardiol 2006; 111(3): 352–357.

  26. 26.

    , , , , , et al. Usefulness of paradoxical systolic blood pressure increase after exercise as a predictor of cardiovascular mortality. Am J Cardiol 2008; 102(5): 518–523.

  27. 27.

    , , , , . Perceived health in 50-year-old women and men and the correlation with risk factors, disease and symptoms. Gend Med 2011; 8(2): 139–149.

Download references

Author information

Affiliations

  1. Department of Emergency and Cardiovascular Medicine, Institute of Medicine, Sahlgrenska University Hospital/Ostra, Sahlgrenska Academy, University of Gothenburg, Göteborg, Sweden

    • L Holmqvist
    • , L Mortensen
    • , C Kanckos
    • , C Ljungman
    •  & K Manhem
  2. School of Public Health and Community Medicine, Sahlgrenska Academy, University of Gothenburg, Göteborg, Sweden

    • K Mehlig

Authors

  1. Search for L Holmqvist in:

  2. Search for L Mortensen in:

  3. Search for C Kanckos in:

  4. Search for C Ljungman in:

  5. Search for K Mehlig in:

  6. Search for K Manhem in:

Competing interests

The authors declare no conflict of interest.

Corresponding author

Correspondence to L Holmqvist.

About this article

Publication history

Received

Revised

Accepted

Published

DOI

https://doi.org/10.1038/jhh.2011.99

Further reading