Role of the Intensity of Habitual Physical Activity in the Maintenance of Normal Blood Pressure: Findings From the SUN Longitudinal Cohort Study

Abstract

Background

Physical activity (PA) is known to protect against incident hypertension, but the preferred intensity of PA to prevent hypertension remains unknown. Energy expenditure (EE) in PA is generally considered the primary determinant of effect, whereas intensity is usually considered nondifferential provided it is moderate or above. However, intensity may produce its own distinct effect.

Methods and Results

We used data from the SUN (Seguimiento Universidad de Navarra) cohort—a large prospective longitudinal cohort in Spain—to investigate the relation between intensity of habitual PA and hypertension incidence. Average intensity of habitual PA was calculated including both leisure time PA and incidental PA (walking and stairclimbing). Hazard ratios (HRs) for incident hypertension and 95% CIs were estimated using Cox regression analyses adjusted for EE, body mass index, and other important covariables. Comparative models explored how duration of time in PA and EE in PA related to hypertension incidence. In the study,10 524 participants without prior diagnosis of hypertension (62.5% women, mean age 36.2 years, mean body mass index 23.3 kg/m²) were followed for 126 876 person‐years. A total of 1504 cases of incident hypertension emerged. After adjustment for EE and other covariables, increasing intensity of PA was monotonically associated with decreased risk for incident hypertension (adjusted HR for Q5 versus Q1, 0.77 [95% CI, 0.64–0.92]). In contrast, increasing time in PA did not appear to affect risk of incident hypertension following adjustment for EE (aHR for Q5 versus Q1, 0.94 [95% CI, 0.57–1.55]).

Conclusions

Intensity of habitual PA is independently and inversely associated with incidence of hypertension.

Registration

URL: https://www.clinicaltrials.gov; Unique Identifier: NCT02669602.

Nonstandard Abbreviations and Acronyms

CoPA

Compendium of Physical Activities

EE

energy expenditure

PA

physical activity

SUN

Seguimiento Universidad de Navarra

Clinical Perspective.

What Is New?

• Our study found that increasing the average intensity of habitual physical activity (PA) was associated with reduced risk of new‐onset hypertension even after controlling for the effect of total energy expenditure; the inverse relationship between PA intensity and risk of hypertension was linear across the entire range of intensities studied (from 2.5 to 10.5 metabolic equivalents of a task).

What Are the Clinical Implications?

• Our results should be used to encourage individuals to increase the average intensity level of their regular PA to protect against hypertension incidence.

• In contrast to current leading guidelines, our findings suggest that individuals should prioritize increasing the intensity of PA over increasing the time they spend in PA, as increasing PA intensity not only allows individuals to increase energy expenditure but also brings additional independent benefit. There is no minimum threshold of intensity that must be attained before additional benefit accrues; any incremental increase in average PA intensity is associated with added protective effect.

Hypertension is a leading modifiable risk factor for cardiovascular disease and death. ¹ , ² , ³ Although medical management of hypertension may mitigate morbidity and mortality, it is associated with significant health care burden and cost. ⁴ , ⁵ Primary prevention must be prioritized.

Physical activity (PA) is a known modifiable protective factor reducing the risk of incident hypertension. ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² Although most studies to date consider PA only in terms of energy expenditure (EE), ¹⁰ , ¹¹ there are many other modifiable aspects of PA. It is worth investigating factors such as intensity, frequency, duration, and type of PA to better understand which elements produce most protective effect. ⁸ , ⁹ , ¹³

The intensity of PA, measured in metabolic equivalents of a task (METs), is a key component of PA and a key determinant of EE. Although some research has identified a protective benefit associated with increasing intensity of PA, it remains unclear whether increasing intensity helps to prevent hypertension and whether any benefit observed results only from the associated increase in EE. Current PA guidelines from the World Health Organization and the US Department of Health and Human Services suggest that moderate intensity PA (3–6 METs) and vigorous intensity PA (>6 METs) have equal benefit provided EE is equal. ¹⁴ , ¹⁵ They recommend a minimum 150 to 300 minutes a week of moderate intensity exercise or 75 to 150 minutes of vigorous intensity exercise, or equivalent combinations of the 2. Nowhere do they suggest any intrinsic benefit of increasing intensity above moderate, except that EE will accumulate faster.

Yet it is possible that the intensity of PA may produce its own differential effect independent of the effect it has on EE. Evidence from the SUN (Seguimiento Universidad de Navarra) cohort has shown that, even after controlling for EE, higher intensity of PA is associated with reduction in incidence of cardiovascular disease, ¹⁶ metabolic syndrome, ¹⁷ and type 2 diabetes, ¹⁸ and an association has been observed between faster self‐perceived walking pace and reduced rates of hypertension. ¹⁹ Other investigations find that engaging in higher intensity PA predicts lower cardiovascular mortality. ²⁰ The aim of our study was to examine how habitual intensity of PA associated with hypertension incidence in the SUN cohort.

METHODS

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Study Population

The SUN study is a multipurpose longitudinal cohort study focused on Mediterranean diet and lifestyle patterns and their relation to cardiovascular and other outcomes. ²¹ It follows a long‐running dynamic cohort, permanently open since December 1999. Participants are graduates of Spanish universities and >50% hold health‐related professional degrees. Data are self‐reported. At entry, a baseline questionnaire collects detailed information on diet, lifestyle, and health‐related characteristics. Follow‐up questionnaires are distributed every 2 years to be completed online or on hardcopy. Voluntary return of completed questionnaires is taken as informed consent in accordance with the Declaration of Helsinki. ²² The study's protocol was approved by the Institutional Review Board of the University of Navarra and registered at clinicaltrials.gov (NCT02669602).

The SUN database as of May 2022 included 23 133 participants. To define our study sample, we excluded those without follow‐up as well as those whose self‐reported data were deemed unreliable because their food frequency questionnaire responses resulted in calculated total energy intake outside a plausible range. ²³ We excluded participants with preexisting hypertension as diagnosed by a physician, participants using antihypertensives (beta blockers, calcium channel blockers, angiotensin‐converting‐enzyme inhibitors, and angiotensin receptor II blockers) and participants reporting hypertensive range blood pressures (systolic BP ≥140 mm Hg or diastolic BP ≥90 mm Hg) at cohort entry. ²⁴ Finally, we excluded participants with incomplete data sets for PA variables. There remained 10 524 participants available for inclusion (Figure 1). ²³

Figure 1. Flow chart of study sample selection from the Seguimiento Universidad de Navarra/University of Navarra follow‐up cohort 1999 to 2022.

*Predefined limits excluded participants with energy intake <500 or >3500 kcal/d for women, or <800 or >4000 kcal/d for men. Limits were defined as in Willett. ²³

Exposure Assessment—Intensity

Intensity of PA can be understood as a rate of EE over time or as a proportion of maximum workload. We approached intensity as a rate quantified in terms of METs. METs are an absolute measure of intensity, based on a scale where 1 MET is equal to the energy consumed over 1 hour at seated rest, and 2 METs are twice that rate and so forth.

Average intensity of habitual PA was calculated for each participant based on self‐reported habitual engagement in PA. We attempted to include both incidental exercise (PA resulting from activities of daily living) and leisure‐time physical activity (LTPA) when calculating average intensity. We were able to incorporate incidental walking and stairclimbing but did not have data on the nature or duration of any other forms of occupational PA or domestic‐chore based PA.

The baseline questionnaire asked, “How much do you usually walk each day?” (answer choices: <10 minutes/10–20 minutes/21–30 minutes/30 minutes–1 hour/1–2 hours/>2 hours) and “How many flights of stairs do you climb each day?” (≤2/3–4/5–9/10–14/≥15). We converted the number of flights of stairs into time climbing stairs (≤2 flights=30 s, 3–4 flights=1 minute 45 s, 5–9 flights=3 minutes 30 s, 10–14 flights=6 minutes, ≥15 flights=8 minutes 45 s).

The question set used to assess LTPA has been validated in our cohort. ²⁵ Participants were asked, “Do you exercise?” (yes/no), then directed to complete a 17‐item LTPA question set titled, “During the last year, how much time on average did you spend doing the following activities?.” Specified activities were “walking”; “jogging”; “running”; “cycling”; “stationary bike”; “swimming”; “racket sports”; “soccer and indoor soccer”; “other team sports (eg, basketball, handball)”; “dance and aerobics”; “hiking and rock‐climbing”; “gym‐based work outs”: “gardening, yard maintenance, pool maintenance and renovations”; “skiing, ice‐skating and rollerblading”; “judo, karate and martial arts”; “sailing” and “other sports or activities not listed.” The seasonal variation of LTPA was also taken into account: for each activity, participants answered, “Average time per week” (Never/1–4 minutes/5–19 minutes/20–59 minutes/<1 hour/1–1.5 hours/2–3 hours/4–6 hours/7–10 hours/≥11 hours) as well as “Months per year” (<3 months/3–6 months/>6 months). Reported time per week was multiplied by 0.125 for <3 months, or 0.375 for 3 to 6 months, or 0.75 for >6 months. Nonresponse for an activity was interpreted as nonparticipation, whereas partial response—that is, answering only “hours per week” or “months per year” but not both—was considered an incomplete answer. Participants with incomplete data for incidental PA or LTPA were excluded from the analysis. PA data were explored for plausibility and no data were identified as implausible.

METs for each activity were adapted from the Compendium of Physical Activities (CoPA), a publicly available reference text for assigning standard METs to different forms of PA. ²⁶ , ²⁷ METs assigned to walking (leisure time walking and incidental walking) depended on self‐reported walking pace (slow/average/fast/very fast). A total of 59 participants (0.6% of our sample) missing responses for walking pace were assigned “average” pace.

Total weekly time in PA was summed and total weekly EE was calculated. Average intensity of PA was calculated dividing average EE per week by average weekly time in PA, generating values for intensity on a continuous scale. Rather than using the predefined categories of light (<3 METs), moderate (3–6 METs), and vigorous (>6 METs) intensity, we used quintiles of continuous intensity as categories for analysis.

Outcome Assessment—Incident Hypertension

Incidence of hypertension was assessed based on participant report of hypertension diagnosis by a physician. Each follow‐up questionnaire asked if participants had been newly diagnosed with hypertension, defined in‐line as systolic BP >130 mm Hg or diastolic BP >85 mm Hg, and the month and year of diagnosis. This assessment of incident hypertension has been validated in a subsample of our cohort. ²⁸ Participants also reported their regular medications at each point of follow‐up and, on 2 questionnaires, reported a recent in‐office BP reading. In a sensitivity analysis (Table S1), participants newly taking antihypertensive medications and those reporting hypertensive range in‐office BPs (systolic BP ≥140 mm Hg or diastolic BP ≥90 mm Hg) were considered as additional positive cases.

Covariables

Covariables for adjustment were sex (male/female), age (years), calendar year of completion of first questionnaire, years of tertiary education (years), total energy intake (kcal/day), special diets (yes/no), Mediterranean Diet score (Trichopoulous score), ²⁹ sodium intake (mg/day), coffee intake (cups/day), alcohol intake (g/day), smoking exposure (pack years), night‐time sleep (hours/night), television viewing time (hours/day), time spent doing domestic chores (hours/day), body mass index (kg/m²), cardiovascular disease (yes/no), diabetes (yes/no), cancer (yes/no), and parental history of hypertension (yes/no). Continuous covariables were reduced to quintiles for the analysis, except age (categorized in decades) and calendar year of first questionnaire (categories in 6 groups). Television viewing time, nighttime sleep and daily time spent in domestic chores were the only covariables with missing values. For these covariables, following categorization of the continuous data into quintiles, a sixth category was created for those missing values.

Statistical Analysis

Cox regression analyses were used to study the relationship between intensity of habitual PA and time to hypertension incidence. The end point was whichever occurred first out of hypertension diagnosis, death or last completed follow‐up questionnaire. The lowest quintile of intensity was used as reference category and the Breslow method was used for tied observations. We performed an unadjusted analysis, then an analysis adjusted for sex and stratified for age and year of cohort entry, then an analysis adjusted for EE as well as all covariables and stratified for age and year of cohort entry.

This same sequence of Cox modeling was used to analyze the association between time habitually spent in PA (in quintiles) and hypertension incidence, and a similar sequence was used to analyze the association between EE in PA (in quintiles) and hypertension incidence (this model differed only in that the multivariable‐adjusted model for EE did not include EE as additional adjustment variable). We also performed multivariable‐adjusted Cox regression modeling examining each PA factor as a continuous variable.

Cubic spline regression modeling with adjustment for all covariables was used to visualize the relationship between PA intensity and incidence of hypertension across the continuous scale for PA intensity (exploring models both with and without adjustment for EE). The same cubic spline modeling was also used to explain the relationship between time in PA on a continuous scale and incident hypertension. These models were evaluated using Wald's test (testparm command in STATA) to determine if the spline terms contributed significantly enough to justify their use. In our sample, the spline terms for the models that adjusted for EE in addition to all other covariables lacked significance. To see whether these spline terms would be significant given a larger sample size, we replicated these models using an expanded sample (17 146 participants) made possible by imputation of missing PA variables.

To compare the effect of habitual PA intensity against the effect of habitual time in PA, we conducted an analysis using categories based on intersecting tertiles of intensity and time. Time to hypertension was analyzed using multivariable‐adjusted Cox regression (without adjustment for EE) taking the intersection of the lowest tertiles of PA intensity and PA time as referent category.

Stratified Cox regression analyses explored the association between PA intensity and incidence of hypertension in subgroups of age (<50 versus ≥50 years), sex (male versus female), overweight (<25 versus ≥25 kg/m²), parental history of hypertension (yes versus no), hours of television viewing (above or below sample median of 1.4 hours), and amount of nighttime sleep (recommended 7–9 hours ³⁰ versus other). The likelihood ratio test of nested models was used to check for interaction by these a priori selected factors.

Multiple sensitivity analyses were performed. In one analysis (Table S1), we counted participants with new antihypertensive use or new self‐report of systolic BP >140 mm Hg or diastolic >90 mm Hg as additional cases (noting that the assessment of hypertension based on self‐report of in‐office BP has previously been validated in our cohort). ³¹ Another analysis excluded participants with hypertension incidence before 2 years of follow‐up or with follow‐up time <2 years (Table S2); another declared maximum end point at 12 years of follow‐up (Table S3). A final sensitivity analysis (Table S4) repeated the main analysis on a larger sample of 17 146 participants made possible using imputation methods to again include the 6622 participants excluded from the main analysis with incomplete PA data sets: assuming missing at random, logistic regression with age and sex as explanatory variables was used to predict missing PA values.

RESULTS

Of the 10 524 participants included in the analysis, 6577 (62.5%) were women and 5924 (56.3%) were health care professionals. The mean age at cohort entry was 36.2 years and the mean body mass index was 23.3 kg/m². During follow‐up (126 876 person‐years), there emerged 1504 cases of incident hypertension, equating to an absolute risk of 14.29%.

Median intensity of habitual PA was 4.2 METs (interquartile range [IQR], 3.8–4.8). Median intensity of habitual LTPA was 4.9 METs (IQR, 4.1–5.7) whereas median intensity of habitual incidental exercise was 3.9 METs (IQR, 3.4–4.1). Median weekly time in PA was 5.3 hours (IQR, 2.2–9.6). Median EE was 23.4 MET‐hours per week (IQR, 9.9–42.5). On average, incidental exercise accounted for 44.5% of EE and 49.3% of time in PA across the week. Walking was also the most common LTPA, with 61.2% of participants engaging in walking for leisure.

The hazard of hypertension increased with age; at 40, the probability of being hypertension free was 98% (95% CI, 97–98), whereas, by 80, the probability of remaining free of hypertension reduced to 22% (95% CI, 19–25). Mean age at time of hypertension diagnosis was 51 years (95% CI, 50–52) and median follow‐up time before diagnosis was 7.4 years (IQR, 3.3–11.5). Ten percent of cases reported hypertension diagnosis within 1.5 years of cohort entry. This raised concern about possible inverse causation, so a sensitivity analysis was performed excluding 226 participants with hypertension incidence before 2 years follow‐up (Table S2), and it confirmed similar results to those of our main analysis.

The baseline characteristics of our study sample across quintiles of PA intensity are presented in Table 1. Women were underrepresented in the uppermost quintile of PA intensity, and participants in the lowest quintile of PA intensity were slightly older on average. Alcohol intake, time in PA, and EE in PA tended to increase across increasing quintiles of PA intensity, whereas baseline prevalence of depression and smoking exposure decreased.

Table 1.

Baseline Characteristics of Study Sample According to Quintiles of Average Intensity of Physical Activity

	Quintiles of average intensity of physical activity (metabolic equivalents)
	Q1 (2.5–3.6 METs)	Q2 (3.6–4.0 METs)	Q3 (4.0–4.4 METs)	Q4 (4.4–5.0 METs)	Q5 (5.0–10.5 METs)
No.	2105	2107	2103	2105	2104
Women, %	69.50	70.38	67.52	62.47	42.59
Age, y	38.2 (11.7)	36.5 (11.4)	35.6 (10.7)	35.5 (10.3)	35.1 (9.8)
Year of cohort entry	2003 (3)	2003 (4)	2003 (4)	2003 (4)	2004 (4)
Years of university, y	5.0 (1.4)	5.0 (1.4)	5.0 (1.4)	5.1 (1.5)	5.3 (1.6)
Energy intake, kcal/d	2304 (617)	2342 (592)	2352 (594)	2353 (601)	2403 (620)
Special diets, %	6.75	6.50	6.28	6.94	7.41
Mediterranean Diet score (out of 9 points)	4.0 (1.7)	4.1 (1.8)	4.1 (1.8)	4.2 (1.8)	4.2 (1.8)
Sodium intake, mg/d	3270 (2054)	3269 (2027)	3325 (2165)	3265 (1986)	3460 (2236)
Coffee intake, cups/d	1.3 (1.3)	1.3 (1.3)	1.3 (1.2)	1.2 (1.3)	1.2 (1.3)
Alcohol intake, g/d	5.6 (9.5)	5.9 (9.6)	6.2 (8.6)	6.5 (9.6)	7.4 (8.9)
Smoking exposure, pack years	7.0 (10.7)	5.9 (9.5)	5.2 (8.8)	4.7 (8.1)	4.4 (7.5)
Nighttime sleep, h/night	7.4 (1.0)	7.3 (0.9)	7.4 (1.0)	7.4 (0.9)	7.3 (0.9)
Television, h/d	1.8 (1.4)	1.8 (1.4)	1.6 (1.3)	1.5 (1.2)	1.5 (1.2)
Time doing domestic chores, h/d	1.8 (1.6)	1.7 (1.5)	1.6 (1.6)	1.5 (1.5)	1.3 (1.5)
Body mass index, kg/m2	24.0 (3.9)	23.1 (3.5)	22.9 (3.2)	22.9 (3.1)	23.3 (2.8)
Prevalent cardiovascular disease, %	0.57	1.04	0.57	0.48	0.71
Prevalent diabetes, %	1.62	1.33	1.33	1.33	0.71
Prevalent cancer, %	2.61	2.71	2.62	1.71	1.85
Prevalent depression, %	12.97	11.96	10.03	11.02	8.65
Parental history of hypertension, %	25.65	24.54	25.44	25.70	24.33
Time spent in physical activity, h/wk	4.8 (4.3)	6.0 (5.0)	6.9 (5.5)	7.6 (5.7)	8.4 (6.3)
Energy expenditure in physical activity, MET‐h/wk	16.2 (14.5)	23.2 (19.4)	28.9 (23.1)	36.0 (26.7)	48.1 (37.3)

Continuous variables are expressed as mean±SD in parentheses and categorical variables are expressed as percentages. MET indicates metabolic equivalents of a task.

Cox regression modeling with multivariable adjustment and adjustment for EE (Table 2) found that greater PA intensity was associated with reduced risk of incident hypertension in a monotonic pattern (P for trend=0.005). The adjusted risk was 23% lower in the highest intensity quintile compared with the lowest intensity quintile. As expected, increasing EE was also associated with reduced risk of incident hypertension: after multivariable adjustment, the fourth and fifth quintiles for EE had relative risk reductions of 19% and 21% respectively compared with the lowest quintile for EE (P for trend=0.003). Cox regression analysis of intensity as a continuous variable estimated that per each 1 MET increase in average intensity, the hazard ratio (HR) for incident hypertension decreased by almost 10% following adjustment for all covariables, including EE (adjusted HR, 0.91 [95% CI, 0.85–0.98], P=0.009). In contrast, following multivariable adjustment including adjustment for EE, additional time in PA was not seen to have any effect on hazard of hypertension (adjusted HR per 5 additional hours of weekly time in PA, 0.99 [95% CI, 0.89–1.10, P=0.84]) and there was no evidence of an association between quintiles of time in PA and risk of incident hypertension (P for trend=0.98).

Table 2.

Hazard Ratios for the Incidence of Hypertension According to Quintiles of Physical Activity Parameters—Intensity of Physical Activity, Time Spent in Physical Activity, and Total Energy Expenditure in Physical Activity

	Q1 (2.5–3.6 METs)	Q2 (3.6–4.0 METs)	Q3 (4.0–4.4 METs)	Q4 (4.4–5.0 METs)	Q5 (5.0–10.5 METs)	P for trend
Average intensity of physical activity (metabolic equivalents)
Person‐years	24 573.7	25 516.5	25 681.8	25 613.5	25 490.5
Cases	355	319	286	293	251
Event rate	14.4 per 1000 person‐years	12.5 per 1000 person‐years	11.1 per 1000 person‐years	11.4 per 1000 person‐years	9.8 per 1000 person‐years
Crude HR	1 (Ref.)	0.94 (0.81–1.09)	0.88 (0.75–1.03)‡	0.92 (0.78–1.07)	0.79 (0.67–0.93)‡	0.006
Age and sex adjusted HR	1 (Ref.)	0.94 (0.80–1.09)	0.85 (0.73–1.00)	0.86 (0.73–1.01)	0.67 (0.56–0.79)‡	<0.001
Multivariable adjusted HR* , †	1 (Ref.)	0.98 (0.84–1.14)	0.93 (0.80–1.10)	0.95 (0.80–1.12)	0.77 (0.64–0.92)‡	0.005

	Q1 (0.8–1.9 h)	Q2 (1.9–3.9 h)	Q3 (3.9–6.7 h)	Q4 (6.7–10.9 h)	Q5 (10.9–53.3 h)	P for trend
Time spent in physical activity per week (h)
Person‐years	25 638.9	25 403.5	25 480.8	25 130.6	25 222.2
Cases	315	264	322	302	301
Event rate	12.3 per 1000 person‐years	10.4 per 1000 person‐years	12.6 per 1000 person‐years	12.0 per 1000 person‐years	11.9 per 1000 person‐years
Crude HR	1 (Ref.)	0.79 (0.67–0.93)‡	0.94 (0.81–1.10)	0.85 (0.72–0.99)‡	0.80 (0.68–0.94)‡	0.040
Age and sex adjusted HR	1 (Ref.)	0.76 (0.64–0.89)‡	0.88 (0.75–1.03)	0.78 (0.66–0.91)‡	0.70 (0.59–0.82)‡	<0.001
Multivariable adjusted HR* , †	1 (Ref.)	0.76 (0.55–1.05)	0.99 (0.67–1.48)	1.01 (0.65–1.58)	0.94 (0.57–1.55)	0.983

	Q1 (2.2–7.5 MET‐h)	Q2 (7.5–17.0 MET‐h)	Q3 (17.0–29.6 MET‐h)	Q4 (29.6–48.8 MET‐h)	Q5 (48.8–354.3 MET‐h)	P for trend
Total energy expenditure in physical activity per week (metabolic equivalent hours)
Person‐years	25 333.3	25 044.4	25 717.0	25 710.0	25 071.4
Cases	307	281	330	291	295
Event rate	12.1 per 1000 person‐years	11.2 per 1000 person‐years	12.8 per 1000 person‐years	11.3 per 1000 person‐years	11.8 per 1000 person‐years
Crude HR	1 (Ref.)	0.86 (0.73–1.01)	0.94 (0.81–1.10)	0.81 (0.69–0.95)‡	0.83 (0.70–0.97)‡	0.032
Age and sex adjusted HR	1 (Ref.)	0.85 (0.72–1.00)	0.89 (0.76–1.04)	0.75 (0.64–0.88)‡	0.72 (0.61–0.85)‡	<0.001
Multivariable adjusted HR*	1 (Ref.)	0.91 (0.77–1.07)	0.96 (0.82–1.13)	0.81 (0.68–0.95)‡	0.79 (0.67–0.94)‡	0.003

^*Adjusted for sex (male/female), age (y), calendar year of entry into the cohort, years of university (y), total energy intake (kcal/d), Mediterranean Diet score (out of 9), special diet (yes/no), sodium intake (mg/d), coffee intake (cups/d), alcohol intake (g/d), smoking exposure (pack years), nighttime sleep (h/night), television viewing time (h/d), time spent doing domestic chores (h/d), body mass index (kg/m²), prevalent cardiovascular disease (yes/no), prevalent diabetes (yes/no), prevalent cancer (yes/no), parental history of hypertension (yes/no), and change in regular physical activity as reported during early follow‐up (no change/increased/decreased). HR indicates hazard ratio; and MET, metabolic equivalents of a task.

^† Also adjusted for total energy expenditure (MET‐h/wk).

^‡ P‐value <0.05.

Regression analysis of intersecting tertiles of PA intensity and time in PA (Figure 2) found that increasing PA intensity and increasing time in PA were both associated with decreased hypertension incidence when EE was not included in multivariable adjustment. However, the reduced risk associated with the highest tertile of PA intensity was significant across all tertiles of time in PA, whereas the risk reduction associated with the highest tertile of time in PA was not significant in the lowest tertile of PA intensity.

Figure 2. Multivariable‐adjusted^‡ hazard ratios for the incidence of hypertension comparing the effect of weekly exercise time (in tertiles) and average exercise intensity (in tertiles).

Statistically significant results marked in bold and underlined. ^‡Adjusted for sex (male/female), age (y), calendar year of entry into the cohort, years of university (y), total energy intake (kcal/d), Mediterranean Diet score (out of 9), special diet (yes/no), sodium intake (mg/d), coffee intake (cups/d), alcohol intake (g/d), smoking exposure (pack years), nighttime sleep (h/night), television viewing time (h/d), time spent doing domestic chores (h/d), body mass index (kg/m²), prevalent cardiovascular disease (yes/no), prevalent diabetes (yes/no), prevalent cancer (yes/no), parental history of hypertension (yes/no), and change in regular physical activity as reported during early follow‐up (no change/increased/decreased). aHR indicates adjusted hazard ratio; and MET, metabolic equivalents of a task.

Multivariable‐adjusted cubic spline modeling of the association between intensity of PA and hazard of incident hypertension revealed an inverse relation that linear across the continuum of intensity, with no minimum threshold for effect (Figure 3). The P value for this model was significant at 0.01, justifying the use of spline terms. There was insufficient evidence to show that the spline terms improved the multivariable‐adjusted model of the relationship between time in PA and hypertension incidence (P=0.07). In our sample, when either of these models was adjusted for EE in addition to other covariables, there was no evidence that the spline terms improved the model (P=0.08 and P=0.68, respectively). However, when we applied cubic spline modeling with multivariable adjustment and adjustment for EE to a larger sample of 17 146 participants facilitated by the imputation of missing PA variables, the terms of the cubic spline models were significant (P=0.004 for PA intensity and P=0.02 for time in PA). The spline model revealed nonlinearity in the relationship between time in PA and hazard of incident hypertension and found that increasing time in PA was associated with increased risk of hypertension. These cubic spline models adjusted for EE are presented in Figure S1.

Figure 3. Cubic regression spline showing multivariable‐adjusted* hazard ratio for the incidence of hypertension according to average intensity of physical activity (METs) as a continuous variable.

*Adjusted for sex (male/female), age (y), calendar year of entry into the cohort, years of university (y), total energy intake (kcal/d), Mediterranean Diet score (out of 9), special diet (yes/no), sodium intake (mg/d), coffee intake (cups/d), alcohol intake (g/d), smoking exposure (pack years), nighttime sleep (h/night), television viewing time (h/d), time spent doing domestic chores (h/d), body mass index (kg/m²), prevalent cardiovascular disease (yes/no), prevalent diabetes (yes/no), prevalent cancer (yes/no), parental history of hypertension (yes/no), and change in regular physical activity as reported during early follow‐up (no change/increased/decreased). aHR indicates adjusted hazard ratio; and MET, metabolic equivalents of a task.

Subgroup analysis (Figure 4) found that the reduced risk of hypertension in the upper quintile of PA intensity compared with the lowest quintile of PA intensity (quintiles defined using the same cut‐points as defined in the principal analysis) was most notable for participants who were aged >50, men, overweight, and those who did not sleep the recommended 7 to 9 hours each night. Sex, overweight, and getting the recommended amount of sleep were identified as interaction variables in the relationship between PA intensity and risk of hypertension incidence. Age (<50 versus ≥50 years) was not identified as an interaction factor.

Figure 4. Multivariable‐adjusted hazard ratios and 95% CIs for incident hypertension in the highest quintile of exercise intensity (5.0–10.5 METs) compared with the lowest quintile (2.5–3.6 METs) analyzed within dichotomous subgroups of the study sample.^§ .

^§Quintiles of exercise intensity were defined using cut‐points as for the total study sample, but the modeling for each subgroup was confined to the subgroup sample alone. *Adjusted for sex (male/female), age (y), calendar year of entry into the cohort, years of university (y), total energy intake (kcal/d), Mediterranean Diet score (out of 9), special diet (yes/no), sodium intake (mg/d), coffee intake (cups/d), alcohol intake (g/d), smoking exposure (pack years), nighttime sleep (h/night), television viewing time (h/d), time spent doing domestic chores (h/d), body mass index (kg/m²), prevalent cardiovascular disease (yes/no), prevalent diabetes (yes/no), prevalent cancer (yes/no), parental history of hypertension (yes/no), and change in regular physical activity as reported during early follow‐up (no change/increased/decreased). ^φStratified analyses for age, BMI, and television viewing time were adjusted for the continuous variable of the self‐same variable under analysis. Stratified analysis for sex was adjusted for all variables listed in *, except sex (male/female). Stratified analysis for parental history of hypertension was adjusted for all variables listed in *, except parental history of hypertension (yes/no). BMI indicates body mass index; HR, hazard ratio; and MET, metabolic equivalents of a task.

DISCUSSION

This study found strong evidence that as intensity of habitual PA increased, incident hypertension decreased, even after controlling for EE. This suggests an intrinsic benefit of intensity beyond its effect on EE. ³² The relationship observed was linear across the range of intensity studied (2.5–10.5 METs), and we did not identify any minimum threshold required for effect. Our findings could be used to encourage individuals to aim to increase the intensity of their habitual PA, rather than focusing on the time spent in PA, whether to maintain or increase their overall EE in PA.

These findings constitute a relatively new perspective in the literature on PA and hypertension. Although the importance of EE in reducing risk of hypertension has been identified by previous investigations, ⁸ , ¹⁰ , ¹³ , ³³ there has been little clarity regarding the role of PA intensity. Current well‐researched PA guidelines suggest no specific benefit is associated with increasing intensity above moderate (3–6 METs) and also equate the benefit derived from all intensities in the moderate range. ¹⁵ , ³⁴ , ³⁵ Moderate intensity PA has been shown to protect against incident hypertension in interventional studies ¹⁰ ; however, these studies compare moderate intensity PA to a reference of little or no PA, not against higher intensities. ¹⁰ Thus, rather than demonstrating that moderate intensity PA offers most protection against hypertension, the findings from these studies may instead point to a dose–response relationship between increasing intensity and decreased hypertension incidence. There are few interventional studies in normotensive subjects comparing different intensities of PA in relation to incident hypertension or ambulatory BP, ¹³ , ³² possibly because randomization to high‐intensity PA intervention could pose safety concerns or prompt participant dropout. Four small studies that compare between different intensities suggest that that greater intensity PA results in greater reduction in BP, ³² but the strength of the evidence is limited by small sample sizes and short follow‐up. Meanwhile, a Cochrane systematic review and meta‐analysis of walking interventions found moderate certainty evidence of a 4.1 mm Hg reduction in systolic BP, ³⁶ and another large systematic review estimated a clinically identical reduction (4.2 mm Hg) following running interventions. ³⁷

The same ambivalence emerges from review of 3 longitudinal cohort studies investigating our question. In a study of 14 998 male Harvard alumni, Paffenbarger et al found that alumni who played vigorous sports were 35% less likely to develop hypertension after 6 to 10 years of follow‐up. ³⁸ However, this study neither characterized the amount of vigorous sport played nor adjusted for important confounders.

A study by Williams and Thompson aimed to compare the effect of moderate versus vigorous intensity PA on incidence of hypertension by comparing across the National Walkers Cohort and the National Runners Cohort, 2 large British cohorts. ³⁹ They found the participants in the runners' cohort 14% less likely to develop hypertension than those in the walkers' cohort. However, their analysis did not control for EE, and they concluded that the difference observed owed to runners' additional EE in PA rather than to any effect of intensity itself. Within both cohorts, faster pace was associated with reduced risk of hypertension; however this effect was attenuated following adjustment for body mass index. This study was limited because each cohort was analyzed separately using a different reference group and there was no ability to adjust for total EE in PA, which remained an important and unmeasured confounder.

A third study followed 11 285 participants from the Australian Women's Longitudinal Study and compared the effect of moderate versus moderate‐high intensity LTPA on BP. ⁴⁰ The analysis adjusted for total EE in PA and found that at all levels of EE, women who engaged in moderate‐vigorous PA had lower risk of developing hypertension than those who engaged in moderate intensity PA alone. However, due to the overlapping CIs between the very similar intensities studied, the authors concluded there was no evidence of a significant difference.

Nonetheless, the hypothesis that intensity of habitual PA could have an independent effect on BP is biologically compelling. The intensity of PA importantly determines the instantaneous cardiovascular response to PA. The higher the metabolic demand of working muscles, the more that heart rate, BP, stroke volume, and respiratory rate must increase to deliver oxygen to muscle mitochondria. The ability to take up, transport, and use oxygen during exercise is known as cardiorespiratory fitness. An individual's cardiorespiratory fitness depends partly on factors like age and sex but can also be conditioned through PA training, with greater intensity PA shown to be more effective at increasing cardiorespiratory fitness. ²⁰ , ⁴¹ Higher cardiorespiratory fitness has been associated with lower rates of cardiovascular mortality overall ⁴² and lower rates of hypertension specifically. ⁴³ Regular high‐intensity PA has been found to cause lasting changes to cardiovascular structures. A series of changes known as the “athlete's heart” or exercise‐induced cardiac remodeling includes increased myocardial mass, enlargement of the 4 chambers, increased cardiac muscle contractility, and improved diastolic function. ⁴⁴ , ⁴⁵ As well as facilitating increased cardiac output during exercise, these changes allow for a lower heart rate at rest and may favor lower BP at rest. ⁴⁴ Other changes occur within vessels exposed to the increased stroke volumes and shearing forces associated with more intense exercise. Whereas hypertension is associated with arterial stiffness, the change in stroke pressures during exercise causes endothelial adaptation favoring increased arterial compliance. ⁴⁵ Athletes have been shown to have lower incidence of hypertension and less arterial stiffness compared with age‐matched healthy controls. ⁴⁶

The strengths of our study include its large sample size, low dropout rate, the longitudinal design, adjustment for confounders, and the use of validated tools to assess PA intensity and hypertension incidence. Compared with previous studies, our analysis was much better able to isolate intensity as study factor because we were able to adjust for multiple confounders, including critical confounders such as EE and body mass index, and compared different PA intensities within the same cohort. Moreover, including incidental PA along with LTPA when calculating PA variables further reduced possible unmeasured confounding.

Several limitations should also be acknowledged. First, our characterization of intensity, although comprehensive, remains imprecise. The LTPA question set has been validated in our cohort and CoPA was used as intended; however, even at best, these tools can only roughly approximate true intensity. Participant responses may be affected by recall bias, our questions may not be valid in the context of physical disability, and our approximation of seasonality is crude. The standard values for intensity listed in the CoPA are generic by nature and the use of any one estimate belies the variety of ways an activity can be performed. In addition, though we attempted to eliminate confounding by unmeasured PA, and though we achieved this much more completely than any previous study into our research question, we still did not capture occupational or domestic PA.

It is important to note that the METs metric itself is not a precise measurement tool because the value of 1 MET varies widely between individuals. It is affected by factors related to fitness, health, and illness in addition to factors such as age, sex, and body weight. The precision of the CoPA is further limited because its standard values assume that, regardless of the true value of 1 MET for any individual, the ratio of change in METs for any activity is the same across all people. However, in women and older people, the ratio of increase is shown to be greater. ⁴⁷ It is more appropriate to understand both METs and MET‐hours as they are found in CoPA and in this study as practical units allowing for the comparative grading of PA intensity and the rough approximation of EE. METs are not in dialogue with relative measures of intensity such as VO₂ max, nor do they provide any insight into physical processes like the aerobic versus anaerobic threshold, although such considerations are important for understandings of intensity. Future studies would do well to use personal activity intelligence and wearable heart rate tracking devices to better capture the totality of PA and more precisely relate intensity to measures like VO₂ max, heart rate reserve, and other metrics of cardiorespiratory effort. ⁴⁸

When appraising our ability to address hypertension as a study outcome, it should be noted that our assessment relies on physician diagnosis of hypertension and risks missing true positive cases who did not attend medical examination. Follow‐up questionnaires do not specifically require that participants see a physician, though we think it likely that most participants in this highly educated and highly medicalized cohort would attend medical review sufficiently often, especially in the Spanish context where there is good access to community medicine. Furthermore, any bias due to physician nonattendance would be nondifferential based on the study factor. The assessment of hypertension based on self‐reported physician diagnosis was previously validated in a subsection of our cohort and our sensitivity analysis (Table S1), including additional cases supported our main results. It should be noted that the question about diagnosis describes hypertension as “systolic >130 or diastolic >85,” a definition that does not correspond to the current American Heart Association definition of hypertension ⁴⁹ and instead describes a BP range the European Society of Cardiology considers “high normal BP.” ²⁴ That said, whether we are measuring prehypertension in addition to true hypertension, this does not change the overall clinical import of our findings.

CONCLUSIONS

Our study provides strong evidence that increasing average intensity of habitual PA is associated with increasing protection against incident hypertension, even when overall EE remains unchanged. Each incremental increase is protective, with no minimum threshold for effect. These findings should encourage individuals to increase the intensity of their habitual PA, whether by increasing their exposure to (and capacity for) high‐intensity LTPA or by increasing the intensity of their incidental PA by increasing the pace or effort or, possibly, choosing to take the stairs.

Sources of Funding

The SUN Project has received funding from the Spanish Government Instituto de Salud Carlos III, and the European Regional Development Fund (RD 06/0045, CIBER‐OBN, Grants PI10/02658, PI10/02293, PI13/00615, PI14/01668, PI14/01798, PI14/01764, PI17/01795, PI20/00564, and G03/140), the Navarra Regional Government (27/2011, 45/2011, 122/2014), and the University of Navarra.

Disclosures

None.

Supporting information

Tables S1–S4

Figure S1