Association between physical activity and hypertension: A retrospective study

Background: Many guidelines recommend physical activity to prevent hypertension. However, the recommendations of different studies and guidelines for physical activity to prevent hypertension are not uniform. The objective of this study is to examine the association between the prevalence of hypertension and physical activity in different patterns, intensities, and durations. Methods: The sample was from a cross-sectional study of 479,842 subjects called the China Hypertension Survey, 2012-2015. We selected participants with physical activity information from the survey results in Sichuan Province (n = 19,277) for this study. As an advantage of our research, we used multiple imputation to supplement the missing data, instead of mean lling or delete missing values directly. The relationship between physical activity and hypertension was mainly analyzed by the Wilcoxon rank-test, Chi-squared tests, and the logistic regression model. Results: There were 2,006 participants eligible. A total of 354 (17.6%) participants were hypertensive. Participants who were older, less educated, had higher BMI, used alcohol regularly, and had a family history of hypertension were more likely to get hypertension (p-value<0.0001). Both walking or cycling ≥ 15 hours/week (OR = 0.58; 95% CI = 0.36-0.95) and the sleeping duration ≥ 10 hours/non-working day (OR = 0.40; 95% CI = 0.19-0.81) are related to the low prevalence of hypertension. The risk factor associated with hypertension is the moderate activity related to leisure ≥ 15 hours/week (OR = 1.98; 95% CI = 1.14-3.40). Conclusions: This study indicates that there is a signicant association between the duration of physical activity and the prevalence of hypertension. Moreover, this association is affected by different patterns of physical activity. those with higher blood pressure were older, less educated, had higher BMI levels, drinking more than once a month, and had a family history of hypertension. We found no signicant differences in gender (p-value = 0.7175), marital status (p-value = 0.6648), and smoking habits (p-value = 0.1133).

breathing and heartbeat, such as carrying heavy loads and digging. Moderate activity was an activity that required moderate physical effort or caused a slight increase in breathing and heartbeat, such as sawing wood, washing clothes, and cleaning. As for leisure-related activity, a vigorous activity referred to, causing a signi cant increase in breathing and heartbeat. Moderate activity referred to, causing a slight increase in breathing and heartbeat. The duration of static behavior per working day was including sitting, working, studying, reading, watching TV, using a computer, and resting, except for sleeping hours. The duration of static behavior per non-working day was including all static action on weekdays except sleeping time and working time.

Statistical analysis Multiple Imputation
Parts of our data were incomplete. To avoid errors caused by deleting missing values, we used multiple imputation (MI) to ll in the missing information. MI is an advanced method for dealing with missing values. Bias and the uncertainty of missing value estimation are the disadvantages of single imputation. If appropriately implemented, MI can be a solution to replace single imputation [17]. Currently, due to a lack of familiarity and computational challenges, MI is not fully utilized in the medical literature [18]. This study used the MICE package in R for data lling. By default, continuous variables use predicted mean matching (PMM), and dichotomous variables use Logistic regression [19]. To minimize the simulation error, we obtained ve datasets and extracted the third one as the interpolation result.

Statistical description
The statistical description refers to the form of data display. The mean ± standard deviation represents the continuous variables in this investigation, and the categorical variables are displayed by frequency (frequency).

Hypothetical test
For continuous data, when the data satis es the normal distribution, and the variances are equal, the T-tests (two-tailed) is used. If not satis ed, the Wilcoxon rank-test is used. Adopt the chi-square test or sher's exact test for categorical variables when necessary. Values are mean ± SD, SD = standard deviation The participants' characteristics strati ed by hypertensive status. Table 2 shows the hypertension distribution of participants in demographics. It indicates that there were signi cant differences between blood pressure categories in age(p-value < 0.0001), education(p-value < 0.0001), BMI(p-value < 0.0001), drinking frequency (p-value < 0.0001), and family history of hypertension(p-value < 0.0001). Speci cally, participants with lower compared to those with higher blood pressure were older, less educated, had higher BMI levels, drinking more than once a month, and had a family history of hypertension. We found no signi cant differences in gender (p-value = 0.7175), marital status (p-value = 0.6648), and smoking habits (p-value = 0.1133). b Values are mean ± SD, SD = standard deviation.
Distribution of hypertension according to physical activity patterns.
There were differences in the distribution of hypertension for moderate activity related to work, agriculture, and housework (p-value = 0.0164) and vigorous activity related to leisure(p-value = 0.0003). Participants in hypertension and non-hypertension have signi cant differences in the duration of vigorous activity related to leisure (p-value = 0.0002) and the sleeping duration per non-working day (p-value < 0.0001). There was no signi cant difference between blood pressure categories in the remaining physical activity patterns (Table 3). The associations of physical activity with hypertension, strati ed by physical activity categories are illustrated in Fig. 4. In model 1, we adjusted for age and sex. For activity related to work, agriculture, and housework and static behavior, time spent doing physical activity and hypertension were not associated. For activity related to transportation, the duration of walking or cycling ≥ 15 hours/week was associated with the low prevalence of hypertension Our study con rmed that the duration of physical activity was associated with the risk of hypertension. Logistic regression indicated that participants who usually took part in the walk or ride over 15 hours every week (OR = 0.58; 95% CI = 0.36-0.95) were less likely to have hypertension. Transportation is a necessary existence in life. We recommend that it is better to choose to walk or ride to prevent hypertension as long as it does not affect life and work.
We found that participants who regularly participated in moderate leisure-related physical activity for over 15 hours every week (OR = 1.98; 95% CI = 1.14-3.40) were more likely to have hypertension, which can be due to temporary physiological stress caused by vigorous exercise [20]. Quinn TJ et al. also found in the longevity study that hypertension is more common in more active men [21], which may be related to overtraining that can cause abnormal heart function Tables   Figure 1 Flow chart of the study population (n=2006).

Figure 2
Distribution of missing data. The red square indicates missing data. The blue square indicates no missing data. The "b411" to "b452m" are the labels for different questions in the questionnaire, where "h" stands for "hours", and "m" stands for "minutes".

Figure 3
Comparison chart of distribution before and after interpolation. A total of 5 data sets were created, in which the red part is the lled data, and the blue part is the original data. The "b411" to "b452m" are the labels for different questions in the questionnaire, where "h" stands for "hours", and "m" stands for "minutes".