Physical Activity Trajectories During Adulthood and Incident Hypertension: A National Longitudinal Cohort Study

Background We aimed to characterize the physical activity (PA) trajectories across adulthood and estimate their association with incident hypertension risk. Methods Data were obtained from the China Health and Nutrition Survey (CHNS) conducted during 2004–2011. Group-based trajectory modeling (GBTM) was used to identify distinct groups of PA trajectories. The Cox proportional hazard model was used to investigate the association between each PA trajectory group and incident hypertension. Results A total of 11,162 participants whose PA was repeatedly measured two to four times in the CHNS during 2004–2011 were included in our study. During the 5.4 years of follow-up, 3,824 incident hypertension cases were identied. Five distinct PA trajectories were identied in men: light and slight decline, light and gradual decline then sharp rise, light to medium-heavy then decline, medium-heavy and persistent decline, and heavy and sharp decline. Two distinct PA trajectories were identied in women: light and stable, and medium and gradual decline. The PA trajectory of medium-heavy and persistent decline was signicantly associated with decreased risk of hypertension in men, with the hazard ratios and 95% condence intervals being 0.80 (0.63, 0.99), 0.74 (0.59, 0.93), 0.76 (0.60, 0.96), and 0.70 (0.55, 0.88) in model 1, model 2, model 3, and model 4, respectively. Conclusions Our study identied ve distinct long ‐ term PA trajectories in men and two distinct trajectories in women. The PA trajectory of medium-heavy PA in early adulthood (at 18–37 years) followed by persistent decline (at 38–70 years) was found to be signicantly associated with a decreased risk of hypertension in later life in men. PA: Physical activity; CHNS: China Health and Nutrition Survey; GBTM: Group-based trajectory modeling; CVD: Cardiovascular disease; MET: Metabolic equivalent of energy; BMI: Body mass index; BIC: Bayesian information criterion; HR: Hazard ratio; CI: Condence intervals; FPG: Fasting plasma glucose; HbA1c: Hemoglobin A1c; Hs-CRP: High-sensitivity C-reactive protein; UA: Uric acid; TG: Triglyceride; HDL-C: High-density lipopolysaccharide-cholesterol.


Introduction
The prevalence of hypertension in China has been increasing dramatically [1]. Its prevalence in adults was 27.9% in 2015-an increase of 9.1% compared with the prevalence reported in 2002 by a national survey [2]. Compelling evidence suggests that hypertension contributes greatly to cardiovascular and cerebrovascular diseases [3]. The alarming rise in the prevalence of hypertension and its subsequent complications indicates urgent need to prevent hypertension.
As a modi able component of lifestyle, regular moderate-intensity physical activity (PA) is negatively correlated with the occurrence of hypertension, which has been con rmed [4][5][6][7]. In addition, regular moderate-to-vigorous PA and leisure-time PA are also related to decreased the risk of incident hypertension [8]. However, the PA-hypertension link over the life course has not been well characterized.
To date, most studies have focused on the measurement of PA at a single time, ignoring the dynamic PA changes throughout the life course [9,10]. As PA varies over the life course [11], assessing within-person trajectories of PA over time would better characterize the association between PA and diseases. The existing literature on PA trajectories is mainly focused on cardiovascular disease (CVD), pancreatic cancer, and physical functioning [12][13][14][15], with limited investigations of the relationship between longterm PA trajectories and incident hypertension risk, especially in the Chinese population.
Using repeated measurements of PA taken two to four times during 2004-2011, we aimed to identify the long-term PA trajectories in a national representative sample of adults  years at baseline and 25-70 years at follow-up), and estimate their associations with incident hypertension using group-based trajectory modeling (GBTM).

Study population
The CHNS is considered as a national, representative study aimed at exploring the impact of socialeconomic transformation on Chinese health and nutritional [16]. It includes multiple samples and cohorts over nine rounds of surveys in nine provinces and three megacities between 1989 and 2011 [17]. The initial round of the CHNS was conducted in 1989, and nine follow-up rounds were carried out respectively in 1991, 1993, 1997, 2000, 2004, 2006, 2009, 2011, and 2015. More details of study design, sampling method as well as eligibility criteria have been published and updated recently [18].
As adults' sedentary leisure time measurements were available after the 2004 survey, the present study included >18-year-old adults from the four surveys conducted between 2004 and 2011 (2004, 2006, 2009, and 2011). Our study included participants aged 18-63 years at baseline and 25-70 years at follow-up.
The following participants were excluded: 875 participants under the age of 18 and 1,203 participants with high blood pressure in the 2004 survey, 4,908 participants whose metabolic equivalent of energy (MET) values were outside the normal range or lost, and 1,792 participants who participated in only one survey. A total of 11,162 participants with PA measurements available from two to four surveys were included in our study. In , 2006In , 2009, and 2011 survey years, the number of participants was 8,293, 8,924, 9,211, and 8,424, respectively.

PA Measurement
In each survey, self-reported PA was collected using a standardized questionnaire [19]. Participants were surveyed the frequency of participation and time spent in different types of PA, which included occupational and domestic activities (such as cleaning, cooking, or washing), leisure activities (various forms of sports), and travel activities, and sedentary leisure activities (such as sleeping, watching TV, reading, writing or drawing, playing video games or computer games, and browsing or chatting online).
The intensity of each activity was expressed as METs, with one MET is de ned as the ratio of a person's working metabolic rate to resting metabolic rate [20]. Vigorous activities (≥6 METs) included running, ball sports, bicycling, dance or wushu classes, and other strenuous exercise. Moderate-intensity activities (3)(4)(5) METs) included walking, driving, doing housework. Light activities (0.9-3 METs) included sleeping, watching TV, reading, and other sedentary activities [21].
The PA level is the product of the speci c MET values multiplied by the time spent in each activity [22].
We multiply the number of minutes spent by each activity by the METs of the activity to calculate each PA score, and defined the total PA score as the sum of METs for all activities [23]. The total PA score ranges from 3,024 to 51,627 METs. The complete questionnaire and scoring system used to calculate the total PA score has been reported in detail elsewhere [19].

Assessment of incident hypertension
Self-reported of a history of hypertension diagnosis and/or consumption of antihypertension medication at baseline is de ned as having hypertension [24]. The incident hypertension cases in the 2006, 2009, and 2011 survey years were collected.

Other measurements
Information on age, body mass index (BMI), carbohydrate, energy, fat and protein intake, urbanization index, education, smoking, drinking, and urban or rural status was collected through a questionnaire in all surveys. Doctor using standard protocols to measured height and weight. The weight and height of people were measured to the nearest 0.1 kg and 0.1 cm, respectively. The BMI was calculated by dividing the weight (kg) by the square of the height (m).

Statistical analysis
GBTM was used to de ne the longitudinal discrete trajectories of PA over the participants' life course by SAS PROC TRAJ [25], which is available at www.andrew.cmu.edu/user/bjones/ [26]. Model t was based on the Bayesian information criterion (BIC), whereby the model with the lower BIC was favored [27].
Participant-years of follow-up were calculated from the date of the initial baseline interview until the date when participants were diagnosed with hypertension, the date of death, or the end of follow-up, whichever occurred rst.
Distributions of covariates at baseline for each PA trajectory group membership were described. Categorical variables were described as percentages (%) and were compared using chi-square tests. Continuous variables were described as the mean ± standard deviation and were compared using oneway analysis of variance. A generalized linear model was used to test differences across PA trajectories.
A Cox proportional hazard model with hazard ratio (HR) and 95% con dence intervals (CI) was used to investigate the relationship between the trajectory group membership and the incident of hypertension. Model 1 was adjusted according to age. Model 2 was adjusted according to smoke, drink, degree of education, urban and rural, and province. Model 3 was further adjusted according to BMI. Model 4 was further adjusted according to protein, energy, fat and carbohydrate intake. Sensitivity analysis excluding participants with hypertension during the rst 2 years of follow-up was conducted to assess whether the results were affected by reverse causation.

Results
Baseline characteristics across different PA trajectories A total of 11,162 participants (5,368 men and 5,794 women) from the 2004-2011 surveys were included in the analyses (Table 1). Table 1 presents the baseline characteristics across different PA trajectories in men and women. The average age of all participants was 38.25 ± 9.9 years, and those of men and women were 38.4 ± 9.1 and 37.25 ± 6.7 years, respectively. During the mean follow-up duration of 5.  Pa Trajectories Over 5.4 years Of Follow-up Figure 1 shows the ve distinct long-term PA trajectories in men and two distinct long-term PA trajectories in women. Detailed description of each group is given in Table 2. Among men, group 1 corresponds to men with light PA throughout adulthood (at 18-70 years) (n = 4,362, 81.26%); group 2 corresponds to men with light PA and gradual decline in PA (at 18-37 years) and then a sharp increase (at 38-70 years) (n = 17, 0.32%); group 3 corresponds to men with light and medium-heavy PA (at 18-55 years) followed by a gradual decline with age (at 56-70 years) (n = 260, 4.84%); group 4 corresponds to men with medium-heavy PA in early adulthood (at 18-37 years) followed by a persistent decline with age (at 38-70 years) (n = 495, 9.22%); and group 5 corresponds to men with heavy PA in early adulthood (at 18-37 years) followed by a persistent decline with age (at 38-70 years) (n = 234, 4.36%).

Association Between Pa Trajectories And Incident Hypertension
The HRs and 95% CIs of the relationship between the PA trajectory strati ed by gender and the incidence of hypertension are listed in Table 3. Among men, compared with the reference group (group 1), trajectory group 4 was signi cantly associated with a decreased risk of hypertension, with the HRs (95% CIs)   Model 2 was further adjusted by smoking, drinking, education, urban or rural status, province status based on model 1.

Model 3 was further adjusted by BMI based on model 2.
Model 4 was further adjusted by energy intake, carbohydrate intake, fat intake, protein intake based on model 3. a case/N: Number of hypertension cases/number of participants in this trajectory group In men, group 1: light and slight decline; group 2: light and gradual decline then sharp raise; group 3: light to medium-heavy; group 4: medium-heavy and persistent decline; group 5: heavy and sharp decline In women, group 1: light and stable; group 2: medium and gradual decline

Discussion
In this national prospective study with repeated measurements of PA over the lifetime of participants, we identi ed ve distinct long-term PA trajectory groups in men and two distinct PA trajectory groups in women. In men, we found that the trajectory group labeled as medium-heavy PA in early adulthood (at 18-37 years) followed by persistent decline (at 38-70 years) was signi cantly associated with incident hypertension risk in later life.
At present, the research on PA trajectory mainly focused on CVD and physical functioning; for example, one study showed that a 20-year PA trajectory (moderately increase in PA level from middle age to old age) was associated with a decreased risk of mortality and CVD in later life, with an observed doseresponse relationship, and that maintaining even a slight PA was helpful [12]. Another study demonstrated that compared with women in the low PA groups, those in the middle and highest PA groups had more than 5% better physical functioning performance in later life [15]. However, the above ndings and our ndings cannot be compared directly owing to the differences in study populations, study design, sample size, methodology, and follow-up time. In addition, the PA calculation in those studies only included sport/exercise and excluded domestic, travel, leisure, and sedentary activities. In contrast, our study incorporated a comprehensive calculation of PA score, and complemented the current evidence of association between PA trajectory and incident hypertension.
Among men, the majority of participants belonged to group 1 (light PA and slight decline in PA from 18 to 70 years of age), indicating the high prevalence of a light, persistently stable PA trajectory in adulthood.
The national representative sample of our study suggests that great effort should be dedicated to promote the PA activities. The Healthy China 2030 advocates several strategies to promote the adoption of PA, such as formulating and implementing extensive national tness campaigns; strengthening the integration of physical and medical (publishing sports and tness activity guidelines); and formulating and implementing physical health intervention plans for special groups (adolescents, women, the elderly, and disabled people).
Notably, group 4, labeled as medium-heavy PA (at 18-37 years) followed by persistent decline (at 38-70 years), accounted for 9.22% of the study sample, and was signi cantly related to decreased risk of incident hypertension. Previous studies have shown that medium-heavy PA in early adulthood is associated with declined hypertension risk [28]. In addition, compared with the reference group, the participants in group 4 were younger; had lower levels of FPG, TG, and UA; had higher levels of HDL-C; had lower fat intake and more protein intake; and included fewer smokers and drinkers. All of these factors were positively correlated to the occurrence of hypertension, with their lower levels contributing to reduced hypertension risk [29]. In this study, we emphasize the importance of maintaining medium-heavy PA in early adulthood, especially for men.
In addition, no signi cant association was observed between the identi ed trajectory groups and incident hypertension in women. Consistent with our ndings, prospective studies also reported non-signi cant correlation between PA and hypertension risk in women [30,31]. Both the intensity of PA and the total amount of energy spent are lower in women than in men. In addition, compared with men, woman spend a greater proportion of time engaged in sedentary and light activities and less time engaged in more strenuous (moderate and intense) PA [32,33]. Furthermore, women's blood pressure is also in uenced by estrogen levels, menstrual cycle, and fertility [34]. Taken together, these ndings suggest that women should be the priority target for PA promotion. Our study has important public health implications. First, it is important to adhere to medium-heavy PA in early adulthood and maintain it over the life course. Second, women should be the prioritized target population for physical health interventions. Public health workers should be involved in distributing publicity materials related to PA (illustrations, small foldouts, desk calendars, CD-ROMs, etc.) and organizing publicity activities on PA lectures and health consultations. PA interventions can also be delivered through avenues such as social/familial events to enhance their effectiveness [35]. For example, people could be encouraged to establish an exercise group within the family or sign exercise contracts with each other to complete a certain amount of PA.

Strengths
To the best of our knowledge, this is the rst study to identify the long-term PA trajectories in a representative national sample of Chinese adults and to investigate the effect of PA trajectory on incident hypertension risk. The strengths of our study also include the large sample size, the availability of repeated measures of PA over time, and the use of GBTM. GBTM is a powerful statistical tool that applies limited hybrid modeling and maximum likelihood estimation to determine different PA trajectories [36]. Furthermore, not only occupational but also domestic, travel, leisure, and sedentary activities were included in the calculation of total PA scores in our study.

Limitations
It should be noted that this study has some limitations. PA was self-reported rather than objectively measured; thus, recall bias cannot be ruled out. In addition, the CHNS only includes the Chinese Han population, so the ndings may not be generalizable to other populations. Next, our study is limited by a short follow-up time (average of 5.4 years). At the end of the follow-up period, some young participants may not have developed hypertension. However, the similar results obtained in sensitivity analysis by excluding participants who developed hypertension during the rst 2 years of follow-up are evidence for the reliability of our ndings.

Conclusions
In conclusion, we identi ed ve distinct long-term PA trajectories in men and two distinct PA trajectories in women. The PA trajectory of medium-heavy PA in early adulthood (at 18-37 years) followed by persistent decline (at 38-70 years) was signi cantly associated with a decreased risk of hypertension in later life in men. Our study emphasizes the preventive effects of medium-heavy PA in early adulthood against incident hypertension in later life, highlighting that medium-heavy PA should be advocated in early adulthood and should be maintained throughout adulthood.  Figure 1 Trajectories of physical activity in men (a) and women (b)

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.