The association between physical activity with incident obesity, coronary heart disease, diabetes and hypertension in adults: a systematic review of longitudinal studies published after 2012

Background A growing body of studies that investigated the longitudinal association between physical activity (PA) and the outcome of incident obesity, coronary heart disease (CHD), diabetes and hypertension has become available in recent years. Thus, the purpose of this systematic review was to provide an update on the association between PA and onset of obesity, CHD, diabetes and hypertension in individuals aged ≥18 years who were free of the respective conditions at baseline. Methods We systematically searched OVID, Pubmed, and Web of Science databases for pertinent literature published between January of 2012 and February of 2019. To ensure that conclusions are based on high quality evidence, we only included longitudinal studies conducted in samples of ≥500 participants and with ≥5 years of follow-up. Result The search yielded 8929 records of which 26 were included in this review. Three studies were conducted on the outcome of incident obesity, eight on incident CHD, nine on incident diabetes, four on incident hypertension, one on the outcome of both diabetes and hypertension, and one on the outcome of CHD, diabetes and hypertension. Overall, there was an association between PA and lower risk of incident obesity, CHD and diabetes, but not hypertension. Higher levels or amount of PA were associated with a reduced risk of new onset of the respective diseases in 20 studies (77%). Whereas four studies reported an elevated risk of incidence of diseases with lower PA levels (15%). PA was not associated with incidence of diseases in two studies (8%). Conclusion Higher levels of PA are likely associated with a lower risk of becoming obese, develop CHD or diabetes. These findings replicate and strengthen conclusions from earlier reviews underlining the importance of promoting PA in adults. The associations between PA and incident hypertension were less consistent. More research, particularly using prospective cohort designs in large population-based samples, is needed to further untangle the association between PA and incident hypertension. Trail registration CRD42019124474 (PROSPERO Protocol registration). Date of registration in PROSPERO 27 February 2019.


Background
The World Health Organization identified noncommunicable diseases (NCDs), such as diabetes mellitus or cardiovascular diseases to be a major threat to economies and societies [1]. NCDs are implicated in 73% of all global deaths in 2017, with 28.8 million deaths attributed to risk factors like high blood pressure, high blood glucose, or high body mass index (BMI) [2]. Furthermore, NCDs are forecasted to account for 81% of all global deaths in 2040 [3].
NCDs usually develop over a long time period and may be impacted by an individual's health behaviors [4]. As such, many NCDs may be preventable by decreasing metabolic risk factors such as hypertension, overweight and obesity, or hyperglycemia, as well as by decreasing behavioral risk factors like tobacco or alcohol use, an unhealthy diet, and physical inactivity [1,5].
The current systematic review presents an update of a previously published review by our group [20], that examined the long-term effects of PA on type 2 diabetes mellitus, CHD, overweight/obesity and dementia by including studies published before 2012. Given the high significance of this topic and since we expected a substantial amount of relevant studies published after 2012, we provide an updated review of longitudinal studies on the association between PA and incident obesity, CHD, diabetes and hypertension over the past 7 years. Hypertension has been added to the current review as it is widely regarded as a major risk factor for several NCDs [21].

Methods
This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guideline [22]. The protocol was registered in the PROSPERO register of systematic reviews (CRD42019124474).

Search strategy
Pertinent articles published between January of 2012 and February of 2019 were searched in electronic databases (PubMed, Web of Science and EMBASE by OVID) by applying a combination of one or more of the following search terms: "longitudinal and/or long-term"; "physical activity/exercise"; "adult"; "overweight and/or obesity", "coronary heart disease and/or coronary artery disease and/or ischemic disease", "diabetes mellitus and/or diabetes type 2", "hypertension and/or blood pressure". Both titles and abstracts were searched. After identification of studies, their bibliographies were searched manually to identify additional relevant studies.

Study inclusion & exclusion criteria
We defined the following inclusion criteria for this systematic review: (1) Longitudinal, i.e. prospective cohort study design; (2) Studies reporting the association between PA and new onset of obesity, CHD, type 2 diabetes mellitus and/or hypertension; (3) Studies providing information on the assessment of PA (predictor variable). PA could be leisure-time/habitual PA, workrelated PA, transportation related PA, organized and unorganized PA, etc.; (4) Only studies with ≥5 years of follow-up were included to allow for a meaningful conclusion on the longitudinal association between PA and selected outcomes of interest; (5) Studies with males and females aged ≥18 years, that were free of the diseases of interest at baseline; (6) Studies with more than 500 participants were included, to improve the probability to capture a substantial amount of incidence cases; and (7) articles written in English.
Excluded from this review were (1) studies investigating the effect of a specific PA intervention, as well as (2) clinical trials, cross-sectional studies, systematic reviews and meta-analyses.

Screening & data extraction
All pertinent studies detected after searching the electronic databases were imported to a reference manager software (Citavi 6) and duplicates were removed. The study selection process was divided into three phases. Two independent reviewers (LC & JKR) screened the titles of the articles, followed by the abstracts, and finally the full-texts based on the inclusion criteria. All studies meeting the eligibility criteria were included in this review. Disagreement was resolved by consensus or by consulting a third author (CN). The following information was extracted by one reviewer (LC): first author's name, publication year, study design, study setting, sample size, follow-up time, participant characteristics (e.g. age, sex, BMI) at baseline, assessment and type of PA (e.g. type, duration, intensity), assessment and type of outcomes of interest (i.e. obesity, CHD, diabetes and hypertension), and main results/ findings of the study (e.g. hazard ratios, relative risk). Extracted data were verified by another author (JKR).

Quality assessment and risk of bias
The quality of included studies was evaluated independently by two authors (LC & JKR) using the 22-item Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement version 4 [23]. Similarly, potential risk of bias of each study included in this systematic review was assessed through the Tool to Assess Risk of Bias in Cohort Studies [24] by the same authors (LC & JKR). Any discrepancies between the two reviewers were resolved by discussion or by consulting a third reviewer (CN).

Results
Overall, we identified 8929 articles, of which 8903 articles were excluded as they did not meet the inclusion criteria as described above. The reader is referred to Fig. 1 for a flow chart summarizing the search process and number of studies at each step. Twenty-six studies were included in this review with a combined N of 1,145,298 participants, and follow-up times ranging between 5 and 34 years. Three articles examined the association between PA and incident obesity [25][26][27], eight studies examined the association between PA and incident CHD [28][29][30][31][32][33][34][35], nine studies examined the association between PA and incident diabetes [36][37][38][39][40][41][42][43][44], and four studies examined the association between PA and incident hypertension [45][46][47][48]. In addition, one study reported the association between both PA and diabetes as well as PA and hypertension [49], and another study reported the association between PA and CHD, diabetes, and hypertension [50].

Association between PA and obesity
The studies included in this review showed an overall association between higher PA and lower risk of incident obesity. Two out of three studies reported a reduced risk of becoming obese for individuals with high PA levels as compared to low PA [25,27]. One study showed an elevated risk (142%) of becoming obese for persons who were physically inactive [26]. The characteristics of the included studies are summarized in Table 1.

Association between PA and CHD
Overall, there was an association between higher levels or amount of PA and a decreased risk of incident CHD. Seven out of nine studies reported a reduced risk of new onset of CHD with increasing PA levels as compared to low or no PA [28,[32][33][34][35], whereas one study revealed an association between PA and decreased CHD risk only for vigorous intensity PA [29]. One study did not find a significant association for occupational PA and CHD risk [30]. Two out of nine studies examined the impact of change in PA levels over time as predictor variable and failed to detect a significant association with incident CHD [31,33]. Please refer to Table 2 for a summary of studies on PA and incident CHD.

Association between PA and diabetes
The studies included in this review provide evidence of an association between increasing PA levels and a decreased risk of incident diabetes. Nine out of 11 studies reported a gradual inverse association between increasing PA levels with up to high/vigorous-intensity and a decreased risk of incident diabetes [36-39, 40-42, 44, 50], whereas one study revealed an association only between moderate intensity PA and reduced diabetes risk [39]. Two out of 11 studies reported an increased risk of incident diabetes (179 and 145%, respectively) for participants engaging in low amount of leisure time PA as compared to the highly active reference group [43,49]. For a summary of included studies on the association between PA and diabetes please refer to Table 3.

Association between PA and hypertension
Overall, there was no consistent association between PA and incident hypertension. Three out of six studies reported a gradual inverse association between PA levels (running and walking, moderate and moderate-vigorous PA) and incident hypertension [46,48,50], whereas one study found an association only for a specific age group (51-60 years) [48]. Two out of six studies found no significant association between PA and incident hypertension [45,47]. One out of six studies reported an increased risk of incident hypertension (137%) for persons with low leisure time PA as compared to the highly active reference group [49]. Please refer to Table 4 for an overview of included studies.

Discussion
The purpose of this research was to review studies published after January of 2012 and up to February of 2019 that investigated the long-term association between PA and new onset of obesity, CHD, diabetes and hypertension. Overall, we observed an association between PA and a decreased risk of incident obesity, CHD and diabetes but not hypertension. This is in line with systematic reviews published by our group and others that also found beneficial associations of PA with overweight/obesity, CHD and diabetes [20,[51][52][53]. Furthermore, it is also consistent with our hypothesis derived from both interventional and observational studies [54][55][56][57].
The included studies that examined the association between PA and incident hypertension reported conflicting results. This is partly in accordance with two other meta-analyses [58,59]. One meta-analysis reported a   Total PA is associated with a reduced risk in incident diabetes: HR [95% CI] for MET level g reduction of incident hypertension by 6% for each 10 metabolic equivalent of task hours per week increment of leisure time PA [58]. However, another meta-analysis detected an inverse association for recreational PA and incident hypertension but not for occupational PA [59]. Of note, the causes of hypertension are multifactorial and the way they interact and ultimately contribute to the development of hypertension is unclear. Thus, potential mechanisms for prevention of hypertension through PA also remain unclear. Some studies included in this review also reported findings stratified by sex and body weight. For example, one study observed a gradual inverse association between frequency, duration and intensity of PA and risk of incident diabetes in males, but only a gradual inverse association between frequency and intensity of PA and risk of incident diabetes in females [37]. Additionally, another study reported that, while overall PA irrespective of body weight was not associated with the outcome of incident hypertension, obese males with high PA had a significantly lower risk of hypertension than obese males with low PA [47].
The quality of included studies was independently assessed by two reviewers and was rated as moderate to good, with scores ranging between 16 and 22 (total range: 0-22). This is not surprising since we only included studies published in or after 2012 that may already have followed quality guidelines on reporting findings of observational studies such as STROBE [23]. The potential risk of bias was rated moderate to poor and there were several concerns that warrant brief discussion: 1) All studies included in this review assessed PA through self-reported questionnaires which may be prone to recall bias. However, given the large sample sizes and since the baseline measurements of PA of many studies took place several years or even decades ago, objective measurement of PA might not have been feasible. There is good reason to believe that more longitudinal studies using novel objective techniques such as accelerometry (e.g. [60,61]) will become available in the near future. 2) The studies differed regarding the assessment of the outcomes of interest, e.g. some studies objectively measured blood glucose levels or blood pressure (e.g. [47,49]) whereas others relied on selfreported information by the study participants and/or medical chart review (e.g. [40,48]).
3) The studies differed in terms of adjustment for potential confounders and mediators which makes a comparison of findings between studies difficult. 4) Five studies were conducted only among males [30,31,38,40,44] and five studies were conducted only among females [28,41,43,46,48].
We did not investigate potential mechanisms underlying the associations between PA and incident obesity,  diabetes and CHD. However, it has previously been postulated that there are acute and chronic effects of PA on insulin resistance, which may account for improvements in insulin action and decreased blood glucose levels as a response to engagement in PA [62]. Additionally, PA impacts energy balance by increasing total energy expenditure, which in turn causes an energy deficit and may lead to lower body weight [63]. Stimulating responses in adipose and body tissues by PA may also influence total energy balance and body composition [63]. Furthermore, studies suggest that regular PA increases capillarization and may reverse endothelial alterations, which is a major risk factor for CHD [64]. The strengths of this review are the rigorous search and selection strategy following published guidelines and conducted by two reviewers. Also, both quality and potential risk of bias were assessed by two authors independently. We deliberately focused on studies published after January of 2012 in order to provide an update of a systematic review previously published by our group [20]. In addition, even though unintentional, our review included studies from various countries such as Sweden, Norway, UK, Greece, Mexico, China, Australia, Italy, US, the Netherlands, Germany, France, Spain and Denmark which may add to the generalizability of our observations. The major limitations of this review pertain to the relatively small number of included studies. This may be due to the fact that we only included studies with large sample size (N ≥ 500) and relatively long follow-up time of ≥5 years. However, we believe that these criteria ensure validity of our conclusions and a higher probability of generalizability of the study findings. In addition, a large body of research on PA and overweight/obesity published after 2012 focused on change in BMI or body weight over time. We opted to not include these studies in our review as we chose our outcome of interest to be incident obesity, and information on change in BMI or body weight over time is thus not sufficient. For instance, a person could be underweight at baseline and an increase in BMI or weight might actually reflect progression to a healthier body constitution. At the same time, we also acknowledge that particularly obesity and hypertension are conditions for which individuals can take action to improve, i.e., a person develops incident obesity but may be able to decrease body weight in order to progress back to overweight or normal weight. Furthermore, our search terms may have been too narrow or not comprehensive enough and there may be published studies that we were not able to identify. However, in order to compensate for this potential shortcoming, we also manually searched bibliographies of included studies. Finally, the studies differed with regard to the depth of investigating PA variables. As such, PA was only one of many predictors in some studies (e.g. [36,42]) and thus only one finding related to PA and the outcome of interest was reported. Whereas in other studies, the association between various PA parameters (e.g. type, intensity, frequency, duration) and the outcome of interest was examined (e.g. [30,47]). More research to untangle the association and potential underlying mechanisms between PA and the outcome of incident overweight, CHD, diabetes and hypertension is needed, preferably using prospective cohort studies with large sample sizes, long follow-up and objective measurement of both predictor variable (i.e. PA) and the outcomes of interest. In addition, metaanalytic approaches to address research questions pertaining the association between PA and various health outcomes are warranted.

Conclusion
Overall, this systematic review replicates, updates, and extends the growing body of research on the associations between PA and incident obesity, CHD and diabetes. No clear association between PA and reduced risk for hypertension was detected. This review emphasises the contribution of PA in the prevention of various chronic diseases. Reducing the risk of new onset of NCDs and thereby reducing the economic burden on health systems is of high importance to societies worldwide. Regional and global action plans and preventive strategies (e.g. [65]) should highlight the beneficial impact of regular PA and support national governments in the implementation of concrete actions towards achieving a higher engagement in PA among individuals across all ages.
Abbreviations BMI: Body mass index; CHD: Coronary heart disease; NCDs: Noncommunicable diseases; PA: Physical activity; PRISMA: Preferred Reporting Items for Systematic Reviews and Meta-Analysis