Association between active commuting and low-grade inflammation: a population-based cross-sectional study

Abstract Background Prior studies suggest that physical activity lowers circulating C-reactive protein (CRP) levels. However, little is known about the association between regular active commuting, i.e. walking or cycling to work, and CRP concentrations. This study examines whether active commuting is associated with lower CRP. Methods We conducted a cross-sectional study using population-based FINRISK data from 1997, 2002, 2007 and 2012. Participants were working adults living in Finland (n = 6208; mean age = 44 years; 53.6% women). We used linear and additive models adjusted for potential confounders to analyze whether daily active commuting, defined as the time spent walking or cycling to work, was associated with lower high-sensitivity (hs-) CRP serum concentrations compared with passive commuting. Results We observed that daily active commuting for 45 min or more (vs. none) was associated with lower hs-CRP [% mean difference in the main model: −16.8%; 95% confidence interval (CI) −25.6% to −7.0%), and results were robust to adjustment for leisure-time and occupational physical activity, as well as diet. Similarly, active commuting for 15–29 min daily was associated with lower hs-CRP in the main model (−7.4; 95% CI −14.1 to −0.2), but the association attenuated to null after further adjustments. In subgroup analyses, associations were only observed for women. Conclusions Active commuting for at least 45 min a day was associated with lower levels of low-grade inflammation. Promoting active modes of transport may lead not only to reduced emissions from motorized traffic but also to population-level health benefits.


C
hronic low-grade inflammation is a mild and generally longlasting inflammatory state characterized by, but not limited to, a slight elevation in levels of circulating inflammatory biomarker Creactive protein (CRP). 1 Such elevated CRP levels are associated with major non-communicable diseases, such as cardiovascular disease, causing death and disability worldwide. 2RP has several roles in the immune response, including cytokine and phagocytosis regulation. 3It also has both pro-inflammatory and anti-inflammatory properties whenever tissue damage occurs.However, the long-term presence of slightly elevated circulating CRP levels may indicate a failure of the inflammatory response and lead to chronic low-grade inflammation. 4High-sensitivity CRP (hs-CRP) tests can help establish the presence or absence of low-grade inflammation, as they have the power to detect CRP plasma values below 10 mg/l.
Evidence is compelling on the potential of physical activity to reduce circulating CRP levels. 5Prior studies have suggested that higher exercise intensities and endurance training may lead to the largest benefits, especially in middle-aged or overweight adults. 6,7However, high-intensity interval training seems insufficient to reduce circulating CRP levels, at least in populations with metabolic disorders. 8t is also unclear whether active commuting, i.e. walking or cycling to work, alone entails a sufficient level of physical activity to lead to noticeable changes in CRP levels.0][11][12][13][14] At the same time, active commuting may also lead to higher exposure to traffic-related air pollution and noise compared with, e.g. car use. 15As air pollution and noise exposure have both been linked to an increase in CRP levels, [16][17][18] this may counteract the health benefits of physical activity.
This study examines whether active commuting is associated with lower hs-CRP, independently from other domains of physical activity and despite commute-related air pollution and noise exposure.We hypothesize that active commuters have lower hs-CRP levels compared with passive commuters.The topic is timely because a change from motorized traffic to sustainable means of commuting is urgently needed to mitigate climate change.

Study population
We used the FINRISK study, a population-based survey coordinated by the Finnish Institute for Health and Welfare, to assess chronic and non-communicable disease risk factors in the population living in Finland.The study has been described in detail elsewhere. 19The analysis was based on data from the cities of Helsinki and Vantaa (area 1) as well as Turku and Loimaa regions (area 2) for the years (n ¼ 2264).The eligible participants were aged 25-74 years and received a questionnaire by mail to fill in as well as an invitation to a health examination.The consent of all the participants has been obtained.Ethical permissions were granted from different ethics committees depending on the survey year, including from the ethical committee of the Finnish Institute of Health and Welfare and the Coordinating Ethics Committee for the Helsinki and Uusimaa Hospital District.Corresponding reference numbers were 38

Exposure variable
Information on active commuting was obtained by asking the participants how many minutes they walk, cycle or exercise to and from work daily.Possible answers were 'I only use motorized vehicles', 'less than 15 min', '15-29 min', '30-44 min', '45-59 min' and 'more than 60 min'.We used the first category corresponding to passive commuting as the reference group.Because of the low number of observations, we combined the two last categories into one category representing daily active commuting for 45 min or more.Since active commuting was expressed by the time spent engaging in it, we considered that it is not limited to its physical activity component, but also involves exposure to environmental stressors like noise and air pollution.Indeed, the time that a participant spends in physical activity to commute is the same time in which the participant is exposed to noise and air pollution during commuting.

Outcome variable
We used serum hs-CRP concentrations as an indicator of low-grade inflammation.The blood samples were drawn during the health examination and the blood analysis methods are described elsewhere. 12Because the distribution of the variable residuals was strongly skewed, we log-transformed hs-CRP values for the main analysis.We then used percentual mean difference in hs-CRP serum concentrations for the interpretation of the results.

Covariates
Based on the literature, plausibility, and data availability, we included age, sex, education level, household income, marital status, smoking status, alcohol consumption and exposure to environmental tobacco smoke as confounders in the main model.We also controlled for occupational physical activity in the main model.On the one hand, we hypothesized that it could be associated with active commuting.For instance, individuals engaged in physically active work might not be willing to include more physical activity during commuting.On the other hand, a possible association between occupational physical activity and CRP has been suggested. 20The variables included in the main model were self-reported (for variable descriptions, see Supplementary table S1).To reduce sampling variability and improve precision, we also took the study year/area and neighbourhood low-income rate into account.The latter corresponds to the percentage of inhabitants aged 18 years or older in the lowest-income quintile in their respective postal code area.We also considered other covariates, including medical history of cardiovascular diseases and diabetes, body mass index (BMI), and measurement season, in further analyses.These variables were available for every survey year.Furthermore, leisure-time physical activity and food intake (meat, fish, raw vegetables and fruits) were available for the years 1997, 2002 and 2007, but not for 2012, which is why they were not included in the main models.Exposure to road-traffic NO 2 was estimated by modelling outdoor concentrations at all residential addresses as described in Supplementary Methods S2.

Statistical analysis
We analyzed the data cross-sectionally by carrying out three linear regression models using the glm() function in R. 21 The basic model included age, sex, study year and area.The intermediate model additionally included education level, household income, smoking status and alcohol consumption.In addition to variables of the intermediate model, the main model included exposure to environmental tobacco smoke, marital status, neighbourhood-level low-income rate and occupational physical activity.We also ran the main model for men and women separately so that we could compare the results with a prior study using the same data. 12e then investigated potential multicollinearities between variables using the vif function in the car package. 22We also explored if the confounders used on a continuous scale had a linear relationship with hs-CRP using the gam function and smoothing terms in the mgcv package. 23We judged that age was not linear, but household income and BMI were.Therefore, we ran the main analyses and additional analyses with a smoothing term for age.In addition, we tested residuals for normality, and the effect of influential datapoints based on the hat matrix, which allows to convert observed values to fitted values using the least squares method. 24Related procedures are described in detail in Supplementary material (R code S3).All these diagnostic tests were run on the main model.
We carried out additional analyses where we separately added BMI, then history of cardiovascular disease and diabetes to the main model.We did not include a medical history of cardiovascular diseases and diabetes, or BMI, in the main analyses because they can both confound and mediate the studied association.We also conducted different sensitivity analyses on the main model.First, we adjusted the main model for additional lifestyle factors (i.e.leisuretime physical activity, and fruit, vegetable, meat, and fish intake).Second, we adjusted the model for residential exposure to roadtraffic NO 2 .Third, we adjusted for measurement season in the main model.Fourth, we only included participants whose serum hs-CRP levels were below 10 mg/l as higher values would indicate acute inflammation or infection.Finally, we excluded year 1997, which was the period with the highest levels of air pollution, to make the results more representative of the present-day air pollution levels. 25or all the analyses, we used R software version 4.2.1 and RStudio version 1.4.1106.0.Results are presented as percentage differences in mean hs-CRP serum levels between active commuting categories and their 95% confidence intervals.

Results
For the survey years and areas used in this study, the participation rate varied between 56% and 68% in men, and between 61% and 75% in women, totalling 10525 participants.For this study, we only included employed individuals (n ¼ 6763).From these, we further excluded participants with missing data on the exposure, associated covariates in the main model, or the outcome.The final analytical sample included 6208 participants (table 1).
The employed individuals who were excluded (n ¼ 555) were slightly older (mean age 46 vs. 44 years) and had a slightly higher BMI (mean BMI 27 vs. 26 kg/m 2 ) and hs-CRP (median 0.98 vs. 0.88 mg/l).They were to some extent less educated (21% vs. 24% in higher education) and less often married or living with a partner (65% vs. 70%).A greater share of them was smoking (32% vs. 28%), but they were to some extent consuming less alcohol (31% vs. 36% drinking more than six drinks weekly).
Compared with participants who were inactive in their leisure time (7.0%), 'very active' individuals (10.9%) belonged to the highest active commuting category (!45 min) more often.Similarly, people Active commuting and low-grade inflammation 293 engaging in active commuting for at least 45 min daily (11.7%) were more often 'very active' in their leisure time compared with passive commuters (8.5%).Moreover, 9.5% (n ¼ 317) of women engaged in active commuting for 45 min or more daily against 5.7% of men (n ¼ 163).Women were also to some extent more physically active in their leisure time, with 8.8% of women (n ¼ 293) and 7.6% of men (n ¼ 219) being 'very active'.In addition, the mean BMI in women was 26 kg/m 2 (n ¼ 3330) against 27 kg/m 2 in men (n ¼ 2878), and women had less cardiovascular disease history (26.4%, n ¼ 879) than men (42.0%, n ¼ 1208).Table 2 represents the main results for the associations between active commuting and hs-CRP.In all the models, except the model including men only, daily active commuting lasting at least 45 min was associated with lower serum hs-CRP concentrations compared with passive commuters.Removing the most influential datapoints with hat value >0.01 (n ¼ 47) did not affect the results.
Table 3 represents the results of the additional analyses in which the association for 45 min or more remained robust.

Discussion
We observed that Finnish adults engaging in high levels of daily active commuting to work (at least 45 min a day) had lower serum hs-CRP concentrations compared with passive commuters.Adjusting for key confounders, including occupational and leisure-time physical activity, diet, and residential exposure to road-traffic NO 2 , as well as additional factors such as BMI or history of cardiovascular disease and diabetes, did not substantially affect the results.No robust associations were observed for lower levels of active commuting.Further, in the subgroup analysis, the association between active commuting and hs-CRP was only observed for women.

Active commuting and low-grade inflammation 295
Active commuting is one way to be physically active, which can decrease hs-CRP.We controlled for both leisure-time and occupational physical activity in the analyses, which turned out to be important because people exercising in their free time were also more likely to use active transportation modes in this study.Similar findings have been previously reported. 26Furthermore, to our knowledge, only one prior study controlled for leisure-time physical activity. 11The cross-sectional study reported that individuals engaging in active commuting for <150 min weekly had significantly lower odds of high hs-CRP levels compared with passive commuters. 11While associations in our study were non-significant at this low level of active commuting, the effect estimates were in the same direction.Unlike in our study, higher levels of active commuting were not associated with hs-CRP concentrations in the abovementioned study.The differences could be explained by the relatively smaller sample size compared with our study.
Our findings align with those from some prior studies that considered the time spent engaging in active commuting.In a small Danish trial, physically inactive, healthy, overweight or obese participants were assigned to a 6-month bicycle commuting intervention.Women had to bike 9-15 km and men 11-17 km daily.These distances were calculated so that related energy expenditures would be in line with the physical activity guidelines (150 min of weekly moderate-intensity physical activity).The trial showed a 30% decrease in the cyclists' CRP levels during the 6-month cycling period. 9,27Furthermore, a Brazilian cross-sectional study found that low-grade inflammation tended to be lower in individuals engaging in active commuting for more than 150 min weekly compared with individuals who spend less time in active commuting. 10ome studies reported no association between the time spent in active commuting and CRP. 13,14Similarly, no associations were found in studies considering commuting physical activity as defined by metabolic equivalent of task. 20,28,29However, many of these studies found that total physical activity that comprised active commuting as one of the physical activity domains was associated with lower circulating CRP concentrations, 13,20,28 while other domains such as leisure-time physical activity were not necessarily associated with CRP when considered separately. 20,28n subgroup analyses, the association between daily active commuting and serum hs-CRP concentration was observed among women but not men.These results are in line with the earlier study using FINRISK data. 12One explanation could be that women in our study were in the highest active commuting category more often than men.They also tended to be more physically active in their leisure time, had lower BMI and fewer cardiovascular diseases.
This study examined the net effect of active commuting on hs-CRP, i.e. the combined effect of beneficial changes from physical activity and harmful changes due to environmental exposures.Indeed, commuters and especially active commuters may be exposed to high levels of traffic-related air pollution and noise during their journey. 15Even though active commuters can be less exposed in some study areas compared with passive commuters, the evidence shows that inhaled dose of air pollution might still be higher in active commuters. 30In both cases, this would favour a CRP increase because even short-term exposures to air pollution during commuting could lead to increased low-grade inflammation. 31Our results were robust after controlling for residential exposure to road-traffic NO 2 .We did the adjustment because residential NO 2 exposure has been observed to correlate with NO 2 exposure during commuting. 32,33Thus, active commuting for 45 min or more might have a beneficial, anti-inflammatory effect on Finnish adults, despite such exposures.Furthermore, active commuting for <45 min a day seems not to be detrimental regarding inflammation.These results are in line with a study where air pollution exposure did not seem to affect circulating CRP levels in cyclists. 34Moreover, health impact assessments also suggest that the net effect of active commuting is beneficial even with detrimental exposures. 35fferent biological mechanisms could explain the potential effect of regular physical activity in reducing low-grade systemic inflammation.For example, a direct reduction in fat mass could disfavour the secretion of pro-inflammatory cytokines, such as interleukin (IL)-6, tumour necrosis factor (TNF) and leptin from adipose tissue. 5,36It could also favour the secretion of anti-inflammatory cytokines, including adiponectin.In addition, IL-10 is known to increase after acute physical activity.This would increase insulin sensitivity and reduce sustained inflammatory state in adipose tissue by reducing macrophage infiltration.Regular exercise also shifts the balance of macrophage recruitment from pro-inflammatory to antiinflammatory phenotypes, which is probably due to a lowered level of tissue inflammation.Furthermore, depending on its intensity and duration, exercise induces the release of cortisol known for its antiinflammatory effects.It is also involved in the release of catecholamines that inhibit the secretion of pro-inflammatory TNF and IL-1b, for instance.Lastly, regular exercise improves endothelial function by reducing local inflammation. 5,36esides the cross-sectional design, the study has some other limitations.For example, we did not control for residential noise exposure even when it might be correlated with commute-related noise exposure (and associated with higher CRP). 17,18,37We were not able to adjust for greenspace exposure.Greenspace has been associated with lower circulating hs-CRP levels 38 and may also be associated with commuting behaviour.In one study, living near green areas was associated with lower levels of active commuting, perhaps because of urban sprawl and long commuting journeys. 39his makes it unclear whether it would have affected our results.In addition, we could not differentiate between walking and cycling to work; thus, it is unclear whether low-intensity walking alone would decrease CRP levels.Besides, serum hs-CRP was measured only once, while this biomarker has a significant intra-individual variability. 40Any subsequent misclassification is likely to bring the association between active commuting and hs-CRP towards the null.Consequently, we could have missed an association between active commuting for <45 min daily and hs-CRP.

Table 2
Association between active commuting and high-sensitivity C-reactive protein (hs-CRP) among Finnish residents who participated in the FINRISK study surveys of1997, 2002, 2007 and 2012Difference (%) in hs-CRP mean serum concentration (vs.reference group) with 95% confidence intervals