Associations between occupational physical activity and left ventricular structure and function over 25 years in CARDIA

Abstract

Aims

Leisure time physical activity (LTPA) confers cardiovascular health benefits, while occupational physical activity (OPA) may have paradoxically negative health associations. This study tested the explanatory hypothesis that unfavourable cardiac remodelling may result from chronic OPA-induced cardiovascular strain.

Methods and results

Longitudinal associations of OPA and left ventricular (LV) structure and function were examined in 1462 participants {50.0% female, 56.4% White, aged 30.4 ± 3.4 years at baseline [Year 5 exam (1990–91)]} from the Coronary Artery Risk Development in Young Adults study. Left ventricular structure and function were measured as LV mass (LVMi), end-diastolic volume (LVEDVi), end-systolic volume (LVESVi), ejection fraction (LVEF), stroke volume (LVSVi), and e/a-wave ratio (EA ratio) via echocardiography at baseline and 25 years later. Occupational physical activity was reported at seven exams during the study period as months/year with ‘vigorous job activities such as lifting, carrying, or digging’ for ≥5 h/week. The 25-year OPA patterns were categorized into three trajectories: no OPA (n = 770), medium OPA (n = 410), and high OPA (n = 282). Linear regression estimated associations between OPA trajectories and echocardiogram variables at follow-up after adjusting for baseline values, individual demographic/health characteristics, and LTPA. Twenty-five-year OPA exposure was not significantly associated with LVMi, LVEDVi, LVSVi, or EA ratio (P > 0.05). However, higher LVESVi (β = 1.84, P < 0.05) and lower LVEF (β = −1.94, P < 0.05) were observed at follow-up among those in the high- vs. no-OPA trajectories.

Conclusion

The paradoxically adverse association of OPA with cardiovascular health was partially supported by null or adverse associations between high OPA and echocardiogram outcomes. Confirmation is needed using more precise OPA measures.

Introduction

While leisure time physical activity (LTPA) is well established as a factor in promoting optimal cardiovascular health,^1–4 recent reports have suggested that occupational physical activity (OPA) may have opposing, adverse associations.^5–9 Specifically, greater exposure to moderate and high levels of self-reported OPA, such as continuous walking or lifting, has been associated with a higher risk of cardiovascular disease and all-cause mortality,^10,11 particularly among those with pre-existing coronary heart disease¹² or low cardiorespiratory fitness.⁶ The conflicting associations of LTPA and OPA have been labelled the ‘physical activity health paradox’. However, research interrogating the hypothesized biological pathways of this paradox is still limited.^8,12 Additionally, most available research measures OPA only once, at midlife, reports associations with distal mortality outcomes, and does not provide empirical evidence for possible explanatory mechanisms.

Several pathways have been proposed through which OPA could negatively impact long-term cardiovascular health. One potential pathway involves adverse cardiac remodelling resulting from chronic cardiovascular strain (e.g. increased ambulatory heart rate and blood pressure) caused by sustained, long-duration OPA exposure accumulated over many years.⁸ Acute (short-term) effects of this pathway have been explored in our previous research, which demonstrated elevated 24 h blood pressure and heart rate during workdays with high levels of OPA compared with non-workdays.^13,14 However, it remains unclear whether long-term exposure to OPA and the accompanying cardiovascular strain is associated with downstream adverse changes to left ventricular (LV) structure and function. If such alterations were associated with chronic exposure to OPA, then such data would support the biological plausibility of the paradoxical associations between OPA and cardiovascular health.

To explore this possibility, the current study examined longitudinal associations of OPA exposure with LV structure and function as measured via echocardiography over a 25-year period as part of the Coronary Artery Risk Development in Young Adults (CARDIA) study.

Methods

Study population

CARDIA is a prospective cohort study that enrolled Black and White men and women aged 18–30 years beginning in 1985–86 and continues today (10th examination in 2020–22) across four study sites (Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA). The study’s overarching aim is to examine cardiovascular disease risk development beginning in young adulthood. The current study used data across 25 years and seven examinations, starting with the exam at Year 5 [Year 5 exam (study baseline): 1990–91, Year 7 exam (study Year 2): 1992–93, Year 10 exam (study Year 5): 1995–96, Year 15 exam (study Year 10): 2000–01, Year 20 exam (study Year 15): 2005–06, Year 25 exam (study Year 20): 2010–11, and Year 30 exam (study Year 25 and final follow-up): 2015–16]. Self-reported OPA exposure was measured at all seven examinations, while the outcomes measured via echocardiography were measured only at baseline and the 25-year follow-up.

Of 5115 original CARDIA participants, 1 withdrew consent and 2011 were excluded because they did not attend both the baseline and follow-up examinations, where echocardiography was completed. Additionally, participants who had poor-quality echocardiograms (n = 510) or did not have valid body weight measurements (n = 15) were excluded. Lastly, those who did not report working full time at baseline and in study Year 2 were excluded (n = 1068) to confirm that participants were working after baseline (avoiding reverse causality). Those who stopped working after study Year 2 were included to limit the impact of the healthy worker effect.¹⁵ Finally, those without valid data for any of the six echocardiogram variables of interest were excluded (n = 48). The final sample characteristics (n = 1462) are summarized in Table 1. The analytic sample had slightly higher age (30.4 vs. 30.0 years), lower female representation (50.0 vs. 54.5%), lower body mass index (BMI) (25.7 vs. 26.2 m/kg²), more White participants (56.4 vs. 48.4%), and lower education level (44.6 vs. 50.8% with high school or less) compared with the complete CARDIA sample.

Table 1.

Sample characteristics by occupational physical activity trajectory^b

	Total (N = 1462)a	Zero OPA (N = 770)a	Medium OPA (N = 410)a	High OPA (N = 282)a	P-valuec
Age	30.4 ± 3.4	30.5 ± 3.3	30.0 ± 3.5	30.5 ± 3.4	0.032
Sex					<0.001
Female	731 (50.0%)	463 (60.1%)	192 (46.8%)	76 (27.0%)
Male	731 (50.0%)	307 (39.9%)	218 (53.2%)	206 (73.0%)
BMI	25.7 ± 5.0	25.4 ± 5.1	25.9 ± 5.3	26.2 ± 4.4	0.053
Race					<0.001
White	824 (56.4%)	474 (61.6%)	215 (52.4%)	135 (47.9%)
Black	638 (43.6%)	296 (38.4%)	195 (47.6%)	147 (52.1%)
Education					<0.001
High school or less	652 (44.6%)	251 (32.6%)	192 (46.8%)	209 (74.1%)
Associate or bachelor’s degree	590 (40.4%)	357 (46.4%)	172 (42.0%)	61 (21.6%)
Graduate or professional degree	208 (14.2%)	155 (20.1%)	44 (10.7%)	9 (3.2%)
Other or no answer	12 (0.8%)	7 (0.9%)	2 (0.5%)	3 (1.1%)
LTPA trajectory					<0.001
Low LTPA	1070 (73.2%)	593 (77.0%)	291 (71.0%)	186 (66.0%)
High LTPA	392 (26.8%)	177 (23.0%)	119 (29.0%)	96 (34.0%)
Treadmill GXT duration	581.2 ± 163.8	576.8 ± 171.9	575.5 ± 163.7	601.8 ± 137.7	0.092
Smoking pack-years	2.6 ± 5.4	1.8 ± 4.3	3.1 ± 5.8	4.0 ± 7.0	<0.001
Alcohol intake (mL/day)	9.6 ± 15.5	7.6 ± 12.1	10.2 ± 15.7	13.9 ± 21.3	<0.001
Year 7 diet score	63.7 ± 12.3	64.9 ± 11.9	63.1 ± 12.7	61.0 ± 12.2	<0.001
Systolic blood pressure (mmHg)	107.2 ± 10.7	106.1 ± 10.4	107.9 ± 10.9	109.3 ± 10.9	<0.001
Diastolic blood pressure (mmHg)	69.1 ± 9.3	68.9 ± 9.2	68.7 ± 9.1	70.5 ± 9.7	0.020
Taking blood pressure medications					0.490
No	1449 (99.1%)	763 (99.1%)	405 (98.8%)	281 (99.6%)
Yes	13 (0.9%)	7 (0.9%)	5 (1.2%)	1 (0.4%)
Resting heart rate	65.3 ± 10.2	65.8 ± 9.4	64.9 ± 10.9	64.6 ± 11.5	0.260
Diabetes	9 (0.6%)	4 (0.5%)	4 (1.0%)	1 (0.4%)	0.520
Centre					0.019
Birmingham	282 (19.3%)	132 (17.1%)	86 (21.0%)	64 (22.7%)
Chicago	362 (24.8%)	214 (27.8%)	87 (21.2%)	61 (21.6%)
Minnesota	334 (22.8%)	161 (20.9%)	98 (23.9%)	75 (26.6%)
Oakland	484 (33.1%)	263 (34.2%)	139 (33.9%)	82 (29.1%)
LVMi (g/m2)	78.3 ± 18.4	76.1 ± 18.0	79.6 ± 18.7	82.4 ± 18.5	<0.001
LVEDVi (mL/m2)	64.8 ± 12.7	63.7 ± 13.2	65.6 ± 12.8	66.6 ± 11.1	0.035
LVESVi (mL/m2)	23.7 ± 6.3	23.2 ± 6.6	24.0 ± 5.9	24.8 ± 5.7	0.022
LVEF (%)	63.0 ± 6.4	63.3 ± 7.0	63.0 ± 5.7	62.5 ± 5.7	0.440
LVSVi (mL/m2)	41.2 ± 9.3	40.7 ± 10.0	41.6 ± 9.0	41.8 ± 7.7	0.340
EA ratio	1.8 ± 0.5	1.8 ± 0.5	1.9 ± 0.5	1.8 ± 0.5	0.110
EA ratio < 1 at baseline	11 (0.8%)	5 (0.6%)	3 (0.7%)	3 (1.1%)	0.790
EA ratio < 1 at follow-up	439 (30.0%)	215 (27.9%)	122 (29.8%)	102 (36.2%)	0.035

OPA, occupational physical activity; BMI, body mass index; LTPA, leisure time physical activity; GXT, graded exercise test; LVMi, left ventricular mass index; LVEDVi, left ventricular end-diastolic volume index; LVEDVi, left ventricular end-systolic volume index; LVEF, left ventricular ejection fraction; LVSVi, left ventricular stroke volume index; EA ratio, E-wave/A-wave ratio.

^aData are presented as mean ± standard deviation or median (interquartile range) for continuous measures, and n (%) for categorical measures.

^bAll time-varying values presented are from the baseline examination.

^c P-values represent a comparison of the characteristic across the three OPA trajectory groups using analysis of variance or χ² tests as appropriate.

Written informed consent was obtained from each participant prior to the study assessment and each site gained approval for all protocols from their respective institutional review boards.

Occupational physical activity

The study’s independent variable, OPA, was measured at all seven exams using three questions from the validated CARDIA Physical Activity History questionnaire.^16,17 First, participants indicated (yes/no) if they did any ‘vigorous job activities such as lifting, carrying, or digging in the past 12 months for at least one-hour total time in any month’. If the participant answered yes, two follow-up questions were asked: (i) ‘How many months did you do any of these activities?’ and (ii) ‘How many of these months were for ≥5 hours/week?’ To capture meaningful OPA exposure, only months with OPA for ≥5 h/week were considered as ‘exposed to OPA’ for each exam. For example, if a person responded ‘yes’ to the first question (they performed OPA in the past year), ‘12’ to the second question (months with any OPA), and ‘7’ to the third question (months with OPA ≥5 h/week), they would be scored as having 7 months of OPA for that exam. Using the number of months with OPA from each of the seven exams, 25-year trajectories of OPA were computed using a group-based multi-trajectory model.¹⁸ To account for the large number of individuals with 0 months of OPA across all study years, a zero-inflated Poisson model was used to calculate the trajectories. The number and polynomial type of the trajectories was chosen based on model fit, where the Bayesian information criterion was examined for every combination of trajectory group number (1–3 groups) and trajectory group polynomial types (linear, quadratic, or cubic). Three trajectory groups reflecting dose were chosen: no OPA (n = 770, 52.6%), medium OPA (n = 410, 28.4%), and high OPA (n = 282, 19.0%), which had linear, quadratic, and quadratic shapes, respectively. The three OPA trajectory groups are displayed in Figure 1.

Figure 1.

Occupational physical activity trajectories over 25 years in CARDIA. Percentages displayed represent the percentage of the sample assigned to the trajectory. OPA, occupational physical activity.

Echocardiography

Two-dimensionally directed M-mode and Doppler echocardiograms were conducted at baseline and follow-up examinations across the four study sites following the recommendations from the American Society of Echocardiography.¹⁹ The full scanning protocols for both years can be accessed on the CARDIA study website²⁰ and were designed to be comparable across the study years. In brief, echocardiograms taken at baseline were measured using a standardized protocol on an Acuson cardiac ultrasound machine (Siemens Medical Solutions, Malvern, PA, USA) and recorded on super-VHS tape.²¹ At follow-up, echocardiograms were measured using the Artida cardiac ultrasound machine (Toshiba Medical Systems, Otawara, Japan) with a sector 30 BT transducer (fundamental frequency 2–5 MHz) to record and store in digital files. Measures from both exams were sent to the core reading centre (via videotape at baseline to the University of California and digitally at the follow-up to Johns Hopkins University). Echocardiograms deemed to be of poor quality by the reading centre were excluded from the analysis.

Left ventricular mass (LVMi, g), LV end-diastolic volume (LVEDVi, mL), LV end-systolic volume (LVESVi, mL), LV ejection fraction (LVEF, %), and LV stroke volume (LVSVi, mL) were captured using M-mode two-dimensional images. Left ventricular mass, LVEDVi, LVESVi, and LVSVi were indexed to the participant’s body surface area (m²) calculated using the Mosteller equation²² and are presented herein as follows: LVMi = left ventricular mass index (g/m²), LVEDVi = left ventricular end-diastolic volume index (mL/m²), LVESVi = left ventricular end-systolic volume index (mL/m²), and LVSVi = left ventricular stroke volume index (mL/m²). E-wave/A-wave ratio (EA ratio) was measured by Doppler echocardiogram and calculated as the mitral peak wave velocity in early diastole (cm/s) divided by the velocity in late diastole, where higher values indicate better diastolic function. All values were calculated using the American Society of Echocardiography recommendations at both exams.²³ All measurements included several quality control steps, including standardized training of technicians and readers, technician observation by a trained echocardiographer, blind duplication readings to establish interreader and intrareader measurement validity, periodic reader review sessions, phantom studies on the ultrasound equipment, and quality control audits.

Echocardiography variables are indexed to body size, as body size strongly influences left ventricular structure and function. Generally, the echocardiography variables are interpreted as follows. Left ventricular mass index is the mass of the left ventricle. It is reflective of the cumulative workload on the heart, where a higher mass is considered less healthy. Left ventricular end-diastolic volume index is the amount of blood in the left ventricle after filling (diastole). Although a higher value can be considered healthier, EDV tends to increase with age. Left ventricular end-systolic volume index is the amount of blood remaining in the left ventricle following contraction (systole), where a lower value is considered healthier. Left ventricular ejection fraction and LVSVi are the percentage and volume of blood, respectively, that is pumped out of the left ventricle with each heartbeat, and higher values are considered healthier. Left ventricular ejection fraction and LVSVi are indicators of systolic heart function. EA ratio is the ratio of the early (E) to late (A) ventricular peak filling velocities, which is considered a marker of left ventricular function. In a healthy heart, the E velocity is greater than the A velocity; therefore, a higher EA ratio is interpreted as indicating better diastolic function.²⁴

Because the availability of valid echocardiography data for each individual outcome was inconsistent, case-wise deletion was employed for the analysis, where only individuals with valid outcome data at baseline and follow-up were included in each respective model to maximize the number of individuals included (LVMi: n = 1272, LVEDVi: n = 697, LVESVi: n = 697, LVEF: n = 695, LVSVi: n = 620, EA ratio: n = 1422).

Covariates

Model covariates included study centre, age, sex, race, education level, smoking history, alcohol intake, diet, LTPA, blood pressure medications, resting systolic and diastolic blood pressure, diabetes status, and BMI measured during baseline and follow-up exams (if time-varying). Missing covariate values were replaced with values from the next nearest exam. Age was measured continuously in years. Sex was categorized as female or male. Race was self-reported throughout CARDIA as Black or White. Education level was self-reported as ≤ high school, associate or bachelor’s degree, graduate or professional degree, or other/no answer. Self-reported smoking history was summarized as accumulated pack-years for each exam (1 pack-year = 20 cigarettes smoked/day for 1 year). Alcohol intake was defined as self-reported average mL/day. Diet was assessed via an interviewer-administered CARDIA Diet History and summarized as an A Priori Diet Quality Score derived from 46 food groups.^25–27 Resting blood pressure was measured three times after a 5 min rest; an average of the second two measurements was used. Diabetes status was self-reported as yes/no. Body mass index (kg/m²) was calculated from height and weight measured to the nearest 0.5 cm and 0.09 kg, respectively.

Leisure time physical activity was measured using the validated CARDIA Physical Activity History questionnaire.^16,17 An LTPA score (in exercise units) was calculated based on a sum of the frequency, consistency, and number of months of participation across 12 physical activities minus the ‘vigorous job activity’ score.^16,28 Of note, due to the objectives of this analysis, the LTPA score did not include self-reported vigorous OPA, as was common practice in previous CARDIA analyses. Leisure time physical activity scores were summarized into two 25-year trajectories reflecting dose [low LTPA (n = 1070) and high LTPA (n = 392)] calculated using a censored normal model with linear shapes for both groups. Supplementary material online, Figure S1 displays the 25-year LTPA patterns for each trajectory.

Cardiorespiratory fitness (i.e. fitness) was measured at study Year 2 using a symptom-limited maximal graded exercise test following a Balke protocol, with nine increasingly difficult 2 min stages. Fitness was estimated in this study based on test duration (seconds). The test was considered valid and was included in the study if the participant reached ≥85% of their age-predicted maximal heart rate using the CARDIA formula after correcting for a protocol deviation at one site.^29,30 Due to its potential to be within the causal pathway of the associations of interest (OPA may be associated with lower fitness which is associated with detrimental cardiac structure and function³¹), fitness was not included as a covariate within the analytic models to limit the possibility for over-fitting. However, because of the potential importance of fitness in explaining the associations, effect modification by individual fitness levels was explored in stratified analyses. For this purpose, fitness was categorized into high and low fitness groups, which were split at the median test duration (551 s).

Analytic approach

The associations between OPA trajectory and the six echocardiogram outcomes were examined longitudinally across the 25-year follow-up period using linear regression. Each echocardiogram variable at 25-year follow-up was modelled as the outcome with adjustment for baseline values. Covariates were included across three progressively adjusted models: Model 1 adjusted for individual characteristics (centre, age, sex, race, education, smoking pack-years, and alcohol intake); Model 2 added adjustment for LTPA trajectory; and Model 3 added adjustment for individual health factors (blood pressure medications, resting systolic and diastolic blood pressure, resting heart rate, diabetes status, and BMI). All models adjusted for baseline values and the associated change for time-varying variables (follow-up minus baseline) for all covariates, including categorical variables (e.g. no change, changed to having disease).

Follow-up interaction and stratified analyses were then conducted in only the fully adjusted models across five relevant factors: sex (female/male), smoking history [never smokers (pack-years = 0)/current or former smokers (pack-years > 0)], education level (high school or less/greater than high school), LTPA trajectory, and fitness (low/high). While all outcome variables were analysed, only LVEF is presented in Figure 3, as it was the most clinically relevant in the primary analyses. However, stratified model results for all other outcomes are presented in Supplementary material online, Figures S2–S6. All analyses were completed in STATA v.17.0 with α = 0.05.

Figure 3.

Stratified longitudinal associations of occupational physical activity trajectories and left ventricular ejection fraction over 25 years in CARDIA. *P < 0.05, **P < 0.01. All beta coefficients (β) represent unit difference in the outcome of interest relative to the reference trajectory of those with no occupational physical activity. Error bars surrounding each point represent the 95% confidence interval. All models are fully adjusted (Model 3) which includes adjustment for centre, age, sex, race, education, smoking pack-years, alcohol intake, diet, leisure time physical activity, blood pressure medications, resting systolic blood pressure, resting diastolic blood pressure, resting heart rate, diabetes status, and body mass index. Analyses stratified by fitness level include only those with valid fitness data at baseline (n=1278). OPA, occupational physical activity; BMI, body mass index; LTPA, leisure time physical activity; BP, blood pressure; HR, heart rate; LVEF, left ventricular ejection fraction.

Results

Table 1 presents the sample characteristics overall and stratified by the three OPA trajectory groups. The total analytic sample (n = 1462) had an average baseline age of 30.4 ± 3.4 years and was 50.0% female and 56.4% White. Almost half of the participants (44.6%) had no more than a high school education. In general, the high-OPA group included more males and more Black participants, had less education, was more likely to smoke, consumed more alcohol, and had higher blood pressure compared with the no-OPA group.

Figure 2 presents the associations of OPA trajectories with echocardiogram outcomes. Left ventricular mass index, LVEDVi, LVSVi, and EA ratio at follow-up were not significantly different across the three OPA trajectory groups (P > 0.05 for all). Left ventricular end-systolic volume index at follow-up was significantly higher in the high-OPA group than in the no-OPA group in all models (β range = 1.84–2.31, P < 0.05 for all), while the medium-OPA group had similar LVESVi to the no-OPA group (P > 0.05 for all). In addition, LVEF was significantly lower in the high-OPA group than in the no-OPA group (β range = −2.20 to −1.94, P < 0.05 for all), whereas LVEF was again not different between the medium- and no-OPA groups (P > 0.05 for all). While EA ratio at follow-up was not significantly different across groups, non-significant associations suggest a lower EA ratio with higher OPA. This pattern is supported by unadjusted descriptive patterns presented in Table 1 where the high-OPA group had a higher frequency of EA ratio < 1, suggestive of diastolic dysfunction, at follow-up than the other two OPA groups.

Figure 2.

Longitudinal associations of occupational physical activity trajectories and left ventricular structure and function over 25 years in CARDIA. *P < 0.05, **P < 0.01. All beta coefficients (β) represent unit difference in the outcome of interest relative to the reference trajectory of those with no occupational physical activity. Model 1 adjusted for centre, age, sex, race, education, smoking pack-years, alcohol intake, and diet. Model 2 added adjustment for leisure time physical activity. Model 3 added adjustment for blood pressure medications, resting systolic blood pressure, resting diastolic blood pressure, resting heart rate, diabetes status, and body mass index. OPA, occupational physical activity; BMI, body mass index; LTPA, leisure time physical activity; BP, blood pressure; HR, heart rate; LVMi, left ventricular mass index; LVEDVi, left ventricular end-diastolic volume index; LVEDVi, left ventricular end-systolic volume index; LVEF, left ventricular ejection fraction; LVSVi, left ventricular stroke volume index; EA ratio, E-wave/A-wave ratio.

Figure 3 presents the longitudinal associations of OPA trajectory and LVEF stratified by five relevant factors (sex, smoking status, education, LTPA, and fitness) in the fully adjusted model only (Model 3). Although the interaction was not statistically significant (P = 0.271), males with high OPA had significantly lower LVEF than males with no OPA (β = −2.12, P < 0.01), whereas no differences were observed among females (P > 0.05). The OPA trajectory by smoking interaction was not statistically significant (P = 0.237). There was no statistically significant interaction between OPA and education level (P = 0.226). However, those with more than a high school education and high OPA had significantly lower LVEF at follow-up than those with no OPA (β = −2.55, P < 0.05). The interaction between OPA and LTPA was not statistically significant (P = 0.265). However, follow-up LVEF in the high-OPA group was lower than in the no-OPA group among individuals reporting low LTPA (β = −2.45, P < 0.05) but not high LTPA (β = −0.44, P > 0.05). Follow-up LVEF was not different between the medium- and no-OPA groups among either of the LTPA-stratified groups (P > 0.05 for all). Similarly, the fitness by OPA interaction was not statistically significant (P = 0.242). However, those with high but not low fitness in the high-OPA group had lower LVEF than those in the no-OPA group at comparable fitness levels (β = −2.04, P < 0.05). The adjusted predictions of ejection fraction by each level of OPA and the stratified groups are presented in Supplementary material online, Figures S7–S11.

Discussion

In this study, high as compared with no OPA exposure over 25 years was associated with lower LVEF and higher LVESVi, yet was not associated with LVMi, LVEDVi, LVSVi, or EA ratio. This analysis provides novel data interrogating potential mechanisms of the physical activity health paradox using longitudinal echocardiogram data from a well-defined US cohort. Overall, the results provide some preliminary support for a mechanism whereby chronic exposure to OPA may result in adverse changes in LVEF and LVESVi but not in other structural characteristics of the left ventricle. While not statistically significant, results were also suggestive of greater diastolic dysfunction in the high-OPA group (lower EA ratio and higher frequency of EA ratio < 1). Additionally, the associations between OPA and LVEF may be modified by sex, education level, LTPA level, and fitness.

Previous studies in CARDIA³² as well as other cohorts^33,34 have generally demonstrated protective longitudinal associations of LTPA on left ventricular structure and function. To date, however, only two cross-sectional studies have examined the relationship between OPA and left ventricular structure and function.^35,36 First, Nde et al.³⁵ compared echocardiographic outcomes similar to those of our current analysis (LVMi, left ventricular end-diastolic diameter, left ventricular end-systolic diameter, and EA ratio) between construction workers with heavy occupational lifting and a matched group of office workers (n = 50 for both groups).³⁵ Results of the Nde study indicated that construction workers had significantly larger LVMi and end-diastolic diameter, but no differences were observed in end-systolic diameter or EA ratio. The second study, conducted by Korshøj et al.³⁶ in 2022, examined cross-sectional data from 2511 participants in the Copenhagen City Heart Study. That study compared several echocardiographic outcomes, including some similar to those of our current study (LVMi, LVEF), between people with and without self-reported occupational lifting (walking, some handling of material, and heavy manual work combined).³⁶ The study found no differences in LVMi or LVEF between the groups. However, when examining linear associations, a positive association between occupational lifting and LVMi was observed in the total sample (β = 0.14, P = 0.03) as well as in the subgroup of normotensive participants (β = 0.20, P = 0.001).

The current longitudinal study findings provide stronger evidence for the biological plausibility of the physical activity health paradox than those of the previously published studies, which had matched and cross-sectional cohort designs. However, our results only partially align with those of these previous studies. For example, we observed no 25-year changes in LVMi in high- vs. no-OPA trajectory groups, whereas previous work has suggested a positive association between OPA and LVMi.^35,36 Additionally, we observed a significantly lower LVEF and higher LVESVi over the 25-year study follow-up in high-OPA compared with no-OPA trajectories—associations that were not supported by the previous, cross-sectional studies.^35,36 The discrepancies between our findings and those of previous reports are likely attributable to cross-sectional vs. longitudinal study designs, differences in comparison groups (i.e. the small samples of construction workers and office workers in Nde et al.³⁵ vs. our comparison of OPA trajectory groups across all workers), or differences in OPA definitions (multi-modality OPA in Korshøj et al.³⁶ vs. our definition including only ‘vigorous job activities’). More research with longitudinal follow-up, careful OPA measurement, thoughtful control of potential confounders, and sound comparison groups is needed to clarify these associations.

Additional comparisons can be made to the literature describing the cardiac structure and function of individuals engaging in large amounts of endurance sports. Aligning in part with our findings in those with high OPA, some previous studies have suggested that high volumes of endurance sports participation (e.g. running, rowing, or cycling) may result in an ‘athlete’s heart’, which is characterized by higher left and right ventricular volumes, increased left ventricle wall thickness and cardiac mass, and increased left atrial size.^37–39 Contrary to our findings, however, the structural changes commonly observed in the athlete heart, taken together, preserve ejection fraction.^37–39

Despite the differences with previous reports, our findings should provoke meaningful discussion and provide novel insight into the potential mechanisms responsible for the physical activity health paradox. Specifically, the observed 25-year reduction in LVEF and elevation in LVESVi in the high- vs. no-OPA trajectory groups supports a hypothesized pathway in which high amounts of OPA over time will result in impaired cardiac function. Importantly, LVEF and by relation, LVESVi, is commonly used and cited as a meaningful measure of systolic function with significant prognostic value in predicting adverse cardiovascular outcomes.⁴⁰ However, it must be noted that the mean differences we observed between the high- and no-OPA groups were small in magnitude. While a ∼2.1% greater reduction in LVEF and a ∼2.3 mL/m² greater increase in LVESVi on average across 25 years is in the detrimental direction and provides a potential explanation for the physical activity health paradox, the effects are unlikely to be of sufficient magnitude to completely explain the additional risk of mortality and cardiovascular disease that have been associated with OPA on a population level.⁴⁰ Therefore, the negative overall and cardiovascular health associations observed among individuals accumulating high amounts of OPA may result from a combination of factors, of which adverse cardiac structural changes may be one. Further research is warranted to interrogate this and other potential mechanistic pathways linking OPA and cardiovascular health.

In addition to our primary findings in the total sample, the stratified results provide some preliminary insight into social, behavioural, and physiological factors that may modify the relationship between OPA and left ventricular structure and/or function. Additionally, the stratified analyses address previous concerns of potential residual confounding by these factors, such as smoking.⁴¹ The negative associations between OPA and cardiovascular health have largely been observed in males only,^10,42 and our results align with this finding. Although unclear, this sex-specific associations may be related to differences in task structure, and therefore OPA exposure, between males and females. In addition, the interaction by smoking status was non-significant and the detrimental associations of high OPA on LVEF were observed equally in both smoking groups. This finding disagrees somewhat with a cross-sectional study using data from the National Health Interview Survey, which identified stronger associations between OPA and cardiovascular disease among non-smokers,¹¹ but both findings support the notion that the observed associations between OPA and health are not simply an artefact of residual confounding by smoking status as some have proposed.⁴¹ The education-stratified results suggest that the negative impacts of OPA on LVEF may be more apparent in those with higher education levels. This supports the notion that the associations are robust and not due to residual confounding by education level. Taken together, the collective evidence from these sensitivity analyses strengthens the argument that cumulative exposure to high levels of OPA may lead to reduced cardiovascular functioning.

The results stratified by LTPA level provide some preliminary support for the hypothesis that the adverse cardiovascular impacts of high OPA are limited to those with low levels of LTPA. This agrees with other studies¹¹ and is unsurprising given that LTPA is understood to have positive cardiovascular health effects. However, it does not support the proposed mechanisms for the physical activity health paradox whereby the deleterious health impacts of OPA are due to cardiovascular overload.⁸ Under the cardiovascular overload framework, greater LTPA accumulation would only further contribute to the acute cardiovascular load experienced by an individual due to OPA and would therefore be expected to increase risk. However, LTPA may catalyse positive fitness adaptations over time that could reduce cardiovascular strain from OPA. At the very least, this result signals that the interrelated pattern of OPA and LTPA accumulation is important to consider and warrants further study.

Lastly, the results stratified by fitness indicate negative impacts of OPA on LVEF only in those with high fitness. However, the coefficients across stratified groups are similar (β = −2.04 and β = −2.61) and the difference in statistical significance is due mainly to the larger confidence interval in the low fitness group. Therefore, it is unlikely that the differences observed by fitness level are clinically meaningful. Nevertheless, while some studies have suggested that low fitness may exaggerate the association between OPA and poor cardiovascular health,^8,43 the current study results do not support that notion.

The strengths of this study include the high-quality measurement of echocardiography outcomes and robust study design, including 25-year longitudinal follow-up; use of a moderately sized US cohort with comprehensive measurement of possible confounders and effect modifiers (allowing for sensitivity analyses) a sample with even distribution across biological sex, race, and socio-economic status groups; and examination across a large portion of a typical working career. Additionally, this study employed repeated OPA measurements across the follow-up period rather than using a single time point as has been done previously. Nevertheless, the study was limited most notably by the somewhat crude self-report measure of OPA (number of months with ≥5 h/week of OPA). Ideally, future studies should collect domain-specific activity using accelerometers and a time-use diary to limit recall and social desirability biases.⁴⁴ Additionally, measurement of LVEF using the M-mode method is limited by only measuring the basal segments of the left ventricle; however, this was the gold standard measurement technique during the 1990–91 baseline assessment and therefore was carried forward in follow-up to facilitate valid longitudinal comparisons. Lastly, this study is limited by selection bias related to the size of the analytic sample. While our total sample was moderately sized overall (n = 1462), the individual analytic samples for each outcome consisted of only a small portion of the larger CARDIA cohort due to exclusions, which stemmed mostly from incomplete or invalid echocardiogram data and could negatively affect the generalizability of our results.

Conclusions

While most echocardiogram outcomes studied were similar across the 25-year follow-up, we identified adverse changes in LVEF and LVESVi in high-OPA vs. no-OPA groups. These relationships may be more pronounced among men, those with greater than high school education levels, and those with low LTPA levels; however, further confirmation is needed. The results overall provide some support for an explanatory mechanism for the physical activity health paradox by which high amounts of OPA over extended periods may catalyse adverse cardiac structural changes due to high cardiovascular strain. Such a mechanism would reinforce the framework for biological plausibility of the paradoxical cardiovascular health associations observed with OPA and could support future policy or intervention strategies that optimize OPA requirements, task structure, or work patterns to promote better long-term cardiovascular health for workers.

Supplementary material

Supplementary material is available at European Journal of Preventive Cardiology.

Author contributions

T.D.Q., A.L., K.P.G., B.S., D.R.J., P.S., and B.B.G.: conception of the work and methods development. T.D.Q. and B.B.G.: data management and processing. T.D.Q. and B.B.G.: data analysis. T.D.Q.: writing of the original draft. All authors participated in the development and editing of the manuscript or providing critical revisions to important intellectual content and agree to be accountable for all aspects of the work.

Funding

This project was supported by the National Institute of General Medical Sciences (5U54GM104942-07). The Coronary Artery Risk Development in Young Adults Study (CARDIA) is conducted and supported by the National Heart Lung and Blood Institute in collaboration with the University of Alabama at Birmingham (HHSN268201800005I and HHSN268201800007I), Northwestern University (HHSN268201800003I), University of Minnesota (HHSN268201800006I), and Kaiser Foundation Research Institute (HHSN268201800004I). This manuscript has been reviewed by CARDIA for scientific content. Additional support for this work was provided by the CARDIA Fitness Study (R01 HL078972 to B.S.) and the CARDIA Activity and Heart Failure Study (R01 HL149796 to K.P.G.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Data availability

Information about the CARDIA study data can be found at https://www.cardia.dopm.uab.edu. Data can be made available upon reasonable request to the CARDIA coordinator centre.