Journal Logo

EPIDEMIOLOGY

Domain-Specific Physical Activity, Pain Interference, and Muscle Pain after Activity

SWAIN, CHRISTOPHER T. V.1; BASSETT, JULIE K.1; HODGE, ALLISON M.1,2; BRUINSMA, FIONA J.1; MAHMOOD, SHAHID2; JAYASEKARA, HARINDRA1,3,4,5; MACINNIS, ROBERT J.1,2; GILES, GRAHAM G.1,2,6; MILNE, ROGER L.1,2,6; ENGLISH, DALLAS R.1,2; LYNCH, BRIGID M.1,2,7

Author Information
Medicine & Science in Sports & Exercise: October 2020 - Volume 52 - Issue 10 - p 2145-2151
doi: 10.1249/MSS.0000000000002358

Abstract

Pain conditions are common and disabling (1,2). They are among the leading contributors to the global disease burden and generate substantial health care cost (2,3). They are also difficult to manage, with few established treatment options offering consistent benefit with minimal side effects (4). Regular physical activity may be effective in preventing or delaying the onset of persistent pain, as well as reducing pain intensity or the disability caused by pain (5–8). There has been considerable interest in physical activity as a tool for the management of pain (9), but associations between physical activity and pain are not simple. There is large variability in pain responses to physical activity interventions (5,7), and in some contexts, physical activity can also trigger or exacerbate pain (10,11).

The context in which physical activity is performed is one factor that may contribute to the variation in physical activity–pain associations (8). Physical activity can be done in occupational, household, transport, and leisure domains (12,13). Each domain differs in terms of the correlates of activity (14,15), the typical movement demands and patterns of activity (16), the physiological and biological responses to the activity (17,18), and association with health outcomes (19–21). As these factors may also affect the development or experience of pain, uniform associations between physical activity domains and pain do not seem likely. Despite this, the total volume of physical activity achieved is often considered more important than the setting within which it is performed (22), and guidelines that promote the analgesic potential of physical activity typically remain focused on achieving the minimum quantities of physical activity required for benefit, regardless of the context (6,16).

To date, evidence regarding physical activity and pain has been derived from studies examining leisure-time physical activity (5), or associations between physically demanding occupations and specific pain conditions, such as low back pain (23). That leisure-time physical activity has been associated with reduced risk of persistent pain (5), whereas physically demanding occupations are associated with increased risk of pain and disability (23), suggests differing associations for different activity domains. For most middle-age and older adults, leisure-time physical activity represents only a small portion of total physical activity (13); physical activity in the transport and household domains comprises a large proportion of their total activity. Our purpose, therefore, was to examine associations between domain specific physical activity and the two pain outcomes, pain interference with normal work and muscle pain after activity.

METHODS

The Melbourne Collaborative Cohort Study

The Melbourne Collaborative Cohort Study (MCCS) is a prospective cohort study designed to investigate relationships between sociodemographic factors, lifestyle patterns, diet, and the risk of developing cancer or other health outcomes. Details of the MCCS have been published previously (24). At baseline (1990–1994), 41,513 participants (24,469 female, 17,044 male) predominantly age between 40 and 69 yr were recruited. Southern European migrants were deliberately oversampled (approximately 25% of the cohort) to extend the range of dietary and lifestyle exposures. As baseline data on physical activity were not domain-specific and did not include information on duration of physical activity or its intensity, only cross-sectional data from wave 2 (2003–2007), which included 27,323 participants, were included in this analysis. The study protocol was approved by Cancer Council Victoria’s Human Research Ethics Committee, and all participants provided written informed consent before participation (25).

Participants

As occupational physical activity data were not available for participants who were not working, we stratified the sample into workers and nonworkers (mostly retired participants). Workers were participants who were of working age at wave 2 (≤65 yr old) and working or volunteering at questionnaire completion. Nonworker participants included participants of any age who were not currently working or volunteering. We excluded participants who did not report work/volunteer status (n = 340) as well as those age above the Australian retirement age (>65 yr) who indicated they were working or volunteering (n = 4627), as half of these participants did not provide occupational physical activity data and the extent to which individuals were representative of a working population was questionable. We further excluded participants who described their general health as poor in the Short-Form 12-Item Health Survey version 2.0 (SF-12) (26) (n = 498; physical activity data were not collected from them) and participants who had incomplete exposure (n = 2436) or outcome data (n = 1473). The final sample included 17,949 participants (Fig. 1).

FIGURE 1
FIGURE 1:
Participant selection for analyses.

Physical activity assessment

Physical activity was ascertained using the long-form International Physical Activity Questionnaire (IPAQ-long) (12), which was administered by trained interviewers during wave 2. Participants provided information on occupational, home (including household and garden work), transport, and leisure-time physical activity completed in the previous 3 months. For the household, transport, and leisure domains, data on the frequency, duration, and intensity of physical activity were collected. As per the IPAQ guidelines, we truncated time spent in transport and in leisure-time physical activity to 180 min·d−1 for any respondent who reported higher durations, resulting in a maximum of 21 h·wk−1 of total activity within each of these two domains. Metabolic equivalents of tasks (METs) within each of these three domains were calculated by multiplying hours per week of each activity by the intensity level assigned by the IPAQ (long form) guidelines for data processing and analysis. For occupational physical activity, the IPAQ questions were modified. In addition to the hours per week of paid or voluntary work, participants were also asked to select their usual occupational activity intensity level from an ordinal scale (“Mainly sitting,” “Mainly sitting with occasional walking and moving about to do tasks,” “Mainly on feet with some light carrying or lifting,” or “Hard physical effort, e.g., scrubbing floors, digging, heavy carrying or lifting”). We used the Compendium of Physical Activities to assign a METs value to the occupational activity intensity level nominated by participants (27). “Mainly sitting” was assigned a value of 1.5 METs; “Mainly sitting with occasional walking and moving about to do tasks” was assigned 1.87 METs (assuming 75% sitting at 1.5 and 25% on feet at 3.0 METs); “Mainly on feet with some light carrying or lifting” was assigned a MET value of 3.0; and “Hard physical effort” was assigned 6.5 METs (27).

Pain assessment

Pain outcomes included pain interference and muscle pain after activity. Pain interference was assessed using the SF-12 (26). Participants were asked “During the past 4 wk, how much did pain interfere with your normal work (including both work outside the home and housework)?” Responses included the following: not at all, a little bit, moderately, quite a bit, and extremely. This item has been used to predict long-term sickness absence due to musculoskeletal complaints (MSK) (28). Pain after muscle activity was assessed using the 12-item Somatic and Psychological Health Report (29), a screening tool used in primary care settings. Participants were asked “Over the past few weeks have you been troubled by muscle pain after activity?” Available responses included the following: none or a little of the time, some of the time, and most of the time. The validity and reliability of each tool have been previously reported for other populations (26,29).

Covariate assessment

A directed acyclic graph was used to graphically represent the structure of the hypothesized physical activity domain–pain association and to determine confounding variables to be included in the multivariable-adjusted models (see Figure, Supplemental Digital Content 1, which provides a directed acyclic graph used for the selection of confounders, http://links.lww.com/MSS/B965) (30). At baseline and at wave 2, participants completed a structured interview including country of birth, education, medical history, and lifestyle factors including smoking status and alcohol consumption. Body mass index (BMI) was calculated using weight measured to the nearest 0.1 kg using Tanita at wave 2 and height measured to the nearest half a centimeter by trained clinical staff using a stadiometer at baseline. Residential postcodes collected at wave 2 were used to assign participants to a quintile of socioeconomic status based on the Index of Relative Socio-Economic Advantage and Disadvantage obtained from Australian Bureau of Statistics census-based Socio-Economic Indexes for Areas. Participants were recorded as having a cardiovascular (CVD) condition/comorbidity if they reported a history of heart attack, heart bypass, angioplasty, stroke, angina, or diabetes. They were recorded as having an MSK condition/comorbidity if they reported a history of arthritis, hip or knee replacement, or bone fracture after the age of 50 yr. Feelings of depression were assessed using the SF-12 at wave 2.

Statistical analysis

Ordered logistic regression, which is an extension of the binary logistic model used when the dependent variable has more than two ordered categories (31), was used to assess associations between continuous exposure measures of domain-specific physical activity (MET·h·wk−1) and pain outcomes (categories), with a separate analysis performed for each exposure–outcome relationship. Adjusting for the covariates listed previously, we estimated adjusted odds ratios (OR) and 95% confidence intervals (CI).

Restricted cubic splines were used to graphically represent the shape of the associations between each physical activity domain and pain outcome (32). To test whether associations were nonlinear, linear models that included continuous measures of each domain were compared with restricted cubic splines with three knots using the likelihood ratio test (33). Knots were placed at the 10th, 50th, and 90th percentiles for occupational, household, and transport physical activity; at the 60th, 75th, and 90th percentiles for leisure physical activity in workers; and at the 70th, 80th, and 90th percentiles for leisure physical activity in nonworkers, to account for the large number of participants who reported no leisure-time physical activity. Reference levels were set at the minimum reported activity level for each domain. For nonlinear associations, the spline models were used to describe the associations, and when the association was determined to be linear, a simpler linear model was used.

To supplement the spline analysis, we analyzed associations between categories of domain-specific physical activity and pain outcomes. For occupational, household, and transport physical activity, participants were categorized into quartiles based on the total MET-hours per week reported. For leisure-time physical activity, participants reporting 0 MET·h·wk−1 were categorized as one group with the remaining participants then divided into three even groups. We report OR and 95% CI from models adjusted for age (years) and sex (male/female; model 1) and from models additionally adjusted for BMI (continuous), socioeconomic status (1st–5th quintile), highest level of education (primary school, some high school or technical school, completed high school or technical school, completed tertiary diploma or degree), CVD comorbidities (yes/no), MSK comorbidities (yes/no), smoking status (never, former, current), alcohol consumption (nondrinker, light drinker, heavy drinker), depression (none of the time, a little, some, most of the time, all of the time), the other physical activity domains (quartiles), and television viewing time (continuous; model 2). The Brant test was used to assess the proportional odds assumption and to ensure the models were valid (31).

Multiple imputation using chained equations was used to handle missing covariate data (34), assuming these data were missing at random, conditional on the variables in the imputation model. All variables from the analysis, as well as auxiliary variables that were highly correlated with variables that had missing data for >5% of participants, were included in the imputation equations. Binary variables (MSK comorbidities) were imputed using logistic regression models, ordinal variables (Socio-Economic Indexes For Areas, smoking status, alcohol consumption, depression) using ordered logistic regression models, and continuous variables (BMI, TV hours per day) using linear regression models. All continuous variables were normally distributed. We generated 15 imputed data sets for analyses of workers and nonworkers, which was based on the rule of thumb that the number of imputations should be approximately equal to the percentage of incomplete cases (which were 11% for workers and 15% for nonworkers) (34). Missing values were sampled and replaced with a set of plausible values randomly drawn from their predicted distribution based on the other observed variables, thus creating the complete data sets. We compared the estimates of association from models using the imputed missing data with the estimates from complete-case analyses. All statistical analyses were performed using Stata version 14.2 (Stata Corporation, College Station, TX).

RESULTS

Participant selection is displayed in Figure 1. The study sample consisted of 7281 working and 10,668 nonworking participants. Descriptive data are presented in Table 1 (see Table 2C, Supplemental Digital Content 2, which provides numbers for missing data and results of multiple imputation using chained equations, http://links.lww.com/MSS/B966). Compared with workers, nonworkers were older, were less educated, had lower socioeconomic status, drank less alcohol, were more likely to have CVD or MSK comorbidity, and were more likely to report pain interference or muscle pain after activity (Table 1). Close to half (49%) of workers and more than half (62%) of nonworkers did not report any leisure physical activity.

TABLE 1 - Participant characteristics.
Workers
(n = 7281) Mean (SD)
Nonworkers
(n = 10,668) Mean (SD)
Age (yr) 58 (4.1) 70 (7.5)
BMI (kg·m−2) 27 (4.5) 28 (4.8)
n (%) n (%)
Sex
 Female 4029 (55) 6700 (63)
 Male 3252 (45) 3968 (37)
Education
 Primary school 196 (2.7) 2176 (20)
 Some high school/technical school 2275 (31) 4728 (44)
 Completed high school/technical school 1833 (25) 2128 (20)
 Tertiary/diploma/degree 2977 (41) 1635 (15)
Socioeconomic Index for Areas
 1st Quintile 693 (10) 1886 (18)
 2nd Quintile 883 (13) 1802 (18)
 3rd Quintile 1179 (17) 2032 (20)
 4th Quintile 1747 (25) 2203 (21)
 5th Quintile 2455 (35) 2384 (23)
Full-time employment
 Working ≥38 h·wk−1 3659 (50)
Smoking status
 Never 4365 (60) 6573 (62)
 Former 2410 (33) 3584 (34)
 Current 506 (7.0) 511 (4.8)
Current alcohol intake
 Nondrinker (0 g·d−1) 1660 (23) 3883 (37)
 Light drinker (0–20 g·d−1) 3695 (51) 4603 (43)
 Heavy drinker (>20 g·d−1) 1879 (26) 2044 (19)
Comorbidities
 CVD comorbidities (yes) 627 (8.6) 2782 (26)
 MSK comorbidities (yes) 2592 (38) 6003 (63)
Felt depressed
 None of the time 3834 (53) 5289 (50)
 A little of the time 2242 (31) 2803 (26)
 Some of the time 1013 (14) 1936 (18)
 Most of the time 161 (2.2) 468 (4.4)
 All of the time 28 (0.4) 162 (1.5)
Pain interference
 Not at all 3936 (54) 4356 (41)
 A little bit 2175 (30) 3066 (29)
 Moderately 684 (9.4) 1845 (17)
 Quite a bit 408 (5.5) 1195 (11)
 Extremely 78 (1.1) 206 (1.9)
Muscle pain after activity
 None or some of the time 5966 (82) 7736 (73)
 A good part of the time 980 (14) 1985 (19)
 Most of the time 335 (4.6) 947 (8.9)
Data are presented as mean (SD) for continuous variables (age, BMI), n (%) for categorical variables (sex, education, socioeconomic index, full-time employment, smoking status, alcohol intake, comorbidities, depression, pain interference, and muscle pain after activity).

Associations between domain-specific physical activity and pain outcomes in workers were nonlinear (P < 0.01) and are presented Figure 2. Workers reporting more household activity were more likely to report more pain. Compared with those reporting the least activity, workers reporting the median level of household activity (16 MET·h·wk−1) reported more pain interference (OR, 1.19; 95% CI, 1.07–1.32) and muscle pain after activity (OR, 1.23; 95% CI, 1.06–1.42). In contrast, reporting the median level of occupational physical activity was associated with less pain interference (OR, 0.82; 95% CI, 0.68–0.99). Higher levels of occupational activity were not associated with less muscle pain after activity. Performing some transport physical activity was associated with less pain, with workers reporting the median level of activity (9.9 MET·h·wk−1), less pain interference (OR, 0.86; 95% CI, 0.77–0.97), and less muscle pain after activity (OR, 0.81; 95% CI, 0.70–0.95). Leisure physical activity was not associated with less pain interference, but some leisure activity was associated with less muscle pain after activity. The lowest odds were observed in workers reporting 20 MET·h·wk−1 (OR, 0.67; 95% CI, 0.56–0.80).

FIGURE 2
FIGURE 2:
Dose–response relations for physical activity domains with pain interference and muscle pain after activity in workers. The solid line represents the proportional OR for a one-unit increase in physical activity domain MET-hours per week on pain levels given that the other variables in the model are held constant, the two dashed lines represent the lower and upper 95% CI, and the dotted line represents an OR of 1.00. The dots represent the 25th, 50th, and 75th percentiles of physical activity (50th and 75th only for leisure physical activity). Median levels of occupational and transport activity were associated with less pain interference (A, C) and median levels of transport and leisure activity were associated with less muscle pain after activity (G, H). More household activity was associated with more pain (B, F). There was no clear association between leisure activity and pain interference (D) and only the highest levels of occupational activity were associated with more muscle pain after activity (E). Associations were adjusted for age, sex, BMI, socioeconomic status, education, CVD comorbidities, MSK comorbidities, smoking, alcohol consumption, depression, the other physical activity domains, and TV viewing time.

Associations between domain-specific physical activity and pain outcomes in nonworkers were nonlinear (P < 0.01) and are presented in Figure 3. Consistent with workers, reporting some transport activity was associated with less pain. Compared with those reporting the least activity, those reporting the median level of transport physical activity (9.9 MET·h·wk−1) reported less pain interference (OR, 0.88; 95% CI, 0.79–0.97) and muscle pain after activity (OR, 0.86; 95% CI, 0.77–0.95). More leisure-time physical activity was associated with less pain interference but not muscle pain after activity. Nonworkers reporting 20 MET·h·wk−1 reported less pain interference (OR, 0.87; 95% CI, 0.77–0.98). Unlike workers, only nonworkers reporting more than 102 MET·h·wk−1 of household activity reported more muscle pain after activity (OR, 1.16; 95% CI, 1.00–1.34).

FIGURE 3
FIGURE 3:
Dose–response relations for physical activity domains with pain interference and muscle pain after activity in nonworkers. The solid line represents the proportional OR for a one-unit increase in physical activity domain MET-hours per week on pain levels given that the other variables in the model are held constant, the two dashed lines represent the lower and upper 95% CI, and the dotted line represents an OR of 1.00. The dots represent the 25th, 50th, and 75th percentiles of physical activity (75th only for leisure physical activity). There was less pain interference in those reporting more transport and leisure activity (B, C), and less muscle pain in those reporting more transport activity (E). Those that reported the highest levels of household activity reported more muscle pain after activity (D). There were no clear associations between household activity and pain interference (A) or leisure activity and muscle pain after activity (F). Associations were adjusted for age, sex, BMI, socioeconomic status, education, CVD comorbidities, MSK comorbidities, smoking, alcohol consumption, depression, the other physical activity domains, and TV viewing time.

Categorical analyses identified similar associations with the cubic splines, except for occupational activity and pain limitation. Multiple imputation of covariates did not change these associations (see Table 2A, B, and D, Supplemental Digital Content 2, which provides categorical analysis for physical activity domains and pain outcomes using complete case as well as imputed data, http://links.lww.com/MSS/B966).

DISCUSSION

Associations between domain-specific physical activity and pain outcomes were not uniform. Transport and leisure-time physical activity were inversely associated with pain interference and muscle pain after activity, whereas increasing household physical activity was associated with increased pain interference and muscle pain after activity in workers and more muscle pain after activity in nonworkers. There were no categories of occupational physical activity that were associated with pain interference or muscle pain after activity; however, the restricted spline analysis indicated workers performing 45 MET·h·wk−1 of activity reported slightly less pain interference compared with the lowest levels of occupational activity.

Our study has several strengths, including the use of a physical activity measure that assessed the frequency, duration, and intensity of activity across multiple domains. Very few studies have assessed physical activity–health outcome associations beyond those of leisure-time physical activity or specific biomechanical work demands (16), and as a result, physical activity guidelines typically encourage achieving set quantities of activity regardless of context (16). Other strengths include the large sample of Australian adults, stratified by working status and assessing nonlinear associations. However, the findings should be interpreted in the context of several limitations. The cross-sectional design is unable to determine whether the physical activity accrued in different domains is a cause or an effect of pain levels. The use of self-reported physical activity data is also a limitation. Although the IPAQ-long has been previously validated (12) and is widely used to examine physical activity and health outcome associations, self-report data can contain systematic and random error and these may attenuate risk estimates (35,36). Furthermore, although this study used two distinct and valid pain outcome measures, pain is a symptom rather than a disease, and the presence of pain does not necessarily provide any information about the underlying factors that caused it. As such, the generalizability of these findings to all conditions associated with pain cannot be assumed.

Nonlinear associations were identified, indicating that the associations between domain-specific physical activity and pain outcomes were not the same across the entire exposure range. The results suggest that for individuals already performing higher levels of transport and leisure activity, increasing activity in these domains may not necessarily result in additional benefit. This is consistent with previous findings from both observational and intervention studies that demonstrate analgesic benefit with only modest levels of physical activity and that increasing activity further may not lead to additional improvement in pain states; rather, there is a suggestion that pain may get worse (8,37).

The nonlinear analysis in general produced results consistent with those from the categorical analysis, with one exception. Restricted cubic splines identified less pain interference for workers reporting 45 MET·h·wk−1 of occupational physical activity. This highlights a limit of categorization, which is commonly used to assess physical activity–pain relationships in observational studies (5). That is, some detail may be lost when all individual cases in a single category are treated as equal (38).

There are several possible explanations for the distinct associations between domain-specific physical activity and pain outcomes. Physical activity differs in each domain by movement type, intensity, and voluntariness (16,18), each of which may affect pain outcomes (7,8,23). For example, transport and leisure physical activity typically involve dynamic movements performed in short bursts that have established benefit for fitness (18), whereas household activity may involve awkward postures (e.g., vacuuming or gardening) is not necessarily voluntary, and may not be performed at an intensity that benefits fitness (16). The potential effect of pain must also be considered when interpreting the findings of this study. Individuals who are limited the most by pain may be less active in transport or leisure domains, thus reporting lower physical activity. Similarly, household activity may not be a direct cause of more pain, but those most affected by pain will accumulate more activity at home by virtue of the decreased time devoted to transport and leisure activities.

There has been considerable interest in the potential of physical activity as a means of preventing or treating pain, and available guidelines outline the value associated with different levels of activity. Although this study can only demonstrate association, because individuals who achieved modest levels of transport and leisure-time physical activity were less likely to report pain interference or muscle pain after activity, these domains should perhaps be encouraged above household or workplace activity. These findings are good for potential physical activity interventions that will naturally be directed to increasing transport and leisure activities more so than housework.

Melbourne Collaborative Cohort Study cohort recruitment was funded by VicHealth and Cancer Council Victoria. The Melbourne Collaborative Cohort Study was further augmented by Australian National Health and Medical Research Council grants 209057, 396414, and 1074383 and by infrastructure provided by Cancer Council Victoria. B. M. L. is supported by a fellowship from the Victorian Cancer Agency (MCRF18005). The funding bodies had no role in the design of the study, data collection, analysis, or interpretation; or writing of the manuscript.

The authors have no conflict of interest to declare. The results of this study have been presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. The results of the present study do not constitute endorsement by the American College of Sports Medicine (ACSM).

REFERENCES

1. Murphy KR, Han JL, Yang S, et al. Prevalence of specific types of pain diagnoses in a sample of United States adults. Pain Physician. 2017;20(2):E257–68.
2. Rice AS, Smith BH, Blyth FM. Pain and the global burden of disease. Pain. 2016;157(4):791–6.
3. Gaskin DJ, Richard P. The economic costs of pain in the United States. J Pain. 2012;13(8):715–24.
4. Califf RM, Woodcock J, Ostroff S. A proactive response to prescription opioid abuse. N Engl J Med. 2016;374(15):1480–5.
5. Shiri R, Falah-Hassani K. Does leisure time physical activity protect against low back pain? Systematic review and meta-analysis of 36 prospective cohort studies. Br J Sports Med. 2017;51(19):1410–8.
6. Kraus VB, Sprow K, Powell KE, et al. Effects of physical activity in knee and hip osteoarthritis: a systematic umbrella review. Med Sci Sports Exerc. 2019;51(6):1324–39.
7. Geneen LJ, Moore RA, Clarke C, Martin D, Colvin LA, Smith BH. Physical activity and exercise for chronic pain in adults: an overview of Cochrane reviews. Cochrane Database Syst Rev. 2017;4.
8. Pitcher MH, Tarum F, Rauf IZ, Low LA, Bushnell C. Modest amounts of voluntary exercise reduce pain- and stress-related outcomes in a rat model of persistent hind limb inflammation. J Pain. 2017;18(6):687–701.
9. Chimenti RL, Frey-Law LA, Sluka KA. A mechanism-based approach to physical therapist management of pain. Phys Ther. 2018;98(5):302–14.
10. Staud R, Robinson ME, Price DD. Isometric exercise has opposite effects on central pain mechanisms in fibromyalgia patients compared to normal controls. Pain. 2005;118(1–2):176–84.
11. Lima LV, Abner TSS, Sluka KA. Does exercise increase or decrease pain? Central mechanisms underlying these two phenomena. J Physiol. 2017;595(13):4141–50.
12. Hagstromer M, Oja P, Sjostrom M. The International Physical Activity Questionnaire (IPAQ): a study of concurrent and construct validity. Public Health Nutr. 2006;9(6):755–62.
13. Rombaldi AJ, Menezes AM, Azevedo MR, Hallal PC. Leisure-time physical activity: association with activity levels in other domains. J Phys Act Health. 2010;7(4):460–4.
14. Bauman AE, Reis RS, Sallis JF, Wells JC, Loos RJF, Martin BW. Correlates of physical activity: why are some people physically active and others not? Lancet. 2012;380(9838):258–71.
15. Chrisman M, Nothwehr F, Yang J, Oleson J. Perceived correlates of domain-specific physical activity in rural adults in the Midwest. J Rural Health. 2014;30(4):352–8.
16. de Souto Barreto P. Time to challenge public health guidelines on physical activity. Sports Med. 2015;45(6):769–73.
17. Autenrieth C, Schneider A, Doring A, et al. Association between different domains of physical activity and markers of inflammation. Med Sci Sports Exerc. 2009;41(9):1706–13.
18. Holtermann A, Krause N, van der Beek AJ, Straker L. The physical activity paradox: six reasons why occupational physical activity (OPA) does not confer the cardiovascular health benefits that leisure time physical activity does. Br J Sports Med. 2018;52(3):149–50.
19. Abu-Omar K, Rutten A. Relation of leisure time, occupational, domestic, and commuting physical activity to health indicators in Europe. Prev Med. 2008;47(3):319–23.
20. Mahmood S, English DR, MacInnis RJ, et al. Domain-specific physical activity and the risk of colorectal cancer: results from the Melbourne Collaborative Cohort Study. BMC Cancer. 2018;18(1):1063.
21. van der Waerden J, Nakamura A, Pryor L, Charles MA, El-Khoury F, Dargent-Molina P. Domain-specific physical activity and sedentary behavior during pregnancy and postpartum depression risk in the French EDEN and ELFE cohorts. Prev Med. 2019;121:33–9.
22. Pate RR, Pratt M, Blair SN, et al. Physical activity and public health. A recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA. 1995;273(5):402–7.
23. Swain CTV, Pan F, Owen PJ, Schmidt H, Belavy DL. No consensus on causality of spine postures or physical exposure and low back pain: a systematic review of systematic reviews. J Biomech. 2019;109312.
24. Milne RL, Fletcher AS, MacInnis RJ, et al. Cohort profile: the Melbourne Collaborative Cohort Study (Health 2020). Int J Epidemiol. 2017;46(6):1757-i.
25. Giles GG, English DR. The Melbourne Collaborative Cohort Study. IARC Sci Publ. 2002;156:69–70.
26. Ware JE, Kosinski M, Keller SD. A 12-item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity. Med Care. 1996;34(3):220–33.
27. Ainsworth BE, Haskell WL, Herrmann SD, et al. 2011 Compendium of Physical Activities: a second update of codes and MET values. Med Sci Sports Exerc. 2011;43(8):1575–81.
28. Roelen CA, Heymans MW, Thun E, et al. Predictive value of the SF-12 for sickness absence because of mental, musculoskeletal, and other somatic disorders. J Occup Environ Med. 2015;57(10):1113–8.
29. Berryman C, McAuley JH, Moseley LG. Sphere 12 screening questionnaire. J Physiother. 2012;58(4):273.
30. Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999;10(1):37–48.
31. Brant R. Assessing proportionality in the proportional odds model for ordinal logistic regression. Biometrics. 1990;46(4):1171–8.
32. Harrell FE Jr, Lee KL, Pollock BG. Regression models in clinical studies: determining relationships between predictors and response. J Natl Cancer Inst. 1988;80(15):1198–202.
33. Kirkwood BR, Sterne JAC. Essential Medical Statistics. 2nd ed. Malden (MA): Blackwell Science; 2003.
34. Nguyen CD, Carlin JB, Lee KJ. Model checking in multiple imputation: an overview and case study. Emerg Themes Epidemiol. 2017;14:8.
35. Matthews CE, Moore SC, George SM, Sampson J, Bowles HR. Improving self-reports of active and sedentary behaviors in large epidemiologic studies. Exerc Sport Sci Rev. 2012;40(3):118–26.
36. Shaw PA, Deffner V, Keogh RH, et al. Epidemiologic analyses with error-prone exposures: review of current practice and recommendations. Ann Epidemiol. 2018;28(11):821–8.
37. Heneweer H, Vanhees L, Picavet HS. Physical activity and low back pain: a U-shaped relation? Pain. 2009;143(1–2):21–5.
38. Schmidt CO, Ittermann T, Schulz A, Grabe HJ, Baumeister SE. Linear, nonlinear or categorical: how to treat complex associations in regression analyses? Polynomial transformations and fractional polynomials. Int J Public Health. 2013;58(1):157–60.
Keywords:

EPIDEMIOLOGY; COHORT; EXERCISE; PHYSICAL ACTIVITY; PAIN

Supplemental Digital Content

Copyright © 2020 by the American College of Sports Medicine