Increased Risk of Incident Knee Osteoarthritis in Those with Greater Work-Related Physical Activity

Abstract

Objective:

Occupations involving greater physical activity may increase risk for knee osteoarthritis (OA). Existing studies have not evaluated work-related physical activity before OA onset. Hence, we aimed to evaluate the association between work-related physical activity and knee OA incidence.

Methods:

We performed a person-based longitudinal study using Osteoarthritis Initiative (OAI) data among people who volunteered or worked for pay without baseline radiographic knee OA or knee pain. Bilateral knee radiographs were obtained at baseline and annual follow-ups. We defined radiographic OA as Kellgren-Lawrence grade ≥ 2. Questions from the Physical Activity Scale for the Elderly at baseline and annual OAI visits provided information about work-related physical activity level and hours. We performed logistic regression with work-related physical activity level (mainly sitting, standing and some walking, walking while handling some materials) and hours as predictors. The outcome was incident person-based radiographic OA within the ensuing 12 months, over 48 months.

Results:

Among 951 participants (2819 observations), higher work-related physical activity levels had greater adjusted odds ratios for incident radiographic OA (people with jobs with standing and some walking: 1.11 (0.60 – 2.08), and walking while handling some materials: 1.90 (1.03 – 3.52), when compared to those with mainly sitting work-related activity). There was no association between number of hours worked and incident radiographic OA.

Conclusions:

People performing work that require walking while handling some materials have greater odds of incident knee OA than those with jobs mostly involving sitting. Strategies are needed to mitigate risk factors predisposing them to radiographic OA.

INTRODUCTION

Osteoarthritis (OA) is one of the leading causes of chronic pain, disability, and reduced quality of life in the United States (US)[1–3]. An estimated 30.8 million US adults have OA [4, 5]. As the aging population increases, the incidence of OA will rise. Thus, OA is a substantial public health problem.

At present, few treatments exist for OA, including non-steroidal anti-inflammatory drugs, physical therapy, and ultimately arthroplasty. Thus, understanding and modifying risk factors are paramount in managing this disease. In a recent meta-analysis, the authors amalgamated data from case-control, cross-sectional studies, and cohort studies and suggested an association between physically labor-intensive occupations and an increased risk of knee OA[6]. They conjectured that the association might be related to the amount of knee bending that occurs with those occupations[6]. Based on this review, 6 cohort studies have assessed this question using longitudinal follow-up. However, only one of the cohort studies ascertained incident knee OA using radiographs, the most convincing way to define OA[7]. Unfortunately, in this study, the occupational assessment only evaluated whether a person participated in manual labor at age 30 or not, evaluating a time point 20 years earlier from when the outcomes were assessed[7]. Thus, high quality longitudinal studies to address this question are still lacking. With few modifiable risk factors that can decrease the risk for knee OA, a better understanding of the role of occupational work-related physical activity level and incident knee OA may have important public health implications.

Therefore, the objective of this study was to determine the association between occupational work-related physical activity level and the development of incident knee OA within the Osteoarthritis Initiative (OAI). The OAI is an observational longitudinal cohort within the US that has collected data on occupational work-related physical activity level and annual knee radiographs, an ideal setting to test this question.

MATERIALS AND METHODS

Study Design and Subjects

We performed a person-based longitudinal study nested within the OAI, a multi-center, prospective study of knee OA. The four clinical sites were Memorial Hospital of Rhode Island (Pawtucket, RI), the Ohio State University (Columbus, OH), University of Pittsburgh (Pittsburgh, PA), and University of Maryland / Johns Hopkins University (Baltimore, MD). At OAI enrollment (February 2004 to May 2006), participants had to meet the following criteria to be included:

General Inclusion Criteria:

• Men and women ages 45 – 79

• With, or at risk for, symptomatic femoral-tibial knee OA

• All ethnic minorities

Major Exclusion Criteria:

• Inflammatory arthritis (RA)

• Contraindication to 3T MRI

• Bilateral end-stage knee OA

For this study, participants without baseline radiographic knee OA or frequent knee pain in either knee who had reported working in the seven days prior to their OAI enrollment visit were included. We restricted the study to people who either volunteered or worked for pay. The last of the 48-month data was collected in 2010.

Assessment of Frequent Knee Pain

We defined the presence of frequent symptoms at the baseline visit using a knees-pecific frequent pain question: “During the past 12 months, have you had pain, aching, or stiffness in or around your right/left knee on most days for at least one month? By most days, we mean more than half the days of a month.”

Acquisition and Assessment of Radiographic Knee Osteoarthritis

Bilateral posterior-anterior semi-flexed knee radiographs[8] were obtained at the OAI baseline and annual clinic visits through 48 months. Participants stood with knees flexed 20–30 degrees with knees internally rotated 10 degrees. A SynaFlexer plexiglass frame positioner was used to fix and standardize the position of feet across follow-up visits[9]. There were 2 central readers, with one supervisor, who assessed OA severity using the Kellgren and Lawrence (KL) grade (0–4)[10]. Kappa coefficient for these measures ranged from 0.70–0.80 [11]. Radiographic OA was defined as KL ≥ 2.

Assessment of Work-Related Physical Activity Levels

Occupation-related information was acquired at baseline and annual follow-up visits using Question 10 of the Physical Activity Scale for the Elderly (PASE)[12]. PASE is a validated [13–15] questionnaire used to measure the occupational, household, and leisure physical activity level over one week for people aged 65 and older. As part of the questionnaire, participants were asked “During the past 7 days, did you work for pay or as a volunteer?” Only those answering affirmative to this question at the OAI baseline visit were included in this study.

Those who responded affirmative to the prior question answered follow-up questions regarding their work. The questions included, “How many hours per week did you work for pay and/or as a volunteer?” Participants provided a free text answer to this question. Participants were also asked to categorize their work-related physical activity into one of four levels. Each category described the type of physical activity involved and provided example occupations. Level 1: jobs requiring “mainly sitting with slight arm movements”; Level 2: jobs requiring “sitting or standing with some walking”; Level 3: jobs requiring “walking, with some handling of materials generally weighing less than 50 pounds” (22.68 kg); and Level 4: jobs requiring “walking and heavy manual work often requiring handling of materials weighing over 50 pounds” (22.68 kg). Because there were limited responses to those with Level 4 jobs, we combined Levels 3 and 4 into one group.

Covariates

Participants’ age and sex were recorded at baseline. Body mass index (BMI) in kg/m² was calculated at each clinic visit from the mean of two height and weight measurements. At the baseline visit, participants answered, “Have you ever injured your right/left knee badly enough to limit your ability to walk for at least two days?”. At each follow-up visit participants were asked, “Since your last annual visit to the OAI clinic about 12 months ago, have you injured your right/left knee badly enough to limit your ability to walk for at least two days?” to assess for the incidence of knee injury between each annual follow up visit.

A question that assessed education for participants at the baseline visits inquired about the highest grade or year of school completed. We divided the educational responses into three groups: (1) some high school or high school graduate, (2) some undergraduate school or undergraduate degree, and (3) some graduate school or graduate school degree.

Assessments of knee bending obtained through the OAI at baseline and each annual visit included three dichotomous (yes or no) questions with the same stem: “The following questions are about the physical activities you may have done on a typical day either at work or outside of work during the past 30 days. These activities do not have to be done all at once. They can be done over the course of the day.”

1. “Did you kneel for 30 minutes or more during a single day?” (kneeling),

2. “Did you squat for 30 minutes or more during a single day?” (squatting), and

3. “Did you get in and out of a squatting position 10 or more times during a single day?” (in-and-out of squatting).

The available answers for these three questions were: yes, no, don’t know, or refused. Those that answered yes were considered those who did perform the knee bending activity. For those who answered yes to these questions, there was one follow-up question to each of these questions that quantified the amount of exposure.

“About how often did you do this in a typical week?”

Level 1 – one day per week or less

Level 2 – 2–3 days per week

Level 3 – 4–5 days per week

Level 4 – nearly every day or every day

At the OAI baseline visit and subsequent three annual follow-up visits, we accounted for 11 other PASE domains. We included five PASE domains that had enough questions to estimate the average number of hours spent doing a respective activity during the prior seven days: 1) walking, 2) light sports or recreational activities, 3) moderate sports or recreational activities, 4) strenuous sport or recreational activities, and 5) muscle strength or endurance exercises. We classified these five PASE domains into tertiles. For the remaining questions, the stem to those questions was “During the past 7 days, have you done…” 1) light housework, 2) heavy housework, 3) home repairs, 4) lawn work, 5) outdoor gardening, and 6) care-giving. Thus, we had dichotomous variables for these six domains.

Statistical Approach

To report baseline characteristics, we grouped individuals based on the highest level of work-related physical activity (levels 1–3) that someone reported at any visit (baseline to 36-months visit). We then performed chi-squared testing to assess differences in sex, education, history of injury in at least one knee before baseline, and knee bending status (i.e., kneeling, squatting, and in and out of squats) among work-related physical activity levels. We used analyses of variance to compare age and BMI among work-related physical activity levels.

We performed repeated-measures logistic regression analysis with the predictor variable being the work-related physical activity level (1–3) at each annual visit (baseline, 12-month, 24-month, and 36- month visit) and the outcome being a person-based definition of incident radiographic OA (KL ≥ 2) by the subsequent annual OAI visit (12-month, 24-month, 36-month, and 48-month visit). Generalized estimating equations (GEE) were used to adjust for correlations within person observations longitudinally[16]. Using this analytic strategy, each participant could have contributed a maximum of three observations periods to the study as long as they did not get censored.

The first observation period was between baseline and 12-month follow up. Each participant started at the baseline visit and was assessed for their work-related physical activity level (predictor/exposure). The outcome of interest was whether or not they developed radiographic OA by the 12-month visit. All those participants who had the outcome of radiographic OA at the 12-month visit were then censored from further analyses. All those participants who remained in the study were then assessed at the 12-month visit for their work-related physical activity level (predictor/exposure). The outcome of interest related to this predictor was whether or not they developed radiographic OA by the 24-month visit. This approach continued until the outcome was assessed at the 48-month visit.

Covariates in the adjusted model included baseline age (continuous variable), sex, baseline BMI (continuous variable), knee injury during the observation period of interest, baseline education, and the response to each of the knee bending questions during the observation period of interest (three dichotomous variables). The observation period’s baseline remaining 11 PASE domains’ data were also included in the model: 1) walking (tertiles), 2) light strenuous sport or recreational activities (tertiles), 3) moderate strenuous sport or recreational activities (tertiles), 4) strenuous sport or recreational activities (tertiles), 5) muscle strength or endurance exercises (tertiles), 6) light housework (dichomotous), 7) heavy housework (dichotomous), 8) home repairs (dichotomous), 9) lawn work (dichotomous), 10) outdoor gardening (dichotomous), and 11) care-giving dichotomous [12]. We also tested for possible interactions between work-related physical activity level with sex, BMI, injury, and squatting.

We repeated the above analyses with the number of hours worked as the predictor (< 20 hours, 20–39 hours, 40–59 hours, and ≥ 60 hours) instead of type of work performed. We categorized work hours into these groups as the variable was not normally distributed. These groups identified two groups working less than full time and two groups representing more than full-time work (i.e., more than half of all participants reported working between 35 and 45 hours). We also tested the relationship of work hours with incident OA, stratified by work-related physical activity level group.

All analyses were performed using SAS version 9.4. An alpha-error=0.05 was accepted.

RESULTS

Figure 1 is the flow diagram that indicates how we selected the 951 OAI participants who contributed 2819 total observations to the analysis. At baseline, 58.3% were female, mean age was 59.5 ± 8.9 years, and mean BMI was 26.9 ± 4.5 kg/m². In the lowest work-related physical activity level, 46% were women compared to the other work-related physical activity levels where most participants were female (Table 1). BMI was similar in all work-related physical activity level groups. There was no statistically significant difference in the rates of knee injury in the different work-related physical activity levels (Table 1). Those who performed the higher levels of work-related physical activity also did more kneeling and squatting (Table 1).

Figure 1.

Osteoarthritis Initiative participants included in the study

Table 1.

Baseline Characteristics of Participants by Highest Level of Work-Related Physical Activity at Any Time (n = 951)

Work-Related Physical Activity Level	Level 1 = Mainly sitting with slight arm movements (n = 153)	Level 2 = Sitting or standing with some walking (n = 427)	Level 3 and 4 = Walking, with some handling of materials (n = 371)	p-value
Participant demographics
Sex (% Female)	46%	61%	60%	0.005a
Education
Some HS/HS graduate	7%	8%	11%	0.03a
Some undergraduate/ undergraduate degree	45%	40%	47%
Some graduate/ graduate degree	48%	52%	41%
History of injury in at least one knee before baseline	1%	4%	4%	0.18a
Age (years) (mean ±SD)	59.4 ± 8.8	58.8 ± 8.9	60.3 ± 8.9	0.06b
Body Mass Index (kg/m2) (mean ±SD)	26.7 ± 4.2	27.0 ± 4.4	26.9 ± 4.7	0.70b
Knee bending assessments
Kneeling	28%	24%	38%	<0.0001a
≤ 1 day per week	19%	15%	17%
2–3 days per week	7%	6%	12%
4–5 days per week	1%	2%	3%
Nearly every day	1%	1%	5%
Squatting	13%	16%	28%	<0.0001a
≤ 1 day per week	10%	8%	10%
2–3 days per week	2%	5%	9%
4–5 days per week	0%	1%	5%
Nearly every day	1%	1%	4%
In and Out of Squats	31%	37%	51%	<0.0001a
≤ 1 day per week	7%	14%	11%
2–3 days per week	13%	13%	19%
4–5 days per week	6%	5%	11%
Nearly every day	6%	5%	11%

Note: We grouped individuals based on the highest level of work-related physical activity (levels 1–3) that someone reported at any visit (baseline to 36-months visit).

HS = high school, SD = standard deviation

^ap-values derived from chi-squared tests

^bp-values derived from analysis of variance

Higher work-related physical activity levels were associated with incident radiographic OA (Table 2). People who walked and handled some materials were about twice as likely to develop incident radiographic knee OA as those who mostly sat at work, reflected as an odds ratio of 1.90 (1.03 – 3.52, 95% confidence interval) (Table 2). There was no significant interaction between occupational work-related physical activity levels and “kneeling”, “squatting”, and “in-and-out of squatting” (p-values were 0.81, 0.18, and 0.13), when evaluating incident radiographic knee OA. There was also no interaction between work-related physical activity levels and sex, BMI, or injury (p-values were 0.50, 0.69, and 0.21).

Table 2.

Associations of Work-Related Physical Activity Levels With Incident Radiographic OA

n = 2819 observations from 951 people		Incidence of Radiographic OA	Unadjusted OR (95% CI)	Adjusted OR (95% CI)*
Work-related physical activity level	Level 1 = Mainly sitting with slight arm movements (e.g., office worker, watchmaker, bus driver)	17/867 (2.0%)	Referent	Referent
	Level 2 = Sitting or standing with some walking (e.g., cashier, light tool and machinery worker)	30/1233 (2.4%)	1.25 (0.69 – 2.26)	1.11 (0.60 – 2.08)
	Level 3 and 4= Walking, with some handling of materials (e.g., mailman, waiter/waitress, construction worker, lumberjack, stonemason, farm laborer)	30/719 (4.2%)	2.18 (1.19 – 3.96)	1.90 (1.03 – 3.52)

OA = osteoarthritis, OR = odds ratio, BMI = body mass index, PASE = Physical Activity Scale for the Elderly, CI = confidence interval

Bold = statistically significant associations with a p-value≤0.05

^*Adjusted for age (continuous variable), sex, knee injury, BMI (continuous variable), the remaining 11 domains of the PASE, kneeling activities, squatting activities, in-and-out of squatting activities, and education.

There was no significant association between the number of hours worked and incident radiographic OA (Table 3). When stratified by work-related physical activity level, the numbers of cases were too small to converge the data. We have provided the prevalence of radiographic OA in the Supplemental Table 1.

Table 3.

Associations of Number of Hours Worked With Incident Radiographic OA

n = 2819 observations from 951 people		Incidence of Radiographic OA	Unadjusted OR (95% CI)	Adjusted OR (95% CI)*
Hours worked	≤20 hours worked	27/1013 (2.7%)	Referent	Referent
	21–40 hours worked	30/1050 (2.9%)	1.07 (0.63 – 1.81)	1.02 (0.55 – 1.89)
	≥41 hours worked	20/756 (2.5%)	0.99 (0.55 – 1.79)	1.03 (0.50 – 2.12)

OA = osteoarthritis, OR = odds ratio, BMI = body mass index, PASE = Physical Activity Scale for the Elderly, CI = confidence interval

^*Adjusted for age (continuous variable), sex, knee injury, BMI (continuous variable), the 11 remaining domains of the PASE, kneeling activities, squatting activities, in-and-out of squatting activities, and education.

DISCUSSION

Individuals within the OAI who engaged in occupations involving greater work-related physical activity that required walking while handling some materials had a greater odds of incident knee OA in the ensuing 12 months than those who performed occupations that required more sitting. Although kneeling and squatting were associated with higher work-related physical activity levels, there were no significant interactions between these activities and incident knee OA. Additionally, adjustment for these activities did not materially change the association between work-related physical activity level and incident knee OA. These findings suggest that occupations requiring walking while handling some materials are a risk factor for incident radiographic knee OA, but that this was not driven by knee bending-related activities. Despite the relationship between work-related physical activity level and incident radiographic OA, there was no association between the number of hours worked and incident radiographic knee OA.

There have been many studies published on this topic, with 71 studies in a recent meta-analysis[6]. However, most of the existing studies are of low to moderate quality, none of which included longitudinal evaluation of this question with ascertainment of OA using radiographs[7]. Using the prior data, the meta-analysis concluded that occupations with greater physical burden were associated with more knee OA[6]. Though some studies have been described as longitudinal[17–21], in all instances, the occupational exposure did not necessarily predate the outcome of incident knee OA. For example, the study by Martin et al.[19], which reported an association between greater occupational physical demand and knee OA, assessed occupational exposure based on jobs held at ages 36, 43, and 53 years old, but OA was only assessed at age 53[19]. Hence, it is unclear if a person developed OA earlier in life, which could have influenced occupational exposure. Our study design is unique compared to prior published literature in that we only included those who did not have any evidence of radiographic OA or frequent knee pain prior to the baseline OAI visit and that we assessed knee OA one year after the occupational exposure.

In our study, the categorical designation of work-related physical activity predicted knee OA. A systematic review of the relationship between occupational loading tasks and knee OA concluded that there was limited evidence that occupational exposure to heavy lifting, kneeling or knee bending, and stair climbing are associated with knee OA[22]. Moreover, the updated meta-analysis by Wang et al. drew a similar conclusion[6]. In our study, we did not find any interactions between categories of work-related physical activity and knee-bending activities, evaluated simultaneously as our assessment of the occupational exposures. Additionally, adjustment for these knee-bending activities did not impact the point estimates of the association between work-related physical activity level and incident knee OA. Because of the simultaneous assessments of the knee bending activities to the occupational assessment, we had high confidence that if knee bending occurred at work, we would have captured it.

Therefore, increased knee bending is unlikely to explain our findings. This is not to say that knee bending is not a risk factor for knee OA, it just does not provide an explanation for why we are seeing an association between work-related physical activity level and incident knee OA in our study. One possibility for why knee-bending activities are not statistically significant confounders in this study is that our cohort was older and the group of people who are at risk for the association between knee bending and OA could be a younger group. Alternatively, another possibility is that the amount of knee bending occurring in our group is at a much lower level than what is required to increase the risk for OA. In a recent meta-analysis, investigators calculated that the incremental odds ratio per 5,000 hours of kneeling was 1.26, peaking at around 24,500 lifetime hours of kneeling at work [23]. This is generally considered a high exposure. Although it is impossible to calculate the lifetime hours of kneeling in our study, the information on the frequency of knee bending activities indicates that most of all the knee-bending activities were performed three or fewer days of the week in the prior seven days (Table 1). If we assumed that people performed a constant level of knee bending through the course of a year, during one year of our study, most people would not exceed 100 hours of knee bending activities. The exposure level of knee bending activity that is needed to increase the risk for knee OA is likely a much higher exposure that is not ever reached in this study population. Despite the low level of knee-bending activities, we still observed an association between work-related physical activity level and incident knee OA, highlighting the importance of other mechanisms in this population.

The association of greater work-related physical activity level and incident knee OA may be explained by the idea that people who walk and handle materials at work are doing something similar to “carrying extra weight” on their joints. When joints have to bear greater weight, even if it is a small amount of weight, this is potentially injurious to joints. Most of the participants whose work involved walking and handling some materials were lifting less than 50 pounds (22.68 kg). Hence, it may not need to be a hefty weight to increase someone’s risk for knee OA. This finding is comparable to past studies showing that small changes in body weight may influence a person’s risk of knee OA incidence or progression[24, 25]. Occupations requiring handling materials, even less than 50 pounds (22.68 kg), may be deleterious, especially if the person must take many steps per day. It may be beneficial to further explore how cumulative joint loading during work (e.g., accounting for weight of materials and steps/day) influences someone’s risk of knee OA. Furthermore, it may be advantageous for future studies to focus on specific occupations instead of grouping occupations because each has a unique combination of weight of materials and steps/day. For example, military service members are a known group that needs special attention because they have physically challenging work exposures unique to those jobs.

There are limitations to our study. Firstly, the PASE is designed to assess activity in those aged 65 and older, but this instrument shows good construct validity even among those <65 [26]. Additionally, the PASE only ascertains information about work performed in the prior seven days. However, to perform our study, we assumed that this question is a good estimate of occupational physical activity exposure over the ensuing year, which may not be correct. Questions about past and current job titles or other instruments to assess work-related physical activity levels may have been beneficial but were not acquired in this large observational study because it was not designed to address our study hypothesis. In addition, there is a risk of selection bias. We included people > 45 years old without radiographic OA or frequent knee pain. Hence, we potentially eliminated people with knee OA who had already stopped working. Therefore, we may be under-estimating the association between radiographic OA and work-related physical activity. Also, this study was conducted more than 15 years ago and it is unclear whether these associations may have been altered over the subsequent decade. Despite these limitations, our study has many strengths, including its large sample size and prospective study design. Furthermore, the quality of the outcome assessments is very high with standardized radiographs that were systematically read with centrally located readers. These features of the Osteoarthritis Initiative make it an unprecedented resource to study risk factors for knee OA.

In conclusion, a higher level of work-related physical activity that involves walking while handling some materials was associated with an increased risk of incident radiographic knee OA. From a public health perspective, there is a need to identify strategies to help mitigate the increased risk for radiographic knee OA in individuals engaged in heavy manual work. These may include strengthening exercises, injury prevention programs, controlling ergonomic hazards, knee braces, or weight loss [27–31]. Future large prospective studies of people in high occupational work-related physical activity levels to evaluate for specific work-related tasks could be used to develop and refine preventative measures.

Key Messages:

1. What is already known about this subject?

Case-control, cross-sectional studies, and cohort studies have suggested an association between physically labor-intensive occupations and an increased risk of knee OA, possibly attributable to the amount of knee bending that occurs with those occupations.

However, we lack high quality longitudinal studies to address this question where the exposure was known to occur before the outcome.

2. What are the new findings?

Occupations requiring walking while handling some materials are a risk factor for incident radiographic knee OA, but this was not driven by knee bending-related activities.

3. How might this impact policy or clinical practice in the foreseeable future?

This study demonstrates the need to identify strategies to help mitigate the increased risk for radiographic knee OA among individuals engaged in occupations involving walking while handling some materials.

Data Availability Statement

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Datasets from the main Osteoarthritis Initiative are available on a public, open access repository: https://nda.nih.gov/oai/