Bicycling over a Lifetime Is Associated with Less Symptomatic Knee Osteoarthritis: Data from the Osteoarthritis Initiative

Abstract

Introduction:

To evaluate the relationship between a history of bicycling and symptomatic and structural outcomes of knee osteoarthritis (OA), the most common form of arthritis.

Methods:

This was a retrospective, cross-sectional study within the Osteoarthritis Initiative (OAI), where we investigated OAI participants with complete data on bicycling, knee pain, and radiographic evidence of knee OA. We used a self-administered questionnaire at the 96-month OAI visit to identify participation in bicycling during four time periods throughout a participant’s lifetime (ages 12–18, 19–34, 35–49, and ≥50 years old). Using logistic regression, we evaluated the influence of prior bicycling status (any history, history for each time period, number of periods cycling) on three outcomes at the 48-month OAI visit: frequent knee pain, radiographic OA (ROA), and symptomatic radiographic OA (SOA), adjusting for age and gender.

Results:

2607 participants were included; 44.2% were male; mean age was 64.3 (SD 9.0) years; body mass index was 28.5 (SD 4.9) kg/m². The adjusted risk ratio for the outcome of frequent knee pain, ROA, and SOA among those who reported any history of bicycling compared to non-bicyclers was 0.83 (0.73–0.92), 0.91 (0.85–0.98), and 0.79 (0.68–0.90), respectively. We observed a dose-response among those who participated in bicycling during more time periods.

Conclusions:

People who participated in bicycling had a lower prevalence of frequent knee pain, ROA, and SOA. The benefit appeared cumulative. This study indicates that bicycling may be favorable to knee health and should be encouraged.

INTRODUCTION

Bicycling is a common activity considered aerobic and moderate to vigorous in intensity as a physical activity (1). Physical activity, such as bicycling, has been recommended by the World Health Organization for people of all ages because intense aerobic activity improves bone and functional health, cardiovascular and muscular health, and brain health and supports cancer prevention, fall prevention, healthy weight, and cognitive outcomes (2).

We were interested to know if participation in bicycling over a lifetime is associated with symptoms and structural damage of knee osteoarthritis. Because the natural history of osteoarthritis is very long, it is a difficult question to study. To date, there have not been any rigorous studies performed evaluating the long-term effects of bicycling on the symptoms and structure of the knee as it relates to osteoarthritis, the most common form of arthritis (3).

Thus, the purpose of this study was to evaluate the relationship between a history of bicycling and symptomatic and structural outcomes of knee osteoarthritis in the Osteoarthritis Initiative (OAI), a cohort with rigorous knee osteoarthritis outcomes of symptoms and structure ascertained systematically, recruited from the community not based on bicycling status.

METHODS

Study Design

We conducted a retrospective, cross-sectional study nested within the OAI. The OAI is a multi-center observational study of knee osteoarthritis including people ages 45 to 79 years old at the time of enrollment (2004 – 2006) who were in one of three groups: (1) had no knee osteoarthritis, (2) at high risk of developing symptomatic knee osteoarthritis, or 3) had existing symptomatic knee osteoarthritis. Participants were recruited at four clinical sites: Memorial Hospital of Rhode Island (Pawtucket, RI,) Ohio State University (Columbus, Ohio), University of Pittsburgh (Pittsburgh, PA), and University of Maryland / Johns Hopkins University (Baltimore, MD).

In this study, we evaluated OAI participants who completed a modified version of the historical physical activity survey instrument at the 96-month visit (the only time point that this instrument was deployed in the OAI), who also had knee-specific pain data and knee radiograph readings at the 48-month visit (the latest time point with the greatest number of readings and data points available) or at a visit proximate to that visit. Institutional Review Board approval was obtained at each participating OAI site and Baylor College of Medicine. All participants provided written informed consent.

Historical Physical Activity Survey Instrument

A self-administered modified version of the historical physical activity survey instrument (4) was sent to participants to complete and bring to their OAI 96-month visit. At the visit, if the survey was incomplete, participants were invited to finish the survey with help from the clinic staff. To incorporate this instrument within the OAI, the instrument was given as a self-administered questionnaire, which was a deviation from the original instrument (4), similar to what was done by others (5). Other modifications implemented to limit response burden included using ordinal categories for each of the frequency/duration selections and only commenting on the three most frequently participated activities in each age period.

In the questionnaire, participants were asked to review 37 leisure physical activities, including “bicycling (outdoor or individual stationary cycling) or spinning”. Participants were then asked to select all activities they performed at least 20 minutes within a given day at least ten times during four age periods: ages 12 – 18, 19 – 34, 35 – 49, and ≥ 50 years old. The participants then identified the three most frequently performed activities during those four age periods. For each period, they answered additional questions regarding the number of years, months per year, and times per month they engaged in their top three activities. Estimates for the number of times the participants engaged in an activity were based on those answers.

Knee Radiographs

We obtained weight-bearing, bilateral, fixed-flexion, posterior-anterior radiographs of knees at the 48-month visit, the most current OAI visit with the largest number of radiographic readings at the time of this analysis. Overall radiographic severity using Kellgren-Lawrence grades (0 – 4) based on the Osteoarthritis Research Society International Atlas (6) were scored by central readers (7). If 48-month visit readings were unavailable, readings from the most proximate timepoint available (baseline, 12-, 24-, or 36-month visits) were used instead. The reliability for these readings (read-reread) was good (8) (weighted kappa for intra-rater reliability = 0.71 (9)). In total, 2289 participants had radiographs from the 48-month visit, 209 (8%) from the 36-month visit, and 109 (4%) from the baseline visit.

Pain Assessment

At the 48-month visit, participants were asked, “During the last 12 months, have you had pain, aching, or stiffness in or around your right knee on most days for at least one month? By most days, we mean more than half the days of a month.” The same question was asked for the left knee. These questions were used to evaluate self-reported knee-specific pain. If the 48-month visit response was unavailable, the response from the most proximate prior in-person visit (baseline, 12-, 24-, 36-month visits) was used in its place. 99% of the frequent knee pain assessments were derived from the 48-month visit (13 were drawn from the 36-month visit, 4 from the 24-month visit, and 3 from the 12-month visit).

Participant Characteristics

Participant ages were calculated using the reported date of birth and date of the 48-month visit. Body mass index (BMI) was calculated as weight divided by height squared (kg/m²), measured at the 36-month OAI visit. This was the closest visit to the 48-month visit where both height and weight were measured. If the BMI was missing at the 36-month visit, the most proximate annual visit BMI was used as an alternative. History of knee injuries and total knee arthroplasties were self-reported at baseline and all annual visits up to the 48-month visit.

Statistical Analysis

We evaluated the association of bicycling with the prevalence of radiographic knee osteoarthritis (ROA), frequent knee pain, and symptomatic radiographic knee osteoarthritis (SOA) using logistic regression, but when the outcome is common, the odds ratios generated can overestimate the relative risk(10). Thus, a correction factor was applied to the logistic regression-generated odds ratios to provide relative risks(11).

Outcome Definitions.

All outcome definitions were person-based definitions. Radiographic osteoarthritis (ROA) was Kellgren and Lawrence ≥ 2 in at least one knee. Frequent knee pain was answering affirmative to the frequent knee pain question for at least one knee. Symptomatic radiographic osteoarthritis (SOA) was having at least one knee with both ROA and frequent knee pain in the same knee. Because we were interested in an assessment of ever having had knee osteoarthritis symptoms or radiographic evidence of knee osteoarthritis, those with a history of total knee arthroplasty were classified as having all three outcomes.

Bicyclers defined

For our initial analysis, we looked at the exposure of bicycling in two ways: (1) dichotomized those who were bicyclers versus non-bicyclers and (2) included four groups: non-bicyclers and three levels of bicycling (low, medium, and high based on tertiles of the number of times people bicycled among those who participated in the activity). We evaluated this over a lifetime and then for each of the four age periods. “Any history of bicycling” included people who participated in this activity in at least one age period. Bicycling at the different age periods (ages 12 – 18, 19 – 34, 35 – 49, and ≥ 50 years old), included people who participated in the activity at the respective specified age periods.

To evaluate the cumulative effect of bicycling over many age periods on the outcomes of frequent knee pain, ROA, and SOA, we created an exposure variable for bicycling based on the number of age periods when a participant identified bicycling as a top three activity. A significant value for the Cochran-Armitage trend test indicated a dose response.

We performed all analyses unadjusted and then adjusted for age and gender. We did not include BMI as a covariate in the primary model as we anticipate that this is likely a mediator of the relationship between bicycling and knee osteoarthritis symptoms and structure. All of the other 36 leisure physical activities evaluated in the study had a correlation with bicycling less than 0.2. Hence, it is unlikely that these activities explained the relationship between bicycling and the outcomes of interest, and therefore, these activities were not included in the model. Because the prevalence of the outcomes was common, which can lead to over-estimation of odds ratios, we corrected the odds ratios generated by logistic regression into relative risks(11).

We performed sensitivity analyses, which we included in the supplemental tables (see Supplemental Digital Content), where we additionally adjusted for BMI and prior injury. Also, to understand if arthroplasty drove our findings, we performed analyses where we excluded participants with an arthroplasty. Finally, to understand the role of race in our results, we restricted the analyses to Whites. We were unable to perform the analyses among underrepresented groups because participation in bicycling was insufficient to support those analyses.

All analyses were performed using SAS version 9.4. P-values <0.05 were considered statistically significant.

RESULTS

Of the 4796 participants enrolled in the OAI, 2607 participants were included. Figure 1 provides a flow diagram indicating the selection process used to identify the final group of participants included in this study.

Figure 1.

Flow Diagram Indicating the Selection Process for the Final Group of Participants Included in This Study.

Of the 2607 participants, 44.2% were male; mean age was 64.3 (SD 9.0) years; mean body mass index was 28.5 (SD 4.9) kg/m². Those who we excluded were a little older, had a mildly higher BMI, and had more frequent knee pain and ROA (Supplemental Table 1, Supplemental Digital Content, Characteristics of those not included in the study). Overall, of the people included in the study, 51.6% had a history of bicycling (Table 1); of those 52.4%, 25.2%, 12.2%, and 10.2% identified bicycling in one, two, three, and four of the time periods, respectively. Among bicyclers, the most common situation was that they either bicycled only during the youngest age group or participated in the activity during all four time periods (Supplemental Table 2, Supplemental Digital Content, Pattern of bicycling participation).

Table 1.

Characteristics of Those With No History of Bicycling, Any History Bicycling, and All Participants

Participant Characteristics	Non-Bicycler (n = 1263)	Bicycler (n =1344)	All Participants (n = 2607)
Age (years)	65.3 (9.1)	64.5 (8.8)	64.3 (9.0)
Male (%)	43.4%	45.0%	44.2%
BMI (kg/m 2 )	28.8 (4.9)	28.3 (4.9)	28.5 (4.9)
Race
Black (%)	24.8%	11.7%	18.0%
White (%)	72.7%	85.9%	79.5%
Knee-based characteristics*
Frequent knee symptoms (%)*	43.1%	35.9%	39.4%
ROA (%)*	61.2%	54.7%	57.8%
SOA (%)*	30.9%	24.3%	27.5%
Total knee arthroplasty (%)*	4.6%	3.6%	4.0%
Prior Injury (%)*	47.4%	50.4%	49.0%

BMI = body mass index, ROA = radiographic osteoarthritis, SOA = symptomatic osteoarthritis

The adjusted risk ratio for the outcome of frequent knee pain, among bicyclers compared to non-bicyclers was 0.83 (0.73–0.92) (Table 2), where bicyclers were 17% less likely to have frequent knee pain than non-bicyclers. For the outcomes of ROA and SOA, the risk ratios were 0.91 (0.85–0.98) and 0.79 (0.68–0.90), respectively (Table 2). The pattern of findings was similar across the outcomes (frequent knee pain, ROA, and SOA), and the findings were mostly seen among participants who bicycled at younger ages (Supplemental Tables 3–5, Supplemental Digital Content, Relative risk of prevalent frequent knee pain, ROA, and SOA compared to non-bicyclers, for bicyclers, and bicyclers divided into three levels of activity: low, middle, and high for the different age ranges).

Table 2.

Risk Ratios of Prevalent Knee Pain, Radiographic Knee OA (ROA) and Symptomatic Knee OA (SOA) in Non-Bicyclers (Referent) compared to Bicyclers (Dichotomous) and Bicyclers Divided into Three Levels of Activity: Low, Middle, and High.

Any History of Bicycling	Frequent Knee Pain	Unadjusted Risk Ratios	Adjusted Risk Ratiosa
Non-Bicyclers (n = 1263)	43.1%	Referent	Referent
Bicyclers (n = 1344)	35.9%	0.83(0.75–0.92)	0.83(0.73–0.92)
Low (n = 448)	35.0%	0.81(0.70–0.93)	0.80(0.69–0.92)
Middle (n = 435)	37.9%	0.88(0.77–1.01)	0.88(0.76–1.00)
High (n = 461)	34.7%	0.80(0.69–0.93)	0.78(0.67–0.91)
	ROA
Non-Bicyclers (n = 1263)	61.2%	Referent	Referent
Bicyclers (n = 1344)	54.7%	0.90(0.83–0.95)	0.91(0.85–0.98)
Low (n = 448)	56.0%	0.92(0.83–1.00)	0.95(0.85–1.03)
Middle (n = 435)	54.5%	0.89(0.80–0.98)	0.90(0.81–0.99)
High (n = 461)	53.6%	0.87(0.79–0.96)	0.90(0.81–0.98)
	SOA
Non-Bicyclers (n = 1263)	30.9%	Referent	Referent
Bicyclers (n = 1344)	24.3%	0.79(0.69–0.90)	0.79(0.68–0.90)
Low (n = 448)	24.5%	0.80(0.66–0.95)	0.80(0.66–0.96)
Middle (n = 435)	24.4%	0.79(0.65–0.95)	0.79(0.65–0.95)
High (n = 461)	24.1%	0.78(0.65–0.94)	0.77(0.64–0.93)

^aAdjusted for age and gender.

When evaluating the number of age periods that people engaged in bicycling in relation to the outcomes of interest, for each increase in the number of age periods engaged in bicycling, people had a lower risk ratio of each outcome (Table 3). For the outcome of SOA, for one, two, three, and four periods of bicycling, the respective risk ratios were 0.83 (0.71–0.96), 0.81 (0.66–0.99), 0.72 (0.52–0.96), and 0.57 (0.39–0.83), with a statistically significant p for trend of <0.0001.

Table 3.

Risk Ratios of Prevalent Frequent Knee Pain, Radiographic Osteoarthritis (ROA), and Symptomatic Osteoarthritis (SOA) Based on the Number of Periods of Life Bicycled, Compared to Never Bicyclers

Number of Age Ranges (Periods) When Participants Cycled	Prev. of Outcome	Unadjusted Risk Ratios	Adjusted Risk Ratiosa
Outcome: Frequent Knee Pain
Never Bicyclers (n = 1263)	43.1%	Referent	Referent
One Period Bicyclers (n = 704)	36.9%	0.85(0.76–0.96)	0.85(0.75–0.95)
Two Periods Bicyclers (n = 339)	36.9%	0.85(0.72–0.99)	0.85(0.72–0.98)
Three Periods Bicyclers (n = 164)	32.3%	0.75(0.59–0.93)	0.72(0.56–0.91)
Four Periods Bicyclers (n = 137)	32.1%	0.75(0.57–0.95)	0.72(0.55–0.92)
		p for trend =0.0002	p for trend <0.0001
Outcome: ROA
Never Bicyclers (n = 1263)	61.2%	Referent	Referent
One Period Bicyclers (n = 704)	57.2%	0.94(0.86–1.01)	0.94(0.86–1.02)
Two Periods Bicyclers (n = 339)	53.4%	0.87(0.77–0.97)	0.90(0.79–0.99)
Three Periods Bicyclers (n = 164)	52.4%	0.86(0.72–0.99)	0.90(0.76–1.03)
Four Periods Bicyclers (n = 137)	47.5%	0.77(0.63–0.92)	0.83(0.69–0.98)
		p for trend <0.0001	p for trend =0.002
Outcome: SOA
Never Bicyclers (n = 1263)	30.9%	Referent	Referent
One Period Bicyclers (n = 704)	25.9%	0.84(0.71–0.97)	0.83(0.71–0.96)
Two Periods Bicyclers (n = 339)	25.1%	0.81(0.66–0.99)	0.81(0.66–0.99)
Three Periods Bicyclers (n = 164)	22.0%	0.71(0.52–0.95)	0.72(0.52–0.96)
Four Periods Bicyclers (n = 137)	17.5%	0.57(0.38–0.81)	0.57(0.39–0.83)
		p for trend <0.0001	p for trend <0.0001

^aAdjusted for age and gender.

Our three sensitivity analyses supported our primary findings: 1) adjusting for BMI or injury (Supplemental Table 6, Supplemental Digital Content, Sensitivity analysis with additional adjustment), 2) among Whites (Supplemental Table 7, Supplemental Digital Content, Sensitivity analysis only including Whites), and 3) excluding participants with arthroplasty (Supplemental Table 8, Supplemental Digital Content, Sensitivity analysis excluding those with arthroplasty).

DISCUSSION

This study is the first epidemiologic study to evaluate a lifetime history of bicycling in a group of people not selected for bicycling status and where the outcomes of knee health were ascertained rigorously with a high level of standardization. Our findings indicate that those who participated in bicycling had a lower prevalence of frequent knee pain, radiographic evidence of osteoarthritis, and the combination of the two outcomes. Additionally, bicycling during more age periods was associated with a lower prevalence of all the outcomes of interest. This observational study indicates that bicycling is favorable to knee health.

Based on data from the Global Burden of Disease, osteoarthritis is the most common form of arthritis, and it affects more than 7% of all people, representing as many as 528 million people (3). Osteoarthritis most commonly affects the knee joint, accounting for 356 million cases globally (12). In our study, 52.3% of the participants identified bicycling as a top three activity that they participated in at least one of the age periods. Of those who participated in bicycling at some time, 67.6% did so during ages 12–18 years. Since the average age of our participants was in their 60’s, this would have been in the 1960’s, and they would not have had access to many of the sophisticated technology available on bicycles today.

Beyond the benefits of bicycling for knee health, there are other benefits to bicycling. The European Prospective Investigation Into Cancer and Nutrition (EPIC) study demonstrated that people with diabetes who participate in bicycling have lower all-cause mortality and cardiovascular-related mortality (13). Also, those who bicycle experience less weight gain over time (14) and have less stress (15) than those who do not. In a small study, there were also suggestions that bicycling was associated with improved executive functioning (16).

Some methodologic decisions warrant discussion. In our selection of covariates to adjust for, we chose not to adjust for BMI in our primary analyses because there was a possibility that BMI is a mediator in the relationship between bicycling and osteoarthritis. As a rule, variables should not be included as a covariate in an analysis if it is likely a mediator. Because there was an interest in knowing the influence of adding BMI into the model, we added it to the model as a sensitivity analysis, presented in Supplemental Table 6 (Supplemental Digital Content). In this model, the point estimate was mildly tempered, though it still showed bicycling to be associated with less knee pain, ROA, and SOA. There was a similar interest in understanding the influence of injury on the association between bicycling and knee osteoarthritis, which did not ultimately change the results much either. In Supplemental Figure 1 (Supplemental Digital Content, A DAG showing the hypothesized causal relationship between bicycling and osteoarthritis outcomes), we have provided a proposed directed acyclic graph, indicating the expected relationship amongst the variables of consideration for this study. The figure also shows that socioeconomic status (income) may relate to bicycling and osteoarthritis-related outcomes; however, we lack a measure of socioeconomic status during bicycling exposure. Instead, we offered a sensitivity analysis among Whites, but future studies may help us understand the association of bicycling and osteoarthritis-related outcomes among underrepresented groups and across socioeconomic states. Another point of discussion is that the instrument used in our study to evaluate the exposure of bicycling has been used to study other leisure physical activities within the Osteoarthritis Initiative. We view this as evidence of the construct validity of the measure. Thus far, it has been used to evaluate running (17), found to be associated with less knee pain and SOA; swimming (18), found to be associated with less ROA and SOA; and strength training(19), found to be associated with less knee pain, ROA and SOA. Meanwhile, American football has been associated with more knee pain, ROA, and SOA (20). We made the decision not to include analyses of all leisure physical activities in one study because the participation in each of the activities is quite different across lifetimes. For example, most of the people who reported participation in football did so in the youngest age range, and 95% were men (20). Therefore, those analyses were restricted to men, focusing on the younger age ranges (20). On the other hand, running as an activity was most common during 35–49 years of age, with a decrease in participation during ages 50 and older (17). Thus, by evaluating these exposures individually, we can customize our analytic approach, best utilizing the data available. We have found both protective and harmful associations for different activities with knee osteoarthritis, adding credence to our approach to using this instrument. The associations that have been described, including those from this study, will inform the need and design of future longitudinal studies evaluating the influence of leisure physical activities as they relate to knee osteoarthritis, a disease with a very extended natural history.

There are some limitations to this study. One is that the exposure of interest, bicycling, has been retrospectively ascertained. The exposure of bicycling in this study is over a long duration of time, thus prospective assessment of this exposure would have been much costlier and logistically challenging than the retrospective ascertainment utilized in this study. While there is a risk of recall bias for bicycling status, we believe this is minimal because participants were unaware of the hypotheses we were testing when they completed the questionnaire. Another limitation is that this is a cross-sectional study, making it difficult to be definitive about assessments of causation. However, in favor of the possibility that bicycling is beneficial from a knee health perspective, those who bicycled for all four age periods were the least likely to have all the outcomes of interest in a dose-dependent manner (Table 3).

Despite the limitations to our study, our findings indicate that those who participated in bicycling had a lower prevalence of frequent knee pain, ROA, and SOA are particularly exciting given the pervasive participation of bicycling already. As of 2015, 42% of households around the world own at least one bicycle, for an estimated 580 million bicycles worldwide (21). Additionally, more than 600 cities worldwide have bike-share systems, with the largest systems being in Chinese large urban areas, also adding to the number of bicycles ridden at any given time (22). This has only been boosted by trends related to the COVID-19 pandemic, affectionally referred to as the “Bike Boom” where bicycle purchases substantially increased, in part because bicycling was viewed as a safer form of exercise and mode of transportation (23).

CONCLUSIONS

In summary, those who participated in bicycling had a lower prevalence of frequent knee pain, ROA, and SOA. The benefit was cumulative if people participated in the activity during more age periods. This study supports the view that bicycling may be favorable to knee health.

DATA AVAILABILITY STATEMENT

The datasets generated during and/or analyzed during the current study are available in the Osteoarthritis Initiative repository, https://nda.nih.gov/oai/.