Patients with osteoarthritis (OA) have increased bone mineral density; however, the association between knee OA and fracture is controversial. Few data exist on the association between knee pain and fracture. We examined the association of knee OA and knee pain with fracture and falls in elderly men and women.
The study group comprised 6,641 men and women ages ≥75 years who participated in a 3-year randomized controlled trial of intramuscular vitamin D therapy. Patients completed a questionnaire about knee pain and OA. Fracture and fall data were collected prospectively every 6 months.
Knee pain prevalence and a clinician diagnosis of knee OA were 35.2% and 6.8%, respectively. A total of 436 incident nonvertebral fractures were reported, and 3,992 patients sustained a fall. Prevalent knee pain was associated with an increased risk of falls (hazard ratio [HR] 1.26, 95% confidence interval [95% CI] 1.17–1.36) and hip fracture (HR 2.0, 95% CI 1.18–3.37). Increasing severity of knee pain was associated with a greater risk of falls and hip fracture. Clinician diagnosis of knee OA was associated with an increased risk of nonvertebral fractures (HR 1.61, 95% CI 1.09–2.36). The increased risk of fracture was not substantially reduced by adjusting for falls, but was attenuated by adjustment for the use of walking aids.
Patients with a clinical diagnosis of knee OA and with knee pain have an increased risk of nonvertebral and hip fracture. This is not explained by the increased risk of falls, but is more likely to be due to the severity of falls sustained. Knee pain and OA should be regarded as independent risk factors for fracture.
Osteoporosis and osteoarthritis (OA) of the knee are both common conditions in the elderly population. A 50-year-old woman and a 50-year-old man have a 53.2% and 20.7% lifetime risk of sustaining an osteoporotic fracture, respectively (1). The prevalence of radiographic knee OA is 33% in persons >63 years of age (2), the prevalence of clinically diagnosed knee OA is 18.1% in those ages ≥55 years (3), and the annual prevalence of significant knee pain is 25% (4). There is a low degree of concordance between radiographic knee OA and knee pain; however, both conditions lead to significant morbidity and health care costs.
The assessment of osteoporotic fracture risk in a patient requires the consideration of a number of different risk factors, including age, sex, bone mineral density (BMD), and previous fracture, in addition to a number of other clinical risk factors. Since the first observations by Foss and Byers (5) that patients with hip fracture rarely had OA, it has been proposed that patients with OA have a reduced risk of osteoporosis and subsequent fracture, and that OA should therefore be regarded as a protective factor when assessing a patient's overall fracture risk. Indeed, numerous studies have demonstrated that patients with OA have increased BMD, even at sites distant to the OA site (6–9). This association appears to be stronger for knee and hip OA (6, 8) than for generalized or hand OA (10).
There is, however, considerable uncertainty as to whether the increase in BMD is translated into a reduced risk of osteoporotic fracture. Although early case-control studies suggested a reduced risk of fracture in patients with OA (11, 12), more recent studies have found either no reduction in risk (9, 13) or an increased risk of fracture (14, 15). There are no data on the risk of fracture in patients with knee pain, the diagnosis most commonly seen in clinical practice.
Patients with knee OA have increased body sway (9) and an increased risk of falling (13), and possibly an increased risk of serious falls resulting in fracture. This may explain the failure of increased BMD to translate into a reduced risk of fracture in these patients. The goal of this study was to explore the risk of falling and fracture in patients with both knee pain and a clinician diagnosis of knee OA in a large clinical trial.
PATIENTS AND METHODS
Study participants were patients who had recently participated in a large, 3-year, randomized, double-blind, placebo-controlled trial assessing the effectiveness of intramuscular vitamin D injections in preventing osteoporotic fractures (16). For this trial, men and women ages ≥75 years were recruited from listings of 111 general practices included in the Wessex Primary Care Research Network (WReN). The WReN includes practices throughout Hampshire, Wiltshire, and Dorset, where lead general practitioners are willing to undertake population-based research and enroll their populations. Previous studies have demonstrated that the age, sex, and socioeconomic status of patients at practices enrolled within the network represent that of England and Wales as a whole (17).
All patients age ≥75 years were invited by their general practitioner to make an appointment with the district or practice nurse for an annual influenza vaccination during the period from September to December. When attending for the influenza immunization, patients were informed of the trial and invited to participate. Patients were excluded if they had a history of renal failure, renal stones, hypercalcemia or sarcoidosis, or if they had current cancer. Patients who were already taking vitamin D supplements (>400 IU daily), who had bilateral hip replacements, or who were being treated for previously diagnosed osteoporosis (bisphosphonates or estrogen) were also excluded.
Patients were randomized at an individual level to either the intervention or control group. Each patient received either 300,000 IU (1 ml) of vitamin D2 (ergocalciferol) or matching placebo (Evans Medical/Celltech, Slough, UK) together with the influenza vaccination by intramuscular injection at annual intervals. Invitations were sent to ∼13,000 patients over a 3-year period. A total of 3,188 patients were recruited during the first year of the study, 4,692 were recruited during the second year, and 1,560 were recruited during the third year.
Fracture and falls.
The principal outcome of the trial was all nonvertebral fracture. Fracture ascertainment was performed at 6-month intervals using the fracture questionnaire validated in the European Prospective Osteoporosis Study (18). This questionnaire enquires about fractures over the intervening 6 months, and is accompanied by a mannequin on which the fracture site is indicated. All fracture sites were coded using the fracture classification adopted by the International Classification of Diseases, Ninth Revision. A history of falls during the intervening 6-month period was also obtained by questionnaire.
Knee pain and osteoarthritis.
Patients were sent a questionnaire enquiring about knee pain and OA with their final fracture questionnaire. The knee pain and OA questionnaire asked whether patients had “ever had pain in or around the knee on most days for at least a month.” This question has been used to classify knee OA in the National Health and Nutrition Examination Survey and is now widely used in epidemiologic studies of knee OA (19). If patients replied yes, they were asked when the pain had first started. They were also asked if they had ever received a clinician diagnosis of knee OA: “Has a doctor ever told you that you have osteoarthritis of the knee?” Patients were asked to quantify their current knee pain using a 5-point ordinal scale.
The study had a power of 80% to detect a difference in fracture rate of 37.5% from 3.6% (i.e., 3.6% versus 2.25%) at the 5% significance level among 4,000 patients with no knee pain and 2,000 patients with knee pain. Patients were classified into never, ever, or prevalent cases of knee pain or clinician diagnosis of knee OA. Prevalent cases were defined as patients reporting knee pain or clinician diagnosis of knee OA who reported the onset as being before entry into the clinical trial. Baseline clinical characteristics were compared using t-tests. Using Cox regression in Stata 8.2 (Stata, College Station, TX), hazard ratios (HRs) were calculated for hip fracture, wrist fracture, and all nonvertebral fractures using each of the above definitions of knee pain or clinician diagnosis of knee OA. All analyses were adjusted for treatment allocation. The 5-year risk of fracture by age and sex was calculated using a life-table approach to build a data set of surviving patients at each year of life for a hypothetical cohort of 100,000 individuals. The mortality data were extracted from UK national death rates (20). Fracture rates from this study were then applied to the theoretical cohort to estimate the 5-year risk of fracture stratified by age and sex.
A total of 6,641 patients completed both the fracture and knee questionnaires. Those who did not complete the knee questionnaires tended to be older than those who completed the questionnaires (median age 80.2 years versus 78.8 years; P < 0.01); however, there was no difference in the treatment allocation, sex, or prevalence of hip, wrist, or all nonvertebral fractures at baseline. At the first study visit, 38.2% of patients had previously sustained a fracture, 35.2% reported knee pain, and 6.8% reported a clinician diagnosis of OA. During the study period, 436 nonvertebral fractures and 3,992 falls were reported.
Patients with knee pain were more likely to be women and use a walking aid (Table 1). Those with prevalent knee pain had an increased risk of any prevalent nonvertebral fracture. Similar results were seen for patients with a clinician diagnosis of knee OA (Table 2). Eighty-nine percent of patients with a clinician diagnosis of knee OA reported knee pain compared with 31% of those without a clinical diagnosis of OA.
Table 1. Baseline clinical characteristics by knee pain status*
No knee pain (n = 4,026)
Ever knee pain (n = 2,186)
Prevalent pain (n = 1,427)
Values are the percentage unless otherwise indicated. IQR = interquartile range.
Patients with prevalent knee pain and, to a lesser degree, prevalent clinician-diagnosed knee OA were more likely to sustain at least 1 fall during the followup period (Table 3). Patients with prevalent knee pain and clinician-diagnosed OA had a high prevalence of walking aid usage (34.1% and 42.2%, respectively), which was itself associated with an increased risk of all falls (HR 1.34, 95% confidence interval [95% CI] 1.25–1.44). After allowing for walking aid usage, patients with prevalent knee pain still had an increased risk of falls, but this was not the case for patients with clinician-diagnosed OA. The risk of falls was greatest in those reporting severe knee pain compared with those with no knee pain (HR 1.51, 95% CI 1.32–1.72). Walking aid use was associated with an increased risk of hip fracture (HR 2.25, 95% CI 1.38–3.68) but not wrist fracture (HR 1.45, 95% CI 0.92–2.30).
Table 3. Hazard ratio (95% confidence interval) of knee pain and osteoarthritis (OA) predicting falls
Prevalent clinician-diagnosed knee OA
Prevalent knee pain
Falls adjusted for walking aid use
Patients with knee pain had an increased risk of sustaining a hip fracture, with a trend towards an increased risk of all nonvertebral fractures, but there was no increased risk in wrist fractures (Table 4). This risk was substantially reduced after adjusting for falls (Figure 1), although the risk of hip fracture remained statistically significant. Additional adjustment for use of walking aids further reduced these risks. The risks were not significantly reduced by further adjustments for prevalent nonvertebral fractures. The risk of all nonvertebral fractures was greatest in patients reporting the greatest pain on the ordinal scale (moderate knee pain and severe knee pain: HR 1.33, 95% CI 1.00–1.78; no knee pain: HR 2.20, 95% CI 1.58–3.07). The equivalent HRs for hip fracture were 1.72 (95% CI 0.79–3.73) and 5.36 (95% CI 2.65–10.83), respectively.
Table 4. Hazard ratio (95% confidence interval) of prevalent knee pain and osteoarthritis (OA) predicting fracture*
All analyses were adjusted for treatment allocation.
All nonvertebral adjusted for falls
All nonvertebral adjusted for falls and walking aids
Hip adjusted for falls
Hip adjusted for falls and walking aids
Wrist adjusted for falls
Wrist adjusted for falls and walking aids
Patients with a clinician diagnosis of knee OA had greater risks of all nonvertebral and hip fracture than patients with knee pain; however, due to the smaller number of patients, the increased risk of hip fracture failed to achieve clinical significance (Table 4). The risks were largely unchanged after adjusting for falls, but were reduced substantially by adjustment for the use of walking aids. The 5-year fracture risks were higher in women than in men (Table 5) and tended to be higher in patients with knee pain and OA.
Table 5. Five-year risks of fracture by age and prevalent knee pain and osteoarthritis (OA) status*
Knee pain/OA status
Comparison of ever, prevalent, and incident with no knee OA.
No knee pain
Clinician diagnosis of knee OA
No clinician diagnosis of knee OA
We have confirmed that patients with a clinical diagnosis of knee OA have an increased risk of nonvertebral fractures. In addition, our study is the first to show that patients with self-reported knee pain also have a significantly increased risk of hip fracture. Neither group of patients had an increased risk of wrist fractures. Patients with knee pain had an increased risk of falling, which was related to the severity of pain; however, this did not fully explain the increased risk of fracture.
Patients with radiographic OA have increased BMD at sites distant from the site of OA compared with age- and sex-matched controls. This finding is most marked for hip OA, with increases in BMD between 3% and 10% depending on the site of BMD estimation (6, 8). Similar results have been published for knee OA, with increases in BMD between 3% and 9% (7, 8, 15, 21). These increases in BMD equate to a reduction in fracture risk of up to 50%, stimulating the interest in the association of OA and fracture. However, several studies have demonstrated that prevalent OA is associated with increased rates of bone loss (8, 22), and other studies have shown that the most severe OA is not associated with an increase in BMD (7). These findings suggest that although patients with early OA have increased BMD, rapid rates of bone loss attenuate this effect as the disease progresses.
The 3 previous studies that examined radiographic OA have all failed to demonstrate a reduced risk of fracture, despite confirming an increased BMD (13–15). In a prospective study of 5,552 elderly women followed for an average of 7.4 years, there was no significant association between radiographic hip or hand OA and incident vertebral or nonvertebral fractures (13). This study also failed to demonstrate an association between either self-reported OA or hip pain and fractures. In a case-control study of 939 women (14), radiographic hip OA was associated with an increased risk of sustaining an osteoporotic fracture (odds ratio 2.38); however, no association was found for radiographic OA at the hand or knee. In a prospective study of 2,773 men and women, radiographic knee OA was associated with an increased risk of vertebral and nonvertebral fractures, with HRs of 1.9 and 1.5, respectively (15). In the 2 studies that collected prospective fall data (13, 15), the association between OA and fracture was not significantly altered by adjusting for falls.
Studies that have examined the association between self-reported OA and fractures have yielded conflicting results. The 2 early case-control studies of hip fracture did demonstrate a reduced risk of fracture in patients with self-reported OA. Dequeker and Johnell, using self-reported physician-diagnosed OA at any site, demonstrated a reduced risk of hip fracture with a relative risk of 0.48 in men and 0.68 in women (11). Cumming and Klineberg, using self-reported arthritis of any type at the hip joint requiring referral to a physician, demonstrated a reduced risk of hip fracture with a relative risk of 0.33 (12). Using the Dubbo Cohort Study, Jones et al examined the effects of self-reported OA on all low-traumatic fractures and found no significant association despite finding increased BMD in these patients (9). Patients with OA had decreased quadriceps strength and increased postural sway; this led the authors to suggest that these patients were at an increased risk of falling, which negated the benefit of the increased BMD.
The discrepancy between the results of the above studies may be due to a number of factors, including the definition of OA, the site of OA, and the study design. The concordance between clinical symptoms and radiographic features of OA is low (23, 24), which may explain the differences in the results of the studies using radiographic and self-reported definitions of OA. Furthermore, the concordance rate between self-reported OA and a confirmed clinician diagnosis of OA can be as low as 15% (25). It is unlikely that the association of OA and fracture will be the same for OA at different sites. Both studies that examined hand OA failed to find a significant association with fracture (13, 14), whereas studies of knee OA found either no association (14) or an increased risk of fracture (15). Studies focusing on hip OA found either no association (13), an increased risk of fracture (14), or a decreased risk of fracture (12). Study design is also an important factor to consider. The only studies that showed a reduced risk of fracture have used the case-control design (11, 12), whereas prospective cohort studies have shown either no association or an increased risk of fracture (9, 13, 15).
This is the first study to specifically address the issue of fracture risk in patients with self-reported knee pain. Although the majority of clinical research is directed at patients with radiographic knee OA, in clinical practice knee pain and clinically diagnosed knee OA are common clinical conditions (4). It is therefore important to address the association of fracture using these diagnoses. This study confirms that both of these diagnoses are associated with an increased risk of fracture. When considering the data from all of the available studies, it seems logical that knee OA and knee pain should be regarded as important clinical risk factors for osteoporotic fractures.
It is well established that lower-limb OA is associated with an increased risk of falling (9, 15, 26, 27) due to a combination of quadriceps weakness, decreased proprioception, and increased postural instability (9, 27, 28). To our knowledge, our study is the first to demonstrate that knee pain is associated with an increased risk of falling in an elderly population. Furthermore, the risk of falling is related to the severity of knee pain, which has important implications for the clinical treatment of these patients. The majority of nonvertebral fractures result from a fall, and it has been suggested that the increased risk of fracture in patients with OA is due to the increased risk of falls. However, our results suggest that the increased risk of falls in patients with knee OA and pain does not appear to fully explain the increased risk of fracture. It is possible that the residual excess risk is due to aspects of bone density/fragility associated with knee pain. Although we had no direct measure of this, we did adjust for previous nonvertebral fractures as a surrogate; however, this had little effect on the residual risk, most likely due to previous fracture being a poor surrogate for osteoporosis/bone fragility.
Only 4–5% of all falls lead to a fracture, with only 1% resulting in a hip fracture (29, 30). It is possible that patients with knee pain and OA have a disproportionate increase in falls likely to lead to a fracture. Our data provide several clues to support this theory. There is increasing evidence that the site of fracture will vary according to the nature of the fall (31). Hip fractures tend to result from a fall that occurs at low walking speed such that patients fall straight down or to the side, landing directly on the hip, whereas distal forearm fractures occur when the patient falls backwards and lands on the outstretched hand. Our results demonstrate an increase in hip fractures but no increase in wrist fractures consistent with a disproportionate increase in hip fracture falls. The addition of walking aid use, in addition to falls, into a multivariate model further attenuated of the risk of hip fracture with both knee OA and knee pain. Walking aids tend to be associated with low-velocity falls and disability, which will increase the risk of a fall onto the hip and reduce the risk of a fall onto the outstretched hand.
This study has several potential limitations. The patients were part of a randomized controlled trial of vitamin D therapy; however, the study did not demonstrate an effect of vitamin D on either falls or nonvertebral fractures. Furthermore, all results in this study were adjusted for treatment status, which had little impact on the results. The knee pain questionnaire was administered at the end of the study, and the timing of the onset of pain may be susceptible to a degree of nondifferential recall bias. We studied an elderly population, and recall of a clinician diagnosis of OA may be reduced. This is supported by the low prevalence of 4.2% compared with a recent survey of general practice databases in the same region but of a younger population, which demonstrated a prevalence of 18.1% (3). However, any bias is likely to be a nondifferential recall bias and unlikely to account for the positive associations found in this study.
In conclusion, we have shown that patients with knee pain or a clinician diagnosis of knee OA have an increased risk of hip and nonvertebral fractures. This risk is independent of falling but may be due to an increased risk of severe falls. These diagnoses should be regarded as a strong risk factor for fractures, and should be considered when assessing a patient's risk of fracture.
We thank the general practitioners and patients who participated. We also thank Julie Smith and Tanya Humphreys for administrative support, Vanessa Cox for managing the data, and Jason Poole for statistical advice. The report was prepared by Lynn Reeves. We are also grateful to the external advisory group: Dr. S. Drew, Dr. S. Clayton, Mr. D. Matthews, Mrs. B. Hughes, and Mrs. Sue O'Regan.