The atrophic type of hip osteoarthritis (OA) is characterized by cartilage degradation without the formation of osteophytes. Individuals with atrophic OA have been less well studied, and it is unknown whether this OA type differs from the osteophytic types with regard to bone tissue. The purpose of this study was to examine bone mineral density (BMD), hip structural properties, and fracture risk in individuals with the atrophic type of OA as compared to those with the osteophytic types (normotrophic/hypertrophic) as well as individuals without OA.
This study is part of the Rotterdam Study, a large prospective population-based cohort study. We examined 5,006 participants who had been assessed for OA, BMD, and geometric measures at baseline and for incident nonvertebral osteoporotic fractures (mean followup 9.6 years). We estimated the differences in bone characteristics between the OA groups and the controls (no joint space narrowing or osteophytes). Cox proportional hazards regression was used to calculate osteoporotic fracture risk.
Participants with atrophic OA had systemically lower BMD as compared to those with normotrophic OA and as compared to the controls (6.5% and 9% for total body BMD; 4% and 5% for skull BMD, respectively). Participants with osteophytic OA had ∼4% and ∼5% higher total body and skull BMD, respectively, a wider femoral neck, and greater bone strength (12% and 5% higher section modulus, respectively) as compared to the controls or to those with atrophic OA. The risk of osteoporotic fractures was almost 50% higher in those with atrophic OA as compared to the controls (hazard risk 1.48, P = 0.008). This difference was not explained by differences in the BMD, number of falls, degree of disability, or use of corticosteroids.
Individuals with atrophic hip OA have an increased risk of osteoporotic fractures that is not fully explained by systemically lower BMD as compared to controls.
Hip osteoarthritis (OA) is a common joint disorder and a major cause of pain and disability in the elderly population. OA is characterized by cartilage degradation, new bone formation, and changes in subchondral bone. Although OA was once considered a primary disorder of articular cartilage, it is now generally thought that bone structure plays a role in the pathologic changes of OA.
Individuals with OA are known to have higher bone mineral density (BMD) and, hence, to be protected against osteoporosis (1–4). However, both conditions can coexist (5–7), and some reports even suggest that patients with OA may have an increased risk of fractures (8–11).
Hip bone geometry has been shown to influence the risk of fracture (12–18). Differences in bone geometry have also been observed in subjects with hip OA as compared with subjects without radiographic evidence of hip OA, presenting with a wider femoral neck and alterations in body mass distribution, which are associated with both incident and prevalent OA (19). Yet, only a few studies have examined the relationship of hip OA to fracture risk, and none of them have considered the different subtypes of hip OA (2, 4, 8, 10, 11, 20–22).
Radiographic hip OA is frequently defined as the presence of both cartilage degeneration (radiographically defined as joint space narrowing) and the formation of new bone spurs at the joint margins (osteophytosis). However, both radiographic features are not always present, which makes it possible to subclassify OA into different types (23–25). The most frequently studied form of hip OA (classic, or normotrophic) presents with both joint space narrowing and osteophytosis. When only osteophytosis is present, it is called hypertrophic hip OA, and when only joint space narrowing is observed, it is called atrophic (23). The atrophic form of hip OA has been studied far less. However, it has been suggested that patients with atrophic hip OA have a higher risk of hip joint destruction (26, 27) and faster progression of disease (28). In addition, patients with atrophic hip OA have been shown to present with microarchitectural disorganization, lower bone volume, and thinner trabeculae than patients with osteophytes, as determined by bone histomorphometry of the iliac crests (27). Therefore, atrophic hip OA might be the result of reduced bone-forming capacity.
The aim of the present study was to examine BMD and structural properties across individuals with different types of hip OA: atrophic and osteophytic (normotrophic/hypertrophic). We also examined the relationship between these hip OA subtypes and the risk of fracture in comparison to control subjects without OA (no joint space narrowing or osteophytes).
PATIENTS AND METHODS
The Rotterdam Study.
The Rotterdam Study is a large prospective population-based cohort study of men and women ages 55 years and older. The design and rationale of the study have been described in detail elsewhere (29). In summary, the objective of the Rotterdam Study is to investigate the determinants, incidence, and progression of chronic disabling diseases in the elderly. The medical ethics committee of Erasmus University Medical School approved the study, and written informed consent was obtained from each participant.
The baseline examination of the original Rotterdam Study cohort included 7,983 participants. For the present study, we used data from 5,006 participants (62.7%) for whom data on both hip OA and BMD were available.
At baseline, trained interviewers and physicians performed an extensive home interview and examination, eliciting information about demographic characteristics, medical history, risk factors for chronic diseases, and use of medication that was verified using computer records generated from the pharmacy network. We also used information about smoking (current and former versus never), use of analgesics (in the last month: yes/no). Falling was assessed by trained medical research nurses using structured personal interviews. Individuals with a history of ≥1 falls in the absence of a precipitating trauma (e.g., car accident or sport injury) during the 12 months preceding the baseline interview were identified. A lower limb disability index was obtained by calculating the mean scores for answers to questions concerning the ability to rise from a chair, walk outdoors on flat ground, bend, and get in and out of a car. The index is represented by a continuous score ranging from 0 to 3, where 0 indicates no impairment and 3 indicates severe impairment. Detailed definitions for lower limb disability are described elsewhere (30). Height and weight were measured at the baseline examination, with the subject in a standing position, wearing indoor clothing, but without shoes.
Use of corticosteroids.
Persons who received a prescription for oral, rectal, parenteral, or inhaled corticosteroids within 1 month before the index date were defined as current users of corticosteroids. All others were considered noncurrent users.
Measurement of C-reactive protein (CRP) levels.
At baseline (1990–1993) and followup (1996–1999), blood was drawn by venous puncture. Samples were initially stored at −20°C, and then thawed and assayed for high-sensitivity CRP levels, using a rate near-infrared particle immunoassay (Immage Immunochemistry System; Beckman Coulter). This method can accurately measure protein concentrations from 0.2 mg/liter to 1,440 mg/liter with a within-run precision of <5.0%, a total precision of <7.5%, and a reliability coefficient of 0.995.
Radiographic assessment and definition of hip OA.
Weight-bearing anteroposterior radiographs of the hip were obtained at 70 kV, with a focus of 1.8 mm2, and a focus-to-film distance of 120 cm, using High-Resolution G 35 × 43–cm film (Fuji Photo Film Company). Radiographs of the pelvis were obtained with both of the patient's feet positioned in 10° of internal rotation and the x-ray beam centered on the umbilicus. Hip radiographs were scored for the presence of OA features (osteophytes and joint space width) by 2 independent observers who were trained by a physician experienced in OA radiography (HJMK) and advised by a radiologist. The trained observers used as a reference an atlas of individual radiographic features in OA (31). After each set of ∼250 radiographs, the scores assigned by the 2 trained observers were evaluated. If the scores differed by >1 grade for osteophytes or by >30% for joint space width (intraclass correlation coefficient [ICC]) a consensus reading was carried out. Readers were blinded with regard to the data for all of the participants. Additionally, there was no indication of the subject's sex or age on the radiographs themselves.
The joint space width of the hip was measured using a 0.5-mm graduated magnifying glass laid directly over the radiograph. The superior and inferior compartments of the femoral head were reviewed in both hips as lateral, superior, and axial assessments (32). We determined joint space narrowing at axial, lateral, and superior sites in all hips. The axial site was the most frequently affected by joint space narrowing; this was the case in normotrophic as well as atrophic hip OA. The Intraclass correlation coefficient (ICC) for the minimal joint space was 0.73 (range 0.69–0.77) and the kappa statistic for femoral osteophytes was 0.76. These estimates were an average that included ICCs taken from the first to the last reading sessions.
We defined joint space narrowing as a joint space width ≤2.5 mm in at least 1 compartment. Osteophytosis was defined as the presence of at least 1 definite osteophyte in the femur. These 2 features, joint space narrowing and osteophytosis, were assessed as dichotomous variables (present/absent) and were used to classify participants as presenting with 1 of the 4 different hip OA groups: no OA, normotrophic OA, hypertrophic OA, and atrophic OA (Table 1).
Table 1. Definition of hip OA according to bone response*
Type of hip OA
Group 1 subjects, who did not have osteoarthritis (OA) of the hip (i.e., no joint space narrowing [JSN] or osteophytes), served as the control group for the study.
Subjects without JSN or osteophytes
Subjects with osteophytes and JSN
Subjects with only osteophytes
Subjects with only JSN
Measurements of BMD and bone geometry.
Dual x-ray absorptiometry (DXA) was used to determine BMD (gm/cm2) of the right proximal femur and lumbar spine at baseline using a Lunar DPX-L densitometer, and data were analyzed with DPX-IQ v.4.7d software, as described previously (33, 34). Total body scans were performed at the third followup visit (mean followup 6.5 years) using a Prodigy fan-beam densitometer (Lunar), and data were analyzed with enCore software. The software uses an algorithm that divides body measurements into areas corresponding to total body, head, trunk, arms, and legs. All analyses were verified by a trained technician, who performed adjustments when necessary. Hip structural analysis software was used to measure bone geometry of the hip on the DXA scan images at the narrow neck region, across the narrowest point of the femoral neck. BMD and width of the narrow neck region (outer diameter) were measured directly from mineral mass distribution, while the section modulus (an index of bending strength) and buckling ratio (an index of cortical bone instability) were calculated using algorithms described previously (33).
Assessment of incident fracture.
All events, including fractures and death, were reported by the general practitioners in the research area (covering 80% of the cohort) by means of a computerized system. Information from general practitioners outside the research area was obtained by regular checking of patient records by research physicians. All reported events were verified by 2 trained research physicians, who independently reviewed and coded the information. All coded events were subsequently reviewed by a medical expert for final classification. Subjects were followed up from their baseline visit until January 1, 2007 or until the occurrence of a first fracture or death (mean ± SD fracture followup duration 9.7 ± 5.1 years). All fractures that were not considered to be osteoporotic (fractures caused by cancer and all hand, foot, skull, and face fractures) were excluded.
We compared the baseline characteristics of the study population and the DXA-derived data on the BMD and geometry parameters from the narrow neck region of the femur between the OA types and the controls by analysis of variance. Categorical variables were tested using the chi-square statistic. All of the BMD measurements and the DXA-derived geometric analyses were adjusted for sex and age. We also examined BMD values obtained at the femoral neck, total body, and skull. Additionally, Z scores were included to compare differences between the different types of hip OA and the controls without hip OA. Analyses of DXA-derived data on hip geometry were adjusted for BMD of the narrow neck region to test its independence. Cox proportional hazards regression was used to study associations between the different types of hip OA and fracture risk. Hazard risks (HRs) with 95% confidence intervals (95% CIs) are also reported. In the Cox regression model for osteoporotic fracture, the type of hip OA was included as a categorical variable. Analyses were adjusted for sex, age, height, and weight, and to evaluate the effect of BMD and corticosteroid use, an additional adjustment for femoral neck BMD was made. All analyses were run using SPSS software version 15.0.
The baseline characteristics of individuals with the 3 different types of hip OA and those without hip OA are shown in Table 2. On average, participants with hip OA were older, and more of them were female, except for those in the hypertrophic hip OA group, which contained more men. There were no significant differences in height, weight, or body mass index between the groups. Those in the hypertrophic and normotrophic groups had more disability in the lower limb as compared to the control subjects without OA.
Table 2. Baseline characteristics of the study subjects, by hip OA type*
Subjects without osteoarthritis (OA) had no definite osteophytes or joint space narrowing. Except where indicated otherwise (categorical variables), values are the mean ± SD (continuous variables).
For the comparison of all subjects with hip OA versus the group without OA, as determined by chi-square test or by analysis of variance.
P < 0.01 versus the group without OA, adjusted as follows: height and body mass index (BMI) were adjusted for sex and age; weight was adjusted for sex, age, and height; lower limb disability was adjusted for sex, age, and weight; and C-reactive protein (CRP) and ≥1 fall per year were adjusted for sex, age, and BMI.
Table 3 presents the results of the adjusted baseline bone and geometry parameters measured by DXA at the narrow neck region of the femur. BMD values at the narrow neck region, the femoral neck, and the lumbar spine were higher in the osteophytic groups compared to controls (Table 3). The atrophic hip OA group had BMD values similar to those in the controls and significantly lower than those in the osteophytic hip OA groups (P < 0.05, not in Table 3). Despite the fact that total body and skull BMD were measured at a later time point, when fewer participants were included in the analysis (n = 2,018), the values for these parameters followed the same trend, being higher in groups with osteophytic OA (P not significant) and significantly lower in the group with atrophic OA as compared to the controls (P = 0.002 for total body BMD and P = 0.007 for skull BMD). Figure 1 shows the Z scores for the BMD values in the different OA groups as compared to the controls.
Table 3. BMD of the femoral neck and lumbar spine and structural parameters of the hip measured at the narrow neck region of the femur, by hip OA type*
Subjects without osteoarthritis (OA) had no definite osteophytes or joint space narrowing. The narrow neck region represents the region across the narrowest point of the femoral neck. Total body bone mineral density (BMD) and skull BMD were measured at a later time point and included 2,018 subjects; structural parameters of the hip were measured in a total of 4,007 subjects who had OA data and dual x-ray absorptiometry (DXA) scans of the narrow neck region. Values are the mean ± SD.
For the comparison of all subjects with hip OA versus the group without OA, as determined by analysis of variance and adjusted for age and sex.
P < 0.01 versus the group without OA, adjusted for age and sex.
P < 0.05 versus the group without OA, adjusted for age and sex.
Subjects with normotrophic or hypertrophic hip OA had 4.1% and 2.5% wider femoral necks and 12.3% and 5% greater bone strength (higher section modulus), respectively, than the control subjects without hip OA. No significant differences in these parameters were found in subjects with atrophic hip OA as compared to the controls.
A total of 1,071 participants with radiographs scored for OA sustained a fracture during the period of investigation. Participants with the atrophic type of hip OA had an almost 50% increased risk of fractures as compared to controls, both with and without adjustment for BMD (HR 1.48 [95% CI 1.11–1.98], P = 0.008 and HR 1.44 [95% CI 1.08–1.92], P = 0.012, respectively) (Figure 2). The normotrophic and hypertrophic hip OA group did not differ significantly in terms of fracture risk from the control group (after adjustment for BMD, HR 1.18 [95% CI 0.84–1.66], P = 0.35 in the normotrophic OA group and HR 0.89 [95% CI 0.74–1.06], P = 0.20 in the hypertrophic OA group (Figure 2). Additional adjustment for lower limb disability did not change the risk estimate, as was expected, since lower limb disability in the atrophic hip OA group was not significantly different from that in the control group (P > 0.05) (Table 2). Similarly, adjustment for falling did not change the risk estimates. The percentages of subjects with ≥1 fall at baseline was not significantly different between the hip OA types and the controls (P > 0.05 for each comparison of OA type versus controls) (Table 2). Finally, adjustment for use of corticosteroids at baseline changed the risk estimates to a slightly higher fracture risk for participants with the atrophic type of hip OA (HR 1.69 [95% CI 1.19–2.41], P = 0.004). The risk of osteoporotic fractures remained not significant for participants with other hip OA types.
In this large, prospective, population-based study, we found differences in the structural geometry of bone across individuals with different types of hip OA. Subjects with osteophytic hip OA (both normotrophic and hypertrophic) had higher BMD, wider femoral necks, and greater bone strength than did those with atrophic hip OA and those without hip OA. While individuals with atrophic OA had lower skull and total body BMD than did those without OA, no differences were observed in the hip geometry parameters. A 50% increase in osteoporotic fracture risk was observed in individuals with atrophic OA as compared to subjects with osteophytic OA as well as those without OA.
Previous studies have found increased BMD and differences in DXA-derived geometry values in OA hips as compared to controls without OA (19, 35–38). However, the majority of those studies used the classic definition of hip OA, which required the presence of cartilage degradation (joint space narrowing) and osteophytes. Such a definition excludes individuals with atrophic forms of OA (only joint space narrowing), despite some reports indicating that these individuals present with clear signs of cartilage degradation (24) and rapid disease progression (39) in the absence of osteophytes. Consistent with our findings, Javaid et al (19), in their Study of Osteoporotic Fractures cohort, found that individuals with an osteophytic type of hip OA had wider femoral necks and displacement of the center of mass. In general, individuals with the atrophic type of hip OA have been excluded from previous studies analyzing the relationship between OA and osteoporosis.
Even though we found no differences in hip structural parameters in subjects with the atrophic form of hip OA, we did find that they had a consistent, systemically lower BMD than did subjects without hip OA and with osteophytic forms of hip OA. The fact that the BMD in subjects with atrophic OA was also lower than that in subjects without OA suggests that this phenomenon is not explained by the absence of osteophytes alone (i.e., artifactual elevation of BMD in those with osteophytic forms of OA). This seems to be the result of a systemic process, considering that total body BMD was also decreased, as well as the fact that this decrease was also evident on skull BMD, a site less prone to being influenced by environmental factors, mechanical strain, and weight bearing (40). Overall, these measurements are less affected by artifactual elevations arising from OA changes.
We evaluated whether the observed differences in skeletal properties observed across the different types of hip OA translated to differences in fracture risk. Even though participants with osteophytic types of OA had higher BMD, wider femoral necks, and greater bone strength than controls, we observed no significant differences in fracture risk. Inversely, subjects with atrophic hip OA who did not differ in DXA-derived geometry values but who had lower BMD, had an ∼50% increased risk of osteoporotic fractures, which was still present after adjustment for BMD. Furthermore, this increased risk is not likely to be explained by lower BMD as a consequence of immobility or by differences in the risk of falling, since correction for lower limb disability and falling did not essentially modify the risk estimate. Finally, adjustment for corticosteroid use made the difference in osteoporotic fracture risk between the atrophic OA group and the controls more pronounced. It is known that prior and current exposure to corticosteroids confers an increased risk of fracture (41).
There are possible explanations for the increased osteoporotic fracture risk in participants with the atrophic type of OA. Subjects with atrophic hip OA might have lower bone quality, and such characteristics are not captured by DXA or DXA-derived geometry measures. Indeed, earlier studies of patients with hip replacements showed that those with OA had lower bone volume and thinner trabeculae than did controls without OA (27). The ones most affected were those with hip joint destruction, which was 3 times more common in subjects with atrophic OA than in those with the other OA types. The study of bone properties using methods other than DXA scans could elucidate bone differences between OA types that might have an effect on fracture risk.
Examination of biomarkers of bone metabolism might also clarify important differences between OA types. In our study, we did not have available data on markers of cartilage and bone resorption and formation, which could have helped to classify the OA subtypes.
Our study has some limitations. Even though we evaluated a large population prospectively, with prospective assessment of fracture, which is less susceptible to bias, dividing the cohort into types of hip OA limited the sample size and, hence, the power of the study to detect differences.
Possible selection biases due to different response rates and to inclusion of participants according to the availability of hip radiographs and DXA images have been discussed previously (42). There was a high response rate in the Rotterdam Study (≥80%), and selection bias will therefore be limited. Yet, people who refused to participate were generally older (mostly >80 years of age) and more often, were seriously ill. Additionally, the fact that subjects had to be mobile enough to visit the research center for radiography and DXA examinations caused a possible health selection bias in our study population. Taking these factors into consideration, it can be expected that our risk estimates are biased toward lower levels of risk.
Other limitations are the inherent properties of DXA scans and the absence of information about markers of bone metabolism. Despite the information about bone structure that is obtained with structural analysis of the hip, the major limitation of such an analysis is imposed by the 2-dimensional nature of DXA (34). The addition of computed tomography to assess 3-dimensional structural configurations would be the ideal situation. Finally, there is a limitation inherent in the reproducibility of manual measurements of joint space width among observers. Interobserver reproducibility of the ICC using manual measurement techniques ranges from 0.71 to 0.78 for joint space width measurements (43). The ICC values for our measures fall within these boundaries. The variability in measuring joint space width was principally due to difficulties in determining the exact points on the femoral head and acetabulum and the directions in which to measure the joint space at the different sites (lateral, superior, and medial) in the hip joint. However, this ICC estimate was an average that included interobserver values for the first reading session, where the observers had less expertise. Joint space width is considered a reliable method of measuring the progression of OA in hip joints, with higher reliability than scoring of osteophytes (25, 44).
In conclusion, we identified differences in BMD and DXA-derived parameters principally between the atrophic and osteophytic subtypes of hip OA. Our findings confirm that there are indeed different types of hip OA with specific structural properties of the bone and different relationships to the risk of fracture. Classifying individuals according to subtypes of hip OA may help to identify distinct causes and pathogenic courses of the disease, which may translate to more appropriate therapeutic interventions.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. van Meurs had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Castaño-Betancourt, Rivadeneira, Bierma-Zeinstra, Kerkhof, Hofman, Uitterlinden, van Meurs.
Acquisition of data. Castaño-Betancourt, Rivadeneira, Bierma-Zeinstra, Kerkhof, Hofman, Uitterlinden, van Meurs.
Analysis and interpretation of data. Castaño-Betancourt, Rivadeneira, Bierma-Zeinstra, van Meurs.
We thank Dr. T. J. Beck for his contribution to the structural analysis of the hip that was included in this study and others. The authors are grateful to the participants, staff, general practitioners, and pharmacists involved in the Rotterdam Study.