To quantify the differences in risk factors influencing total hip replacement (THR) and total knee replacement (TKR) based on the presence versus absence of multiple interphalangeal nodes in 2 or more rays of the fingers of each hand in patients with large joint osteoarthritis (OA).
A group of 3,800 patients with large joint OA who underwent total joint replacement (1,201 of whom had the nodal phenotype) and 1,906 control subjects from 2 case–control studies and a population-based cohort in the UK were studied. Odds ratios (ORs) and 95% confidence intervals (95% CIs) were calculated for the risk of total joint replacement in association with age, sex, body mass index (BMI), height, and prevalence of the T allele in the GDF5 rs143383 polymorphism. ORs for total joint replacement were compared between cases of nodal OA and cases of non-nodal OA and between patients who underwent TKR and those who underwent THR.
Age, sex, and BMI had significantly higher ORs for an association with total joint replacement in nodal OA cases than in non-nodal OA cases. The GDF5 polymorphism was significantly associated with THR in cases of nodal OA, but not in cases of non-nodal OA, and increased height was a risk factor for THR in non-nodal OA cases only. Female sex was a protective risk factor for TKR in non-nodal OA cases (OR 0.60, 95% CI 0.52–0.70) but was predisposing for TKR in the nodal form of OA (OR 1.83, 95% CI 1.49–2.26). The nodal phenotype was associated with a significantly higher risk of undergoing both THR and TKR (OR 1.46, 95% CI 1.09–1.94) and also a significantly higher risk of bilateral TKR (OR 1.70, 95% CI 1.37–2.11), but, paradoxically, was associated with a lower risk of bilateral THR (OR 0.72, 95% CI 0.56–0.91).
Nodal and non-nodal forms of large joint OA have significantly different risk factors and outcomes, indicating a different etiology for the 2 forms of OA. With regard to the likelihood of undergoing THR, this appears to be, at least in part, genetically determined.
Severe osteoarthritis (OA) of the hip and the knee represent 2 of the most significant causes of pain and physical disability in adults (1). OA is multifactorial, and patients are most likely affected by a combination of both genetic and environmental factors, characterized by a continuous distribution between the extremes of predominantly genetic and predominantly environmental (2).
Generalized OA refers to the involvement of at least 3 joints or of a group of joints (e.g., the interphalangeal [IP] joints). Two types of generalized OA have been described: nodal and non-nodal. The nodal type, characterized by Heberden's and Bouchard's nodes (firm, posterolateral swellings) of multiple distal and proximal IP joints, is predominant in women and is associated with underlying radiographic OA of the IP joints (3). Genetic factors play a role in both the incidence and, very likely, the outcome of OA. For instance, the presence of Heberden's nodes has been reported to confer a 6-fold increased risk of progression of knee OA (4). It has been proposed that generalized OA may be a distinct disease in which systemic (genetic) predisposition is more important than local environmental (e.g., mechanical) factors (5). Indeed, the presence of OA in the IP joints of the hand at baseline has been found to confer an increased risk of future hip OA (6).
The hereditary nature of Heberden's nodes was noted as early as the 19th century (7), and by the 1940s, it was concluded that the phenotype was inherited as a dominant trait (8). Further studies have established that nodal OA often occurs in the context of OA at multiple other joint sites, including the knee and hip, and shows familial clustering (7). It is thus likely that environmental and constitutional factors may play a larger role in the development of the non-nodal form of large joint OA, whereas genetic factors may have a more prominent role in large joint OA accompanied by IP nodes or radiographic hand OA.
The most widely reproduced genetic factor found to be associated with OA is the single-nucleotide polymorphism (SNP) rs143383 in the promoter region of the growth differentiation factor 5 (GDF5) gene. An association between this SNP and OA of the hip or knee was reported originally in Japanese and Chinese case–control cohorts (9). The T allele of this SNP in Asians is significantly increased in both patients with knee OA and those with hip OA, and in vitro cell transfection studies revealed that this allele mediates a significant reduction in the activity of the GDF5 promoter. A recent large-scale meta-analysis demonstrated that this SNP is also strongly and consistently associated with knee OA in European cohorts (10). However, in patients with hip OA, very large between-study heterogeneity in this association was observed, and statistical support for the involvement of this SNP in hip OA was only borderline (P = 0.016). One possible explanation is that there is substantial genetic heterogeneity between different forms of hip OA, and the presence of IP nodes may be one of the factors involved.
The objective of the present study was to investigate possible differences, according to nodal versus non-nodal status of generalized OA, in some of the genetic and nongenetic risk factors for the development of large joint OA of sufficient clinical severity to warrant total hip replacement (THR) or total knee replacement (TKR).
PATIENTS AND METHODS
Cases of THR and TKR were recruited from 2 case–control studies in Nottingham, UK, the Nottingham OA case–control study and the Genetics of Osteoarthritis and Lifestyle (GOAL) study. In addition, control subjects were obtained from both of these studies. In order to increase the sample size, and thus the statistical power, and, more importantly, to include individuals over the age of 60 years who were representative of the general population, we also included control subjects from a separate UK population-based cohort, the Hertfordshire study. (The descriptive characteristics of the cases and controls from all 3 cohorts, as well as the distribution of grades of radiographic severity of OA, are available at http://www.treatoa.eu/publicpdfs/SupplementaryTablesAr-10-0362.pdf.)
Nottingham OA and GOAL case–control studies.
Patients with hip or knee OA were recruited from hospital orthopaedic surgery lists (current cases as well as those in the previous 5 years) in the Nottingham area. Some of the participants for this study were originally recruited as part of a sibling cohort study (11, 12). All participants gave their written informed consent to take part. Approval for recruitment of knee and hip OA cases was obtained from the research ethics committees of Nottingham City Hospital and North Nottinghamshire. In all of the cases in the present study, patients had symptomatic, clinically severe hip or knee OA and had been referred to the hospital for joint replacement surgery, and the majority had undergone unilateral or bilateral THR or TKR within the previous 5 years.
Preoperative radiographs of the knee or pelvis of patients with knee or hip OA were examined to confirm the diagnosis and to grade for changes of OA (11–13). All pelvis and knee radiographs were scored by a single observer (SAD) for individual radiographic features of OA according to a standard atlas, with a grade range of 0–3, using the Kellgren/Lawrence (K/L) grading system for each knee or each hip joint (14). Self-reported ethnicity was assessed by a nurse-administered questionnaire, and only individuals of European descent were included in the genetic study. The presence of IP nodes in the fingers of each hand was ascertained by trained metrologists (SAD and MW) in all patients who underwent total joint replacement. Heberden's nodes were determined by clinical inspection and palpation for firm/hard posterolateral rounded swellings and/or joined dorsal bars, irrespective of any tenderness or symptom reporting, and classified as present or absent in each finger of both hands. Bouchard's nodes were similarly determined by inspection and palpation, and classified as present or absent. Further details on these clinical assessments are described elsewhere (3, 15). The nodal phenotype was defined as the presence of Heberden's and/or Bouchard's nodes that affected at least 2 rays of each hand.
Subjects ages 45–85 years who had undergone intravenous urography (IVU) in the same hospital were recruited as unrelated controls and underwent clinical examination and joint radiography. Only individuals with no symptoms and no clinical evidence of large joint OA (defined as the absence of swelling, duration of stiffness, if present, <30 minutes, and no crepitus) or no radiographic evidence of large joint OA were included as controls. In addition, for some of the proband joint replacement cases, unaffected siblings who were free from radiographic OA and whose age was older than 45 years were considered as controls. A maximum of 1 unaffected sibling per family was included among the controls. The allele frequencies between unaffected siblings and unrelated controls were compared, and no differences were detected.
For the GOAL study, patients with clinically severe knee or hip OA were recruited from joint replacement lists in an identical manner as that described above for the Nottingham case–control study. Subjects ages 45–85 years who had undergone IVU in the same hospital and who had no hip or knee symptoms were recruited as unrelated controls and underwent clinical examination and joint radiography. Only control subjects who had no clinical or radiographic signs (in both hips or both knees) of large joint OA were included in the present study. Nodal status was again assessed clinically by trained metrologists for all cases and controls from this cohort.
Medical Research Council (MRC) Hertfordshire cohort study.
The MRC Hertfordshire cohort study is a large population-based study designed to investigate the relationship between growth in infancy and the development of adult disease. Details of the study design have been published previously (16, 17). In brief, 3,000 men and women were recruited to the MRC Hertfordshire study, which included a home interview. In addition, a subgroup of patients (498 men and 468 women) underwent knee radiography. Ethics approval was obtained from the East and North Hertfordshire ethics committees, and all participants gave their written informed consent. Subjects who were taking or had previously taken bisphosphonate treatment were excluded.
Knee radiographs were graded individually for osteophytes and scleroses using a standard atlas, from which the K/L score was determined (14). For the present study, only subjects who had no radiographic evidence of knee OA (K/L grade <2 in both knees) and who had not reported any significant pain in or around the knees on most days of any given month over the last year were included as controls. Nodal status was not available for the Hertfordshire study subjects, but the presence or absence of Heberden's or Bouchard's nodes was reported.
Clinical and laboratory analyses.
For all study participants, age at the time of the clinic visit was recorded, and weight and height (determined with a simple standard tape measure) were determined to compute the body mass index (BMI). For 3,096 total joint replacement cases and 1,168 controls, both height and BMI were recorded, but for 827 of the control subjects (from the Nottingham and GOAL studies) and 704 of the cases of total joint replacement, only information on BMI was available in the records.
Genomic DNA was extracted from the peripheral blood leukocytes of patients and control subjects using standard protocols. Genotyping was carried out at Kbioscience in Hertfordshire, UK. SNPs were genotyped using KASPar chemistry, which is a competitive, allele-specific polymerase chain reaction SNP genotyping system that utilizes fluorescence resonance energy transfer quencher cassette oligonucleotides.
For continuous variables, the mean ± SD values were compared between patients with nodal OA and patients with non-nodal OA who underwent total joint replacement, using analysis of variance. The proportions of each sex and the frequencies of the GDF5 polymorphism rs143383 (T allele) were computed, and statistically significant differences were assessed using Pearson's chi-square test. The odds ratios (ORs) were calculated for comparisons of the risk of total joint replacement between cases and controls, with cases classified according to the nodal OA status, i.e., nodal versus non-nodal OA. Logistic regression was used to adjust for confounding factors. Risk factors included in this analysis were the presence of the GDF5 polymorphism, age, sex, and BMI. The OR for height, which was derived from a smaller number of samples (see Table 1), was adjusted for age and sex only. Given the strong correlation between height and BMI, the BMI value was not included, to avoid the problem of multicollinearity.
Table 1. Risk factors for total joint replacement in nodal versus non-nodal large joint osteoarthritis (OA) cases compared with controls from the 3 studies*
Except where indicated otherwise, values are the mean ± SD (95% confidence interval). GOAL = Genetics of Osteoarthritis and Lifestyle (study). P values for comparison between nodal and non-nodal cases were derived by analysis of variance for body mass index (BMI), age, and height, and by Pearson's chi-square test for female sex and presence of the GDF5 polymorphism.
Adjusted for age and sex.
Only a subset of patients had available data on height: for controls, n = 1,168; for total hip replacement (THR), nodal cases n = 473 and non-nodal cases n = 959; for total knee replacement (TKR), nodal cases n = 598 and non-nodal cases n = 1,260.
All analyses were carried out using S-Plus 2000 (Insightful Inc.). To assess the statistical significance of the difference in ORs between cases of nodal OA and cases of non-nodal OA or between patients who underwent TKR and those who underwent THR, the P value for the Z statistic was computed, as follows: z = |LnORi − LnORj|/√(SE2i − SE2j), where SEi and SEj denote the standard error of the log OR (the LnORi and LnORj) for nodal and non-nodal disease, respectively or for TKR and THR, respectively.
In total, 2,142 patients who underwent TKR (673 nodal OA and 1,469 non-nodal OA cases) and 1,879 patients who underwent THR (625 nodal OA and 1,254 non-nodal OA cases) were studied and compared with 1,906 control subjects. The numbers of samples from each study are shown in Table 1.
All 5 risk factors considered (BMI, age, sex, height, and presence of the GDF5 T allele) had significantly different mean values in patients with nodal OA who underwent THR when compared with patients with non-nodal OA who underwent THR. In contrast, among patients who underwent TKR, only age, sex, and BMI had significantly different mean values between the nodal and non-nodal forms of the disease (Table 1).
We explored which of these factors influenced the risk of total joint replacement in patients with nodal OA and patients with non-nodal OA who underwent either TKR or THR as compared with that in controls. The results, as shown in Table 2, indicated differences in the association of the risk factors with total joint replacement between each form of OA. Age was associated with total joint replacement in both the nodal and non-nodal forms of OA, but the association was significantly greater in patients with nodal OA. Female sex was also associated with the nodal OA form in those undergoing total joint replacement, but was not associated with an increased risk of THR in non-nodal cases. Female sex was actually a protective factor for TKR in non-nodal cases. BMI was a strong risk factor for both types of total joint replacement, but the OR for either form of OA in TKR was significantly higher than that for either form of OA in THR, and the OR in patients with the non-nodal form was significantly higher than that in patients with nodal disease. Age was a significantly stronger risk factor for TKR than for THR.
Table 2. Odds ratios (ORs) for the likelihood of total joint replacement in nodal versus non-nodal large joint osteoarthritis (OA) cases compared with controls*
ORs with 95% confidence intervals (95% CIs) from the multivariable model with age, height, and body mass index (BMI) are per standard deviation, while those for the GDF5 polymorphism are per T allele carried.
P values represent the Z-statistic comparison between nodal and non-nodal disease or between patients who underwent total hip replacement (THR) and those who underwent total knee replacement (TKR).
Increased height appeared to significantly increase the risk of non-nodal hip OA, but this did not appear to be a factor in either the nodal or non-nodal forms of knee OA. A technical point to consider is the estimate of effect size for height as a risk factor, which was not adjusted for BMI to avoid introducing multicollinearity into the model. We observed a difference in height as a risk factor for THR in patients with the non-nodal form of OA, but not in those with the nodal form. The effect size estimates for height were very similar when BMI was included in the model and when BMI was not included, for both forms of OA in patients undergoing THR (P > 0.35). In fact, the OR for height as a risk factor for THR was 1.30 (95% confidence interval [95% CI] 1.15–1.48) in non-nodal OA cases and 0.97 (95% CI 0.82–1.15) in nodal OA cases, when adjusted for age, sex, and BMI, and the difference between these 2 estimates was significant (P < 0.007), i.e., stronger than when not adjusted for BMI. Thus, the difference observed is not a statistical artifact due to the inclusion or exclusion of BMI in the statistical model.
Finally, carriage of the T allele at the rs143383 SNP was a significant risk factor in patients with either the nodal or non-nodal form of OA who underwent TKR. However, in patients who underwent THR, this was only significant in those with the nodal form, showing no effect on THR in non-nodal OA cases. Carriage of the T allele was not associated with nodal status in controls (OR 0.99, 95% CI 0.60–1.61; P < 0.96).
We further investigated whether the nodal and non-nodal forms of large joint OA had different levels of severity, as measured by the prevalence of bilateral joint replacement and by the prevalence of total joint replacement at both the hip and the knee (regardless of bilateral status). The results, as shown in Figure 1, indicated that, as expected, patients who underwent total joint replacement and had the nodal phenotype were at a higher risk of undergoing both THR and TKR compared with those with non-nodal OA. Similarly, patients who underwent TKR and had the nodal form of the disease were significantly more likely than those with non-nodal OA to have undergone a bilateral knee replacement. In contrast, patients who underwent THR and had the nodal form of OA were significantly less likely than those with non-nodal OA to have undergone a bilateral hip replacement.
Our results indicate that nodal and non-nodal forms of severe hip OA have different genetic and nongenetic risk factor profiles and differ in the likelihood of the requirement for bilateral THR. There are also differences in risk profiles between patients with nodal OA and patients with non-nodal OA who underwent TKR, but only with respect to age, sex, and BMI. Our results provide evidence that the nodal phenotype is associated with a higher prevalence of OA at more than 1 joint, as shown by 1) the higher risk of undergoing both a hip and a knee replacement in nodal cases of large joint OA, and 2) the higher risk of bilateral TKR in nodal cases.
However, some of our findings appear counterintuitive. For example, bilateral THR was associated more with non-nodal hip OA than with nodal hip OA, and in spite of the presumed larger genetic component of nodal OA, both patients with nodal disease undergoing TKR and those with nodal disease undergoing THR were older at the time of recruitment than their counterparts with non-nodal OA. Such findings suggest that the relationship between nodal status, severity, and age at onset is complex. It is possible that the risk of both bilateral THR and younger age at surgery is also associated with genetic traits, but these traits may differ from those associated with nodal OA. We speculate that part of the reason for this observation could be that hip OA is strongly governed by local morphologic characteristics that result in biomechanical stress (e.g., mild acetabular dysplasia, nonspherical femoral head shape, neck-shaft angle) and that the genetic and programming constituents of these parameters are likely to differ from those of generalized OA, which may reflect more widespread abnormalities of joint tissue constituents (e.g., bone, cartilage, muscle). Consistent with this hypothesis, morphologic traits have been implicated in the risk of hip OA independent of nodal status (18).
Contrary to expectation, we found no differences, at least for the GDF5 gene, in the risk conferred between patients with nodal OA and patients with non-nodal OA who underwent TKR, suggesting that the non-nodal form of knee OA is affected by genes similar to those affecting the nodal forms. In addition, our findings show that height can be a risk factor for some forms of OA. These data are consistent with results from a Swedish population-based prospective cohort study in which height was shown to be a significant risk factor in hip and knee OA (19). Patients with non-nodal OA were taller than those with nodal OA. This is not surprising, given the higher proportion of men and the younger age among the non-nodal cases. However, in the case of THR, this difference in height was statistically significant, even after adjusting for age and sex. In addition, height was a significant risk factor in patients with non-nodal OA who underwent THR as compared with that in controls, in spite of the fact that controls were 3.5–5 years younger than the patients with non-nodal OA who underwent THR. The sex proportions were the same between the groups. This association with height further suggests a different etiology from that of other forms of severe OA.
The main strength of the present study is the large number of cases that were identified and assessed in the same way. However, there are some potential limitations. Not all controls could be excluded on the basis of radiographic evidence of hip OA and not all were characterized for nodal status. Nevertheless, this would seem unlikely to have had an effect on the main conclusion regarding the differences between nodal and non-nodal OA. The end point in this study was large joint OA that was severe enough to require a total joint arthroplasty, and therefore, in that sense, all controls were truly free from disease as defined by our inclusion criteria. Furthermore, such misclassification would have the effect of underestimating risk factors but would not have an impact on the differences between nodal and non-nodal cases, which is the focus of this study. When the same analyses of the risk of THR were run with exclusion of controls for whom no assessment of hip radiographic OA was done, essentially the same results as those presented herein were obtained. Moreover, exclusion from the analysis of controls who were siblings of proband cases and exclusion of all controls with the nodal phenotype or with no nodal characterization all yielded the same qualitative results as those presented herein. Thus, the differences in recruitment criteria for the various control sets were not a source of bias with regard to the validity of our conclusions.
A second potential limitation is the choice of total joint replacement as a clinical end point, which was used to select a cohort of patients with clinically relevant OA. Indications for total joint replacement may vary between centers, doctors, and care settings, but in the present study, the cases clearly were representative of an OA population with clinically relevant disease (details available at http://www.treatoa.eu/publicpdfs/SupplementaryTablesAr-10-0362.pdf), which is suitable for the primary analysis of differences between OA cases with or without the nodal phenotype. Nevertheless, the fact that the study focused on patients who underwent total joint replacement may limit the generalizability of our results. Cases of total joint replacement represent only a fraction of the overall population with severe, symptomatic large joint OA. Among subjects over the age of 60 years in the UK, the prevalence of the need for total joint replacement (excluding those who actually receive total joint replacement) is between 12.4% and 16% (20, 21), and the prevalence of receipt of total joint replacement is ∼6% (20), which, when added to the prevalence of the need for total joint replacement, yields a prevalence of total joint replacement with or without receipt of surgery of between 18.4% and 22%. Thus, 27–33% of individuals who require a total joint replacement receive one.
Our results should be applicable to at least one-fourth of individuals with severe, symptomatic large joint OA, which is a high enough proportion of such cases to be of merit and is consistent with the rates of transition from hip pain to being listed for total joint replacement (23%) in primary care patients over a 4-year period in a prospective UK cohort study (22). Factors such as wealth, education, geographic region, smoking status, obesity, and sex (21, 23) have been reported to influence rates of access to total joint replacement, and recent studies have revealed considerable heterogeneity in the radiographic severity, functional disability, and pain in candidates for total joint replacement (24). Thus, for example, patients undergoing total joint replacement may be generally fitter than people with both clinically severe large joint OA and severe comorbidity, and our conclusions may not hold among nonsurgical severe large joint OA cases. However, it is recommended that comorbidity should not be a reason to not refer for surgery (25), and advances in anesthetic technology and perioperative care have resulted in more patients with comorbidity undergoing surgery. A population-based study with broad inclusion is required to examine the validity of the results in nonsurgical severe large joint OA.
A major difference from previous studies is the large sample size, with 3,800 severe large joint replacement cases being evaluated. As such, it is the first large study to attempt a detailed comparison of nodal versus non-nodal severe large joint OA. Nevertheless, a number of smaller studies have tried to describe nodal generalized OA. Kellgren and Moore were the first to describe generalized OA, noting its predominance in middle-aged women and the apparent absence of marked obesity as a feature (7). The higher prevalence among women was confirmed herein, but we found that the age at which patients with nodal OA underwent total joint replacement was older than that in patients with non-nodal OA, and that age was a stronger risk factor for total joint replacement in nodal disease. In the GOAL study, the average age at first total joint replacement was 66.8 years for nodal OA cases and 64.7 years for non-nodal OA cases. Furthermore, although BMI is a strong risk factor for all forms of large joint OA, it is greater in knee OA than in hip OA and also is a stronger risk factor in non-nodal cases than in nodal cases.
The fact that nodal and non-nodal large joint OA may have a different pathogenesis has been suggested by data showing that chondroitin sulfate disaccharide levels in synovial fluid are significantly different in patients who have nodal OA of the hand in addition to knee OA as compared with those with non-nodal knee OA (26). However, further studies investigating the differences in markers of joint damage, inflammation, or other metabolic processes between the 2 forms of OA have not been reported.
It has been suggested that generalized OA has a stronger genetic component than other forms of OA (27). The current data support a different genetic contribution, at least for one genetic variant, for hip OA but not for knee OA. It is worth noting that radiographic hip OA has a strong familial aggregation (λs 4.27–5.07) (12) and a strong heritability (60%) (28), whereas radiographic knee OA has a weaker familial aggregation (λs 1.66–2.13) (11) and a weaker heritability (39%) (29). In spite of this, the strongest genetic associations with large joint OA reported to date (GDF5, the 7q22 cluster) are all with knee OA (30), and the reported associations with hip OA (31) are much more modest. These observations suggest that there may be a considerable amount of heterogeneity in the hip OA phenotype that is not captured by simple definitions like total hip replacement or radiographic grade.
Although the association between hand and knee OA in people with nodal generalized OA is well accepted, the inclusion of the hip as a target site within the generalized OA concept is controversial. Kellgren and Lawrence initially noted that the hip was uncommonly involved in nodal generalized OA, and several subsequent studies also failed to show a clear association between the presence of nodes and hip OA (32–35). However, in contrast, one study has demonstrated an association between the presence of nodes and hip OA (36), and 2 studies have shown an association between radiographic hand OA and hip OA (37, 38).
The data presented herein indicate that 2 forms of severe large joint OA have a different etiology. Treatment options for severe OA are very limited. There are no disease-modifying drugs, and use of biochemical markers and imaging to assess progression are suboptimal. If pharmacologic interventions or monitoring tools are going to work, it is important to understand that severe large joint OA is not a single disease and that the pathogenesis, and thus the molecular and environmental factors determining progression, is likely to be different. Further studies to characterize subsets within large joint OA seem warranted.
In conclusion, our findings confirm that severe large joint OA is not a single disease entity, but rather represents a group of disorders with a similar pathologic end point. Classification according to nodal status appears to be useful for separating 2 groups with different risk factor profiles and different clinical outcomes.
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. Valdes 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. Valdes, Arden, Muir, Maciewicz, M. Doherty.
Acquisition of data. Valdes, Arden, S. A. Doherty, Wheeler, Muir, Zhang, Cooper, M. Doherty.
Analysis and interpretation of data. Valdes, McWilliams, Arden, S. A. Doherty, Zhang, M. Doherty.