As populations age and the prevalence of hip osteoarthritis (OA) increases, health care providers must manage increasing demands for services. Evidence regarding the progression of hip OA can assist health care practitioners in determining expected patient prognosis and planning care. This systematic review of prospective cohort studies examines prognostic variables in patients with hip OA.
Articles were selected following a comprehensive search of Medline, EMBase, CINAHL, and Allied and Complementary Medicine from database inception to October 2008 and hand searches of the reference lists of retrieved articles. Inclusion criteria involved 1) estimates of the association between prognostic variables and progression of OA, 2) prospective cohort design, 3) patients diagnosed with hip OA based on established criteria, 4) at least 1 year of followup, and 5) access to the full published text. Two independent reviewers assessed the methodologic quality of each study and the association between prognostic variables and OA progression.
Eighteen articles met the inclusion criteria; 17 were considered to be of high quality. Strong evidence of progression was associated with age, joint space width at entry, femoral head migration, femoral osteophytes, bony sclerosis, Kellgren/Lawrence hip grade 3, baseline hip pain, and Lequesne index score ≥10. Strong evidence of no association with progression was associated with acetabular osteophytes. Evidence was weak or inconclusive regarding associations between various other radiographic or clinical variables, molecular biomarkers, or use of nonsteroidal inflammatory drugs.
Overall, few variables were found to be strongly associated with the progression of hip OA, and a variety of other variables were weakly predictive of outcome.
Osteoarthritis (OA) is the most common form of arthropathy (1), and is second only to heart disease as the predominant cause of functional decline in the elderly (2). OA of the hip affects approximately 5% of the population age >65 years, resulting in nearly 200,000 total hip replacements per year in the US (1). It has been estimated that the aging population will give rise to a higher prevalence of disabling OA because the number of people age >60 years is expected to increase by 20–33% by 2030 (3). An aging population, along with escalations in obesity and physical inactivity (4, 5), would increase the economic burden to society of disablement due to OA. It has been estimated that loss of consumer and occupational productivity, diagnostic services, pharmacologic and nonpharmacologic therapies, and surgical interventions due to OA cost approximately 0.7% of the US gross domestic product (5, 6).
OA is a chronic disease that manifests inconsistently in those afflicted. Recent prognostic studies have investigated variables that are associated with accelerated or delayed progression of joint destruction or functional loss due to OA (7–9). Prognostic studies provide patients, physicians, and third-party payers with expectations in regard to the course of symptoms, and help distinguish between patients who are at high risk for worsening pain and disability versus those with a more favorable clinical course.
The objective of this study was to systematically review the evidence regarding useful prognostic variables associated with the progression of hip OA. This review involved a systematic review of available articles using contemporary methods of identification and assessment of the available evidence (10, 11). The findings may assist in outlining effective diagnostic and intervention strategies for patients with OA.
MATERIALS AND METHODS
Identification and selection of the literature
We conducted a systematic, computerized search of the literature based on recommendations by Wilczynski and Haynes (10, 11) in Medline (1950 to October 2008), EMBase (1988 to October 2008), CINAHL (1951 to October 2008), and Allied and Complementary Medicine (AMED) (1985 to October 2008) (the search strategy is shown in Supplementary Appendix A, available in the online version of this article at http://www3.interscience.wiley.com/journal/77005015/home). The reference lists of all selected publications were checked to retrieve relevant publications that were not identified in the computerized search. To identify relevant articles, titles and abstracts of all identified citations were independently screened by 2 reviewers (AAW, JHA). Full-text articles were retrieved if the abstract provided insufficient information to establish eligibility or if the article had passed the first eligibility screening.
An article was eligible if it met all of the following criteria: 1) the statistical association of at least 1 prognostic variable with the outcome of interest reported, 2) derived from a prospective cohort of subjects, 3) included radiographic or clinical evidence of hip OA based on established criteria, 4) the outcome of interest was radiologic and/or clinical progression of hip OA, 5) the followup period was at least 1 year, and 6) the article was available in full text. No language restrictions were imposed.
An article was excluded if 1) the study population principally included patients with secondary OA, including major congenital or developmental diseases and bone dysplasias; metabolic diseases associated with joint disease such as hemochromatosis, rheumatoid arthritis, tuberculosis, ankylosing spondylitis, sickle cell disease, and Cushing's syndrome; infection; other bone diseases such as Perthes disease and osteochondritis; femoral head necrosis; and trauma (fractures); or 2) the article discussed prognosis following joint replacement surgery, osteotomy, or other specific surgery.
All criteria were independently applied by 2 reviewers (AAW, JHA) to the full text of the articles that passed the first eligibility screening. In case of disagreement, a consensus method was used to discuss and solve the disagreement.
The methodologic quality of each of the studies was independently assessed by the same 2 reviewers (AAW, JHA). Reviewers were not masked to trial identifiers such as author and journal names. To our knowledge, there is presently no consensus standard for the assessment of prognostic studies; therefore, we used a modified version of checklists used in other systematic reviews of prognostic variables in musculoskeletal disorders (12–14) that reflected the important methodologic aspects (15). The final checklist consisted of 18 items (Table 1), with each having a “yes”/“no”/“don't know” answer option. A “yes” score indicated sufficient information and a positive assessment, with bias considered unlikely. A “no” score indicated sufficient information, but with potential bias from inadequate design or conduct. A “don't know” score indicated that insufficient information was provided in the article or the methodology was unclear. Disagreements among the reviewers were discussed during a consensus meeting and, where unresolved, were resolved by a third reviewer (CC). A detailed explanation of each of the criteria is available from the corresponding author.
Table 1. Criteria list for the methodologic quality assessment of studies on prognostic factors in patients with hip osteoarthritis
+ = positive (sufficient information and a positive assessment); − = negative (sufficient information, but potential bias due to inadequate design or conduct); ? = unclear (insufficient information).
Description of study population
Description of inclusion and exclusion criteria
Response of ≥75% for cohorts and controls
Followup of at least 12 months
Dropouts/loss to followup <20%
Information completers vs. loss to followup/dropouts
Prospective data collection
Treatment in cohort is fully described/standardized
Clinically relevant potential prognostic factors
Standardized or valid measurements
Data presentation of the most important prognostic factors
Clinically relevant outcome measures
Standardized or valid measurements
Data presentation of the most important outcome measures
Appropriate analysis techniques
Prognostic model is presented
The maximum attainable score on the criteria list was 18. The total score was the count of all of the criteria that scored “yes.” “No” and “don't know” scores carried a zero score value. For each study, a total quality score was given based on the information from all of the available publications. A priori, we chose to consider a study to be of high quality if it scored ≥12 points (≥66.6% of the maximum attainable score), and of low quality if its score was <12 points. The cut point score was arbitrary but similar to other previously mentioned systematic reviews on prognosis (13, 14).
One reviewer (AAW) independently extracted information and data regarding study population, setting, outcome measures, prognostic variables, and strength of association statistics associated with the prognostic variables. Another reviewer (JHA) reviewed and confirmed the abstracted results. The second reviewer (JHA) was not blinded to the results abstracted by the first reviewer (AAW).
Interobserver agreement of quality assessment was assessed using kappa statistics. Because unadjusted kappa can provide misleading results when the sample size is small (16) or if the data are highly symmetric or unbalanced, despite high observed agreement (17), we calculated an adjusted kappa. The adjusted kappa was calculated from the unadjusted kappa divided by the maximum kappa (16).
We tabulated the available evidence for each prognostic variable by reporting the number of articles evaluating each variable, the methodologic quality of the articles, and the strength and grade of the available evidence. The strength of evidence for prognostic variables associated with progression of OA of the hip was assessed by defining 4 levels of evidence based on those by Phillips et al (18) (Tables 2 and 3). Findings were reported as relative risks, odds ratios, hazard ratios, or P values.
Table 2. Levels of evidence for prognostic factors on hip osteoarthritis
1. Evidence obtained from high-quality cohort studies
2. Evidence obtained from lesser-quality cohort studies
3. Case–controlled or retrospective studies
4. Case series
5. Expert opinion
Table 3. Grades of evidence for prognostic factors on hip osteoarthritis*
Grades of recommendation
Modified from the Oxford Centre for Evidence-Based Medicine (18).
A. Strong evidence
Consistent findings (≥75%) in ≥2 high-quality cohort studies
B. Moderate evidence
Consistent findings (≥75%) in 1 high-quality cohort and ≥1 low-quality cohort
C. Weak evidence
Findings in 1 high-quality cohort study or consistent findings (≥75%) in ≥3 low-quality cohorts
D. Conflicting evidence
Inconsistent or inconclusive studies of any level or 1 low-quality study
E. Theoretical/foundational evidence
No data presented
F. Expert opinion
No data presented
Selection of studies
Initially, the search yielded 3,018 citations (Medline 1,298, EMBase 1,530, CINAHL 124, and AMED 66). Of these, 363 duplicates were deleted, leaving 2,655 titles with abstracts for review. After the first screening, the full-text studies of 40 potentially eligible citations were retrieved. Following a consensus meeting, a total of 18 studies were included in the review (7–9, 19–33) (Figure 1). Reference checking did not provide any additional studies.
The 2 primary reviewers (AAW, JHA) required clarification from a third reviewer (CC) for the interpretation of 2 quality assessment criteria (item A and item R). Following clarification, the adjusted interobserver agreement was κ = 0.68, calculated from the unadjusted κ = 0.51 and maximum κ = 0.75. This finding represents substantial agreement (34). Items E and H were not included because all studies scored “yes” based on our inclusion criteria.
Disagreements occurred mainly because of reading errors and misinterpretation of the methodologic criteria list and were readily resolved, with only 1 disagreement persisting (item I: Ledingham et al ). The third reviewer (CC) made the final decision in this case. The results of the quality assessment are shown in Table 4. The studies are ranked in descending order by methodologic quality score. The overall quality score ranged from 10 to 17 points and 17 studies were classified as high-quality studies. The median score was 13 points (72.2%).
Most methodologic shortcomings concerned the following items: description of inception cohort, failure to describe response rate, failure to report numbers lost to followup, failure to provide information on completers versus those lost to followup, and failure to determine a collection of prognostic variables with the highest prognostic value using multivariate techniques.
Supplementary Appendix B outlines the characteristics of the articles, including study population, outcome measures, followup, prognostic variables, and the strength of association with outcome (estimates and 95% confidence intervals) (Supplementary Appendix B, available in the online version of this article at http://www3.interscience.wiley.com/journal/77005015/home). We found that several studies were conducted in a common cohort of subjects, and so grouped according to the study cohorts (Evaluation of the Chondromodulating Effect of Diacerein in Osteoarthritis of the Hip [7,8,19,20,22], Study of Osteoporotic Fractures [9, 24, 25, 27, 32, 35], and the Rotterdam Study [28–31]). Of the 18 different articles, 7 included patients referred from rheumatologists, 10 from the general population, and 1 from both rheumatology and orthopedic clinics.
There was considerable variation among the studies with respect to sample size, length of followup, diagnostic criteria, and definition of progression. The sample size ranged from 48 to 1,904, with 12 studies enrolling >300 subjects and 4 enrolling >1,000 subjects (28–31). The length of followup ranged from 12 months to 8 years. OA diagnosis at baseline was determined based on 3 different criteria: American College of Rheumatology (36), Croft summary grade, and Kellgren/Lawrence (K/L) grade. Four studies defined the progression of OA as joint space narrowing (JSN) alone, although the studies varied in their definition of progression (0.5, 0.6, 1.0, or 1.5 mm/year). Two studies defined OA progression as total hip arthroplasty (THA) alone, and 3 used a combination of JSN or THA. The remaining 9 studies used a combination of JSN, THA, increased osteophyte score, bony sclerosis, or an increase in a Croft summary grade or K/L grade to define OA progression. The percentage lost to followup varied between 0% and 40%. The most frequently reported prognostic variables were age, sex, and joint space width (JSW).
Table 5 shows a summary of the evidence for the prognostic variables reported with regard to each associated outcome and overall level of evidence.
Table 5. Overall level of evidence for prognostic factors and their association with long-term outcome*
Eight variables demonstrated the strongest level of evidence (Level 1A) and were predictive of the progression of hip OA: age (7, 8, 19, 29), JSW at entry (7, 8, 19, 20, 29), femoral head migration (7–9, 19, 20, 32, 33, 35), femoral osteophytes (9, 32, 35), bony sclerosis (9, 20), K/L hip grade 3 (8, 21, 33), baseline hip pain (8, 9, 29, 32, 33, 35), and Lequesne index score ≥10 (8, 19).
Acetabular osteophytes (9, 20, 32, 35) demonstrated strong (Level 1A) evidence for no association with the progression of hip OA.
Level 1C evidence: weak.
The following variables were assigned the score of Level 1C because each was reported in only 1 high-quality study: acetabular and femoral osteophytes (9), K/L hip grade ≥2 (29), K/L hip grade ≥2 plus hip pain (29), K/L hip grade 4 (21), a combination of radiographic grade 3 or 4 plus previous nonsteroidal antiinflammatory drug (NSAID) use and global assessment above the median (21), functional impairment ≥2 (7), mean global patient assessment >47 over the first 6 months (21), disability index score ≥0.5 (29), restricted flexion >20% (29), molecular biomarker C-terminal crosslinking telopeptide of type II collagen (CTX-II) >346 ng/mmoles of urinary creatinine (7), molecular biomarker serum hyaluronic acid >137 mg/ml (7), CTX-II >346 ng/mmoles of urinary creatinine plus serum hyaluronic acid >137 mg/ml (7), and baseline hip pain plus diclofenac >180 days (31) are predictive of the progression of hip OA.
Body mass index (30), the presence of cysts (9), molecular biomarker CTX-II (28), molecular biomarker Frizzled-related protein (27), baseline serum N-telopeptide crosslinks (NTX) (26), percent change in serum NTX (25), statin user (24), previous NSAID intake (21), ibuprofen (>30 days) (31), naproxen (>30 days) (31), diclofenac (>30 days) (31), and piroxicam (>30 days) (31) demonstrated weak (Level 1C) evidence of no association with the progression of hip OA.
Molecular biomarker Dkk-1 (27) and percent change in serum cartilage oligomeric matrix protein (COMP) (25) demonstrated weak (Level 1C) evidence of reduced progression of hip OA.
Level 1D evidence: inconclusive.
There is inconclusive (Level 1D) evidence that female sex (7, 8, 19, 22, 29, 33) and baseline serum COMP (23, 26) are associated with the progression of hip OA.
Level 2D evidence: inconclusive.
Seven variables demonstrated inconclusive (Level 2D) evidence and were predictive for the progression of hip OA: atrophic bone response (33), mild OA (33), any radiographic change (33), rapid radiographic change (33), night pain (33), patient assessment of change (33), and a decrease in exercise tolerance (33). One variable, male sex (33), demonstrated inconclusive (Level 2D) evidence for no progression of hip OA. Intermediate bone response (33) demonstrated inconclusive (Level 2D) evidence for no association with the progression of hip OA.
This systematic review has summarized the results of 18 prospective cohort studies reporting the prognostic value of various variables for the progression of radiographic JSN or THA. The evidence is weak or inconclusive on most prognostic variables because of inconsistencies in the way data is collected and measured, dissimilar outcome measures, and when variables were reported in few articles. Heterogeneity in diagnostic criteria, definitions of progression, and variations in the definitions of the prognostic variables themselves prevent us from performing a meta-analysis; therefore, each article must be described separately. This resulted in weak or inconclusive evidence for many variables. However, there was strong evidence that age, JSW at entry, femoral head migration, femoral osteophytes, bony sclerosis, K/L hip grade 3, baseline hip pain, and Lequesne index score ≥10 are predictive of poorer outcome or progression of hip OA. Similarly, there was strong evidence that acetabular osteophytes demonstrate no association with the progression of hip OA.
Our findings updated those reported in an earlier review performed in 2002 (12). In contrast to the previously reported review, we did not restrict our search by publication language, we used the search strategy guidelines by Wilczynski and Haynes (10, 11), and we used contemporary methods for the identification and assessment of available evidence. In contrast to the findings of Lievense et al (12), we found a higher age at baseline assessment and JSW at entry to demonstrate strong evidence of poorer outcome or progression of hip OA. We could not confirm the strong prognostic value of atrophic bone response for progression. Furthermore, we found that body mass index demonstrated weak (Level 1C) versus strong evidence for no association with progression. These discrepancies are likely associated with the more recently available evidence and our inclusion and exclusion criteria, which included only prospective study designs and excluded patients with congenital or developmental bone dysplasias. We only included studies with a prospective study design because this is considered to be the optimal design to minimize bias in the association between prognostic variables and outcome. We believed that a followup period of >1 year was needed in order to study the course of chronic hip OA. Therefore, some otherwise relevant studies (37, 38) were not eligible for our review.
There were some limitations of this review. Of the 18 articles included, 15 were derived from 3 patient cohorts. Therefore, we decided to evaluate and report the findings from each individual article. Separate reporting may risk bias resulting from statistical associations reported twice from the same cohort. The lack of independent study cohorts highlights that prognostic research regarding variables associated with the progression of hip OA is understudied and/or is a focus of few research teams. As a result, we advise caution in interpreting the level of evidence for the Lequesne index (8, 19) and femoral osteophytes (9, 32, 35). Furthermore, publications from common cohorts may receive higher-quality scores due to the reliability of information reported in earlier publications that may have otherwise not been obtained within each individual article.
Consensus on a validated methodologic quality checklist for prognostic studies is yet to be established, and therefore our list was based on previous reviews and recommendations described by Scholten-Peeters et al (14), Kuijpers et al (13), Hayden et al (39), and Altman (15). Also, our cut point of ≥12 (66.6%) to define a high-quality study was arbitrary. Although this is similar to or higher than other prognostic studies, only one of the included articles was rated as low quality.
Items E (followup >12 months) and H (prospective study design) of our methodologic quality list were part of our inclusion criteria, so perhaps they could have been deleted from scoring. If these items had been excluded, 3 articles would have been below the 66.6% quality assessment cut point, and would therefore be rated as low quality. Eliminating items E and H would have downgraded some of our conclusions, with NSAID variables then assigned inconclusive (Level 1D) evidence versus weak (Level 1C) evidence, and femoral osteophytes assigned moderate (Level 1B) evidence versus strong (Level 1A) evidence. A higher cut point such as 75% would have had a similar effect. However, our cut point was consistent with other studies (13, 14), and we reasoned that deleting those methodologic quality items would have artificially deflated the quality score because those studies included earned the quality points of items E and H on merit. We believe our quality assessment criteria were rigorous (12–14) and our strict inclusion criteria strengthens our findings because retrospective studies, which carry a greater risk of bias, were eliminated from the review.
Information about prognostic variables can guide physicians and other health care providers toward providing the most appropriate treatment options. It has been recommended that clinicians conduct a holistic assessment of patients and provide patients with information about the condition, its management, and prognosis (40). The knowledge that age, JSW at entry, femoral head migration, femoral osteophytes, bony sclerosis, K/L hip grade 3, baseline hip pain, and Lequesne index score ≥10 are predictive of poorer outcome will help clinicians provide realistic, individually tailored advice, dispel myths, and better plan care for their patients.
The results of our review suggest that patients with more severe joint damage at study entry, hip pain at baseline, and who report poorer function progress more rapidly. These results are consistent with the contemporary understanding that loss of cartilage results in uneven and localized excess loading, which in turn accelerates damage to the joint (41). Joint damage includes JSN, osteophytosis, and malalignment, which can then lead to inflammation, inhibition of muscle activity, and pain (41). This information may suggest a nonlinear relationship in the progression of hip OA, in that it is the additive effect of coexisting features that lead to accelerated progression.
However, predicting outcome from hip pain and joint damage at baseline remains inexact, because although there is strong evidence that baseline pain and progressive joint deterioration are associated, the strength of that association is modest in magnitude (Table 5). It is not uncommon for people with evidence of joint pathology to be pain-free (41). Also relatively weak in magnitude was the association of JSW at baseline with JSN at followup, which is surprising in light of the potential circularity in using as a predictive variable the same variable used to define outcome. This form of incorporation bias may highlight the consistent use of radiographic markers as predictors of morbidity progression and demonstrates a need to avoid confounding the relationship between the prognostic variables and the outcomes in research design. Adjusted odds ratios from multivariate statistical models were seen in disappointingly few of the included studies, and only 2 studies sought to estimate the incremental prognostic power of combining radiographic data with other clinical data (22, 29).
JSN is considered the gold standard for monitoring the progression of hip OA (42); however, given the variable relationship between radiographic appearance, pain, and function (41), better prognostic information may result from a focus on what matters to patients, such as loss of physical function and independence rather than simply measures of joint structure damage. Botha-Scheepers et al (43) suggest that in order to understand the entire impact OA has on the lives of patients, a multiperspective approach is needed that evaluates not only the impairment of body structures (JSW, JSN) but also the components of body functions (pain, joint range of motion, muscle strength), limitation of activities (walking, occupation), and participation in societal activities (social involvement).
A variety of molecular biomarkers have been investigated for potential prognostic value; however, most have been studied in only one cohort. Even in those that have been studied in more than one cohort, such as CTX-II and COMP, heterogeneous analysis methods and cut points, as well as weak, inconclusive, and inconsistent evidence, fail to demonstrate convincing prognostic value. It is possible that other predictors exist, but at present evidence does not exist to support this assumption. We found little or no evidence regarding the prognostic value of common clinical information such as psychosocial variables, clinical tests and measures such as strength and range of motion, and physical performance measures.
The current understanding of prognostic variables influencing the clinical course and progression of hip OA is not comprehensive. We suggest that further high-quality prognostic studies, i.e., prospective cohort studies with a followup period of at least 1 year, are needed to generate valid and informative prognostic variables. Currently, the available data are at risk of bias through multiple reports from a few common cohorts, and the addition of one new study can have a significant impact on the results. We recommend that future studies investigate the predictive value of psychosocial variables, clinical tests, and physical performance measures such as range of motion, strength, and flexibility, and address outcome in patient-centered terms, such as physical function decline and health-related quality of life.
There is strong evidence that age, JSW at entry, femoral head migration, femoral osteophytes, bony sclerosis, K/L hip grade 3, baseline hip pain, and Lequesne index score ≥10 are predictive of the progression of hip OA and that acetabular osteophytes have no association with the progression of hip OA. There is weak or inconclusive evidence or no available high-quality evidence regarding the prognostic value of a range of other potential prognostic variables. These conclusions demonstrate a need for further high-quality research in the area of prognosis and hip OA.
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. Wright 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. Wright, Abbott.
Acquisition of data. Wright.
Analysis and interpretation of data. Wright, Cook, Abbott.