Objective To develop a radiographic score for assessment of hand osteoarthritis (OA) that is based on histopathological alterations of the distal (DIP) and proximal (PIP) interphalangeal joints.
Methods DIP and PIP joints were obtained from corpses (n=40). Plain radiographies of these joints were taken. Joint samples were prepared for histological analysis; cartilage damage was graded according to the Mankin scoring system. A 2×2 Fisher's exact test was applied to define those radiographic features most likely to be associated with histological alterations. Receiver operating characteristic curves were analysed to determine radiographic thresholds. Intraclass correlation coefficients (ICC) estimated intra- and inter-reader variability. Spearman's correlation was applied to examine the relationship between our score and histopathological changes. Differences between groups were determined by a Student's t test.
Results The Interphalangeal Osteoarthritis Radiographic Simplified (iOARS) score is presented. The score is based on histopathological changes of DIP and PIP joints and follows a simple dichotomy whether OA is present or not. The iOARS score relies on three equally ranked radiographic features (osteophytes, joint space narrowing and subchondral sclerosis). For both DIP and PIP joints, the presence of one x-ray features reflects interphalangeal OA. Sensitivity and specificity for DIP joints were 92.3% and 90.9%, respectively, and 75% and 100% for PIP joints. All readers were able to reproduce their own readings in DIP and PIP joints after 4 weeks. The overall agreement between the three readers was good; ICCs ranged from 0.945 to 0.586. Additionally, outcomes of the iOARS score in a hand OA cohort revealed a higher prevalence of interphalangeal joint OA compared with the Kellgren and Lawrence score.
Conclusions The iOARS score is uniquely based on histopathological alterations of the interphalangeal joints in order to reliably determine OA of the DIP and PIP joints radiographically. Its high specificity and sensitivity together with the dichotomous approach renders the iOARS score reliable, fast to perform and easy to apply. This tool may not only be valuable in daily clinical practice but also in clinical and epidemiological trials.
- Hand Osteoarthritis
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Osteoarthritis (OA) of the hand is a very common disease. Radiographic prevalence lies between 29% and 76% of the population.1 ,2 As hand OA will be symptomatic in up to 16%,1 ,3 it is an important differential diagnosis if joint pain, aches, joint tenderness or impairment of joint function are presenting symptoms. The diagnosis of hand OA can be made clinically4; however, plain radiographs are recommended measures for morphological assessment.5 ,6 Additionally, as no structure modifying treatments exist for OA,2 ,7 the development of such therapies implies the need all the more for reliable and sensitive measurement tools for structural assessment.8
Yet there is no consensus on which radiographic score to prefer.6 ,9 ,10 Various radiographic scores have been developed to determine the presence of OA and to assess the severity and progression of structural joint damage of the hands.11–17 They differ with regard to number and region of assessed hand joints, radiographic features scored, grading of abnormalities or application of a global score. While all of these scores have been successfully applied in clinical studies, none was developed or confirmed by morphological findings as obtained by histopathological assessment of joints.
In the present study, we developed and validated a radiographic score that is based on histopathological changes in hand OA. We wished to establish a score that could be rapidly performed, was reliable and easy to apply, with only a limited number of radiographic features to score. X-rays and histological samples of the same distal (DIP) and proximal (PIP) interphalangeal joints were used to develop the Interphalangeal Osteoarthritis Radiographic Simplified (iOARS) score that provides a simple dichotomy of whether or not OA is present in a distinct interphalangeal joint, depending on the presence/absence of distinct radiographic features.
Patients and methods
Post mortem specimens
Joint specimens were obtained post mortem at the Department of Pathology, Medical University of Vienna, and comprised 80 interphalangeal (40 DIP and 40 PIP) joints from 40 consecutive individuals (18 women and 22 men). Patients with a documented history of inflammatory joint disease, such as rheumatoid arthritis or psoriatic arthritis, were excluded. This cohort comprises the same specimens that were used to determine the role of plain radiography in hand OA.5
Joints were obtained within 24 h of the subject’s demise and were derived from the second or third finger, according to the high prevalence of OA at these locations.7 ,18 ,19 To obtain these specimens, the skin and subcutaneous tissues were carefully dissected until the DIP and PIP joints, as well as the phalangeal bones, became visible. Then, the first and third phalangeal bones were cut above the DIP joint and below the PIP, respectively, so that the PIP and DIP joints could be obtained in toto. Of these 80 joints, 76 (37 DIP and 39 PIP joints) could be processed for histological and radiographic investigations.
Prior to dissection, both hands were clinically examined for Heberden and Bouchard nodes (palpation for firm/hard posterolateral rounded swelling and/or joined dorsal bars). Heberden and Bouchard nodes were classified as present or absent. If bony swelling was present, we selected the finger that clinically displayed the worst changes either on the left or on the right hand. If no nodes were present, the right hand was always used in accordance with the higher prevalence of OA on this side,20 and a computer program was employed to randomise which finger to dissect. The joint assessment was performed by a physician with longstanding rheumatological experience (KB). This study was approved by the ethics committee of the Medical University of Vienna (No. 409/2005).
Plain radiography of the interphalangeal joints (posterior-anterior views) was performed using a Philips Optimus 80 x-ray generator. A blinded assessment was carried out by a musculoskeletal radiologist (FK), who defined the presence of distinct radiographic features of hand OA (joint space narrowing (JSN), osteophytes, subchondral sclerosis, subchondral cysts, erosions, malalignment), as described elsewhere.11 ,21 Severity of the radiographic changes was not graded.
A new simple radiographic score, which was related to morphological abnormalities, was developed in this study, as will be described in detail in the results section below. Each joint was graded according to this new iOARS score. Scoring was performed by displaying the images in DICOM format on a workstation equipped with Agfa Impax software suitable for diagnostic use.
Entire finger joints were prepared for histological analysis and stained with Safranin-O/Fast green and toluidin blue, as previously described.5
The grade of structural damage of each sample was histomorphologically evaluated (I-GS), according to the modified Mankin score.22 This scoring system is composed of four categories: cartilage structure (0–6 points), cartilage cells (0–3 points), staining (0–4 points) and tidemark integrity (0–1 point). Scores of each category are summed up to a total score, with a possible maximum of 14 points. The samples displayed total Mankin scores ranging from 2 to 14 for DIP joints and from 0 to 14 for PIP joints. The Mankin score mainly focuses on the integrity of articular cartilage and does not comprise bony alterations such as osteophytes. However, it is significantly correlated with the presence and extent of radiographic osteophytes in interphalangeal joints.5 Moreover, the Mankin score correlated very well with radiographic subchondral sclerosis in DIP (rs=0.639, p<0.0001) and in PIP joints (rs=0.604, p<0.0001); thus we regarded an additional histological evaluation of bony changes not necessary.
Clinical hand OA patient cohort
Over a 2 year period (May 2006–March 2008), unselected patients from our outpatients clinic, who presented themselves with persistent or transient hand pain, aching or stiffness, were prescreened. Patients who either fulfilled the American College of Rheumatology classification criteria for hand OA4 or displayed a Kellgren–Lawrence score of ≥2 in at least one DIP or PIP joint on plain radiographs of the hands were included. Individuals with a documented history of inflammatory joint disease, such as rheumatoid arthritis or psoriatic arthritis, were excluded. Patients with a positive rheumatoid factor and/or positive anticitrullinated protein antibody test were also excluded. A total of 206 patients were enrolled in the study. Mean patient age was 61.8 years (range 36–85); 87.9% of the cohort were women.
Participating patients gave their written informed consent according to the Declaration of Helsinki 1996.
Plain radiographs of both hands (posterior-anterior views) of all 206 patients (1648 DIP and PIP joints) were scored according to the Kellgren and Lawrence system, as well as the iOARS score, by I-GS. In order to estimate inter- and intra-reader reliability, hand x-rays of 41 randomly selected patients were independently graded by three readers (GS, I-GS and KB) according to the iOARS score.
A 2×2 Fisher's exact test was applied to examine the significance of the association between individual radiographic features and histopathological alterations. Effect sizes are provided by Pearson r. Receiver operating characteristic (ROC) curves were analysed in order to determine cut-off points for radiographic examinations. To examine the relationship between histological (Mankin score) and radiographic changes (iOARS score), as well as the correlation between the iOARS score and the Kellgren–Lawrence scale, Spearman's rank order correlation was used and expressed as r values (rs). A Student's t test was used to determine the significance of differences between groups (non-OA vs OA joints). To evaluate inter- and intra-reader reliability for the iOARS score, intraclass correlation coefficients (ICC) were estimated. A p value <0.05 was considered significant.
Development of the iOARS score
A histopathological cut-off, beyond which a sample can be seen as definitive osteoarthritic, was determined at a Mankin score of 5, as reported previously.5 A 2×2 Fisher's exact test was then applied to elucidate the association between the presence of individual radiographic features (JSN, osteophytes, subchondral sclerosis, subchondral cysts, erosions, malalignment) and the Mankin score. For both DIP and PIP joints, JSN, osteophytes and subchondral sclerosis were the features most strongly associated with histopathological changes (table 1).
In order to keep our score as simple and practicable as possible, and avoid any bias towards a single feature, we assigned 1 point for each feature present, regardless of severity, allowing for a minimum score of 0 and a maximum score of 3 points. Then we analysed ROC curves (figure 1) to determine the radiographic cut-off that would indicate the presence of OA based on the histopathological scores (threshold 5 for DIP and PIP joints). The analysis elucidated a radiographic cut-off of ≥1 for both DIP (sensitivity 92.31% (95% CI 74.9 to 99.1), specificity 90.91% (95% CI 58.7 to 99.8)) and PIP (sensitivity 75% (95% CI 53.3 to 90.2), specificity 100% (95% CI 78.2 to 100)) joints. Thus our score provides a simple dichotomy whether or not OA is present, depending on the presence of at least one radiographic feature (table 2).
Construct and criterion validity
To assess construct validity, we correlated the iOARS score with patient age, hypothesising that with increasing age, an increasing number of individuals will display an iOARS score ≥1. Figure 2 demonstrates a significant correlation between increasing age and increasing number of radiographic features. This correlation was, however, more pronounced for DIP (rs=0.727, p<0.0001) than for PIP (rs=0.377, p<0.02) joints.
Criterion validity was assessed in the clinical hand OA patient cohort by correlating the number of OA joints per patient, determined by both the iOARS and the most frequently applied radiographic hand scale, the Kellgren–Lawrence score. A moderate but significant correlation was found for both DIP and PIP joints (figure 3). To further elucidate these findings, we compared the number of non-OA and OA joints determined separately by both scoring systems. We found a considerable difference between both scoring systems with regard to non-OA joints (DIP: Kellgren–Lawrence, n=682 and iOARS, n=167; PIP: Kellgren–Lawrence, n=1065 and iOARS, n=231) and OA joints (DIP: Kellgren–Lawrence, n=966 and iOARS, n=1481; PIP: Kellgren–Lawrence: n=583 and iOARS: n=1417) (figure 4A, B). Interestingly, out of the Kellgren–Lawrence non-OA group, 515 (76%) DIP joints and 794 (75%) PIP joints were classified as ‘doubtful OA’. The prevalence of OA changes in the DIP and PIP joints in accordance with the iOARS and Kellgren–Lawrence scores is provided in figure 5.
Intra- and inter-reader variability was assessed by comparing the independent results of three readers (I-GS, GS, KB). DIP joints II–V as well as PIP joints II–V from 41 randomly selected radiographs from the hand OA cohort were scored according to the iOARS score. All readers were able to reproduce their own readings in DIP and PIP joints after 4 weeks. Overall, there was good to moderate agreement between the three readers, depending on the individual joints, ranging from 0.945 (PIP VI, right hand) to 0.586 (PIP III, left hand). The inter-reader reliability was generally similar for DIP (ICC mean±SD 0.794±0.08) and PIP joints (ICC 0.783±0.08).
The average (±SD) time for a trained reader required to score one anteroposterior hand radiograph displaying one hand was 24.09±7.7 s and was mainly dependent on the severity of the radiographic changes (the heavier the damage, the faster the scoring procedure). The average time to score the same images according to the Kellgren–Lawrence scale was 37.44±6.7 s.
In daily clinical practice as well as for clinical trials, radiographic definition of hand OA is important to confirm or reduce the likelihood of this diagnosis. A number of x-ray scoring systems exist to evaluate joint damage; however, none was developed on the basis of histopathological changes. We aimed to define a simple, yet exact radiographic scoring system for PIP and DIP joints that is based on histopathological alterations and relies on a minimum of distinct x-ray features.
This simplified x-ray scoring system, iOARS, for DIP and PIP joints, follows a simple dichotomy (presence or absence of OA), depending on the absence/presence of distinct radiographic changes. This radiographic scoring system was developed on the basis of histopathological changes of individual joints.
Our score was created pursuing an iterative approach starting with (1) the definition of a histopathological cut-off, continuing with (2) the identification of radiographic markers most likely to be associated with histological alterations and finally (3) determining the critical number of x-ray changes that would imply the presence of OA.
There is no general agreement on a histopathological cut-off that distinguishes normal from OA cartilage. In general, thresholds were arbitrarily defined within the lower range of the Mankin scale.23–26 One could argue that every deviation from a Mankin grade of 0 would represent OA cartilage changes in a more or less pronounced manner,27 however, correlations of arthroscopy and MRI with histopathoplogy showed that Mankin scores of normal and mild OA cartilage overlapped within a range of 0–5.28 ,29 It seems that the lower third of the Mankin scale represents some form of grey zone where boundaries between normal and OA cartilage are blurred and not every minor change in cartilage structure has to be considered as OA. To take this issue into account, we defined a threshold of Mankin grade 5, beyond which DIP or PIP cartilage can be regarded as definitively osteoarthritic.
Once the histopathological threshold was defined, we determined radiographic features to be incorporated into our score. We aimed to identify those features that would most likely be associated with histological changes. Classic radiographic changes include osteophytes, JSN, subchondral sclerosis, subchondral cysts, erosions and malalignment. Kellgren and Lawrence, for instance, rely on osteophytes as a hallmark of radiographic changes in hand OA14 and in fact osteophytes were reported to be the earliest radiographic alteration.30 Compared with JSN, osteophytes also had a slightly higher incidence.31 ,32 In contrast, radiographic severity of JSN was more pronounced,33 ,34 thereby emphasising a similar or even greater importance of this feature. We found that both JSN and osteophytes were strongly associated with histopathological alterations. A highly significant association was also found for subchondral sclerosis, which can be seen as a hallmark and indisputable sign of progressing OA.35 The three radiographic features were not weighted against each other to avoid any bias towards one single feature, as seen in the Kellgren and Lawrence score14 or in the scale presented by Kessler et al15; the more so as no evidence exists on whether one feature (osteophyte, JSN or subchondral sclerosis) is more important than the other for defining radiographic hand OA. However, all three features correlated significantly with the Mankin score.
Finally, we determined the critical number of radiographic features (osteophytes and/or JSN and/or subchondral sclerosis) whose presence would reflect significant histological cartilage damage and thus definite OA. X-ray feature sum scores were related to the extent of histopathological damage, and ROC analyses revealed a cut-off of ≥1 for DIP and for PIP joints. The sensitivity of the iOARS score for the PIP joints is lower than that for DIP joints, which may be dependent on distinct anatomical and mechanical properties of the PIP joints that might influence tissue remodelling and mask structural damage. However, the cut-off of ≥1 is somewhat unexpected, in particular since a single diagnostic feature is generally believed to be less valuable for the diagnosis of hand OA than a composite of two or more features.6 Nevertheless, according to our findings, a single radiographic feature would already reflect substantial histological damage in interphalangeal joints. This was particularly obvious when we compared the outcomes of the Kellgren–Lawrence scoring method with the findings of the iOARS score in our hand OA patient cohort. In contrast with the iOARS score, the Kellgren–Lawrence scale identified 41% of the DIP and 65% of the PIP joints as non-OA; however, 76% of the non-OA DIP and 75% of the non-OA PIP joints were classified as doubtfully OA, thereby potentially underestimating the morphological changes in the interphalangeal joints. Moreover, and possibly more important, our data revealed a higher prevalence of radiographic OA of DIP and PIP joints than generally believed.3 ,20 ,36–39 With regard to x-ray changes, it is known that a considerable amount of damage has to have occurred in order to detect OA radiographically,40 suggesting that joint x-ray would not reliably reflect minor to moderate histopathological alterations.
Our findings need to be tested in a larger patient cohorts, and cohorts reflecting the general population, which was not the case for our hand OA cohort. Nevertheless, as the majority of trials used the Kellgren–Lawrence score as a radiographic readout,41 our data suggest that the prevalence rate of radiographic OA in DIP and PIP joints is up to two times higher than implied when using the Kellgren–Lawrence score.
In conclusion, we established a radiographic score for DIP and PIP joints that is uniquely based on histopathological changes and comprises only three distinct x-ray features (osteophytes, JSN, subchondral sclerosis). The score, however, provides a simple dichotomy (whether or not OA is present) and hence was not developed to determine the severity or progression of OA changes over the course of time, as might be necessary for some interventional clinical trials. Eventually, as the iOARS score provides good sensitivity and specificity and can easily be applied without time consuming grading of different x-ray features, it might be a useful tool for rapidly and reliably determining the presence of hand OA in daily clinical practice and may be used as an inclusion criterion in trials, cohorts or surveys where a radiographic method is mandatory for disease classification.
The opportunity to carry out this investigation was provided under the auspices of the Joint and Bone Centre (Schwerpunkt muskuloskelettale Erkrankungen) of the Medical University of Vienna.
Handling editor Tore K Kvien
Contributors I-GS: study design, data collection, analysis and interpretation, and writing. LA-M, BN and AS: data collection. TS and SH: data collection, analysis and interpretation. GS: data collection and analysis. FK: data collection and interpretation. JSS: study design, data analysis and interpretation, and writing. KB: study design, data analysis and interpretation, figures and writing.
Funding Our trial was an independent, academic study funded by and performed at the Medical University of Vienna.
Competing interests None.
Ethics approval This study was approved by the ethics committee of the Medical University of Vienna (No. 409/2005).
Provenance and peer review Not commissioned; externally peer reviewed.