Objective The authors have previously reported on the relationship between activity and subsequent damage at the patient level for patients with psoriatic arthritis (PsA). The aim of this study was to identify key predictors of damage to individual joints in the hands and feet of patients with PsA, in particular those that capture previous activity.
Methods Data from patients followed prospectively at the University of Toronto PsA clinic between 1978 and 2006 were available for analysis. Logistic regression was used to relate the probability of a joint developing damage, within a specified time interval after the most recent clinic visit, to potential predictor variables. The predictor variables considered encompassed the history of disease activity of the joint and elsewhere, previous damage and the timing of clinical assessments.
Results 511 patients with no hand damage at clinic entry and 552 patients with no foot damage at clinic entry were included in the analysis of the hand and foot joints, respectively. The analysis of the hand and foot joints demonstrated that the activity (tenderness and/or swelling) history of the specific joint is associated with subsequent damage. For the joints of the feet, activity observations elsewhere in the same foot, and in particular the same toe, were also shown to be associated with subsequent damage.
Conclusions Both joint tenderness and swelling are important predictors of joint damage in PsA.
Statistics from Altmetric.com
Psoriatic arthritis (PsA) is an inflammatory arthritis associated with psoriasis, usually seronegative for rheumatoid factor.1 PsA leads to progressive joint damage, disability and increased mortality risk.2,–,8 Joint damage in PsA is a marker of severity because it is associated with and predicts loss of function and mortality.9,–,12 Therefore, determining predictors of joint damage is important.
We have previously demonstrated that age, time in clinic, initial erythrocyte sedimentation rate, number of tender and swollen joints at previous visit and number of deformed joints at previous visit are risk factors for progression of joint damage.13 These analyses were based on total joint counts but not on the presence of activity or damage at the individual joint level. We now aimed to investigate the relationship between inflammatory activity and clinical damage at the individual joint level. We chose clinical rather than radiographic damage as the outcome measure because clinical damage is measured at each clinic visit, whereas radiographic damage is assessed only every 2 years. We have previously shown that clinical damage is related to radiological damage.14
Our PsA cohort data are longitudinal with correlated observations on the same joints over several clinic visits.15 A suitable statistical analysis needs to reflect this and the irregularity in observation times. Nevertheless, this longitudinal dataset allows for a comprehensive investigation into disease progression as observed in routine clinical practice, in contrast to that in clinical trials. Therefore, our aim was to investigate the relationship between disease activity and clinical damage at the individual joint level in PsA using data from the University of Toronto PsA clinic.
The University of Toronto PsA clinic is an observational cohort of patients with PsA followed prospectively since 1978.4 Patients are followed at approximately 6–12-month intervals according to a standard protocol. The clinical evaluation includes assessment of disease activity and clinical damage of 64 individual joints. The study was approved by the University Health Network Research Ethics Board. A total of 705 patients registered in the clinic between 1978 and 2006 and observed for at least two clinic visits was included in the study. Data from 511 patients without clinical damage in their hand joints (wrists, metacarpophalangeal, proximal interphalangeal (PIP) and distal interphalangeal joints (DIP) (total 30)) at clinic entry were included in the analysis of the hand joints; similarly, data from 552 patients without clinical damage observed in the joints of the feet (ankles, metatarsophalangeal, interphalangeal joints (total 22; PIP and DIP joints of each of the second to fifth toes were considered to represent only one joint because it is difficult to determine reliably tenderness and swelling of the PIP and DIP separately in the toes)) at clinic entry contributed data to the analysis of the foot joints.
Assessment of activity and clinical damage
An active joint is described as either tender (presence of stress pain and/or joint line tenderness) or swollen (joint swelling with or without tenderness). Clinical damage is determined by the presence of a limitation of range of movement of more than 20% of the range not related to the presence of joint effusion, the presence of joint deformities, subluxation, flail joints or ankylosis. The reliability of these measures has been demonstrated.16,–,19
A pragmatic modelling approach was used.20 Logistic regression was used to model the probability of a joint developing damage within 1, 2, 3 and over 3 years after a clinic visit, with times between visits coded into these four categories. Previous clinical information, including information on activity, was incorporated by means of explanatory variables and patient-specific random effects modelled correlation between multiple observations from the same patient.
Predictor variables initially considered included indicators for activity in the individual joint, both recent and over the course of observation, together with an indicator for previous joint activity elsewhere in the hands/feet, and an interval variable for the total time in clinic. Results are reported in terms of OR.
We also developed a range of concise summary measures of previous activity in the joint. The first was a simple measure, termed proportion (PROP), corresponding to the proportion of visits with activity observed in the joint. A second measure, termed the adjusted mean (ADJ) that takes into account the duration between clinic visits, provides an estimate of the proportion of total time, since the first clinic visit, with activity present in the joint. ADJ was defined as in Boers et al21 and Ibañez et al.22 The value for ADJ is calculated as:
where ai is an indicator for activity in the joint at the ith visit and ti is the time between the ith and (i−1)st clinic visits.
Since recent activity is likely to be more important in causing joint damage than remote activity in that joint, an estimate of the probability of activity in a joint was based on an exponentially weighted moving average (EWMA) that gives more weight to recent observations without discarding older observations entirely.23 For a joint that has been observed at v visits the EWMA is:
where μ̂v is the current probability estimate, μ̂v-1 is the probability estimate from the first v−1 visits, av is an indicator for activity in the joint at the vth visit and κ is the weighting parameter. A range of values for κ were investigated, and 0.4 emerged as most appropriate. The initial estimate, μ̂0, was taken as the proportion of active hand or foot joints at the first visit for the patient.
The demographic and disease characteristics of study patients are given in table 1. The subjects had a median of eight (range two to 48) clinic visits over a median of 8 years (range 0.1–33). At the most recent clinic visit, 170 (33%) of the patients in the hand study had developed hand damage with a median number of three (range one to 30) damaged hand joints. Likewise, 225 (41%) of the patients in the foot study developed foot damage with a median number of five (range one to 22) damaged foot joints.
The key finding from initial analyses of onset of damage in the hands is that the prime driver for progression to damage is recent activity in that specific joint. Recent activity was defined as the last three clinic visits, although the length of time that these visits span differs between patients. The estimated OR (2.29, 95% CI 1.90 to 2.77) for the indicator of activity at the most recent clinic visit was much greater than the estimated OR for activity indicators for two or three visits ago (1.33, 95% CI 1.09 to 1.61 and 1.66, 95% CI 1.36 to 2.03).
To summarise concisely the effect of joint-specific activity history, summary measures for individual joint activity, defined previously in the Methods section, were explored. PROP and EWMA emerged as important predictors of subsequent damage, but not ADJ. The results for a final multivariate model are given in table 2. Both PROP and EWMA achieve high significance. The addition of binary indicators for recent activity in the joint does not lead to a significant improvement in fit (likelihood ratio test p value 0.061) indicating that EWMA provides a better representation of recent activity. Past joint swelling is highly significant but past joint tenderness, not included in the model for parsimony, adds less predictive information (p=0.035) as it is also reflected in the PROP and EWMA variables. In addition, although global activity is less important than joint-specific activity, recent swelling elsewhere in the same hand also emerged as an important predictor of progression to joint damage.
An investigation into the effect of previous damage revealed that the probability of a joint progressing to damage is increased if the equivalent joint on the opposite hand, termed the ‘symmetric joint’, is damaged, and increased further if a joint of the same type (metacarpophalangeal, PIP or DIP) of an adjacent finger, termed an ‘adjacent joint’, is damaged. Interactions of activity variables with damage were also examined revealing that the effect of ‘PROP’ and ‘Swelling at any recent visit’ is reduced if an adjacent joint is damaged.
Further details are available in a supplementary document available online only.
In all models fitted, the total time in clinic and time between pairs of clinic visits were included. Total time in clinic was incorporated by means of an interval variable with five intervals: 0–5, 5–10, 10–15, 15–20 and greatet than 20 years. The results suggest that the longer the patient has been in clinic the more likely it is that damage will be observed. Results based on an interval variable for time between visits indicate that the longer the time between visits, the more likely is the observation of damage. Intervals of less than 3 months were seen in only 0.35% of observations. Exclusion of these pairs of visits did not alter the results. Similarly, the qualitative results were unchanged if intervisit times of more than 2 years, 6.8% of observations, were dropped.
A Hosmer–Lemeshow goodness-of-fit test based on deciles of risk generated a χ2 test statistic of 25.6 (9 df, p=0.001). A significant result is expected given the number of observations. Qualitatively, lack of fit arose due to some underestimation of risk in the highest categories of predicted risk.24
Initial analyses identify a key predictor of subsequent damage in a foot joint as previous activity in that joint, in particular previous swelling. However, in striking contrast to the results for hand joints, a large effect of activity at the most recent clinic visit is not demonstrated. However, if ankles are included in the analysis this variable achieves high significance. This differing effect led us to exclude ankle joints from the analysis. The results for a final multivariate model are given in table 3. As compelling evidence for the importance of recent joint-specific activity observations was lacking, the average measures for disease severity, PROP and ADJ, were considered but not EWMA. The estimated effect for ADJ was significant in addition to other variables, whereas PROP was not.
In contrast with the results for hands, the global activity variable defined by previous activity elsewhere in the feet achieves high significance in initial analyses. The results identify recent activity in the feet and previous activity in the other joint of the same toe as important predictors for damage. This has been captured by two binary indicators, ‘Activity in the feet at any of the three most recent visits’ and ‘Activity in the other toe joint at any visit’. Strikingly, the odds for damage progression in a foot joint are more than doubled if the other joint in the same toe is observed as active (assuming that the values of the other variables are fixed). On the contrary, for hand joints, previous activity in a joint of the same finger was not an important predictor.
The model in table 3 contains an indicator for previous damage in the symmetric joint. As for hands, the results indicate that the odds of a joint progressing to damage are increased if the equivalent joint on the opposite foot is damaged. The final model also contains an indicator variable for damage in an adjacent joint (a joint of the same type (metatarsophalangeal or interphalangeal) of an adjacent toe), but previous damage in an adjacent joint was not found to be important. Significant interactions between the symmetric damage variable and activity variables were not found.
The inclusion of the time variables suggests that, as for the hands, the longer the time in clinic, the more likely it is that the foot joint becomes damaged and that the longer the time between visits the more likely it is that damage will be observed at the next visit. However, the trend lessens with increasing time with interval estimates for parameters associated with intervisit times of 2–3 and more than 3 years overlapping considerably. The exclusion of short (0.40%) and long (6.9%) intervisit times did not alter the results.
A Hosmer–Lemeshow goodness-of-fit test based on deciles of risk generated a χ2 test statistic of 31.0 (9 df, p<0.001).24 Qualitatively, lack of fit arose due to some underestimation of risk in the third and fourth quantiles.
The results of this study using data collected over almost 30 years in a longitudinal cohort of PsA patients show that after a median follow-up period of 8 years, 33% of patients who did not have clinical damage in the hand joints at study entry developed damage to one or more hand joints, and 41% of patients without damage in the foot joints developed damage to one or more foot joints. Predictors of joint damage to a particular joint in the hands or feet are previous activity in that joint, with additional risk associated with activity elsewhere in the hand or foot. The presence of joint damage in a symmetric joint and longer duration in the clinic also increases risk.
The study confirms the progressive nature of joint damage in PsA and the clinical impression of a strong link between activity in a particular joint and subsequent damage to that joint. This has not previously been demonstrated in PsA, although a smaller study from the Combinatietherapie Bij Reumatoide Artritis trial showed similar results in rheumatoid arthritis (RA).21 Importantly, our study demonstrates that the presence of joint tenderness at the time of the clinic visit even in the absence of joint swelling significantly increases the risk of future damage. As patients with PsA have less tenderness than patients with RA, when tenderness is found, it probably represents inflammation.25 The presence of joint swelling further increases the risk, suggesting that swelling reflects more significant inflammation. Our analyses are focused on data collected at clinic visits. Information is not available on the activity status of a joint between clinic visits. It is possible that joint tenderness indicates the presence of active inflammation at the clinic visit and that joint swelling was missed on clinical examination, or that swelling occurred in that joint between clinic visits. Assessment of joint swelling is known to be less reliable than the assessment of joint tenderness.18 19 As one of the major goals of treatment of PsA is to prevent damage, our study suggests that clinicians should aim to achieve complete ‘remission’ in the joints (joints neither tender nor swollen). The risk of damage to a joint of the hands or feet increased significantly if the duration between clinic visits was longer. Patients with longer duration between visits may have activity in the joints between visits. Therefore, patients need to be evaluated at shorter regular intervals so that joint damage is prevented.
Although in general terms the results from the hand joints and foot joints were similar, important differences were noted. Compared with hand joints, recent activity in the foot joints was not found to be as important; activity in other joints of the feet was more important. The ankle joint is more affected by recent joint-specific activity than the other foot joints. Activity in the other joint in a toe was a predictor of damage, whereas activity in other joints in a finger was not. Although joint damage in an adjacent joint at the same ‘level’ (metacarpophalangeal, PIP or DIP) was an important predictor in the hand joints, this was not the case in the joints of the feet. We speculate that one major reason for these differences is the anatomical differences between hands and feet. The toe joints are relatively close to each other compared with the finger joints. It is conceivable that inflammation in a toe joint could lead to ‘collateral damage’ to a neighbouring joint and this is more likely to occur in toes compared with fingers. Clinicians recognise that assessment of the foot joints, especially that of the toes, is more difficult than assessment of the hand joints. This is why we club each of the PIP and DIP joints of the second to the fifth toes into one interphalangeal joint. It is also possible that the clinical assessment of joint tenderness and swelling is less reliable in foot than in hand joints. In addition, combining the PIP and DIP toe joints may have contributed to differences observed between the results for feet and hands.
In this study, in contrast to most studies on joint damage, the outcome chosen was clinical and not radiographic damage. Assessment of clinical damage is reliable.16 17 There is a strong correlation between the onset of radiographic and clinical damage, with radiographic damage usually preceding clinical damage.14 The consistency of results relating active joint counts to damaged joint counts whether damage is measured clinically or radiographically has been shown based on detailed investigations in the clinical and statistical literature.13 14 26 The advantage, particularly for the study of individual joints, of clinical over radiographic damage, is that it is measured at each clinic visit, whereas radiographs are undertaken, usually, once every 2 years and provide relatively sparse data. In addition, in contrast to only 42 joints being assessed using radiographs, clinical joint damage can be assessed in all 64 joints.4 27 Radiographic scoring of the foot joints does not include interphalangeal joints of the toes and the ankles.27 Therefore, comparison between the models developed for hand joints with those of the foot joints would not be possible. The main disadvantages of using clinical damage over radiographic damage is that comparison of results across studies is difficult because most centres do not collect information on clinical damage and rely only on radiographs for damage assessment. Moreover, damage progression in a particular joint after the onset of damage is better assessed using radiographs; clinical damage progression (from deformity to flail/ankylosed joint) is slower. However, our aim was to determine predictors of the ‘onset’ of damage and not of ‘worsening’ damage. Therefore, given the advantages of using data on clinical damage, we chose clinical rather than radiographic damage as the outcome measure.
In contrast to RA, we do not have strong predictors of joint damage at the patient level. Acute phase reactants predict overall joint damage progression.13 The relationship between activity in individual joints and risk factors at the ‘patient level’, such as erythrocyte sedimentation rate, were not investigated. These studies are planned for the near future.
To conclude, the practical clinical inferences from these results are that clinicians must strive to achieve the absence of both joint tenderness and swelling so that joint damage is prevented. Frequent and regular follow-up assessments may contribute to the prevention of joint damage.
Funding VC is supported by a Canadian Institutes of Health Research Clinical Research Initiative Fellowship, the Henry A Beatty Scholarship from the University Health Network and the Krembil Foundation. VTF and LC are supported by MRC funding U.1052.00.009. The University of Toronto Psoriatic Arthritis clinic is supported by the Krembil Foundation and by the Arthritis Society, Canada, through a national research initiative grant.
Competing interests None.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the University Health Network Research Ethics Board.
Provenance and peer review Not commissioned; externally peer reviewed.
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.