Description and prediction of physical functional disability in psoriatic arthritis: A longitudinal analysis using a Markov model approach

Abstract

Objective

To describe the longitudinal course of physical functioning in patients with psoriatic arthritis.

Methods

Between June 1993 and June 2003, 341 patients attending the University of Toronto Psoriatic Arthritis Clinic completed 2 or more Health Assessment Questionnaires (HAQs). At the time of administration of each HAQ, patients were assigned to 1 of 3 physical functional disability states, based on their HAQ score. A Markov model that allowed for transitions to and from these 3 disability states was used to characterize the longitudinal course of physical functioning, as well as to identify factors for both progression and regression of disability.

Results

Despite patient variability in the course of physical functioning, the following 3 longitudinal patterns were observed: 1) a stable state of disability throughout the entire study period, with 28%, 12%, and 6% of patients experiencing no, moderate, or severe disability, respectively; 2) a steady improvement or deterioration in disability over time (this pattern was observed in 27% of patients); and 3) a fluctuating state of disability, occurring in 27% of the patients. Sex, age, disease duration, number of actively inflamed joints, and number of deformed joints predicted transitions between disability states.

Conclusion

Although 28% of patients appeared resistant to becoming disabled over the duration of this study, the remaining patients were observed either to experience enduring disability or to move between disability states.

INTRODUCTION

Psoriatic arthritis (PsA) is an inflammatory arthritis that is associated with psoriasis; patients are usually seronegative for rheumatoid factor. Prior to the mid 1980s, PsA was considered a benign disease, with short-lived synovitis that did not lead to residual damage in most patients. Since 1987, however, we and other investigators have shown that in many patients with PsA, deformities, damage, and disease progression develop over time (1–8). Furthermore, cross-sectional studies have shown that physical functioning among patients with PsA is significantly lower than that among healthy controls and comparable with that of patients with rheumatoid arthritis (RA) (9–11). However, it is unknown whether the physical limitations observed cross-sectionally reflect short-lived, episodic, or chronic disability. From a clinical perspective, it would be useful to understand the pattern of physical disability over time, with particular attention to the factors associated with persistent or chronic disability.

The aim of this study was to describe the course of physical functional disability in PsA. Although it is expected that some patients will experience a steady decline in function over time, it is also expected that others will experience either steady improvement or a highly variable course. To capture all of these possible changes over time, we adopted a reversible multistate Markov model that allows for transitions to and from physical functional disability states (12). Although Markovian models have been used to study transitions between functional states among elderly residing in the community (13), thus far they have not been used to study similar changes in arthritis populations.

PATIENTS AND METHODS

Patient population.

Established in 1978, the University of Toronto PsA clinic at the Toronto Western Hospital represents the largest population of patients with PsA ever collected and followed prospectively. It is both a primary, secondary, and tertiary referral center, including patients with mild to severe disease as well as patients with newly diagnosed disease and those with existing disease.

Patient assessment at first clinic visit and followup.

At each visit, a complete history was obtained and an examination was performed according to a standard protocol, and data were recorded according to data retrieval protocol. Briefly, inquiry was made into demographic features (e.g., age, marital status, and education), the presence of joint inflammation, medications, the American College of Rheumatology (formerly the American Rheumatism Association) functional level (14), and general medical history. The examination consisted of a general medical examination with particular attention to the skin, nails, eye, and heart, as well as both peripheral and axial joints. Psoriasis severity was rated according to the Psoriasis Area and Severity Index (PASI) (15). The number of actively inflamed joints (stress pain, joint line tenderness, and effusion) and deformed joints (ankylosis, subluxation, or decreased range of motion >20%, attributable to joint damage rather than inflammation), were recorded (16). Patients were followed at regular intervals (6–12 months).

Since June 1993, the Health Assessment Questionnaire (HAQ) (17) has been administered annually to clinic patients. The HAQ assesses physical functional status over the past week and includes questions related to fine movements of the upper extremity, locomotor activities of the lower extremity, and activities that involve both upper and lower extremities. The HAQ consists of 20 questions that cover 8 categories of daily living (i.e., dressing and grooming, arising, eating, walking, hygiene, reach, grip, and activities, including errands and chores). Patients rate their ability to perform a particular task within a category on a scale from 0 (no difficulty) to 3 (unable to do). The highest score for any task within a category determines the score for that category, with an adjustment for the use of devices or the need for assistance. The scores for all 8 categories are then averaged to obtain an overall score on a scale from 0 (no disability) to 3 (severe disability). The HAQ has been shown to be reliable in our clinic population (18–20).

Statistical methods.

Overall HAQ scores were divided into 3 categories: 0–0.49, 0.5–1.5, and 1.51–3, representing physical functional disability states 1, 2, and 3, respectively. State 1 represented the absence of disability, while states 2 and 3 represented “moderate” and “severe” disability, respectively. These cut-offs were consistent with past studies, where HAQ scores less than 0.50 represented no disability (i.e., few difficulties in performing daily activities), and scores above 1.50 reflected a higher or more severe level of disability (i.e., considerable difficulties or assistance required in performing daily activities) (21–24). At each HAQ administration, a patient was assigned to 1 of these 3 disability states. Patients could show deterioration or improvement in physical functional disability over their followup periods.

For modeling this type of staged data, multistate models based on Markov processes provided a natural framework, because interest lay in the course of physical functional disability over time and, in addition, patients were only under intermittent observations. This allowed estimation of rates of transitions between the 3 states of functional disability and easily incorporated the effects of covariates (both time-independent and time-dependent) on transition rates. Here, correlation among the states of a patient at the different HAQ assessment visits was directly modeled through the Markov assumption that the future evolution of the patient's disability process depended only on his/her current state and not on his/her previous history.

An alternative approach that does not require categorizing the HAQ scores is based on random-effects models for longitudinal data. Here correlation is induced through the random effects. We preferred the multistate approach, because it allowed us to more directly address our objectives. In addition, distributional assumptions concerning the HAQ score were avoided.

Initially, a multistate Markov model that allowed forward and backward transitions from each of the transient disability states was fit. However, the simplified model that eliminated the direct transitions from state 1 to state 3 and from state 3 to state 1 was sufficient to characterize our data (see Results). Therefore, results reported in this study are obtained using this “simplified” model (Figure 1).

Figure 1. Reversible multistate Markovian model for observed transitions between physical functional disability states for 341 patients with psoriatic arthritis.

We assumed that each baseline transition rate remained constant throughout the followup period. We further assumed that the times of the HAQ assessments were noninformative for the disability process. Note that sampling schemes in which patients are assessed at regular intervals or assessed at randomly sampled times, or in which the doctor monitoring the patient's progress chooses the next assessment time for the patient depending on the state the patient is in at the current HAQ assessment visit are all noninformative. A sampling scheme is informative when a patient who feels unwell and/or whose symptoms suggest that the disability process is advancing may “self-select” to have a HAQ assessment visit (25).

We examined the separate (univariate) and joint (multivariate) effects of select demographic and clinical variables on the transition rates. These covariates were incorporated into the models through the proportional hazards assumption. The variables included were sex, age, duration of PsA, psoriasis severity as measured by the PASI, the number of clinically deformed or damaged joints, and the number of actively inflamed joints updated at each HAQ visit. Preliminary model investigations showed that it was adequate to make the simplifying assumption that each variable had a common effect on all forward transitions and another common effect on all backward transitions. Descriptive information on the patients was also provided.

RESULTS

Between June 1993 and June 2003, 395 clinic patients had completed at least 1 HAQ. Of these, 341 patients (86.3%) had completed 2 or more HAQs and comprised the study group for our analysis. Demographic and clinical features of these 341 patients at their first HAQ administration are shown in Table 1. Briefly, the majority of the patients were male (59%). The mean age and mean disease duration at the first HAQ administration were 45.9 years and 10.6 years, respectively. PASI scores ranged from 0 to 57.4, with a mean of 7.1. The mean number of actively inflamed and clinically deformed joints (8.3 and 6, respectively) reflected moderate disease activity and disease severity. The majority of patients (62%) had polyarthritis, with or without a spondylarthritis, whereas 41% demonstrated evidence of spondylarthritis. With 1 exception, there were no significant differences in demographic and clinical features between the 341 patients with 2 or more HAQs and the 54 patients with 1 HAQ only. At their first HAQ administration, patients with 1 HAQ had, on average, longer disease durations (mean duration of PsA 13.9 years).

Table 1. Demographic and clinical features of 341 PsA patients at their first HAQ assessment*

Variable	Value	Range
* PsA = psoriatic arthritis; HAQ = Health Assessment Questionnaire; PASI = Psoriasis Area and Severity Index.
No. (%) men	201 (58.9)
No. (%) women	140 (41.1)
Mean ± SD age, years	45.9 ± 12.4	17.7–93.6
Mean ± SD duration of PsA, years	10.6 ± 8.4	0.2–60.6
Mean ± SD number of active joints	8.3 ± 9.3	0–55
Mean ± SD number of joint effusions	3.0 ± 4.5	0–30
Mean ± SD number of deformed joints	6.0 ± 11	0–59
Mean ± SD PASI score	7.1 ± 9.7	0–57.4
Arthritis pattern, no. (%)
Distal	9 (2.6)
Oligoarthritis	53 (15.5)
Polyarthritis	121 (35.5)
Back alone	14 (4.1)
Back + distal	7 (2.1)
Back + oligoarthritis	27 (7.9)
Back + polyarthritis	90 (26.4)
Remission	15 (4.4)
Missing	5 (1.5)

The mean number of HAQs administered per patient was 5 (range 2–11), and the mean ± SD length of followup with the HAQ was 5.2 ± 3.04 years. The mean ± SD duration between HAQ administrations was 1.29 ± 0.70 years. At the first HAQ assessment, the mean ± SD HAQ score was 0.69 ± 0.67, reflecting moderate disability. One hundred fifty-seven patients (46%) had a HAQ score of less than 0.5 and thus were assigned an initial disability state of 1, 134 patients (39%) had a score between 0.5 and 1.5 (inclusive) and were assigned to disability state 2, and the remaining 50 patients (15%) had a score greater than 1.5 and were assigned to disability state 3.

Table 2 summarizes the number and type of transitions in disability states that were observed over the followup period. Note that not all patients were observed to experience a transition. Of the 341 patients, 95 (28%) were in state 1 (no disability) throughout the entire followup period, while 42 (12%) and 20 (6%) remained only in state 2 (moderate disability) or state 3 (severe disability), respectively. One hundred eighty-four of the 341 patients were observed to experience a transition in disability state. Ninety-one patients (26.7%) encountered a single transition to either a lower or higher disability state. Ninety-three patients (27.3%) experienced 2 or more observed transitions, with 1 patient moving progressively to a lower state and 92 fluctuating between higher and lower states of disability. The mean ± SD changes in HAQ score for patients who moved from a higher to a lower disability state in consecutive visits (i.e., improved) and for those who moved from a lower state to a higher state (i.e., deteriorated) were −0.57 ± 0.36 and 0.55 ± 0.32, respectively. The mean ± SD change in HAQ score for patients who were not observed to change disability state in consecutive visits was −0.002 ± 0.215.

Table 2. Number and type of observed transitions between disability states for 341 PsA patients*

Number and type	Disability state at time of first HAQ assessment
	State 1 (n = 157)	State 2 (n = 134)	State 3 (n = 50)
* PsA = psoriatic arthritis; HAQ = Health Assessment Questionnaire.
No transitions (n = 157)	95	42	20
One transition (n = 91)
Deterioration
State 1 → 2	25	0	0
State 1 → 3	1	0	0
State 2 → 3	0	10	0
Improvement
State 2 → 1	0	36	0
State 3 → 1	0	0	3
State 3 → 2	0	0	16
Two or more transitions (n = 93)
Steady observed deterioration
State 1 → 2 → 3	0	0	0
Steady observed improvement
State 3 → 2 → 1	0	0	1
Fluctuating course, both deterioration and improvement
State 1 ⇔ 2	36	29	0
State 2 ⇔ 3	0	14	8
State 1 ⇔ 2, 1 ⇔ 3, 2 ⇔ 3	0	3	2

For the vast majority of patients, the observed transitions in disability occurred either between states 1 and 2 or between states 2 and 3. Relatively few individuals were observed moving directly to and from states 1 and 3. Figure 1 characterizes the disability process of patients in this study and assumes that an observed transition from state 1 to state 3 implies passage through state 2 (and vice versa), even if the time spent in state 2 is brief and unobserved.

Our estimated multistate model, with no covariates, provided estimates of 5.50, 2.26, and 2.61 years for the mean number of years in each of the 3 disability states, respectively. Their respective 95% confidence intervals (95% CIs) were 4.61–6.79, 1.96–2.68, and 2.05–3.58. As expected, Table 2 shows that observed transitions in disability state were more frequent among patients in state 2 and state 3 at their first HAQ assessment (69% and 60%, respectively) compared with those in state 1 (39.5%).

We examined the univariate effects of age, sex, duration of PsA, psoriasis severity as measured by the PASI, the number of clinically deformed joints, and the number of actively inflamed joints on transition rates in our model. The following results were obtained. The older a patient was at the time of the HAQ assessment, the slower the patient was to improve if he or she was currently in either state 2 (moderate disability) or state 3 (severe disability). The transition rate for moving from a higher state to a lower one is reduced by 8.5% (95% CI 3.1–13.5%) for a patient 5 years older than another patient, all else being the same. Male patients appeared to have a slower rate of decline in disability than female patients. Patients with duration of PsA less than 2 years were found to have more frequent transitions to different states (either to better or worse states). Patients who had PsA for 2–5 years and more than 5 years had a reduction in transition rates of 56–70% compared with those patients with PsA of less than 2 years' duration. There was no evidence to suggest an association between PASI scores and the transition rates (P = 0.08 and P = 0.79 for moving from a better to worse state, and for moving from a worse to better state, respectively). Finally, patients with a higher number of clinically deformed joints had, on average, a lower transition rate for improving (relative risk [RR] per joint increase 0.98; 95% CI 0.96–0.99), while those patients with a higher number of actively inflamed joints were quicker to show deterioration (RR per joint increase 1.04; 95% CI 1.02–1.07). When all of these variables (with the exception of PASI) were adjusted for in a multistate Markov model, the results shown in Table 3 were obtained and are similar to those obtained univariately.

Table 3. Results from the multivariate analysis that identified predictors of transitions between disability states*

Variable	Transitions
	1 → 2 2 → 3	2 → 1 3 → 2
	Relative risk (95% CI)	Relative risk (95% CI)
* 95% CI = 95% confidence interval; PsA = psoriatic arthritis.
Sex
Male	0.54 (0.38–0.76)	0.92 (0.66–1.28)
Female	1.00	1.00
Age	1.01 (0.99–1.03)	0.99 (0.97–1.00)
Duration of PsA, years
<2	1.00	1.00
2–5	0.42 (0.16–1.09)	0.33 (0.14–0.77)
>5	0.33 (0.14–0.76)	0.44 (0.21–0.90)
No. of clinically deformed joints	1.00 (0.99–1.01)	0.98 (0.97–0.99)
No. of actively inflamed joints	1.03 (1.01–1.06)	0.99 (0.97–1.01)
−2 log-likelihood	1,716.885

Finally, we investigated the impact of excluding the 78 patients with only 2 HAQ assessments (for whom a fluctuating course could not be observed). The results of this reanalysis were consistent with the reported results. We also examined the impact of redefining disability states into 5 categories: 0–0.49 (state 1), 0.5–0.99 (state 2), 1.0–1.5 (state 3), 1.51–1.99 (state 4), and 2–3 (state 5). For this analysis, we used the total group of patients and found that the reported findings did not materially change, with a single exception. Men were now found to move more rapidly than women back and forth between the newly defined states 2 and 3 (i.e., within the moderate disability state defined earlier for the 3-state model). However, men moved less frequently than women from state 1 to state 2 and between state 3 and state 4.

DISCUSSION

This study is the first to examine the longitudinal course of physical functional disability in PsA. Although several reports have indicated that patients with PsA experience reduced physical functioning (9–11), all of these studies have used a cross-sectional design. Because physical functional disability is known to fluctuate over time (26, 27), cross-sectional measures of physical function may provide a misleading picture of the burden of disability.

Our results indicated that although there was patient variability in the course of physical functional disability, 3 longitudinal patterns could be observed. The first reflected a stable state of disability throughout the study period, with 28% of the patients experiencing no disability over the study duration, and 12% and 6% experiencing moderate or severe disability, respectively. The second pattern was one of either steady improvement or deterioration and occurred in ∼27% of patients. With 1 exception, these individuals were observed to undergo a single transition only. The third pattern was characterized by multiple transitions and fluctuating states of disability over the study period and was also observed in 27% of patients.

Figure 1 illustrates the multistate Markov model used to characterize the disability process in PsA. This model allows for direct transitions only between no disability and moderate disability and between moderate disability and severe disability. Observed transitions from no disability to severe disability (and vice versa) involved passage through moderate disability. The model also provided estimates of time spent in each of the 3 disability states. The mean number of years spent in either the state of moderate or severe disability (2.26 and 2.61 years, respectively) was lower than that spent in the state of no disability (5.50 years), reflecting the fact that observed transitions in disability occurred more frequently in patients who entered the study with moderate or severe disability.

In the multivariate analysis, sex, age, disease duration, number of actively inflamed joints, and number of deformed joints significantly influenced transition rates. In the literature on disability (27–29), female sex has been consistently associated with higher levels of physical disability. Similar to these past studies, we found that being female increased the likelihood for progression of disability. In terms of age, increasing age was found to decrease the likelihood of improvement among patients with moderate or severe disability. Older patients were therefore more likely to experience persistent disability than were younger patients. Sherrer et al (30) found that age was the most powerful single predictor of subsequent disability in RA but showed that the predictive power of age is partially dependent on its interrelationships with other factors related to long-term disability, namely disease duration, comorbid conditions, and frailty. In this study, the effect of increasing age remained after adjusting for disease duration. With respect to disease duration, our results are consistent with those from longitudinal studies in RA (26, 31), reporting more variability in levels of disability during the early course of RA compared with that in the later course of disease. There are several possible explanations for this observation. Wiles et al (26) argue that in early RA, functional disability may fluctuate considerably due to spontaneous changes in disease activity, variability in timing and response to disease-modifying drugs, coping strategies, and adaptation to RA. In contrast, joint damage starts to accumulate in later disease, leading to an increase in persistent disability. It has also been suggested that the efficacy of treatment may be reduced over time. This, too, may result in an increase of enduring disability in later disease. Alternatively, patients with longer disease duration may represent a group of patients who failed to respond to past treatment and consequently experienced enduring disability. Interestingly, in the multivariate model, the effect of disease duration remained after adjusting for both number of actively inflamed joints and number of deformed joints. Because we did not include either treatment response or measures of coping and illness adaptation in our modeling approach, these variables might partially explain the observed relationship between disease duration and transition rates. As expected, a higher number of actively inflamed joints was associated with subsequent deterioration in disability, and the number of deformed joints reduced the likelihood of improvement in functional disability state.

Some methodologic issues related to this study need to be addressed in future research. To observe a fluctuating course in disability over time, all patients ideally should have at least 3 HAQ assessments. This was not the case for 78 patients who had only 2 HAQ assessments, although the findings did not materially change when these patients were excluded from the analysis. Ongoing followup of this sample will allow us not only to assess the robustness of our characterization of physical disability in PsA, but also to investigate additional factors that predict changes in the level of disability, such as medication history, comorbid conditions, and radiologically detected joint damage.

In summary, although 28% of patients appeared resistant to becoming disabled over the study duration, the remaining patients were observed either to have enduring disability or to move between disability states. Future research should identify factors in addition to age, sex, disease duration, and number of actively inflamed and deformed joints that predict changes in disability over the course of illness.