Patients and methods

Patient population

The study sample originated from the University of Toronto Psoriatic Arthritis Clinic, Toronto, Ontario, Canada, established in 1978 to follow patients with PsA prospectively. Patients are included in the clinic if they have an inflammatory arthritis associated with psoriasis, and more than 98% fulfilled the CASPAR criteria for the classification of PsA.33 34 Patients are seen at 6–12-month intervals and at each visit undergo a complete assessment according to a standard protocol.33 Demographic features, medications and general medical history are obtained. Physical examination includes: average grip strength (right and left hand, scales 0–300 mm Hg); total number of actively inflamed joints (stress pain, joint line tenderness, and/or swelling, scale 0–68); number of swollen joints (0–66), number of deformed joints (anklyosis, subluxation or decreased range of motion of more than 20%, attributable to joint damage rather than inflammation, scale 0–66); number of fibromyalgia tender points (American College of Rheumatology criteria, scale 0–18) and severity of psoriasis (psoriasis area and severity index (PASI), scale 0–72).35 36 Laboratory measures included erythrocyte sedimentation rate (ESR) and haemoglobin. Anaemia was defined as haemoglobin of less than 120 g/l for women and less than 130 g/l for men. Increased ESR was defined as more than 20 mm/h for women and over 15 mm/h for men.17

Patient reported outcomes

In the early 1990s patients were asked to complete both the HAQ and medical outcome survey short form 36 (SF-36) on an annual basis.22 37 The HAQ assesses physical function and includes questions related to fine movements of the upper extremity, locomotor activities of the lower extremity and activities involving both upper and lower extremities. It consists of 20 questions that cover eight categories of daily living (eg, dressing and grooming, eating, walking and activities including errands and chores). The scores for all eight categories are averaged to obtain an overall score on a scale from 0 (no disability) to 3 (severe disability). The SF-36 questionnaire assesses functional status, wellbeing and general health perceptions and consists of eight subscales. The bodily pain and mental subscales were used to measure pain and psychological distress, respectively. The SF-36 bodily pain subscale was selected over the HAQ visual analogue pain due to a lower percentage of missing data (6% vs 13%, respectively) and the high correlation (r  =  −0.81) between the two measures. There was also a high correlation (r  =  −0.77) between the mFSS and the SF-36 vitality subscale, suggesting that these scales may measure the same construct. In a previous study we showed that the mental health subscale of the SF-36 provided information that was additive to the traditional clinical measures of disease activity, severity and function.10 The subscales are calculated by summing the individual items within each scale and then transforming the individual scale score to a 0–100 range, with higher scores indicating better functioning or wellbeing. Both the HAQ and SF-36 have been shown to be reliable in our clinic population.8 10 Since 1998, PsA patients have been asked to complete the mFSS at yearly intervals.30 The mFSS includes nine items asking about the extent to which fatigue influences motivation, exercise, physical functioning, duties and responsibilities, work, family and social life. Patients rate each item on a scale from 0 (not at all) to 10 (entirely), with an average overall score (0–10) being computed.31 38 A higher score indicates more severe dysfunctional fatigue. Moderate to severe fatigue is defined as FSS scores of 5 or over; severe fatigue is defined as FSS scores of 7 or over.24 The study sample consisted of 499 patients who completed at least one FSS questionnaire between 1998 and 2006. We concentrate here on the data from these patients at or before the first FSS assessment.

Statistical analysis

Descriptive information on the 499 patients at the first FSS assessment was provided. Initially, univariate linear regression analyses were performed to investigate separately the cross-sectional associations of the various demographic, clinical, laboratory and medication variables with FSS. Variables with p values below or on the borderline of the 0.05 significance level were considered for inclusion in multivariate analyses. Because there were large numbers of variables examined, some of which may be highly correlated with each other and/or possibly belong to the same underlying latent domain, a factor analysis on all the variables, using varimax rotation, was performed. The factor analysis was used to determine whether variables would cluster into separate and clinically sensible domains based on the factor loadings.26 Moreover, it was used to assist in reducing the number of variables for inclusion in the final multiple linear regression analyses. Initially, we restricted multivariate analyses to within domains, to assess the importance of within-domain-specific variables multivariately, but in isolation of variables in other domains. Based on the results of these domain-specific analyses and the univariate analyses (R² values and p values), as well as the percentage of missing information in variables and existing previous knowledge about the association between certain variables (eg, the strong negative correlation (r  =  −0.61) between grip strength and the HAQ), further multiple linear regression analyses were performed to construct an overall final multivariate model.

Results

Of the 499 patients, there were 214 women and 285 men, with a mean age at first FSS assessment of 48.5 years and a mean disease duration of 12.7 years for arthritis and 20.5 years for psoriasis. The baseline clinical characteristics at the time of initial FSS assessment are shown in table 1. PASI scores ranged from 0 to 60, with a mean of 6.0. There was a mean of 8.1 actively inflamed joints, 2.8 swollen joints and 7.0 clinically damaged joints, reflecting both moderate disease activity and severity. Twenty per cent of women and 14% of men had anaemia (p = 0.13), whereas 53% of women and 39% of men had raised ESR (p = 0.01). The mean HAQ score was 0.7 and the mean SF-36 bodily pain score (lower scores indicate a higher degree of pain) was 48.6, indicating moderate physical disability and a fair degree of pain. Patients reported moderate psychological functioning (mean SF-36 mental health score 68.7), and among the 233 patients with data on antidepressant medication, 18% reported “ever use”.

The mean FSS score was 5.0 (SD 2.9) (women 5.9 (SD 2.8) and men 4.3 ( SD 2.7)), similar in magnitude to that observed in patients with SLE (mean 5.1, SD 3.1) and higher than that seen in normal healthy adults (mean 3.9, SD 2.1).26 31 38 Approximately 49.5% and 28.7% had moderate to severe (FSS ⩾5) and severe (FSS ⩾7) fatigue, respectively.

The majority of patients had taken non-steroidal anti-inflammatory drugs (NSAID), disease-modifying antirheumatic drugs (DMARD), an immunosuppressive drug, or methotrexate. Seventeen per cent has used an oral steroid. Only 24 (5%) of patients had “ever used” biological agents.

Table 2 shows the results of the univariate regression analysis that examined the relationship between the individual correlates and level of fatigue, as measured by the FSS. As shown, the vast majority of variables were significantly associated with fatigue. Female patients and older patients at FSS assessment were more likely to report fatigue. Relative to normotensive patients, patients with hypertension were more likely to report higher fatigue levels. Clinical indicators of increasing disease activity and damage were associated with increasing fatigue levels, as were clinical and self-reported measures of physical limitations and pain. Higher levels of fatigue were also associated with poorer psychological functioning, as measured by the SF-36 mental health scale. In terms of laboratory variables, higher ESR was associated with higher fatigue levels, whereas higher haemoglobin levels were associated with less fatigue. Based on the R² values, the HAQ, the SF-36 pain and mental health scales and the number of fibromyalgia points were the most strongly associated with level of fatigue (R² 38.1%, 36.8%, 27.4% and 20.7%, respectively), followed by average grip strength, number of active joints and morning stiffness (R² 12.9%, 12.7% and 11.6%, respectively). Finally, present and “ever use” of PsA medications (with the exception of DMARD) were significantly related to higher fatigue. Variables not associated univariately with fatigue were duration of arthritis, psoriasis activity or severity (PASI) and history of cancer.

The results of the factor analysis on all the variables are shown in table 3. As expected, there was evidence of strong correlations among some variables, leading to nine generally clinically sensible domains. Consistent with previous studies, the mental health dimension appeared to provide information additional to that typically obtained from other self-reported health status and clinical measures of function and pain.9 10 Multivariate analyses within domains were then conducted to assess the importance of within-domain-specific variables simultaneously.

The final multivariate model included 6 variables and explained 54.5% of the variance in FSS scores (see table 4).

The p values for all the variables were below or on the borderline of the 0.05 significance level. The findings suggest that among the variables studied—female sex, higher HAQ scores (increased physical disability), lower SF-36 pain and mental health scores (reflecting higher levels of pain and psychological distress), higher number of fibromyalgia tender points and “ever use” of methotrexate—were most related to fatigue. The level of psychological distress accounted uniquely for approximately 6.6% of the R² explained, whereas HAQ, the SF-36 pain score and the number of fibromyalgia tender points each uniquely accounted for 2.3%, 1.3% and 1.9%, respectively of the R² explained. Sex and “ever use” of methotrexate only accounted uniquely for 0.5% and 0.9% of the R² explained. The remaining 41% of the R² explained was shared among the six variables. In particular, 24.5% of the R² explained could be attributed entirely to having the SF-36 mental and pain scores and HAQ together in the model (ie, not shared with sex and/or “ever use” of methotrexate and/or number of fibromyalgia tender points).

The number of active joints failed to remain in the final model due to the presence of HAQ. In multivariate models that entered HAQ and active joints together, the number of active joints became non-significant. Indeed, most of the disease-related factors found to be significant at the univariate level did not achieve statistical significance in the context of HAQ and pain measures. We also ran multivariate models that replaced “ever use” of medication with their “since last clinic visit” counterparts, and that included “ever use” of immunosuppressive medication instead of “ever use” of methotrexate. The results are essentially the same as those shown in table 4 (data not shown).

Discussion

This study confirmed the importance of fatigue, as measured by the FSS, in a large, cross-sectional sample of PsA patients. Approximately 49.5% of patients reported at least moderate fatigue and 28.7% experienced severe fatigue. This was higher than that previously observed in healthy controls (34% and 8%, respectively) but comparable to that reported in other rheumatic conditions.26 31 38 At the univariate level, pain as measured by the SF-36 pain scale and the number of fibromyalgia tender points was significantly associated with the level of fatigue and explained 36.8% and 21.2% of the variance in FSS scores, respectively. However, pain was not the only factor to play a significant role in fatigue. Most of the demographic, disease-related and health status variables studied were significantly related to the level of fatigue univariately, with physical disability (HAQ) and psychological distress (SF-36 mental health scale) explaining 38.1% and 27.4% of the variance in FSS scores, respectively. Disease duration, psoriasis activity and severity (PASI) and a history of cancer were the few exceptions. Six variables—female sex, the SF-36 pain and mental health scales, the number of fibromyalgia tender points, HAQ and “ever used” methotrexate—comprised the final multivariate model and jointly explained 54.5% of the variation in FSS scores. This suggested that the FSS was not a redundant measure with the core set of outcome measures in PsA (ie, pain, physical function HAQ and disease-related factors).17 In the multivariate model, the unique contribution of pain measures (SF-36 pain and the number of fibromyalgia tender points) and of the HAQ to the explained variance in fatigue scores was smaller than that of the SF-36 mental health scale. Most of the disease-related factors (number of active joints, morning stiffness, ESR, average grip strength), although significant at the univariate level, did not achieve statistical significance in the context of other variables, namely the HAQ and pain measures.

Our findings are consistent with the results of studies in other rheumatic conditions, indicating that the primary correlates of fatigue are pain, psychological distress, female sex and physical disabilities.26 27 28 30 It has been suggested that fatigue may predominantly reflect psychosocial distress, including an inability to cope with disease, rather than a true indicator of inflammatory disease.25 The use of a cross-sectional design in this study and the strong correlations among some variables make it difficult to draw conclusions about the impact of inflammatory disease on self-reported fatigue in PsA. Indeed, disease activity and severity and the psychosocial correlates may be in the same causal chain, with increased disease activity or severity antedating and leading to diverse manifestations, including psychological distress, functional impairments, pain and fatigue. According to this theory and consistent with our results, the strongest correlates of fatigue would be those closest to fatigue in the causal chain (ie, other disease manifestations such as pain and functional limitations). Also the significant association between “ever used” methotrexate and fatigue in our multivariate model may reflect treatment effects or alternatively past or present disease severity. Moreover, the effect of female sex may partly reflect gender differences in laboratory parameters of inflammatory activity, with women having higher rates of raised ESR than men (p = 0.01). Longitudinal assessments are needed to describe the temporal sequence among disease-related variables, physical limitations, pain and fatigue. Indeed, in a preliminary study changes in fatigue have been shown to reflect changes in disease activity in PsA over time.32 Longitudinal studies could also clarify the causal link between psychological distress and fatigue in PsA and other rheumatic conditions. In the RA literature, patients with a history of psychological distress or depression report higher levels of fatigue over the illness course than patients without such a history. It is believed that a depressive cognitive set plays an important role in the experience and manifestation of a chronic illness such as RA.39 In a related article, it was shown that respondents with depression comorbid with one or more chronic diseases (angina, arthritis, asthma and diabetes) had the worst health scores of all the disease states.40 In this study, in the subgroup of patients with data on antidepressant medication (n  =  233), 18% reported “ever use”, and at the univariate level there was a significant association between “ever use” and fatigue (p<0.001 and R² 11.5%).

Future studies also need to address some additional limitations of this study. We did not measure some potentially important determinants of fatigue, including sleep quality, comorbidities other than cancer and hypertension, smoking status and body mass index. Moreover, there was no evidence of a relationship between fatigue and psoriasis activity and severity, despite some evidence in the literature.21 This may reflect either the relatively low level of psoriasis in the majority of the sample or the insensitivity of the FSS to psoriasis-related fatigue. Future studies are required to replicate our findings in other PsA populations using additional measures of fatigue.

In closing, fatigue is a common symptom in PsA, and is associated with pain, female sex, physical disability, past and current medication status and psychological distress. Fatigue appears to provide some information that does not overlap with the core set of outcome domains in PsA. Longitudinal studies will improve our understanding of the temporal relationship between demographic variables, health status, disease-related factors and fatigue.