Features of the 394 patients at the baseline SF-36 administration are shown in Table 1. At baseline, 74% of patients were known to have ever used DMARDs. The corresponding percentages for those who ever used immunosuppressive drugs, methotrexate, oral steroids, nonsteroidal antiinflammatory drugs, and biologic agents were 64.2%, 62.2%, 12.2%, 70%, and 26.0%, respectively.
Table 1. Demographic and clinical features of the 394 patients at their baseline SF-36 administration*
| ||No. (%)||Mean ± SD||Median (IQR)|
|Demographics|| || || |
| Sex|| || || |
| Female||114 (28.9)|| || |
| Male||280 (71.1)|| || |
| Age, years|| ||44.88 ± 12.72||44.63 (35.24–54.07)|
|Clinical history|| || || |
| Age at psoriasis diagnosis, years|| ||27.78 ± 13.93||26 (17–38)|
| Age at PsA diagnosis, years|| ||36.10 ± 12.61||34 (26.50–45.00)|
| Psoriasis duration, years|| ||17.07 ± 11.48||15.64 (7.71–24.65)|
| Arthritis duration, years|| ||8.70 ± 7.95||6.65 (2.02–13.17)|
| History of HAQ pain†|| ||1.047 ± 0.773||0.880 (0.403–1.595)|
| History of SJC†|| ||3.34 ± 4.67||2.00 (0.56–4.00)|
|Baseline clinical measures|| || || |
| Active joint count|| ||7.25 ± 8.01||5 (2–11)|
| SJC|| ||3.28 ± 4.91||2 (0–4)|
| Damaged joint count|| ||5.08 ± 9.72||0 (0–5)|
| ESR, mm/hour|| ||21.43 ± 18.90||15 (8–28)|
| Psoriasis Activity and Severity Index|| ||5.94 ± 8.79||3 (1.2–6.9)|
|Baseline patient-reported measures of pain and psychological and physical functional status|| || || |
| HAQ pain|| ||1.032 ± 0.789||0.860 (0.360–1.640)|
| Mild pain‡||190 (55.9)|| || |
| Moderate pain‡||98 (28.8)|| || |
| Severe pain‡||52 (15.3)|| || |
| No. missing||54|| || |
| SF-36 MCS score§|| ||49.44 ± 10.45||52.03 (41.94–57.53)|
| Probable depressive disorder: MCS score ≤42||99 (25.1)|| || |
| Probable depressive disorder: MCS score ≤35||47 (11.9)|| || |
| SF-36 PCS score§|| ||40.52 ± 10.94||40.86 (32.76–49.71)|
Baseline and longitudinal characteristics of depressive symptoms and pain.
At baseline, the mean ± SD SF-36 MCS score in this predominantly male sample of 394 patients was 49.44 ± 10.45 (Table 1), somewhat lower than published norms for the US male general population (51.21) but similar to that of the US female population (48.89). Baseline means ± SDs of the SF-36 MCS score by men and women in our study sample of 394 patients were similar (49.43 ± 10.36 and 49.46 ± 10.71, respectively). Based on suggested cutoffs of 42 for medical patients and 35 for chronic pain patients (27, 28), approximately 25% and 12%, respectively, had a probable depressive disorder. The mean ± SD HAQ pain score at baseline was 1.03 ± 0.79, indicating moderate pain levels. Approximately 56%, 29%, and 15% of the sample reported mild, moderate, and severe pain, respectively. Pain history indicated moderate levels of chronic or episodic pain (mean ± SD 1.03 ± 0.79).
Of 16 baseline comparisons made between the 317 study patients used in the final multivariate analyses and the 77 patients excluded because of missing information, only age at diagnosis of psoriasis, psoriasis duration at the first SF-36 assessment, the PASI, and the SF-36 physical component summary (PCS) score at the first assessment were found to be statistically significant (P < 0.05), with the 317 patients on average younger at psoriasis diagnosis, having a longer mean psoriasis duration, and having higher mean PASI and mean PCS scores. The mean baseline PCS scores in these 2 groups were 41.25 and 37.51, respectively.
At baseline, pairwise complete Pearson's correlations between the actual levels of HAQ pain and the SF-36 MCS, HAQ pain and SJC, and the SF-36 MCS and SJC were modest, at −0.26, 0.24, and −0.13, respectively. There were sizeable within-patient variations in the actual levels of depressive symptoms, pain, and SJC during the followup period. That is, for depressive symptoms, a within-patient SD of 6.95 compared to a between-patient SD of 7.38, which corresponds to 47.0% of the total variation in the SF-36 MCS after adjustment for arthritis duration, was found. For pain, the within- and between-patient SDs were 0.56 and 0.52, respectively, corresponding to 52.8% of the total variation in pain after adjustment for arthritis duration. The corresponding numbers for swollen joint counts were 2.48, 2.16, and 56.8%, respectively.
Longitudinal associations between depressive symptoms and pain, controlling for inflammatory disease activity.
Table 2 shows the final multivariate analyses that modeled the interplay between changes in depressive symptoms, pain, and SJC between clinic visits, controlling for other covariates and based on the 317 patients described earlier. These models included both lagged and the other concurrent change measures and explained 37.4%, 14.1%, and 50.6% of the variance in changes in HAQ pain, MCS scores, and SJC, respectively. The strongest correlates of changes in pain, depressive symptoms, and SJC between visits were previous visit scores of the corresponding variables, with standardized regression coefficients exceeding 0.75 in absolute value (Table 2 and black lines in Figure 1). Higher previous HAQ pain (or the worse the pain at the previous visit) was associated with a larger improvement in pain (or lesser worsening in pain between visits). A similar pattern was observed for SJC. Conversely, higher previous MCS scores (or few depressive symptoms at the previous visit) were associated with worsening of depressive symptoms (or a smaller improvement in depressive symptoms between visits).
Table 2. Final multivariate linear mixed-effects models for the 3 change outcomes of HAQ pain, SF-36 MCS, and SJC*
| ||Change in HAQ pain, β (SE)†||Change in SF-36 MCS score, β (SE)†||Change in SJC, β (SE)†|
|Intercept||−0.248 (0.187)||0.223 (0.227)||0.367 (0.159)|
|Sex (male vs. female)||−0.096 (0.070)||−0.119 (0.086)||0.037 (0.060)|
|Age at PsA diagnosis, years||0.005 (0.003)||0.001 (0.003)||−0.008 (0.002)|
|HAQ pain history at first SF-36 administration||0.338 (0.048)||−0.071 (0.059)||−0.092 (0.042)|
|SJC history at first SF-36 administration||−0.044 (0.010)||−0.023 (0.012)||0.027 (0.008)|
|Arthritis duration at previous visit, years||0.006 (0.003)||0.004 (0.004)||−0.010 (0.003)|
|HAQ pain at previous visit†||−0.856 (0.028)‡||−0.166 (0.040)||0.281 (0.030)|
|SF-36 MCS score at previous visit†||−0.251 (0.031)||−0.794 (0.029)‡||0.079 (0.027)|
|SJC at previous visit†||0.385 (0.040)||0.083 (0.046)||−0.920 (0.025)‡|
|Change in HAQ pain†|| ||−0.210 (0.028)||0.248 (0.022)|
|Change in SF-36 MCS score†||−0.185 (0.024)|| ||0.034 (0.021)|
|Change in SJC†||0.342 (0.031)||0.061 (0.034)|| |
|Unadjusted R2, %||37.4||14.1||50.6|
|Adjusted R2, %||35.3||13.8||49.6|
Figure 1. Diagram displaying the clinically important standardized lagged and cross-association effects between Health Assessment Questionnaire pain (Pain), depressive symptoms (mental component summary [MCS]), and disease activity (swollen joint count [SJC]) variables.
Download figure to PowerPoint
These results will partly reflect pure regression to the mean, because even if there were no association between the previous and current scores, a negative correlation would be seen between the change and previous scores. To assess the theoretical relationships between the previous and current scores for the 3 outcomes, the (unstandardized) effect estimates of previous on current scores for HAQ pain, MCS, and SJC were calculated from the effect estimates of −0.86, −0.79, and −0.92, respectively, in Table 2 by rescaling using known factors (based on the equivalence between the 2 types of linear mixed-effects models as mentioned earlier) of 0.98, 0.91, and 0.95, respectively, and then adding 1. These unstandardized regression estimates are 0.16 (SE 0.03), 0.28 (SE 0.03), and 0.12 (SE 0.03), respectively, and reflect the autoregressive nature of the 3 outcomes, with statistically significant positive associations existing between previous and current scores for the 3 outcome variables, and thus reflecting, over the time period, more than just pure regression to the mean. That is, when previous scores are high the current scores also tend to be high.
There was evidence of a bidirectional association between depressive symptoms and pain over the followup period, although the effects were small in magnitude (absolute effect sizes <0.3) relative to those found for the association of change with previous scores (Table 2 and Figure 1). Both previous and change scores between clinic visits in MCS scores were associated with change in pain between visits, even after adjusting for SJC and other covariates (blue lines in Figure 1). Specifically, higher MCS scores (or fewer depressive symptoms) at the previous visit were associated with improvement in pain between visits (or lesser worsening of pain symptoms; β = −0.25 [SE 0.03]), and a positive change in the MCS score (diminishing depressive symptoms) was associated with an improvement in pain between visits (β = −0.19 [SE 0.02]). Conversely, both previous and change between clinic visits in pain scores were associated with change in depressive symptoms between visits, after adjusting for SJC and other covariates (red lines in Figure 1). Higher HAQ pain scores (more pain) at the previous visit were associated with decreasing MCS scores or increasing depressive symptoms (β = −0.17 [SE 0.04]) between visits, and a larger positive change in HAQ pain (increasing pain symptoms) between visits was associated with a worsening in depressive symptoms between visits (β = −0.21 [SE 0.03]).
Previous SJC and change in SJC had moderate effects (absolute β effect sizes between 0.3 and 0.75) on change in pain between visits (green lines in Figure 1). Both a higher SJC at the previous visit and a positive change in SJC (increasing disease activity; β = 0.39 [SE 0.04] and β = 0.34 [SE 0.03], respectively) were associated with worsening of pain symptoms between visits. There were no significant relationships between SJC (both previous and change measures) and change in MCS scores after adjusting for HAQ pain measures.
Of further note, there were weak effects between pain and change in SJC, with higher HAQ pain (more pain) at the previous visit and a positive change in HAQ pain score (increasing pain symptoms) between visits being associated with increasing inflammatory joint activity (or lesser improvement of joint activity) between visits (β = 0.25 [SE 0.02] and β = 0.28 [SE 0.03], respectively) (Table 2 and red lines in Figure 1). There was no evidence that illness duration was associated with change in pain or in depressive symptoms between visits or modified the relationship between pain and depression.
Similar results were obtained when the SF-36 bodily pain measure was entered into the multivariate model, replacing HAQ pain, as was the case when the acute-phase reactant measure erythrocyte sedimentation rate replaced SJC. Moreover, there was no evidence of effects for medications ever used up to the previous visit (both DMARDs and biologics) on any of the change outcomes in the final models. Furthermore, although some interaction effects of lagged variables with sex, arthritis duration, age at PsA diagnosis, and history measures were found, these had no clinically significant impact on change outcomes beyond their main effects. For example, the interaction effect estimate of sex with previous MCS was −0.11 (SE 0.09), whereas the main effect estimate of the previous MCS was −1.16 (SE 0.11), an order 10 times higher.