- Top of page
- PATIENTS AND METHODS
Pain is an important symptom in children with chronic arthritis. However, only a small number of studies have explored factors influencing pain perception in children with juvenile rheumatoid arthritis (JRA) or juvenile chronic arthritis (1–14). These studies, mainly from single centers and involving small numbers of patients with variable disease durations, have suggested that only a small proportion of the variation in pain measures can be explained by disease activity. Most of the pain variance demonstrated in these studies could not be explained by any of the demographic or disease-related variables examined.
In an attempt to better understand the factors influencing pain in children with JRA, we examined pain scores in a large, well-characterized, multicenter cohort of children who had been diagnosed with JRA at least 5 years prior to the study. Pain experienced over the previous week, measured on a 10-cm visual analog scale (VAS), was analyzed for its association with demographic and disease-related variables. Long-term data about this cohort have been reported previously (15).
- Top of page
- PATIENTS AND METHODS
Patient characteristics and pain scores are shown in Table 1. Several variables were not available for all 388 patients; the total number available for each variable is shown in Table 1. Univariate analyses (Table 2) showed positive correlations with pain for active disease duration, active joint count, morning stiffness, and age at study. Among categorical variables, significant differences in pain scores were apparent for PGA scores, race categories, and JRA onset subtypes. For PGA, a score of 0 was associated with the lowest median pain score, and a score of 3 with the highest median pain score. For race, the highest median value was obtained for the aboriginal and part-aboriginal group. Among onset subtypes, the pauciarticular group had the lowest and the polyarticular RF-positive group had the highest median value. Place of residence and sex of the patient were the only variables that were not significantly associated with pain.
Table 1. Patients studied: clinical characteristics, assessment of disease activity, and pain scores by age groups*
| ||Age 8–15 years n = 183||Age ≥16 years n = 205||P|
|Onset subtype, %|| || || |
| Pauciarticular||66.7||47.8|| |
| RF-negative polyarticular||18.0||22.9|| |
| RF-positive polyarticular||2.7||17.1|| |
|Race, white/aboriginal or part aboriginal/other, %||84.7/10.4/4.9||81.5/14.6/3.9||NS†|
|Residence, urban or suburban/rural/reserve, %||62.2/35.0/2.8||62.9/30.3/6.8||NS†|
|Active disease duration, median (range) years||5.4 (0.2–14.3)||6.8 (0.2–22.5)||0.001‡|
|Pain score by VAS, median (range) cm||0.2 (0–10)||1.4 (0–10)||<0.0001‡|
|Morning stiffness|| || || |
| Patients assessed, no.||122||105|| |
| Morning stiffness, median (range) hours||0 (0–5)||0 (0–14)||0.002‡|
|Active joint count|| || || |
| Patients assessed, no.||155||156|| |
| Median (range)||0 (0–42)||0 (0–40)||NS‡|
|PGA score|| || || |
| Patients assessed, no.||165||157|| |
| PGA score 0, %||57.0||52.9||NS†|
| PGA score 1, %||27.3||28.7|| |
| PGA score 2, %||13.9||15.9|| |
| PGA score 3, %||1.8||2.5|| |
Table 2. Univariate correlations with pain*
|Variable||Category||Number in category||Median pain score for category (range)||Spearman's correlation coefficient||P|
|Active disease duration||NA||NA||NA||0.459||<0.0001|
|Active joint count||NA||NA||NA||0.424||<0.0001|
| ||1||90||2.0 (0–10)|| || |
| ||2||48||3.1 (0–9.3)|| || |
| ||3||7||4.6 (0–7.4)|| || |
|Morning stiffness||NA|| ||NA||0.517||<0.0001†|
|Age at study||NA|| ||NA||0.258||<0.0001|
| ||Male||87||0.2 (0–9.7)|| || |
| ||Aboriginal or part aboriginal||49||2.3 (0–10.0)|| || |
| ||Other||17||0.2 (0–6.0)|| || |
|Residence||Urban or suburban||241||0.5 (0–10.0)||NA||NS†|
| ||Rural||126||0.6 (0–9.8)|| || |
| ||Reserve||19||1.5 (0–8.3)|| || |
|Onset subtype||Systemic||48||0.7 (0–9.3)||NA||0.001†|
| ||Pauciarticular||220||0.2 (0–10.0)|| || |
| ||Polyarticular RF−||80||1.1 (0–10.0)|| || |
| ||Polyarticular RF+||40||2.6 (0–10.0)|| || |
For the multivariate analysis, complete data excluding morning stiffness were available for 301 patients (Table 3). If morning stiffness had been included as an independent variable, fewer patients could have been entered into the model (see below). Independent correlations with pain were determined for active disease duration, PGA, and age. The final regression model explained 22% of the variation in pain scores (Table 3). Collectively, disease activity as determined by PGA scores >0 could explain 6.5% of the variation in pain scores. Higher PGA scores had the greatest effects. For example, when disease duration and age at study are kept constant, an increase in PGA score from 0 to 2 or 3 increased pain scores by 2.1 cm. Active disease duration had a moderate effect resulting in a 1.28-cm increase for every 10 years, when the other variables are controlled; however it explained most of the variation of pain scores in the regression.
Table 3. Multivariate analysis for independent correlations with pain scores for all patients and for patients stratified by age group*
|N R2||All patients (n = 301) (R2 = 22.0)||Age 8–15 years (n = 153) (R2 = 11.0)||Age ≥16 years (n = 148) (R2 = 25.0)|
|B (95% CI)||P||% explained†||B (95% CI)||P||% explained||B (95% CI)||P||% explained|
|Active disease duration||0.128 (0.063, 0.192)||<0.0001||14.3||E|| || ||0.177 (0.106, 0.248)||<0.0001||16.6|
|PGA 2 or 3||2.093 (1.287, 2.899)||<0.0001||5.3||E|| || ||2.269 (1.158, 3.380)‡||<0.0001||8.4|
|PGA 1||0.756 (0.046, 1.465)||0.037||1.2||E|| || ||E|| || |
|Active joint count||E|| || ||0.100 (0.049, 0.151)||<0.0001||8.4||E|| || |
|Age at study||0.064 (0.011, 0.117)||0.018||1.3||0.159 (0.008, 0.311)||0.040||2.6||E|| || |
|Constant||−0.533 (−1.428, 0.362)||0.242|| ||−0.773 (−2.671, 1.124)||0.422|| ||0.696 (−0.042, 1.435)||0.064|| |
To further evaluate the effect of age at study, the patient population was stratified into 2 age groups (Table 3). An effect of age was evident in the age group 8–15 years but not in the ≥16-year group. Age had more than twice the effect in the 8–15-year group than found for the entire group (B = 0.159 versus 0.064; Table 3), although the proportion of pain variation explained by age was small (2.6% and 1.3%, respectively). Disease activity remained an independent predictor for both groups. However in the younger age group, the significant correlation was with active joint counts, whereas in the older age group it was with global scores (Table 3). Active disease duration was not an independent variable in the younger age group but was in the older age group.
Data on morning stiffness were available for 227 patients. Morning stiffness had a very high correlation with pain in the univariate analysis (Table 2). Complete data for inclusion in stepwise linear regression analyses were available for 195 patients. In this model, a total of 27% of the pain variance could be explained, and morning stiffness proved to be an independent variable (B = 0.0122; 95% confidence interval 0.006, 0.018; P < 0.0001, where morning stiffness is measured in minutes), accounting for an estimated 19% of the variation. Active disease duration and a PGA of 2 or 3 accounted for the remaining 8% of the variation (analysis not shown).
In a separate analysis for the entire patient group, no interactions between onset subtype and disease duration or between onset subtype and active joint count were found when pain was the outcome variable (data not shown).
- Top of page
- PATIENTS AND METHODS
This study of pain, in perhaps the largest number of children with JRA reported in the literature, confirms the findings of several smaller studies that a small amount of pain variance can be explained by demographic or disease-related variables. In a study of 23 children with JRA, Varni et al (3) found that present pain intensity measured by a VAS completed by the child, a parent, or a physician correlated positively with the physician's assessment of disease activity. In a study of 18 children with JRA, Ilowite et al (7) found that only between one-quarter and one-third of present pain intensity (using a VAS) could be explained by disease activity (using thermography as the measure of joint inflammation). Hagglund et al (8) studied 60 children with JRA and found an inverse correlation between disease duration and children's rating of average pain over the last month in univariate analysis. However, disease duration was not a significant factor in the hierarchical regression model in which demographic data (age and socioeconomic status) and disease status (disease duration and articular severity index) explained only 8% and 10% of the pain variance, respectively. A study by Schanberg et al (12) of 56 children with juvenile chronic arthritis found that disease activity explained 28% of pain variance, with patient age and disease duration explaining only 2% and 1%, respectively.
Disease activity accounted for only 6.5% of the pain variance in our study, whereas disease duration explained a relatively larger proportion. A possible explanation for the latter is that our patient population included older patients with longer disease durations who may have reached a threshold effect with time. This is borne out in the analysis of the 8–15-year-old group in which disease duration was not significant. However, direct comparisons are difficult because of differences in the independent variables and the characteristics of the patient populations examined. For example, >50% of patients in our study had inactive disease compared with only 16% in Schanberg's study. Another explanation for the disparity between studies is the assessment of present rather than worst pain. It has been pointed out that better correlations are found with scores for worst pain during the previous week than with present pain scores obtained at the time of assessment of disease activity (9). In the study by Hagglund et al (8), only 10% of the pain variance could be explained by disease activity; but young children may have had difficulty with memory recall for rating average pain over the previous month.
The effect of age on pain has been much discussed in the literature, and the data is conflicting. Earlier studies suggested children with JRA reported less pain than adults with rheumatoid arthritis (1), but it is generally accepted that these results were due to failure to ask about pain in a developmentally appropriate manner. It has also been reported that children with JRA aged 8–11 years had lower scores compared with those 12–17 years of age (2). Conversely, the study by Ilowite et al (7) found a strong correlation between pain scores and joint inflammation for 9 children <7.86 years of age, with >50% of the pain variance being explained by inflammation. For 9 older children, however, there was only a weak and negative correlation between pain and joint inflammation. Other studies have found no difference in scores between these age groups (5, 9). Compared with other studies, this present report included subjects with a wider age range (8–32 years). An effect of age was noted only in the 8–15-year age group, supporting a difference in reports of pain between middle childhood and adolescence when using a VAS.
Unfortunately, information on morning stiffness was not available for about one-quarter of the patients. Morning stiffness was correlated strongly with pain intensity in univariate analyses. As far as we are aware, no other study has investigated the role of morning stiffness in pain in JRA. Although morning stiffness was an independent variable in these analyses, it should be noted that it is possible that stiffness may not be truly independent of pain, but may be a component of pain in arthritis, and children may not really be able to discriminate between the sensations of pain and stiffness.
We have not discussed the relationship between pain and function in this study. In our previous analysis (18) and in other studies (19–21) of both JRA and rheumatoid arthritis, however, pain has been shown to have a significant impact on activities of daily living and other measures of physical function.
Several of the reports discussed above have investigated other nondemographic and non–disease-related factors that might contribute to pain variance. The model that has to date explained the greatest amount of the pain variance in children with JRA was proposed by Thompson et al (4). Their model included the Family Relationship Index of the Family Environment Scale; the internalizing, externalizing, and social subscales of the Child Behavior Checklist; and the disease parameters JRA onset subtype and disease activity. Thirty-four percent of the variation in present pain and 72% of the worst pain intensity could be explained by this model. Hagglund et al (8) hypothesized that, in addition to being positively correlated with disease status, pain would be positively correlated with depressed mood and feelings of hopelessness, and negatively correlated with social support. However, they found that this model could account for only 24% of the pain variance, and that the psychosocial factors accounted for only 6% of the variance. As already discussed, the fact that the authors asked the children to rate their average pain over the previous month may have adversely affected the study results. Schanberg et al (12) were able to explain 54% of the pain variance using a model that included age, disease duration, disease activity, and pain coping factors. The pain coping factors explained 26% of the variance compared with 28% for disease activity. In other words, the children who reported a high ability to control and decrease pain experienced less pain. However, as the authors point out, it will need longitudinal and interventional studies to know the direction of the relationship between coping and pain, e.g., whether the coping strategies truly affect pain perception or whether children with less pain are more likely to believe that they can control the pain than children experiencing more severe pain.
Our study was originally designed to determine predictors of long-term outcome in children with JRA, with pain being one outcome measure. As a consequence, it lacks information on many of the psychosocial factors, in particular coping strategies, that other studies have indicated probably contribute to pain perception. In addition, the number of disease-associated variables collected was limited. For example, articular disease course, rather than onset subtype, might have given additional insights. Additional information about joints with limited mobility or joint damage might possibly have increased the proportion of pain explained. The study also used a uniform VAS for all age groups. Although the reading level of all instructions was adjusted to a grade 3 level, it is possible that the VAS used may not have been appropriate for the younger patients in the study. As noted by Varni et al (3), an ideal assessment of pain in children requires an interdisciplinary, multidimensional, and comprehensive approach, combining self-report, behavioral, cognitive, socioenvironmental, medical, and biologic parameters. We acknowledge that our study lacks many of these components and would have been strengthened by the use of more than one pain instrument, with the measurement of pain on more than one occasion.
This study does have the strength of studying a large number of children followed over a long period. Despite its limitations, we believe that it provides useful data about pain in children with JRA. It confirms that pain is an important symptom in many children with long-standing JRA, and that pain in JRA is a multifaceted condition that cannot be simply explained by demographic or disease-related factors. These findings stress the importance of looking further at other factors, such as coping, mood, and family structures, both in research studies and in day-to-day clinical work when evaluating pain in children with JRA and presumably with other forms of chronic arthropathy.