ARTHRITIS & RHEUMATISMVol. 43, No. 11, November 2000, pp 2402–2409© 2000, American College of Rheumatology

EARLY PREDICTORS OF POOR FUNCTIONAL OUTCOME INSYSTEMIC-ONSET JUVENILE RHEUMATOID ARTHRITIS

A Multicenter Cohort Study

LYNN R. SPIEGEL, RAYFEL SCHNEIDER, BIANCA A. LANG, NINA BIRDI, EARL D. SILVERMAN,RONALD M. LAXER, DEREK STEPHENS, and BRIAN M. FELDMAN

Objective. To examine the ability of a previouslydescribed set of criteria to predict poor functionaloutcome in a large, multicenter cohort of children withsystemic-onset juvenile rheumatoid arthritis (JRA).

Methods. All children who were diagnosed withsystemic-onset JRA since 1980 at the Hospital for SickChildren (Toronto), since 1983 at the Isaac WaltonKillam Hospital for Children (Halifax), and since 1981at the Children’s Hospital of Eastern Ontario (Ottawa)were evaluated. Patients were included in the study ifthey had been evaluated clinically within 6 months ofdiagnosis and had been followed up for at least 2 years.Patients were divided into 4 cohorts according to theirlength of followup: 2–4 years, 4–7 years, 7–10 years, and>10 years. Using previously described criteria for de-structive arthritis in children with systemic-onset JRA,the patients were classified as either high risk or lowrisk for poor functional outcome based on the data fromtheir 6-month visit. High-risk patients had active sys-temic disease (persistent fever or corticosteroid require-ment for control of systemic disease) and a plateletcount >600 3 109/liter. Poor outcome was defined asmoderate or severe disability (defined as a score of>0.75 on the Childhood Health Assessment Question-naire) or disease-associated death.

Results. Among 122 eligible patients withsystemic-onset JRA, we were able to contact 111 (91%)for outcome data. The mean followup period was 7.7years (SD 3.7). The mean age at outcome assessmentwas 13.5 years (SD 5.3). There were 51 boys and 60 girls.Twenty-four patients (22%) had moderate-to-severe dis-ability and 2 patients died; these 26 patients wereconsidered to have had a poor outcome. We coulddetermine risk classification for 104 patients. Twenty-four patients (23%) met the criteria for high risk at the6-month visit. Overall, the risk of a poor functionaloutcome was significantly higher in the high-risk group(relative risk 3.3, 95% confidence interval [95% CI]1.73–6.43, P 5 0.0004). This risk was most marked inthe cohort with >10 years of followup (relative risk 4.3,95% CI 1.82–10.29, P 5 0.006).

Conclusion. The presence of active systemic dis-ease at 6 months, as characterized by fever or the needfor corticosteroids, and thrombocytosis strongly pre-dicted the development of a poor functional outcome inthese patients. This was especially apparent with long-term followup. Our study validates the previously devel-oped prognostic criteria for systemic-onset JRA.

The systemic-onset subset of juvenile rheumatoidarthritis (JRA) accounts for ;15% of children diag-nosed with JRA. It is characterized by spiking fevers, anevanescent macular erythematous rash, arthritis, andadditional extraarticular features, which include serosi-tis, hepatosplenomegaly, lymphadenopathy, leukocyto-sis, thrombocytosis, and anemia (1). The systemic fea-tures of the disease will eventually remit in the majorityof patients, but the course of arthritis is variable (2).Long-term followup studies have shown that 15–37% ofpatients develop a debilitating, chronic, destructive ar-

Lynn R. Spiegel, MD, FRCPC, Rayfel Schneider,MBBch, FRCPC, Earl D. Silverman, MD, FRCPC, Ronald M. Laxer,MD, FRCPC, Derek Stephens, MSc, Brian M. Feldman, MD, MSc,FRCPC: The Hospital for Sick Children, University of Toronto,Toronto, Ontario, Canada; Bianca A. Lang, MD, FRCPC: IsaacWalton Killam Hospital, Grace Health Center, Halifax, Nova Scotia,Canada; Nina Birdi, MD, FRCPC: Children’s Hospital of EasternOntario, Ottawa, Ontario, Canada.

Address reprint requests to Lynn R. Spiegel, MD, FRCPC,Hospital for Sick Children, University of Toronto, 555 UniversityAvenue, Toronto, Ontario M5G 1X8, Canada.

Submitted for publication September 20, 1999; accepted inrevised form July 28, 2000.

2402

thritis (3–7). This subset of patients may have thepoorest outcome of all children with JRA.

Early aggressive therapy may modify the diseaseoutcome in systemic-onset JRA. Ravelli et al (8) retro-spectively studied 19 children with systemic-onset JRAwho had been treated with methotrexate (MTX). Theyreported that earlier treatment with MTX, possiblybefore the appearance of radiographic changes, mayfavorably influence the response to therapy. If earlyintervention can indeed influence prognosis, it is criticalto find early clinical and laboratory measures that canreliably predict the subset of patients who will develop apoor outcome.

In a previous study, we assessed one aspect ofpoor outcome in children with systemic-onset JRA. Wereported that persistent systemic symptoms (defined byprolonged fever or sustained corticosteroid require-ment) and thrombocytosis (platelet count $600 3 109/liter) at 6 months after disease onset were predictive ofa chronic, polyarticular arthritis course and early radio-graphic evidence of joint damage (9,10). Another impor-tant aspect of poor outcome in patients with systemic-onset JRA is poor physical function. However, thesedifferent measures of poor outcome are not necessarilycorrelated. For example, significant roentographic jointdisease does not necessarily reflect functional disability,as demonstrated by Svantesson et al, who showed that of33 systemic-onset JRA patients with advanced radio-graphic changes, half were in functional classes I and II(3). Therefore, in order to assess the predictive validity ofour previously defined criteria for poor outcome insystemic-onset JRA, we applied them to a different aspectof outcome, specifically, functional ability. We wanted tosee if we could predict how patients will ultimatelyperform in their activities of daily living as measured bythe Childhood Health Assessment Questionnaire(CHAQ). The CHAQ has been validated for assessmentof functional outcome in children with JRA (11).

PATIENTS AND METHODS

Patients. We performed a multicenter cohort study ofpatients diagnosed with systemic-onset JRA from 1980 to 1994.We established an inception cohort of all patients diagnosedwith systemic-onset JRA between 1980 and 1994 at the Hos-pital for Sick Children in Toronto, Ontario, between 1983 and1994 at the Isaac Walton Killam Children’s Hospital in Halifax,Nova Scotia, and between 1981 and 1994 at Children’s Hospi-tal of Eastern Ontario in Ottawa. The different time frames fordata collection at the different centers are a reflection of theavailability of reliable patient information. The study wasapproved by the ethics review boards at all 3 centers.

To be included in the study, patients had to meet theAmerican College of Rheumatology (ACR) classification cri-teria for systemic-onset JRA (12), be evaluated within the first6 months of disease onset, and after initial evaluation, befollowed up for a minimum of 2 years. Data were collectedfrom the medical records at the time of diagnosis, 6 monthsafter disease onset, and at the most recent clinic visit. Infor-mation collected included demographic data (age, sex, geo-graphic location), clinical data (fever, rash, hepatomegaly,splenomegaly, lymphadenopathy, arthritis, pericarditis, pleuri-tis), treatment data (systemic corticosteroids, intraarticularcorticosteroids, MTX, intravenous gammaglobulin [IVIG],hydroxychloroquine, sulfasalazine, gold, azathioprine,D-penicillamine, cyclophosphamide, cyclosporine), and labo-ratory data (hemoglobin level, white blood cell count, plateletcount, erythrocyte sedimentation rate, antinuclear antibodytiter, presence of rheumatoid factor).

Definition of risk groups. Children were classified ashigh or low risk for poor functional outcome based on datafrom their 6-month visit (61 month). According to the previ-ously established ACR criteria, high-risk patients had activesystemic disease (persistent fever or corticosteroid require-ment for control of systemic disease) and a platelet count$600 3 109/liter.

Assessment of functional outcome. All subjects com-pleted the CHAQ between 1996 and 1997, the time period duringwhich the study was undertaken. The questionnaire was eitherself-administered at the time of a hospital visit or conductedover the telephone by a study investigator (LRS). The inves-tigator was blinded to each patient’s risk classification.

The CHAQ is a 30-item self-report or parent-reportquestionnaire. It examines 8 domains of physical function,which include dressing and grooming, arising, eating, walking,hygiene, reach, grip, and activities. In addition, it assesses theneed for aids and devices. For each activity, subjects are askedwhether they can perform the task without any difficulty, withsome difficulty, or with much difficulty, or are unable toperform it. These options are scored 0, 1, 2, and 3, respectively.A fifth option—not applicable—is included for those instanceswhen the activity is age dependent and the child is too youngto do it. An overall score is obtained by averaging the scores ofthe worst activity in each of the 8 domains (11).

We determined that patients had moderate-to-severedisability if their CHAQ score was $0.75. This was calculatedas follows: if patients have some difficulty (mild disability) in 6domains, much difficulty (moderate disability) in 3 domains, orwere unable to perform the task (severe disability) in 2domains, their CHAQ score would be 0.75. This has beenvalidated in a study in which parents of children with JRAcompleted questionnaires that assessed disability (13); cutofflevels for the CHAQ scores, corresponding with mild, moder-ate, and severe levels of disability, were determined for thesepatients. Thus, for the purposes of this study, poor functionaloutcome was defined by a CHAQ score $0.75 or the occur-rence of a disease-associated death.

Statistical analysis. Patients were divided into 4 co-horts according to length of followup (i.e., the number of yearsfrom the time of disease onset to completion of the CHAQ), asfollows: 2–4 years, 4–7 years, 7–10 years, and .10 years.Statistical tests were performed on each followup stratum.Data from the 3 clinical centers were analyzed separately and

PREDICTORS OF OUTCOME IN SYSTEMIC-ONSET JRA 2403

then as a group. The primary analysis compared functional-outcome status (dependent variable) with risk-group status(independent variable) as determined above. As a secondaryanalysis, we evaluated a number of other variables to assesstheir ability to predict functional outcome. Bivariable compar-isons were performed to screen for potential predictors. Lo-gistic regression modeling using backward stepwise eliminationtechniques was then performed.

RESULTS

Of the 159 systemic-onset JRA patients identi-fied, 122 met our inclusion criteria. Thirty-seven patientsdid not meet our inclusion criteria because they had notbeen seen within the first 6 months of diagnosis. Wewere able to contact 29 of these 37 patients to administerCHAQs. We found no differences in CHAQ scoresbetween the study patients and the non–study patients.

Of the 122 eligible patients, we were able tocontact 111 (91%) for outcome data. There were 70patients (63%) from Toronto, 26 (23%) from Halifax,and 15 (14%) from Ottawa. Demographic, clinical,laboratory, and treatment data were reviewed for allpatients. There were 51 boys and 60 girls. Most featureswere similar among the patients from the 3 centers.However, significantly more patients from Halifax con-tinued to have fever at 6 months compared with children

from the other 2 centers (Table 1). Therefore, based onour risk criteria, there were significantly more patientsfrom Halifax who were classified as being at high risk(46%, compared with 7% from Ottawa and 16% fromToronto; P 5 0.001).

There were 21 patients followed up for 2–4 years,32 patients followed up 4–7 years, 25 patients followedup 7–10 years, and 33 patients followed up .10 years.The mean followup was 7.7 years (SD 3.7) and wassimilar for all 3 centers. The mean age at outcomeassessment was 13.5 years (SD 5.3). Twenty-four pa-tients (22%) had moderate-to-severe disability as de-fined by a CHAQ score $0.75, and 2 patients haddisease-associated death. These 26 patients, of whom 17were from Toronto (including the 2 patients who died),4 from Halifax, and 5 from Ottawa, were considered tohave had a poor outcome.

We could determine risk classification for 104patients (7 patients did not have platelet countsmeasured at the 6-month visit). Twenty-four patientsmet the criteria for high risk at the 6-month visit: 11were from Toronto, 12 from Halifax, and 1 fromOttawa. Overall, the likelihood of a poor functionaloutcome was significantly higher in the high-riskgroup (relative risk 3.3, 95% confidence interval [95%

Table 1. Clinical and laboratory data at 6 months’ followup and subsequent followup in patients withsystemic-onset juvenile rheumatoid arthritis*

Total(n 5 111)

Halifax(n 5 26)

Ottawa(n 5 15)

Toronto(n 5 70)

Descriptive dataMean age at diagnosis, years 6.1 5.5 4.3 6.7Mean age at time of CHAQ, years 13.5 11.9 12.0 15.0Mean followup, years 7.7 6.4 7.7 8.0Classified at high risk, % patients 22.0 46.0 7.0 16.0Median (range) CHAQ score 0.125 (0–3) 0.125 (0–2.125) 0.375 (0–2.5) 0 (0–3)Good outcome, % patients 77.0 85.0 67.0 76.0Poor outcome, % patients 23.0 15.0 33.0 24.0

Clinical data at 6 months, % patientsFever 29.0 58.0 20.0 20.0Rash 40.5 58.0 27.0 37.0Arthritis 62.0 73.0 67.0 57.0Lymphadenopathy 21.0 27.0 20.0 19.0Hepatomegaly 12.5 19.0 13.0 10.0Splenomegaly 12.5 15.0 13.0 11.0Pericarditis 2.0 4.0 1.0 1.0Pleuritis 1.0 4.0 0.0 0.0Corticosteroids (any dose) 41.0 38.0 47.0 40.0

Laboratory data at 6 months, meanWBC, 3 109/liter (n 5 106) 13.4 15.5 11.7 13.0Hemoglobin, gm/dl (n 5 106) 10.5 10.0 10.5 10.7Platelets, 3 109/liter (n 5 104) 527.1 596.0 546.0 495.0ESR, mm/hour (n 5 101) 34.1 34.0 41.0 33.0

* CHAQ 5 Childhood Health Assessment Questionnaire; WBC 5 white blood cells; ESR 5 erythrocytesedimentation rate.

2404 SPIEGEL ET AL

CI] 1.73–6.43) (Figure 1 and Table 2). This risk wasmost apparent in the cohort with .10 years’ followup(relative risk 4.3, 95% CI 1.82–10.29) (Figure 1). The

sensitivity of the risk classification for determining apoor outcome was 0.5, and the specificity was 0.85.

We were able to collect treatment data on 109

Figure 1. Outcome measured by Childhood Health Assessment Questionnaire (CHAQ) in high-risk andlow-risk patients with systemic-onset juvenile rheumatoid arthritis (as defined in Patients and Methods).Box plots show the CHAQ results in all study patients, patients with 2–4 years’ followup, patients with 4–7years’ followup, patients with 7–10 years’ followup, and patients with followup .10 years. The limits of theboxes represent the 25th and 75th percentile values, while the line across the box is the median value.The whiskers depict the range of the results, the circles show the outliers, and the asterisks represent theextreme outliers. The shaded area represents the 95% confidence interval around the median value.

PREDICTORS OF OUTCOME IN SYSTEMIC-ONSET JRA 2405

patients (Table 3). We assessed whether differences intherapy over the time period of the study could haveaffected the predictive ability of our risk criteria. Wedefined early therapy as treatment with DMARDswithin the first 2 years of disease. Sixty-five patients(60%) were started on at least 1 DMARD early in theirdisease. Of these patients, 36 were administered MTX.We found that overall, the frequency of early use ofDMARDs was similar for the 4 followup cohorts. How-ever, in the cohort followed up for .10 years, signifi-cantly fewer patients were treated with MTX. Since mostof the patients followed up for .10 years were fromToronto, there were significantly fewer Toronto patientswho received MTX early, compared with patients in theother centers.

We were able to collect treatment data on 102 ofthe 104 patients for whom we had assigned risk classifi-cation (Table 3). As expected, there were significantlymore high-risk patients than low-risk patients treatedearly with DMARDs. Significantly more patients treated

early with DMARDs had a poor outcome. For thosehigh-risk patients started early on DMARDs, the rela-tive risk for a poor outcome was 2.24; for the subsetstarted on MTX alone or in combination with otherDMARDs, the relative risk was 3.2; for those who werenot started on MTX, but received other DMARDs, therelative risk was 2.96. Therefore, criteria for high riskwere highly predictive of poor outcome despite earlytreatment with DMARDs and, specifically, MTX.

As a secondary analysis, we sought other predic-tive variables in addition to our previously definedpredictive criteria. We wanted to determine if therewere additional prognostic factors that could predictpoor outcome that had not been identified in our earlierstudy. These included demographic data (age, sex, geo-graphic location), clinical data (fever, rash, hepatomeg-aly, splenomegaly, lymphadenopathy, monarticular arthri-tis, pauciarticular arthritis, or polyarthritis, pericarditis,pleuritis), treatment data (use of systemic cortico-steroids, intraarticular corticosteroid injections, MTX,IVIG, hydroxychloroquine, sulfasalazine, gold, azathio-prine, D-penicillamine, cyclophosphamide, cyclo-sporine), and laboratory data (hemoglobin level, whiteblood cell count, platelet count, erythrocyte sedimenta-tion rate, antinuclear antibody titer, rheumatoid factorpositivity).

Exploratory logistic regression showed that at 6months, the presence of fever, use of corticosteroids,and thrombocytosis were the only statistically significantpredictors of functional outcome. Study patients diag-

Table 2. Outcome as measured by Childhood Health AssessmentQuestionnaire (CHAQ) in high- and low-risk patients with systemic-onset juvenile rheumatoid arthritis*

Poor outcome(CHAQ $0.75)

Good outcome(CHAQ ,0.75) Total

Low risk 12 68 80High risk 12 12 24Total 24 80 104

* Values are the number of patients.

Table 3. Early therapy with disease-modifying antirheumatic drugs (DMARDs)*

Any DMARD withinfirst 2 years

Methotrexate withinfirst 2 years†

Yes No P Yes No P

Outcome cohort (n 5 109) 0.152 0.00142–4 years (n 5 21) 9 12 9 124–7 years (n 5 32) 23 9 16 167–10 years (n 5 24) 16 8 9 15.10 years (n 5 32) 17 15 2 30

Study site (n 5 109) 0.152 0.017Toronto (n 5 68) 36 32 16 52Halifax (n 5 26) 19 7 13 13Ottawa (n 5 15) 10 5 7 8

Risk group (n 5 102) 0.0001 0.0002High 22 2 15 9Low 37 41 18 60

Outcome group (n 5 109) 0.0005 0.2158Poor 23 3 11 15Good 42 41 25 58

* No correction for multiple hypothesis testing. Values are the number of patients.† Methotrexate alone or in combination with other DMARDs.

2406 SPIEGEL ET AL

nosed before age 5 had a worse outcome compared withthose diagnosed later; however, this difference was notstatistically significant (relative risk 1.86, P 5 0.08). Theneed for corticosteroids and the platelet count at 6months were the most important prognostic factors(Figure 2). The risk of having a poor outcome wasmultiplied by 1.45 for every 100 3 109/liter increase inplatelets (95% CI 1.14–1.85). The relative risk of havinga poor outcome for patients undergoing treatment withcorticosteroids was 2.63 (95% CI 1.52–4.55).

Of the 111 patients included in the present study,25 were included in our earlier study (9). Exclusion ofthese patients did not change our results. In addition,when the 2 patients who died were removed from theanalysis, our results were unchanged. The 11 patientslost to followup were analyzed in a best case/worst casescenario, with no difference in our results. There was 1patient in the high-risk group with severe developmentaldelay, and 2 patients in the low-risk group with coexist-ing fibromyalgia. When these patients were excludedfrom the analysis, there was no change in our results.

DISCUSSION

The results of this multicenter cohort study of 111patients with systemic-onset JRA show that it is possible

to predict poor functional outcome in patients withsystemic-onset JRA early in their disease course, withthe use of a set of simple criteria. Specifically, thepresence of active systemic disease (persistent fever orcorticosteroid requirement) and a platelet count$600 3 109/liter at 6 months following presentation isassociated with the development of a poor functionaloutcome as measured by the CHAQ. This is especiallyapparent with long-term followup, since the risk of apoor outcome was greatest in the high-risk cohort ofchildren who were followed up for .10 years.

There have been a number of predictors of poorprognosis in systemic-onset JRA cited in the literature;however, these predictors have varied from study tostudy. They have included age ,6 years at disease onset,persistent disease of .5 years’ duration, presence ofcardiac disease, and elevated IgA level (3–5,14). Anselland Wood retrospectively studied 79 patients withsystemic-onset JRA who were followed up over a 15-year period. They found that patients whose diseaseonset was observed at ,6 years of age fared worse (4).Calabro et al prospectively studied 20 patients withsystemic-onset JRA. They found that patients with dis-ease that persisted .5 years had the worst functionaloutcome (5). Svantesson et al prospectively studied 33patients with systemic-onset JRA, with followup rangingfrom 3 years to 24 years. They concluded that risk factorsfor disability included cardiac disease, rapid progressionwithin the first 5 years of diagnosis, and age ,5 years atonset (3). We found that patients diagnosed before age5 tended to have a poorer outcome; however, thistendency was not statistically significant. Cassidy et almeasured immunoglobulin concentrations in 200 pa-tients with systemic-onset JRA. They reported that thereappeared to be a direct correlation between IgA levelsand the extent of articular and systemic disease inchildren (14).

In an earlier study (9), our group demonstratedthat persistent systemic symptoms and thrombocytosis 6months after disease onset predict one aspect of pooroutcome, namely, persistent polyarthritis and early ra-diographic evidence of joint damage. The present studyvalidates these predictive criteria and uses physicalfunction as another measure of poor outcome. Weanalyzed multiple variables, including demographic,clinical, therapeutic, and laboratory data. There were noadditional factors that were significant predictors offunctional outcome in this study.

The risk of a poor functional outcome was mostmarked in the cohort followed up for .10 years. Wefound that there were no statistically significant differ-

Figure 2. Predicted likelihood of a good outcome with or withouttreatment with corticosteroids at 6 months in patients with systemic-onset juvenile rheumatoid arthritis. The y-axis shows the probability ofa good functional outcome given our definition (see Patients andMethods). The x-axis shows the platelet count at 6 months.

PREDICTORS OF OUTCOME IN SYSTEMIC-ONSET JRA 2407

ences between the 4 groups in terms of overall use ofDMARDs in the first 2 years of disease. However, veryfew patients in the cohort followed up for .10 yearswere treated with MTX early in their disease course.This may explain, in part, why this group had thegreatest risk for a poor outcome. This study, however,was not designed to assess the efficacy of differenttherapeutic agents. Therefore, we are unable to properlyevaluate whether MTX does indeed improve functionaloutcome in these patients.

We considered the possibility that differences intherapeutic regimens over time could have affected thepredictive ability of our criteria. Overall, significantlymore patients in the high-risk group and the poor-outcome group were started on DMARDs, and specifi-cally, MTX, within the first 2 years of diagnosis. Thislikely reflects the severity of disease and the decision totreat this group more aggressively. Among high-riskpatients, the relative risk for a poor outcome was notsignificantly different in those treated early withDMARDs. This held true for patients treated with MTXalone or in combination with other DMARDs. There-fore, our analysis showed that the criteria remainedhighly predictive for those patients treated early withDMARDs, including MTX.

While the combination of active systemic diseaseand thrombocytosis was an important prognostic indica-tor, we found that the single most important predictor ofpoor functional outcome was the need for cortico-steroids for the control of systemic symptoms at 6months. A possible explanation for this is that patientswho have more systemic inflammation are more likely tobe treated with corticosteroids. Many of the clinical andlaboratory variables analyzed are interrelated. Further-more, they tend to improve with corticosteroid therapy,thus minimizing their contribution as significant inde-pendent predictors of outcome.

We found that these criteria have high specificity,but low sensitivity. Traditional management of patientswith systemic-onset JRA has avoided early aggressivetherapy because of the risk of toxicity in a large numberof patients who have a very good prognosis. With highlyspecific prognostic criteria, those patients identified asbeing at high risk and very likely to have a poor outcomecan therefore be considered for early, more aggressivetreatment. However, because of the low sensitivity ofthese criteria, some patients who will have a pooroutcome will not be identified.

There were some potential limitations to thisstudy. Ideally, newly incident cases should be used tostudy prognosis. However, given the low incidence of

systemic-onset JRA in children, we used a retrospec-tively identified inception cohort. For this reason, wewere missing some risk data for 7 patients. Of 159systemic-onset JRA patients identified, 122 met ourinclusion criteria. Twenty-nine patients were excludedbecause we had not examined them within the first 6months of disease onset. We were concerned that theexclusion of this group could potentially lead to sam-pling bias. Therefore, we administered CHAQs to thesepatients to permit comparison with the study population.There was no difference between the 2 groups. Thus, webelieve that our results are likely to be generalizable.

There were a number of issues with regard toindividual patients that could have confounded ourresults. There was a patient in the high-risk group withsevere developmental delay and 2 patients in the low-risk group with fibromyalgia. There was concern thattheir CHAQ results would be influenced by these addi-tional diagnoses. Therefore, 2 separate analyses weredone that first included, and then excluded these pa-tients. There were no differences in our findings. Inaddition, of the 111 patients included in the study, 25were included in our earlier study. We performed aseparate analysis without this subgroup and obtainedsimilar results.

In analyzing the data for each center, we discov-ered that there were significantly more high-risk patientsfrom Halifax compared with Toronto or Ottawa. Fur-thermore, patients from this center tended to have abetter outcome. We noted that patients from Halifaxwere more likely to be treated with MTX within the first2 years of disease. Although this study was not designedto evaluate treatment efficacy, these results suggest thatearly therapy with MTX may have contributed to abetter functional outcome in this group. However, therisk criteria were still highly predictive in the Halifaxpatients. When we compared demographic and clinicaldata between centers, we could not find any othersignificant differences.

In determining risk classification, we includedpatients receiving corticosteroids at any dosage. There-fore, at 6 months following diagnosis, some patients mayhave been receiving a tapering dose of corticosteroidsbecause they were improving clinically. If they also hadthrombocytosis, they would have been included in thehigh-risk group. This may explain why some high-riskpatients did not have significant disability when outcomewas assessed.

In conclusion, this study shows that persistence ofsystemic features and a high platelet count at 6 monthspredict a poor functional outcome in children with

2408 SPIEGEL ET AL

systemic-onset JRA. It will now be important to prospec-tively study the value of using early, more aggressivetherapy in the group of patients identified as being athigh risk for chronic destructive arthritis and poorfunction.

REFERENCES

1. Cassidy JT, Petty RE. Textbook of pediatric rheumatology. 3rd ed.Philadelphia: WB Saunders; 1995. p. 145–6.

2. Schaller JG. Juvenile rheumatoid arthritis: series 1. ArthritisRheum 1977;20:165–70.

3. Svantesson H, Akesson A, Eberhardt K, Elborgh R. Prognosis injuvenile rheumatoid arthritis with systemic onset: a follow-upstudy. Scand J Rheumatol 1983;12:139–44.

4. Ansell BM, Wood PHN. Prognosis in juvenile chronic polyarthri-tis. Clin Rheum Dis 1976;2:397–412.

5. Calabro JJ, Holgerson WB, Sonpal GM, Khoury MI. Juvenilerheumatoid arthritis: a general review and report of 100 patientsobserved for 15 years. Semin Arthritis Rheum 1976;5:257–98.

6. Hanson V, Kornreich HK, Bernstein B, King KK, Singsen BH.Three subtypes of juvenile rheumatoid arthritis: correlations ofage at onset, sex, and serologic factors. Arthritis Rheum 1977;20:184–6.

7. Mozziconnacci P, Prieur AM, Hayem F, Oury C. Prognosticarticulaire de la forme systemique de l’arthrite chronique juvenile(cent cas). Ann Pediatr 1983;30:553–6.

8. Ravelli A, Romenghi B, di Fuccia G, Ruperto N, Zonta L, MartiniA. Factors associated with response to methotrexate in systemiconset juvenile chronic arthritis. Acta Pediatr 1994;83:428–32.

9. Schneider R, Lang BA, Reilly BJ, Laxer RM, Silverman ED,Ibanez D, et al. Prognostic indicators of joint destruction insystemic onset juvenile rheumatoid arthritis. J Pediatr 1992;120:200–5.

10. Lang BA, Schneider R, Reilly BJ, Silverman ED, Laxer RM.Radiologic features of systemic onset juvenile rheumatoid arthri-tis. J Rheumatol 1995;22:168–73.

11. Singh G, Athreya BH, Fries JF, Goldsmith DP. Measurement ofhealth status in children with juvenile rheumatoid arthritis. Arthri-tis Rheum 1994;37:1761–9.

12. Cassidy JT, Levinson JE, Bass JC, Baum J, Brewer EJ Jr, FinkCW, et al. A study of classification criteria for a diagnosis ofjuvenile rheumatoid arthritis. Arthritis Rheum 1986;29:274–81.

13. Dempster HT, Porepa M, Young N, Feldman B. Clinical relevanceof functional measures in children with juvenile rheumatoidarthritis and other musculoskeletal disability [abstract]. UnivToronto Med J 2000;77:S29.

14. Cassidy JT, Petty RE, Sullivan DB. Abnormalities in the distribu-tion of serum immunoglobulin concentrations in juvenile rheuma-toid arthritis. J Clin Invest 1973;52:1931–6.

PREDICTORS OF OUTCOME IN SYSTEMIC-ONSET JRA 2409