Article Text

Extended report
Early predictors of prognosis in juvenile idiopathic arthritis: a systematic literature review
  1. E H Pieter van Dijkhuizen1,2,
  2. Nico M Wulffraat2
  1. 1Pediatria II, Reumatologia, IRCCS G. Gaslini, Genoa, Italy
  2. 2Department of Paediatric Immunology, University Medical Centre Utrecht, Wilhelmina Children's Hospital, Utrecht, The Netherlands
  1. Correspondence to E H Pieter van Dijkhuizen, Pediatria II, Reumatologia, IRCCS G. Gaslini, Largo Gaslini, 5, 16147 Genova, Italy; E.H.P.Dijkhuizen{at}


Objectives Juvenile idiopathic arthritis (JIA) is subdivided into seven categories. Even within these categories, the prognosis varies markedly. To start appropriate treatment in patients with JIA and to inform patients and their parents correctly, it is essential to know the individual prognosis, preferably at the time of diagnosis. The aim of this study was to identify variables that predict disease activity, joint damage, functional ability and quality of life (QoL) early in the disease course.

Methods A systematic literature review was performed, and 3679 articles were identified. The results were screened and critically appraised using predefined criteria. Articles that described validated outcomes, such as the Wallace criteria, the childhood health assessment questionnaire (CHAQ) and the juvenile arthritis damage index (JADI), and that determined predictors in the first 6 months of disease were selected.

Results Forty mostly retrospective articles were selected. Polyarticular onset predicted a worse prognosis for all outcomes, except QoL. A diagnostic delay and the systemic category predicted continuation of active disease. Notably, antinuclear antibodies (ANA) did not predict disease activity. Symmetric involvement and rheumatoid factor positivity predicted less damage. More disease activity was mainly associated with worse functional outcome. However, most predictors were not validated.

Conclusions Few predictors for the selected outcomes were found. Prospective, longitudinal studies using standardised outcome measurements, and evaluating a broader range of predictors, such as genetics, immunological and imaging data, should be performed. For the outcomes joint assessment and quality of life, standardised and validated outcomes should be developed.

  • Juvenile Idiopathic Arthritis
  • Disease Activity
  • Treatment
  • Outcomes research
View Full Text

Statistics from


Juvenile idiopathic arthritis (JIA) is characterised by joint inflammation of unknown aetiology, lasting for at least 6 weeks with an onset before 16 years of age.1 It is classified into seven categories based on the number of joints involved in the first 6 months and the presence of extra-articular symptoms, such as fever, rash, enthesitis or psoriasis.1 ,2 JIA is heterogeneous because of differing symptoms and a varying prognosis even within the same category, ranging from a short and modest inflammation of a single joint to long-lasting, debilitating inflammation of multiple joints, leading to joint destruction and lifelong incapacity.3–6

Nowadays, a range of effective medication stands at the physician's disposal, from non-steroidal anti-inflammatory drugs (NSAIDs) and steroidal intra-articular injections, to disease-modifying antirheumatic drugs (DMARDs) like methotrexate, and biologicals such as etanercept and adalimumab.7–13 In adult rheumatology, early and effective treatment in the so-called window of opportunity improves the outcome.14–19 In paediatric rheumatology, the time to initiation of methotrexate therapy was shown to be an important factor for the response.20 In a recent study of early aggressive therapy in polyarticular JIA, fairly high response rates were achieved,21 while early treatment of systemic JIA with anti-interleukin (IL)1 therapy was very successful and reduced the use of steroids markedly.22 It seems therefore likely that starting early and aggressive therapy in children with JIA can prevent long-term sequelae. On the other hand, it is desirable to spare milder cases aggressive treatment for both safety and economic reasons. Consequently, it is essential to know the prognosis for the individual patient early in the course of disease and preferentially at the time of diagnosis in order to start the right treatment immediately. Furthermore, patients and their parents not only want to know what kind of disease JIA is in general, but especially how it will affect their personal lives and prospects, for which, too, it is essential to know the individual prognosis.

Much effort has already been done to elucidate predictors of the prognosis. These studies have been reviewed by Adib et al in 2005.3 ,4 Major finding at that time was the variability among studies owing to a lack of standardised criteria. Since then, however, the Wallace criteria for disease remission have been developed and validated,23–26 potentially leading to a reduced variability among studies and the ability to draw consistent conclusions.

Therefore, the aim of this study was to systematically review the literature for studies that determine predictors for the prognosis within 6 months after the diagnosis of JIA. We focused on the outcome measurements disease activity, joint damage, functional ability and quality of life (QoL).


On 3 October 2013, a systematic literature search was performed in PubMed, Embase, The Cochrane Library and PsycInfo. Using the algorithm in table 1, 3679 articles were identified (figure 1). All titles and abstracts were screened for applicability to the research subject using predefined criteria. Studies that evaluated systemic JIA only were excluded because this category is thought to have a different pathogenesis, warranting a separate review.1 ,27–31 Likewise, studies only assessing predictors for the development of uveitis were excluded.

Table 1

Search strategy*

Figure 1

Flow chart. DDO, domain, determinant and outcome; sJIA, systemic juvenile idiopathic arthritis.

Articles that seemed relevant were selected for in-depth full-text screening. If a full text could not be retrieved, the author was contacted. The references of selected articles were screened as well. If a relevant article had been missed, the title and abstract of that article were screened to identify the relevant keywords. These were then added to the search algorithm and the search was repeated until no more relevant articles were found. Doing so, the articles were screened several times. Table 1 shows the final search algorithm.

Selected articles were critically appraised using predefined criteria (see online supplementary table S1). In order to reduce the expected variability among the articles,4 the outcome definition and the time of determination of the predictors were emphasised. Outcome measurement regarding disease activity was focused on the validated Wallace criteria for clinical remission.23–26 Studies in which remission was defined much like the Wallace criteria were included as well, but other definitions were excluded. Regarding the other outcomes, validated measurements were emphasised, such as the juvenile arthritis damage index (JADI)32 and the childhood health assessment questionnaire (CHAQ),33 although we were less strict regarding the outcomes joint damage and QoL since no single, broadly recognised evaluation is available in these fields. As an additional criterion, we selected articles in which the predictors were determined within 6 months after the diagnosis to enable timely prediction of the prognosis.

Finally, the two authors discussed which articles would be included for analysis based on the critical appraisal.


Forty articles with original research and seven reviews were identified after full-text screening (figure 1), of which the original research articles were critically appraised (see online supplementary table S1). Of those, 27 assessed disease activity, 14 joint damage, 22 functional ability and 4 QoL (multiple outcomes possible). Almost none of the articles described the process of blinding and none described whether the group with the unfavourable outcome was eligible to receive the same care as the group with a favourable outcome. Studies using the validated outcomes and determining predictors within 6 months after diagnosis were selected.

Disease activity

Ten studies assessed disease activity (table 2). Due to our selection, all studies used Wallace criteria or very similar criteria as outcome measurement. Nonetheless, there was still some heterogeneity since these criteria were applied in different ways. For example, some studies used the percentage of patients achieving either clinical inactive disease (CID), clinical remission on medication (CRM) or clinical remission off all medications (CR),24 ,34 ,36–38 ,40 ,41 whereas others assessed the percentage or cumulative amount of time spent in CID, CRM or CR.24 ,35 ,39 Other study parameters differed as well, most notably the duration of follow-up that varied from 6 months to a median of 15.3 years. Only two studies were prospective studies.34 ,36 One cross-sectional study was included, which assessed only a genetic predictor.35 Two studies assessed only a single category of JIA.37 ,38

Table 2

Characteristics of included studies

Table 3 shows the results the studies obtained, demonstrating that most of the variables were clinical and routine laboratory predictors. Only few were genetic, there were hardly any imaging-related and no immunological (ie, cytokine, cellular characteristics) predictors. Only demographics, JIA category, antinuclear antibodies (ANA) and few others were evaluated more than once. The results show that demographics were not able to predict disease activity, with the notable exception of a longer duration between disease onset and diagnosis, which predicted a worse prognosis in a multivariate analysis in two studies.36 ,40 The results of JIA category reflected the known worse prognosis in the systemic and polyarticular categories. Because of our selection criteria, only onset categories could be evaluated, making it impossible to note differences between the oligoarticular persistent and extended categories. Disease activity parameters showed a doubtful ability to predict a worse prognosis. Finally, ANA positivity did not predict disease activity.

Table 3

Results for outcome disease activity*

Joint damage

Seven studies evaluated joint damage, none of them using the JADI (table 2).32 Instead, other outcome measures were used, such as the Dale or Steinbrocker classification.57 ,58 Other studies assessed the presence of joint space narrowing, erosions or ankylosis, without specifying a classification or scoring system.43 ,47 ,48 Follow-up ranged from 5 years to a median of 14.9 years, but was unclear in one study.43 Three studies were prospective cohort studies.43–45 Two studies assessed oligoarticular onset JIA patients only.46 ,47

Table 4 shows the results. Again, it is notable that almost all predictors were clinical and laboratory values. Only a few genetic and no imaging or immunological predictors were studied. Due to the paucity of studies, many predictors were assessed only once. Nonetheless, the results revealed a worse prognosis in the polyarticular categories. Of note is the negative predictive value of symmetric joint involvement41 ,47 and of rheumatoid factor (RF) positivity41 ,45 in two studies. This may be due to correlations with the polyarticular categories because the independent effect of these predictors was not studied in multivariate analysis together with JIA category. Other negative predictors included disease activity parameters, which, however, were mostly not validated.

Table 4

Results for outcome absence of joint damage*

Functional ability

Fifteen studies were selected, assessing functional ability (table 2). All but two used the CHAQ33 or the HAQ when appropriate as outcome measurement. However, in some studies, the score was dichotomised, mostly using a score of 0.75 as the cut-off point, whereas in other studies it was assessed as a continuous score. Two studies used the 36-item Short-Form health survey (SF-36).37 ,41 Follow-up ranged from 6 months to as much as 23 years. In total, 6 out of 15 studies were prospective cohort studies36 ,43–45 ,49 ,50 and one was a cross-sectional study,51 which assessed genetic predictors only.

Table 5 shows the results. Demographic variables showed varying results. JIA category showed the worse prognosis for polyarticular JIA again. Among disease activity parameters, quite some significantly associated predictors were identified. In general, more disease activity at diagnosis was predictive for a worse prognosis. The two most notable predictors were a higher CHAQ score at baseline and, again, the presence of symmetric joint involvement. The laboratory parameters showed some mixed results, making it difficult to draw an overall conclusion. Some genetic factors, being human leucocyte antigens (HLA) and single nucleotide polymorphisms (SNPs) in cytokine genes, were analysed. However, almost all of these were analysed only once and were insignificant.

Table 5

Results for outcome functional ability*

Quality of life

Only four studies assessing QoL were retrieved (table 2). Different outcome measurements were used, namely, the SF-36 mental component summary scale, the Juvenile Arthritis Quality of Life Questionnaire (JAQQ)59 and the Quality of Life Scales (QOLS).60 ,61 Follow-up ranged from 6 months to 17 years. Two studies were prospective cohort studies.36 ,50

Table 6 shows the results. A comparison is difficult to make due to the paucity of studies and the difference in outcome measurements. Most of the predictors were studied only once. Generally, only disease activity parameters and potentially a longer doctor's delay showed some prospects as predictors of QoL.

Table 6

Results for outcome quality of life*


The aim of our study was to identify early predictors of disease activity, joint damage, functional ability and QoL in JIA. The most consistent predictor was polyarticular onset of JIA, which was identified in multiple studies for three of the outcomes assessed. This reflects the known worse prognosis for the polyarticular categories of JIA.1 ,4 ,5 However, as was already noted by Adib et al4 in their review in 2005, there is considerable variance in the prognosis also within categories and it is hence useful to identify other factors that can predict that variance.

Many of the assessed variables did not show that ability. For example, demographic and laboratory data showed too much variability to be valuable predictors, with the notable exception of a diagnostic delay, predicting continuation of disease activity. Other valuable results were shown by disease activity parameters, such as the number of active joints at baseline, the physician's global assessment, the parent or patient global assessment, the CHAQ score and especially symmetric joint involvement. However, most of these parameters were assessed in only a few studies, leaving the question of reproducibility open. Moreover, the issue of confounding arises, since the number of affected joints defines the distinction between oligoarticular and polyarticular onset JIA, and other disease activity parameters may correlate with this predictor. To what extent disease activity parameters show individual predictive capabilities should be evaluated in multivariate analysis. Lastly, it remains to be elucidated whether the involvement of specific joints has any relationship with the outcome. Here too, confounding might obscure the relationships, since, for example, the finger joints are more frequently affected in the polyarticular categories.1 ,6

Overall, demographic, clinical and laboratory values were insufficient to predict the individual prognosis timely. It is noteworthy that hardly any other potential predictors were taken into account, such as cytokine levels, cell characteristics, results of imaging obtained early in the disease course or genetic evaluations, such as HLA and SNPs in genes with a known function in the immune system. Future studies should address this issue.

There were some concerns about the validity of the studies. Many of the identified studies were retrospective. Since children with a worse disease course tend to visit their physician more often and for a longer period, these retrospective studies are prone to a selection bias. Moreover, many of the studies did not perform a multivariate analysis in order to elucidate individual predictors and try to rule out confounding. Furthermore, almost none of the studies described the process of blinding, a lack of which could lead to information bias, especially in retrospective studies, where the outcome is known at the time of predictor determination. Finally, the selected studies were performed in different periods and made use of different treatment regimens, which potentially confounded our results.

In 2011, the American College of Rheumatology (ACR) issued their recommendations for the treatment of JIA, which included some features of poor prognosis.12 It was difficult to compare these features with our results since the paper did not mention their outcome criteria and these features were not all determined in the first 6 months after diagnosis, as in our review. Furthermore, they included predictors that associated only in univariate analysis, but not in multivariate analysis, that were only based on the authors' clinical experience or that were shown to correlate with the outcome instead of predicting the prognosis. Consequently, the evidence for some of these prognostic features was wanting.

In 2005, Adib et al noted the variability among studies and recommended the development of standardised outcome criteria. Since then, the Wallace criteria have been developed.23–26 In short, each visit is scored as CID if there are no joints with active arthritis, no systemic features, no uveitis, erythrocyte sedimentation rate and/or C-reactive protein are normal, physician's global assessment is the best possible on the scale used and duration of morning stiffness is less than 15 min. A child is regarded to be in clinical remission on medication (CRM) when these criteria are met during 6 consecutive months and she is regarded to be in clinical remission off medications (CR) when these criteria are met during 12 consecutive months while using no antiarthritis or antiuveitis medication.23–26 In our results, Wallace criteria were used in different ways. For example, outcome was defined as the percentage of patients in CID, CRM or CR at the last visit, or as the cumulative amount of time patients spent in CID, CRM or CR during follow-up. To avoid variability, the outcome definition should be standardized. Since JIA has a fluctuating disease course,24 ,40 ,62 the longitudinal aspect of disease activity should be taken into account. Hence, the best outcome measurement seems to be the cumulative amount of time in CID, CRM or CR. A core set of data should routinely be collected in daily clinical practice allowing the evaluation of JIA according to these standardised outcomes such as Wallace criteria or the juvenile arthritis disease activity score (JADAS),63 thus facilitating clinical studies.

Regarding joint damage, none of the identified studies used the JADI, which is a validated assessment of both articular and extra-articular damage.32 Instead, they used the visual assessment of radiographic images for joint space narrowing, erosions and ankylosis, which is regarded as the gold standard.64 However, its reliability, validity and reproducibility are limited due to the lack of validated scoring systems in paediatric rheumatology.32 ,64 Those that exist focus on the knee and wrist joints, whereas damage can occur in any joint.32 ,58 ,65–67 Consequently, such validated scoring systems should be developed for other joints as well and longitudinal outcome studies should be performed with either the JADI or these scoring systems.

The CHAQ is a standardised and validated questionnaire for functional ability.33 It measures the abilities and disabilities in various day-to-day activities, such as dressing, grooming, eating and grip, on a 0–3 point scale. It is still a widely used questionnaire, although there are concerns about its ceiling effect and sensitivity to change.68–71 Since most of the analysed studies used the CHAQ, comparability among them was good.

Finally, for QoL, many validated questionnaires exist, such as the SF-36, JAQQ and QOLS. Because all of these were used in the studies we analysed, comparability was very limited. If QoL is to be an important outcome measurement, one validated measurement should be settled on.

All selected studies focused on finding parameters, which predicted the outcome significantly (p<0.05). Alternatively, one could construct a clinically usable prediction rule, which yields an individual score indicating the probability of a certain outcome. In such an approach, statistical significance is not important, but the focus is on the predictive value instead.72 So far, no such prediction rules for the outcome in JIA have been developed. This hiatus should be addressed in future studies.

We chose to make a strict selection of articles to be analysed in order to focus on early predictors, homogenise the results and to be able to draw a consistent conclusion. The drawback of this approach is that we may have missed some predictors by excluding articles that used other outcome definitions. We feel our approach is justified because if such studies would have been added to the analysis, comparability would be very limited. In order to be not too strict, we did include some articles with outcome measurements very close to the Wallace criteria, some of which were, indeed, also used by Wallace et al25 to test their criteria against.

In conclusion, demographic, clinical and laboratory values are insufficient as early predictors for long-term outcome. Studies that evaluate other early predictors, such as genetics or immunological parameters, hardly exist. Since standardised criteria for disease remission and functional ability have been developed, future efforts should be directed at performing prospective studies, using these validated outcomes, assessing not only clinical but also immunological, genetic and imaging variables within 6 months of the disease course. These validated outcomes should therefore routinely be registered in daily clinical practice. Finally, validated and standardised outcome measures should be developed for joint damage and QoL if these are to be important outcome measures in longitudinal studies.


View Abstract

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • Contributors Both authors were involved in drafting the article and revising it critically for important intellectual content, and both authors approved the final version to be submitted for publication. Study conception and design: EHPVD, NMW. Literature search: EHPVD. Analysis and interpretation of results: EHPVD, NMW.

  • Funding This project has received funding from the 7th Framework programme of the EU, SP3-People, support for training and career development for researchers (Marie Curie), Network for Initial Training (ITN), FP7-PEOPLE-2011-ITN, under the Marie Skłodowska-Curie grant agreement No 289903 [grant to EHPvD] and SHARE project, EAHC grant number 2011 1202 [grant to NMW]. The sponsors had no role in the study planning, design, management or data analysis.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.