Successful completion of clinical trials requires validated outcome measures. There are several measures used in adult patients with axial spondyloarthritis (axSpA), including the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) and the Ankylosing Spondylitis Disease Activity Score (ASDAS).1,2 However, these have not been studied in pediatric patients. Although there are obvious similarities between pediatric and adult SpA, there are important clinical differences as well, including the long-recognized observation of more frequent peripheral involvement in children as compared to adults.3 Thus, the development of a disease activity index specific to the juvenile SpA population by Weiss and colleagues in 2014, called the Juvenile Spondyloarthritis Disease Activity Index (JSpADA; Table 1), was a welcome addition.4 Although the JSpADA includes features that are also present in the ASDAS and BASDAI, including morning stiffness and enthesitis, it is unique in that it also incorporates active peripheral arthritis as well as uveitis. The JSpADA has been accepted by the pediatric rheumatology community, as evidenced by usage in several subsequent studies.5-7
A limitation to any data collection instrument is that it is only as good as the completeness of the data that are collected. For example, a recent study using the Pediatric Rheumatology International Trials Organisation registry documented that only 52% to 61% of critical data elements were collected in the database.8 The JSpADA requires complete data, as missing data in a given domain would result in a score of zero for that domain and thus would have the same statistical impact as the known absence of disease features in that domain.
The current edition of The Journal of Rheumatology includes an assessment of the JSpADA in which either or both of the most frequently missed items, measurement of back mobility and inflammatory markers (ie, C-reactive protein [CRP], erythrocyte sedimentation rate [ESR]), are excluded.9 To do this, Srinivasalu and colleagues reviewed 1961 visits of patients with SpA captured in the Childhood Arthritis and Rheumatology Research Alliance (CARRA) Registry, a multicenter observational pediatric rheumatology disease registry inclusive of, although not limited to, all of the juvenile idiopathic arthritis categories.10 They collected data on patients with SpA with complete data for the entire JSpADA, as well as patients missing only the modified Schober (mSchober) test result (JSpADA7 – no Schober), patients missing only an inflammatory marker (JSpADA7 – no CRP/ESR), and patients lacking both a mSchober and an inflammatory marker (JSpADA6). Convergent validity of the abbreviated versions of the JSpADA was assessed by comparing their scores with that of additional outcome measures as well as by assessing their respective sensitivity to change among subjects with 2 visits separated by at least 6 months.
Their results showed that the results of the modified versions of the JSpADA were similar to that of the original. Specifically, all 4 versions of the JSpADA (original and 3 proposed modified versions) demonstrated similar correlations with the Childhood Juvenile Arthritis Disease Activity Score in 10 joints (cJADAS10), Childhood Health Assessment Questionnaire (CHAQ), and physician global assessment of disease activity (PGA; Table 2), although CHAQ scores did not generally correlate well with any versions of the JSpADA. Additionally, all 4 versions discriminated between active and inactive disease as defined by a PGA of zero vs higher than 0 and demonstrated similar sensitivity to change as compared to the gold standard assessment of improved/unchanged/worsened PGA. Importantly, the fold increases in numbers of registry visits with complete data for each version of the modified JSpADA as compared to the original JSpADA were 2.2, 3.8, and 7.9 for the JSpADA7 – no CRP/ESR, the JSpADA7 – no Schober, and the JSpADA6, respectively.
Pearson correlation coefficient between original and proposed variants of the JSpADA and existing outcome measures.
The study does have its limitations. The number of visits with complete data was a very small percentage of the overall number of visits (3.2%); even the JSpADA6 was available in only 26% of the visits. Clearly, with such small numbers, there is a potential for type II error, in which larger sample sizes would have identified meaningful differences not detected in the smaller set. Additionally, the subjects with complete data may not be representative of the full population. Specifically, the authors did show that the subjects with complete data had higher indicators of active disease (clinically active sacroiliitis, active joint count, cJADAS10, PGA) as compared to those with incomplete data, indicating bias in collection or reporting of the data. Further, although demonstrating convergent validity might bolster the case for use of the modified JSpADA as an outcome measure in research, it does not address the potential performance of the JSpADA in predicting response to therapy. Such use of the JSpADA would require a different analytic approach not yet reported by Weiss and colleagues.4
Additionally, it would have been of interest if the authors had assessed whether the mSchober test and inflammatory markers—as individual variables and not as part of JSpADA—correlated with the CHAQ, PGA, and cJADAS10. A lack of association would have provided additional evidence to support their exclusion from the modified JSpADA.
How important are the missing items?
Modified Schober. The mSchober test is a measure of spinal flexion that is reflective of radiographic progression due to axSpA. This is a widely used measure and typically worsens over time in patients with AS.11 However, the mSchober does not predict radiographic progression from nonradiographic axSpA to ankylosing spondylitis (AS) meeting modified New York criteria,12 thus limiting its applicability to children, most of whom do not meet modified AS criteria at baseline.
Studies in children with SpA do support its exclusion as an outcome measure. Although 1 study showed that an abnormal mSchober test correlated with multiple variables associated with active disease (including ESR and active joint count) as well as damage (including Juvenile Arthritis Damage Index),13 a study published by Weiss et al showed that the mSchober test does not correlate with magnetic resonance imaging (MRI) findings, as an abnormal mSchober test was present in 1/8 of patients with sacroiliitis vs 0/32 of patients without.14 This study was admittedly underpowered, so it should not be the final word on the matter. However, an additional investigation reported the sensitivity of the mSchober test for detecting abnormal MRIs was only 35% to 45% and specificity was only 65% to 73%.15 Most recently, Akdeniz et al showed a statistically insignificant difference in frequency of an abnormal mSchober test (15% vs 8.4%, P = 0.24) in a cohort of 124 patients with juvenile SpA with vs without sacroiliitis.16 These studies argue that the mSchober may not be crucial.
Additionally, to the extent that the JSpADA will be used in clinical trials to assess response to treatment, the elements should show responsivity to change. Here, the mSchober test has not performed well, with Puhakka et al showing that among 34 patients with inflammatory type back pain—of whom 32 had MRI findings indicative of sacroiliitis—who underwent clinical assessment and MRI at baseline and 1 year thereafter, there was no correlation between changes in the mSchober test result and changes in active or chronic MRI findings.17 Likewise, an early randomized trial of etanercept in patients with AS revealed that the mSchober test did not differentiate the placebo and treatment groups.18 Recent randomized trials in patients with nonradiographic axSpA did not include the Schober test as an isolated outcome measure, although it is incorporated in the metrology index reported in one of the studies.19-21
Inflammatory markers. Inflammatory markers are valuable, albeit nonspecific, indicators of disease activity, as evidenced by their incorporation into the aforementioned ASDAS,2 among other scores. In adult patients with axSpA, an elevated CRP is a well-recognized risk factor for sacroiliitis as well as for radiographic progression among those with sacroiliitis.22,23 This has not been studied as thoroughly in children with SpA. However, a recent study demonstrated that baseline elevated CRP, although not ESR, predicted hip involvement in children with SpA, a major cause of functional loss24; a small study also demonstrated that among 40 children with juvenile SpA, of whom 8 had sacroiliitis, only a positive HLA-B27 and an elevated CRP predicted the presence of sacroiliitis.14 Therefore, CRP may be an important predictor in children. Of note, there is likely to be variability among CARRA sites with respect to use of high-sensitivity vs routine CRP, as these may not perform identically, based on data from adults.25
The issue with inflammatory markers as an outcome in retrospective studies is bias in the nature of the missing data. Practitioners may be more likely to measure ESR and CRP in patients suspected to have active arthritis compared to those suspected to be in remission. This may be particularly true in children with axSpA, for whom there is often no indication to treat with traditional disease-modifying antirheumatic drugs and who, therefore, may have less need for therapeutic drug monitoring. Thus, although this information may be valuable, it is clearly better to exclude it on all subjects than to incorporate it only in those with complete data. That is, it would be ill-advised to use both the original and the modified JSpADA in the same dataset.
In summary, the current study9 shows that modifying the JSpADA with exclusion of the mSchober test and inflammatory marker requirement did not impair its performance in a population of children with SpA, while increasing the number of children available for evaluation. Although CRP remains an important disease activity biomarker, the value of the mSchober test in children is uncertain, especially in the setting of a clinical trial.
Footnotes
MLS has consulted for Novartis. EH and LC declare no conflicts of interest relevant to this article.
See Modified Juvenile SpA Disease Activity Index, page 532
- Copyright © 2023 by the Journal of Rheumatology
