|Home | About | Journals | Submit | Contact Us | Français|
Paediatricians often order laboratory and radiological tests to identify children with potential rheumatological disease prior to subspeciality referral. However, the pattern of testing suggests inadequate understanding of their diagnostic utility and limitations. Herein we will address some of the most common rheumatological diagnoses encountered in the subspeciality clinic – juvenile idiopathic arthritis (JIA), juvenile spondyloarthritis (JSpA) and systemic lupus erythematosus (SLE), and related connective tissue diseases – and the tests most frequently ordered to diagnose them: anti-nuclear antibodies (ANA), rheumatoid factor (RF), human leukocyte antigen (HLA)-B27 and radiological tests. This article will highlight the sensitivity, specificity and positive predictive value of the tests. In general, none of these tests were appropriate to use as rheumatological ‘screens’, as no individual test was diagnostic. Specific tests should be ordered only when there is a high clinical index of suspicion for a particular disease entity. Greater understanding of a test’s diagnostic utility should decrease unnecessary testing, anxiety and expense and aid in interpretation.
Musculoskeletal pain, fatigue and malaise are common complaints encountered in primary paediatrics that prompt evaluation for rheumatologic aetiologies. Unfortunately, the search for ‘answers’ often leads to the indiscriminate ordering of tests, in particular antinuclear antibodies (ANA) and rheumatoid factor (RF), without a clear understanding of their diagnostic utility or limitations. False-negative tests can be misleading and delay diagnosis. False-positives, on the other hand, can cause undue anxiety and lead to unnecessary referrals, testing and expense. The most common rheumatological diagnoses seen in the paediatric subspeciality setting are juvenile idiopathic arthritis (JIA), juvenile spondyloarthritis (JSpA), and systemic lupus erythematosus (SLE) and related connective tissue diseases, accounting for 10–20, 5–11 and 2–9% of all diagnoses, respectively.1–4 Therefore, this article will focus on the tests most frequently ordered to evaluate these particular diagnoses: the ANA, RF, human leukocyte antigen (HLA)-B27 and radiological tests.
In order for a test to be diagnostically useful, it should be highly sensitive and specific (both ideally >90%) and have a high positive predictive value (PPV). If a disease is much less frequent in the population than the positive test, the PPV will be low. Consider the ANA, for instance; a test widely used to screen for rheumatological conditions. The prevalence of a positive ANA among healthy children has been reported to be as high as 13–18%, depending on geographical location and cut-off titre.5–7 However, rheumatological diseases are relatively rare in children: annual incidence rates for JIA and SLE have been reported at 0.8–23/100,000 and 0.3–0.4/100,000, respectively, with prevalences of 7–400/100,000 and 6–37/100,000.1,2,8–11 Moreover, ANA are not specific to rheumatological conditions, but may also be positive in the setting of malignancy, infection, and drug reaction (e.g. minocycline-induced lupus).7,12–14 Thus, even though the ANA is a highly sensitive test for SLE (98%), its PPV is low (0.10).15
Several studies and reviews have highlighted the futility of using a positive ANA as a rheumatological screen. In one study of 245 children referred to a paediatric service for a positive ANA (≥1:40), only 55% had a rheumatological diagnosis; thus, children with a positive ANA were not more likely to have a rheumatological diagnosis.16 In another study of all ANA tests performed at the British Columbia Children’s Hospital (1,369), the ANA was positive (>1:20) in 67% of those with rheumatic disease and 64% without a rheumatic diagnosis, leading the authors to conclude that “a positive test has little or no predictive value”. Even in children with an unequivocally high titre of 1:320, only 31% had a rheumatological disease (see Table 1).15 Among children referred to a rheumatology service for musculoskeletal pain, children referred for a positive ANA or RF were “no more likely to have a chronic inflammatory disease” than those who were not referred for these tests.17 Some evidence suggests that a positive ANA in a child without rheumatological disease is not predictive of future disease: in 24 children referred for non-rheumatologic musculoskeletal pain with a positive ANA but normal complement levels and negative extractable nuclear antigen (ENA) and double-stranded DNA (dsDNA), 21 remained ANA-positive over a mean follow-up of 61 months. Titres were variable over time, with 14 (58%) having titres of 1:320 or greater. Although this study was small, no children developed overt autoimmune disease.18 The one clinical setting where a positive ANA may be useful for identifying children at risk is in idiopathic thrombocytopenia purpura (ITP). In a study of 87 children with ITP, 36% of the 25 children with a positive ANA (>1:40) developed further autoimmune disease compared with 0% of the ANA-negative children. Of note, 75% (six of eight) of those who progressed tested positively for more specific autoantibodies.19
Therefore, a positive ANA does not distinguish children with rheumatological disease and should not be used as a general screen for rheumatical conditions.
SLE is a protean autoimmune disease that may affect any organ system. To aid in the diagnosis of SLE, a core set of 11 criteria was developed in 1982. Establishing four of 11 criteria provides a sensitivity and specificity of 96%, and thus having four positive criteria has become accepted as diagnostic.20 In 1997, the immunological component was revised to include anti-dsDNA, anti-Sm or anti-phospholipid antibodies (see Table 2 for modified criteria).21 The effect of this modification on sensitivity or specificity is unclear.22
There are some data to suggest that very high ANA titres are more strongly associated with SLE: in 110 children referred for a positive ANA, the mean titre in SLE patients was 1:1,080 compared with 1:240 for JIA and 1:160 in controls. All 10 patients with a titre ≥1:1,080 had SLE. However, a titre of ≤1:360 had a negative predictive value (NPV) for SLE of 0.84.23 In another study a higher titre (≥1:640) correlated with more specific ANAs and autoimmune disease, in particular SLE.16
However, high titres are not diagnostic in and of themselves: in one study, even at a titre of 1:1,280, 20% had non-rheumatic diagnoses and only 50% had SLE or related conditions (see Table 1).15 Rather, a high titre should prompt further laboratory work-up aimed at fulfilling the 11 criteria, including a complete blood count (CBC), urinalysis, ENA and dsDNA. An erythrocyte sedimentation rate (ESR) may aid in detecting the presence of significant systemic inflammation. The ENA and dsDNA are less sensitive than an ANA but provide specificity for autoimmune disease: in one study, 100% of children with a positive ANA profile had an autoimmune disorder. Of the 22 SLE patients, 73% had a positive dsDNA, and 68% had anti-Sm and anti-RNP.16 A discussion of anti-phospholipid antibodies is outside the scope of this review, but testing should be pursued with caution given the lack of specificity of anti-phospholipid antibodies for SLE.24
The immunofluorescence pattern of nuclear staining is not specific: one group reported a lack of association between pattern and ENA and dsDNA except for a higher frequency of dsDNA in samples with the homogeneous pattern. Anti-DNA was also detected in samples with speckled and nucleolar pattern.25 For a high-titre ANA (1:640), a mitotic or mitotic and homogeneous staining pattern only marginally increased the PPV for SLE (72–77 versus 69%).26
Therefore, high-titre ANA or a strong clinical index of suspicion for SLE should prompt further laboratory work-up targeted at evaluating the 11 diagnostic criteria.
JIA includes a heterogeneous group of arthritic conditions with distinctive laboratory and clinical appearances occurring in children <16 years of age, lasting longer than six weeks, and without any other known cause.27 JIA is not equivalent to adult rheumatoid arthritis, except in a small subgroup of patients.28,29 The diagnosis of JIA is fundamentally clinical: there is no specific laboratory test for JIA. In considering a child with a swollen joint, the differential diagnosis includes non-JIA autoimmune conditions, autoinflammatory diseases (e.g. neonatal-onset multisystem inflammatory disease [NOMID]), and infectious, post-infectious, orthopaedic/mechanical, haematological/oncological, serum sickness/hypersensitivity, genetic and psychiatric/mechanical causes. Thus, prior to any testing, the history and physical are critical for narrowing down the possible aetiologies. Since JIA is a diagnosis of exclusion, laboratory and radiologic tests should be used to aid in ruling out other diagnoses. For instance, Lyme titres should be ordered in endemic areas prior to rheumatologic referral. Streptococcal titres (anti-DNaseB, ASO) should be sent if there is a history of preceding illness. A CBC (low white blood cells [WBC] plus low to normal platelets) in the face of night-time pain is useful for distinguishing JIA and leukaemia.12
Despite the lack of specific diagnostic tests for JIA, the ‘arthritis panel’ makes a regular appearance in rheumatology referrals. These tests are problematic for the following reasons: although a higher proportion of children with JIA have positive ANAs than the general population (18–77%, depending on JIA subtype), a significant percentage have a negative ANA.30 One group analysing the clinical utility of the ANA test concluded that “Neither the presence nor the titre of ANA served to distinguish children with (JIA) from children with other musculoskeletal conditions”.17
Immunoglobulin M (IgM)-RF is positive in over 60% of adult rheumatoid arthritis patients but only 5–10% of children with JIA.31–33RF may also be positive in the setting of general immune activation (e.g. infection), subacute bacterial endocarditis, mixed connective tissue disease, SLE and sarcoidosis.11,34,35 In one study of 437 children, RF had a sensitivity of 5%, specificity of 98%, but an estimated PPV in the primary care setting of only 0.7%. Furthermore, “In no case was rheumatoid factor testing helpful in establishing a diagnosis of juvenile rheumatoid arthritis or in ruling it out.”35 Anti-cyclic citrullinated peptide (CCP) antibodies have a comparable sensitivity to RF in adults and higher specificity.31,32 Fewer data are available regarding anti-CCP antibodies in JIA but in general they do not have greater sensitivity than RF, being present in only 4–15% of children with JIA (mostly polyarticular RF-positive children).33,36,37
Therefore, the ANA and RF tests have no utility in establishing a diagnosis of JIA and should not be ordered in the primary care setting with this goal in mind. JIA remains essentially a clinical diagnosis.
One of the earliest characteristic abnormalities on plain radiographs is periarticular osteopenia due to inflammatory hyperaemia. Later abnormalities include joint space narrowing and bony erosions. Longstanding disease can result in joint ankylosis and fusion.38 However, at the time of diagnosis, most radiographs in JIA are normal: even in the group with the highest likelihood of early aggressive and erosive disease (polyarticular JIA), only 61% of radiographs were abnormal, with 44% showing periarticular osteopenia and 28% erosions or joint space narrowing.39 Interpreting radiographs in young patients with JIA is particularly challenging given the immature skeleton and thick cartilage. Ultimately, plain radiographs may be more helpful in excluding other bony causes of joint pain and swelling such as trauma, tumours and skeletal dysplasias.
Compared with plain radiographs, ultrasound and magnetic resonance imaging (MRI) are more sensitive at detecting effusion, synovial thickening and erosions.38,40,41 These alternative modalities have also proven more sensitive than clinical exam.40,42,43 For instance, in a study of 20 children with JIA, 25% of clinically ‘normal’ metacarpophalangeal joints were abnormal by ultrasound.44 However, ultrasound is highly dependent upon the operator and the cooperation of a potentially young patient. MRI is excellent at defining soft tissue anatomy and less dependent upon these variables. To evaluate synovial thickening and inflammation by MRI, gadolinium contrast should always be given.40 MRI and ultrasound are especially useful for evaluating joints that are difficult to examine clinically, such as the shoulder, hip, and axial spine. Another potential application would be the confirmation of effusion in a morbidly obese patient.
Ultimately, these radiologic methods cannot substitute for clinical diagnosis. Joint effusion and synovial hypertrophy are non-specific findings. Radiologic ‘mimics’ of JIA include viral arthritis, reactive arthritis, Lyme disease, septic arthritis, tuberculous arthritis, acute transient synovitis of the hip, sarcoidosis, haemophilic arthropathy, inherited synovial disorders, leukaemic arthritis, pigmented villonodular synovitis and others.45 Granted, some of these disorders may have identifying characteristics on MRI, but MRI seems to be particularly insufficient at differentiating infectious aetiologies. In one series comparing 40 patients with clinically diagnosed JIA versus 40 with other sources of knee pain, the sensitivity and specificity of MRI for diagnosing JIA were 93.5 and 92.5%, respectively, but both cases of septic arthritis were misdiagnosed.46 Moreover, in younger patients, MRI may require sedation or general anaesthesia, providing unnecessary exposures. Finally, given the relative ease of evaluating swelling in peripheral joints it is difficult to justify the considerable expense of an MRI on a regular basis just to confirm synovial inflammation.
Therefore, radiologic tests may be more useful for ruling out other causes of joint pain and swelling than for establishing a diagnosis of JIA.
The spondyloarthritides (SpA) comprise a group of disease entities, including ankylosing spondylitis, enthesitis-associated arthritis, psoriatic arthritis, reactive arthritis, Reiter’s disease and arthritis associated with inflammatory bowel disease (IBD). These conditions are unified by a strong genetic linkage to the major histocompatibility (MHC) class I allele HLA-B27 and arthritic involvement of the axial skeleton. The prevalence of juvenile SpA is estimated at 1–2/100,000.2,9 Because of the wide prevalence of low-back pain, even among children, difficulty in examining the axial skeleton, and insidiously progressive nature of these diseases, the delay between symptom onset and diagnosis is typically eight to 11 years.47 The diagnosis of ankylosing spondylitis, the prototypic SpA, requires radiographic evidence of sacroiliitis, which is a relatively late finding.48 Effective treatment is available (e.g. tumour necrosis factor [TNF] blockade), and evidence suggests greater efficacy early in disease.49 Therefore, early diagnosis has become a critical goal of clinical and basic research in this field. In this regard, a combination of clinical findings, HLA-B27 and MRI-identified sacroiliitis may be diagnostically useful.
Although the onset of ankylosing-spondylitis-related symptoms first occurs typically in adolescents and young adults (before 30 years of age), most of the data available have been gathered from adult studies.50 The prevalence of HLA-B27 among the general population varies tremendously depending on ethnicity and geographical location: for instance, a high proportion of North American Indians and Lapps are HLA-B27-positive (26–50%), but HLA-B27 is virtually absent among African-Americans. The prevalence of HLA-B27 positivity in European populations is reported at 7–10%. Among the spondyloarthritides, HLA-B27 has the highest prevalence in ankylosing spondylitis at 90–95%, with the linkages to other spondyloarthritides ranging from 36 to 100% (see Table 3). In considering ankylosing spondylitis, an estimated disease prevalence of 0.1–1% and a positive HLA-B27 frequency of 10% would yield a specificity of 90–91% and a PPV of 1–10%. This calculation correlates well with the observed incidence of ankylosing spondylitis of 1–6% among HLA-B27-positive subjects.51 Given the low PPV, HLA-B27 should not be used as a general screen for SpA.
Limiting testing to patients with inflammatory back pain may improve the usefulness of the HLA-B27 test. Inflammatory back pain is characterised by onset at <40 years of age, duration greater than three months, insidious onset, morning stiffness, and improvement with exercise. Among adults presenting with chronic low-back pain, the presence of these qualities raises the pre-test probability of axial SpA from 5 to 14%. The presence of two–three other clinical or laboratory features suggestive of axial SpA raises the post-test probability (PPV) to >90%. Among the features examined (including family history, heel pain, uveitis, dactylitis, positive response to non-steroidal anti-inflammatories, acute-phase reactants, etc.), HLA-B27 remained the best single predictor: the combination of inflammatory back pain and HLA-B27 alone raised the probability to 59%. A negative HLA-B27 was valuable for decreasing the probability to <2% (a strong NPV).50,52
Therefore, HLA-B27 should not be used as a general population screen; however, positivity in the presence of inflammatory back pain should prompt further evaluation.
The sacroiliac joints are difficult to assess clinically. MRI can identify sacroiliitis earlier than radiographs and can evaluate for both chronic and active disease.40 However, even the finding of sacroiliitis by MRI has sub-optimal predictive value for the development of radiographically apparent disease: in one report, the PPV was only 60%, with a sensitivity of 85% and specificity of 47%.53 In the presence of inflammatory back pain, the combination of HLA-B27 and MRI findings yielded the highest likelihood ratios for inflammatory axial spondyloarthritis.49 In this regard, the severity of sacroiliitis by MRI may be important for predicting the development of radiographic disease. In one study of patients with recent-onset inflammatory back pain (less than two years), the development of radiographic disease was evaluated at eight years: in HLA-B27+ patients with severe sacroiliitis on MRI, the likelihood ratio (LR) of developing subsequent radiographic change was was 8.0, sensitivity was 62% and specificity was 92%. Including moderate sacroiliitis increased the sensitivity to 77%, but decreased the specificity (62% for all patients and 77% for HLA-B27+). Mild or no sacroiliitis by MRI gave an LR of 0.4. These subjects were 20-fold less likely to develop ankylosing spondylitis regardless of HLA-B27 status.54
Therefore, the combination of inflammatory back pain, HLA-B27 status and degree of sacroiliitis detected by MRI may be useful for identifying patients at high risk of having or developing axial spondyloarthritis (see Figure 1 for proposed diagnostic algorithm).
Some of the most frequently ordered rheumatology tests are not useful for screening general paediatrics patients because of the high prevalence of positive tests in healthy children (e.g. ANA) or the low prevalence in disease (e.g. rheumatoid factor). None of the tests reviewed here are diagnostic in and of themselves but, in the appropriately selected patient, where there is a high clinical index of suspicion, should prompt further work-up and rheumatology referral. A summary of more specific recommendations is:
Judith A. Smith M.D. Ph.D. received her Pediatric Rheumatology training at Cincinnati Children’s Hospital Medical Center. She is currently an Assistant Professor of Pediatrics at the University of Wisconsin School of Medicine and Public Health in Madison. Dr. Smith belongs to the Society for Pediatric Research. Her current research interests include the regulation of inflammatory cytokine production in macrophages and the pathogenesis of ankylosing spondylitis.
Disclosure: The author has no conflicts of interest to declare.