Anti‐platelet antibody immunoassays in childhood immune thrombocytopenia: a systematic review

Background In adult immune thrombocytopenia (ITP), an acquired autoimmune bleeding disorder, anti‐platelet autoantibody testing may be useful as a rule‐in test. Childhood ITP has different disease characteristics, and the diagnostic and prognostic value of anti‐platelet antibody testing remains uncertain. Objective To systematically review the diagnostic accuracy of anti‐platelet autoantibody testing in childhood ITP. Methods PubMed and EMBASE were searched for studies evaluating immunoassays in childhood ITP. Study quality was assessed (QUADAS2), and evidence was synthesized descriptively. Results In total, 40 studies (1606 patients) were identified. Nine studies reported sufficient data to determine diagnostic accuracy measures. Anti‐platelet IgG antibody testing showed a moderate sensitivity (0·36–0·80 platelet‐associated IgG [direct test]; 0·19–0·39 circulating IgG [indirect test]). In studies that reported control data, including patients with non‐immune thrombocytopenia, specificity was very good (0·80–1·00). Glycoprotein‐specific immunoassays showed comparable sensitivity (three studies) and predominantly identified IgG anti‐GP IIb/IIIa antibodies, with few IgG anti‐GP Ib/IX antibodies. Anti‐platelet IgM antibodies were identified in a substantial proportion of children (sensitivity 0·62–0·64 for direct and indirect tests). Conclusion The diagnostic evaluation of IgG and IgM anti‐platelet antibodies may be useful as a rule‐in test for ITP. In children with insufficient platelets for a direct test, indirect tests may be performed instead. A negative test does not rule out the diagnosis of ITP. Future studies should evaluate the value of anti‐platelet antibody tests in thrombocytopenic children with suspected ITP.


Introduction
Childhood immune thrombocytopenia (ITP) is an autoimmune bleeding disorder [1]. Given a lack of laboratory tests to rule-in or rule-out the diagnosis of ITP, the disease is diagnosed clinically by exclusion of alternative causes of thrombocytopenia [2,3]. This may lead to misdiagnoses, as much as 12% in adult ITP [4], for example for patients with hereditary platelet and immune disorders [5,6]. Misdiagnosed patients are not only exposed to unsuitable ITP-specific treatments, but also the treatmentassociated side-effects and costs. Misdiagnoses also lead to delayed diagnosis and adequate management of the actual underlying disorder. On the other hand, with new treatment options, for example the early administration of TPO agonists that are currently being investigated [7], clinicians may wish to ascertain an ITP diagnosis when considering such treatments. Thus, there is an unmet clinical need for laboratory tests to support the diagnosis of ITP [2].
A pathophysiological hallmark of ITP are anti-platelet autoantibodies specific to platelet glycoproteins [8,9]. Soon after the discovery of these platelet autoantibodies, assays to measure anti-platelet antibodies were developed [10,11] and several studies identified anti-platelet antibodies in childhood ITP [11][12][13]. Nonetheless, anti-platelet antibody testing currently remains of unclear clinical benefit [2,3], for several reasons. The results of early studies were dismissed by the observation of non-specific adsorption of plasma protein to platelets and potential false-positive results in healthy controls and patients with non-immune thrombocytopenia [14][15][16][17][18][19][20]. Later, to overcome these challenges, antigen-specific assays were developed that are now considered standard for assessment of anti-platelet antibodies in ITP [16,21,22]. A recent systematic review suggested that such antigenspecific anti-platelet antibody testing may be a useful rule-in test for adult ITP [23]. However, it is unknown whether this finding is transferable for childhood ITP. In contrast to adult ITP, childhood ITP has a large proportion of patients with preceding infections and mostly self-limiting disease courses, and the clinical context of suspected ITP in children is clearly different. A further reason for uncertainty is the heterogeneous background of ITP, which represents a mix of cases with distinct underlying pathophysiology. For instance, ITP may also be caused by T-cell-mediated immunity directed towards platelet autoantigens and megakaryocytes [24]. It is unknown how this influences the diagnostic and prognostic value of anti-platelet antibody testing. In sum, the role of anti-platelet antibody testing in the diagnosis of childhood ITP remains uncertain.
In the present study, we aimed to synthesize the available evidence to determine the potential diagnostic accuracy of anti-platelet antibody testing in childhood ITP. In addition, we describe some data on the prognostic significance of anti-platelet antibody testing.

Study identification and quality assessment
Reporting standards of the PRISMA guidelines were followed. PubMed and EMBASE were searched from inception until 4 April 2019 to determine the diagnostic accuracy of various immunoassays measuring platelet autoantibodies in childhood ITP (Table S1). The search string contained three elements: domain (children, 3 months to 18 years), disease (ITP) and diagnostic tests for comparison (platelet autoantibody tests). The sensitivity of the search strategy was assessed by the inclusion of pre-determined index publications. Screening of abstracts, full-text assessment and data extraction was performed independently by two investigators (D.S. and A.L.; Fig. 1). Studies were included if they were published in English language and evaluated a platelet autoantibody assay in children with immune thrombocytopenia and fully described the anti-platelet antibody immunoassay in the manuscript or a previous publication. Studies were excluded if they did not separate adult and childhood ITP data, assessed neonates or used anti-platelet antibodies as criteria for the diagnosis of ITP. Case series (less than 10 patients) were excluded. Data collection was validated independently by a third investigator (L.P.). The methodological quality of included articles was assessed using a standardized protocol for quality assessment of diagnostic accuracy studies (QUADAS2) [25].
Methodological details are described in the Supplementary Methods.

Synthesis of results
Given substantive heterogeneity in study populations and methodological quality, study results were not pooled for meta-analysis of diagnostic accuracy.

Study identification and characteristics
After screening and full-text assessment, 40 studies (1606 patients) were identified for this systematic review ( Fig. 1; Table S2). The studies assessed either platelet-associated antibodies on autologous patient platelets (direct test) or circulating anti-platelet antibodies (indirect test).
inclusion of patients (patient selection), (2) the evaluated anti-platelet antibody test (index test), (3) the criteria for diagnosis of ITP (reference standard) and (4) included disease stages and treatments (flow and timing). Overall, this assessment indicated that the risk of bias was unclear or high for a large proportion of studies (Fig. S1). Three factors were primarily responsible for this. First, although studies usually assessed the assay in at least some controls (either in the same study or historically), a substantial analysis was often lacking for control subjects, or this was incompletely reported. Second, studies did not report sensitivity and specificity or dichotomized test results from which these could be derived, that is true-positive, true-negative, false-negative and false-positive results. Third, many studies included heterogeneous disease stages with varying time from diagnosis, as well as mixing patients with prior or current treatments, both of which are known to influence anti-platelet antibody testing [11,13,26]. Nine of the 40 studies reported sufficient data to determine diagnostic accuracy of anti-platelet antibody testing; all nine were of unclear or high risk of bias ( Fig. 2; top panel) [27][28][29][30][31][32][33][34][35]. Eight of the 40 included studies were judged to be of low risk of bias regarding the inclusion of ITP patients, but none of them explicitly reported control data ( Fig. 2; lower panel) [36][37][38][39][40][41][42][43]. We subsequently focused our analysis on the nine studies which allowed assessment of diagnostic accuracy data and describe data of the eight studies with low risk of bias and the remaining studies, as applicable.

Diagnostic accuracy of anti-platelet antibody testing in childhood ITP
Nine studies allowed an assessment of the diagnostic accuracy of anti-platelet antibody testing ( Table 1). The studies had a median sample size for cases and controls of 21 and 20, respectively. When we assessed these studies together, the primary finding was that most of the nine studies showed moderate sensitivity of immunoassays, with very good to excellent specificity ( Table 2). These results were consistent across small study cohorts (N < 20) and larger studies, as well as amongst studies that included vs. did not include nonimmune thrombocytopenic controls. Indirect tests potentially have a lower sensitivity than direct tests, but a direct comparison was lacking ( Table 2). Specificity was high across the studies (Table 2). Importantly, although false-positive results may be missed by small studies, specificity results were the same in studies which included small or large control groups. Next, we assessed some of the methodological concerns of these studies with unclear or high risk of bias. Treatment with steroids (two studies) may have lowered the sensitivity of the tests; one small study included IVIg-treated patients that may lead to false-positive test results [33], and another study measured anti-platelet antibody levels in IVIg-treated patients, but by an antigen-specific immunobead assay [35]. The inclusion of patients later in the disease course, as opposed to assessment at diagnosis, may skew results, but the studies at low risk of bias which included patients within a week of diagnosis showed a similar sensitivity (discussed below). We calculated positive likelihood ratios that indicate how much more likely it is that a positive test result occurs in a patient, compared to controls, where ratios of ≥10 are considered clinically useful to rule-in disease [44]. Amongst four of the nine studies (92 patients) where positive likelihood ratios (LR) could be calculated, the positive LR was at least 8Á8 in three studies, indicating that the assays could be used as rule-in tests for childhood ITP. In the other studies, only patients showed positive test results. On the other hand, with the given sensitivity and negative LR ranging between 0Á14 and 0Á73 (170 patients), the assays were insufficient as a rule-out criterion for ITP, for which a negative LR of ≤0Á1 would be considered useful. In sum, there was weak-to-moderate evidence that anti-platelet antibody testing could potentially be useful as a rule-in test of ITP.

Supporting evidence
We next compared these data to results of the other studies that were identified by our systematic review. First, amongst the eight studies with low risk of bias (Table 1), six studies assessed anti-platelet IgG in patients with newly diagnosed ITP within a week of diagnosis, and two studies included patients with chronic ITP (Table 3; results of all 40 identified studies are shown in Table S5). We observed a similar moderate sensitivity as for the studies discussed above. Moreover, sensitivity was the same for newly diagnosed, as well as chronic ITP, and for glycoprotein-specific assays [39,40,42,43]. The results were unchanged when only the largest studies were considered [37,42,43].
For the remainder of the studies, a further eight reported quantitative results (i.e. mean or median and distribution) of platelet antibody immunoassays in cases and controls (Table S3). These eight studies unanimously indicated distinct anti-platelet antibody levels in patients with ITP compared to controls (Table S3). Thus, even though it was not possible to calculate the diagnostic accuracy, these studies confirmed that anti-platelet antibody levels are elevated in ITP patients compared to controls.
Together, these findings provide support that IgG antiplatelet antibodies can be detected in children with ITP and that results could be relatively specific for ITP.

Detection of anti-platelet IgM and IgA antibodies
Three of the eight studies with low risk of bias assessed immunoassays for evaluation of anti-platelet IgM antibodies, two in newly diagnosed ITP and one in chronic ITP (Table 3; full results, Table S6). The sensitivity for detection of IgM autoantibodies was 0Á62-0Á69, indicating that IgM class autoantibodies could potentially be as prevalent as anti-platelet IgG antibodies [37,39,41]. Due to a lack of control data, the specificity could not be determined. Nonetheless, these findings in studies with low risk of bias suggest that IgM anti-platelet antibodies could be of importance in childhood ITP.
© 2020 The Authors.  Only studies were included that presented dichotomized results or quantitative data in a graph. CI, confidence interval; LR, likelihood ratio.    [37]. Taken together, some studies indicate a role of anti-platelet antibody testing to determine prognosis, but in the presence of conflicting evidence further data are required.

Discussion
In this study, we systematically reviewed 40 studies and summarized the current knowledge regarding the diagnostic accuracy of anti-platelet antibody testing in childhood ITP. Although many studies suffered from insufficient reporting and a lack of controls, we carefully analysed and weighted the available data. The main finding of this review is the overall good diagnostic accuracy of anti-platelet antibody testing, based on multiple casecontrol and prospective cohort studies, supported by evidence from low risk of bias studies. Sensitivity was moderate in multiple studies, and this could be due to low levels of circulating antibodies during active clearance of platelet-antibody immune complexes, low avidity of the antibodies or insensitivity of the used tests or heterogeneity in the involved pathomechanisms, such as antibodymediated or T-cell-mediated anti-platelet clearance [24]. Even though evidence has been published that indicated potential false-positive results in patients with thrombocytopenia by other causes than ITP, particularly with PA-Ig assays [14][15][16], studies included in this review showed good specificity in healthy children and patients with non-immune thrombocytopenia (multiple case-control studies), also in the largest included studies [28,31,32]. It might be argued that studies that did not report control data showed a high rate of false-positive results. Regardless, studies which used the more recently developed antigen-specific assays [16] showed similar sensitivity as whole-platelet testing (three studies; two with low risk of bias), which provides a compelling argument against skewing to false-positive results. Thus, when balancing the available data on diagnostic accuracy, anti-platelet antibody assays may potentially be used as a rule-in test, as also indicated by a high positive likelihood ratio. This is in line with results of a recent systematic review of anti-platelet antibody testing in adult ITP [23]. On the other hand, given the limited sensitivity, the data show that anti-platelet antibody assays cannot be used as a rule-out test to exclude a diagnosis of ITP, which is also in agreement with the conclusions of the systematic review in adult ITP [23]. A secondary finding of our analysis in childhood ITP was that IgM anti-platelet antibodies may be as prevalent as IgG anti-platelet antibodies (three studies), which should be assessed in future studies. Moreover, tests for circulating anti-platelet antibodies are interesting alternatives when children have insufficient platelets available for a direct test. We found that although indirect tests might have a reduced sensitivity, they could still be of diagnostic value. Several studies described lower anti-platelet antibody levels when patients were assessed later in their disease course or after treatment [11,13,28,32,48], indicating that delayed testing may affect the diagnostic accuracy. Finally, with regard to relevant platelet antigens, in two studies, anti-GPIb/IX antibodies were not detected at all, and GP IIb/IIIa was the predominant antiplatelet antibody [35,49]. Conversely, antibodies directed against GP IIb/IIIa and GP Ib/IX were found at about the same rate by one study [43]. Thus, the role of antibodies against specific platelet glycoproteins for the diagnosis and prognosis of childhood ITP remains elusive.
The primary limitation of this review was the reference test of our study (gold standard), which was a clinical diagnosis of ITP. As surrogates of patient heterogeneity between studies, we extracted data on age, sex and preceding infection, which are associated with ITP prognosis [50]. Moreover, the reported diagnostic accuracy may be optimistic estimates, since none of the studies investigated anti-platelet antibody assays in children with suspected ITP. However, in the context of a rare disease and a paediatric study population, such studies are notoriously difficult to conduct. Two further shortcomings regarding the methodological quality of included studies limited our review: (1) insufficient reporting of control data to determine specificity for the majority of studies and (2) the inclusion of heterogeneous patient populations at various disease stages and divergent treatments. Nonetheless, we recognized that the identified studies represent the best evidence to date in a disease setting that is challenging to investigate, and carefully assessed and compared the available data. We suggest that to improve the quality of data in the field, future studies published in childhood ITP should be explicitly required to disclose clinical characteristics, inclusion criteria, time from diagnosis and previous and current treatments. Finally, during routine diagnostic testing, not all childhood patients have sufficient platelet numbers to perform direct tests; thus, results could be biased to older children, for whom more material is available. None of the studies disclosed the number of patients that could not be evaluated.
For future directions, direct tests should be developed that allow the use of low number of platelets. The assessment of anti-platelet antibodies may be of prognostic significance, potentially indicating a subgroup amongst heterogenous ITP patients, and this should be investigated further. A key current area of uncertainty is the evaluation of anti-platelet antibody assays in children with suspected ITP.
In conclusion, this systematic review indicates that anti-platelet antibody testing could potentially be used as a rule-in test for childhood ITP, although anti-platelet antibody testing cannot be used to exclude a diagnosis of ITP. 48

Supporting Information
Additional Supporting Information may be found in the online version of this article: