We retrospectively studied 568 patients with thrombocytopenia (platelet count < 100 × 10⁹/L) of various causes who presented to Korea University Guro Hospital from April 2013 to December 2017. We classified the patients into two groups: those diagnosed with ITP (ITP group) and those diagnosed with other causes of thrombocytopenia apart from ITP (non-ITP group). ITP was diagnosed on the basis of clinical manifestations and laboratory results, according to the International Working Group diagnostic criteria [1], but some patients underwent bone marrow examination for differential diagnosis. Causes of thrombocytopenia other than ITP included the following: primary bone marrow diseases like aplastic anemia, acute leukemia, multiple myeloma, and myelodysplastic syndrome, infection, drugs, disseminated intravascular coagulation, and liver diseases, etc. The Korea University Hospital Institutional Review Board approved the protocol of this study (2019GR0072). Informed consent was waived by the board.

We collected 3 mL of peripheral blood from each patient in K2-EDTA anticoagulant bottles (Becton Dickinson, Franklin Lakes, NJ, USA) on the day of thrombocytopenia confirmation. Samples were analyzed using the Sysmex XE-2100 to quantify IPF and routine complete blood count within 8 hours after collection. The Sysmex XE-2100 is a fully-automated hematology analyzer using fluorescent dye and a semiconductor diode laser beam system. Two fluorescent dyes (polymethine and ozazine) penetrate into the cells, staining the RNA in the platelets. The stained cells are passed through a semiconductor diode laser beam and resulting forward scatter light (cell volume) and fluorescence intensity (RNA content) measured. The immature platelet fractions are distinguished by the intensity of their fluorescence because the youngest platelets contain a greater amount of RNA [15]. The IPF is expressed as a proportional value of the total optical platelet count. Several studies have demonstrated the stability and reproducibility of the IPF.

We also estimated other variables that could affect thrombocytopenia by logistic regression analysis, including hemoglobin, platelet count, white blood cell counts, reticulocyte count, protein, albumin, bilirubin, prothrombin, activated partial thromboplastin time, ferritin, lactate dehydrogenase (LDH), blood urea nitrogen, creatinine, and C-reactive protein.

All the data were statistically analyzed using SPSS version 20 (IBM Corp., Armonk, NY, USA). For continuous variables, the Student’s t test and the Mann-Whitney test were applied to compare statistical differences between the two groups. We used a chi-square test or Fisher’s exact test for comparisons of categorical variables. We used receiver operating characteristic (ROC) curves to determine the cut-off value of IPF at the best sensitivity and specificity and the area under the curve was calculated. Logistic regression analysis with the laboratory variables was performed to access their diagnostic contribution. A p values less than 0.05 were considered statistically significant.

A total of 568 patients with thrombocytopenia (platelet count less than 100 × 10⁹/L) were evaluated in this study. Of these, 206 were diagnosed with ITP according to the definitions proposed by the International Working Group, and 362 were classified as having thrombocytopenia due to causes other than ITP. The most common cause of thrombocytopenia in the non-ITP group was primary bone marrow diseases (n = 132), followed by infection (n = 69), drugs (n = 56), liver disease (n = 40), disseminated intravascular coagulation (n = 30), and others (n = 35) including chronic kidney disease, other autoimmune diseases, megaloblastic anemia, and bone marrow involvement of other cancers (Table 1).

The baseline laboratory data of the ITP group and non-ITP group were compared in Table 2. The ITP group had a significantly higher hemoglobin concentration (p < 0.001), white blood cell count (p < 0.001), and lower LDH (p < 0.001), ferritin level (p < 0.001). The median IPF (%) was significantly higher in the ITP group; the median IPF (%) (range) of the ITP group was 8.7% (1.3% to 47.5%) while that of the non-ITP group was 5.1% (0.7% to 31.5%) (p < 0.001). In order to obtain the optimal cut-off value of IPF (%) for the diagnosis of ITP, ROC curve analysis was carried out. As a result, the optimal cut-off value of IPF (%) for differentiating ITP was 7.0% with a sensitivity of 61% and a specificity of 70% (Fig. 1A). Bone marrow examinations are not mandatory for the diagnosis of ITP, but some patients who have clinically diagnosed ITP without a bone marrow examination may be difficult to differentiate from those with early bone marrow disease. To exclude this possibility of misclassification, we performed subgroup analysis with the patients who had an accurate diagnosis with bone marrow examination. A total of 200 patients underwent bone marrow examination; 77 patients were classified into the ITP group and 123 patients into the non-ITP group. The median IPF (%) was significantly higher in the ITP group. The median IPF (%) (range) was 12.5% (1.9% to 45.7%) for the ITP group and 4.7% (0.7% to 31.5%) for the non-ITP group. The optimal cut-off value was 7.8%, with a sensitivity of 75% and a specificity of 76.7% (Fig. 1B). The median IPF was higher in this subgroup and the sensitivity and specificity of the cut-off value were also higher than the former group. We confirmed that IPF could be a useful parameter for diagnosing ITP, but since IPF alone could not be used to diagnose ITP, we also evaluated other laboratory variables via logistic regression analysis. Hemoglobin, LDH, and ferritin showed statistical significance with p < 0.001, p = 0.011, and p < 0.001, respectively. The optimal cut-off value of hemoglobin for distinguishing between ITP and non-ITP was 12 g/dL with a sensitivity of 60% and specificity of 81.3%. The optimal cut-off value of LDH was 420 IU/L (upper limit of normal [ULN]) with sensitivity 57.9%, specificity 70.5% and the cut-off value of ferritin was 177 mg/mL with sensitivity 67%, specificity 83.3% (Fig. 2).

These four independent variables which were identified through the logistic regression analysis comprised a simple diagnostic predictive scoring model (Table 3). To adequately reflect sensitivity and specificity, we divided patients into three groups according to IPF; IPF < 7, ≥ 7 and < 10, ≥ 10. Our model gave point to each of variables; 1 to high hemoglobin level (> 12 g/dL), low ferritin level (≤ 177 ng/mL), normal LDH level (≤ ULN) and IPF ≥ 7 and < 10; and 2 to IPF ≥ 10. The final score was obtained by summing the points (Table 4). The scoring model underwent validation. In the internal validation cohort consisting of patients who underwent bone marrow examination, a total of 166 patients had all variables of the scoring system. Sixty-four patients were diagnosed with ITP and 102 patients were diagnosed with non-ITP diseases. In this cohort, none of the 64 patients with ITP were assigned a score of 0; 55 patients (85.9%) were assigned a score of 3–5. We defined that ITP could be highly predicted in patients with more than 3 points. Accuracy, precision and recall were 87%, 82%, and 86%, respectively. We also applied this scoring model to the external validation cohort. The external validation cohort included patients with thrombocytopenia (platelet count < 100 × 10⁹/L) of various causes who presented to Korea University Guro Hospital from January 2018 to December 2018 or Korea University Anam Hospital from January 2017 to December 2018. A total of 171 patients who underwent bone marrow examination had all variables and could be scored. Twenty-nine patients were diagnosed with ITP and 142 patients were diagnosed as non-ITP. None of 29 patients with ITP were assigned a score of 0 to 1; 26 patients (89.7%) were assigned a score of 3 to 5. We also found that ITP could be highly predicted in patients with more than 3 points. Accuracy, precision and recall were 91%, 67%, and 90%, respectively.