A single-center retrospective cohort study published in Frontiers in Endocrinology has identified associations between three derived metabolic indices and a poor early response to corticosteroid therapy in adults with primary immune thrombocytopenia (ITP). The research, conducted at Shandong Provincial Hospital Affiliated to Shandong First Medical University, analyzed data from 259 patients to evaluate whether routine laboratory measurements could help predict which individuals are less likely to achieve a platelet response by day seven of treatment. The study focused on the atherogenic index of plasma (AIP), the triglyceride-glucose index (TyG), and the log-transformed blood urea nitrogen-to-high-density lipoprotein cholesterol ratio (logBHR).
Immune thrombocytopenia is an acquired autoimmune disorder characterized by increased platelet destruction and impaired platelet production. While corticosteroids remain the most commonly used first-line treatment for adults with newly diagnosed ITP, early responses vary considerably among patients. A substantial proportion of patients show an inadequate initial response or experience rapid relapse, which can increase exposure to corticosteroid-related adverse effects and delay the initiation of subsequent treatment strategies. Therefore, identifying patients at high risk of a poor response early in the treatment course is clinically important for optimizing individualized treatment decisions and minimizing ineffective corticosteroid exposure.
The study included adults aged 18 to 84 years who were hospitalized between January 2019 and February 2026. Participants were required to have newly diagnosed primary ITP and to receive a corticosteroid-containing first-line regimen, such as prednisone, methylprednisolone, or high-dose dexamethasone. The primary outcome was day-7 nonresponse, defined as failure to meet the criteria for complete response or response according to the Chinese Guideline for the Diagnosis and Treatment of Adult Primary Immune Thrombocytopenia. Complete response was defined as a platelet count of at least 100 × 10^9/L with no bleeding, while response required a platelet count of at least 30 × 10^9/L, a twofold increase from baseline, and no bleeding.
Among the 259 patients included in the analysis, 66 (25.5%) had a poor early response, while 193 (74.5%) achieved an early treatment response by day seven. The median age of the participants was 51 years, and 63.3% were women. Baseline characteristics showed that patients in the poor-response group had higher levels of AIP, TyG, and logBHR compared to those in the early-response group. The study used multivariable logistic regression to evaluate the associations of these indices with early treatment response, adjusting for prespecified covariates including age, sex, and viral markers such as Epstein-Barr virus status and hepatitis B surface antigen status.
After adjustment for these covariates, the odds ratios for day-7 nonresponse per one-unit increase in AIP, TyG, and logBHR were 5.726, 2.105, and 2.331, respectively. This indicates that higher levels of these metabolic indices were associated with greater odds of nonresponse. When analyzed as categorical variables, patients in the highest quartile of each index had significantly greater odds of poor early response compared to those in the lowest quartile. The adjusted odds ratios for the highest quartile versus the lowest quartile were 3.247 for AIP, 2.363 for TyG, and 4.272 for logBHR. Significant linear trends across quartiles were observed for all three indices, reinforcing the association between higher metabolic index values and poorer treatment outcomes.
Restricted cubic spline analyses were used to examine the functional forms of these associations. The results showed overall associations of AIP, TyG, and logBHR with day-7 nonresponse, with P values for overall association of 0.021, 0.034, and 0.047, respectively. No clear evidence of nonlinearity was found for any of the three indices, suggesting that the relationship between these metabolic markers and treatment response is largely linear. Subgroup analyses indicated that the associations were generally consistent in direction across various subgroups defined by sex, immunological markers, and viral status. However, a significant interaction was observed when patients were stratified by granular megakaryocyte proportion, with a stronger association between logBHR and nonresponse in patients with a proportion greater than 0.8.
The discriminatory performance of the derived indices was evaluated using receiver operating characteristic curve analysis. The adjusted models incorporating AIP, TyG, or logBHR yielded area under the curve values ranging from 0.658 to 0.675. The AIP model had the highest AUC at 0.675, followed by the TyG model at 0.662 and the logBHR model at 0.658. These values indicate limited ability to discriminate between patients with poor early response and those with early response. The authors note that these findings are exploratory and require prospective validation before clinical application. The study highlights the potential of routine laboratory measurements to reflect metabolic and inflammatory states associated with early treatment response in ITP.
The biological rationale for these associations lies in the interplay between metabolic regulation and immune function. AIP reflects the characteristics of the atherogenic lipoprotein profile, while TyG is a surrogate marker of insulin resistance and glucose-lipid metabolic dysfunction. BHR integrates blood urea nitrogen and high-density lipoprotein cholesterol, potentially reflecting nitrogen metabolism and lipid metabolic abnormalities. Disturbances in lipid metabolism may contribute to the development and progression of autoimmune diseases by altering immune-cell energy supply and inflammatory signaling. Elevated triglyceride and reduced high-density lipoprotein cholesterol levels may promote oxidative stress and chronic low-grade inflammation, potentially exacerbating platelet destruction in ITP.
The study has several limitations that must be considered when interpreting the results. The single-center retrospective design may have introduced selection bias, information bias, and unmeasured confounding. The use of a single baseline measurement for each index limited the assessment of within-person variability. Additionally, the regression models did not adjust for corticosteroid regimen, dose, or concomitant therapy, which may have influenced day-7 response status. The number of nonresponse events was limited, which restricted the precision of the multivariable models and subgroup analyses. The ROC results represent apparent performance estimated from the same dataset, without correction for optimism through internal validation or confirmation in an external cohort.
Despite these limitations, the study provides valuable insights into the potential role of metabolic indices in predicting early treatment response in ITP. The findings suggest that indices derived from routine laboratory measurements may be useful for identifying patients who warrant closer early monitoring. However, these indices should not be used alone to guide treatment decisions. Further research is needed to validate these associations in independent prospective cohorts with standardized treatment protocols and outcome assessments.
Methodological details of the investigation outline how laboratory values were curated and transformed prior to statistical evaluation. To address incomplete records, missing data points were imputed using a random forest technique across five datasets, with the initial completed dataset selected for analysis, followed by winsorization of selected variables at the first and ninety-ninth percentiles to minimize distortion from extreme outliers. The atherogenic index of plasma was derived as the base-ten logarithm of the ratio of triglycerides to high-density lipoprotein cholesterol in millimoles per liter. The triglyceride-glucose index was calculated using natural logarithms after converting triglycerides and fasting plasma glucose into milligrams per deciliter, dividing their product by two. For the blood urea nitrogen-to-high-density lipoprotein ratio, right-skewed raw values necessitated natural-log transformation, meaning each one-unit increase in logBHR corresponded to an e-fold change, or approximately a 2.72-fold rise in the baseline ratio.
Beyond aggregate discrimination statistics, the investigation established operational cutoff thresholds for each adjusted model using the Youden index. For the atherogenic index model, a probability threshold of 0.276 yielded a sensitivity of 0.576, a specificity of 0.741, an overall accuracy of 0.699, a positive predictive value of 0.432, and a negative predictive value of 0.836, with a Youden index of 0.317. In contrast, the triglyceride-glucose model applied a probability cutoff of 0.206, achieving higher sensitivity at 0.848 and a negative predictive value of 0.896, but lower specificity at 0.446, an accuracy of 0.548, and a positive predictive value of 0.344. The log-transformed ratio model demonstrated comparable test properties at a threshold of 0.204, producing a sensitivity of 0.833, a specificity of 0.446, an accuracy of 0.544, a positive predictive value of 0.340, and a negative predictive value of 0.887. These operating profiles demonstrate that while the indices captured most poor responders, false-positive rates remained substantial.
Crucially, the researchers emphasize that failure to respond by day seven represents an early milestone rather than definitive treatment resistance or refractory disease. Guidelines from the International Working Group on immune thrombocytopenia recommend assessing clinical responses in accordance with the expected onset of therapy, recognizing that certain patients who miss early response criteria may still recover acceptable platelet levels later. Day-seven status remains clinically meaningful because delayed platelet recovery leaves patients vulnerable to prolonged, severe thrombocytopenia and bleeding hazards. In such scenarios, recognizing elevated baseline metabolic markers may alert clinicians to monitor platelet trajectories and bleeding signs more vigilantly, or to prepare timely rescue interventions, such as intravenous immunoglobulin, when severe bleeding risks demand urgent platelet elevation.
Subject of Research: Association between derived metabolic indices and early corticosteroid response in primary immune thrombocytopenia
Article Title: Derived metabolic indices and poor early response to corticosteroids in adults with primary immune thrombocytopenia: a retrospective cohort study
Article References: Zhou, X., Lu, G., Song, J., Tian, X., Zhang, L., & Shan, N. (2026). Derived metabolic indices and poor early response to corticosteroids in adults with primary immune thrombocytopenia: a retrospective cohort study. Frontiers in Endocrinology, 17, Article 1981987. https://doi.org/10.3389/fendo.2026.1981987
Image Credits: AI Generated
DOI: 10.3389/fendo.2026.1981987
Keywords: immune thrombocytopenia, corticosteroid therapy, metabolic indices, atherogenic index of plasma, triglyceride-glucose index, retrospective cohort study, treatment response, hematology, Derived, metabolic, indices, poor
