Machine learning algorithm as a prognostic tool for Epstein-Barr virus reactivation after haploidentical hematopoietic stem cell transplantation

Peripheral blood stem cell transplantation from haploidentical related donor could achieve satisfactory clinical outcomes for intermediate- or high-risk adult acute myeloid leukemia patients

Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is the most important curative method for intermediate- and high-risk adult acute myeloid leukemia (AML) patients. We aimed to identify the clinical outcomes of haploidentical related donor (HID) peripheral blood stem cell transplantation (PBSCT) who receiving peripheral blood (G-PB) harvest, and the patients receiving bone marrow (BM) plus G-PB harvest (BM + PB) as grafts were enrolled as control. The engraftments of neutrophil and platelet in G-PB group were both faster than those in BM + PB group. The cumulative incidences of grade II-IV acute graft-versus-host disease (aGVHD), and moderate to severe chronic GVHD (cGVHD) were all comparable between G-PB and BM + PB groups. The cumulative incidence of relapse and non-relapse mortality at 3 years after HID HSCT was 12.6% versus 13.7% (p = 0.899) and 3.6% versus 7.3% (p = 0.295), respectively, in G-PB and BM + PB group. While the probabilities of GVHD-free/relapse-free survival, leukemia-free survival, and overall survival at 3 years after HID HSCT were 60.6% versus 53.4% (p = 0.333), 83.8% versus 79.0% (p = 0.603), and were 87.3% versus 82.9% (p = 0.670), respectively. We confirmed the safety and efficacy of HID PBSCT in intermediate- and high-risk AML patients in a large cohort.

Early T-cell reconstitution predicts risk of EBV reactivation after allogeneic hematopoietic stem cell transplantation

The quality of immune reconstitution (IR) is crucial for the outcome of patients who received allogeneic hematopoietic stem cell transplantation (allo-HSCT), and is closely connected with infection, relapse and graft-versus-host disease (GvHD) which are the most important causes for transplantation failure. However, the IR pattern in the early stage after allo-HSCT, particularly haploidentical (HID) HSCT, remains unclear. In this retrospective study, we examined the T cell reconstitution of patients within the initial 30 days (n = 173) and 100 days (n = 122) after allo-HSCT with myeloablative condition (MAC), of which > 70% were HID HSCT, to assess the influence of IR on the transplant outcomes. By comparing 78 patients with good IR (GIR) to 44 patients with poor IR (PIR), we observed that GIR was associated with lower risk for Epstein-Barr virus (EBV) reactivation and cytomegalovirus (CMV) reactivation, but had no significant impacts on the survival outcomes (i.e., overall survival, event-free survival) and cumulative incidences of GvHD. Importantly, we found lymphocyte reconstitution pattern at day 30 after allo-HSCT would be a surrogate for IR evaluated at day 100. In the Cox proportional hazard model, early reconstitution of CD4+, CD4+CD25+, CD4+CD45RO+, CD4+CD25+CD27low, and CD8+ T cells at day 30 was reversely correlated with risk of EBV reactivation. Finally, we constructed a predictive model for EBV reactivation with CD8+ and CD4+CD45RO+ T cell proportions of the training cohort (n = 102), which was validated with a validation cohort (n = 37). In summary, our study found that the quality of IR at day 30 had a predictive value for the risk of EBV reactivation, and might provide guidance for close monitoring for EBV reactivation.

Outcomes of haploidentical haematopoietic stem cell transplantation for adolescent and young adults with acute myeloid leukaemia

We aimed to identify the efficacy of haploidentical related donor (HID) haematopoietic stem cell transplantation (HSCT) in adolescent and young adults (AYAs) with acute myeloid leukaemia (AML) in a large cohort. Consecutive AML AYAs (15-39 years old, n = 599) receiving HID HSCT in complete remission (CR) were included. The 3-year cumulative incidence of measurable residual disease occurrence, relapse and non-relapse mortality after HID HSCT was 28.6% (95% CI: 25.0-32.2), 11.6% (95% CI: 9.0-14.2) and 6.7% (95% CI: 4.7-8.7) respectively. The 3-year probability of event-free survival, leukaemia-free survival (LFS) and overall survival (OS) after HID HSCT was 60.7% (95% CI: 56.9-64.8), 81.7% (95% CI: 78.7-84.9) and 85.6% (95% CI: 82.8-88.4) respectively. In multivariable analysis, AML risk category at diagnosis and comorbidity burdens before HID HSCT were independently associated with LFS and OS. Compared to the older adults (≥ 40 years, n = 355) with AML receiving HID HSCT in CR during the same time period, AYAs have a lower incidence of non-relapse mortality and higher probabilities of LFS and OS. Thus, we firstly confirmed the safety and efficacy of HID HSCT in AYAs with AML-CR.

simpleNomo: A Python Package of Making Nomograms for Visualizable Calculation of Logistic Regression Models

Background

Logistic regression models are widely used in clinical prediction, but their application in resource-poor settings or areas without internet access can be challenging. Nomograms can serve as a useful visualization tool to speed up the calculation procedure, but existing nomogram generators often require the input of raw data, inhibiting the transformation of established logistic regression models that only provide coefficients. Developing a tool that can generate nomograms directly from logistic regression coefficients would greatly increase usability and facilitate the translation of research findings into patient care.

Methods

We designed and developed simpleNomo, an open-source Python toolbox that enables the construction of nomograms for logistic regression models. Uniquely, simpleNomo allows for the creation of nomograms using only the coefficients of the model. Further, we also devoloped an online website for nomogram generation.

Results

simpleNomo properly maintains the predictive ability of the original logistic regression model and easy to follow. simpleNomo is compatible with Python 3 and can be installed through Python Package Index (PyPI) or https://github.com/Hhy096/nomogram.

Conclusion

This paper presents simpleNomo, an open-source Python toolbox for generating nomograms for logistic regression models. It facilitates the process of transferring established logistic regression models to nomograms and can further convert more existing works into practical use.