Analysis of Risk Factors and Development of a Nomogram Prediction Model for Stem Cell Mobilization Failure in Lymphoma Patients

Background

The mobilization and harvest of a sufficient amount of peripheral blood stem cells are essential to autologous hematopoietic stem cell transplantation (ASCT). However, up to 40% lymphoma patients failed with current stem cell mobilization. Risk factors which predicting mobilization failure in real-world situations has not been fully assessed. This study aimed to explore risk factors for stem cell mobilization failure in these patients and develop a nomogram prediction model to optimize future mobilization regimens.

Methods

The mobilization for stem cell collection was conducted in lymphoma patients who were eligible for ASCT after completing 3-4 cycles of first-line treatment or 1-2 cycles of salvage treatment. Clinical characteristics and laboratory data of 159 mobilizations from 146 lymphoma patients were randomly divided into training and testing sets. A multivariate logistic regression analysis was conducted to determine the risk factors of mobilization failure in these patients and was used to construct the nomogram model based on the training set. Moreover, the testing set evaluated the model's prediction efficiency.

Results

A total of 146 patients experiencing 159 mobilizations were included. The overall cohort had a median age of 58 years (range, 16-81 years) at the time of diagnosis. The number of male and female patients was similar (51.6% vs.48.4%). Ninety-eight (61.6%) patients were diagnosed with diffuse large B cell lymphoma, 11 (6.9%) with angioimmunoblastic T-cell lymphoma, and 11 (6.9%) with peripheral T-cell lymphoma. Other diseases include follicular lymphoma and NK/T cell lymphoma. Only 13 (8.2%) patients were diagnosed with type 2 diabetes, and 35 (22.0%) patients had chronic HBV infection. They had received a median of 4 courses (range, 2-22 courses). The median platelet count was 211×10⁹/L (range, (12-648) × 10⁹/L) before mobilization. All patients received a mobilization regimen of granulocyte-colony-stimulating factor (G-CSF) in combination with disease-specific chemotherapy, one hundred (62.9%) patients obtained at least 2×10⁶/kg CD34+ cells, and 46 (28.9%) achieved optimal collection.

All cases (N=159) were randomly partitioned into a training set (N=96) and a testing set (N=63). There was no discernible difference between training and testing sets in terms of clinical characteristics. In the training set, 12 variables were compared between the successful and unsuccessful groups. In the univariate analysis, intervals from diagnosis >140 days (p=0.011) were significant predictors of mobilization failure, while platelet count before mobilization>150×10⁹/L (p=0.001) was a protective factor. In a multiple logistic regression analysis, the interval from diagnosis >140 days and platelet count before mobilization ≤150×10⁹/L remained risk factors for mobilization failure. The odds ratios for the interval from diagnosis >140 days were 3.33 (1.03-10.75; P = 0.044), and for platelet count before mobilization ≤150×10⁹/L were 0.21 (0.07-0.6; P = 0.003).

The multiple logistic regression analysis revealed that failure of mobilization was associated with an interval from diagnosis to mobilization >140 days and a platelet count before mobilization ≤150×10⁹/L (P < 0.05). The nomogram presented a prediction model for mobilization failure: logit(p)=-0.01794+1.204*if interval from diagnosis >140 days-1.581*if platelet count before mobilization >150×10⁹/L. The prediction model displayed acceptable predictive accuracy, with area under the curve (AUC) values of 0.726 in both the training and testing sets. The calibration curve validation indicated good agreement between predicted and actual outcomes.

Conclusion

Identifying variables leading to mobilization failure in lymphoma patients can help identify poor mobilizers in advance and develop timely interventions. We showed that a diagnosis to mobilization interval of more than 140 days and a platelet count of less than or equal to 150×109/L before mobilization were independent risk factors for mobilization failure in lymphoma patients. In this study, we developed a risk prediction model for mobilization failure in lymphoma patients. The model could predict the risk of mobilization failure before the start of the mobilization regimen and implement individualized mobilization strategies, therefore reducing the cost of mobilization failure.

No relevant conflicts of interest to declare.