TL;DR
This study validates three predictive models for early severe infections in newly diagnosed multiple myeloma (NDMM) patients. The FIRST score, a simple combination of ECOG performance status, β2-microglobulin, hemoglobin, and lactate dehydrogenase, proved to be a robust and practical tool for stratifying infection risk, especially in frail patients and those treated with bortezomib.
What Problem Does This Paper Address?
Multiple myeloma (MM) is a blood cancer where patients are highly susceptible to infections, which are a major cause of early mortality and treatment complications. While survival rates have improved, early mortality due to infections remains a significant challenge. Accurately predicting which patients are at highest risk for severe infections (grade ≥3) is crucial for implementing timely preventive strategies. Existing predictive models, developed in controlled clinical trial settings, need validation in real-world, diverse patient populations, particularly considering factors like frailty and common treatment regimens. This paper addresses this by evaluating and comparing three established infection risk prediction models in a real-world, multicenter cohort of NDMM patients.
Methods
The researchers conducted a retrospective, multicenter study involving 306 patients diagnosed with NDMM. They evaluated three existing predictive models for early grade ≥3 infections:
- FIRST score: Based on Eastern Cooperative Oncology Group (ECOG) performance status, serum β2-microglobulin, lactate dehydrogenase, and hemoglobin levels.
- GEM-PETHEMA score: Incorporates serum albumin, ECOG performance status, male sex, and MM subtype.
- Infection Risk model of Multiple Myeloma (IRMM) score: Utilizes ECOG performance status, serum β2-microglobulin, globulin, and hemoglobin.
The study retrospectively analyzed patient data to estimate the cumulative incidence of early grade ≥3 infections using the Kaplan-Meier method and log-rank tests. To compare the predictive performance of the models, Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC) were calculated, with DeLong's test used for statistical comparisons. Subgroup analyses were performed to assess the models' performance in specific patient groups, such as those treated with bortezomib-based regimens and frail patients.
Datasets
The study utilized data from 306 patients with newly diagnosed multiple myeloma (NDMM) collected from three hospitals in China between May 2016 and April 2022. This represents a real-world dataset, distinct from the controlled environments of clinical trials where some of the predictive models were initially developed.
Key Insights and Ideas
- Real-world validation is critical: Models developed in clinical trials may not perform identically in the complex, heterogeneous real-world setting. This study highlights the importance of such validation.
- FIRST score shines in practice: Despite AUC values being statistically similar across models, the FIRST score's simplicity in calculation and its robust performance in stratifying risk, particularly in vulnerable subgroups, makes it a strong candidate for routine clinical use.
- Frailty and treatment matter: The FIRST score's continued effectiveness in frail patients and those receiving common bortezomib-based therapies underscores its adaptability to diverse clinical scenarios.
- Infection burden is high: The study reported a substantial rate of early infections (66.01%) and severe infections (40.20%) in NDMM patients, emphasizing the ongoing need for effective risk prediction.
Key Results
- High infection rates: 66.01% of NDMM patients experienced infections within the first 4 months, with 40.20% experiencing grade ≥3 infections.
- FIRST score effectiveness: The FIRST score effectively stratified patients into low-risk (25.81% grade ≥3 infection) and high-risk (50.00% grade ≥3 infection) groups, with a statistically significant difference (p<0.001).
- GEM-PETHEMA score limitations: While showing some statistical differences, the GEM-PETHEMA score's performance in discriminating risk groups was less pronounced (p=0.045) and did not show statistically significant differences in cumulative incidence between groups (p=0.090).
- IRMM score performance: The IRMM score also demonstrated statistically significant differences in infection probability between risk groups (p<0.001), though the distinction between moderate and high-risk groups was not significant.
- AUC comparison: While the AUC values for the three models were not statistically different (FIRST score: 0.650, GEM-PETHEMA: 0.619, IRMM: 0.630), the FIRST score was deemed slightly better and easier to implement.
- Subgroup analysis: The FIRST score maintained its predictive power in frail patients and those treated with bortezomib-based regimens.
Limitations and Future Work
The authors acknowledge several limitations:
- Small sample size: Despite being multicenter, the sample size was considered relatively small.
- Retrospective design: This design inherently carries the risk of biases.
- Limited use of novel therapies: The study cohort had limited exposure to newer treatments like daratumumab, which may influence infection rates. Future studies should investigate the performance of these models with such regimens.
- COVID-19 impact: The study did not specifically account for COVID-19 infections, which could be a factor in infection rates and risk. Further investigation into COVID-19's role in NDMM patients is suggested.
Future work should focus on prospective validation of the FIRST score, its integration into clinical decision-making, and its evaluation in patient cohorts receiving newer therapeutic agents and in the context of ongoing pandemics.
Summary updated