The treatment timing of a procedure in patients with hematologic neoplasms is crucial, especially when the therapy concerns the Hematopoietic Stem-Cell Transplantation (HSCT), which is the only effective treatment in patients affected by Chronic Myelomonocytic Leukemia (CMML), a rare disease with a high mortality rate. Using multi-state models, cure models and microsimulation, we identified the optimal HSCT timing for each patient profile and quantified its curative potential. Moreover, since patients affected by this particular leukemia have widely heterogeneous outcomes, a risk-based strategy is implemented, integrating our decision support system to define the optimal HSCT timing with clinical, molecular and genomic information, as provided by three risk scores: cpss_risk, cpssmol_risk and i_cpss_risk. The model selection procedure, which represents the biggest original contribution of our thesis, has been performed by relying on the standard BIC approach and on an innovative goodness-of-fit index, the Adapted Brier Score, which captures the efficacy of a multi-state model by comparing the transition probability estimated by it with the empirical one observed in real data. Moreover, the same new metric represents a valuable method to validate our multi-state model by comparing its value on the train set and on the test set. This work is highly clinically relevant because, by relying on microsimulation techniques, it provides an alternative to randomized trials that, even if very effective in generating clinical evidence, are ethically debatable and often practically constrained. Moreover, even if our approach is built on observational data of patients affected by CMML, it can be easily extended, representing a non-invasive approach useful for defining the optimal timing of general once-in-a-lifetime treatments. The results of the methodology and their relevance are validated by comparing the proportion of "cured" patients undergoing HSCT within the optimal window to those transplanted outside it. Our findings demonstrate that adherence to the optimal HSCT window significantly improves survival and reduces the risk of relapse, providing clinicians with a robust and personalized decision-support tool for CMML management.
La tempistica del trattamento nei pazienti con neoplasie ematologiche è un aspetto critico, in particolare quando la terapia riguarda il trapianto di cellule staminali ematopoietiche (HSCT), che rappresenta l’unico trattamento efficace nei pazienti affetti da leucemia mielomonocitica cronica (CMML), rara malattia caratterizzata da un’elevata mortalità. Utilizzando modelli multi-stato, modelli di cura e microsimulazioni, abbiamo identificato il momento ottimale per l’HSCT in base al profilo di ciascun paziente e ne abbiamo quantificato il potenziale curativo. Inoltre, poiché i pazienti affetti da questa particolare leucemia presentano esiti estremamente eterogenei, è stata implementata una strategia basata sul rischio, che integra il nostro sistema di supporto decisionale con informazioni cliniche, molecolari e genomiche, fornite da tre indici prognostici: cpss_risk, cpssmol_risk e i_cpss_risk. L’innovativa procedure di selezione del modello, che rappresenta il maggiore contributo di questa tesi, è stata effettuata utilizzando l’approccio standard basato sul BIC e un nuovo indice di bontà di adattamento, l’Adapted Brier Score, che misura l’efficacia di un modello multi-stato confrontando la probabilità di transizione stimata con quella empirica osservata nei dati reali. Inoltre, la stessa nuova metrica si è rivelata uno strumento utile per validare il modello multi-stato, confrontando i valori ottenuti sull’insieme di training e su quello di test. Questo lavoro ha una forte rilevanza clinica, poiché l’uso delle tecniche di microsimulazione rappresenta un’alternativa ai trial randomizzati che, sebbene molto efficaci nel fornire evidenze cliniche, risultano spesso discutibili dal punto di vista etico e difficilmente realizzabili nella pratica. Inoltre, sebbene il nostro approccio sia costruito su dati osservazionali di pazienti affetti da CMML, esso può essere facilmente esteso, configurandosi come una strategia non invasiva utile per definire il momento ottimale di trattamenti somministrabili una sola volta durante la vita del paziente. I risultati della metodologia e la loro rilevanza vengono validati confrontando la proporzione di pazienti "guariti" sottoposti a HSCT entro la finestra ottimale con quella dei pazienti trapiantati al di fuori di essa. I nostri risultati dimostrano che aderire alla tempistica ottimale dell’HSCT migliora significativamente la sopravvivenza e riduce il rischio di recidiva, offrendo ai clinici un potente e personalizzato strumento di supporto decisionale per la gestione della CMML.
Multi-state and cure models for personalized optimal timing in HSCT in patients affected by CMML
Flamigni, Alice
2024/2025
Abstract
The treatment timing of a procedure in patients with hematologic neoplasms is crucial, especially when the therapy concerns the Hematopoietic Stem-Cell Transplantation (HSCT), which is the only effective treatment in patients affected by Chronic Myelomonocytic Leukemia (CMML), a rare disease with a high mortality rate. Using multi-state models, cure models and microsimulation, we identified the optimal HSCT timing for each patient profile and quantified its curative potential. Moreover, since patients affected by this particular leukemia have widely heterogeneous outcomes, a risk-based strategy is implemented, integrating our decision support system to define the optimal HSCT timing with clinical, molecular and genomic information, as provided by three risk scores: cpss_risk, cpssmol_risk and i_cpss_risk. The model selection procedure, which represents the biggest original contribution of our thesis, has been performed by relying on the standard BIC approach and on an innovative goodness-of-fit index, the Adapted Brier Score, which captures the efficacy of a multi-state model by comparing the transition probability estimated by it with the empirical one observed in real data. Moreover, the same new metric represents a valuable method to validate our multi-state model by comparing its value on the train set and on the test set. This work is highly clinically relevant because, by relying on microsimulation techniques, it provides an alternative to randomized trials that, even if very effective in generating clinical evidence, are ethically debatable and often practically constrained. Moreover, even if our approach is built on observational data of patients affected by CMML, it can be easily extended, representing a non-invasive approach useful for defining the optimal timing of general once-in-a-lifetime treatments. The results of the methodology and their relevance are validated by comparing the proportion of "cured" patients undergoing HSCT within the optimal window to those transplanted outside it. Our findings demonstrate that adherence to the optimal HSCT window significantly improves survival and reduces the risk of relapse, providing clinicians with a robust and personalized decision-support tool for CMML management.| File | Dimensione | Formato | |
|---|---|---|---|
|
2025_10_Flamigni_Executive_summary_02.pdf
accessibile in internet per tutti a partire dal 29/09/2026
Descrizione: Executive Summary
Dimensione
833.43 kB
Formato
Adobe PDF
|
833.43 kB | Adobe PDF | Visualizza/Apri |
|
2025_10_Flamigni_Tesi_01.pdf
accessibile in internet per tutti a partire dal 29/09/2026
Descrizione: Tesi
Dimensione
2.42 MB
Formato
Adobe PDF
|
2.42 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/243527