CLINICAL EFFICACY OF A NOVEL VALIDATED PROGNOSTIC INDEX FOR TRIAL DESIGN IN ADULT ACUTE LYMPHOBLASTIC LEUKAEMIA.
Author(s): ,
Anthony Moorman
Affiliations:
Northern Institute for Cancer Research,Newcastle University,Newcastle upon Tyne,United Kingdom
,
Amy Kirkwood
Affiliations:
CRUK and UCL Cancer Trials Center,UCL Cancer Institute,London,United Kingdom
,
Amir Enshaei
Affiliations:
Northern Institute for Cancer Research,Newcastle University,Newcastle upon Tyne,United Kingdom
,
Laura Clifton-Hadley
Affiliations:
CRUK and UCL Cancer Trials Center,UCL Cancer Institute,London,United Kingdom
,
Emma Lawrie
Affiliations:
CRUK and UCL Cancer Trials Center,UCL Cancer Institute,London,United Kingdom
,
David Marks
Affiliations:
Haematology,United Hospitals Bristol,Bristol,United Kingdom
,
Andrew McMillan
Affiliations:
Haematology,Nottingham University Hospitals,Nottingham,United Kingdom
,
Tobias Menne
Affiliations:
Haematology,Freeman Hospital,Newcastle upon Tyne,United Kingdom
,
Pip Patrick
Affiliations:
CRUK and UCL Cancer Trials Center,UCL Cancer Institute,London,United Kingdom
,
Bela Wrench
Affiliations:
Haematology,Barts Cancer Institute,London,United Kingdom
,
Krisztina Zuborne-Alapi
Affiliations:
UCL Cancer Institute,UCL,London,United Kingdom
,
Clare Rowntree
Affiliations:
Haematology,University Hospital of Wales,Cardiff,United Kingdom
Adele Fielding
Affiliations:
UCL Cancer Institute,UCL,London,United Kingdom
EHA Library. Moorman A. 06/16/19; 267375; S1621
Prof. Anthony Moorman
Prof. Anthony Moorman
Contributions
Abstract

Abstract: S1621

Type: Oral Presentation

Presentation during EHA24: On Sunday, June 16, 2019 from 09:00 - 09:15

Location: Elicium 1

Background

The majority of clinical trials in ALL use a combination of binary risk factors to assign patients to different treatment pathways. Examples include age, white cell count (WCC), minimal residual disease (MRD) and genetics. Typically, variables are dichotomised and used sequentially. This approach reduces predictive power by failing to leverage the power of continuous data and predetermines the size of the risk groups due to the skewed distribution of variables. A prognostic index (PIUKALL) using four paediatric clinical trials (UKALL2003, DCOG-ALL10, NOPHO-ALL2008 and CoALL-07-03) was recently developed in childhood ALL (AVM & AE). PIUKALL is based on two, weighted, log-transformed continuous variables (WCC and end of induction MRD) as well as two weighted binary variables – the presence of good risk and high risk cytogenetics. PIUKALL is readily calculated from a linear model and produces patient specific risk scores.

Aims

We aimed to determine whether PIUKALL was a valid prognostic marker in adult ALL and, if so, how it could be optimally ultilised to inform future risk stratification algorithms. 

Methods

Using the original linear model, we calculated patient specific risk scores for a representative cohort of 367 Philadelphia-negative B-cell precursor and T-ALL patients, aged 25-65 years, treated on UKALL14. UKALL14 used two induction phases, so we calculated risk scores for each time point: end of phase 1 (PI1, n=343) and phase 2 (PI2, n=295). We used the PIUKALL good risk cytogenetic abnormalities (ETV6-RUNX1, high hyperdiploidy) but added complex karyotype to the list of high risk cytogenetic abnormalities (low hypodiploidy, KMT2A gene fusions and iAMP21). We used standard endpoints and statistical methods to explore the prognostic impact and clinical utility of PIUKALL.

Results

PIUKALL is a continuous variable ranging, in this study, from -3.37 to 1.82 with higher scores indicating poorer outcome. The median score for PI1 was higher than PI2 (-0.99 v -1.99, p<0.001) reflecting the lower MRD levels after phase 2. As expected the median PI1 and PI2 scores differed according by baseline risk groups: PI1 standard -2.03 v high -0.77 p<0.0001; PI2 standard -2.47 v high -1.68 p<0.0001. Univariate analysis showed that each unit increase in PI1/PI2 resulted in a 50-60% greater risk of event or relapse. Both scores retained significance when adjusting for patient age. Dividing PI1 and PI2 into tertiles produced groups with distinct EFS rates (Figure). On UKALL14, standard risk patients were allocated maintenance chemotherapy while high risk patients were eligible for alloSCT. To gain insight into how PIUKALL could be used prospectively to inform treatement decisions, we examined the prognostic impact of both PI1 and PI2, using a variety of thresholds, according to post induction therapy. The table highlights examples of how PI1 and/or PI2 can be used to identify patients with differential outcomes despite receivng the same therapy. For example, patients with a >90% EFS with maintenance therapy or a 42% relapse risk despite receiving a myeloablative alloSCT.

Conclusion

PIUKALL, which was developed and validated using paediatric ALL data, is a valid and powerful biomarker in adult ALL. This intriguing observation highlights the benefit of integrating risk factors as continuous variables. We have demonstrated how PIUKALL could be used to provide additional prognostic information in treatment scenarios previously allocated by binary decision. We plan to use this risk score when designing our next adult ALL trial, UKALL15.

Session topic: 2. Acute lymphoblastic leukemia - Clinical

Abstract: S1621

Type: Oral Presentation

Presentation during EHA24: On Sunday, June 16, 2019 from 09:00 - 09:15

Location: Elicium 1

Background

The majority of clinical trials in ALL use a combination of binary risk factors to assign patients to different treatment pathways. Examples include age, white cell count (WCC), minimal residual disease (MRD) and genetics. Typically, variables are dichotomised and used sequentially. This approach reduces predictive power by failing to leverage the power of continuous data and predetermines the size of the risk groups due to the skewed distribution of variables. A prognostic index (PIUKALL) using four paediatric clinical trials (UKALL2003, DCOG-ALL10, NOPHO-ALL2008 and CoALL-07-03) was recently developed in childhood ALL (AVM & AE). PIUKALL is based on two, weighted, log-transformed continuous variables (WCC and end of induction MRD) as well as two weighted binary variables – the presence of good risk and high risk cytogenetics. PIUKALL is readily calculated from a linear model and produces patient specific risk scores.

Aims

We aimed to determine whether PIUKALL was a valid prognostic marker in adult ALL and, if so, how it could be optimally ultilised to inform future risk stratification algorithms. 

Methods

Using the original linear model, we calculated patient specific risk scores for a representative cohort of 367 Philadelphia-negative B-cell precursor and T-ALL patients, aged 25-65 years, treated on UKALL14. UKALL14 used two induction phases, so we calculated risk scores for each time point: end of phase 1 (PI1, n=343) and phase 2 (PI2, n=295). We used the PIUKALL good risk cytogenetic abnormalities (ETV6-RUNX1, high hyperdiploidy) but added complex karyotype to the list of high risk cytogenetic abnormalities (low hypodiploidy, KMT2A gene fusions and iAMP21). We used standard endpoints and statistical methods to explore the prognostic impact and clinical utility of PIUKALL.

Results

PIUKALL is a continuous variable ranging, in this study, from -3.37 to 1.82 with higher scores indicating poorer outcome. The median score for PI1 was higher than PI2 (-0.99 v -1.99, p<0.001) reflecting the lower MRD levels after phase 2. As expected the median PI1 and PI2 scores differed according by baseline risk groups: PI1 standard -2.03 v high -0.77 p<0.0001; PI2 standard -2.47 v high -1.68 p<0.0001. Univariate analysis showed that each unit increase in PI1/PI2 resulted in a 50-60% greater risk of event or relapse. Both scores retained significance when adjusting for patient age. Dividing PI1 and PI2 into tertiles produced groups with distinct EFS rates (Figure). On UKALL14, standard risk patients were allocated maintenance chemotherapy while high risk patients were eligible for alloSCT. To gain insight into how PIUKALL could be used prospectively to inform treatement decisions, we examined the prognostic impact of both PI1 and PI2, using a variety of thresholds, according to post induction therapy. The table highlights examples of how PI1 and/or PI2 can be used to identify patients with differential outcomes despite receivng the same therapy. For example, patients with a >90% EFS with maintenance therapy or a 42% relapse risk despite receiving a myeloablative alloSCT.

Conclusion

PIUKALL, which was developed and validated using paediatric ALL data, is a valid and powerful biomarker in adult ALL. This intriguing observation highlights the benefit of integrating risk factors as continuous variables. We have demonstrated how PIUKALL could be used to provide additional prognostic information in treatment scenarios previously allocated by binary decision. We plan to use this risk score when designing our next adult ALL trial, UKALL15.

Session topic: 2. Acute lymphoblastic leukemia - Clinical

By clicking “Accept Terms & all Cookies” or by continuing to browse, you agree to the storing of third-party cookies on your device to enhance your user experience and agree to the user terms and conditions of this learning management system (LMS).

Cookie Settings
Accept Terms & all Cookies