Novel algorithmic approach to generate consensus guidelines in AML
Coats, T. Basset, A. Bean, D. Bowen, D. Hunter, H. Mehta, P. Raj, K. Knapper, S. Dennis, Michael Khwaja, A.
Coats, T.
Basset, A.
Bean, D.
Bowen, D.
Hunter, H.
Mehta, P.
Raj, K.
Knapper, S.
Dennis, Michael
Khwaja, A.
Citations
Altmetric:
Abstract
Abstract Content: Treatment options for acute myeloid leukaemia
(AML) have become more complex with the licensing of 4 new
drugs for 1st line treatment. Clinicians have the dual challenge of
establishing a patients’ eligibility for each drug and assessing their
relative effectiveness in different clinical scenarios.
To help solve this problem, we developed an algorithmic
approach to identify the different treatment paradigms for AML and
surveyed UK experts to understand the degree of consensus for each
scenario.
We created a series of decision trees (DTs) to convert the eligibility
criteria for upfront AML treatments into a digital format, based
on NICE guidance. A DT was also designed to replicate the ELN risk
groupings (CBF, Favourable, Intermediate, Adverse) based on molecular
and cytogenetic features. All DTs were designed using open
source software esyN (www.esyn.org). 1000 in silico AML cases were
created to cover a variety of AML clinical and genetic features. Cases were classified by the DTs and assigned to one of 20 paradigms,
based on ELN risk group and drug eligibility. All cases were eligible
for daunorubucin + cytarabine (DA).
One representative case from each paradigm was identified for
review by 9 AML experts who were asked to select their preferred
induction chemotherapy for a 40- and 65-year-old patient, both with
good performance status and no major comorbidities. A second
question asked if FLAG-IDA was preferred over the initial choice. A
threshold for establishing a strong consensus was arbitrarily set as
>=85% agreement on 1st line choice and a weak consensus of >=
75% agreement. To compare the outcomes of the survey to an existing
guideline, the ESMO Clinical Practice Guideline was converted
into a DT as above and applied to the selected cases to give a recommended
ESMO treatment.
The survey revealed that for a 40-year-old patient, there is a
strong consensus in 13/20 paradigms, a weak consensus in 2/20 paradigms
and no consensus in 5/20 paradigms. For a 65-year-old, there
is a strong consensus in 11/20 paradigms, a weak consensus in 3/20
paradigms and no consensus in 6/20 paradigms.
In 16/24 of the paradigms with a strong consensus in the survey,
the ESMO recommendation is the same. In 4 paradigms where
DA+gemtuzumab (GO) is the preferred option, the ESMO guidance
is less specific with DA+/ GO recommended. ESMO recommended
treatment is different in 2 paradigms (DA over CPX in a 40-year-old
with intermediate risk AML and prior MDS, and CPX over
DA+Midostaurin in a 65-year-old with intermediate risk AML). For
2 paradigms, there is no recommendation by ESMO as that clinical
scenario is not included in the guideline.
At least 3 of the clinicians surveyed preferred FLAG-IDA as 1st
line therapy in 5 paradigms for a 40-year-old and in one scenario for
a 65-year-old. In the ESMO guidance FLAG-IDA was suggested as
an option in 2 and one paradigms, respectively.
Our algorithmic approach successfully assigned all cases to the 20
treatment paradigms and highlighted 2 paradigms that were not covered
by an existing guideline. Consensus was established in the
majority of scenarios, but barriers included rigidity of drug approval
or lack of evidence for rarer clinical/genetic paradigms. The survey
suggests there are some differences in current treatment practices in
the UK compared to those published by ESMO, and formal guidelines
are needed to reflect this. This survey is the first part of a Delphi
method approach to generate a UK consensus guideline.
Description
Date
2021
Publisher
Collections
Keywords
Type
Meetings and Proceedings
Citation
Coats T, Basset A, Bean D, Dobson R, Bowen D, Hunter H, et al. Novel algorithmic approach to generate consensus guidelines in AML. British Journal of Haematology. 2021;193:5.