ADAM: A clinical decision support system to standardise management of the maintenance treatment phase of acute lymphoblastic leukaemia

About the project

(Figures cited in the write-up below are uploaded in the Gallery photo section of the Media tab)

A ssisted Dose A dvicein ALL M aintenance (ADAM)

Proponents

1. Tata Medical Center (TMC) is a specialty not-for-profit tertiary care cancer hospital in Kolkata. The Tata Translational Cancer Research Center (TTCRC) is a research institution embedded within TMC. TTCRC’s mission is to address disparities in cancer care for patients in India. Strategies include the use of high-throughput technologies, innovative laboratory models, integrated information systems, collaborative partnerships, and hub-spoke networks to provide equitable cancer care and to improve cancer outcomes.
2. Tata Consultancy Services (TCS) is an Indian multinational company providing services and solutions in information technology and management, advanced analytics and business development. TCS is part of the Tata Group of companies.
3. TTCRC and TCS established a research partnership in 2016 to advance the goal of addressing inequities in cancer care . These programs focused initially on one cancer, acute lymphoblastic leukemia (ALL).

Background and context of the initiative

ALL (acute lymphoblastic leukaemia), a cancer of white blood cells, is the commonest cancer in children. An estimated 15000 children (age 1-18 years) are diagnosed with ALL annually in India, representing 25% of all pediatric cancers. In high income countries, cure rates approach 90%. Cure rates are lower (65-70%) in India and other low-middle income countries (LMICs). Treatment is the most important determinant of outcome. Limited availability of contemporary treatment accounts for the poor ALL outcomes in LMICs.

Treatment of ALL: Intensive treatment phase

Treatment of ALL spans 2.5 to 3 years. Patients receive an intensive phase of treatment (6-8 months) followed by a continuation or maintenance phase of treatment for 24 months. The intensive phase of treatment involves use of drugs developed in the 1960s-70s that are available widely at affordable cost. Treatment in the intensive phase is tailored to the predicted risk of ALL relapse (‘risk stratification’) ( Figure 1.1-1.3 ). Patients predicted to be at highest risk of relapse receive additional drugs, additional drug doses and higher doses of drugs. Conversely, patients predicted to be at the lowest risk of relapse receive fewer drugs and fewer drug doses. Risk stratification of ALL is based on clinical features, tumor genetics and response to treatment ( Figure 2.1-2.3  ). The intensive phase requires close supervision and frequent visits to hospital for treatment.

Standardizing the intensive treatment phase of ALL in India

The Indian Collaborative Childhood Leukaemia (ICiCLe) study group was formed in 2013 to improve childhood ALL outcomes in India. The group involves five major pediatric cancer centers (Delhi, Chandigarh, Mumbai, Chennai, Kolkata) drawn from all four geographic zones in the country. The group developed a treatment protocol (‘ICiCLe-ALL-14’) for contemporary risk-stratified management of patients with newly-diagnosed ALL. The clinical research unit at TTCRC designed and coordinated the multicentre clinical study and analyzed study findings. TCS developed the study’s web-based data system for remote data capture, patient randomisation, study notifications and decision assists (patient enrolment, risk stratification). The ICiCLe-ALL-14 study (Clinical Trials Registry-India CTRI/2015/12/006434) [1] , the first randomized pediatric cancer clinical trial in India, confirmed the capability of Indian treatment centers to participate in multicentre clinical trials and record high quality data. Over the course of the study (pre-trial phase, 2013-2017; randomized trial, 2017-2022; cumulative, ~5000 patients), standardization of treatment through collaborative practice and contemporary risk stratification halved treatment related deaths (from 15-20% prior to ICiCLe-ALL-14 to 8% in the randomized trial) and improved survival rates significantly (from ~65% prior to ICiCLe-ALL-14 to ~72% in the trial cohort at a median 3 years). This strategy is very cost effective by WHO-CHOICE guidelines [2] .

Treatment of ALL: maintenance treatment phase

During the 2-year maintenance treatment (MT) phase of ALL, patients are prescribed the oral antimetabolites, daily 6-mercaptopurine (6MP) and weekly methotrexate (MTX), to be taken at home. These drugs work in combination. To ensure that patients receive the right doses of 6MP and MTX during MT, doses are increased serially (‘treatment to tolerance’) until no further dose increase is possible due to drug-related side effects (‘tolerance limit’). This tolerance limit varies from patient to patient due to host factors that influence drug metabolism and bone marrow function. Treatment to tolerance requires patients to visit cancer treatment centers frequently (at least once every 2 weeks) for blood count tests and dose prescriptions ( Figure 3.1-3.2 ).

Improving ALL outcomes by standardizing maintenance phase treatment

The MT phase requires steady escalation in the dose of oral drugs through the 2 years of treatment. Failure to treat patients to tolerance during MT increases the risk of ALL relapse [3] . In a cohort of 210 paediatric ALL patients treated at our center (diagnosed between 2013-2017), drug doses were not increased in 40% patients while in the other 60%, doses were increased late (at a median, 18 months) into treatment . Drug doses were frequently reduced (85% of dose reduce decisions) or stopped (36% of dose stop decisions) even though blood count values ​​did not support these dose decisions.We subsequently introduced measures (2018-) for electronic data capture of longitudinal information required for dosing decisions in ALL maintenance and strengthened supervision of the maintenance treatment phase to ensure adherence to protocol recommendations for timely drug dose increases. Review of the impact of these measures in 253 patients showed that weighted mean drug doses were prescribed closer to protocol-recommended doses: 93% and 95% for 6MP and MTX respectively compared to 82% and 81% for 6MP and MTX respectively in an earlier cohort (2013-2017, 163 patients). Significantly the increase in drug doses with these measures was not accompanied by increase in drug-related toxicity ( Figure 4.1-4.2 ).These observations indicate that standardizing ALL maintenance is feasible. This standardization requires access to longitudinal clinical records and adherence to protocol recommendations for drug dose adjustments to ensure treatment to tolerance.

Developing an assisted dose advice system to standardize ALL maintenance treatment

Two factors appear to contribute to sub-optimal drug dosing during ALL maintenance: (a) restricted access to longitudinal records of blood counts and drug doses when making dose decisions and (b) physician inexperience, and consequent diffidence in increasing drug doses serially. An automated assisted dose advice system in ALL maintenance (ADAM) has the potential to address these limitations [4] . The system allows capture of longitudinal data required for drug dose adjustments in individual patients and uses inbuilt rules for drug dose adjustments to assist physicians of varying levels of experience with dosing advice during ALL maintenance. Implemented as a web-based system, ADAM enables remote prescribing, precluding the need for patients to travel frequently to treatment centers for dosing advice.

Rationale and objectives

rationale

Experience with the ICiCLe-ALL-14 clinical trial indicates that standardization of ALL treatment in line with contemporary international practice improves ALL outcomes. Additional standardization of treatment during the ALL maintenance phase is expected to further advance ALL cure rates in India (Figure 5). An important element of treatment standardization during ALL maintenance is to treat with the highest tolerated doses of oral anti-cancer drugs. This requires serial increase in drug doses over the course of ALL maintenance treatment until the limit of tolerance. Our observations indicate that this dose increase principle is practiced poorly in the clinic.We postulate that this is a consequence of limited availability of longitudinal clinical data required for drug dose decisions, coupled with physician inexperience. ( Figure 6.1-6.3 )

1. Objective 1
   Design, develop and test an assisted dose advice system for dose adjustment decisions in ALL Maintenance (ADAM) [September 2022-March 2023]

2. Objective 2
   Systematically evaluate the performance of ADAM in terms of accuracy of dose adjustment predictions, discordance with physician dose prescriptions & reasons for dose advice discordance, and ease of use [April-September 2023]

3. Objective 3
   Introduce and evaluate refinements in ADAM based on findings from Objective 2 [October-December 2023]. Electronic prescribing will be introduced by integrating the system securely with the hospital’s electronic medical records.

4. Objective 4
   Implement as a standalone system to assist with dose advice decisions in ALL maintenance at the Tata Medical Center Kolkata. Continue audit of performance [January 2024].

Design of the assisted dose advice system in ALL Maintenance (ADAM)

ADAM is developed using the Python Django programming language to develop decision assist modules, supported on a PostgresSQL database ( Figures 7.1-7.6 [ Figures 7A-7B ]). It also uses multiple python modules for the graphical plots. Following log-in, the clinical user interface includes tabs to enter the date of prescription and blood counts for each patient visit; functionality to upload patient historical data for patients already into maintenance treatment; display information from the last prescription date (blood counts and drug doses) and highlight (where available) the highest tolerated doses of drugs;options to provide physicians with information on serial blood counts and drug doses by displaying the information both in tabular format and as longitudinal line plots; and a button that when clicked, engages the decision assist module to recommend drug doses and distribution of the recommended drug dose over the course of a week.

The decision assist module incorporates pre-specified rules for drug dose adjustments, including decisions to start drug treatment; continue treatment at previous doses; interrupt drug treatment (either reduce drug doses or stop drug treatment); resume drug treatment (either at previously prescribed doses, at previous highest tolerated doses, or at doses different from these two dose levels); and increase drug doses (including the magnitude of dose increase). Doses are rounded off to the nearest available tablet strength of the oral drugs. Clinical users can accept or disagree with the drug dose recommendations; a ‘comment’ section records the user’s reason for disagreement.A downloadable output tabulates ADAM’s serial dose recommendations, the corresponding clinical decisions,

ADAM: Development and Testing

Plans for developing ADAM were initiated in July 2021 and the first prototype was developed by TCS in January 2022. The clinical user interface was subsequently enhanced, and the decision assist module updated to incorporate more recent dosing rules (February to August 2022). A two-stage user acceptance testing is underway with the updated version. In the first stage (September to November 2022), simulated case scenarios were used to evaluate the decision assist module. The second stage of testing is in progress at the ALL maintenance clinic (December 2022-March 2023) to evaluate performance of the user interface and decision assist module in a real-world setting. ADAM is deployed in the TTCRC server and is not currently linked to the hospital’s information systems.To ensure privacy, the system uses patient hospital registration numbers alone.

References

1. Das N, Banavali S, Bakhshi S, Trehan A, Radhakrishnan V, Seth R, Arora B, Narula G, Sinha S, Roy P, Gogoi MP, Chatterjee S, Abraham B, Das P, Saha V, Krishnan S. Protocol for ICiCLe-ALL-14 (InPOG-ALL-15-01): a prospective, risk stratified, randomised, multicentre, open label, controlled therapeutic trial for newly diagnosed childhood acute lymphoblastic leukemia in India. Trials. 2022;23:102
2. Mungle, T, Das N, Pal S, Gogoi MP, Das P, Ghara N, Ghosh D, Arora RS, Bhakta N, Saha V, Krishnan S. Comparative treatment costs of risk‐stratified therapy for childhood acute lymphoblastic leukemia in India. Cancer Medicine. 2022. https://doi.org/10.1002/cam4.5140
3. Toksvang LN, Lee SHR, Yang JJ, Schmiegelow K. Maintenance therapy for acute lymphoblastic leukemia: basic science and clinical translations. Leukemia. 2022;36:1749-1758
4. Mungle T, Gogoi MP, Mitra S, et al. Developing an automated dose advice program to assist adaptive antimetabolite dose decisions during maintenance therapy in acute lymphoblastic leukaemia. Pediatric Hematology Oncology Journal. 2020; 5:p.S10

The purpose of the ADAM initiative is to  optimize and standardize treatment during the maintenance treatment phase of acute lymphoblastic leukemia (ALL). Treatment during this phase requires practitioners to modify drug doses periodically through the 96 weeks (~2 years) of treatment to ensure that patients are treated with the highest tolerated doses of the anti-cancer antimetabolite drugs 6-mercaptopurine and methotrexate. Our findings indicate that practitioners comply poorly with dose modification rules during ALL maintenance. Suboptimal drug treatment during ALL maintenance has been consistently shown to increase risk of ALL relapse. Our observations highlight three key factors for suboptimal treatment in ALL maintenance:(a) limited access to longitudinal records of historical blood counts and drug doses required for dose modification decisions in patients (b) insufficient and variable practitioner experience with drug dose modification decisions in ALL maintenance (c) inadequate frequency of supervision during ALL maintenance, since monitoring during this phase requires outpatient visits to treatment centers at least once every two weeks for blood count checks and dose advice prescriptions. A web-enabled automated assisted dose advice system (ADAM) has the potential to address all three limitations by enabling systematic data capture of blood counts and drug doses during this treatment phase, assisting practitioners with timely and appropriate dose modification decisions in individual patients,and supporting remote clinical supervision during the maintenance treatment phase. By ensuring optimal drug treatment and enabling more rigorous supervision of ALL maintenance through remote monitoring, the ADAM initiative has the potential to standardize treatment during the ALL maintenance phase across treatment centres, translating directly to improved ALL cure rates.

The Covid-19 pandemic compelled treatment centers to introduce remote monitoring of ALL maintenance, using e-mail and newly-built apps to exchange information on blood counts and drug dose advice. There is at present no system similar to ADAM that is designed to support and standardize dose adjustment decisions and practice in ALL maintenance. The availability of longitudinal clinical data for dose adjustment decisions and ADAM’s decision assist module are expected to ensure timely iterative increases in maintenance drug doses. The decision assist module will enable clinical personnel of varying experience to prescribe treatment in ALL maintenance, including junior physicians and nurse practitioners.This was illustrated by LISA (Leukemia Intervention Scheduling & Advice), a proof-of-concept web-based decision support system developed similarly for drug dose decisions in ALL maintenance in the UK (2005) [1] . As a web-based system, ADAM will enable remote drug dosing of patients in ALL maintenance, decreasing the need for patients to travel to treatment centers for dose prescriptions. The availability of remote dosing via ADAM will also encourage more rigorous supervision and monitoring of the ALL maintenance treatment phase. ADAM may be available as a standalone web-based system or may be integrated with multicentre clinical trial databases for the treatment of ALL, enabling standardization of ALL maintenance across the network of study centres.Standardized treatment in ALL maintenance is expected to improve ALL cure rates. Although developed to optimize ALL maintenance treatment in Indian patients, the ADAM initiative and its potential benefits are applicable for patients with ALL across the globe.

Reference

1. Bury J, Hurt C, Roy A, Cheesman L, Bradburn M, Cross S, Fox J, Saha V. LISA: a web-based decision-support system for trial management of childhood acute lymphoblastic leukaemia. Br J Haematol. 2005;129:746-754