Presentation is loading. Please wait.

Presentation is loading. Please wait.

Using Q-learning Method in Identify Optimal Treatment Regime

Similar presentations


Presentation on theme: "Using Q-learning Method in Identify Optimal Treatment Regime"— Presentation transcript:

1 Using Q-learning Method in Identify Optimal Treatment Regime
Haocheng Li Vincent Shen Hao Xu Sylvia Hu

2 Background Lung cancer is one of the most commonly diagnosed cancers and is a leading cause of cancer death worldwide Non-small-cell lung cancer (NSCLC) is the most common type of lung cancer Bevacizumab (Bev), sold under the trade name Avastin®, is a recombinant, humanized monoclonal antibody, which acts against vascular endothelial growth factor (VEGF) When used in combination with chemotherapy as first-line treatment in patients with advanced NSCLC, bevacizumab significantly improves patient’s survival outcomes

3 Line of Therapy in Cancer Treatment
Sequential systemic therapies are commonly used for advanced cancer Death Advanced NSCLC Diagnosed Start Treatment for 1L Therapy Disease Progression/ Toxicity Treatment Change to 2L Therapy to 3L Therapy

4 Induction and Post-Induction Phase
First-line treatment could also be divided into induction phase (IP) and maintenance phase (or post-induction phase, post-IP ) Post-IP is a continuation of one or more drugs used in the induction regimen …… Advanced NSCLC Diagnosed Start Treatment for IP in First-line Disease Progression/ Toxicity Treatment Change to post-IP in First-line to 2L Therapy No progression or Death

5 ARIES Observational Cohort Study
Avastin® Registry—Investigation of Effectiveness and Safety (ARIES) observational cohort study Patient cohort had bevacizumab (Bev) and chemotherapy as IP therapy Patient can be Bev exposure or non-Bev exposure as Post-IP therapy After disease progression, patient can be Bev exposure or non-Bev exposure as second-line therapy Potential confounding factors such as age, gender, smoking status, diabetes status, alk phosphatase level and albumin, are recorded Overall survival is set to be primary outcome in this study Later line of therapies depending on the response to the previous treatment and patient status

6 Patient flowchart and analysis cohorts
post-IP (n=1050) post-IP: Bev exposure (n=502) post-IP: Non-Bev exposure (n=548) Death or Censoring (n=102) Death or Censoring (n=164) Progression Progression 2L: Bev exposure (n=118) 2L: Non-Bev exposure (n=282) 2L: Bev exposure (n=75) 2L: Non-Bev exposure (n=309)

7 Methods Proposed methods G-estimation Dynamic weighted OLS Q-learning

8 Future Extension With the development of electronic medical record, the treatment history of cancer patients can be recorded in convenience. Instead of Bev exposure vs Non-Bev exposure selection, multiple treatment selections are available such as chemotherapy, immunotherapy and targeted therapy Instead of two treatment lines, patients may also get third, fourth or later lines in real world practice High dimensional biomarker information can be utilized for dynamic treatment decision

9 Doing now what patients need next


Download ppt "Using Q-learning Method in Identify Optimal Treatment Regime"

Similar presentations


Ads by Google