Mathematical prognostic model utilizing multi-platform dynamics in patients with metastatic urothelial carcinoma receiving post-platinum PD1/L1 inhibitors

Lead Investigator: Guru Sonpavde, AdventHealth Cancer Institute
Title of Proposal Research: Mathematical prognostic model utilizing multi-platform dynamics in patients with metastatic urothelial carcinoma receiving post-platinum PD1/L1 inhibitors
Vivli Data Request: 7601
Funding Source: None
Potential Conflicts of Interest: My conflicts and Disclosures in the past 36 months are as follows, but do not impact my ability to conduct this project in my judgment: Advisory Board: BMS, Genentech, EMD Serono, Merck, Sanofi, Seattle Genetics/Astellas, Astrazeneca, Exelixis, Janssen, Bicycle Therapeutics, Pfizer, Gilead, Scholar Rock, G1 Therapeutics, Eli Lilly/Loxo Oncology, Infinity Pharmaceuticals. Research Support to Institution: Sanofi, Astrazeneca, Gilead, QED, Predicine, BMS, EMD Serono. Steering committee of studies: BMS, Bavarian Nordic, Seattle Genetics, QED, G1 Therapeutics (all unpaid), and Astrazeneca, EMD Serono, Debiopharm (paid). Data safety monitoring committee: Mereo. Employment: Spouse employed by Myriad. Travel costs: BMS, Astrazeneca. Writing/Editor fees: Uptodate, Editor of Elsevier Practice Update Bladder Cancer Center of Excellence. Speaking fees: Physicians Education Resource (PER), Onclive, Research to Practice, Medscape, Cancer Network, Masters Lecture Series (MLS). We will manage the potential Conflicts of Interest (COI) in our proposal. The organisations/individuals involved in funding other projects above are not connected to this specific project and will have no involvement in research development, analysis or interpretation in this project. All the potential conflicts are disclosed to the institution and on publications and presentations as required by institutions.

Summary of the Proposed Research:

Bladder cancer affects ~80,000 individuals every year in the USA. Approximately 1 in 4 patients have aggressive cancer that is not always curable. Patients with bladder cancer that has spread outside the bladder (metastatic or stage-4 bladder cancer) receiving immunotherapy have different responses and outcomes. Only 1 in 5 patients have shrinkage of the cancer, while the remaining have no change or growth of cancer on immunotherapy. Patients who receive chemotherapy typically have more brief benefits compared to immunotherapy. A large number of patients are not fit for further treatment when the cancer grows. Hence, it would be useful to be able to discriminate at an early time point between those who have growth and those who have benefit for a long time on immunotherapy. A user-friendly model that discriminates outcomes of these patients with high accuracy can be exploited to improve therapeutic strategies. For those predicted to have poor long-term outcomes on immunotherapy, earlier switch to other new treatments or intensification of therapy by combinations of new agents with immunotherapy may improve outcomes. This study proposes to construct a mathematical model utilizing early changes in readily available clinical (e.g. organs of spread of cancer, fitness of patient, age, weight, gender, side effects) and laboratory variables (blood cell counts, blood albumin, cancer size measurement). We will analyze 931 patients with metastatic bladder cancer who were enrolled in the IMvigor211 trial, which was a phase 3 clinical trial comparing atezolizumab immunotherapy versus chemotherapy in those who had cancer growing after previous chemotherapy.

Statistical Analysis Plan:

Both arms of IMvigor211 phase III trial comparing post-platinum atezolizumab vs historical chemotherapy (taxane or vinflunine) will be studied retrospectively. The control arm of chemotherapy will provide a reference control when examining the atezolizumab arm. A mathematical model of dynamic changes in multiple variables in response to post-platinum PD1/L1 inhibitors will be developed to predict time to progression (TTP). Sensitivity and correlation analysis will be used to identify parameters to be used for model calibration. The data will be divided into training and testing cohorts and the model will be calibrated to and validated against the training and testing cohorts, respectively. Model parameters will be analyzed to determine appropriate simulated dynamics of partial response versus stable disease. Changes in predicted dynamics will be used to predict time to progression and flag high-risk patients. Youden’s J Index will be used to assess the model’s predictive power. The positive predictive value (PPV) and negative predictive value (NPV) will also be used to assess clinical utility. After the model has been proven to be predictive, it will be used to investigate alternative therapeutic options in those patients predicted to have suboptimal response to PD1/L1 inhibitors alone. The model will also be used to propose treatment escalation or de-escalation in those patients predicted to have a prolonged or brief TTP. We will develop and calibrate a non-parametric mixed effects tumor growth inhibition (TGI) ordinary differential equations (ODE) model, describing treatment sensitivity at treatment initiation (γ_0) and pre-treatment growth rate (λ) as fixed (uniform) effects and sum of longest diameters (SLD) at treatment initiation (SLD_0) and treatment efficacy decay rate (ε) as random (patient-specific) effects. This model can replicate tumor dynamics with short-, medium-, and long-term outcomes. We will then perform a receiver operator characteristic (ROC) analysis to assess the prognostic value random effects on risk level. These analyses can be repeated for all of the other selected variables under the design section, i.e. performance status (PS), sites of metastasis status, neutrophil/lymphocyte ratio (NLR), platelet count, hemoglobin, albumin, immune adverse events, and weight. Missing values will be excluded.

Requested Studies:

A Phase III, Open-Label, Multicenter, Randomized Study to Investigate the Efficacy and Safety of Atezolizumab (Anti-PD-L1 Antibody) Compared With Chemotherapy in Patients With Locally Advanced or Metastatic Urothelial Bladder Cancer After Failure With Platinum-Containing Chemotherapy
Data Contributor: Roche
Study ID: NCT02302807
Sponsor ID: GO29294

Public Disclosures:

Graser, C., McDonald, T.O., Sonpavde, G. and Michor, F., 2024. Use of early dynamics of clinical and laboratory parameters to predict resistance in patients with metastatic urothelial carcinoma receiving post-platinum atezolizumab. Cancer Research, 84(6_Supplement), pp.2392-2392. Doi : 10.1158/1538-7445.AM2024-2392