Lead Investigator: Ashley Hopkins, Flinders University
Title of Proposal Research: Predicting therapeutic and adverse outcomes to immune checkpoint inhibitors used in cancer treatment
Vivli Data Request: 7595
Funding Source: National Health and Medical Research Council (NHMRC) of Australia, Flinders University
Potential Conflicts of Interest: None
Summary of the Proposed Research:
Immune checkpoint inhibitors (anticancer medicines that aim to help your immune system fight cancer) are important emerging treatments option for cancer which are resulting in significant benefit for many patients. However, response and toxicity to immune checkpoint inhibitors is highly unpredictable. For example, up to 40% of patients who initiate immune checkpoint inhibitors do not respond, while 40% may experience serious toxicity.
Using the diverse range of data collected from clinical trials, it is possible to unlock clinical predictions that enable an improved understanding of the therapeutic and adverse outcomes from medicines. Specifically, this program will use data science techniques to assess the association between available clinicopathological data with therapeutic and adverse events outcomes from immune checkpoint inhibitors. Evaluated predictors will be prioritised according to biological/clinical plausibility and prior evidence. This includes emerging evidence that outcomes following immune checkpoint inhibitor initiation are based on an individual’s demographic, clinical, laboratory, disease, and genetic characteristics. Using this data from a diverse range of patients may enable expected response and adverse effect profiles to be predicted, which will help facilitate patients and clinicians to make better decisions regarding whether to commence certain immune checkpoint inhibitor treatments.
Statistical Analysis Plan:
Study:
Our research team has systematically collated information on clinical trials evaluating immune checkpoint inhibitor therapies eligible for sharing on the Vivli platform. At this stage available data largely includes clinical trial data investigating the immune checkpoint atezolizumab – although we hope to be able to expand this in the future. This study will endeavour to gather the individual participant data from these clinical trials which have been indicated to the team as within scope for sharing with the purpose to identify predictors of therapeutic and adverse outcomes in patients. This study will utilise time-to-event and penalised regression techniques to identify, rank and validate the strongest predictors of therapeutic and adverse outcomes in the appropriate patient cohorts described below.
Software:
The R Software (R Core Team) will be used for data preparation, modelling and graphical output.
Outcomes:
Data are required for the outcomes including response (early, depth, best overall), overall survival, progression-free survival, and adverse event outcomes (clinician / patient reported adverse effects that have been defined according to the international common toxicity criteria, and adverse events requiring medication changes).
Predictors:
Predictors of adverse event and therapeutic outcomes will be screened according to biological and clinical plausibility and empirical evidence based on prior research – this project will be seeking to identify and publish research on predictors of therapeutic and adverse outcomes to immune checkpoint inhibitor therapies used in the treatment of various cancer types. As most of the data collected within a clinical trial contains some information on the immune system, disease severity and prognosis, toxicity risk or drug exposure, it is important to have access to all the baseline and follow-up clinical/biological/patient characteristic data collected on an individual during a trial (i.e., as the purpose of this study is to optimise prediction of adverse events or therapeutic response via the use of collected clinicopathological data, we need to be able to access the collected clinicopathological data). Covariates to be investigated include:
• Basic patient characteristics – e.g., age, sex, race / ethnicity, body mass measuers, weight, weight loss prior to diagnosis and therapy initiation, smoking status, alcohol consumption, family history of disorders, and measures of performance status.
• Laboratory data – e.g. levels of lactate dehydrogenase, alkaline phosphatase, albumin, bilirubin, leucocyte and leucocyte subtype counts (e.g. total white blood cells (WBC), lymphocyte, monocyte or eosinophils, neutrophil to lymphocyte ration (NLR), lymphocyte to monocyte ratio (LMR), platelets to lymphocyte ration (PLR)), haemoglobin, platelets, glucose, hemoglobin A1c (HBA1C), creatinine, C-reactive protein, circulating tumour cells, calcium, total protein, total triglycerides, cholesterol, blood urea nitrogen, international normalized ratio, and anaemia.
• Disease classification data – e.g. tumour stage and grade, site and histology/subtype of primary tumour, prior therapy, prior surgery, time to response / progression for previous therapies, sum of longest tumour diameter, time since diagnosis, number and sites of metastases, mutation and expression status of disease specific oncogenes (e.g. programmed death protein 1 (PD-1) / programmed death-ligand 1 (PD-L1)), line of therapy, pathological features of tumour cells, tumour infiltrating lymphocytes, number of nodes with tumour, cancer antigen level.
• Other common predictors – e.g. concomitant medications (e.g. antibiotics, proton pump inhibitors, corticosteroids, etc), respiratory comorbidity (e.g. asthma / chronic obstructive pulmonary disease (COPD)), other comorbid diseases (e.g. peripheral vascular disease, cerebrovascular disease, diabetes, Hepatitis C infection), simplified comorbidity score, organ dysfunction (e.g. liver, lung or renal impairment), and other clinical, biological, vital statistics, laboratory, imaging, pharmacokinetic and patient-reported outcomes measures.
General model development:
Time-to-event models will be used to assess the association between potential predictors and the time to an adverse effect or survival time. Associations will be reported primarily as hazard/risk ratios with 95% confidence intervals. The association of potential predictors with binary outcomes (e.g., best overall response) will be modelled using logistic regression and will be reported as odds ratios with 95% confidence intervals. Longitudinal analysis (e.g., linear, and non-linear mixed effect modelling) will be used to assess the nature and patterns of longitudinal changes of key continuous variables (e.g., drug concentration, tumour size, neutrophil counts).
Crude associations will be reported based on univariable analysis (albeit, as appropriate, univariable analyses will still be adjusted/stratified by clinical trial, cancer medicine, and trial arms), and adjusted associations based on a multivariable analysis (e.g., adjusting for cancer type, stage, performance status, age, sex, race, line of therapy, presence of liver metastases and programmed death-ligand 1 (PD-L1) status). Continuous variables will be assessed for non-linearity of association with outcomes using restricted cubic splines, STEPP, and fractional polynomial techniques. Should multiple values of an assessed covariate be recorded for a single visit (e.g. blood pressure) the mean of the multiple reads will be used. Where appropriate, the varying performance of clinical prediction models developed using multivariable analysis techniques including stepwise regressions, penalised methods [which minimise the risk of overfitting (e.g., elastic net analysis)], and machine learning [which excel in optimising prediction with high-dimensional data (e.g., random forest and gradient boosted methods)] will be assessed. Early markers of exposure, response and toxicity will be evaluated using landmark approaches, with sensitivity analyses based on the use of time-dependent covariates. Landmark times will be dependent on the time points available in individual studies, and the time frame of changes in each specific predictor variable. As our analyses are primarily hypothesis generating and they will require subsequent validation, no formal adjustment for multiple testing is intended. However, this limitation will be clearly stated in any publications of results. As it is expected that < 5% of data will be missing for assessed variables, complete case analyses are planned. Should variables with substantial missing data be present, the pattern and likely cause of the missing data will be evaluated and if missing at random is reasonable to assume then single regression imputation will be undertaken.
Predictors of adverse event and therapeutic outcomes will be screened according to biological/ clinical plausibility and empirical evidence based on prior research. Where informed by plausibility and the literature, analyses will evaluate potential predictors according to specific immune checkpoint inhibitor drugs within specific cancer types. Where appropriate, generalisability across class and cancer types will be assessed in further work. Any analyses using data from multiple clinical trials/arms will be adjusted/stratified by these discriminators. Where appropriate two-stage approaches will also be undertaken to further assess for consistency between trials/arms. Analyses will include evaluating the heterogeneity in outcome predictors and risk profiles for immune checkpoint inhibitors as compared to relevant comparators. Such evaluations will allow a better understanding of whether the relationships identified are specific to immune checkpoint inhibitors or are common to standard treatment (i.e., common prognostic factors).
Power:
Predictors that have a clinically meaningful (e.g., double the risk) effect on mortality and adverse effects will be of primary interest. Based upon an event rate of 40% during trial follow-up (e.g., for progression), approximately 450 participants are required for 80% power to detect a predictor (with a 10% frequency within the population) associated with a two-fold hazard of the event (α=0.05). Reference: Chow S, Shao J, Wang H. 2008. Sample Size Calculations in Clinical Research. 2nd Ed. Chapman & Hall/CRC Biostatistics Series. page 177. These samples sizes are well within scope for this study. Sample sizes greater than this will allow the investigation of more complex relationships with greater predictive performance.
Quality Control:
A major objective of the project is to deliver a coding infrastructure which facilitates the formation of a standardised database which enables the identification and validation of predictors of therapeutic and adverse outcomes to immune checkpoint inhibitor therapies – which will be of benefit to future research as new clinical trial data on immune checkpoint inhibitors emerge. Considerable effort will be put into exploring inconsistencies in time recordings, physiologically unreasonable covariate values, and unit errors. Prior to beginning analyses, individual data values will be extracted/constructed based on the raw and analysis datasets provided. To ensure that each variable has been correctly extracted/constructed from the data provided, basic analyses and descriptive statistics will be reproduced to check for consistency with pertinent results in published manuscripts or clinical study reports (CSRs) relating to the specific trial. Where there are insufficient published results to confirm the proper extraction of the variable, the extracted values will be manually checked against a random sample of the original dataset values.
Countries where analysis will be conducted
Requested Studies:
A Phase III, Open-Label, Randomized Study of Atezolizumab in Combination With Bevacizumab Compared With Sorafenib in Patients With Untreated Locally Advanced or Metastatic Hepatocellular Carcinoma
Data Contributor: Roche
Study ID: NCT03434379
Sponsor ID: YO40245
A Phase II, Multicenter, Single-Arm Study of Atezolizumab in Patients With Locally Advanced or Metastatic Urothelial Bladder Cancer
Data Contributor: Roche
Study ID: NCT02951767
Sponsor ID: GO29293 (Cohort 1)
A Randomized, Multicenter, Double-Blind, Placebo-Controlled Phase II Study of the Efficacy and Safety of Trastuzumab Emtansine in Combination With Atezolizumab or Atezolizumab-Placebo in Patients With HER2-Positive Locally Advanced or Metastatic Breast Cancer Who Have Received Prior Trastuzumab and Taxane Based Therapy
Data Contributor: Roche
Study ID: NCT02924883
Sponsor ID: WO30085
A Phase III, Double-Blinded, Randomized, Placebo-Controlled Study of Atezolizumab Plus Cobimetinib and Vemurafenib Versus Placebo Plus Cobimetinib and Vemurafenib in Previously Untreated BRAFV600 Mutation-Positive Patients With Unresectable Locally Advanced or Metastatic Melanoma
Data Contributor: Roche
Study ID: NCT02908672
Sponsor ID: CO39262
A Phase I/III, Randomized, Double-Blind, Placebo-Controlled Study of Carboplatin Plus Etoposide With or Without Atezolizumab (Anti-PD-L1 Antibody) in Patients With Untreated Extensive-Stage Small Cell Lung Cancer
Data Contributor: Roche
Study ID: NCT02763579
Sponsor ID: GO30081
A Phase III, Open-Label, Randomized Study of Atezolizumab (Anti-PD-L1 Antibody) in Combination With Bevacizumab Versus Sunitinib in Patients With Untreated Advanced Renal Cell Carcinoma
Data Contributor: Roche
Study ID: NCT02420821
Sponsor ID: WO29637
A Phase III, Open-Label, Multicenter, Randomized Study Evaluating the Efficacy and Safety of Atezolizumab (MPDL3280A, Anti-PD-L1 Antibody) in Combination With Carboplatin+Paclitaxel or Atezolizumab in Combination With Carboplatin+Nab-Paclitaxel Versus Carboplatin+Nab-Paclitaxel in Chemotherapy-Naive Patients With Stage IV Squamous Non-Small Cell Lung Cancer
Data Contributor: Roche
Study ID: NCT02367794
Sponsor ID: GO29437
A Phase III Multicenter, Randomized, Open-Label Study Evaluating the Efficacy and Safety of Atezolizumab (MPDL3280A, Anti-PD-L1 Antibody) in Combination With Carboplatin+Nab-Paclitaxel for Chemotherapy-Naive Patients With Stage IV Non-Squamous Non-Small Cell Lung Cancer
Data Contributor: Roche
Study ID: NCT02367781
Sponsor ID: GO29537
A Phase III, Open-Label, Randomized Study of Atezolizumab (MPDL3280A, Anti-PD-L1 Antibody) in Combination With Carboplatin+Paclitaxel With or Without Bevacizumab Compared With Carboplatin + Paclitaxel + Bevacizumab in Chemotherapy-Naïve Patients With Stage IV Non-Squamous Non-Small Cell Lung Cancer
Data Contributor: Roche
Data Request ID: 00007595
Sponsor ID: GO29436
A Multistage, Phase II Study Evaluating the Safety and Efficacy of Cobimetinib Plus Paclitaxel, Cobimetinib Plus Atezolizumab Plus Paclitaxel, or Cobimetinib Plus Atezolizumab Plus Nab-Paclitaxel as First-Line Treatment for Patients With Metastatic Triple-Negative Breast Cancer
Data Contributor: Roche
Study ID: NCT02322814
Sponsor ID: WO29479
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
A Phase II, Multicenter, Single-Arm Study of Atezolizumab in Patients With Locally Advanced or Metastatic Urothelial Bladder Cancer
Data Contributor: Roche
Study ID: NCT02108652
Sponsor ID: GO29293 (Cohort 2)
A Phase II, Multicenter, Single-Arm Study OF Atezolizumab In Patients With PD-L1-Positive Locally Advanced Or Metastatic Non-Small Cell Lung Cancer
Data Contributor: Roche
Study ID: NCT02031458
Sponsor ID: GO28754
A Phase III, Open-Label, Multicenter, Randomized Study to Investigate the Efficacy and Safety of Atezolizumab (Anti-PD-L1 Antibody) Compared With Docetaxel in Patients With Non-Small Cell Lung Cancer After Failure With Platinum Containing Chemotherapy
Data Contributor: Roche
Study ID: NCT02008227
Sponsor ID: GO28915
A Phase II, Randomized Study of Atezolizumab (Anti-PD-L1 Antibody) Administered as Monotherapy or in Combination With Bevacizumab Versus Sunitinib in Patients With Untreated Advanced Renal Cell Carcinoma
Data Contributor: Roche
Study ID: NCT01984242
Sponsor ID: WO29074
A Phase II, Open-label, Multicenter, Randomized Study to Investigate the Efficacy and Safety of MPDL3280A (Anti−PD-L1 Antibody) Compared With Docetaxel in Patients With Non−Small Cell Lung Cancer After Platinum Failure
Data Contributor: Roche
Study ID: NCT01903993
Sponsor ID: GO28753
A Phase II, Multicenter, Single-arm Study of MPDL3280A in Patients With PD-L1-Positive Locally Advanced or Metastatic Non-small Cell Lung Cancer
Data Contributor: Roche
Study ID: NCT01846416
Sponsor ID: GO28625
A Phase III, Open-Label, Randomized Study of Atezolizumab (MPDL3280A, Anti-Pd-L1 Antibody) in Combination With Carboplatin or Cisplatin + Pemetrexed Compared With Carboplatin or Cisplatin + Pemetrexed in Patients Who Are Chemotherapy-Naive and Have Stage IV Non-Squamous Non-Small Cell Lung Cancer
Data Contributor: Roche
Study ID: NCT02657434
Sponsor ID: GO29438
Public Disclosures:
- Parent N, Hopkins AM, Sorich MJ. Machine learning for enhanced survival prediction from tumour growth inhibition data. Population Approach Group of Australia and New Zealand (PAGANZ). Oral Presentation. 2024.
- Li LX, Hopkins AM, Sorich MJ. Methodologically appropriate evaluation of continuous BMI as a clinical predictor of chemoimmunotherapy efficacy in advanced non-small cell lung cancer. Population Approach Group of Australia and New Zealand (PAGANZ). Oral Presentation. 2024.