Using zanamivir trial data to evaluate baseline characteristics and disease progression by various subgroups to inform clinical development of novel therapies in hospitalized patients with influenza

Proposal
1934

Title of Proposed Research

Lead Researcher

Lesley Butler

Affiliation

Genentech, Inc., A member of the Roche Group

Funding Source

The proposed research study will be internally funded by Roche/Genentech per FTE support for the Research Team (see Section B).

Potential Conflicts of Interest

An employee of Roche/Genentech and involved in the clinical development of a novel monoclonal antibody for treatment of severe influenza.

Data Sharing Agreement Date

29 August 2017

Lay Summary

Illness due to severe influenza represents a tremendous burden to patients and health care systems worldwide. This is due in part to the relatively high hospitalization rates, as well as the morbidity and mortality associated with severe influenza. The lack of novel therapies being developed for severe influenza is partly due to challenges, such as a heterogeneous patient population and the lack of a validated endpoint (EMA Concept Paper, 2017; FDA Guidance, 2011). The proposed research study will diminish barriers to designing effective and safe novel therapies for patients with severe influenza by using the zanamivir trial data to address these challenges.

Severe influenza causes 140,000 to 710,000 hospitalizations per year, with an average annual rate of 95.9/100,000 (Rolfes, 2016; Thompson, 2004). Mortality exceeds 12% amongst patients that are admitted to the ICU. Worldwide, the number of deaths due to severe influenza range from 250,000 to 500,000 annually (WHO, Influenza Factsheet, 2016). The proposed research study will inform the design of clinical trials to test the efficacy of novel treatments in patients hospitalized with influenza.

The research objectives of the proposed study are:
1. To identify subgroups of patients meeting selected demographic and clinical characteristics at baseline.
2. To quantify proportions of patients from identified subgroups meeting clinically defined disease-related outcomes.

The proposed research will be led by a cross-functional team at Roche-Genentech. The project lead and statistician will review interim results on a weekly basis, and the team will meet via teleconference on a bimonthly basis to communicate interim results, align on interpretation and discuss next steps. Once the requested zanamivir trial data are made available for analyses, the entire project period from data acquisition to submitting a manuscript for peer review will be accomplished within 18 months.

The focus of this research study is on descriptive analyses of the GSK-114373 phase 3 trial data to evaluate the extent of patient heterogeneity by indicators of disease severity and its impact on clinical outcomes. The main analyses will be done by combining the two zanamivir treatment arms and the standard-of-care oseltamivir arm. The rationale for this approach is that intravenous zanamivir was not found to be superior to oseltamivir, as reported by Marty, et al. (2017).

Findings from descriptive analyses will be interpreted based on overall distributions and trends, the direction of difference between subgroups, and the degree of overlap of the 95% confidence intervals between the subgroups being compared. Where statistical tests are performed, P values less than 0.05 will be interpreted as statistically significant. The findings from this research project will be communicated through forums, such as infectious disease- or epidemiology-focused congresses, as well as in research publications.

Study Data Provided

[{ "PostingID": 4740, "Title": "GSK-NAI114373", "Description": "A Phase III international, randomized, double-blind, double-dummy study to evaluate the efficacy and safety of 300 mg or 600 mg of intravenous zanamivir twice daily compared to 75 mg of oral oseltamivir twice daily in the treatment of hospitalized adults and adolescents with influenza" }]

Statistical Analysis Plan

Overall ApproachAll analyses will be descriptive in nature. Summary statistics (e.g., mean, standard deviation, median, inter-quartile range and percentage) and graphical displays will be made, as appropriate and confidence intervals will be calculated if data permits. No formal hypothesis testing will be conducted in this study. In addition, as indicated in Section E.1, we are requesting information pertaining to post-hoc analyses that were performed by the GSK statistical team to investigate a potential ordinal scale endpoint. The statistical methods and derived variables used in the post-hoc analyses will be explored in this proposed project to evaluate potential differences in clinical outcomes in patient subgroups defined by severity of disease. Sensitivity and Subgroup AnalysesSensitivity analyses will be performed for findings of interest among the ITT population and the oseltamivir-only arm. Subgroup analyses will be performed across groups defined by severity of disease and by country and/or geographic region. The following factors will be considered in defining severity-of-disease subgroups: • age• tobacco use• chronic illness at baseline• chronic supplemental oxygen use• baseline oxygen use • chest x-ray showing presence of infiltrates (as an indicator of bacterial co-infection)• time to hospital admission from symptom onset• admitting ward (ICU or non-ICU)The geographic region subgroups will be defined as: • US/Canada• Mexico/Brazil• Australia/New Zealand/South Africa• Western Europe• Eastern Europe• East Asia/India. Statistical Power CalculationsNo formal hypothesis testing will be conducted in this study. Thus, statistical power calculations were not performed. Handling of Missing DataImputation for missing data will not be performed.Statistical Analyses Continuous and categorical variables will be summarized for all patients and by subgroups using descriptive statistics, as appropriate (e.g., mean, standard deviation, median, inter-quartile range, percentage, 95% confidence intervals). Graphical display of the data will be used to assess trends (e.g., box plots, scatter plots, density plots). Analysis of Covariance methods (after appropriate transformation of data, if needed) will be used to estimate subgroup differences and 95% confidence intervals. For binary variables, subgroup estimates will be assessed using stratum-adjusted Mantel-Haenszel methodology, if data permits, and 95% confidence intervals will be presented (Koch, et al., 1989).Time to event variables will be analyzed using Kaplan-Meier methodology (if data permits), stratified Cox proportional hazards model, and summaries of n, median when estimable, hazard ratios, and 95% confidence intervals will be presented. The following covariates will be considered for inclusion in PH models to adjust for potential confounding: age, sex, ethnic origin, and study site country. Patients who are lost to follow-up (while event free) will be censored at the time that they are last known to be event free.The proportional odds model will be used to compare disease severity subgroups for ordinal outcomes. To supplement the overall summary odds ratio, separate odds ratios will be estimated for each dichotomized definition of improvement that can be formulated from the components of the ordinal outcome (e.g. ICU not on oxygen vs. ICU on any non-invasive oxygen). A test for the proportionality assumption will also be made. The following covariates will be considered for inclusion in logistic regression models to adjust for potential confounding: age, sex, ethnic origin and study site country. As defined in section A.5., we are proposing the following definition for clinical status categories, but understand that modification may be necessary depending on the data available:• Death (all-cause and influenza related)• ICU on mechanical ventilation• ICU on any non-invasive oxygen• ICU not on oxygen• General ward on any non-invasive ventilation• General ward not on oxygen• Discharged

Publication Citation

https://clinicalstudydatarequest.com/Documents/All%20Results_19March2018.xlsx