Masterclass: Best Statistical Practices for the Life Sciences

This 3-hour masterclass during which the key elements of the recommendations and regulations of the various government agencies (Health Canada, FDA, EMA, etc.) are presented with their concrete implications on the practices to be respected. Aimed at both management teams and scientific staff, the workshop allows stakeholders to be made aware of these issues. In itself, the masterclass is sufficient to allow startups to self-assess their practices and offer them tools to improve the points they consider problematic.

Course Outline

1. Introduction: Scientific & Economic Context

The Principles of Scientific Research and the Links with Statistical Tools

    • The experiment: Transitioning from the sample to the target population
    • Representativeness and reproducibility of studies and experiments
    • Analysis and interpretation of results: p-value, a scientific and economic holy grail to be demystified

The Constraints of Startups

    • Scientific principles put to the test of funding and deadlines
    • Regulatory constraints
    • A long-term vision: building the scientific history of society

Good Practices – 3 Guiding Principles

    • Rigorous test planning
    • A predictable, adapted and adequate statistical analysis of results
    • Effective reporting and communication of results

2. Best Practices for Planning Experiments & Collecting Data

Clarity of the research question: type of comparison, endpoint(s), factors, etc.

Use the knowledge already acquired to streamline each test

Adequate & Efficient Representation of Variability

    • Biological variation
    • Replication vs replicate

Control the Risk of Bias

    • Randomization
    • Blinding

Detection of Relevant Clinical/Biological Effects

    • Establish and justify the desired clinical effect
    • Calculate the chances of detecting the desired effect
    • Statistical power and sample size

Choosing a Scientifically & Economically Efficient Experimental Structure

3. Adequate Data Analysis

Preparing the Collected Data

    • Pre-processing and manipulation of raw data
    • The treatment of extreme values ​​“outliers”
    • Handling missing data
    • Impacts on power and risk of bias

The Choice of a Statistical Data Processing Method

    • Reflect the structure of experience
    • Verify the underlying conditions of use of statistical methods
    • Control and model sources of variability to maximize the chances of detecting important differences

Controlling Multiplicity

    • The basic principle: rationalize the analysis of results or risk reaching erroneous conclusions
    • Corrective measures
      • The statistical approach: effective but expensive tools
      • The scientific method: “primary endpoints”, proof of mastery of the research topic

Management of experimental hazards: when everything does not go as planned…

    • Adapt to unexpected events
    • Anticipate potential problems
    • Making the most of an experimental “failure”

4. Reporting of Studies & Results

The Research Protocol

    • A somewhat trying but rewarding task
    • A guarantee of quality for investors

The Data Analysis Report

    • A realistic assessment
    • Update the advancement of the company’s scientific knowledge

Scientific Communications

    • Highlight rigor to better convince
    • Aim for high-profile journals

Conclusion: Meeting the Challenges of Startups

    • Good statistical practices: binding rules or efficiency gains?
    • Find acceptable compromises
    • Some possible solutions

Course Duration

The duration of this masterclass is 3 hours.

Target Audience

The masterclass is aimed at both the management team and the scientific team, ideally both simultaneously to stimulate discussions on the subject. Aimed at enabling startups to assess the adequacy of their scientific approach with regulatory requirements, it can stand on its own or be supplemented by different interventions for further reflection.