Hardware Podcasts

How AI Can Help You Manage Risk

Global Medical Device Podcast · 26 May 2023 · 33m · Episode 319

What this episode covers

Tyler Foxworthy, Chief Scientist at Greenlight Guru, explains how his company applies Bayesian statistics and machine learning to transform medical device risk management from qualitative guesswork into rigorous, probabilistic assessment. He details how the platform mines FDA data and clinical literature to update risk probabilities, addressing the industry's struggle with false precision and the limitations of human intuition regarding rare events. Foxworthy argues that adopting these models, pioneered in nuclear and aerospace sectors, allows companies to right-size clinical sample sizes, reduce uncertainty, and accelerate time to market by providing statistically defensible confidence intervals.

Most relevant to

early-stage companies

Topics

Mentioned

Companies
Greenlight Guru, Resultant, Demand Jump, Vertex
Products
Greenlight Guru Academy
Also
ISO 14971, FDA, NASA

Related episodes from other shows

Other shows on the same ground, closest first.

  • Let's Combinate · 40m

    098 - MedTech vs. Aerospace Risk Management, AFAP vs. ALARP/ALARA, Quantifying Benefit, Risk Control Measures, and Risk Management Reports with Bijan Elahi

    Bijan Elahi, a Technical Fellow and Medtronic corporate Advisor on safety risk management, explains the fundamental differences between MedTech and Aerospace risk management, noting that MedTech balances risk against benefit while Aerospace focuses on elimination. He details the challenges of quantifying risk using ISO 14971, advocating for quantitative methods over qualitative guesswork to provide objective evidence of safety. Elahi also discusses the regulatory requirements of the EU MDR, specifically the 'as far as possible' (AFAP) standard, and how companies determine when risk reduction is sufficient without adversely affecting the benefit-risk ratio.

    Quality & yieldCertification & compliance
    7 chapters

  • Let's Combinate · 32m

    202 - Becoming an Expert in the Work Nobody Wants to Do/Design, Risk, and the Messy Middle of Combination Products w/Joey Frechin

    Joey Frechin, a Senior Technical Program Manager in med tech consulting, explains the complexities of aligning risk assessments for combination products, noting that while drug and device teams are starting to integrate their processes, significant methodological gaps remain. He details the P1/P2 probability scoring methodology used in ISO 14971 to distinguish between device failure and patient harm, and shares his career strategy of becoming an expert in 'unwanted' but high-value tasks like DFMEA and PFMEA to bridge engineering and quality teams. Frechin also discusses current industry trends, where companies often pivot from complex bespoke designs to simplified or off-the-shelf subsystems to meet clinical trial timelines, highlighting the critical importance of early sterilization and accelerated aging testing to avoid late-stage failures.

    Quality & yieldDesign for manufacturing & NPI
    5 chapters

  • Let's Combinate · 1h 12m

    014A - "The First P: Product" with Larry Mager

    Larry Mager, Principal and Founder at Mgmt-Ctrl, explains the 'Product' phase of his Predictive Quality Management (PQM) framework, arguing that companies must first ensure their marketed devices are safe and effective before remediating broader quality systems. He details how to use post-market surveillance data, specifically FDA MAUDE and upcoming EU EUDAMED databases, to identify safety signals and compare them against competitive products to establish the 'State of the Art' (SOTA) benefit of patient safety. Mager describes a workflow where statistical analytics identify concerning product groups, triggering post-market risk assessments and CAPAs to improve risk control measures, thereby aligning the product with regulatory expectations and preventing recalls.

    Quality & yieldSourcing & supply chain
    10 chapters

  • Let's Combinate · 32m

    107 - Distilling Decades of MedTech Data, Generative AI, and Doing Without Tribal Knowledge with Michelle Wu

    Michelle Wu, CEO and Founder of NyquistData, explains how her company uses generative AI to distill decades of MedTech data, removing the 'tribal knowledge' barrier for quality and regulatory professionals. She details the distinction between pre-trained models and fine-tuned algorithms, emphasizing that Nyquist’s data is validated and specific to FDA and global medical device regulations, unlike generic models like OpenAI. Wu discusses the challenge of changing organizational culture and overcoming skepticism in life sciences, highlighting features like the 'Device Explorer' and 'Data Writer' that democratize access to complex regulatory insights.

    AI & machine learningQuality & yield
    10 chapters

  • Let's Combinate · 1h 31m

    025 - “The Frying Pan to the Fire” with Edwin Bills

    Edwin Bills, a Risk Management Expert and former committee member for ISO 14971, ISO TR 24971, and AAMI TIR 105, explains the evolution of risk management standards in medical devices and combination products. He details the philosophical differences between ISO 14971 and ICH Q9, emphasizing that risk management must focus on patient safety rather than just regulatory compliance. Bills advises engineers to use qualitative methods early in development and highlights the importance of analyzing interactions between drug and device constituents in combination products. He also discusses the expanded requirements for post-market surveillance in the 2019 standard and the upcoming harmonization of ISO 14971 in Europe.

    Medical & health devicesCertification & compliance
    12 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by Global Medical Device Podcast. The episode itself belongs to its publisher.