AI Impact on Drug Development in Clinical Trials:

• patient-first approach is important here: most patients enroll because most treatments have failed → not all diagnosed patients are elibible for trail participation (finding eligible patients is hard)
• most clinical processes last 9 years and 1.3B on average → major companies like Novartis attributed 15% drops in Q1'17 net income to a failed phase III drug intended to treat heart failure
  • biggest target area for our solution should surround smaller biopharmaceutical startups where they require one-hit wonders (limited cash and are dependent on the clinical trial)

• AI can help in finding clinical trials → matching right trial with right patient is time-consuming and challenging process for clinical study team and patient

• for oncology, “In fact, only 3 percent of cancer patients today are enrolled in clinical trials.” — WhiteHouse.gov, May 2018

  • roughly 80% of clinical trials fail to meet enrolment timelines and most Phase III clinical studies are terminated because of enrolment difficulties

  • 22l clinical studies in the US that are currently recruiting patients, heatmap below shows the location of 1000 active breast cancer studies

    

    • most patients get trials recommended → most databases fail patients and no backstream

  • aquisitoin strategies to get patient data are oncology-focused go back to collecting electronic medical records which are often difficult and not streamlined (several AI startups are working on this right now so not that much opportunity there)

    Flatiron Health tackled the interoperability problem by acquiring an oncology-focused electronic medical record (EMR) company Altos Solutions in 2014. At the time, Flatiron was selling its cloud analytics platform to healthcare and life science companies, and Altos’ EMR was being used by oncology institutions like Florida Cancer Specialists. The deal gave Flatiron direct access to raw patient data, instead of relying solely on access to third-party EMRs. Roche subsequently acquired Flatiron for $2B+ in 2018 to gain access to its real-world evidence — insights generated from EHRs, claims, and wearable sensors. Roche plans to use this data to improve its cancer treatment pipeline development.

  • having open-source pipelines and access to key datapoints as such could be a great way to market a solution → cloud-based analytics platform that connects to several oncology institutions and patient records to generate key insights and matchups on patients

    • for cancer specifically (breast-cancer), most eligibility terms are hard to actually see on the customer-side, no way to actually see whether eligibility terms are met by patients → in general, examination tests are difficult

  • medical adherence is difficult for digital phenotyping, especially for cancer → most of this is being worked on for visual data collection but BenchSci could form strategic partnerships where it can store and

  • examples of this involve Google's Verily where they map human health to research → tracking specific diseases is critical

  

  3 Types of Trials: Active, Completed, Terminated → focusing on terminated but perhaps also failed completes

  Possible failure points for oncology trials in clinicaltrials.gov include:

  • poor accrual, interim results, toxicity/adverse events, study agent unavailable, canceled by the sponsor, inadequate budget, logistics, trial no longer needed, principal investigator left, no reason given, or other
  • KEY FINDING FOR ALL ONCOLOGY TRIALS:
    • for Phase 3 terminations (failed but not due to some unavoidable circumstance) were primarily due to poor accrual of patients and members in study, prevented them from demonstrating efficacy

  Failure in Completed Trials:

  • more drug targets discovered in fight to treat diseases such as cancer is the hardest since producing right drugs for targets is key for providing the best treatment
    • most FDA issues with approval for oncology is at 97% failure rate at clinical trials for oncology with issues in drug efficacy and toxicity
    • most cancer drugs work via off-target effects, meaning that original targets in drug are nonessential for cancer cell survival → failure to understand the drug's Mechanism of Action
    • caused by:
      • Mischaracterization of target-specific inhibitors
      • Misidentification of biomarkers (unable to decipher drug's true target)
      • Interlaboratory variability between cell lines without reasonable cause
    • most of the previously thought target proteins are actually nonessential

  Off-target effects:

  • off-target effects common for small molecule drugs and expect minor side effects with most
    • off-target interactions are actually MOA through which drug works to block cancer progression
    • this is common problem that effect cancer trials
  • failure to recognize off-target effects causes mischaracterization of target-specific inhibitors, and likely contributes to low FDA approval

  True drug target:

  • lack of understanding of drug's MOA hinders efforts in uncovering biomarkers capable of predicting therapeutic responses, further highlighting importance of understanding drug-target interaction
  • protein interaction analysis is vital precilnical research necessity and access to cutting-edge technology like protein interaction analysis can help make bigger impact in clinical applications
    • need to use more accurate and stringent studies early on to confirm drug MOA apart from offset effects
    • Fluidic Analytics working on this as of now

  Challenges in Oncology Studies (Not points of failure but things to consider across global markets)

  1. Study Design:
     • dose escalation designs are based on safety and efficacy considerations and incorporated with overlapping dose groups
     • Phase 1 studies based on patients due to anticipated toxicities (why there is a typical high failure rate)
     • rate in Phase 1 to have placebo due to ethical considerations but some design random placebo insertions
     • biggest challenge across all phases is keeping recruitment length to minimum → much more difficult for rare cancers and is compounded later on where large number of patients are required and no need to balance both recruitment and length of follow-up with large numbers of sites
     • GOLD STANDARD: endpoint of overall survival (OS) in a randomized double-blind study or studies demonstrating the required clinical superiority compared to the current standard therapy in the chosen indication.
       • to determine efficacy, cancer trials require years of monitoring and quality/number of patients dramatically goes down