Maximilian Schons

Resurrecting MeCell Maps with Claude Code

Sat, 25 Jan 2025 00:00:00 GMT

Last year I wrote about putting MeCell in the public domain and closing that chapter. The illustration was preserved, but the interactive Google Maps-style website had been dead for years. Symfony 3.2 end-of-life. PHP 5.5.9. Hardcoded credentials. A full LAMP stack required to run what was essentially a static visualization.

I assumed it was gone.

Then, one evening with Claude Code, it wasn't.

What happened

I pointed Claude Code at the old repository and asked it to think through a refactoring plan. What followed was genuinely surprising: 44 commits, a complete architecture migration from a PHP/MySQL backend to a modern static-first stack, 276 tests written, and a working prototype deployed to mecellmaps.mxschons.com.

The original codebase was roughly 65,000 lines of legacy PHP and Twig templates. The new version is around 12,700 lines of vanilla JavaScript with OpenLayers 8, offline PWA support, and accessibility features the 2016 version never had.

I gave maybe 10-15 prompts total. The rest was autonomous execution against a plan it generated.

Why this matters to me

I spent over 1,000 hours creating that cell map during medical school. Watching it become inaccessible felt like losing something. A month-long refactor was never going to happen alongside actual work.

One evening did happen.

There's something poetic about working on AI risk mitigation by day, then watching an AI agent resurrect a passion project by night. The capability cuts both ways, and I think we need to hold both truths: this technology is powerful, it can do beautiful things, and that's exactly why getting it right matters.

Try it

Explore the interactive cell map at mecellmaps.mxschons.com. All 537 structures, 10 metabolic pathways, Wikipedia integration, and a guided tour.

The illustration itself remains in the public domain on Wikimedia Commons.

Appendix: Technical Migration

	2016	2025
Backend	Symfony 3.2 + PHP 5.5.9	None (static-first)
Database	MySQL 5.7 + Doctrine ORM	Static JSON files
Frontend	jQuery + Bootstrap 3 + OpenLayers 3	Vanilla ES2022+ + OpenLayers 8
Build	Assetic + Bower	Vite 5
Tests	None	276 (Vitest + Playwright)

Claude Code: 44 commits, 1,388,451 lines added, 3,682 removed.

Comparing AI Labs and Pharmaceutical Companies

Fri, 01 Nov 2024 00:00:00 GMT

import aiPharmaImg from '../../../assets/blog/ai-pharma-comparison.jpg'; import randOcLevels from './rand-operational-capacity-levels.png'; import lakeraGandalf from './lakera-gandalf-game.png'; import { Figure } from '@components/mdx';

Over the last year, I spent a lot of time thinking about what the field of AI could learn from the pharmaceutical industry. As a physician who has worked in clinical trials, I am intimately familiar with the processes there and learned to deeply appreciate them over time. I also spent a few months upskilling in AI, from learning how transformers work and attending conferences to project managing ML runs and cyber security workshops.

Here, I juxtapose AI labs and Pharma, showcase some of my favorite examples of discrepancies, and discuss more rigorous assessments of AI models inspired by clinical trial frameworks.

Thanks to LightSpeed Grants for providing some funding to investigate this topic and to the many individuals with whom I have discussed these thoughts over the past year.

AI Labs vs. Pharma

"The industry produces products that can provide great benefits and can pose severe risks to individuals. Extensive technical expertise is needed to develop products, and there are high costs ($100M+) involved in development."

This statement applies to both AI labs and pharmaceutical companies. Moreover, regulators are starting to gravitate toward a similar "mostly unregulated" and "seriously regulated" dichotomy in AI, similar to what is already in place for drug regulation. Using indicators such as application context and invested computational resources (~10^26 FLOPs) is similar to distinguishing between the regulation of drugs and dietary supplements.

In the tables below, I juxtapose R&D processes and stakeholders. Early phases of research and testing look quite similar. Later in the R&D process, clear differences become apparent, such as the commonly discussed difference in path to approval/deployment/licensing, which is far less systematized for AI (at this point) and often lacks complete workstreams.

There are many differences beyond that. One core difference is that the phases of drug development are built to rigorously test the safety of candidate drugs. Safety is defined as preventing a system or product from impacting its environment in an undesirable or harmful way, typically to protect human lives, the natural environment, or assets. The drug development process is not built to identify and tackle threat models where someone would use a drug detrimentally against others, i.e., security risks. Security aims to prevent often-adversarial agents or conditions from impacting a system in an undesirable or harmful way, e.g., product weaponization, autonomous replication, security breaches, or broad societal risks. Indeed, a 2010 National Research Council report concluded that predicting traits such as virulence or pathogenicity to better deal with the risks from bioengineered potential pandemic pathogens—with the degree of certainty necessary for regulatory purposes—would be impossible in the foreseeable future.

While safety of AI systems is clearly a concern, most experts are even more nervous about its security implications, i.e., how this tool can be abused. The AI industry can learn extensively about adverse event detection and management from drug development, but when it comes to protecting groups or even whole societies from technology abuse, the AI industry needs to look elsewhere for guidance.

If anything, this means that AI should go beyond what is expected in drug development, right?

Despite clear differences, many authors have written about various lessons that could be drawn from the FDA in particular regarding regulation and licensing in the AI industry.

The four lessons closely related to pharma as opposed to drug regulators I rarely see discussed are:

Budget
Scientific approach
Scope
Ecosystem checks and balances

Key areas of improvement for AI

Domain	Drug Development	AI
(Context)	Drug development and other industries have spent the better part of a century developing methodologies to evaluate, test, and assure the quality and safety of their products. Industry best practices have made huge leaps over the years. However, these standards were not built in a day, and the lack of proper measures had severe consequences for both companies and patients.	AI is a nascent field with little to no track record in safety and security-related efforts. Given the limited adoption from related fields (and minimal attention from academia), we should be very concerned if AI labs are not meeting the societal expectations we have for other high-risk sectors.
Budget	Today, approximately 50-90% of total R&D costs in drug development are invested in quality and safety measures (including overhead for good manufacturing practices, animal experiments, clinical trials, etc.).	AI companies appear to spend only low single-digit percentages on assurance measures (and increasingly less in relative terms due to rising training costs). This represents an inversion of what we see in mature industries such as pharma or aviation.
Ecosystem Checks and Balances	Drug development occurs within an ecosystem and is far from being run by a single entity. This enables numerous checks and balances.	AI companies are mostly autonomous and have few checks and balances.
Scope	Testing depends on scoping and must be repeated if the scope or product changes. You must be very careful when applying a drug outside the distribution of what you tested.	Application of AI does not seem to recognize being "out of scope." Safety results are often overly generalized.
Scientific Approach	Drug evaluations are based on quantifying acceptable risks and obtaining statistically reliable information about real-world risk.	AI labs currently pursue a more qualitative approach to risk assessment without proper hypothesis testing.

Budget: Putting Money Where Your Quality & Safety Is

This was one of the most compelling insights for me. In short, 50-90% of total drug development costs go to safety testing and quality assurance. Rigorous testing, including abandoning unsafe or inefficacious products, is what makes drug development so expensive—not the engineering challenges of manufacturing the drug itself. With investments of hundreds of millions spent on safety evaluations, they represent probably the biggest single R&D cost factor in a drug development program, followed by effectiveness evaluation, failed candidates, and process development/manufacturing.

To provide some figures: R&D expenditure per approved drug varies between under $1 billion to over $2 billion. Failed candidates account for a third of these costs. Process development and manufacturing consume approximately 15% of the R&D budget from pre-clinical trials to approval. The bulk (~60%) of expenses for successful candidates arise during clinical testing, with the combination of Phase 1 and 4 trials being the most costly. Frankly, it's difficult to disambiguate the costs as virtually everything revolves around the central dogma of producing a safe and reliable drug.

For a $1 billion drug approval:

$330M (33%): Failed candidates abandoned mostly for efficacy or safety reasons
$100M (10%): Process development and manufacturing costs

$66M (6%): Stringent quality and testing criteria in GMP (⅔) and actual manufacturing costs (⅓) (estimate from expert experience)

Small-scale initial GMP batch (up to Phase 1): $3-10M Medium/large-scale GMP batch (Phase 2 to Approval): $25-50M
$225M (22.5%): Preclinical development efforts, including many iterations of

in-vitro at less than 0.5M,

animal experiments (0.5-5M) and
small scale initial non-GMP drug batches at 0.5-1.5M
$335M (33.5%): Clinical trials

Costs vary dramatically, but can be cited at 4, 13, 20 and 20M dollars respectively.

Average number of trials: Phase 1 (1.7), Phase 2 (2.0), Phase 3 (2.8), Phase 4 (3.2)
Tallying them up gets you to average costs of $155M
Capital opportunity costs make up 25-50% of costs in each category due to the decade-long development timeline:

Preclinical phase: ~31 months

Clinical phase: 5.9-7.2 years (non-oncology), 13.1 years (oncology) Individual trial durations: Phase 1 (1.6 years), Phase 2 (2.9 years), Phase 3 (3.8 years)

Looking at drug development costs, safety and quality assurance account for over 50% of total R&D expenditure. When analyzing clinical trials, animal studies in preclinical phases, failed candidates, and GMP overhead, the ratio of development to safety costs ranges from 1:1 to 1:10, depending on the therapeutic area and definitions used. Novel drug designs require additional mandatory safety measures due to their uncharted nature, meaning safety and quality assurance can comprise 50-90% of total costs.

In aviation, my brief investigation of R&D cost breakdown for passenger airplanes shows total costs about ten times higher than drug development, around $10+ billion / new airplane. Safety and quality assurance are integrated into every R&D step, and the testing/certification process alone spans multiple years, consuming hundreds of millions of dollars. The ratio appears closer to 1:1 between testing/quality assurance and actual manufacturing costs, though this warrants deeper investigation. (One helpful intuition pump: A new passenger airplane is in the low $100M range, i.e. 1-5% the development costs)

AI developers appear to be at the opposite extreme, with an estimated 95%+ of total deployment costs going to what's essentially manufacturing—training the model. For example, ChatGPT's estimated $100M cost included only a trivial amount for safety measures. Industry insiders confirm that safety testing typically receives only low single-digit percentages of investment.

Given AI's elevated stakes, a minimum ratio of 1:10 (training costs to assurance measures) seems appropriate for AI models—and possibly more. However, even this suggested ratio falls short of recommendations from concerned AI experts, who call for at least one-third of R&D budgets to be allocated to safety and ethical use. It's also below some AI labs' proclaimed safety budgets, such as OpenAI's former superalignment team at 20%.

Ecosystem of checks and balances: Multi-party development where stakeholders have skin in the game

Drug development has many mechanisms to involve independent parties into the process and provide them with leverage. Many production and manufacturing services are provided by external vendors, contract research organizations oversee testing, pharmacists handling the product and phyisicans recruiting (and later treating) patients. Every single one of them is accountable for their work and expected to ensure that best-practices are accounted for. If not they are legally liable, can lose their license to operate and even end up in jail.

In the following table I tried to juxtapose different categories of stakeholders between drug development and AI. In terms of getting things done, having to synchronize between more people is certainly a challenge. But splitting responsibility across actors and holding each of them accountable by law seems like a substantially more resilient way of advancing a field. Incentives can be better aligned against recklessness and fraud.

A world where AI Labs not only need to find a data center with enough power and GPUs, but also have to engage deeply with multiple committees, independent boards, sensitive applications specialists etc. would be way more appealing to me.

Table: Stakeholders in Drug Development vs. High-Risk AI Development

The following compares the "actors" in drug development with those in high-risk AI model development. Note that while plausible equivalents exist for each drug development actor in high-risk AI development, most have not yet been established, though early efforts are underway. The establishment of independent boards, certifications, and public reporting systems has previously been identified as crucial for AI governance.

Domain	Category	Drug Development Implementation	High-risk AI Model Development Implementation
Regulatory body	International Guidelines Development	ICH / OECD / WHO Good X Practice (GxP, x = variable) guidelines with local implementation, e.g., Good Clinical Practice (GCP) - for the design and conduct of clinical trials; Good Laboratory Practice (GLP) - for design and conduct of nonclinical (animal) studies; Good Manufacturing Practice (GMP) - for the manufacture of the drug product. "GMP material" means a drug produced in compliance with GMP standards.	e.g., NIST AI Risk Management Framework and ISO/IEC FDIS 23894.
	Local Law and Regulatory Guidance	Compliance with these standards is required by FDA regulations for US submissions (e.g., 21CFR210 for GMP, 21CFR58 for GLP, several parts of 21CFR for GCP), and often by other countries.	EU AI Act, guidelines on ethical AI and similar laws
	Regulatory Agencies	FDA, EMA, …	Unclear
Independent bodies	Independent advisory committees	E.g., Expert groups providing evaluations for regulators	Not a standard
	Independent Ethics committees	Institutional review boards (IRBs)	Not a standard
	Independent Safety committees	Safety databases and Data and safety monitoring boards (DSMBs)	Not a standard
	Independent Auditors	Review by independent external professionals as well as federal agencies	Not a standard
	Public engagement	Patient Organizations in respective boards	Not a standard
	Non-Profits	Public advocacy groups and non-profit watchdog groups	GovAI and think tanks such as RAND
Reporting	Public databases	Clinicaltrials.gov, other national trial registrations	Not a standard
	Publication of results in peer reviewed journals	Medical Journals (mandatory publication strategy)	Not a standard, Arxiv (voluntary)
Company	Sponsor	Pharmaceutical company (often the Sponsor)	AI Lab
Providers	Operational Services	Contract research organizations (CROs) for project management	Not a standard
	Subject Matter Experts	Consultants for diseases, regulation, applications, etc	E.g. Consultants for chemical / biological dual use and other harmful application
	Manufacturing	Contract Development and Manufacturing Organization (CDMO)	Data Centers / Cloud providers / Energy providers
	Early product evaluation	Animal study providers, specialized labs	Organizations like METR
	Training and Certification organizations	Medical/pharmaceutical licenses, GxP training, audits, etc	Not a standard
	Logistic / Distribution Providers	Supply chain/pharmacies	Cloud providers running models (e.g. legal access to AI Models only available over certified providers)
	Advanced product evaluation	Study sites	Not a standard. Potentially: Companies with specially certified cybersecurity systems that can participate in controlled real world evaluation runs
	Locally responsible	Physician Principal investigator (PI)	Not a standard. Potentially: AI Officers in companies
	Study subjects	Patients	Customers / IT Systems / data structures / etc.
Real-world Application	Real world testing	Phase 4 Studies	Public Beta of ChatGPT and other LLMs
	Real world monitoring	E.g. FDA's Adverse Event Reporting System (FAERS)	Not a standard; OECD Proposal for incident reporting database
	Customers	Patients	Customers
Bystanders		Access control via physicians. For drug development there is little to no history of harms to non-customers	Strong suspicion of potential high-stake complications for bystanders

Scope: Implications of Different Customers and Ongoing Modification

Approved drugs always have an intended target population. Prescribing outside this population enters a legal gray zone and is approached with extreme caution. Similarly, drug modifications (and even generic drugs replicating the substance) require comprehensive new studies to prove that a) past evidence remains relevant and b) new evidence demonstrates safety in a similar target population. This creates an inherent tension between deploying the latest innovations to all potential patients worldwide and maintaining safety and security standards for each target group and version.

AI developers appear to be inadequately scoping their safety and security assessments, deploying to customers and contexts far removed from their testing scenarios, without adequate tracking of consequences.

While generative AI models' safety testing is considered unusually complex due to their broad capabilities, this shouldn't lead to abandoning safety testing. Instead, it suggests that capabilities are too broad and need to be scoped to allow for conclusive safety data collection. The solution is to narrow down use cases for generative AI models. Many drug candidates similarly show wide-ranging potential benefits across different diseases or even in healthy individuals. However, due to the cost of establishing product safety confidence, companies can only market them to targeted populations where convincing evidence for safety (and effectiveness) has been collected.

Whether for society at large, companies, or specific individuals, different use cases require different types and amounts of conclusive and reliable evidence to determine product safety, either through company testing or regulatory oversight. This necessitates a structured approach and external validation of experimental planning, execution, and analysis.

Scientific Testing: AI Safety Studies for Assessing Dangerous Capabilities Reliably - Scale Matters

In drug development, all stakeholders share a singular obsession: do no harm to humans. This applies both to test subjects and the millions who will eventually receive the drug. Massive resources are invested in ensuring safety, with every aspect of development, testing, and rollout designed to meet predefined conditions.

There are various ways to quantify acceptable risks, and risk tolerance always depends on context and potential positive effects. For example, the acceptable risk profile for a potentially life-saving drug in a terminally ill patient differs significantly from that of a vaccine against a mild virus in healthy individuals. The most serious adverse events in drug development are termed "serious adverse events" (SAE), encompassing seven specific event types (see here for definition). For most drugs, even single-digit occurrences of these events can halt an entire development program or restrict the drug from certain patient populations. Event rates of serious adverse events between 1:100 to 1:10,000 are often considered unacceptable.

This stringency exists because rolling out drugs across an entire population could potentially cause thousands of deaths or other SAEs—outcomes deemed absolutely unacceptable in most cases. Preventing such situations requires rigorous evidence collection, monitoring, and quality assurance.

The upper bound of harm envisioned in drug development appears to be the lower bound of what AI experts consider catastrophic risks. Beyond immediate individual harms, AI experts acknowledge potential catastrophic societal risks, ranging from "thousands of deaths or hundreds of billions of dollars in damage" (Anthropic's RSP) to extinction-level events.

Current primary threat models focus on:

Misuse by bad actors using AI to cause harm via cyber, bio, nuclear, or social unrest
Similar threats from autonomous, uncontrollable AI agents

Current frameworks for assessing frontier AI model risks typically involve teams of 3-10 domain experts attempting to elicit specific capabilities, benchmarked against fair comparators like search engines or human performance (as seen in biological evaluations reports and various system cards).

While individual teams test various prompts and paths, a single team's inability to elicit dangerous capabilities with given resources doesn't guarantee that other equally capable teams wouldn't succeed—if only by chance. Capability elicitation in AI models remains a nascent field, and ongoing demonstrations of jailbreaks and other attacks by amateur individuals highlight the importance of scale in testing.

If we want to claim "individuals of type X cannot elicit this capability given Y resources," we need experiments with acceptable event rates (e.g., 1:1000) and power thresholds (e.g., 90% power) repeated multiple independent times.

With sufficient repetition (guided by sample size calculations), failing to detect any team eliciting a dangerous capability could support statements like: "we are 95% confident that fewer than 1:1000 teams of X individuals with resources Y will be able to elicit these capabilities from the model." While this doesn't completely protect against catastrophic risks, it provides substantially more reliable data than assessments from AI lab employees who have obvious conflicts of interest.

AI labs should assess and implement mitigations for frontier models to obtain statistically reliable data on whether individuals or teams at different expertise and resource levels can misuse their models. Similar evidence is needed for dangerous autonomy risks, though this discussion focuses on misuse contexts. For understanding different threat actors, RAND's recent report on securing AI model weights defines five operational capacity (OC) levels from OC1: Amateur level (~$1,000) to OC5: State actor level ($1 billion budget). These resource levels are also relevant for dangerous capability assessments.

<Figure src={randOcLevels} alt="RAND operational capacity levels for AI threat actors" caption="RAND's operational capacity levels (OC1-OC5) for AI threat actors. Source: RAND Research Report RRA2849-1" /> To demonstrate reliable hypothesis testing in misuse scenarios, let's consider the global distribution of actors across operational capacity levels:

OC1 (Amateur): 100,000 actors
OC2: 10,000 actors
OC3: 1,000 actors
OC4: 100 actors
OC5 (State-level): 2 actors

Using the rule of three in statistics, to reliably rule out unwanted capability elicitation at any OC level, you need approximately 3x the number of test instances. For example, 300,000 OC1 attempts would allow you to state with 95% confidence that fewer than 1:100,000 amateur coders with ≤$1k resources can elicit dangerous capabilities.

The elicitation of dangerous capabilities could be measured through aggregate scores across domains or individual metrics, with careful consideration given to appropriate primary endpoints.

Importantly, this scale of testing is achievable: Lakera demonstrated this by creating "Gandalf," an online jailbreaking game challenging users to devise clever prompts of increasing difficulty. Their community has invested an aggregate of 25 years across over 1 million sessions attempting to jailbreak their system.

<Figure src={lakeraGandalf} alt="Lakera Gandalf jailbreaking game interface" caption="Lakera's Gandalf game - a crowdsourced jailbreaking challenge that has accumulated over 1 million sessions. Source: gandalf.lakera.ai" /> In their responsible scaling policy Anthropic defines that they would consider actors who are able to elicit dangerous capabilities with “1% of the total training cost” as concerning such that they would initiate substantial additional security mitigations. 1% of the total training costs is roughly equivalent to OC3, benchmarked at ~1 million dollars. If we wanted to get 95% confident in assuring that less than 1:1,000 of these groups are able to do that, we would need ~3,000 trials, very similar to the minimum number of participants in vaccine drug development where data to detect adverse events of 1:1,000 need to be provided. This would be equivalent to run a phase 3 clinical trial, but likely with 100x the costs (approximately $30M / trial vs $3B). Importantly, one might make the assertion that while 1,000 such actors exist globally, not all of them will decide to pursue misusing the model / will spend that much money on it. One could assume that the outlined experiment of 3,000 trials with an average of $1M per trial would include level OC4 (actors with $10M in funding) as well. Similar to drug development one would assume that multiple such trials might be necessary, in particular for general AI models. General models have many threat models making testing them all exceedingly hard. Scoping down the model’s capabilities, e.g. by purely training it on code, or purely on biology etc. would remove the necessity of multiple paths of testing.

Misuse from OC5 actors is likely out of scope for safety testing of models, as funding (≥ $1B) of these actors is sufficient to create frontier models themselves. Additionally studies on OC3 & 4 level would provide indicators for crucial mitigations that would also apply to even better resourced actors. Additionally safety buffers could serve as a proxy for what is assumed to be possible for state actors (e.g. shifting the threshold towards a more conservative end)

The proposals above necessitate a substantial investment in testing infrastructure. Similar to the interactions of pharmaceutical companies, contract research organizations and quality assured study sites that recruit patients, frontier AI labs would likely need to partner with companies that provide the platform to engage with quality controlled teams and partners securely and with very clear monitoring and reporting requirements.

Appendix

Comparison of Drug and AI model development process

This table lists all the essential steps of the drug development process and maps respective steps and terminology from the AI model development space. In the early stages, substantial overlap exists, whereas no equivalent for clinical testing exists at later stages for AI Models.

Drug Development	Milestones	AI Model Development
Target population and disease literature review; patient sampling; expert consultation	Discovery	Goal: General Artificial intelligence
In-silico / in-vitro prototyping. Identification of product. Correspondence with regulators on how to best assess and determine safety/efficacy	Research	Prototyping (test runs, hypothesis testing, early algorithm evaluations)
Generating initial drug substance via process development run	Early Manufacturing	Generating initial model weights via iterative training runs (iteratively with evaluations at predefined intervals)
Formulation & Fill Finish to arrive at initial Drug Product	Training (Testruns)	User interface attached to model
Preclinical (multiple in-vitro, animals), Phase-0 trials. Ethical approval	Early drug evaluation and iterative drug improvement in controlled environments
Toxicology. Start of assessment of characteristics, safety, and efficacy in highly controlled low stake environment	Safety Evaluations	Safety Evaluations of dangerous capabilities (controllability, autonomous replication, lying, etc) via standard tests and red-teaming
Pharmacology (Pharmacodynamics & Kinetics)		Interpretability runs, incl. capability testing
Multiple model organisms		Fairness assessment
Various temperatures, timepoints and analysis methods	Stability testing	Accuracy, Robustness and jailbreaking testing
Evaluation of generated data by SMEs. Proof of concept established - additional investments in upscaling	Consulting with external experts	Evaluations done by external experts
Clinical Batch	Production Manufacturing run	Model trained with additional data from evaluation runs now meeting all the requirements. Model Fine Tuning
Regulatory and Ethics approval (for each study individually)	Extensive Real World Drug Evaluation: Clinical Trials in controlled environments	No controlled real world evaluation as part of the AI Model development process
Phase 1 (~2x): Assess the drug's safety and dosage in a small group of healthy volunteers. This helps determine the safe dosage range and identifies potential side effects.		Start testing characterization, safety, and efficacy in a highly controlled high stake environment
Phase 2 (~2x): Evaluate the drug's efficacy and further assess its safety in a larger group of patients who have the condition that the drug aims to treat.
Phase 3 (~3x): Test the drug in an even larger group of patients to confirm its effectiveness, compare it to commonly used treatments, and monitor side effects in a more diverse population.
Audits throughout the development process proportional to the criticality of the materials or services provided. All necessary data on safety and efficacy collected and ready evaluation by regulators	Audits and inspections by sponsor and regulators
Application	Approval & Post Market Drug Monitoring	Gradual public Rollout
Phase 4 (~3x): Access to the public		Beta testing programs
Full market release		Public release
Adverse Event Reporting		Withdrawal
		Access controls
New application process with new data collected	Updated Versions	Deployment of updated model. Updates

Overview of safety principles in clinical trials

Category	Description	Documents
Preparation	Research Prior to Real-World Testing: Literature, animal, and later human data indicating concerning effects and their frequency are consolidated in continuously updated reports that define plausibly related adverse events to monitor.	International guidelines for various preclinical safety testing. Example (mandatory) preclinical safety brochure and package insert (consumer) (physician) Pfizer / BioNTech Comirnaty.
	Optimized Experimental Designs: Employ strategies like sequential testing, simulations, and sample size calculations to minimize harm, ensure resource efficiency, and detect adverse events at specific frequencies (e.g., 1 adverse event per 10,000,000 user interactions).	FDA and EMA guidance on statistical principles for clinical trials based on ICH E9 guideline.
	Optimized Detection Environments: Specialized sites employing personnel with mandatory training for optimal detection, response, and reporting of adverse events.	See section 4. of Good Clinical Practice guidelines and FDA guidance on physician responsibilities.
Detection	Dedicated Experiments: Clinical trials, particularly Phase 1, 3, and 4 studies, are large-scale experiments deliberately designed to detect adverse events.	International guidelines for safety in human testing (see section E1 and E2A-F) and FDA guidance for post market evaluation.
	Continuous Adverse Event Evaluations: Extensive structured data collection during development and after market access. Automated signal detection is commonly implemented through pharmacovigilance.	"Pharmacovigilance is the science and activities relating to the detection, assessment, understanding and prevention of adverse effects or any other medicine/vaccine related problem." WHO definition
	Recording All Negative Events: Ensure a holistic view by recording any unexpected, undesired, or harmful occurrences, regardless of their causal relationship.	"An adverse event (AE) can therefore be any unfavourable and unintended sign (including an abnormal laboratory finding, for example), symptom, or disease temporally associated with the use of a medicinal product, whether or not considered related to the medicinal product." ICH E2A
	Maintaining Reliable Data: Employ auditable safety databases, uphold good documentation practices, and implement various forms of monitoring.	Good Documentation Practice (GDocP) as outlined by section 4.9 of GCP.
Management	Categorization and Prioritization of adverse events: Relatedness (causal relationship between treatment and effect), Severity (intensity of affected individuals experience), Expectedness (consistency with anticipated adverse events), Outcome (seriousness of consequences), Clusters (hierarchical mapping of events into clusters).	International Clinical Safety Data Management: Definitions and Standards for reporting, and the FDA implementation. MedDRA provides mapping of over 70,000 medical terms.
	Prespecified Event Handling: Type-specific prespecified procedures and notifications.	Safety event procedures for physicians and pharmaceutical companies (GCP Section 4.11, 5.16, 5.17).
	Unbiased Assessment: Use independent monitors/safety boards and multiparty decision-making to eliminate conflicts of interest.	FDA on Establishment and Operation of Data Monitoring Committees.

Documents in a typical clinical trial application

The minimal set of documents will include
Study protocol

Introduction (incl Study Rationale, Background, Risk/Benefit Assessment)

Objectives and Endpoints
Study Design (incl. Scientific rational, end of study definition, stopping criteria)
Study Object (incl. Inclusion / exclusion criteria)
AI model interventions and management (incl. Data infrastructure, accountability, preparation, etc)
Discontinuation of Study
Study Assessments and Procedures (incl. benchmarks, assessments, interpretability work, definition of adverse events, independent monitoring etc.)
Statistical considerations (incl. Statistical Hypotheses, Sample Size Determination, endpoint analysis etc)
Administrative Matters (incl Ethics, data privacy, informed consent, records, etc)
Investigators Brochure (similar to Model Cards in AI)
Informed Consent
Vendor Evaluation Plan
Monitoring Plan
Project Management Plan (incl comms, escalation)
Vendor Management Plan
Trial Master File Plan (includes Investigator Site Files)
Safety Management Plan
Risk Assessment and Categorisation
Data Management Plan
Statistical Analysis Plan

My human cell map illustration MeCell now in the public domain

Mon, 29 Apr 2024 00:00:00 GMT

In 2016 I created a highly detailed DIN-A1 illustration of the human cell and its biochemestry. The project was called MeCell ("Me" was short for "menschliche" which is German for human). MeCell was driven by my enthusiasm for how the human cell is uniquely suited to be explained on a big sheet of paper - and some annoyances the usefulness for students of other illustrations that existed during my medical studies.

Creating the MeCell cell map easily took more than 1000 hours. I sat in our university library looking through large stacks of biochemistrry books to find the best illustrations. I puzzled hours how to best arrange them so that everything fit well together. This video captures the process with some epic background music:

https://youtu.be/DfmhDEUFDLQ

I'm super greatful for the collaboration with my brother, my dad, and my med-school friend Linn on this project who helped with illustration and shipping of the final product across the country. I also coded up a Google Maps version where you could zoom into the cell map and click on sections that were linked to Wikipedia.

https://youtu.be/DYXTRJ5SXRI

I stopped the project in in 2019 after ~5,000 MeCell maps were distributed to students and cell enthusiasts in Germany. You can check-out our old website here. It was just too time consuming to handle all the administrative overhead and when I eventually started as a clinician scientist.

Today I announce that all of this work is now under public domain. The full illustration is available on Wikimedia Commons now. To increase the accessibility I got the cell map translated in English, Spanish and Chinese in addition to the original German version. (2024-04-29 Note: The uploader is causing me headaches and I'll try again in a couple of days to upload all versions to Wikimedia).

Wikimedia Commons does not allow for Affinity Designer file formats, so I provide them here on my website: English, Chinese, Spanish, German.

This post officially closes this MeCell as a project for me and I am deeply grateful for all our supporters in the past 🙏

The 100h clinical trial

Fri, 26 Apr 2024 00:00:00 GMT

import burgerKingImg from '../../../assets/blog/burger-king-military-base.jpg'; import { Figure } from '@components/mdx';

At Alvea we thought a lot about how to run faster clinical trials. At one point, we even considered whether we could fit an animal testing facility, a manufacturing pipeline, a lab and a clinical trial site into one big aircraft. While you may think this sounds ludicrous, humanity is capable of crazy things when we decide we want to achieve something, including having burger kings in overseas military bases. (I haven't fact checked this twitter post, but it illustrates well that humanity can get even much more mundane things done)

The Nucleic acids On-demand Worldwide (NOW) program by DARPA is one of the only efforts to fund a manufacturing platform to rapidly produce medical countermeasures (MCM). However, rapid manufacturing is insufficient: we also need to evaluate the efficacy and safety of the MCMs, and administer the MCMs to the relevant populations. Completing this is an immense operational challenge, and isn’t addressed by NOW. To illustrate, a recent Ebola outbreak had two vaccine candidates, but was too slow to get the testing started and recruited no one, despite more than 100 cases.

Bill Gates outlines in his GERM (Global Epidemic Response and Mobilization) team proposal the importance of an international team that has fire-fighter spirit when new pathogens with pandemic potential emerge. To my knowledge GERM is not pursued with any seriousness at this point.

Within days of detecting an outbreak, I believe a small team could distribute ~1,000 doses of any stockpiled vaccine candidate and set up ring-vaccination trials to evaluate the effectiveness of the candidates. They’d operate with pre-approved documents and no dependencies to external vendors, reducing timelines to days or weeks, rather than months. And of course, we’d put everything into an A380 aircraft. If the US had ice cream ships during WW2, we should have vaccine airplanes in the 2020s.

Here is the blueprint we developed at Alvea for how to make this a reality. Our “local emergency response” approach is complementary to CEPI’s 100 Days Mission, which is focused on having vaccines ready to be produced at scale in under 100 days after a dangerous pathogen is detected.

100h Pharmaceutical Response and Preparedness Teams

Ambulances, firefighters, and police benchmark themselves to take less than 15 minutes to arrive at an emergency scene. Substantial investments in decentralized and mobile preparedness, structured and repeated training, algorithmic execution, and extensive technology R&D enable this.

Humanity needs an equivalent pharmaceutical preparedness and response ability to stop pandemics in their tracks: with an exponentially growing threat, speed is the decisive factor.

When a pathogen attacks the human body, the innate immune system rapidly provides the first response, buying time and suppressing pathogenicity before the more specific processes of the adaptive immune system become effective. For humanity to be prepared for a new pandemic threat, development is needed in two analogous areas:

Broadly effective vaccines and therapeutics that are stockpiled for immediate deployment and evaluation when an outbreak begins
Extremely rapid targeted product development and deployment after the outbreak starts

Here we argue that a 100h timeline is an audacious, but realistic goal for a first-response to an outbreak. Within 100h of identification of a potential outbreak (T0 ), stockpiled broad pharmaceutical medical countermeasures (MCMs) should be deployed in pre-approved clinical trials and targeted candidate vaccines and therapeutics produced for testing, with scaled-up manufacture beginning in parallel. Nucleic-acid medicines represent a highly promising technology available currently for pharmaceutical pandemic preparedness and response. Unlike any other modality, nucleic-acids like RNA and DNA can form the basis of therapeutics, vaccines, and combination products, enabling a great diversity of medicines to be created from a single manufacturing process. Crucially, it is a ‘plug and play’ technology with extensive safety data and enormous manufacturing capacity globally, which could be adopted in disaster scenarios to quickly manufacture effective vaccines.

A 100h Pharmaceutical Response and Preparedness (PREP) team needs to be established with the mission to contain outbreaks and save lives. It would be available 24/7 to manufacture, deploy and evaluate the efficacy of medicines against pathogens, at the site of the outbreak. A 100h timeline for pharmaceutical emergency response could be the difference between a prospering society and millions of deaths.

Broadly effective vaccines and therapeutics that are stockpiled for immediate deployment and evaluation when an outbreak begins

We argue that the optimal approach for pandemic suppression requires many different MCMs to cover a range of threats. These must be accessible quickly and ready for human use. Many currently stockpiled vaccines target specific strains of pathogens and will either not be useful or need to be updated in the event of a pandemic. Existing stockpiles contain large quantities of few vaccines (until recently, around half of the US national stockpile budget on anthrax vaccines) and therapies.

Broad pharmaceutical MCMs that cover ten or more strains of a viral family have the potential to be an effective immediate response at the earliest stages of an outbreak, enabling suppression or buying valuable time in combination with other non-pharmaceutical interventions (NPIs) until more targeted approaches are available.

Extremely rapid targeted product development after the outbreak starts

Targeted product development needs to begin as soon as a threat is identified. End-to-end integrated and automated drug development and manufacturing processes can generate experimental batches suitable for human use within days rather than weeks or months.

Pre-approved, virus agnostic manufacturing processes, batch release specifications and analytical rather than preclinical testing would facilitate this. Development needs to be independent of external service providers, supply chains, and not reliant on slow biological processes (such as cell-based manufacture). In peace-time, a constant state of readiness will be maintained with regular equipment testing and process validation. Notably it is far more viable to store ingredients to produce the first 1,000 doses of a drug then tens of millions.

Other ambitious proposals to accelerate pandemic product development exist. CEPI’s 100d timeline for a vaccine ready for authorisation allows ~30 days to develop and manufacture pathogen specific vaccines for clinical use, with design being possible in as little as 2 days. In COVID-19 we were not far from this: Moderna manufactured the first batches of their vaccine in 42 days after the sequence was released. With current technology it only takes ~ a week to produce an experimental batch of mRNA vaccine from the vaccine sequence. This timeline can be compressed to days by stacking and automating existing technologies.

100h Pharmaceutical Response and Preparedness (PREP) Team / Special forces

Even within hospitals and despite being health care experts themselves non-emergency doctors and nurses hand off patients who need CPR quickly to specialized teams to ensure best possible care. Time and expertise matter. Similarly, the value of pharmaceutical response is exponentially greater, the closer to T0 it is available. Every hour of exponential growth reduces the probability of containment. The full value of broad pharmaceutical MCMs and rapid development will therefore only be realized with resources dedicated to deploying and developing MCMs at the earliest signals of outbreaks. The Pharmaceutical Response and Preparedness (PREP) Team will fulfill this function having the following capabilities:

Deployment of stockpiled broad pharmaceutical MCMs against all viral families:_ _immediate access to decentralized stockpiles of pre-approved pharmaceutical MCMs and capacity to deploy and evaluate them in pre-approved studies at the outbreak location via plane and helicopter
**End-to-end sequencing, automated drug design and development systems: **technology to allow ‘point of care’ design and development of novel RNA medicines with minimal human involvement
Ultra-rapid cell-free GMP-grade mRNA manufacturing: fully automated manufacture including process control and batch release, allowing manufacture of any stockpiled RNA MCM in an approved process remotely (in case e.g. transport fails) and manufacture of targeted MCMs produced by AI-guided design and development process for immediate testing in humans
**Containerized infrastructure deployment to the outbreak: **containerized full infrastructure stack (sequencing technology, manufacturing equipment, stockpiled MCM and ingredients depot, clinical trial site, software stacks etc.) ready for transport 24/7 via plane, boat, or road
Comprehensively trained and certified execution algorithms: the full “new pathogen to first-in-human” process codified in standard operating procedures and algorithms, similar to those used by CPR professionals
**Deliver on the 100h timeline: **A high-level sequence of events in the first 100h of deployment is listed in the below table.

Milestones of the 100h timeline

Hour	Milestone
1	Reception and verification of outbreak signal. Alert to PREP team members.
2	Confirmation of outbreak and immediate deployment of on-call scouting team carrying an Illumina NextSeq (or similar) metagenomic sequencing device by air. A battery of rapid PCR/lateral flow tests for known pathogen and pathogen families. Activation of additional staff members. Coordination and remote instructions for local non-PREP staff (sampling, contact tracing, etc.)
5	Scouting team arrives at outbreak center and conducts further confirmation steps incl. comprehensive battery of rapid tests and sets up local headquarter. Deployment of first PREP sub-unit team to the outbreak center. Continued remote training of local non-PREP staff on upcoming procedures.
10	First sub-unit team starts enrolment in pre-approved ring-vaccination and therapeutic trial with suspected infected individuals using stockpiled MCMs based on early characterization of pathogen. Early characterization communicated to global RNA manufacturing sites to begin scaled manufacture of stockpiled relevant MCM. Automated drug-development process begins. In case of false positive operations are adjusted and teams called back.
20	Deployment of second PREP subunit team delivering further specialized equipment. Local headquarter in outbreak location established. Start of GMP manufacturing of first targeted candidate.
30	Ring vaccination with stockpiled vaccines & therapeutics completed for all known (contact) cases in pre-approved trial.
50	Vaccination of all local health care personnel with stockpiled therapeutics completed. Start of GMP manufacturing second targeted candidate drug.
75	Start of GMP manufacturing third targeted candidate drug. First results of ring vaccination trial published in continuously updated open access database.
100	Early outcome data from stockpiled efforts available. 1000 doses of first targeted candidate available for administration and use in local trial. Technology transfer to scaled up manufacturing sites. Vaccination center set up with local authorities.

Alvea: Case Study of the fastest biotech to go to in-human trials (5/6 Drug Development Series)

Thu, 02 Nov 2023 00:00:00 GMT

To connect the somewhat abstract information provided in our series so far to a real world use case, we want to outline what made it possible for a team to be the fastest transition of a pharmaceutical company to the clinical stage so far.

Alvea Story & Culture

End of 2021, at the beginning of the SARS-CoV-2 Omicron wave, the biotech startup Alvea was founded with the objective of delivering a booster vaccine with characteristics suitable for low- and middle-income countries. Those included strong shelf stability, straightforward manufacturing and easy adaptability to future variants. Citing from the publication of the trial (currently under review) that tested their vaccine candidate:

“Against the widely held public conception that drug development is slow, bureaucratic and unattainable for new companies, 174 days after Alvea was founded at the onset of the Omicron wave Alveavax-v1.2 was dosed in humans, without skipping any preclinical safety precautions or shortcuts in good manufacturing practice.”

Alvea validated the earliest versions of the vaccine design during a one-week pre-Christmas 2021 hackathon with a team of ~8. Subsequently the org grew to 35 people in the matter of weeks following the mantra: hire the best people anyone has in their network and convince them to collaborate on a three-month sprint. The company ended up with a team spanning from Europe, the U.S., and Canada, to a person living in New Zealand. The following describe core culture principles and activities that we believe enabled the team to achieve these results.

Understand/verify industry best-practices

Alvea spent a lot of time and effort to educate themselves about how to best navigate a very ambitious drug development program. Part of what motivated the authors of this article to write it was not finding a comprehensive end-to-end overview of the essential parts of drug development that could be given to staff members without a history in drug development. Many ways get you there including: reading the GxP guidelines; reading regulatory documents of jurisdictions; investigating documents of other clinical trials; talking to a variety of consultants or doing training programs. Don’t be afraid to ask dumb questionsIt turns out that you’ll end up getting a non-trivial amount of contradicting evidence and resolving is on you - the final responsibility always resides with the sponsor. Combine understanding/verifying what you ought to be doing with an unwavering focus on a primary goal (“Submit our injectable Alveavax-v1.2 clinical trial application to South Africa for conditional approval before March 18th 2022, as long as GLP BA2 studies show positive results.”), relentless work and quick turnarounds, and you are kind of set up for success. For many at the company it was the closest to lived operational excellence they had experienced yet.

Determine your jurisdiction and exact specification

While ICH provides an exceptional improvement in terms of harmonisation across countries’ drug development regulation, and for vaccines in particular many refer to the WHO guidelines, countries differ in nuances that eventually matter. Alvea scanned the world and deeply considered at least 10 different options to find a jurisdiction that provided the right mix of right target population, regulatory clarity, fast timelines and reputable regulators. The top candidates for a Phase 1 study ended up being Australia, South Africa, and Canada, and the decision was made to proceed with South Africa.

Once Alvea knew the target jursidiction, processes could be optimised. The exact specification kick off all the downstream processes: animal testing (which, how many, how long, etc.), GMP material (how stable, what exact GMP specification, etc), submission criteria (timelines for review cycles, do you need to have all the data for submission or can you provide certain documents such as drug stability data later). CROs often like to claim that they can run trials anywhere, but sponsors benefit heavily from local expertise gained by past trial run with the same study sites, ethic boards, laboratories and supply depots. Going with a locally experienced CRO and having exceptionally high standards for the CRO’s project manager was key to the success of the study.

Find, evaluate, and parallelize vendors (or do it yourself if you can).

For manufacturing, animal testing, regulatory application, trial setup, and all the respective subdomains highly specialised suppliers or contractors exist. Towards the goal of conducting the clinical trial, all of them constitute single points of failures. Alvea was very concerned about this and therefore followed a couple of strategies:

Whenever you can reliably do it yourself, you should. Communication across orgs is harder and in most scenarios you don’t know how relentless another party will ensure that they deliver the right results on time, even if it means more hours, a different process, etc. Examples include inhousing initial steps for creating starting material for manufacturing yourself or designing the clinical trial protocol.
If you can afford it, parallelize vendors. Alvea had to rely on contractors for manufacturing, animal testing, and parts of clinical trial setup. For all of them at least initially multiple parties were hired and thoroughly evaluated. The trial was also submitted to multiple jurisdictions.
Build a quality management system and audit contractors. This is not regulation slowing you down, but regulation saving your ass from future disasters.
Be able and willing to part from vendors. Alvea was dissatisfied with the performance of multiple CROs and even while running the trial decided to stop the collaboration and find new partners or in-house the activity.
Build partnerships with contractors that deliver exceptional results.

Alvea winded down mid of 2023. You can read more about the story here and here.

Origins of the Drug Regulation Industry (3/6 Drug Development Series)

Thu, 02 Nov 2023 00:00:00 GMT

In our last pieces in this series, we looked into the people involved and what must be done to get a drug to market. One begins to wonder – why does drug development necessitate all this rigour, time, and money? Understanding drug development's history might offer context and foresight for us to see its current and future direction. As you examine the history, you’ll notice change has often been attributed to reactions to tragedies, though in many cases, a closer examination shows reforms were already underway.

We’ll also see that as international attempts to harmonise laws became successful, the regulatory histories of different countries began to overlap, especially in the second half of the 20th century. Note that the perspective below focuses on an overview primarily from a US perspective – we encourage readers to explore Dan Carpenter’s historical analysis of the FDA for a more detailed understanding and other authors for respective jurisdictions.

US: Tracing the origins of drug regulation

We didn't know it yet, but buying drugs in the 1900s was like rolling a die. At that time, many "patent medicines" said they could fix everything from headaches to heartache. For example, Dr. D. Jayne's Tonic Vermifuge was advertised as a cure-all for throat and respiratory diseases and a worm remedy in 1896.

An advertisement for Dr. D. Jayne's Tonic Vermifuge

Another iconic representation of patent medicines in the 19th century was "snake oil," touted for its supposed healing properties ranging from curing headaches to alleviating chronic pain. The lore of snake oil symbolises the archetypal patent medicine, embodying the unchecked claims and placebo remedies that were commonplace in an era before rigorous drug regulation and testing.

A photo of famous "snake liniment" flyers and bottles by Clark Stanley in the 1800s

This was the landscape before drug regulation. Imagine you're a scientist in this narrative, eager to test and market your new drug. Maybe this lack of regulation might have felt liberating, but as time unfolds, you start seeing how the crucial need for standardisation and safety protocols becomes apparent. As time passes, evolving regulatory frameworks will increasingly shape your work.

Since the 1880s, there have been efforts to regulate the proper labelling of food and drugs and prevent contaminations in either category. Harm due to unregulated food and medicines, such as the contaminated diphtheria antitoxin incident, which was prevalent, leading to multiple children’s deaths. For the first time, the Biologics Control Act (1902) institutionalised inspections and purity testing in the US, marking a move towards formal regulation in ensuring drug safety and efficacy. Shortly after, the American Medical Association (AMA) pioneered self-regulation with a voluntary drug approval program, where companies were required to have proof of efficacy for their drugs before advertising in AMA’s journals. These set precedents for further regulatory measures like **the Pure Food and Drug Act (1906) **for the US Federal Drug Administration (FDA), making selling contaminated food or meat illegal. The law required truthful labelling – no one could “promise the moon and the stars” on a label anymore – at least, that was the goal.

US: Reactivity, Proactivity, and Regulatory Measures

Many regulations resulted from the public outcry that followed health disasters.

Federal Food, Drug, and Cosmetic Act (1938): Mandated safety proof before approval, introduced following a devastating tragedy involving a poorly formulated antibiotic, leading to over a hundred deaths.
Good Manufacturing Practices (GMP, early 1940s): Developed following an incident with a batch of sulfa drugs contaminated with the sedative phenobarbital that went awry (with hundreds of deaths and injuries).
**Nuremberg Code of 1947: **Served as a guideline for clinical trials (establishing principles for ethical research like voluntary consent) in the wake of inhuman experiments by Nazis in Germany, built upon by the Declaration of Helsinki in 1964.
FDA’s ban on GAS vaccines (1979), An arguably overreactive prohibition on using Group A Streptococcus (GAS) organisms and their derivatives in vaccines after two or three out of 21 healthy subjects developed rheumatic fever after vaccination in a study in 1969. Despite the long history of successful M protein vaccines—targeting components of bacteria like GAS—the FDA's ban effectively stifled research in this area for nearly three decades. The decision has been considered particularly significant given the high unmet medical need related to GAS. It wasn't until 2006 that the ban was lifted, allowing for a resumption of vital research and prospective breakthroughs.
Faster Access to Life-saving Medications (1990s): Albeit carefully monitored, access to life-saving medications after AIDS activists criticised the FDA for blocking access to potentially life-saving drugs and slow drug approvals during the 1990s.
The FDA Gender Guideline (1993)**: **Required medication testing in both sexes and overturned a 1977 guideline that excluded fertile women from early clinical studies (which was a response to the Thalidomide disaster, an event described below) .

_However, _it’s important to note that while it is tempting to think that drug regulation emerged purely from public outcry, a sort of legislative knee-jerk reaction to high-profile disasters, that's only part of the story. Many reforms were already simmering on the backburner before tragedy turned up the heat. Also, the policy evolution shifting from reactive to proactive regulation is notable. If legislative changes used to be more crisis-driven, today, there's an apparent effort to anticipate risks and implement preventive measures. Voluntary guidelines have also played a role, sometimes originating from industry best practices that later become regulatory expectations.

As early as 1933 – to exhibit shortcomings of the 1906 Pure Food and Drug Act and illustrate the need for a new law – the FDA launched the “American Chamber of Horrors,” an exhibit featuring problematic products like unsafe medical devices (e.g. harmful womb supporter and a deadly weight-loss drug) and toxic cosmetics. First Lady Eleanor Roosevelt, who had toured the exhibit, leveraged it to push for stronger consumer protections. This was a precise instance of the FDA taking the initiative to push for legislative change.
FDA "batch certification"** (Early 1940s): **Mandated for certain key drugs, starting with insulin in 1941 and then penicillin in 1945. Companies were required to submit samples from each lot for FDA testing before they could be released to the market. The policy was later expanded to include all antibiotics but was discontinued in 1983, a rare example of regulations getting less stringent.
The Belmont Report (1979): Laid out ethical standards for human subject research, emphasising respect for persons, beneficence, and justice. If you’re the scientist, the responsibility would now fall on you to ensure that every step of the drug discovery and clinical testing adheres to these ethical guidelines, a testament to the evolving integrity of the field. It profoundly impacted clinical research regulations in the US, guiding ethical considerations in medical trials. Around the same period, bioethical committees in other countries also made strides, publishing comparable guidelines and reinforcing a global commitment to ethical standards..

The Thalidomide disaster (Late 1950s - Early 1960s) is a significant chapter in drug development history, highlighting the importance of historical tension between reactivity and proactivity. Thalidomide – a drug initially approved to treat morning sickness in pregnant women, used a lot in the 1950s – led to severe congenital disabilities that were apparent in the 1960s and ignited public outcry for stronger drug laws. US Senator Kefauver, already advocating for drug reform, seized this crisis to reinvigorate a languishing bill in Congress. In July 1962, Kefauver publicised the drug's harmful effects, breathing new life into his bill, which he felt had lost momentum.

Apart from having lots of reactions, the tragedy catalysed broad reforms in drug regulation in the US and globally. In the US, the disaster led to a mandate requiring any sponsor planning a clinical investigation of a drug to provide a detailed study outline to the FDA and demonstrate its efficacy, not just its safety. Internationally, the event spurred centralised authorisation procedures in the EU for new drug assessments. For instance, the UK passed the 1968 Medicines Act, establishing comprehensive drug classifications. Generally, the Thalidomide disaster served as a wake-up call for the industry, catalysing reforms that were not just reactive but also comprehensive and robust.

Global Harmonisation for Drug Development

The International Conference on Harmonisation (ICH) aimed to standardise the drug development process starting in the 1990s (similar to how a USB-C attempts to standardise connectors). The ICH initiative represents a collective stride towards a more streamlined and unified global drug development paradigm, ushering in a new era of collaborative efforts among drug manufacturers across borders. By bringing together regulatory bodies, from Europe, Japan, and the US, the ICH ensures that a drug tested and approved in one region can more easily gain acceptance in others, accelerating global access to vital medications.

From their website: “ICH's mission is to achieve greater harmonisation worldwide to ensure that safe, effective, and high quality medicines are developed and registered in the most resource-efficient manner. Harmonisation is achieved through the development of ICH Guidelines via a process of scientific consensus with regulatory and industry experts working side-by-side.”

ICH has developed extensive processes for new harmonisations, clarifications, revisions, and maintenance and has ~150 Guideline documents on safety, efficacy, quality, and miscellaneous. At Step 5 of the ICH process, harmonised ICH Guidelines are implemented by ICH Regulatory Members and Observers within their respective country/region. Implementation and adherence to ICH Guidelines within Regulatory Member and Observer countries/regions are monitored via independent third-party surveys (see 2021 Project Report). ICH has a total of 21 member organisations, each with two representatives. Many regulatory authorities incorporate ICH guidelines into national law, making compliance a legal requirement.

The ICH Process.

An example of the ICH process at work: As we progress into 2023, the ICH’s Good Clinical Practice (GCP) guidelines are under ongoing revision. They were initiated in 1996, last updated in 2016, and their newest revision is currently accessible as a draft. These guidelines continue to evolve, focusing on new trial types and data sources, reflecting the regulator’s and industry's commitment to staying ahead of the curve.

Global Milestones in Regulation: A Tour from Beijing to Russia

While the history is approximately similar to the US for many Western countries, this section provides a very short peek across borders and overseas; the narrative of drug regulation also unfolds with distinct chapters written in different nations. Each country has its historical events and reforms, sculpting its regulatory landscape. This is meant to provide a concise overview and to shift away from an US centric view on drug development – we’ve linked sources for deeper dives into specific countries and don't at all claim to have significant expertise for any of these regions. Let's explore a few key milestones in certain countries.

Brazil

The National Health Surveillance Agency or ANVISA (1999): Established to protect the population's health through sanitary control.
Law No. 13.411 (2016): Modified the pharma regulatory environment, lowering bureaucratic barriers in clinical research and improving pharmaceutical registration and post-market monitoring processes.
Read more: Drug registration in Brazil

China

**The Pharmaceutical Company of China (1950): **assigned to take charge of the nation’s wholesale trade of pharmaceuticals, at a time when hospitals/clinics were transferred from the private sector to direct government control.
**GCP Launch **by the State Drug Administration or SDA (2001)
CFDA Improvements (2015)**: **Announced by the China Food and Medication Administration to speed up medication approval processes and give Chinese patients faster access to new drugs.
ICH Membership (2018): China became a full member of the ICH, demonstrating its commitment to harmonising its pharmaceutical rules with global norms (and the reach of ICH).
Read more: China’s drug regulation history

India

Pre-Regulation

Practicing of traditional Indian medicine systems (Pre-20th Century) Using natural ingredients like Ayurveda, Siddha, and Unani for treatment.

**Importation Era (20th Century) **Most drugs were imported from Europe during their colonial period.
The early pharmaceutical industry in India began with a few indigenous companies manufacturing basic medicines in the early 20th century, limiting local pharmaceutical development.
Drugs and Cosmetics Act (1940)**: **Established regulatory control over the import, manufacture, distribution, and sale of drugs and cosmetics in India, making the sale of substandard drugs a serious offence.

In 1945, the government established the Drugs and Cosmetics Rules to classify drugs under given “schedules” and present guidelines for the storage, sale, display, and prescription of drugs.

In 1988, the Act was reformed to integrate WHO's GMP principles.
In 2005, The Act's “Schedule Y” (schedule providing clinical trial guidelines) was updated to incorporate more precise standards and instructions for conducting clinical trials in India.
The Patents Act (1970): Allowed the development of India's generic drug industry by recognising only process patents, not product patents.
Modernisation and Global Expansion (2000s-Present):

Becoming a significant player in “Contract Research and Manufacturing Services” (CRAMS)

Formation of regulatory bodies like the Central Drugs Standard Control Organization (CDSCO) and the National Pharmaceutical Pricing Authority (NPPA) to improve drug quality, safety standards, and ensure affordability
Changes in India's patent laws due to the introduction of the World Trade Organization’s Trade-Related Aspects of Intellectual Property Rights (TRIPS) agreement, affecting the approach to generic drug production.
The New Drugs and Clinical Trials Rules (2019): Reformed the regulatory framework by reducing complexity in clinical research and improving the processes for medicine registration and post-market surveillance.

Russia

Soviet Regulation (Post-1917 Revolution - 1930): Following the 1917 Revolution, the government established a state monopoly controlling all aspects of pharmaceutical production, distribution, and quality assurance, including the creation of the Pharmacopoeia Commission in 1923 (that controlled pharmaceutical quality) and comprehensive management of pharmaceutical activities under the People's Commissariat of Health by 1930.
**Recent regulation (2010-11): **Enactment of federal laws aimed at improving healthcare and drug provision, alongside the Law “On Circulation Medicinal Products” in 2010 to ensure the quality and safety of medicines, reflecting the evolving healthcare landscape.

Zimbabwe

The Drugs and Allied Substances Control Act (1969): An act beginning systematic control and management of medicines in the country.
Establishment of the Medicines Control Authority of Zimbabwe or MCAZ (1997): Created to oversee and regulate medicines in the country.
Regulations on Homeopathic Remedies (2015)**: **An act representing an official cancellation or revocation of a previous set of regulations governing homoeopathic remedies in Zimbabwe.
Good Distribution Practice (GDP) Guidelines Adoption (2018): Recently, the MCAZ adopted GDP Guidelines to ensure the quality and safety of medical products during all aspects of the distribution process.

What about countries that don’t have an independent drug regulatory system?

Despite these strategies, research has shown how developing countries still meet resistance in adopting global standards for pharma quality and regulatory infrastructure. Countries that don’t have their own independent regulatory system for drugs (generally smaller countries that don’t have their own explicit regulatory authorities) use certain strategies:

WHO Prequalification: These countries regularly rely on the WHO prequalification system, which analyses and ensures the quality, safety, and efficacy of vital medications.
The Collaborative Procedure for Accelerated Registration**: **spearheaded by the WHO, expedites product registration in numerous nations by leveraging the work of strict regulatory bodies.
**Reference to neighbouring frameworks: **Some countries adopt or refer to the regulatory frameworks of adjacent countries or regions with more established systems.
Reliance on international guidelines: Countries can use international guidelines from orgs like the ICH and the WHO.

Recommended Literature and Courses (4/6 Drug Development Series)

Thu, 02 Nov 2023 00:00:00 GMT

Certified training is mandatory for most activities in drug development for staff at sponsors, trial sites, vendors or regulatory agencies. A wide range of online and in-person courses exist. The following lists some established providers we have used for training colleagues in the past and a list of strongly recommended complementary reading (think of “most important references used in the series”):

Manufacturing: Video on the 10 principles of GMP (not strongly recommended, but it’s OK; note that it’s not certified)

**Quality Assurance: **Good overview of quality management systems by one QMS vendor (not certified):

Pre-clinical: This is a solid GLP resource (not certified)

Clinical Trials:

Excellent resource for an overview of clinical trial documents by smartsheets.
Basic training: Global Health Training Centre: Free high-quality courses on basic topics such as GCP, Data Management, Adverse Event Management and much more.
Advanced Training: Association of Clinical Research Professionals (ACRP): Paid courses cover a wide range of topics, including advanced courses on quality management, event reporting, etc.

Complementary reading to the Drug development Series :

General

Research and Development in the Pharmaceutical Industry, Congressional Budget Office

Costs and barriers to drug development by the Assistant Secretary for Planning and Evaluation (ASPE)
A History of the FDA and Drug Regulation in the United States
Manufacturing

COVID-19 and Beyond: Oversight of the FDA's Foreign Drug Manufacturing Inspection Process (provides a good overview of FDA inspections)

Clinical trials

ICH efficacy guidelines: Particularly E6:GCP

Why are clinical trials so complicated? Short article from the Science Magazine
Study other sponsor's clinical trial protocols, informed consents, and investigator's brochures

Conclusions Drug Development Explainer Series (6/6)

Thu, 02 Nov 2023 00:00:00 GMT

Below are my personal core conclusions from this series (and my work in drug development). I think many of these lessons are not limited to pharma but describe general principles for working on high-stakes products, finetuned over decades of discourse and experience.

History

Many dramatic incidents related to drugs happened in the past

Regulators have often been reactive and but also proactive in shaping the drug development landscape
Harmonisation and standardisation across nations is a big trend
All Stakeholders

... do everything possible to prevent harm to participants (and to a lesser extent animals)

... Train everyone involved constantly
... Get independent opinions and have a system of checks and balances - audit and inspect each other
Quality

Have a process that reproduceably and reliably creates what you want and verify that

Document everything such that anyone can reconstruct exactly what has happened
Prepare for risks and tackle issues with comprehensive measures
Don’t take shortcuts and invest the time and resources necessary
Operations

While highly regulated, drug development can be an extremely agile process depending on the involved individuals.

This concludes this series. Hope you enjoyed the read 🙌!

The Blueprint of Drug Development (2/6 Drug Development Explainer Series)

Wed, 18 Oct 2023 00:00:00 GMT

**Author Note: **This series was written in collaboration with James Smith and Kirsten Angeles.

Overview

In our last piece, we looked into the different stakeholders involved in drug development and the roles they play in bringing a drug from concept to market. To appreciate the meticulous nature of the journey of these stakeholders, let’s look at the actual process they participate in, starting from idea to drug.

It is commonly known that drug development happens across multiple phases and is subject to diligent protocols and documentation before market approval. While companies don’t share their R&D budgets, estimates put the cost of bringing a new drug to market from less than $1 billion to more than $2 billion. That includes the cost of failure; approximately a third of this budget is spent on candidates that never make it to market. Developing a new drug isn’t for the shallow of pocket.

In the following, we not only break down the main phases – manufacturing, discovery, preclinical testing, clinical testing, and market approval – but also want to provide a behind-the-scenes view with examples. We also go into some of the various Good Practices _(GxP) guidelines _– i.e., Good Manufacturing Practices (GMP), Good Laboratory Practices (GLP) and Good Clinical Practices (GCP).

Phase	Key Aspects	Average Duration / Data	Key Regulatory Guidance	Costs and Investments
Manufacturing	GMP Compliance, Quality Assurance, Record-keeping	N/A	GMP (Good Manufacturing Practices)	15% of R&D budget; Single to double-digit million dollar runs
Discovery	Compound Screening, In-vitro, In-silico Testing	Years	Biosafety standards	A third of overall spend on failed candidates
Pre-clinical Research	Pharmacokinetics, Toxicology, Pharmacodynamics, Efficacy	~31 months on average	GLP (Good Laboratory Practices)	~$75k for single 100 mice study, varies widely based on specifics
Clinical Development	Phase I to IV Trials, Trial Operations	5.9 to 13.1 years (varies by drug type)	GCP (Good Clinical Practices), IND submission	~$5-30M per trial on average - varies widely
Market Approval	NDA/BLA Submission, FDA Review	6-10 months for first review, often with resubmissions; ~400 IND applications/year; 38 new approvals/year	NDA (New Drug Application), BLA (Biologics License Application)	Approximately $2-3M as filing fee
Post-Market Surveillance	Post-Marketing Surveillance Reports, Efficacy Studies	Reports initially every 6 months, extending to 3+ years	Post-Marketing Surveillance Reports	N/A

Manufacturing

Manufacturing is defined by the ICH as to “include all operations of receipt of materials, production, packaging, repackaging, labeling, relabelling, quality control, release, storage, and distribution of [drugs] and the related controls.”

The respective global standard outlining how drugs should be made and ensuring that quality is “baked into” _every step of the manufacturing process – extending far beyond testing of the final product – is called GMP. In the US, it's called _Current Good Manufacturing Practices (cGMP), emphasising that manufacturers need to use up-to-date technologies and approaches to remain compliant.

In practice, this means there are requirements for every conceivable part of manufacture – look at this GMP guide to get a sense – you’ll see that even personnel hygiene has a few lines. Compliance with the requirement processes involves written protocols (e.g. in the form of SOPs, Master Formulae), employee training, internal audits, equipment testing, good quality materials being used (e.g., ensuring the traceability and quality of materials used in production), quality assurance, quality control measures, and many other things, all documented in a quality management system (QMS). GMP emphasises thorough record-keeping making tracing any activity during a manufacturing batch possible. (For more details on quality assurance, this paper provides a readable overview – it’s focused on cell therapy, but the principles are the same.)

How does GMP look in practice?

Processes are defined and execution-checked. Every step in the manufacturing process and testing of the drug product, which consists of the actual active ingredient (drug substance) and all other adjuvants, binders, coatings, etc., must be defined by a recipe/procedure, execution against which is recorded and checked.

This recipe will be highly detailed: for example, there might be three different steps.

Measure 1 ml of reagent A using a specific instrument and model.

Combine with reagent B into a specific vessel of a specific size and model.
Record the volume of the resulting mixture using a technique defined in a standard operating procedure.
Person A follows the procedure.
After every small step, person A, who completed it, has to initial next to that step on a piece of paper (or electronic equivalent), confirming that they did what was stated there. If there is a result to record (e.g., as in step iii), they write that result.
Person B observes Person A doing the step, and Person B also initials next to the step to confirm that Person A did it correctly. If Person A recorded a result, Person B confirms they recorded it correctly.
Person A must have been trained to follow that procedure and demonstrate that they were trained to follow it. Same for Person B, the checker.
All of this is done for every “quality-critical" step.
**Intermediate products are tested and quality-verified. **There are tests of intermediate products throughout the manufacturing process to ensure that it is within pre-defined specifications and, e.g., there has been no contamination. Often, new assays must be developed to test whether the intermediaries and final drug substance you created according to specification. It is insufficient to rely on testing the drug at the end of the process.
**Records are reviewed and approval-secured by a dedicated quality function. **Before patients can use a drug,, all of the documents recording every step of the process must be reviewed by people from a department whose sole role is quality assurance.

When all the manufacturing is completed, you have a set of documents that have been initialled by persons A and B, each one representing a part of the overall manufacturing process.

Collectively, that set of records is referred to as the batch record. (They record the manufacture of the _batch _of the drug.)
The quality function (quality assurance department) of the manufacturing facility must review the batch record.

GMP requires that the quality function be operationally independent of the team directly doing the manufacturing to avoid conflicts of interest. The quality function typically reports to the CEO.

In some countries, e.g., in the EU, this “batch release” can only be done by “qualified persons” (QPs) – accredited people with specific expertise in drug manufacturing.
If a sponsor (the company who owns the drug, essentially – read more here) outsources manufacturing, this step will be done by the sponsor andthe contractor directly doing the manufacturing.
The review will include, for example, checking that the product meets pre-defined specifications and that procedures were properly followed.
**Documents are retained for potential audits. **The documents resulting from this process, and those recording the review of them that took place, will be stored securely and version controlled, and accessible to regulators during audits.

Manufacturing a drug comes with a lot of overhead. Small quantities of drug prototypes for very early experiments might be synthesised in-house by sponsors, but due to all the processes listed above, collaborations with CDMOs are often needed – and they commonly have waiting times of 6-12 months. Interestingly, only about 15% of the total R&D budget is allocated to manufacturing, with the bulk going toward quality assurance and testing. The cost definitely differs by therapeutic area, but a GMP manufacturing run often demands a single to double-digit million-dollar investment, while non-GMP batches usually are well below the million dollar mark.

Drug Discovery

There are many ways of arriving at a hypothetical candidate, some of which involve in-silico screening of thousands of compounds for efficacy depending on the therapeutic approach (e.g., High Throughput Screening (HTS), others are simple in-vitro test tubes or cell culture experiments. There is a significant reduction in the multitude of compounds as they are screened - see the following figure (source) of the failure rate at each step of drug discovery and development.

While no “Good Discovery Practices” exist and don’t require formal regulatory approval, scientists in laboratories are always subjected to biosafety standards under the purview of agencies — state agencies for lower-level labs (BSL-2 and below), and the CDC in the US for higher-level labs (BSL-3 and BSL-4). Additional regulation can apply in special circumstances, such as pathogens of specific concern (Federal Select Agent Program) or due to international collaborations.

Failed candidates consume about a third of a drug development program's overall spend, i.e. hundreds of millions of dollars, and generally spans over the years. “Aha moments,” where scientists successfully identify compounds that target key biological mechanisms, are rarer than one would hope.

Pre-clinical research

Once a candidate is identified, it goes on to preclinical research. This stage used to require animal testing, but the FDA announced in 2022 that they now also accept non-animal organoid models.

Different parameters are studied in preclinical animal studies. While the details of how to conduct the testing of drug safety and efficacy in animals are described in GLP, with additional specifics on what to test are provided in different regulatory documents such as the Preclinical Safety Evaluation of Biotechnology-Derived Pharmaceuticals or the WHO guidelines on non-clinical evaluation of vaccines. In general, the following components are relevant:

Pharmacokinetics: This area investigates the absorption, distribution, metabolism, and excretion of the drug to understand its behaviour in the biological system.
Toxicology: Defines dose-related effects on different organ systems and developmental stages of organisms. They help to establish the safe starting dose in human trials, where the doses at the No Observed Adverse Effect Level (NOAEL) in animals get multiplied by safety factors. (A detailed elaboration on extrapolation from animal to human doses can be found here.)
Pharmacodynamics: Determines the effects of the drug on the body, including the mechanism of action and the relationship between drug concentration and effect.
**Efficacy studies: **These determine how well a drug performs its intended therapeutic function. It includes investigations into dose-response relationships to understand the optimal dosage required for the desired therapeutic effect without producing harmful side effects. This helps define the therapeutic window of a drug, a critical aspect in further clinical development.

Some of these tests can be combined; others warrant additional studies. For instance, a single toxicity study usually tests~100 animals from at least two species, and can also be used to capture additional pharmacological aspects of the drug. The total number and type of animals can rise far beyond that when additional testing in different model organisms is warranted. Depending on factors like the drug's chemical makeup, treatment area, dose-levels, and effects being tested.

There are rarely hard requirements from regulators regarding which specific animal model to use, but there might be _expectations _in guidelines, and it is worth speaking to people with experience in the field (e.g., while not explicitly mentioned in guidelines, Syrian hamsters have been a standard for COVID-19 vaccine studies, sheep are a common organism for inhalation studies, and non-human primates are more and more in demand for almost any type of research, despite often being actively discouraged by regulators).

Unintuitively, the cost drivers in preclinical studies are not the number of animals but the drug itself, human labour at the study provider (number of doses, drawing blood, etc.) and analysis/pathology costs. In recent US industry quotes studies with ~100 mice with multiple injections and blood draws over 35 days, the cost per animal was around $750, totaling $75k. The price tag for manufacturing the necessary non-GMP drug can easily double this, with analysis costs ranging from 0.5 to 1.5 times the study costs. These numbers obviously can vary a lot depending on the concrete animal study plan, typically provided by the study provider once they know all the relevant variables. Preclinical testing takes about 31 months and constitutes roughly 43% of a drug development program’s total R&D spending.

Clinical development

Once a potential drug clears the preclinical studies, it moves into clinical trials in humans. Clinical development is segmented into Phase I, II, III, and IV trials, each with distinct objectives and escalating scales of operation. A single drug approval commonly involves multiple trials across different phases, testing in different patient subgroups, dose levels, or endpoints. In the EU, ~3,700 clinical trials are authorised annually, making up 46% of nearly 8,000 applications.

As a rough guide (combining different sources here, here, and here):

Phase I: Engages up to 100 healthy participants; ~2 studies/drug development program
**Phase II: **Involves ~100-500 patients and aims to establish efficacy proof of concept; ~2 studies/drug development program
**Phase III: **Tested on a group of ~500-5000+ patients to generate further safety and efficacy data sufficient to convince regulators to approve the drug; ~3 studies/drug development program
Phase IV: ~100-1500+ patients depending on the trial and drug being studied; ~3 studies/drug development program.

The total participant count of a clinical development program can range from hundreds to over 30,000 before full drug approval. Note that for certain indications, especially less common diseases, the number of patients in each phase can be much smaller than stated. The figure (source) shows the associated average costs of trial (note that substantially more expensive outliers exist)

Clinical testing takes about twice as long as the discovery and preclinical stage together. Non-oncology drugs’ median duration in clinical trials ranges from approximately 5.9 to 7.2 years. For oncology drugs, that timeframe extends to 13.1 years on average. Breaking clinical development by phase, expect to spend ~1.6, 2.9, and 3.8 years in Phase I, II, and II, respectively.

The unprecedented speed of the COVID-19 vaccine development sparked hope of many for shorter timelines in the future. There often exists an impression that this is due to slow pharma companies and regulatory hurdles. Sponsor make mistakes like poor study design, ineffective site selection, poor recruitment, patient burden/safety, and trial execution and induce delays or even failures. However, it is important to note that some factors are really hard to accelerate. Sourcing certain types of participants (e.g. rare diseases) can be very challenging. While it is easy to get surrogate blood markers, endpoints we care about the most – mortality and morbidity – require longer observation periods. For instance, we could ask, “Do less people die due to cardiovascular benefits over one, three, or up to five years?” – this would be inherently impossible to expedite.

GCP is the single most important guideline for conducting clinical trials. It covers ethical considerations, informed consent, roles with their responsibilities, quality assurance and data management protocols and much more. Compliance with and training for GCP is mandatory at clinical trial sites, and principal investigators can be personally liable for non-compliance. A clear and comprehensive explanation of how all these requirements will be met must be submitted to regulators (also known as Investigational New Drug (IND) submission in the US) and local IRBs. Three of the most relevant documents are:

The study protocol includes all primary/secondary/exploratory endpoints, statistical analysis plan, and other applicable descriptions, e.g., randomisation strategies. The clinical trial has to be performed according to protocol. Deviations from the protocol have to be recorded and can threaten the validity of the trial if mishandled. Many study protocols are published and accessible either via Clinicaltrials.gov or via the supplements of publications of the respective trial.
The investigator’s brochure (IB) is aimed at educating physicians to conduct the clinical trial and being able to educate themselves, staff, and patients about the investigated product. This includes information on the drug's composition and production, nonclinical study findings (usually from in vitro and in vivo testing) that demonstrate the drug's safety profile and warrant human trials. The GCP guideline in its last chapter outlines what sections should be included in an IB. While less often publicly available, a good example is the Pfizer mRNA COVID-19 vaccine Investigator’s Brochure.
**Patient Informed Consent **IRBs evaluate that human participants' rights and well-being are safeguarded, the research's potential benefits outweigh its possible hazards, and the investigation is carried out ethically and in accordance with recognized regulations. IRBs often provide checklists or templates what information and formatting they expect. CROs commonly know about special preferences of IRBs and can advise accordingly.

Both the regulatory authority and the IRB evaluate the document package upon receipt to assess whether the proposed trials are safe to conduct. There can be very different timelines across jurisdictions/IRBs and expert regulatory consultation can sometimes save months. Approval for a Phase 1 trial in Australia for instance can be granted in a matter of weeks, whereas the US or EU easily can take 6 months.

Similar to the GMP section, the following provides a concrete example of all activities involved in a clinical trial a sponsor needs to keep track of:

Trial Operations

CRO and site selection (in consultation with stakeholders - Clinical, regulatory)

Compiling and submission of documents for CTAs
Compiling and submission of documents for Ethics
Study closeouts
Project management plans for the trials, incl. trial timeline management
High level project management of the clinical trial
Management of vendors not associated with other SMEs
Trial insurances
Legal agreements related to the trial and trial vendors
Risk log
Action log
Identification and Continuous collaboration with clinical trial sites and CROs
Coordination of site visits in collaboration with relevant SMEs from Alvea
Trial Supply Chain

Getting the IMP to the study sites

Procurement of Placebos / Comparators
Management and oversight of Depots pre/during/post trial
Ensuring adequate procurement of non-IMP supplies, either directly or in collaboration with the respective SMEs
Data Management/processing/analysis

Oversight to medical monitors

IRT oversight and management
EDC setup, monitoring, management and oversight
Data cleaning
Statistical analysis
Medical Writing

Development, revision and maintenance of

Protocols

IBs
ICFs and other patient facing materials
Lab Manuals and Pharmacy manual
relevant additional material necessary for trial applications and execution
Respective translations for materials when necessary
Publication of study results
Participant well-being / Pharmacovigilance

Redteaming of any safety concerns that could arise during the trial for participants

Development, revision and maintenance of SMP / MMP
Setup and monitoring of Safety Databases
Safety report writing
Management of DSMBs / IMMs
Review and continuous monitoring of any sort of adverse events in clinical trials
Trial testing

Identification of capable vendors who can run testing necessary for our trial.

Definition of the tests necessary, including physical examination, immunogenicity / safety labs, genetic testing, etc.
Definition of materials / procedures needed for appropriate testing (e.g. tubes, swabs, measurement devices, etc.)
Verification of the validity of the data delivered by testing vendors
Cross-department interfaces

Close collaboration with other departments across the whole company (e.g. manufacturing, preclinical science, operations, etc.)

Trial Quality Assurance & Monitoring

Oversight on development of relevant SOPs and WI in collaboration with QA team

Verification of ongoing use of SOPs / WI across the team
Regulatory Compliance

Continuous mutual updating regulatory teams as well as senior leadership

Collaboration on the definition of roadmaps for the trial team’s trajectory together with regulatory teams and senior leadership
Informing the rest of the team about progress made between the respective parties

The list above illustrates that you’ll need an interdisciplinary team. For many companies, multiple of these interdisciplinary teams are stacked on top of each other:

Global development teams think on the level of a therapeutic area (such as “Which vaccine type or blood pressure medication should we advance?”)
Clinical program teams focus on the evaluation of drug candidates (“How do we collect all the necessary data for getting this drug candidate approved?”)
Clinical study teams that execute individual clinical trials (“How do we get the study done efficiently and with the most valuable data?”)

Market approval

When a drug passes all these phases with good results, an application for market approval is submitted. In the US, this is submitting a New Drug Application (NDA) or Biologics License Application (BLA). Similar approval applications elsewhere are typically split into three main parts: preclinical data, detailed manufacturing methodology ensuring reproducibility, and and, of course, clinical trial information. Each section is approximately the same length, contributing to a “dossier” spanning _thousands _of pages. The FDA thoroughly reviews all the data and decides whether to approve the drug. This comprehensive nature is epitomised by this extensive table of contents spanning ten pages in the NDA.

Every year, the FDA processes ~400 Original Investigational New Drug applications, with an average of 38 new drugs gaining approval each year from 2010 to 2019. Of these, 3-4% are later withdrawn. Globally, the withdrawal rate hovers around 10% post-approval. Any changes to the manufacturing process, indications, or dosing require additional scrutiny and re-approval. The odds of a drug development program attaining approval is, one in seven, with rates ranging from 3.4% for oncology to a maximum of 33.4% for vaccines.

Post-Market Surveillance

The drug developer is expected by regulators to continue monitoring drug safety as it enters public use. This involves reviewing problem reports, regulating prescription drug advertisements, conducting routine inspections, etc. All data is summarised in Post-Marketing Surveillance Reports submitted by the sponsor, initially at six-month intervals, extending to three or more years as time progresses. The FDA can also require further efficacy studies while also conducting audits themselves to ensure ongoing compliance and drug safety. Both physicians and patients are encouraged to report any side effects they notice with a drug to regulatory agencies – see here for FDA’s and EMA’s contact information.

In our next piece, we will look into the origins of drug regulation, examining how historical events and key regulatory milestones have contributed to the current stringent practices in drug development. Understanding the evolution of drug regulation will provide a context to the existing rigorous processes and how they continue to shape the pharmaceutical landscape.

Drug Development Explainer Series

Tue, 17 Oct 2023 00:00:00 GMT

You’re standing in line at the pharmacy, impatiently tapping your foot as you await your meds. Little did you know, an army of regulators, scientists, and lawmakers have orchestrated this moment, having spent years making this errand possible for you. Welcome to the world of drug development, where safety measures and legalese are as ubiquitous as white lab coats.

Knowledge about drug development is often fragmented, expensive, and jargon-heavy. With this series we don’t want to debate the pros and cons of the system’s overall value; we want to explain how it works and how it came to work like this. The articles below aim to provide a comprehensive but concise introduction to all pharma's essential stakeholders, processes, and principles, connecting them to actionable insights and further resources:

This series was written in collaboration with James Smith and Kirsten Angeles. It is not a typical SEO article about pharma. Rather, it details the experience me and my ex-colleague James accumulated over the years. We have actively engaged with most of the references cited in the articles to inform decisions we made while being involved in drug development.

We hope you find this series valuable. Let’s dive in.

Stakeholders in Drug Development (1/6 Drug Development Explainer Series)

Tue, 17 Oct 2023 00:00:00 GMT

We discuss stakeholders in drug development, their checks and balances, and share recommendations for getting in touch with them:

Regulators like the FDA who set standards and conduct inspections to ensure compliance;
Sponsors such as pharmaceutical companies that oversee the process and assume responsibility for bringing a drug to market;
Vendors like Contract Research Organisations (CROs) and Contract Development and Manufacturing Organisation (CDMOs)that provide specialised services in areas like clinical trials and manufacturing and study sites who run trials and recruit participants;
Independent parties, particularly institutional review boards that evaluate ethics and safety.

**Author Note: **This series was written in collaboration with James Smith and Kirsten Angeles.

Overview

There are many layers to drug development, from clinical trials to chemical compounds, to long lists of regulations… But, at the centre, you’ll find people—teams of experts and individual patients, each playing a part in a complex symphony of innovation and regulation.

The interactions of these stakeholders are governed by regulatory frameworks ensuring checks and balances at every stage. The system is designed with the intention that it will lead to good results and is not dependent on a single company or even individual doing a good job – whether it's guidelines made by regulatory bodies, how sponsors qualify their vendors, the involvement of independent expert boards, or routine audits and inspections.

Working and maintaining good relationships with these varied, important stakeholders is critical for a drug developer. In the following sections, we discuss the key stakeholders, the people involved, and how you might contact them.

Note: The focus of this piece is on “regulated” stakeholders leading up to approval. Obviously other players that we don’t discuss here play important roles too, e.g. people who take part in trials and purchase the drug, medical journals that publish research studies, and downstream entities like pharmacies that sell the drug.

The table provides an overview of all stakeholders discussed in this article.

Domain	Stakeholder Category	Details
Regulatory body	International Guidelines Development	ICH / OECD / WHO Good X Practice (GxP, x = variable) guidelines with local implementation, e.g., Good Clinical Practice (GCP) - for the design and conduct of clinical trials; Good Laboratory Practice (GLP) - for design and conduct of nonclinical (animal) studies; Good Manufacturing Practice (GMP) - for the manufacture of the drug product. "GMP material" means a drug produced in compliance with GMP standards.
	Local Law and Regulatory Agencies	Federal law and regulatory bodies such as FDA, EMA, enforce compliance with standards, e.g. for the US (e.g., 21CFR210 for GMP, 21CFR58 for GLP, several parts of 21CFR for GCP).
Drug Developer ('Sponsor')	Pharmaceutical Companies	Biotech and established pharmaceutical companies. Pfizer and BioNTech, for example, "sponsored" the development of Comirnaty, the mRNA COVID-19 vaccine.
	Government Institutions	Institutions such as National Institutes of Health or NIH
	Investigator Initiated Trial	Individuals (commonly associated with an institution) exploring, for instance, new dosing regimens for a drug.
Vendors	Contract Development and Manufacturing Organisations (CDMOs)	Manufacturing facilities providing some or all components for the final drug product.
	Contract Research Organizations (CROs)	Animal study providers, specialised labs, Logistic / Distribution Providers, Consultants for diseases, regulation, applications, etc
	Study Sites	Physician Principal investigator (PI)
Independent boards	Institutional Review Boards	Institutional Review boards (IRBs)
	Independent Monitoring committees	Safety databases and Data and safety monitoring boards (DSMBs)
	Scientific Advisory Committees	E.g., Expert groups providing evaluations for regulators
	Patient Advocacy	Patient Advocacy Organizations and non-profit watchdog groups

Regulatory Bodies

The foundation of the drug development process is the regulatory framework provided by national and international organisations.

The World Health Organization (WHO),the International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) play significant roles in standardising practices across countries. This standardisation is intended to make it easier for drugs to reach a global market.

The WHO develops norms, standards, and guidelines to promote pharmaceutical quality assurance as part of its Constitution. For instance, it published its Laboratory Biosafety Manual in 1983, which is updated regularly and guides biosafety standards globally.
The ICH defines guidelines that are then implemented as laws in respective countries, providing the boilerplates for regulation. An example is the Good Clinical Practice (GCP) guideline (ICH E6), which sets international standards for clinical trials involving human subjects and is widely adopted by regulatory agencies worldwide.

Based on the WHO and ICH standards (such as GCP for clinical trials and Good Manufacturing Practice (GMP) for drug manufacturing) and other organisations which define adjacent standards such as Good Laboratory Practice (GLP) federal agencies like the Food and Drug Administration (FDA) in the United States and the European Medicines Agency (EMA) in Europe establish national guidelines that dictate the proper conduct of clinical trials, manufacturing, and marketing of medical products. .

These guidelines are often not just recommendations but codified into law, requiring strict compliance. For example, in the U.S., the FDA mandates compliance with some standards in Title 21 of the Code of Federal Regulations (CFR) with regulations outlining minimum manufacturing or laboratory practices and requirements.

How do regulatory bodies oversee standards?

Regulatory agencies conduct inspections and audits of ~all the other stakeholders mentioned below. Audits are often unannounced and thorough, so it is challenging to game them. A risk-based approach is used to determine who to audit, such that facilities selected for audit have the greatest potential for public health risk should they not be compliant. This includes earlier-stage manufacturing facilities or earlier-stage companies where a problem has been reported (e.g., if there was an unexpected safety issue in a trial). The FDA inspects pharmaceutical facilities globally, including those involved in manufacturing ingredients and finished products. Inspections will often take place during clinical trials, pre-approval, and then regularly post-approval (e.g., every few years). FDA also tests some drugs in their own labs or via contractors to check that they meet the specifications they claim to/should meet (about one percent of tested drugs fail to meet the established quality specifications). After approval, safety monitoring with real-world systems comes into play (like the FDA's Adverse Event Reporting System).

The FDA also conducts regular inspections of clinical trial sites, pharmaceutical companies, contractors, and Institutional Review Boards (IRBs) to ensure compliance with GCP guidelines. The FDA’s Office of Bioresearch Monitoring Operations (OBIMO) oversees domestic and foreign inspections for clinical and non-clinical research. In a 2022 report, they shared summaries from 766 inspections across multiple stakeholders that have to adhere to GCP, looking at failures to comply with protocol requirements, inadequate record-keeping, and lack of proper monitoring clinical investigations. IRBs were observed to have issues with meeting minutes, review procedures, and documentation.

Non-compliance to these regulations can be costly. From 2003 to 2016, pharmaceutical companies paid fines of $33B in the US. Violations of cGMP (Current Good Manufacturing Practices, the U.S. equivalent of GCP) can result in fines and jail time. A database lists the fines paid by pharmaceutical companies since 2000 at ~$112B for 1,217 cases. Most of these were due to false promotion and were settled under the False Claims Act. A recent example of a GCP violation and its consequences was Pfizer replacing half of its trial sites in one of its vaccine trials.

Vendors

Vendors are external organisations or service providers that assist in various stages of drug development. Many aspects of pharmaceutical development are outsourced to them, particularly by smaller biotechnology companies that can’t develop sufficient internal infrastructure for large-scale drug development (similar to other industries, e.g. smartphones). Their jobs dovetail neatly – while CDMOs bring the drug to life, and CROs evaluate its efficacy with the help of study sites who enrol participants:

Contract Research Organisations (CROs): CROs roles can be incredibly diverse, from pre-clinical research involving in vitro and animal studies to Phase I-IV clinical trials and post-marketing surveillance. Beyond doing the work, CROs bring expertise that is helpful for smaller companies without experience with drug development. They will know how to talk to regulators, and know what regulators tend to expect.
Contract Development and Manufacturing Organisations (CDMOs): CDMOs, on the other hand, handle, as the name suggests, the manufacture of the drug and the necessary manufacturing development. The term development might seem mysterious: it means creating the manufacturing process, any tests that need to be done on the drug to check its function, quality, etc, and testing all of that out. CDMOs can provide formulation, analytical services, blending, coating, converting, packaging, serialisation and shipment, though unfortunately, a single organisation sometimes (maybe even often) can’t do everything needed. Just to give one example, Lonza, a well known CDMO, helped develop Moderna’s mRNA COVID-19 vaccine.
Study sites: Study sites are locations where participants are recruited, and the clinical trial protocol is executed. A Principal Investigator (PI) leads study sites supported by a team of doctors, nurses, and other healthcare professionals. Typically, study sites are part of hospitals or outpatient practices, but some specialised sites exist that only do clinical trials. Recruitment usually happens at many study sites in parallel. Some study sites collaborate as Clinical Trial Network (CTN), which aims to pool resources under unified standard operating procedures and thereby streamline trial procedures.

There are also specialised contractors for all sorts of services we won’t go into detail here, such as animal testing facilities, clinical trial depots, specialised manufacturing techniques, or regulatory consulting in certain jurisdictions.

Some notes on vendors:

As a small biotech, your business is is not worth much to large CROs, whose main clients are big pharma, so expect to be working with their C or D teams – this means even if they have a good reputation, they might not be good for you. Working with a smaller CRO often means they are likely more invested in your project but might be unable to run larger trials later in development. It’s a bit of a trade-off.
All vendors can have really long waiting timelines (6-12 months for CDMOs and 1-6 months for CROs, similar scales for certain clinical trial sites).

How do vendors ensure the quality and safety of drug development?

It’s worth noting that CDMOs and CROs have an inherent economic interest in delivering services according to the sponsor's specifications (as contractors their reputation and future business depends on delivering). Some billion-dollar company CROs even specialise in providing monitoring services for sponsors, overseeing all their vendors’ activities. So how do sponsors and vendors guarantee the quality and safety of drug development?

Different risk management tools are routinely employed, e.g. Failure Mode and Effect Analysis (FMEA) and risk analysis plans to foresee potential risks.
To ensure accountability and compliance, every relevant action and document is time-stamped and stored for future scrutiny in a Quality Management System (QMS) to provide a transparent trail that can be audited for compliance at any time in the future.
When risks turn into issues, they are systematically addressed through Corrective and Preventive Measures (CAPAs). (We list some good resources to consult in another post if you want to learn more about quality management procedures.)
It’s worth noting that PIs are also professionally obliged to ensure that participants are cared for optimally and treated ethically.

Independent Parties

We just talked about how the previous stakeholders ensure and oversee quality, ethics, and safety in drug development – well, this is the sole purpose of independent parties.

Institutional Review Boards are the first stakeholder mentioned in the GCP guidelines. In the U.S., the National Research Act of 1974 mandates IRBs for any biomedical or behavioural research involving human subjects and the FDA for any drug development in humans.

Institutional Review Boards (IRBs): Are composed of at least five experts in pharmacology, medicine, bioethics, and patient representatives. They provide independent scientific review and ethical oversight, ensuring participants' dignity, rights, and welfare are respected and risks minimised. IRBs, either independent or affiliated with academic institutions or hospitals.
**Monitoring Boards: **In addition, monitoring committees put together by the sponsor for a single trial have become increasingly prevalent. In the simplest form, Independent Medical Monitors (IMMs) provide an independent examination of the ongoing safety events of a clinical trial. Data and Safety Monitoring Boards (DSMBs) are independent groups of experts reviewing both the safety events as well as data produced in a clinical trial.
Scientific Advisory Boards (SABs)**: **They form a temporary or permanent body associated with regulators or companies consisting of independent scientific experts. SABs provide independent overviews, analyses, and critiques for new technologies, approval considerations or regulatory procedures.
Patient organisations are commonly not considered part of the formal drug development process in guidelines or law. Still, they are heavily involved in advocacy and oversight of all other stakeholders, including regulators. Examples include the Patient-Centered Outcomes Research Institute (PCORI) or public advocacy groups and non-profit watchdogs like the Public Citizen’s Health Research Group.

How to get in touch with stakeholders?

Regulators: Interactions with regulators are not easy to broker, and you should aim only to discuss issues with regulators that you really need to.

Guidance documents: The first place to look is regulatory guidance documents which describe their (sometimes remarkably specific) thinking on a topic (see the FDA Guidance Document Database for specific examples). Though technically only guidance – that is, not legally binding, assume that regulators will expect you to follow the recommendations unless you have a strong scientific justification for not doing so.

**Consultants: **There are a lot of ex-regulators or ex-regulatory professionals from pharmaceutical companies or biotechs that you can engage as consultants. This can spare you from wasting precious time with regulators. The best way to find the specific person you need is to talk to other companies and get referrals. Check for experience with jurisdiction, indication, and modality (e.g. small molecule vs. biologic). Expect to pay between 200 and 500 USD per hour, sometimes significantly more for more senior people.
Official channels: When you’re ready to interact with regulators, there are different ways of doing so depending on the country you’re targeting and the stage of development. Ensure you check early on when it’s possible to interact with regulators and aim to take every opportunity to interact with them. Some are optional (like the INTERACT meeting), while others (such as the pre-IND meeting) are essentially required.
There are also various regulatory conferences worth attending:

FDA Conferences and Workshops,

BIO International Convention,
Drug Information Association Conferences,
Miscellaneous other regulatory conferences
**Sponsors: **Probably the most effective way of engaging with sponsors is through therapeutic-specific conferences. Ask around before booking the next conference that just ended up in your inbox – there are many conferences, and unfortunately, a large number are not valuable. JP Morgan Health Care Conference is one of many investment bank events, some of which are hard to get into.
**CROs & CDMOs: **Conferences, industry directories, references (strongly recommended), or direct website outreach may work best. They are almost always eager to talk to you. Ideally, you should get to non-business development people as quickly as possible so you can evaluate their actual expertise, and bring or talk to an expert on your side to help with that evaluation.
**Study Sites: **CROs will almost always be able to provide a list of study sites they have collaborated with in the past. In addition, databases such as Clinicaltrials.gov (or national equivalents) list the study sites that have enrolled participants for a given clinical trial. Beyond that, look for professional associations with member directories, look at notable faculty and alumni profiles on university and research institutions, or attend industry events where medical experts may be invited as speakers.
**Independent boards: **While directories exist, both study sites and CROs should be consulted for recommendations on which IRB to engage with. Given their independent nature, they can have vastly different timelines and be more conservative/progressive despite covering the same participant population and intervention. DSMBs are best staffed with senior experts in the field, e.g. from medical societies, academic publications, or from the personal network. Connect through professional bodies, educational institutes, or their official online avenues.
**Public engagement and non-profits: **Engage in community events, visit the organisations' websites, or even go to their social media.

I ran two donation workshops to determine where to give €50k

Sat, 12 Aug 2023 00:00:00 GMT

import groupPhotoImg from '../../../assets/blog/IMG_3635-scaled.jpeg'; import { Figure } from '@components/mdx';

Picking the right charity can be overwhelming - so many exist focusing on a myriad of causes. In this post I argue that collaborating with others to determine a donation portfolio is an underrated activity.

Story

In the past two years, I have invited friends to join me in determining the best way to allocate my annual donations. This workshop donation approach is sometimes referred to as a "Giving Game": For 1-2 hours 10-20 people compare charities on characteristics such as effectiveness, neglectedness, and room for funding. At the end, people vote which charity should receive the prize money - typically a few hundred dollars.

I first heard of Giving Games in 2018 and was enamored with the concept. However, I felt that 1-2 hours were insufficient to really dive into the important concepts and considerations. I wanted to create space for a more elaborate version and also increase the stakes. So in 2021 and 2022 I organized such “immersive” Giving Games and I am happy to report that the two events were a huge success on many levels!

We allocated a total of ~€50k, 30% of which came from participants. More notably, everyone agreed that they had significantly improved their understanding of how to do good in the world with money, regardless of their prior experience with donating. And I am pretty certain that my donations were allocated substantially better than without the events.

Absorb

Core lessons learned from two donation workshops (Giving Game Events) allocating ~€50k with groups of 10-15 people over 1-3 days:

Charity Portfolios Over Single Choices: If you allocate a lot of money between charities I encourage the use of portfolios instead of single choices. This approach made it easier to change minds and allow participants to accommodate their personal preferences.
You can actually change people's minds: Almost everyone moved at least 20% of their portfolio to a different organization during the event, many changed 50% or more across sectors.
Charity Assessment is hard: Charities should make it easier for donors to assess them by providing summarized yearly reports with quantified theories of change, but also very concrete and up-to-date funding needs for projects.
Tax Deductibility: This was a major issue, but we managed to mitigate some problems via donation swaps (I applaud organizations like Effektiv Spenden for expanding the ability to donate to charities worldwide from Germany.)

In November 2021, we as a group of 15 people spent a weekend in an AirBnb. I seeded the event with €10k, but my friends threw in money as well and we reached €20,080 (later we got an additional €10k from individuals I shared the results with). Discussions happened about preselected charities during four parallel world cafe sessions in the morning and afternoon. Each session was guided by someone with relevant experience in either the organization or the charitable sector. Participants then announced their personal portfolio , i.e. distributing the money across charities rather than being limited to one (you can see this in the picture at the top of the post). We then had time to convince each other to re-optimize our respective portfolios during a change-my-view discussion - and it worked! In the end we averaged out all portfolios (giving equal weight to each participant) to determine how the donation pool would be allocated.

In 2022, on very short notice, I decided to do another giving game. With the help of Thomas at Effektiv Spenden I designed it as a full day workshop with a seed amount of €25k. Although the structure was similar, I increased the number of charities from 9 to 14 and skipped the "Charity X" category from 2021, where individuals could propose an additional charity.

At both events, roughly half of the participants had previously investigated donating themselves, while the other half had not. However, everyone agreed (myself included) that they had never thought so deeply about donating money. I learned a lot, even through preparing for the events. Topics we discussed included marginal impact, funding constraints, personal relationship biases, established orgs vs experimental charity startups, and more.

If you're looking for a truly fun and unique experience, go for a giving game workshop.

Adopt & Adapt

Recommendations if you want to set-up a Giving Game yourself:

Utilize Existing Resources: Feel free to use the documents we used as a template for inspiration or those of others.
Timing and Scheduling: Aim for late November, as this is typically the giving season. Schedule early; perhaps you might even [send a quick message to a friend right now](http://[email protected]?subject=Planning%20a%20Giving%20Game%20(Weekend&body=Hi!I%20just%20read%20this%3A%20) who could help you organize a Giving Game Weekend. The total effort / event for me was probably around 5-10h.
Seek Expertise: Try to include people knowledgeable about different donating areas. The Effective Altruism movement has some really good resources on donating and charity evaluation, as well as local groups worldwide one can reach out to.
Have fun: Good food, walks, and games like the Trolley Problem card game are highly recommended.

This read was time well spent? Let me know in 10 seconds.

2-click-action: Sharing on 𝕏 twitter

“Appreciate @mxschons for writing an excellent post on Giving Games.”

AND / OR

[email protected]?subject=Blog%20Post%20Giving%20Game%202023-08&body=Hey%20Max%2C%20thank%20you%20so%20much%20for%20writing%20this%20blog%20post%20on%20Giving%20Games.">2-click-action: Send an 📨 email

“Hey Max, thank you so much for writing this blog post on Giving Games.”

Maximilian Schons

Resurrecting MeCell Maps with Claude Code

What happened

Why this matters to me

Try it

Appendix: Technical Migration

Comparing AI Labs and Pharmaceutical Companies

AI Labs vs. Pharma

Key areas of improvement for AI

Budget: Putting Money Where Your Quality & Safety Is

Ecosystem of checks and balances: Multi-party development where stakeholders have skin in the game

Table: Stakeholders in Drug Development vs. High-Risk AI Development

Scope: Implications of Different Customers and Ongoing Modification

Scientific Testing: AI Safety Studies for Assessing Dangerous Capabilities Reliably - Scale Matters

Appendix

Comparison of Drug and AI model development process

Overview of safety principles in clinical trials

Documents in a typical clinical trial application

My human cell map illustration MeCell now in the public domain

The 100h clinical trial

100h Pharmaceutical Response and Preparedness Teams

Broadly effective vaccines and therapeutics that are stockpiled for immediate deployment and evaluation when an outbreak begins

Extremely rapid targeted product development after the outbreak starts

100h Pharmaceutical Response and Preparedness (PREP) Team / Special forces

Milestones of the 100h timeline

Alvea: Case Study of the fastest biotech to go to in-human trials (5/6 Drug Development Series)

Alvea Story & Culture

Understand/verify industry best-practices

Determine your jurisdiction and exact specification

Find, evaluate, and parallelize vendors (or do it yourself if you can).

Origins of the Drug Regulation Industry (3/6 Drug Development Series)

US: Tracing the origins of drug regulation

US: Reactivity, Proactivity, and Regulatory Measures

Global Harmonisation for Drug Development

Global Milestones in Regulation: A Tour from Beijing to Russia

Brazil

China

India

Russia

Zimbabwe

What about countries that don’t have an independent drug regulatory system?

Recommended Literature and Courses (4/6 Drug Development Series)

Conclusions Drug Development Explainer Series (6/6)

The Blueprint of Drug Development (2/6 Drug Development Explainer Series)

Overview

Manufacturing

How does GMP look in practice?

Drug Discovery

Pre-clinical research

Clinical development

Market approval

Post-Market Surveillance

Drug Development Explainer Series

Stakeholders in Drug Development (1/6 Drug Development Explainer Series)

Overview

Regulatory Bodies

How do regulatory bodies oversee standards?

Sponsors

How do sponsors oversee the drug development process?

Vendors

How do vendors ensure the quality and safety of drug development?

Independent Parties

How to get in touch with stakeholders?

I ran two donation workshops to determine where to give €50k

Story

Absorb

Adopt & Adapt