How Is Medicine Made: Discovery, Trials, and Production

Every medicine you pick up at a pharmacy has traveled a path that typically spans a decade or more, moving from a molecular idea in a research lab through years of testing and regulatory review before arriving on a production line. The journey involves thousands of scientists, billions of dollars, and a failure rate that would bankrupt most industries outside of pharma. Understanding how this process works helps explain why new drugs cost what they do, why some promising treatments never make it to patients, and why the pills and injections that do reach you are overwhelmingly safe and effective.

Finding a Target

Drug development usually begins not with a molecule but with a question: what specific biological process is causing the disease? Researchers look for a “target,” which is a particular protein, enzyme, receptor, or gene that plays a key role in the illness they want to treat. Identifying that target is one of the most consequential decisions in the entire pipeline, because everything that follows depends on it. Modern strategies for finding targets range from computational modeling and gene-editing screens to large-scale analysis of proteins and metabolic pathways across tissues and body fluids.¹

Once a target is identified, researchers need to confirm it actually matters. A protein might look involved in cancer cell growth, for example, but blocking it might do nothing useful, or worse, might harm healthy cells too. Pinning down both on-target effects and off-target effects often requires combining multiple approaches: biochemical experiments, genetic studies, and computational analysis together, rather than relying on any one method alone.²

With a validated target in hand, chemists begin designing or screening molecules that can interact with it. In one typical example, researchers synthesized dozens of indole-based compounds and tested them against more than 60 types of human tumor cells, eventually identifying candidates that inhibited cell growth at very low concentrations and pinpointing the enzyme dihydrofolate reductase as the likely target.³ This kind of iterative chemistry, synthesizing a batch of related molecules, testing them, refining the best performers, and repeating, is how “lead compounds” emerge. A lead compound is simply a molecule that shows enough promise to justify investing more resources into developing it further.

Preclinical Testing

Before any experimental drug is given to a person, it goes through preclinical studies designed to assess whether the compound is safe enough to test in humans. This stage usually involves cell-based (in vitro) experiments and animal studies. Researchers examine how the drug is absorbed, distributed through the body, metabolized, and excreted, collectively known as its ADME properties. They also look for signs of toxicity at various doses and check whether the drug actually does what it is supposed to do in living organisms.

Animal testing has been a cornerstone of preclinical work for decades, but its predictive value is a source of ongoing debate. Analysis suggests that animal models are poor predictors of drug safety in humans, raising questions about whether the field needs to rethink its reliance on them.⁴ Drugs can behave very differently in mice or monkeys than they do in people, and a clean safety profile in animals does not guarantee safety in humans. Still, animal studies catch enough problems early to remain a regulatory requirement in most countries, and for certain products like cell-based therapies, they remain the only way to monitor for immediate adverse reactions such as respiratory distress or organ damage before moving into people.⁵

Clinical Trials

If a drug survives preclinical testing, it enters clinical trials, where it is tested in humans for the first time. Clinical trials proceed in phases, each designed to answer a different set of questions. The progression is deliberately cautious: early phases involve a handful of volunteers, later phases involve thousands.

Phase 1 trials are primarily about safety. A small group of healthy volunteers or, in oncology, patients with advanced disease, receive the drug at escalating doses. The goal is to find the highest dose the body can tolerate without unacceptable side effects. Traditional methods for dose escalation, like treating small groups of three patients and watching for toxicity before moving up, are widely used but have known limitations, including a tendency to expose a significant number of participants to doses that are too low to be effective.⁶ Newer statistical models are gradually replacing these older designs, especially in immunotherapy trials where the relationship between dose and toxicity does not follow the straightforward pattern seen with traditional chemotherapy drugs.

Phase 2 trials expand the number of participants and begin testing whether the drug actually works against the disease. Researchers measure specific endpoints, such as tumor shrinkage, symptom improvement, or viral load reduction. These trials also continue to monitor safety at the doses selected in Phase 1. Phase 2 is where most drug candidates fail: the compound either does not produce a strong enough therapeutic effect or causes side effects that were not apparent at Phase 1’s smaller scale.

There is also a well-documented pattern in which Phase 2 results look more impressive than what later Phase 3 trials show. A meta-analysis of rheumatoid arthritis drugs found that Phase 2 outcomes systematically overestimated subsequent Phase 3 results, largely because of differences in how patients were selected for the trials.⁷ Phase 2 patients tend to be more carefully screened and homogeneous, making the drug look better than it will perform in the broader, messier population enrolled in Phase 3.

Phase 3 trials are the large-scale, randomized, controlled studies that regulators rely on most heavily when deciding whether to approve a drug. They typically involve hundreds to thousands of patients, compare the new drug to either a placebo or the current standard treatment, and generate the safety and efficacy data that appear on the drug’s label. These trials are expensive and can take years to complete.

Getting Through Regulatory Review

After Phase 3 trials, the company submits a massive application to a regulatory agency, such as the FDA in the United States or the EMA in Europe, containing all the preclinical and clinical data. Regulators comb through the evidence to determine whether the drug’s benefits outweigh its risks for the intended use. Standard review can take a year or more.

For drugs that address serious or life-threatening conditions, several expedited pathways exist. In the US, these include fast track, breakthrough therapy, accelerated approval, and priority review designations. Expedited approval does reduce the time to market, and evidence suggests that drugs approved under these designations tend to be more innovative and deliver higher clinical benefit than drugs receiving standard approval.⁸ But faster timelines come with trade-offs: the clinical trials supporting these approvals more often use small, non-randomized, open-label designs, and required post-approval studies to monitor for safety problems are frequently delayed or never started.

These expedited pathways now contribute to a majority of all new drug approvals in the US.⁹ That shift has practical consequences: more drugs are reaching patients faster, but the evidence base at the time of approval is thinner than it once was for many products. The assumption is that post-market surveillance will fill in the gaps, though as we will see, that system has its own limitations.

How Small-Molecule Drugs Are Manufactured

Once a drug is approved, actually making it at scale is its own engineering challenge. Most traditional medicines are small molecules, meaning their active ingredient is a relatively simple chemical compound that can be synthesized through chemical reactions. Scaling up from a few grams in a lab to tons in a factory is not as straightforward as running the same recipe in bigger equipment. Chemical reactions that work perfectly at bench scale can behave differently at industrial volumes, where heat dissipation, mixing efficiency, and impurity profiles all change.

One particularly demanding step in manufacturing is maintaining the drug’s stereochemistry, the three-dimensional arrangement of atoms in the molecule, because even a mirror-image version of the same compound can have completely different biological effects. Continuous-flow processes, where reactions run through specialized equipment nonstop rather than in large batches, have shown promise for handling this. In one case, a key asymmetric hydrogenation step in an active pharmaceutical ingredient was scaled to kilograms using continuous-flow catalysis, achieving over 95% conversion and over 98.6% enantioselectivity on every pass, with no detectable catalyst contamination in the product.¹⁰

Beyond synthesis, the physical form of the drug matters enormously. Many drug molecules can crystallize into different arrangements called polymorphs, and the particular polymorph used affects how well the drug dissolves and how much of it your body absorbs. Drugs with low water solubility are especially prone to variable bioavailability depending on their crystal form, which creates both an opportunity (choosing the right polymorph to improve absorption) and a headache (keeping that form stable throughout the product’s shelf life).¹¹

How Biologics Are Made

Biologics, which include monoclonal antibodies, vaccines, and mRNA therapies, are manufactured very differently from traditional pills. They are produced by living cells rather than chemical synthesis, which makes the process more complex and more sensitive to small variations in conditions.

Monoclonal antibodies, for instance, are typically produced in large bioreactors filled with mammalian cells, often Chinese hamster ovary (CHO) cells, that have been genetically engineered to secrete the desired protein. Getting high yields from these cells requires careful optimization of the cell line itself, the nutrients in the culture medium, bioreactor conditions like temperature and oxygen levels, and the purification steps that extract the antibody from the cell broth.¹² Because the product is made by living organisms, even minor changes in manufacturing can alter the drug’s properties, which is why regulators treat manufacturing changes for biologics far more carefully than for small-molecule drugs.

mRNA-based products, like the COVID-19 vaccines, follow a different route. Production starts with a DNA template that serves as the blueprint. An enzyme called RNA polymerase reads that template and generates the mRNA strand. The residual DNA is then digested away, the mRNA molecules are chemically capped (a modification that helps cells read the instructions), and the final product is purified to remove truncated fragments and other contaminants before being dissolved in a storage buffer and frozen.¹³ The mRNA then needs to be packaged inside lipid nanoparticles, tiny fat-based bubbles that protect it and help it enter cells. This formulation step uses rapid mixing of an ethanol phase containing the lipids with an aqueous phase containing the mRNA, a technique that produces uniform particles with high encapsulation efficiency and is scalable to commercial volumes.¹³

Quality Control and the Cold Chain

Every step of manufacturing operates under strict Good Manufacturing Practice (GMP) regulations, which are designed to ensure that the final product is safe, pure, and consistently what the label says it is. GMP guidelines are harmonized across countries through cooperation between regulatory authorities and the pharmaceutical industry, aiming to achieve exceptional levels of quality, security, and effectiveness in health products.

Quality control does not end when the drug leaves the factory. Many medicines, especially biologics and vaccines, must be kept within narrow temperature ranges during shipping and storage. Improper handling at any point can degrade the product and reduce its therapeutic properties.¹⁴ Different vaccines and biologic products have different temperature requirements, and the pharmaceutical cold chain, the network of refrigerated trucks, warehouses, and pharmacy storage, must be carefully managed at every step.¹⁵ A broken cold chain is one of the most common reasons vaccines lose potency in parts of the world with unreliable infrastructure.

What Happens After Approval

A common misconception is that once a drug is approved, the testing is over. In reality, post-market surveillance, sometimes called Phase 4, is a critical part of the drug’s lifecycle. Clinical trials, even large Phase 3 studies, typically involve only a few thousand people and last a few years. Rare side effects that affect one in ten thousand patients, or problems that emerge only after years of use, simply cannot be detected in trials of that size and duration.

Spontaneous reporting systems, where doctors and patients voluntarily report suspected adverse reactions, are one of the main tools for catching these problems. In the US, the FDA’s MedWatch system serves this function. These systems are effective at revealing unusual or rare adverse events, and reports can often be sufficient to establish causality for distinctive reactions. But they have blind spots: they are not reliable at detecting side effects that occur long after the drug was taken, or effects that look like diseases people get anyway.¹⁶ If a heart medication slightly increases the risk of a type of stroke that already happens in older adults, spontaneous reports alone will struggle to pick that up, because doctors may not connect the two events.

More sophisticated methods supplement voluntary reports. Large healthcare databases that track which medications patients are taking and what diagnoses they receive over time can reveal statistical signals that spontaneous reporting misses. Researchers have used the FDA’s Adverse Event Reporting System database to perform disproportionality and Bayesian analyses, which essentially ask: is this particular side effect being reported for this drug more often than you would expect by chance?¹⁷

What a New Drug Actually Costs to Develop

The cost of bringing a new drug to market is one of the most debated numbers in healthcare, and the estimates vary widely depending on what you count. The headline figures you hear, often in the range of one to two billion dollars, include not just the direct spending on the successful drug but also the cost of all the failed candidates that never made it, plus the “cost of capital,” meaning the money the company could have earned if it had invested those funds elsewhere instead of tying them up in research for a decade.

A 2024 study estimated the mean direct cost of developing a new drug at roughly $173 million. Once the cost of failed programs was factored in, that rose to about $516 million, and adding the cost of capital pushed it to roughly $879 million.¹⁸ A separate analysis using a different methodology found that the median total cost, including adjustments for inflation, capitalization, and discontinued products, was about $708 million, while the mean was substantially higher at $1.31 billion, reflecting that a few extremely expensive programs pull the average up.¹⁹ An earlier study estimated the median capitalized investment at about $985 million.²⁰

The range across disease areas is also enormous. Drug development for some fields, such as genitourinary conditions, can be done for well under $100 million in direct costs, while drugs for pain and anesthesia can exceed $290 million before even accounting for failures.¹⁸ These variations matter for understanding why the pharmaceutical industry gravitates toward certain diseases and away from others: the economic return has to justify the investment risk, which is one reason rare diseases and tropical infections have historically attracted less commercial interest.

Where the Raw Materials Come From

The active pharmaceutical ingredients (APIs) in your medicine may have traveled through several countries before reaching you, and the supply chain for these raw materials is more concentrated and fragile than most people realize. For generic drugs in the US market, India, China, and Italy are the top producers of APIs, while only about 14% are manufactured domestically. Roughly a third of APIs come from a single manufacturing facility worldwide, and another third come from just two or three facilities.²¹

For antibiotics specifically, the concentration is even starker. Between 2020 and 2024, China was the source of roughly 63% of the total imported volume of antibiotic APIs entering the US.²² This kind of concentration creates vulnerability: a factory closure, a natural disaster, or a geopolitical dispute in a single country can ripple through the global drug supply and cause shortages thousands of miles away. The COVID-19 pandemic exposed exactly these weaknesses when lockdowns in key manufacturing regions disrupted supply chains for both APIs and finished drug products.

Artificial Intelligence in Drug Discovery

One of the biggest shifts happening across the pharmaceutical industry is the growing use of artificial intelligence and machine learning to speed up the earliest and most uncertain stages of drug development. AI is being applied to virtual screening, the process of computationally evaluating millions or billions of chemical compounds to find ones likely to bind to a particular drug target, before ever synthesizing them in a lab.

A recently developed platform called RosettaVS uses structure-based virtual screening that models how both the drug candidate and its protein target flex and move, and it outperforms many existing methods on standard benchmarks. In real-world tests, the platform screened multi-billion compound libraries against two unrelated protein targets and found active hit compounds for both, with hit rates of 14% and 44% respectively, all with binding strengths in the low micromolar range.²³ For context, traditional high-throughput screening campaigns in the lab often have hit rates well under 1%, so these numbers represent a meaningful improvement in efficiency.

Deep learning approaches for predicting how a drug docks into its target protein have also been accelerated by breakthroughs in protein structure prediction. Tools that predict protein shapes from amino acid sequences alone have enhanced both the accuracy and speed of virtual screening, opening the door to screening compound libraries against targets whose three-dimensional structure was previously unknown or poorly resolved.²⁴

Meanwhile, on the manufacturing side, continuous production processes, where material flows through equipment without stopping between stages, are gaining ground as an alternative to traditional batch manufacturing. Continuous methods have been shown to be as reliable as batch production while saving time and money.²⁵ They also allow for real-time quality monitoring, which is harder to do when you are mixing up a large vat and testing it only at the end.

Companion Diagnostics and Targeted Medicine

An increasingly important trend in drug development is the co-development of companion diagnostics: tests that identify which patients are most likely to benefit from a specific drug based on their individual biology. This is most advanced in oncology, where therapies targeting specific molecular pathways have become standard. Companion diagnostic-associated approvals have been growing steadily, making up 43% of all oncology drug indications by 2022.²⁶

The appeal is intuitive: if you can test a tumor for a specific genetic mutation and then prescribe a drug designed to exploit that mutation, the treatment should work better and cause fewer unnecessary side effects. The data support this. A systematic review and meta-analysis found that targeted cancer drugs paired with companion diagnostics had significantly lower rates of severe side effects and treatment discontinuation compared to targeted drugs without them.²⁷ The improvement in tolerability was striking: the odds of grade 3 or 4 adverse events roughly doubled for targeted drugs without companion diagnostics, compared to only a modest increase when companion diagnostics were used to select patients.

Companion diagnostics also reshape clinical trials themselves. By enrolling only patients whose tumors carry the relevant biomarker, trials can be smaller and still detect meaningful effects, because the “noise” from patients who were never going to respond has been removed. This is especially useful when the biomarker is relatively rare. Co-development of the drug and the diagnostic adds logistical complexity to the approval process, but it enables initial approvals for molecularly defined solid tumors that would otherwise be drowned out in a broader, unselected patient population.²⁶

Ethics and Informed Consent in Drug Testing

Every clinical trial participant today must give informed consent before enrolling, meaning they receive a clear explanation of what the study involves, its potential risks and benefits, and their right to withdraw at any time. This is such a routine part of modern medicine that it is easy to forget it was not always the norm. The concept of informed consent has evolved significantly across both medical practice and research over centuries.²⁸

Today, consent forms are overseen by independent ethics committees or institutional review boards that evaluate whether the trial design is fair, whether participants are adequately informed, and whether vulnerable populations are being protected. There is a growing movement toward “tailored consent,” which adapts the format and level of detail to the individual participant rather than handing everyone the same dense legal document.²⁹ This is especially relevant for trials involving elderly patients, people with cognitive impairments, or communities with limited health literacy, where the standard 20-page consent form often fails to actually inform.

1
PubMed Central. The Art of Finding the Right Drug Target: Emerging Methods and Strategies
2
PubMed Central. Target identification and mechanism of action in chemical biology and drug discovery
3
PubMed. Structural optimization of indole based compounds for highly promising anti-cancer activities: structure activity relationship studies and identification of lead molecules
4
PubMed Central. Limitations of Animal Studies for Predicting Toxicity in Clinical Trials: Is it Time to Rethink Our Current Approach?
5
PubMed Central. Preclinical safety testing of Mreg_UKR
6
PubMed. Contemporary dose-escalation methods for early phase studies in the immunotherapeutics era
7
PubMed Central. Efficacy outcomes in phase 2 and phase 3 randomized controlled trials in rheumatology
8
PubMed Central. Special FDA designations for drug development: orphan, fast track, accelerated approval, priority review, and breakthrough therapy
9
PubMed. FDA designations for therapeutics and their impact on drug development and regulatory review outcomes
10
Organic Process Research & Development. Enabling the Scale-Up of a Key Asymmetric Hydrogenation Step in the Synthesis of an API Using Continuous Flow Solid-Supported Catalysis
11
PubMed Central. Polymorph Impact on the Bioavailability and Stability of Poorly Soluble Drugs
12
PubMed Central. Cell culture processes for monoclonal antibody production
13
Nature Reviews Materials. Lipid nanoparticles for mRNA delivery
14
ECS Transactions. Temperature Excursion Management in Cold Supply Chain of Pharmaceutical Products
15
International Journal of Pharmaceutical and Healthcare Marketing. Investigating the performance of the sustainable cold supply chain in the pharmaceutical industry
16
PubMed. Postmarketing surveillance and adverse drug reactions: current perspectives and future needs
17
PubMed Central. Exploring and comparing renal adverse effects between PARP inhibitors based on a real-world analysis of post-marketing surveillance data
18
PubMed Central. Costs of Drug Development and Research and Development Intensity in the US, 2000-2018
19
JAMA Network Open. Use of Clinical Trial Characteristics to Estimate Costs of New Drug Development
20
PubMed Central. Estimated Research and Development Investment Needed to Bring a New Medicine to Market, 2009-2018
21
PubMed. Competition And Vulnerabilities In The Global Supply Chain For US Generic Active Pharmaceutical Ingredients
22
JAMA Health Forum. US Antibiotic Importation and Supply Chain Vulnerabilities
23
PubMed Central. An artificial intelligence accelerated virtual screening platform for drug discovery
24
Accounts of Chemical Research. Advancing Ligand Docking through Deep Learning: Challenges and Prospects in Virtual Screening
25
PubMed. Analytical comparison between batch and continuous direct compression processes for pharmaceutical manufacturing using an innovative UV-Vis reflectance method and chemometrics
26
PubMed. Strategic Integration of Companion Diagnostics in Precision Oncology: Key Factors, Drug Development Timelines, and Regulatory Insights from U.S. FDA Approvals
27
PubMed Central. Influence of companion diagnostics on efficacy and safety of targeted anti-cancer drugs: systematic review and meta-analyses
28
PubMed Central. A Modern History of Informed Consent and the Role of Key Information
29
PubMed Central. Evolution of informed consent in research: From the Hippocratic Oath to the tailored consent