From Promising Discovery to Prescribed Medicine

A promising drug discovery usually begins as an answer to a narrow question. Can this molecule block an enzyme that a tumour depends on? Can an antibody silence an inflammatory signal? Can a modified strand of genetic material replace an instruction that a cell has lost? At that moment, the object in the laboratory is a candidate, not a medicine. Its effect may be striking in a dish and still disappear in a human body. It may reach the right organ but damage another. It may work at a dose that cannot be manufactured reliably. The long route to a prescription exists because each of those possibilities has to be confronted with evidence [1].

The familiar image of a pipeline makes the process look smoother than it is. Real development behaves more like a series of locked gates. A candidate must have a plausible biological target, a usable chemical or biological form, acceptable exposure in the body, tolerable toxicity, convincing clinical activity, and a manufacturing process that produces the same material time after time. Failure at any gate can close the programme. An analysis of candidates from four large pharmaceutical companies found that properties such as solubility, dose, and tissue exposure helped explain why projects survived or failed, while lack of efficacy and safety problems remained central causes of attrition [2].

Discovery can start in universities, hospitals, biotechnology firms, large companies, or partnerships among them. Researchers may screen thousands of compounds for an effect, design a molecule around the structure of a target, adapt an existing drug, or follow a clue from human genetics. A laboratory hit is then altered repeatedly. Medicinal chemists try to strengthen its useful action while reducing unwanted interactions. Pharmacologists ask how much reaches the bloodstream, how long it remains, what breaks it down, and whether enough arrives at the intended tissue. Formulation scientists face a practical question that can decide the whole project: can the active substance become a stable tablet, injection, inhaler, infusion, or other product that people can actually use?

The target itself may be wrong. A protein can be associated with a disease without controlling it. Animal models can reproduce one feature of a human illness while missing the feature that matters most. Cells grown under laboratory conditions lack the immune system, metabolism, age, and competing diseases of a person. A candidate can therefore do exactly what its designers intended and still fail to improve health. This is one reason human genetic evidence has become valuable in target selection: variants that naturally alter a biological pathway can offer clues about the consequences of changing it over a lifetime. Even then, genetics strengthens a hypothesis rather than issuing a guarantee.

Before first use in people, developers assemble a non-clinical package. Studies examine pharmacology, the way the body absorbs and clears the candidate, and toxic effects in appropriate laboratory systems or animals. The programme depends on the product and the proposed human exposure. A one-off treatment for advanced cancer poses different questions from a medicine intended for daily use over decades. International guidance links the duration and design of non-clinical studies to the clinical trial being proposed, with the aim of identifying hazards early enough to protect participants without demanding every conceivable experiment before human research begins [3].

Permission to start a clinical trial does not mean that the candidate has been shown to work. In the UK, a clinical trial of an investigational medicinal product requires regulatory authorisation and a favourable ethical opinion. The application covers the scientific protocol, safety information, manufacture of the trial product, and arrangements for participants. The rules introduced in April 2026 retain the basic bargain of clinical research: risks should be proportionate and managed, consent should be informed, and the study should be capable of answering a worthwhile question [4]. A weak trial exposes people to inconvenience and risk while leaving the original uncertainty intact.

Clinical development is divided into phases, though modern programmes often blur their boundaries. First-in-human studies usually begin with a small number of participants and focus on tolerability, exposure, and dose. Healthy volunteers are common, but patients may be enrolled from the outset when toxicity is expected or the disease is severe, as in many cancer trials. Phase 2 studies look more directly for biological or clinical activity and refine dose and population. Phase 3 studies are usually larger and are designed to provide confirmatory evidence against placebo, current care, or another relevant comparator. Random allocation helps make the groups comparable. Blinding, where feasible, reduces the chance that expectations influence care or assessment.

The phase labels can hide the hardest intellectual work. Investigators must decide whom to enrol, what outcome to measure, how long to follow participants, and what size of difference would matter. A treatment tested only against placebo may be less informative when effective care already exists. A composite outcome can increase the number of events while combining consequences of very different importance. A biomarker may give an early answer while leaving uncertainty about pain, disability, hospital admission, or survival. Statistical significance shows how compatible the data are with a specified null model. It does not, on its own, establish that the benefit is large enough for patients to notice or accept the harms.

Most clinical candidates never become approved medicines. In a large analysis linking clinical trials and drug development programmes from 2000 to 2015, the estimated probability of a candidate moving from phase 1 to approval was 13.8 percent, with wide differences among disease areas and types of product [5]. Such estimates depend on how programmes, failures, and missing outcomes are classified, so they should not be read as a timeless law. Their message is sturdier than any single percentage: early promise is common, while convincing translation into a useful medicine is uncommon.

Failure also has several meanings. A candidate may cause liver injury, heart rhythm changes, infection, or another unacceptable harm. It may be safe enough but ineffective. A dose that engages the target may be intolerable, while a comfortable dose does too little. Recruitment can stall, the trial design can become obsolete as care changes, or a competitor can reach the same patients first. In one study of 640 novel therapeutics that entered late-stage development between 1998 and 2008, 54 percent failed during pivotal development, mainly because of efficacy or safety concerns; some negative results also remained unpublished, weakening the collective ability to learn from failure [6].

If the pivotal evidence is persuasive, the developer submits a marketing authorisation application. For a medicine assessed through the European Union's centralised procedure, the dossier includes how the product is made and controlled, what non-clinical studies found, the benefits and harms observed in people, and a plan for managing risks. Independent rapporteurs and expert committees interrogate the evidence, request analyses, inspect manufacturing and trial compliance where needed, and judge the benefit-risk balance for the proposed use [7]. In the UK, the Medicines and Healthcare products Regulatory Agency performs the licensing role through national procedures.

Safety is never an isolated score. A severe adverse effect might be unacceptable for a mild, self-limiting condition yet tolerable when a medicine offers a meaningful chance against an otherwise fatal disease. Regulators also examine whether uncertainty can be reduced through warnings, laboratory monitoring, restricted prescribing, pregnancy prevention, or further studies. The authorised indication matters as much as the molecule. Evidence may support use only after other treatments fail, only in people with a particular biomarker, or only at a dose lower than the one first imagined. The label is therefore a boundary around the claim the evidence can sustain.

Quality carries equal legal weight with safety and efficacy. Regulators need evidence that the commercial product matches the material studied in trials and can be made consistently under good manufacturing practice. For a simple small molecule, that includes identity, purity, stability, and control of impurities. A biological medicine produced in living cells can vary with changes in cell line, culture, purification, or storage. Gene and cell therapies add chain-of-identity and handling problems. A spectacular trial result cannot rescue a product whose contents or potency are unreliable. Medicine manufacture is part of the scientific claim, because every patient should receive a product sufficiently similar to the one that generated the evidence.

A licence permits marketing. It does not automatically place a medicine in routine NHS care. In England, the National Institute for Health and Care Excellence appraises clinical effectiveness and cost effectiveness. Its committees compare the new treatment with relevant current practice, consider patient and professional evidence, and estimate health effects using quality-adjusted life years when a full cost-utility analysis is appropriate. The question is distributive as well as clinical: what health might the NHS gain with the medicine, and what care might be displaced by spending the same money here [8]? Scotland has its own medicines consortium, and arrangements differ across the devolved health systems.

This gap between authorisation and access can puzzle patients. The regulator asks whether evidence supports a positive benefit-risk balance for a defined population. A health technology assessment body asks how the treatment compares with available options and whether its price represents reasonable value within a finite service. Both may reach internally coherent conclusions from the same trial record because their responsibilities differ. Price discounts, managed access agreements, or further data collection can sometimes reduce the uncertainty. Prescribing guidance then translates a population decision into choices for one person, whose other conditions, preferences, previous treatments, and tolerance of risk may differ from the trial average.

Approval is also an evidence milestone rather than the end of inquiry. Trials may include too few people to detect a very rare harm, last too briefly to reveal delayed effects, and enrol participants healthier or less diverse than those treated later. Pharmacovigilance combines suspected adverse-reaction reports, scientific literature, company data, electronic health records, registries, and post-authorisation studies. European regulators can require risk-management measures or additional studies and can update product information, restrict use, suspend a medicine, or withdraw its authorisation if the balance changes [9]. Effectiveness can shift too, as resistance develops, standards of care improve, or longer follow-up changes the estimate of benefit.

The development story is therefore less a triumphant march from bench to bedside than an organised reduction of uncertainty. Laboratory work tests whether an idea is biologically and technically plausible. Clinical trials ask what the intervention does in people under specified conditions. Regulators decide whether its demonstrated benefits justify its known and potential risks. Health systems judge comparative value, and surveillance watches what happens at scale. At no point does uncertainty reach zero. A prescribed medicine is a candidate that has accumulated enough reliable evidence for a particular use, together with a system designed to notice when the evidence needs to change. Its remaining uncertainties are named, monitored, and weighed against the alternatives available to the patient.