Can a Drug Be Approved Before We Know Whether It Helps People Live Longer?

Yes. A medicine can be approved before anyone knows whether it lengthens life. The answer sounds alarming only until the alternative is made visible. For a rapidly fatal disease with no effective treatment, waiting several more years for a survival result can deny present patients access to a therapy with strong early evidence. Yet acting early carries its own danger: a treatment may change a scan, laboratory value, or short-term event without helping people feel better, function better, or survive longer. Modern regulation sometimes accepts that uncertainty. The scientific and ethical question is how much uncertainty, for which patients, and under what obligation to resolve it.

The key device is a surrogate endpoint. An endpoint is the outcome used to judge a trial. Death, stroke, hospital admission, pain, mobility, and quality of life are clinical outcomes because they directly describe how a person survives, feels, or functions. A surrogate is an earlier measure used in their place. Blood pressure can stand in for future cardiovascular events in some settings. Sustained suppression of HIV can stand in for progression of infection. In oncology, tumour shrinkage or progression-free survival, the time before a cancer grows or a patient dies, may be measured sooner than overall survival. A biomarker becomes a useful surrogate only when evidence connects the treatment-induced change in that marker to the clinical outcome of interest [1].

That last clause does heavy work. A marker may be prognostic, meaning that people with a better value tend to fare better, without being a valid surrogate for treatment benefit. Tumour response illustrates the distinction. Patients whose tumours shrink may, on average, live longer than those whose tumours do not. A drug could still shrink tumours while causing lethal toxicity, selecting resistant cells, or producing a response too brief to affect survival. To validate a surrogate, researchers need to show across relevant trials that treatments which improve the surrogate also produce predictable improvements in the clinical outcome. The relationship can vary by disease stage, drug class, and treatment mechanism [2].

Surrogates can be extraordinarily successful. Trials of treatments for severe hypertension do not have to wait for enough strokes and deaths whenever a new medicine is tested, because a large body of randomised evidence has established blood-pressure lowering as a meaningful predictor of reduced cardiovascular risk for relevant drug classes. HIV research advanced rapidly once viral load and immune-cell measures could reveal treatment activity well before AIDS or death. In chronic myeloid leukaemia, molecular response provides powerful information. These examples show why a blanket rejection of surrogate endpoints would be perverse. The responsible question concerns the strength and transportability of the validation for the exact context in which it is used.

Cancer exposes the tension most sharply. Overall survival is clear and important, but it can require long follow-up and can be blurred when patients receive several later treatments or cross from the control group to the experimental drug. Progression-free survival can be observed sooner and is unaffected by some later therapies. It may also matter directly when progression brings symptoms or forces toxic treatment. Its interpretation still depends on the size of the delay, the burden of treatment, scan schedules, masking, and whether better disease control improves life as patients experience it. European guidance accepts progression-free survival as a possible endpoint in confirmatory cancer studies while requiring the choice and analysis to be justified [3].

How often does approval arrive without direct evidence of longer or better life? A cohort study examined 68 cancer indications authorised by the European Medicines Agency from 2009 through 2013. At the time of market entry, 39 lacked evidence that the drug improved survival or quality of life. After a median of 5.4 years from authorisation, only six of those 39 had subsequently shown such a gain, while uncertainties remained for the rest [4]. The study judged published evidence, so it could not capture every confidential or still-developing analysis. It nevertheless revealed the practical consequence of accepting intermediate endpoints: unanswered patient-centred questions can persist long after a medicine becomes available.

In the European Union, conditional marketing authorisation is one formal way to act earlier. It is reserved for medicines addressing serious or life-threatening illness, emergencies, or orphan conditions when comprehensive data are incomplete. Regulators must judge that the medicine addresses an unmet need, that a positive benefit-risk balance is established, that complete data can be provided later, and that immediate availability brings a benefit greater than the risk of waiting for those data. The authorisation lasts one year at a time. The company receives specific obligations with deadlines, often including completion of ongoing or new studies [5].

The UK's system is related but legally distinct. Since January 2025, the MHRA can grant a UK-wide conditional marketing authorisation under the Human Medicines Regulations when an unmet medical need exists and comprehensive data are expected soon. The agency can require further studies, impose conditions, and renew the authorisation annually [6]. This should not be confused with accelerated assessment, which shortens the regulator's review timetable without necessarily lowering the evidential standard, or with authorisation under exceptional circumstances, used when comprehensive evidence is unlikely ever to be obtainable. Labels that sound similar can describe different bargains.

The United States uses the term Accelerated Approval for a pathway that commonly relies on a surrogate endpoint considered reasonably likely to predict clinical benefit in a serious condition with unmet need. Traditional approval can follow if required trials verify the benefit; withdrawal procedures are available if they do not [7]. The American system dominates public discussion because it has authorised many cancer medicines this way, yet it is not a universal template. EMA conditional authorisation does not require a surrogate, and an EU medicine supported by a surrogate may receive a standard authorisation. The pathway, review speed, endpoint, and completeness of evidence are separate dimensions.

Conditional approval is best understood as a loan of uncertainty. Patients gain earlier access, and the developer acquires revenue and clinical experience. In return, the promised evidence must arrive. Starting the confirmatory trial before approval can shorten the period of doubt and avoid a later problem: once the medicine is available, patients and clinicians may be reluctant to accept randomisation, particularly to a control arm. The trial should also answer the unresolved question. Repeating the same weak surrogate may confirm biological activity while leaving survival, symptoms, or function uncertain.

Post-authorisation obligations do not always proceed exactly as first written. A study of 26 medicines conditionally authorised in the EU between 2006 and 2016 examined 69 specific obligations. Regulators changed the description or due date of 39 percent of the obligations after authorisation, and 23 percent of fulfilled obligations with an available submission date arrived after the latest specified deadline [8]. Amendments can be scientifically reasonable when recruitment, standards of care, or methods change. Frequent delay still matters because people are already receiving the product while the evidential debt remains outstanding.

Olaratumab, sold as Lartruvo, shows the system doing something painful but necessary. In 2016 the medicine received an EU conditional authorisation for certain adults with soft-tissue sarcoma. An early randomised study had suggested that adding it to doxorubicin substantially extended survival, and the company was required to confirm the result. The larger ANNOUNCE trial did not. Median survival was 20.4 months with olaratumab plus doxorubicin and 19.8 months with placebo plus doxorubicin; the hazard ratio was 1.05, which offered no evidence of a survival advantage. EMA recommended withdrawal of the marketing authorisation in 2019 [9].

Withdrawal can sound like proof that early access was a mistake. That conclusion is too simple. A conditional system has to permit two honest outcomes. Some medicines will confirm their benefit and move to standard authorisation. Others will fail, leading to restriction or withdrawal. A pathway in which nothing ever leaves the market would suggest that the test after approval has lost force. The ethical failure arises when uncertainty is hidden, confirmatory studies are poorly designed or delayed without good reason, or ineffective treatment continues after the evidence has turned.

The consequences of delay extend beyond clinical uncertainty. A medicine can become embedded in guidelines and clinical habits. Patients may endure toxicity, travel, monitoring, and opportunity costs. Health systems may pay a high price while foregoing other care. Evidence from the US programme has shown that confirmatory studies have sometimes continued to use surrogate outcomes rather than measuring clinical benefit directly [10]. Recent US reforms have strengthened the FDA's authority to require confirmatory trials to be underway and to expedite withdrawal. Those reforms respond to a general design problem that matters on both sides of the Atlantic: post-market evidence needs a credible clock and a consequence.

Survival itself is not always the only outcome that counts. A medicine might relieve severe symptoms, prevent irreversible disability, keep someone out of hospital, or allow treatment at home without extending life. Conversely, a small survival gain accompanied by debilitating toxicity may be a poor trade for some patients. Trials should therefore measure outcomes appropriate to the disease and ask patients which changes matter. Patient-reported outcomes can capture pain, fatigue, or daily function, provided they are collected rigorously and missing data are handled honestly. The phrase clinical benefit should never become a euphemism for survival alone.

Regulatory approval and NHS funding add another layer. NICE may evaluate a newly licensed medicine using immature survival data and a model that projects what happens beyond the observed trial. The farther that model travels beyond the data, the more assumptions drive the result. Managed access can allow time-limited NHS use while additional evidence is collected, with a planned review and commercial terms to manage financial risk. That arrangement can reduce uncertainty if the new data are fit for purpose. Routine records cannot automatically repair a trial that lacks a valid comparator or a meaningful outcome [11].

For clinicians and patients, the crucial information is more specific than whether a medicine is approved. What outcome improved? By how much, compared with what, and for how long? Was the endpoint directly meaningful or a surrogate? How strong is the evidence linking that surrogate to the hoped-for benefit in this disease and drug class? What serious harms were observed? Which confirmatory study is due, and when? A clear consent conversation can accommodate uncertainty, but only when the uncertainty has been made legible.

A society that insists on perfect knowledge before any treatment reaches patients will sometimes wait too long. A society satisfied by an encouraging scan can expose patients to costly, harmful, or ineffective care. Conditional and accelerated pathways occupy the difficult space between those errors. Their legitimacy rests less on the speed of the first decision than on the discipline of the second: a serious surrogate, a rigorous confirmatory trial, a public deadline, and the willingness to change course when the promised benefit does not appear.