A Q&A with Geert Jan Groeneveld, CEO of Centre for Human Drug Research, on why the strongest clinical development programmes are organised around questions, not simply phases.
Drug development is usually described as a progression through phases. Phase I establishes initial safety and pharmacokinetics. Phase II explores efficacy and dose. Phase III confirms benefit and risk in a larger patient population.
It is a useful structure. But it can also create the impression that successful development means moving smoothly from one phase to the next.
The scientific reality is different.
A drug can advance through several studies while important uncertainties remain unresolved. Does sufficient drug reach its intended site of action? Does it produce the biological effect it was designed to produce? Does that translate into a clinically meaningful effect? Is there enough separation between desirable and undesirable effects? And how much will all of this vary across the intended patient population?
At CHDR, we have long advocated approaching clinical development through questions such as these. We call this Question-Based Drug Development. The central principle is simple: rather than asking which conventional study comes next, identify the most important uncertainty in the programme and design the experiment that can resolve it.
Clinical Leader spoke with CHDR’s CEO about five questions that reveal whether a development programme is actually accumulating the knowledge required for success.
What is wrong with organising drug development around Phase I, II and III?
Nothing, provided we remember what the phases are for. They are useful regulatory and organisational descriptions. The problem starts when they become the scientific strategy. You sometimes hear a development team say, “We have completed Phase I; now we need a Phase II study.” My first question would be: what did Phase I actually tell you, and what is the most important thing you need to learn next? Those are not necessarily the same question.
Drug development is fundamentally a succession of decisions under uncertainty. At each point, you are deciding whether to invest more money, expose more participants or patients, manufacture more drug, expand into additional countries, or perhaps stop the programme. The quality of those decisions depends on the quality of the knowledge you have generated. A phase label tells you relatively little about that.
So what is Question-Based Drug Development?
It means starting with the scientific uncertainties rather than the conventional study sequence. In our earlier work on question-based clinical development, we described five broad categories of questions that recur across many development programmes:
Does the drug reach its intended site of action?
Does it produce the intended functional effect?
Does it produce the intended clinical effect?
Does an effective exposure have an acceptable safety profile, in other words, is there a therapeutic window?
How will variability in the target population influence the drug’s effects and its development?
The exact formulation will differ between compounds. A monoclonal antibody, a CNS-active small molecule and a gene therapy will not have identical development questions. But these five categories are a useful stress test. If a programme cannot answer them convincingly, it is worth asking whether progression into larger trials is really reducing uncertainty or simply postponing it.
Question 1: Does the drug reach its intended site of action?
This sounds straightforward, yet it is one of the most important questions in clinical pharmacology. We often measure drug concentrations in plasma because plasma is accessible. But the pharmacological target may be in the brain, a tumour, the skin, a specific immune compartment or another tissue altogether. Plasma exposure is therefore not necessarily the same as exposure at the site where the drug must act.
That distinction matters enormously when interpreting a negative study. Imagine that a compound does not produce the expected pharmacological effect. One explanation is that the biological hypothesis is wrong. Another is that the compound did not sufficiently engage its target. But there is a third possibility: sufficient active drug never reached the target in the first place.
Those explanations have very different consequences for the future of the programme. The first question should therefore establish, as far as possible, the relationship between dose, systemic exposure and exposure at the relevant site of action. Sometimes that can be measured directly. In other situations, imaging, tissue sampling, biomarkers, active-metabolite measurements, receptor occupancy studies or pharmacokinetic modelling may be needed.
The methods differ. The question does not. Before interpreting what a drug does, we should have confidence that enough of it gets to where it needs to be.
Question 2: Once it gets there, does it do what we think it does?
Reaching the site of action is necessary, but it is not sufficient. The next question concerns functional pharmacology. Does the compound inhibit the enzyme, occupy the receptor, neutralise the cytokine, activate the pathway or produce the physiological change predicted from the preclinical work?
Ideally, we do not want just a yes-or-no answer. We want to understand the relationship between exposure and effect. At what concentration does the pharmacological effect begin? Does it increase with exposure? Does it plateau? How quickly does it occur? How long does it persist after concentrations begin to decline?
This is where a scientifically rich early clinical study can generate much more information than a study designed predominantly around safety and plasma pharmacokinetics. Appropriate pharmacodynamic biomarkers, experimental medicine models, imaging, physiological assessments or functional tests can sometimes demonstrate human pharmacology very early.
This is also where unexpected findings become informative. A drug may produce effects that were not predicted from preclinical experiments. Those effects can be important both scientifically and for understanding future tolerability.
The objective is to move from saying, “We administered 100 mg and it was tolerated,” to saying, “At these exposures, the compound produces this magnitude and duration of the intended pharmacological effect in humans.” That is a much stronger foundation for the next decision.
Question 3: Does that pharmacology translate into the intended clinical effect?
This is the critical translational step. A drug can reach its target and demonstrate convincing pharmacology, and still fail as a medicine. Changing a biomarker is not the same as helping a patient. The third question therefore asks whether the functional effect actually influences the disease in the way the therapeutic hypothesis predicts. That does not necessarily mean jumping immediately to a large conventional efficacy trial.
Depending on the mechanism and indication, there may be opportunities to investigate disease-relevant effects in relatively small, carefully selected patient populations. Experimental medicine models, sensitive clinical endpoints, challenge tests, physiological measures or translational biomarkers can sometimes provide important evidence before a large proof-of-concept study.
What we are trying to build is an increasingly coherent chain:
dose → exposure at the site of action → functional pharmacology → clinical effect
The more of that chain we understand, the more interpretable subsequent results become.
Suppose the clinical effect is disappointing, but we know the drug reached the target and produced essentially maximal functional pharmacology. That result says something important about the therapeutic hypothesis. If the clinical effect is disappointing and we do not know whether the drug ever produced adequate pharmacology, we have learned much less.
This distinction is important because a negative result should be informative. A failed study that leaves the team unable to distinguish a failed molecule from a failed mechanism or an inadequate dose is an expensive way to remain uncertain.
Question 4: Is there a useful therapeutic window?
For a medicine, pharmacological activity alone is not enough. The desired effects must occur at exposures that are acceptably safe and tolerable. That is why we prefer to think in terms of a therapeutic window, rather than treating efficacy and safety as completely separate development streams.
On one side of that window is the exposure-response relationship for the pharmacology and clinical effects we want. On the other are the unwanted pharmacological effects, adverse events, physiological changes, laboratory abnormalities or other limitations that emerge as exposure increases. The relationship between the two is what matters.
A compound that produces meaningful pharmacology well below the exposures associated with unacceptable effects is very different from one in which efficacy and toxicity emerge at almost the same concentrations. This is also why “maximum tolerated dose” is not always the most interesting concept. The important dose may be considerably lower: the dose that produces sufficient target engagement or functional effect while preserving an adequate margin to undesirable effects. And again, exposure is often more informative than nominal dose. Two individuals receiving the same dose may achieve substantially different concentrations.
Understanding those concentration-effect relationships early can tell us whether a compound has room to become a useful medicine before very large numbers of patients are exposed.
Question 5: What happens when we introduce the variability of the real target population?
This question is frequently treated as something for later development. That can be a mistake. Once a compound moves from highly selected participants into a broader patient population, variability appears everywhere.
Patients differ in pharmacokinetics because of body composition, organ function, age, genetics, concomitant medication and many other factors. They may differ in target expression or disease biology. The same exposure may not produce the same pharmacodynamic effect in everyone. And patients who carry the same clinical diagnosis may have substantially different underlying mechanisms.
The question is therefore not simply, “Does the average patient respond?” It is: what determines who responds, who does not respond and who experiences unwanted effects?
A therapeutic window that appears comfortable in a homogeneous early study may become much narrower when variability in exposure and pharmacodynamics is introduced. Biomarkers, pharmacogenetics, deep phenotyping, PK/PD modelling and careful analysis of individual concentration-effect relationships can therefore be valuable much earlier than many development plans assume.
We should not expect to explain every source of variation. But we need to understand enough of the important ones to judge whether the proposed dose, population and treatment strategy remain realistic.
Do these five questions have to be answered in that order?
No, and this is a very important part of the concept. Question-Based Drug Development should not become another rigid five-stage system. The right order depends on the compound.
For one drug, site-of-action exposure may be the major uncertainty and should be resolved as early as possible. For another, safety may dominate. For a highly heterogeneous disease, understanding the target population could become important surprisingly early.
In the work in which we formally developed the question-based approach, we demonstrated exactly this point. When a hypothetical development team estimated the cost and probability of successfully answering each question, the mathematically optimal route was different from the sequence the team had initially planned. In particular, variability in the target population needed to be considered much earlier.
That is the real point. The five questions are not a new set of phases. They are a way of identifying where the uncertainty and risk actually reside.
Does every question require its own clinical study?
Again, no. One study can answer parts of several questions, and one question may require evidence from several studies. That distinction can lead to much more efficient development. For example, an early clinical study could simultaneously characterise pharmacokinetics, assess target engagement, quantify a functional pharmacodynamic effect, explore concentration-related adverse effects and investigate potential sources of variability. The value comes from integrating those measurements.
Rather than simply generating separate safety, PK and PD outputs, the study begins to describe how exposure connects to both desirable and undesirable effects over time. That is particularly valuable in early development, where relatively small, data-dense experiments can sometimes answer questions that become much harder and more expensive to disentangle later. The question should determine the study design, not the other way around.
Isn’t this simply what good drug development teams already do?
To some extent, yes. Most experienced drug developers instinctively think about these questions. The problem is that the formal development plan and its investment decisions are still often organised predominantly around studies and phases. That can have subtle consequences.
If the objective is “complete Phase I”, the team naturally optimises the programme for completing Phase I. If the objective is “establish whether therapeutically relevant target engagement can be achieved with an acceptable safety margin”, the study may look quite different.
The question-based formulation forces the team to be explicit. What exactly are we uncertain about? How important is that uncertainty to the eventual success of the programme? What experiment could resolve it? How much will that experiment cost? And what will we do differently depending on the result? Those are scientific questions, but they are also investment questions.
How does this approach improve the economics of development?
The usual assumption is that the cheapest development path is the one that minimises expenditure today. That is not necessarily true. An additional early experiment can seem expensive if it is viewed only as an extra study. But if it resolves the uncertainty most likely to stop the programme, it can prevent a much larger investment later.
Conversely, delaying a difficult question does not remove the risk. It simply allows more capital to accumulate behind it. This was an important reason we originally linked question-based development to project valuation. A programme has greater value when its development route takes account of the specific scientific risks of that compound rather than relying only on average probabilities of progressing from Phase I to Phase II and Phase III.
The objective is therefore not to perform more experiments. It is to answer consequential questions at the point where the answers have the greatest impact on future decisions.
What does a drug programme that is “built to succeed” actually look like?
It is not a programme in which every study is positive. That would be an unrealistic definition of success. A well-designed programme should sometimes produce an early, convincing reason to stop. If the drug cannot adequately reach its site of action, cannot produce the expected functional pharmacology, lacks the intended clinical effect, has an unusably narrow therapeutic window or is too unpredictable in the target population, discovering that early is valuable.
The costly outcome is not necessarily failure. The costly outcome is late failure after the programme had opportunities to resolve the critical uncertainty earlier. So when looking at a development programme, I would ask five questions:
Does it get there?
Does it do what we expect?
Does that help the patient?
Can it do so safely?
And can we predict what will happen across the patients we ultimately want to treat?
If those questions are being answered progressively and deliberately, the programme is building knowledge. And that is a much better definition of progress than simply entering the next phase.
Conclusion
Phase I, Phase II and Phase III remain useful ways of describing clinical development. But they should not determine its scientific logic. Drug development is not inherently linear. Different compounds fail for different reasons, and the most important uncertainty at any given moment depends on the drug, mechanism, indication and target population.
Question-Based Drug Development starts from that reality. It asks whether the drug reaches the intended site of action, produces the intended functional effect, generates the desired clinical effect, has a workable therapeutic window and behaves predictably enough in the target population.
It then asks a second, equally important question: which of those uncertainties should we resolve next? That shift may appear subtle, but it changes the purpose of clinical studies. Instead of designing a trial because it is the conventional next step, we design it because its result will enable a better decision.
The strongest development programmes are therefore not necessarily those that move most smoothly from Phase I to Phase II to Phase III.
They are the ones that make it progressively harder to be surprised.
