How to stop clinical trials failing before they start

Most people working in drug development have lived through a trial that failed. Not failed in the scientific sense – the molecule simply did not work – but operationally: enrolment missed by 40%, a protocol amendment at month eight, or a primary endpoint shifted because the original was unachievable. These failures are expensive, demoralising, and often avoidable.

The uncomfortable truth is that the majority of late-stage operational failures leave clear fingerprints at the design stage. Protocol complexity is the single most consistent predictor of dropout rates, amendment frequency, and timeline slippage. Sponsors who treat design review as a rubber-stamp exercise are, statistically, setting themselves up for problems they will be scrambling to fix two years later.

This post is about what rigorous up-front design looks like in practice, why it is a competitive advantage rather than a bureaucratic overhead, and what sponsors consistently get wrong – and right – at the point when a protocol can still be changed at minimal cost.

The complexity problem

There is a well-established relationship between protocol complexity and operational performance. The table below, drawn from published benchmarks across Phase II and Phase III trials, illustrates what happens as endpoint count and procedural burden increase.

Sources: Tufts Center for the Study of Drug Development (2016); Sertkaya et al., Clinical Trials (2016). Endpoint count used as a proxy for overall protocol complexity.

The data is striking because the relationship is not linear – it is exponential. A trial with 16 or more endpoints is not twice as likely to run into trouble as one with 8; it is substantially more so. Each additional endpoint adds investigator training time, data-management complexity, screen-failure risk, and patient burden. The cumulative effect compounds quickly.

The Tufts Center for the Study of Drug Development found that the average Phase II protocol in 2015 contained 14.8 endpoints—more than double the number reported in 1999. That escalation has continued. Understanding why it happens is the first step to reversing it.

Why protocols get too complex

It is rarely scientific intent that inflates protocols. In most cases, complexity accumulates through a familiar organisational process: each internal stakeholder – clinical, commercial, regulatory, medical affairs – adds one or two items that matter to their function. No individual request seems unreasonable. The aggregate, however, is a protocol that pressurises site teams on reliable execution without deviation, and pressurises patients, making dropping out more common.

The scale of this problem is confirmed by Tufts CSDD data: approximately one in four procedures in Phase II and III protocols support non-core endpoints – assessments that are not essential to answering the primary scientific question.

Industry protocol reviewers consistently identify three categories of redundant endpoints:

1. Inaccurate Measurements: Assessments that cannot be measured with sufficient accuracy in the trial setting.
2. Duplicative Items: Endpoints that duplicate information already captured elsewhere in the protocol.
3. Internal Wishlists: Items added to satisfy internal stakeholder requests rather than scientific or regulatory necessity.

Each category adds procedural burden without adding evidential value.

“It is not the science that fails. It is the organisational process that builds the protocol around it.”

Challenging these inclusions requires diplomatic courage – and a clear framework. The question to ask for each endpoint is not “Is this useful?” but “Is this worth the operational cost of including it?” When that cost is made visible – increased screen-failure rates, longer visit windows, more site training, higher dropout risk – the conversation changes.

Feasibility: The step most sponsors skip

Feasibility assessment is the most consistently underinvested activity in Phase II development. Bell et al. (2019) found that fewer than half of sponsors conducted structured feasibility assessments before site selection in Phase II trials, yet inadequate feasibility is cited as a contributing factor in the majority of enrolment failures.

The consequences are well documented:

• Shortfalls in Recruitment: A 2023 analysis of over 2,500 randomised controlled trials found that fewer than half met their pre-specified enrolment target, with the median shortfall exceeding 30% of the planned sample size.
• Over-Optimistic Forecasts: A 2022 peer-reviewed study confirmed that principal investigators show poor calibration across recruitment, timelines, and outcomes—consistently overestimating favourable results for their own trials.

Meaningful feasibility goes beyond asking potential sites whether they have seen patients with the relevant diagnosis. It requires validating that:

• The patient population exists in sufficient density within the catchment area.
• The proposed visit schedule is compatible with how patients actually live and travel.
• Sites have the staffing, equipment, and capacity to run the protocol as written.
• The standard of care at each site is aligned with protocol assumptions.

The last point is underappreciated. A protocol that is feasible at a major academic medical center may be operationally impossible at a community site – not because the site is less capable, but because the protocol was designed with the academic center in mind. Sponsors who do not explicitly design for their full site range end up concentrating enrollment in a small number of sites, which concentrates risk, drives up per-patient costs, and reduces the generalizability of their data.

The Site capacity crisis

That risk is compounded by a structural capacity problem at sites. The global clinical research coordinator workforce shrank by approximately 28% between 2017 and 2024 (from ~56,000 to 40,500) as experienced staff left for sponsors and CROs offering higher pay and flexible working. While turnover at well-managed institutions declined to ~15% by 2024 following acute pandemic-era spikes of up to 61%, the structural headcount loss remains. Protocols designed without accounting for constrained site capacity land on already-strained teams – compounding the dropout and amendment risk that complexity creates.

Adaptive designs: Flexibility without compromise

Adaptive trial designs – where pre-specified rules allow modifications based on interim data – have moved from theoretical frameworks to practical tools. Both the FDA and EMA have issued guidance supporting their use in Phase II, showing that well-designed adaptive trials reach primary endpoints faster and at a lower cost than traditional fixed designs.

Key adaptive features for Phase II include:

• Sample-Size Re-estimation: Allows early adjustment if effect size assumptions prove optimistic.
• Seamless Phase II/III Designs: Automatically rolls a successful Phase II into Phase III without a protocol restart.
• Response-Adaptive Randomization: Shifts allocation toward arms showing early efficacy signals.

These designs require careful statistical pre-specification, early regulatory engagement, and robust data infrastructure. However, sponsors who build adaptive options into their initial design process deliver faster development timelines and more efficient budget usage.

What sponsors should do before protocol lock

Structural decisions about protocol design are easiest – and cheapest – when made before submission, not after enrollment has started. Three disciplines reduce risk materially:

1. Stress-test every endpoint against operational cost. Ask: “What does this cost in screen-failure risk, site training, visit duration, and amendment probability?” Any endpoint without a clear answer is a candidate for removal.
2. Conduct structured feasibility. Validate the visit schedule against patient lifestyles and confirm the planned country mix has the actual staffing capacity to execute.
3. Design the home/site visit split before lock. Work with an expert team like MRN to evaluate every visit. Visits requiring specialized imaging, infusions, or physician-scored measures belong at the site; most observations, sampling, and data collection can be delivered at home by a trained nurse. Moving viable visits to the home reduces patient burden, extends site reach, and removes a major source of dropout risk.

Rigorous up-front design is not a trade-off between speed and quality; it is a commercial decision. Sponsors who design well move faster, spend less, and generate cleaner data. Those who do not pay the difference – with interest – in amendments, delays, and dropout management.