Human behaviour in cancer screening trials: A variable that goes unmeasured
Cancer screening guidelines – which tests to use, when to begin screening, how often to screen – are built on the evidence generated by clinical trials. That evidence is typically presented as a question: ‘Does earlier detection improve outcomes?’
But clinical trial data is not shaped by the answer to this question alone. The behaviour of the patients enrolled in those trials – how consistently they follow study protocols, whether they remain engaged over months or years, and how they respond to the experience of participation itself – directly influences what the data shows. In oncology research, where screening recommendations can shape care decisions across large populations, that influence deserves more attention than it currently receives.
Why behavioural variability matters in clinical research
When patients enrol in a clinical trial, they bring with them a range of psychological traits, expectations, and life circumstances, all of which shape how they interact with the study. These factors determine whether participants complete the full study period, adhere to testing schedules, and accurately report their health status – affecting the quality and interpretability of the resulting data.
Three behavioural phenomena are particularly consequential in screening and prevention research:
Patient dropout is the most visible form of behavioural variability. When patients drop out of a clinical trial before it ends, it creates a reliability problem. In later-stage trials across the industry, a large chunk of participants don't make it to the finish line – and, in cancer research specifically, that number has been getting worse over time. The issue isn't just the raw number of dropouts; it's where they come from. Clinical trials work by splitting patients into comparable groups and comparing outcomes between them. When dropout happens unevenly – more patients leaving one group than the other – those groups stop being truly comparable. Researchers can apply statistical methods to account for the missing data, but those are workarounds, not solutions. The underlying problem remains.
Patient non-adherence compounds this challenge. In screening trials specifically, the benefit of any intervention depends on patients showing up reliably over time. Those who miss visits, skip preparation requirements, or disengage partway through a multi-year study create gaps in the data that can systematically underestimate how effective screening is, not because the test failed, but because it was never applied to a complete, representative population.
Observation-driven behaviour change – commonly referred to as the Hawthorne effect – adds a further layer of complexity that is rarely quantified in published trial results. When participants know they are being observed as part of a study, they often modify their behaviour in ways that reflect researcher expectations, rather than their normal patterns. In prevention and screening trials, where health behaviour is closely intertwined with the outcomes being measured, this can introduce additional noise into the data that is difficult to separate from genuine intervention effects.
The specific challenge for oncology screening evidence
Cancer screening trials face a particular version of this problem. These studies typically enrol participants for years, require repeated procedures, and depend on consistent protocol adherence across diverse patient populations. The landmark trials that generated current screening evidence – across lung, colorectal, breast, and cervical cancer – reflect not just the biology of detection, but the behavioural reality of who stayed in those studies and who did not.
That distinction matters when guidelines derived from trial data are applied broadly across clinical practice. If patients who complete a trial differ systematically from those who dropout in terms of health literacy, socioeconomic status, or baseline health behaviours, the resulting evidence may not represent the broader population for whom guidelines are written. The screening recommendations that emerge from that evidence may be accurate for the people who finished the study, but less applicable to the people who would have dropped out.
The limits of conventional approaches
The clinical research field has long recognised that dropout and non-adherence create methodological problems. The standard response has been to address these challenges indirectly: enrol more participants to preserve statistical power, extend study timelines to allow for attrition, or apply imputation methods to account for missing data.
These approaches are useful, but incomplete. They compensate for behavioural variability without explaining its sources. They preserve the statistical validity of a study without clarifying whether the people who completed it are representative of the people for whom the resulting guidelines will be written. And they do not address the observation-driven behaviour changes that may quietly inflate outcome measures throughout the study period.
As a result, the published evidence base for cancer screening contains an unmeasured margin of uncertainty that is rarely acknowledged in guideline development or clinical communication.
The role of behavioural science in addressing this gap
Advances in behavioural science and predictive modelling offer a more direct path to understanding the human factors that influence clinical trial data. Rather than treating dropout, non-adherence, and observation effects as unavoidable background variability, these approaches attempt to identify, measure, and account for the psychological and situational drivers behind them.
This includes identifying patient characteristics at baseline that predict engagement risk, developing models that flag participants likely to disengage before that happens, and building behavioural measurement into trial protocols from the start, rather than treating it as an afterthought. When behavioural data is incorporated into trial design and statistical analysis, researchers gain a more complete picture of what the evidence actually represents – and where its boundaries lie.
For oncology screening research specifically, this kind of behavioural insight could improve the reliability of the evidence that shapes public health recommendations. It could also support more accurate communication with clinicians and patients about what screening guidelines are based on and how confidently they can be applied across diverse populations.
A more complete picture of clinical trial evidence
Clinical trials remain the gold standard for generating evidence in oncology and across medicine. Their value depends on the quality of the data they produce, and the quality of that data depends on more than biological measurement alone.
Understanding behavioural variability – who drops out, why, and how participation itself shapes outcomes – is not a peripheral methodological concern; it is part of what it means to interpret clinical trial evidence responsibly.
As cancer screening guidelines continue to evolve and as trials grow larger and more operationally complex, integrating behavioural science into the research process is one of the clearest opportunities available to improve the evidence base on which clinical decisions are made.
About the author
Dr Dominique Demolle is the co-founder and CEO of Cognivia, a company leveraging machine learning and patient psychology to reduce data variability and improve the conduct of clinical trials. With decades of experience in clinical development, including her role as associate director of global early phase operations at Eli Lilly, and her efforts in building partnerships with pharmaceutical and biotech companies, she demonstrated the importance of understanding patient behaviour in drug development. Her commitment to creating innovative technology aims to prevent therapy failures due to treatment nonadherence, dropout, patient heterogeneity, and placebo response. Dr Demolle earned her PhD in Biochemistry from the University of Brussels and was honoured as one of the Most Inspiring People in Life Sciences by the 2022 PharmaVoice 100 awards.
