FAQs

TrialKey employs sophisticated AI models and machine learning algorithms to simulate clinical trials and predict their success probabilities. By analysing over 1,300 variables from historical trial data, the platform offers predictions with over 92% accuracy. The AI continuously refines its models based on new data, ensuring up-to-date and reliable predictions.

TrialKey’s model simulates each clinical trial 10,000 times, utilising machine learning and deep learning techniques. Due to the adaptive nature of these models, the probability of success may vary slightly with each simulation as the system continuously refines its predictions. This results in a probability distribution, with over 92% accuracy, providing a more comprehensive and reliable prediction of success.

TrialKey offers comprehensive competitor analyses, including success probability scores, benchmarking graphs, and detailed trial conclusions. These insights allow sponsors to understand market trends, compare their trials with those of competitors, and make data-driven decisions to improve trial strategies. This competitive intelligence is crucial for optimising trial outcomes and staying ahead in the clinical research landscape.

TrialKey has been refined based on feedback from experts in the CRO, pharmaceutical, and medical investment sectors. The platform’s proprietary search algorithm and precise condition-based searches, enhanced by natural language processing, provide comprehensive insights into competitor activities. These include technology extraction, heatmaps of research areas, and success probability estimates, enabling stakeholders to allocate resources effectively and anticipate outcomes.

TrialKey categorises variables into distinct groups based on their relevance to clinical trial outcomes. These categories include patient demographics, treatment protocols, study designs, and more. Each category contains specific variables, which are then analysed using advanced AI models. We present this information in a summarised form to ensure our intellectual property remains protected while providing transparency. To explore the full list of our +1300 variable categories, visit our methodology page.”

TrialKey’s Explainable AI framework effectively distinguishes between causality and correlation by elucidating marginal effects. These insights are further validated by subject matter experts, and various causality tests, such as pairwise analyses, are employed to ensure the robustness of our findings.

Recognising the gaps in data reporting (with only a 16.5% reporting rate), TrialKey leverages Natural Language Processing (NLP) to analyse journal articles and link trials across different phases. This enhances target coverage, allowing us to infer high confidence in trial outcomes even when earlier phase data is missing, ensuring accurate and meaningful AI-driven insights.

While ClinicalTrials.gov relies on user updates, TrialKey proactively monitors the change history of trials to maintain data freshness and relevance. Continuous updates ensure that insights provided are based on the most current data available.

To ensure statistical rigour and backtest its prediction models across different trial phases, TrialKey employs several advanced techniques grounded in machine learning and real-world data analysis:

1. Extensive Historical Data and Real-World Evidence (RWE): TrialKey draws on data from over 350,000 clinical trials, spanning two decades. This extensive dataset incorporates a wide range of variables, allowing the model to reflect real-world conditions more accurately than traditional statistical methods. By utilising RWE, TrialKey’s predictions align closely with the outcomes observed in actual patient populations  

1. Comprehensive Simulations: Each clinical trial is simulated thousands of times, with up to 10,000 iterations. This thorough process accounts for the inherent variability in trial conditions, ensuring robust statistical accuracy across different phases of trials, from Phase I (safety) through to Phase III (efficacy). These simulations help refine the model and validate its predictions under different scenarios  

1. Backtesting with Real-World Results: TrialKey’s model is continuously backtested against real-world trial outcomes. The model’s success in predicting the results of high-profile COVID-19 vaccine trials, such as those for Pfizer and Moderna, highlights its reliability. This backtesting process ensures that TrialKey’s predictions are accurate and reflect real-world trial performance  

1. Explainable AI for Transparency: TrialKey employs explainable AI techniques, such as Shapley contributions, which make the model’s predictions transparent. These techniques show which variables (such as patient demographics or dosage levels) most affect trial outcomes, allowing researchers to understand and validate the model’s predictions. This transparency also helps to refine trial designs  

1. Cross-Phase Validation: TrialKey applies its rigorous methodology across all phases of clinical trials. Whether in the early stages of Phase I or the more complex Phase III trials, the model adapts to different phases while maintaining accuracy. This cross-phase consistency ensures that the predictions are reliable at every stage of the trial process  

Through these techniques, TrialKey guarantees that its prediction models maintain statistical rigour, providing trustworthy insights across various clinical trial phases. 

At TrialKey, we determine confidence intervals and margins of error through rigorous statistical methods, ensuring that our predictions are both accurate and reliable. Each simulation is run 10,000 times, capturing a wide range of possible outcomes based on the data inputs and trial design variables. From this, we calculate a 95% confidence interval, which represents the range within which the true trial outcome is likely to fall, given the data and assumptions used. 

By generating this confidence interval, we quantify the margin of error associated with our predictions, giving users a clear understanding of the variability and uncertainty in the trial design. This helps trial managers and sponsors make informed decisions with a deeper insight into the potential outcomes and risks. 

These methods ensure that our predictions are statistically sound, providing confidence that the simulated trial outcomes will closely align with real-world results. 

TrialKey has developed a set of data visualisations specifically designed for clinical trials, helping users easily explore and interpret analysed data. These visualisations are tailored to the unique requirements of clinical trials, providing insights into critical trial parameters like recruitment rates, patient demographics, success probabilities, and timelines.

The visualisation suite includes:

• Trial Simulation Outputs: Interactive graphs and charts that let users compare different trial designs and outcomes based on various scenarios.

• Patient Recruitment and Retention: Visual tools that display projected recruitment and retention rates over time, assisting trial managers in making informed decisions.

• Success Probability Trends: Graphical representations of success probabilities across various phases and conditions, offering a clear view of potential outcomes.

• Comparative Analysis: Visuals that enable side-by-side comparisons of multiple trial arms or different trial designs, aiding decision-making.

These tools are designed to provide clear, actionable insights, making complex data easier to digest and allowing users to optimise their trial designs efficiently.