How accurate is the Clinical Trial Risk Tool?

How accurate is the Clinical Trial Risk Tool?

In order to develop the Clinical Trial Risk Tool, we had to conduct a quality control exercise on the components. Each parameter which is fed into the complexity model is trained and evaluated independently. For an overview of how the tool works, please read this blog post.

Datasets

I used two datasets to train and evaluate the tool:

  1. Manual dataset – this was a set of between 100 and 300 protocols which I read through individually and annotated key parameters such as the sample size. The number annotated per parameter varied between 100 and 300.
  2. ClinicalTrials.gov dataset – this was a much larger dataset of 11925 protocols downloaded from ClinicalTrials.gov. These came together with NCT ID, phase, pathology, SAP, number of arms and number of subjects, but the data was voluntarily provided by the researchers and in many cases is out of date or inaccurate.

By combining the two datasets I was able to get some of the advantages of a large dataset and some of the advantages of a smaller, more accurate dataset.

Validation

For validation on the manual dataset, I used cross-validation. For validation on the ClinicalTrials.gov dataset, I took the third digit of the trial’s NCT ID. Trials with values 0-7 were used for training, with value 8 were used for validation, and those with value 9 are held out as a future test set.

Validation scores for manual dataset

Validation scores on small manually labelled dataset (about 100 protocols labelled, but 300 labelled for number of subjects). You can reproduce my experiments using the notebooks from this folder.

ComponentAccuracy – manual validation datasetAUC – manual validation datasetTechnique
Condition (Naive Bayes: HIV vs TB)88%100%Naive Bayes
SAP (Naive Bayes)85%87%Naive Bayes
Effect Estimate73%95%Naive Bayes
Number of Subjects69% (71% within 10% margin)N/ARule based combined with Random Forest
Simulation94%98%Naive Bayes

Validation scores for ClinicalTrials.gov dataset

You can reproduce my experiments using the notebooks from this folder. As a sanity check I also trained a Naive Bayes classifier for some of these components to check that our models are outperforming a reasonable baseline.

ComponentAccuracy – ClinicalTrials.gov validation datasetBaseline Accuracy (Naive Bayes) – ClinicalTrials.gov validation datasetTechnique
Phase75%45%Ensemble – rule based + random forest
SAP82%Naive Bayes
Number of Subjects13%6%Rule based combined with Random Forest
Number of Arms58%52%Ensemble
Countries of InvestigationAUC 87%N/AEnsemble – rule based + random forest + Naive Bayes

In particular I found that the ClinicalTrials.gov value for the sample size was particularly inaccurate, hence the very low performance of the model on that value.

Other results of the validation

By far the most difficult model was the number of subjects (sample size) classifier.

I designed this component as a stage of manually defined features to identify candidate sample sizes (numeric values in the text), combined with a random forest using these features to identify the most likely candidate. Here is an output of the feature importances of the random forest model.

Similarly, the simulation classifier is a random forest that uses manually defined features of key words:

For any of the components, we also plotted a Confusion Matrix.

Confusion matrix for the baseline (Naive Bayes) phase extractor on the ClinicalTrials.gov validation dataset

Conclusions

Since each component is designed differently, it has been complex to validate the performance for clinical trial risk assessment.

However I have provided some Jupyter notebooks in the repository to run the validation and reproduce my results.

There is still much scope for improvement of several features, especially sample size.

Some parameters, such as simulation, were not available in the ClinicalTrials.gov dataset and so could only be trained and validated manually. We hope to be able to annotate more data for these areas.

5 Ways AI is Transforming Clinical Trial Risk Assessment

5 Ways AI is Transforming Clinical Trial Risk Assessment

This post originally appeared on Fast Data Science’s blog on LinkedIn. Clinical trials are vital for advancing medical innovation, yet they often face significant hurdles, including ensuring patient safety, adhering to regulatory requirements, controlling costs, and maintaining efficiency. Traditional risk assessment methods frequently need to be revised to address these complexities. Artificial Intelligence (AI) is transforming clinical trial management, offering data-driven solutions to predict and mitigate risks. AI-powered tools like the Clinical Trial Risk Tool have revolutionised trial planning and execution.

How AI and NLP Are Transforming Clinical Trial Risk Assessments

How AI and NLP Are Transforming Clinical Trial Risk Assessments

This post originally appeared on Fast Data Science’s blog on LinkedIn. Clinical trial protocols are often long, detailed documents—sometimes 200 pages—filled with vital information about sample size, treatment methods, and statistical plans. These protocols ensure the effective conduct of trials, but their complexity increases the time needed for manual reviews and the risk of human error. This is where Natural Language Processing (NLP) steps in. NLP enables machines to “read” unstructured data, such as clinical trial protocols, and extract key insights.

Transforming Clinical Trials with Fast Clinical AI

Transforming Clinical Trials with Fast Clinical AI

This post originally appeared on Fast Data Science’s blog on LinkedIn. Clinical trials, the backbone of medical science advancement, often grapple with high costs, complexity, and lengthy timelines. Fast Data Science presents Fast Clinical AI, a game-changing solution that harnesses the power of Natural Language Processing (NLP) and predictive modelling to tackle these challenges head-on. Streamlined Data Extraction and Analysis: Fast Clinical AI automates the extraction of critical information from trial protocols, significantly reducing manual efforts.