A simple application that tracks major trial sponsors with unreported trials on ClinicalTrials.gov.

Application structure

The application is in two parts:

• In the data directory, a Jupyter notebook called demonstrates how we identify the trials that we think should have reported results, and how we check whether results have been reported either there or on PubMed.
• In the app directory, the front-end application presents the data we have found.

Get the data

The data/all.csv file contains our full results for all trials. This is a huge file: the easiest way to obtain the raw version may be to download the zipfile of the whole repo.

The Jupyter notebook has full details how we obtain and classify data (as does our paper). In summary, if the is_completed column is TRUE then we regard the trial as eligible for our study (based on phase, completion date, study type, disposition etc) and if the is_overdue column is TRUE then we have not found linked results either on ClinicalTrials.gov or on PubMed.

ClinicalTrials.gov is the source of the data used in this application. Data is used according to ClinicalTrials.gov's standard terms of use.

We make no guarantees that the data is current. Please refer to GitHub commit history to check when data was last processed. For up to date data, please refer to ClinicalTrials.gov.

All our code is available under the MIT licence. If you reuse any of our work, please cite us as follows: Powell-Smith A and Goldacre B. The TrialsTracker: Automated ongoing monitoring of failure to share clinical trial results by all major companies and research institutions. F1000Research 2016, 5:2629 (doi: 10.12688/f1000research.10010.1).

Update the app

• To update the data: first download raw data from ClinicalTrials.gov (instructions are in the Jupyter notebook). Then run the notebook, toggling REGENERATE_SUMMARY and REGENERATE_PUBMED_LINKS to regenerate the data from scratch. Running this notebook will automatically update the data used in the app.

• You can run the notebook from the command line with

  jupyter nbconvert --ExecutePreprocessor.timeout=-1 --to notebook --execute Examine\ unreported\ trials\ on\ ClinicalTrials.gov.ipynb

  • If that executes successfully, it will create a new file with the extension nbcovert.ipynb; copy this over the input notebook and commit
  • The notebook will create a number of files in data/; copy data/completed.csv to app/data/
  • Commit that; upload it to Webfaction (see pass for credentials)
  • Change "last updated" text on home page
  • Deploy by copying to Webfaction
  • Purge https://trialstracker.ebmdatalab.net/data/completed.csv in Cloudflare

• To work on the app: run tests with (in the data directory): nosetests tests/test_utils.py. To test the JavaScript: npm run test. To build the JavaScript: npm run watch (development) and npm run build (production).