Open Source version of the time-to-onset signal detection algorithm

Almost a decade ago, I designed a signal detection method that screens the time-to-onset (TTO) distribution of events reported post vaccination and flags those presenting an unexpected pattern (see here). As far as I know, the method is not implemented in most of the pharmacovigilance softwares and consequently not accessible to most of the companies, universities or regulatory authorities. Publishing this algorithm in open source meets several objectives:

1. Spread the use of this method in Pharmacovigilance to improve quantitative signal detection standards. It is complimentary to disproportionality and screens in a systematic fashion all TTO distributions, something ignored by disproportionality.
2. Make the method evolve. Originally, the method published was different as it was performing two comparisons:

• against the TTO distribution of the event of interest post other vaccines
• against the TTO distribution of other events after the vaccine of interest.

However, further evaluation showed that the comparison to other events was not a good causality predictor. Plus, usage of the method tended to show that significance of the test was mainly driven by the sample size. Frequently reported events tended to have significant p-values for the two-sample Kolmogorov-Smirnov test.Consequently, the implementation of the TTOSD_function performs only the most meaningful comparison to the TTO distribution of the event of interest post other vaccines. And it displays the distance between the two distributions (a value between 0 and 1) that can be used as an additional filter to focus on TTO distribution significantly different AND with a substantial difference in distribution.

3. Allow collaboration with other researchers. It is my hope that the OpenSourcePV package will receive contribution for other researchers who tried alternate tests for flagging product-event pairs.
4. Get tools ready for the massive amount of pharmacovigilance data that will flood starting early 2021 with the availability of a coronavirus vaccine that will be provided worldwidely to a large fraction of the population.

Finally, a word on the choice of releasing it under GPL license. This means that any software that is written based on any GPL component must be released as open source. The result is that any software that uses any GPL open source component is required to release its full source code and all of rights to modify and distribute the entire code. That sounds restrictive but actually that’s only restrictive to software companies that would intend to monopolize open source code without being open on their own. I personnally hope that the future of Pharmacovigilance will be made of more open source initiatives making the PV world less depend on the software industry. It is to be noted that the OpenEBGM is also under GPL license, as is the PhViD package. 

To install the library locally, please run devtools::install_github(“Lion666/OpenSourcePV”) in your R console.