Skip to contents

Modified Detecting Deviating Cells Algorithm in Pharmacovigilance

Source: R/MDDC-package.R
MDDC-package.Rd

In this package, we present the Modified Detecting Deviating Cells (MDDC) algorithm for adverse event identification. For a certain time period, the spontaneous reports can be extracted from the safety database and depicted as an \(I \times J\) contingency table, where:

  • \(I\) denotes the total number of adverse events (AEs).

  • \(J\) denotes the total number of drugs or vaccines.

The cell counts \(n_{ij}\) represent the total number of reported cases corresponding to the \(j\)-th drug/vaccine and the \(i\)-th AE. We are interested in identifying which (AE, drug or vaccine) pairs are signals. Signals refer to potential adverse events that may be caused by a drug or vaccine. In the contingency table setting, signals refer to the cells where \(n_{ij}\) is abnormally higher than the expected values.

The Detecting Deviating Cells (DDC) algorithm, originally proposed by Rousseeuw and Bossche (2018), was designed for outlier identification in multivariate datasets. However, the original DDC algorithm assumes multivariate normality of the data, with cutoffs based on this assumption. In contrast, the MDDC algorithm is designed for the discrete nature of adverse event data in pharmacovigilance, which clearly do not follow a multivariate normal distribution.

Our Modified Detecting Deviating Cells (MDDC) algorithm has the following characteristics:

  1. It is easy to compute.

  2. It considers AE relationships.

  3. It uses data-driven cutoffs.

  4. It is independent of the use of ontologies.

The MDDC algorithm consists of five steps. The first two steps identify univariate outliers via cutoffs, and the next three steps evaluate the signals using AE correlations. More details can be found in the MDDC algorithm documentation.

For an introduction to the `MDDC` package, see the vignette: Usage Examples for MDDC in R.

Details

This work has been supported by the Food and Drug Administration and the Kaleida Health Foundation.

References

Liu, A., Mukhopadhyay, R., and Markatou, M. (2024). MDDC: An R and Python package for adverse event identification in pharmacovigilance data. arXiv preprint. arXiv:2410.01168.

Liu, A., Markatou, M., Dang, O., and Ball, R. (2024). Pattern discovery in pharmacovigilance through the Modified Detecting Deviating Cells (MDDC) algorithm. Technical Report, Department of Biostatistics, University at Buffalo.

Rousseeuw, P. J., and Van den Bossche, W. (2018). Detecting deviating data cells. Technometrics, 60(2), 135-145.

See also

Useful links:

Author

Anran Liu, Raktim Mukhopadhyay, and Marianthi Markatou

Maintainer: Anran Liu anranliu@buffalo.edu