Using pvOmega with DiAna

Introduction

Omega is a measure of disproportionality designed to identify potential drug-drug interactions in collections of individual case safety reports. It compares the observed reporting of an event for a pair of drugs with the reporting expected under an additive model.

The omega_analysis() function provides an interface for performing an Omega analysis directly from drug and reaction data. This vignette illustrates the use of the functions coming from pvOmega with the sample data distributed with DiAna and with locally stored FAERS data.

Example using DiAna sample data

The sample datasets included with DiAna can be used to explore omega_analysis() without importing an external pharmacovigilance database.

library(DiAna)
library(pvOmega)

omega_analysis(
  drug1_selected = "paracetamol",
  drug2_selected = "ibuprofen",
  reac_selected = "overdose",
  temp_drug = DiAna::sample_Drug,
  temp_reac = DiAna::sample_Reac
)

Here, paracetamol and ibuprofen define the drug pair and overdose is the event of interest. The temp_drug and temp_reac arguments specify the drug and reaction tables used to construct the counts required for the Omega calculation.

Using FAERS data

The following examples demonstrate how the same analysis can be performed using locally prepared FAERS data.

We import the data as usual, once the setup_DiAna function has been used to download them locally.

library(data.table)

import("DRUG")
import("REAC")

The examples below are not evaluated when the vignette is built because the FAERS data are not distributed with the package.

Gemfibrozil, cerivastatin, and rhabdomyolysis

A classic drug-drug-event combination can be investigated by selecting gemfibrozil and cerivastatin as the drug pair and rhabdomyolysis as the event:

omega_analysis(
  drug1_selected = "gemfibrozil",
  drug2_selected = "cerivastatin",
  reac_selected = "rhabdomyolysis"
)

Identifying reports containing the triplet

The reports contributing to a particular drug-drug-event combination can also be identified directly from the underlying tables.

For example, restricting drug exposure to primary suspect (PS), secondary suspect (SS), and interacting (I) drug roles:

omega_analysis(
  drug1_selected = "gemfibrozil",
  drug2_selected = "cerivastatin",
  reac_selected = "rhabdomyolysis",
  temp_drug = Drug[role_cod %in% c("PS", "SS", "I")]
)

Other drug-drug-event combinations

The same approach can be applied to other combinations. For example, digoxin, clarithromycin, and increased drug level:

omega_analysis(
  drug1_selected = "digoxin",
  drug2_selected = "clarithromycin",
  reac_selected = "drug level increased"
)

or ipilimumab, nivolumab, and hepatitis:

omega_analysis(
  drug1_selected = "ipilimumab",
  drug2_selected = "nivolumab",
  reac_selected = "hepatitis"
)

Selecting multiple drugs

More than one drug can be included in a selection by supplying a character vector.

For example:

omega_analysis(
  drug1_selected = c(
    "ipilimumab",
    "paracetamol"
  ),
  drug2_selected = "nivolumab",
  reac_selected = "hepatitis"
)

This allows an analysis to be run separately on multiple substances rather than a single drug name.

Defining a drug group

Related substances can also be combined into a named group by supplying a named list.

For example, several PD-1/PD-L1 inhibitors can be grouped as PD1PDL1 and considered as a single substance:

omega_analysis(
  drug1_selected = c(
    "ipilimumab",
    "paracetamol"
  ),
  drug2_selected = list(
    "PD1PDL1" = c(
      "nivolumab",
      "pembrolizumab",
      "atezolizumab",
      "avelumab",
      "durvalumab",
      "cemiplimab",
      "tislelizumab",
      "dostarlimab",
      "retifanlimab",
      "toripalimab",
      "cosibelimab"
    )
  ),
  reac_selected = "hepatitis"
)

Using a named group can be useful when the scientific question concerns a drug class or another predefined collection of substances rather than an individual product.

The substances included in a group should be selected on the basis of the analysis question rather than the observed Omega results.

Defining an event group

The same principle can be applied to reactions. A named list can be used to combine related reaction terms into a single event definition.

For example:

omega_analysis(
  drug1_selected = c(
    "ipilimumab",
    "paracetamol"
  ),
  drug2_selected = list(
    "PD1PDL1" = c(
      "nivolumab",
      "pembrolizumab",
      "atezolizumab",
      "avelumab",
      "durvalumab",
      "cemiplimab",
      "tislelizumab",
      "dostarlimab",
      "retifanlimab",
      "toripalimab",
      "cosibelimab"
    )
  ),
  reac_selected = list(
    "hepatitis" = c(
      "hepatitis",
      "liver injury"
    )
  )
)

This defines the event of interest using both hepatitis and liver injury, while retaining hepatitis as the label for the grouped event.

Interpretation

Omega is intended as a tool for signal detection and hypothesis generation. An elevated Omega value should not, by itself, be interpreted as evidence of a causal or clinically relevant drug-drug interaction.

Results depend on the underlying reporting data and on analytical choices such as the definition of drug exposure, drug roles, event terms, report population, and grouping of drugs or reactions.

In particular, grouping multiple drugs or reaction terms changes the question being addressed by the analysis. Such groupings should therefore be defined on clinical, pharmacological, or other substantive grounds before interpreting the resulting Omega estimates.