Skip to contents

Omega drug-drug interaction analysis on DiAna CDM

Source: R/omega.R
omega_analysis.Rd

Computes the Omega interaction measure (Noren et al. 2008) for every combination of a first drug, a second drug and an adverse event, starting from the Drug and Reac tables of the DiAna CDM. The interface mirrors disproportionality_analysis() and calls the pvOmega package for the calculation.

Usage

omega_analysis(
  drug1_selected,
  drug2_selected,
  reac_selected,
  temp_drug = Drug[role_cod %in% c("PS", "SS", "I")],
  temp_reac = Reac,
  restriction = "none",
  minimum_cases = 3,
  log2_threshold = 0,
  store_pids = FALSE,
  save_in_excel = FALSE,
  file_name = "omega_results"
)

Arguments

drug1_selected

First drug(s). A character vector, or a (named) list whose elements are character vectors of terms to collapse into one group.

drug2_selected

Second drug(s), same format as drug1_selected.

reac_selected

Adverse event(s), same format as drug1_selected.

temp_drug

Drug dataset (default Drug[role_cod%in%c("PS","SS","I")], as created by DiAna::import("DRUG")). Can be set to DiAna::sample_Drug for testing.

temp_reac

Reac dataset (default Reac). Can be set to DiAna::sample_Reac for testing.

restriction

Primary IDs to consider (default "none", the entire population). For example Demo[!RB_duplicates_only_susp]$primaryid excludes duplicates according to one of DiAna's deduplication algorithms.

minimum_cases

Minimum number of cases (n111) to flag a signal (default 3). With alpha = 0.5, credibility interval 0.95 and the Gamma interval, fewer than 3 cases cannot produce omega_lower > 0 anyway.

log2_threshold

Threshold on omega_lower for flagging a signal (default 0).

store_pids

Logical. If TRUE, adds a list column pids_cases with the primary IDs of the reports listing both drugs and the event, useful for case-series review. Default FALSE.

save_in_excel

Logical. If TRUE, writes the results to an Excel file (requires the 'writexl' package). Default FALSE.

file_name

Name of the Excel file (without extension). Only used if save_in_excel = TRUE.

Value

A data.table with one row per drug1-drug2-event combination: the labels drug1, drug2, event, all columns returned by omega_from_counts(), label_omega and omega_signal (an ordered factor: "not enough cases", "no SDR", "SDR").

Details

Report universe

The total number of reports (n...) is the number of distinct primaryid in temp_reac, intersected with restriction when supplied. All drug and event report sets are intersected with this universe, so every count refers to the same population. When both orderings of a drug pair occur (e.g. A-B and B-A), only the first is kept. Omega is symmetric in the two drugs.

Drug roles

By default only drugs recorded as primary suspect, secondary suspect or interacting (role_cod in PS, SS, I) count as exposure, following the approach used at Uppsala Monitoring Centre (Hult et al. 2020). The universe is not filtered by role: a report where a drug is only concomitant contributes to the background stratum. Set temp_drug = Drug to count all roles.

Interpretation

Omega compares the observed number of reports with the number expected if the two drugs contributed additive, independent risks. omega_lower > 0 (Omega025 > 0 at the default 95% level) indicates that the combination is reported more often than expected under no interaction. As for any disproportionality measure, this is a hypothesis for clinical review, not evidence of a causal interaction. Always inspect f00, f10, f01, f11 and g11 and the case series, alongside Omega, as recommended in the original paper.

References

Noren GN, Sundberg R, Bate A, Edwards IR. A statistical methodology for drug-drug interaction surveillance. Stat Med. 2008;27(16):3057-70. doi:10.1002/sim.3247

Hult S, Sartori D, Bergvall T, et al. A feasibility study of drug-drug interaction signal detection in regular pharmacovigilance. Drug Saf. 2020;43:775-85. doi:10.1007/s40264-020-00939-y

Examples

omega_analysis(
  drug1_selected = "paracetamol",
  drug2_selected = "ibuprofen",
  reac_selected = "overdose",
  temp_drug = sample_Drug,
  temp_reac = sample_Reac
)
#>          drug1     drug2    event  n1.1  n.11  n1..  n.1.  n..1  n...  n111
#>         <char>    <char>   <char> <num> <num> <num> <num> <num> <num> <num>
#> 1: paracetamol ibuprofen overdose     5     0    51    15    15  1000     0
#>     n110  n101  n100  n011  n010  n001  n000  n11.  n10.  n01.  n00.       f00
#>    <num> <num> <num> <num> <num> <num> <num> <num> <num> <num> <num>     <num>
#> 1:     6     5    40     0     9    10   930     6    45     9   940 0.0106383
#>          f10   f01   f11       g11      E111     omega omega_lower omega_upper
#>        <num> <num> <num>     <num>     <num>     <num>       <num>       <num>
#> 1: 0.1111111     0     0 0.1111111 0.6666667 -1.222392   -11.21428    1.106411
#>                 label_omega     omega_signal
#>                      <char>            <ord>
#> 1: -1.22 (-11.21; 1.11) [0] not enough cases