ADEX - dose adjustments

[kabis-ops]

Hi @bundfussr , could you please point me to the spec where the dose adjustment calculation uses FA or EC. I had a look at our spec, and cannot find anything related to this.

[bundfussr]

Hi @kabis-ops , please look at the ADJ parameter in the "ADEX Parameters" tab. (You have to enlarge the cell to see the complete derivation.)

[kabis-ops]

Thanks, so the definition is (this is for the 1:1 mapping)

• Set to 'Y' if the Reason for Dose Adjustment [EX.EXADJ] is mapped and is not null
• Else, set to 'Y' if it has at least 1 related record in EC (matched on the Link Identifiers [EX.EXLNKID] and [EC.ECLNKID]) where a dose adjustment is indicated (Reason for Dose Adjustment [EC.ECADJ] is mapped and is not null and Mood [EC.ECMOOD] = 'PERFORMED')
• Else, set to 'Y' if it has at least 1 related record in FA (matched on the Link Identifiers [EX.EXLNKID] and [FA.FALNKID]) where a dose adjustment is indicated (Findings About Test Short Name [FA.FATESTCD] = 'OCCUR' and Reason for Dose Modification [SUPPFA.FAECADJ] is mapped and is not null and Object of the Observation [FA.FAOBJ] = 'Dose Adjusted' and Result or Finding in Original Units [FA.FAORRES] = 'Y')
• Else, set to 'N'.

I don't think this definition is standard and i guess it is also dependent on the data collection. To derive this, I would expect the user to first add the required variables from EC and FA in EX (to add the relevant variables, we could offer a function that would link the domains and add the required info based on the LINKID? Maybe what is used in ADCM, for the WHO Drug (FA) could be used too here?) and then adjust the derivation in the 1:1 mapping.

adex0<-<ex +EC.ADJ where ECMOD=PERFORMED and FA.FAORRES where .... >
adex1_1 <- bind_rows(
  adex0 %>% mutate(PARAMCD = "DURD", AVAL = EXDUR),
  adex0 %>% mutate(PARAMCD = "DOSE", AVAL = EXDOSE),
  adex0 %>% mutate(PARAMCD = "ADJ", AVALC = if_else(!is.na(EXADJ) | !is.na(ECADJ) | FAECADJ=="Y"), "Y", NA_character_)),
...

Then the flag for an overall/periodic adjustment would be handled as described in the adex script:

# Any Dose adjustment?
tadj <- exposure_var_spec(
  new_parameter = "TADJ",
  input_parameter = "ADJ",
  fns = list(AVALC ~ if_else(sum(!is.na(.)) > 0, "Y", NA_character_))
)

At the end, of the code, we should have the metadata function which should select only the relevant variables and get rif odd these extra variables.

@aehmann-gsk , @thomas-neitmann, please let us know your thoughts :) Thanks

[bundfussr]

In principle this workflow could work. But it is not really modular. The derivation is separated in three steps:

1. derive variables ECADJ and FAECADJ (this requires more than merging a variable because the relationship between EC/FA and EX may be many to one.)
2. derive the ADJ parameter
3. remove the ECADJ and FAECADJ variable

[thomas-neitmann]

For now we require the user to merge any variables from other datasets onto EX themselves.