here::i_am("Model_implementation_validation/quarto/valid_amox_carlier_2013.qmd")
# Functions used for simulation-----
# export functions contained in the "functions" folder
list_of_function_path = list.files(
path = here::here("Model_implementation_validation/R_functions/"),
pattern = ".*\\.R$",
full.names = TRUE
)
purrr::walk(.x=list_of_function_path,.f=source)The same model describe both amoxicillin and clavulanic acid concentrations.
readr::read_csv(
file = here::here("Model_implementation_validation/data/carlier_2013/model_description.csv"),
show_col_types = FALSE
) |>
dplyr::rename("Structural model" = structural_model,
"Variability model" = variability_model,
"Covariates effects" = covariate_effects,
"Number of patients" = patients_number,
"Free or total concentrations" = free_or_total_concentrations,
"Unbound fraction" = unbound_fraction) |>
knitr::kable()| DOI | Structural model | Variability model | Covariates effects | Number of patients | Free or total concentrations | Unbound fraction |
|---|---|---|---|---|---|---|
| 10.1093/jac/dkt240 | 2-comp | log-normal clearance | creatinine clearance on clearance | 13 | free | 0.83 |
readr::read_csv(
file = here::here("Model_implementation_validation/data/carlier_2013/model_parameters.csv"),
show_col_types = F
) |>
dplyr::select(-DOI) |>
knitr::kable()| parameter_name | parameter_description | unit | mean |
|---|---|---|---|
| CLpop | Typical clearance (for a patient with creatinin clearance of 102 mL/min) | L/h | 10.0 |
| Q | Intercompartmental clearance | L/h | 15.6 |
| Vcpop | Typical central volume of distribution (for a patient with creatinin clearance of 102 mL/min) | L | 13.7 |
| Vp | Peripheral volume of distribution | L | 13.7 |
| cv_iiv_CL | Coefficient of variation of the inter individual variability on clearance (%) | unitless | 39.9 |
| cv_iiv_Vc | Coefficient of variation of the inter individual variability on central volume of distribution (%) | unitless | 38.7 |
| ruv | Coefficient of variation of the residual variability | unitless | 22.0 |
Covariate effect : \[
TVCL_{i} = CLpop*\frac{CLCR_{i}}{102}*e^{\eta_{i}}
\] With :
\(TVCL_{i}\) : Amoxicillin clearance for individual i
\(CLpop\) : Typical amoxicillin clearance (L/h)
\(CLCR_{i}\) : 24h urinary creatinin clearance (mL/min) for individual i
\(\eta_{i}\) : Normal variable with mean 0 and variance \(\omega^2_{CL}\)
\(102\) : population’s median urinary creatinine clearance in mL/min
To validate that our implementation of the model is correct we will attempt to reproduce figure 2 and table 5 of the original article.
First we check that simulations for a single dosing regimen and creatinine clearance value make sense.
path_dosing_regimens = here::here("Model_implementation_validation/data/carlier_2013/dosing_regimen_table_amox.csv")
path_parameters = here::here("Model_implementation_validation/data/carlier_2013/model_parameters.csv")
model = mrgsolve::mread(model = here::here("Model_implementation_validation/mrgsolve/amox_carlier.cpp"))
dosing_reg_6g_q24h = select_dosing_regimens(selected_regimen = "ci 6g q24h",
path_to_the_file = path_dosing_regimens)
patients = tibble::tibble(ID = 1)
sim_dur = 48 #hours
sim_step = 1/6 #1 every 10 min
omega_mat = mrgsolve::dmat(0,0)
cov_default = list(MIC = 4,SEX = 0, AGE = 60, BSA = 1.73,CREAT = 2.24)
# simulation for a patient with CLCR = 30 mL/min dosing regimen ci 6g q24h
sim_6g_q24h_clcr_30 <- simulate_mrgsolve(
dosing_regimen = dosing_reg_6g_q24h,
sim_dur = sim_dur,
model = model,
patients = patients,
sim_step = sim_step,
param_file = path_parameters,
cov_default = cov_default,
omega_mat = omega_mat)
plot_sim_6g_q24h_clcr_30 = sim_6g_q24h_clcr_30 |>
ggplot2::ggplot(mapping = ggplot2::aes(x=time)) +
ggplot2::geom_line(mapping=ggplot2::aes(y=IPRED)) +
ggplot2::scale_x_continuous(name = "Time (h)",
breaks = seq(0,sim_dur,by=12)) +
ggplot2::scale_y_continuous(name = "Unbound concentration (mg/L)")+
ggplot2::theme_classic()+
ggplot2::ggtitle(label = paste0("CLCR = 30 mL/min"))
plot_sim_6g_q24h_clcr_30
pta_ft_over_mic_6g_q24h_clcr_30_MIC_4 = compute_ft_over_mic(
sim_tbl = sim_6g_q24h_clcr_30,
from = 24,
to = 48) |>
compute_pta_t_over_mic(target = 0.5)
pta_ft_over_mic_6g_q24h_clcr_30_MIC_4 |>
knitr::kable()| MIC | mean_ft_over_mic | sd_ft_over_mic | median_ft_over_mic | PTA_0.5 |
|---|---|---|---|---|
| 4 | 1 | NA | 1 | 1 |
Looks ok
This table indicate wether the target is attained for a typical patient receiving amoxicillin, for various dosing regimen and various creatinin clearances.
readr::read_csv(path_dosing_regimens, show_col_types = FALSE) |>
dplyr::rename("Dose (mg)" = amt,
"Interdose interval (h)" = ii,
"Infusion duration (h)" = tinf,
"Dose regimen" = regimen_id,
"Loading dose ? " = loading_dose,
"Infusion type" = infusion_type) |>
knitr::kable()| Infusion type | Loading dose ? | Dose regimen | time | Infusion duration (h) | Dose (mg) | Interdose interval (h) | ss | evid | cmt |
|---|---|---|---|---|---|---|---|---|---|
| intermittent | no | ii 0.5g q4h | 0.0 | 0.5 | 500 | 4 | 0 | 1 | 1 |
| intermittent | no | ii 0.5g q6h | 0.0 | 0.5 | 500 | 6 | 0 | 1 | 1 |
| intermittent | no | ii 0.5g q8h | 0.0 | 0.5 | 500 | 8 | 0 | 1 | 1 |
| intermittent | no | ii 1g q4h | 0.0 | 0.5 | 1000 | 4 | 0 | 1 | 1 |
| intermittent | no | ii 1g q6h | 0.0 | 0.5 | 1000 | 6 | 0 | 1 | 1 |
| intermittent | no | ii 1g q8h | 0.0 | 0.5 | 1000 | 8 | 0 | 1 | 1 |
| intermittent | no | ii 2g q6h | 0.0 | 0.5 | 2000 | 6 | 0 | 1 | 1 |
| intermittent | no | ii 2g q8h | 0.0 | 0.5 | 2000 | 8 | 0 | 1 | 1 |
| extended | no | ei 0.5g q4h | 0.0 | 2.0 | 500 | 4 | 0 | 1 | 1 |
| extended | no | ei 0.5g q6h | 0.0 | 3.0 | 500 | 6 | 0 | 1 | 1 |
| extended | no | ei 0.5g q8h | 0.0 | 4.0 | 500 | 8 | 0 | 1 | 1 |
| extended | no | ei 1g q4h | 0.0 | 2.0 | 1000 | 4 | 0 | 1 | 1 |
| extended | no | ei 1g q6h | 0.0 | 3.0 | 1000 | 6 | 0 | 1 | 1 |
| extended | no | ei 1g q8h | 0.0 | 4.0 | 1000 | 8 | 0 | 1 | 1 |
| extended | no | ei 2g q6h | 0.0 | 3.0 | 2000 | 6 | 0 | 1 | 1 |
| extended | no | ei 2g q8h | 0.0 | 4.0 | 2000 | 8 | 0 | 1 | 1 |
| extended | no | ei 3g q6h | 0.0 | 3.0 | 3000 | 6 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 6g q24h | 0.0 | 0.5 | 1000 | 0 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 6g q24h | 0.5 | 24.0 | 6000 | 24 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 4g q24h | 0.0 | 0.5 | 1000 | 0 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 4g q24h | 0.5 | 24.0 | 4000 | 24 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 3g q24h | 0.0 | 0.5 | 1000 | 0 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 3g q24h | 0.5 | 24.0 | 3000 | 24 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 8g q24h | 0.0 | 0.5 | 1000 | 0 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 8g q24h | 0.5 | 24.0 | 8000 | 24 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 12g q24h | 0.0 | 0.5 | 1000 | 0 | 0 | 1 | 1 |
| continuous | yes 1g over 0.5h | ci 12g q24h | 0.5 | 24.0 | 12000 | 24 | 0 | 1 | 1 |
convert_clcr_to_creat_mdrd <- function(clcr,
sex = 0,
bsa = 1.73,
age = 60) {
sex_factor <- ifelse(sex==0, 1, 0.742)
creat <- (clcr / (175 * (age^-0.203) * sex_factor * (bsa / 1.73)))^(-1 /
1.154)
creat
}
clcr_values <- c(30,50,130,190)
creat_values <- convert_clcr_to_creat_mdrd(clcr_values) #convert using mdrd assuming male, 60 yo, and base size
pretty_clcr_values <- paste0(clcr_values," mL/min")Simulations were performed for 4 creatinine clearance values (30 mL/min, 50 mL/min, 130 mL/min, 190 mL/min).
pkpd_targets <- c(0.5,1)
mics <- c(4,8,16)
pretty_pkpd_targets <- paste0(pkpd_targets*100,"%")Two different fT>MIC were studied (50%, 100%) They were computed for a range of MICs (4, 8, 16 mg/L).
First, \(fT>MIC\) was calculated at steady state. Amoxicillin concentrations were considered at steady state after nine days.
table_5_pta_results_path <-
here::here(
"Model_implementation_validation/data/carlier_2013/sim_results/table_5_pta_results.csv"
)
#run only if results does not exist yet
if (!file.exists(table_5_pta_results_path)) {
#Function to #compute PTA for 1 dosing regimen, 1 CLCR, 1 MIC, enables looping through them
wrapper_compute_pta <- function(dosing_reg_id,
path_dosing_regimens,
sim_dur,
sim_step,
model,
patients,
path_parameters,
cov_default,
omega_mat,
from,
to,
targets) {
dosing_reg = select_dosing_regimens(selected_regimen = dosing_reg_id,
path_to_the_file = path_dosing_regimens)
sim <- simulate_mrgsolve(
dosing_regimen = dosing_reg,
sim_dur = sim_dur,
model = model,
patients = patients,
sim_step = sim_step,
param_file = path_parameters,
cov_default = cov_default,
omega_mat = omega_mat
) |>
dplyr::rename(CLCR = CRCL)
pta =
purrr::map(
targets,
~ compute_ft_over_mic(
sim_tbl = sim,
from = from,
to = to
) |>
compute_pta_t_over_mic(target = .x)
) |>
purrr::reduce(dplyr::left_join) |>
dplyr::mutate(regimen_id = dosing_reg_id,
CLCR = round(unique(sim$CLCR),0))
pta
}
#### Dosing regimens to be simulated ####
dosing_regimens_ids = readr::read_csv(path_dosing_regimens) |>
dplyr::pull(regimen_id) |>
unique()
#### Patient level covariates ####
# None used here
patients = tibble::tibble(ID = 1)
omega_mat = mrgsolve::dmat(0,0) # no variability
#### Simulation time parameters ####
sim_dur = 264 #hours
sim_step = 1 / 6 #1 every 10 min
#### PK/PD targets and PTA parameters ####
from = 240
to = 264
targets = c(0.5, 1)
#### Setting up the loop for simulations ####
regimen_cov_grid = expand.grid(
dosing_reg_id = dosing_regimens_ids,
CREAT = creat_values, #defined in an earlier chunk
MIC = mics,#defined in an earlier chunk
BSA = 1.73,
SEX = 0,
AGE = 60
)
dosing_reg_ids = regimen_cov_grid$dosing_reg_id
cov_list = dplyr::select(regimen_cov_grid, -dosing_reg_id) |>
split(x = _, f = 1:nrow(regimen_cov_grid))
#### Perform the simulations and compute PTA ####
pta_results <- purrr::map2(
.x = dosing_reg_ids,
.y = cov_list,
.f = ~ wrapper_compute_pta(
dosing_reg_id = .x,
path_dosing_regimens = path_dosing_regimens,
sim_dur = sim_dur,
sim_step = sim_step,
model = model,
patients = patients,
path_parameters = path_parameters,
cov_default = .y,
omega_mat = omega_mat,
from = from,
to = to,
targets = targets
)
) |>
purrr::list_rbind()
readr::write_csv(
x = pta_results,
file = table_5_pta_results_path
)
}
table_5_sim <- readr::read_csv(table_5_pta_results_path)table_5_data_paper <- readr::read_csv(here::here("Model_implementation_validation/data/carlier_2013/data_table_5_paper.csv"))
comparison_table <- dplyr::left_join(
table_5_data_paper,
dplyr::relocate(table_5_sim,CLCR,regimen_id,MIC,PTA_0.5,PTA_1),
by = c("MIC", "CLCR", "regimen_id"),
suffix = c("_paper", "_sim")
) |>
dplyr::select(-sd_ft_over_mic, -mean_ft_over_mic) |>
dplyr::rename(sim_ftovermic = median_ft_over_mic) |>
dplyr::mutate(dplyr::across(tidyselect::starts_with("PTA"),~round(.x,digits=4))) |>
dplyr::mutate(dplyr::across(tidyselect::starts_with("sim_"),~round(.x,digits=4))) |>
dplyr::mutate(
PTA_0.5_agreement = PTA_0.5_paper == PTA_0.5_sim,
PTA_1_agreement = PTA_1_paper == PTA_1_sim
) |>
dplyr::rename_with(
~ stringr::str_replace(., "PTA_", "PTA") |> vctrs::vec_as_names(),
.cols = tidyselect::starts_with("PTA")
) |>
dplyr::select(sort(tidyselect::peek_vars())) |>
tidyr::pivot_wider(
names_from = MIC,
names_glue = "{MIC}_{.value}",
names_vary = "slowest",
names_sort = F,
values_from = c(tidyselect::starts_with("PTA"),"sim_ftovermic"),
id_cols = c(CLCR, regimen_id)
)
comparison_table |>
gt::gt(rowname_col = "regimen_id",groupname_col = 'CLCR') |>
gt::tab_spanner_delim(delim = "_",
columns = tidyselect::matches("^[0-9].*")) |>
gt::text_transform(
fn = function(x)
gsub("PTA", " fT>MIC > ", x),
locations = gt::cells_column_spanners()
) |>
gt::text_transform(fn = function(x)
paste0("MIC = ",x," mg/L"),
locations = gt::cells_column_spanners(spanners =
c("spanner-4_PTA0.5_agreement",
"spanner-8_PTA0.5_agreement",
"spanner-16_PTA0.5_agreement")
)
) |>
gt::fmt_tf(columns = tidyselect::matches("paper$||sim$"),
true_val = "+",
false_val = "-") |>
gt::fmt_tf(columns = tidyselect::matches("agreement"),
tf_style = "check-mark") |>
gt::fmt_percent(columns = tidyselect::matches("sim_ftovermic"),
decimals = 1) |>
gt::text_transform(fn = function(x) paste0("Creatinine clearance ",x," mL/min"),
locations = gt::cells_row_groups()
)
### Code to find the spanner names
### better function is a work in progress at : https://github.com/rstudio/gt/issues/1628
# test = dplyr::left_join(
# table_5_data_paper,
# table_5_sim,
# by = c("MIC", "CLCR", "regimen_id"),
# suffix = c("_paper", "_sim")
# ) |>
# dplyr::select(-sd_ft_over_mic, -mean_ft_over_mic) |>
# dplyr::rename(sim_ft_over_mic = median_ft_over_mic) |>
# dplyr::mutate(
# PTA_0.5_agreement = PTA_0.5_paper == PTA_0.5_sim,
# PTA_1_agreement = PTA_1_paper == PTA_1_sim
# ) |>
# dplyr::rename_with(
# ~ stringr::str_replace(., "PTA_", "PTA") |> vctrs::vec_as_names(),
# .cols = tidyselect::starts_with("PTA")
# ) |>
# dplyr::select(sort(tidyselect::peek_vars()))|>
# tidyr::pivot_wider(
# names_from = MIC,
# names_glue = "{MIC}_{.value}",
# names_vary = "slowest",
# names_sort = F,
# values_from = tidyselect::starts_with("PTA"),
# id_cols = c(CLCR, regimen_id)
# ) |>
# gt::gt(groupname_col = 'CLCR') |>
# gt::tab_spanner_delim(delim = "_",
# columns = tidyselect::matches("^[0-9].*")) |>
# gt::text_transform(
# fn = function(x) gsub("PTA", " fT>MIC > ", x),
# locations = gt::cells_column_spanners()
# ) |> gt::tab_info()
# test$`_data` |>
# dplyr::filter(location == "Spanners" & i == 2)
MIC = 4 mg/L
|
MIC = 8 mg/L
|
MIC = 16 mg/L
|
|||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
fT>MIC > 0.5
|
fT>MIC > 1
|
sim
|
fT>MIC > 0.5
|
fT>MIC > 1
|
sim
|
fT>MIC > 0.5
|
fT>MIC > 1
|
sim
|
|||||||||||||
| agreement | paper | sim | agreement | paper | sim | ftovermic | agreement | paper | sim | agreement | paper | sim | ftovermic | agreement | paper | sim | agreement | paper | sim | ftovermic | |
| Creatinine clearance 30 mL/min | |||||||||||||||||||||
| ii 0.5g q8h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | - | - | 77.8% |
| ii 1g q8h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% |
| ii 1g q6h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% |
| Creatinine clearance 50 mL/min | |||||||||||||||||||||
| ii 0.5g q6h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✘ | + | - | ✔ | - | - | 46.0% |
| ii 1g q8h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | - | - | 76.0% |
| ei 1g q8h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✘ | + | - | 99.6% |
| ii 1g q6h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% |
| ci 4g q24h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% |
| Creatinine clearance 130 mL/min | |||||||||||||||||||||
| ii 1g q8h | ✔ | + | + | ✔ | - | - | 62.7% | ✔ | - | - | ✔ | - | - | 39.2% | ✔ | - | - | ✔ | - | - | 18.1% |
| ii 1g q6h | ✔ | + | + | ✘ | + | - | 86.5% | ✔ | + | + | ✔ | - | - | 55.2% | ✔ | - | - | ✔ | - | - | 26.0% |
| ei 1g q6h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | - | - | 76.4% | ✘ | + | - | ✔ | - | - | 36.1% |
| ci 4g q24h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | - | - | ✔ | - | - | 0.0% |
| ii 1g q4h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✘ | + | - | 93.0% | ✘ | + | - | ✔ | - | - | 47.0% |
| ci 6g q24h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% |
| Creatinine clearance 190 mL/min | |||||||||||||||||||||
| ii 1g q6h | ✔ | + | + | ✔ | - | - | 56.3% | ✔ | - | - | ✔ | - | - | 33.2% | ✔ | - | - | ✔ | - | - | 16.9% |
| ei 1g q6h | ✔ | + | + | ✘ | + | - | 79.1% | ✔ | + | + | ✔ | - | - | 54.1% | ✔ | - | - | ✔ | - | - | 0.0% |
| ci 4g q24h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✘ | - | + | 100.0% | ✔ | - | - | ✔ | - | - | 0.0% |
| ii 1g q4h | ✔ | + | + | ✘ | + | - | 89.7% | ✔ | + | + | ✔ | - | - | 54.5% | ✔ | - | - | ✔ | - | - | 27.6% |
| ci 6g q24h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✘ | + | - | ✔ | - | - | 0.0% |
| ei 2g q6h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | - | - | 79.1% | ✔ | + | + | ✔ | - | - | 54.1% |
| ci 8g q24h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✔ | + | + | 100.0% |
| ei 3g q6h | ✔ | + | + | ✔ | + | + | 100.0% | ✔ | + | + | ✘ | + | - | 93.8% | ✔ | + | + | ✔ | - | - | 68.6% |
Table 4.3 below shows creatinine clearances and dosing regimens for which there is at least one disagreement with the paper.
dplyr::left_join(
table_5_data_paper,
table_5_sim,
by = c("MIC", "CLCR", "regimen_id"),
suffix = c("_paper", "_sim")
) |>
dplyr::select(-sd_ft_over_mic, -mean_ft_over_mic) |>
dplyr::rename(sim_ftovermic = median_ft_over_mic) |>
dplyr::mutate(dplyr::across(tidyselect::starts_with("PTA"),~round(.x,digits=4))) |>
dplyr::mutate(dplyr::across(tidyselect::starts_with("sim_"),~round(.x,digits=4))) |>
dplyr::mutate(
PTA_0.5_agreement = PTA_0.5_paper == PTA_0.5_sim,
PTA_1_agreement = PTA_1_paper == PTA_1_sim
) |>
dplyr::filter(dplyr::if_any(tidyselect::ends_with('agreement'), ~ . == FALSE)) |>
dplyr::rename_with(
~ stringr::str_replace(., "PTA_", "PTA") |> vctrs::vec_as_names(),
.cols = tidyselect::starts_with("PTA")
) |>
dplyr::select(sort(tidyselect::peek_vars())) |>
tidyr::pivot_wider(
names_from = MIC,
names_glue = "{MIC}_{.value}",
names_vary = "slowest",
names_sort = F,
values_from = c(tidyselect::starts_with("PTA"),"sim_ftovermic"),
id_cols = c(CLCR, regimen_id)
) |>
gt::gt(rowname_col = "regimen_id",groupname_col = 'CLCR') |>
gt::tab_spanner_delim(delim = "_",
columns = tidyselect::matches("^[0-9].*")) |>
gt::text_transform(
fn = function(x)
gsub("PTA", " fT>MIC > ", x),
locations = gt::cells_column_spanners()
) |>
gt::text_transform(fn = function(x)
paste0("MIC = ",x," mg/L"),
locations = gt::cells_column_spanners(spanners =
c("spanner-4_PTA0.5_agreement",
"spanner-8_PTA0.5_agreement",
"spanner-16_PTA0.5_agreement")
)
) |>
gt::fmt_tf(columns = tidyselect::matches("paper$||sim$"),
true_val = "+",
false_val = "-") |>
gt::fmt_tf(columns = tidyselect::matches("agreement"),
tf_style = "check-mark") |>
gt::fmt_percent(columns = tidyselect::matches("sim_ftovermic"),
decimals = 1) |>
gt::text_transform(fn = function(x) paste0("Creatinine clearance ",x," mL/min"),
locations = gt::cells_row_groups()
) |>
gt::sub_missing(missing_text = "")
MIC = 4 mg/L
|
MIC = 8 mg/L
|
MIC = 16 mg/L
|
|||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
fT>MIC > 0.5
|
fT>MIC > 1
|
sim
|
fT>MIC > 0.5
|
fT>MIC > 1
|
sim
|
fT>MIC > 0.5
|
fT>MIC > 1
|
sim
|
|||||||||||||
| agreement | paper | sim | agreement | paper | sim | ftovermic | agreement | paper | sim | agreement | paper | sim | ftovermic | agreement | paper | sim | agreement | paper | sim | ftovermic | |
| Creatinine clearance 130 mL/min | |||||||||||||||||||||
| ii 1g q6h | ✔ | + | + | ✘ | + | - | 86.5% | ||||||||||||||
| ii 1g q4h | ✔ | + | + | ✘ | + | - | 93.0% | ✘ | + | - | ✔ | - | - | 47.0% | |||||||
| ei 1g q6h | ✘ | + | - | ✔ | - | - | 36.1% | ||||||||||||||
| Creatinine clearance 190 mL/min | |||||||||||||||||||||
| ei 1g q6h | ✔ | + | + | ✘ | + | - | 79.1% | ||||||||||||||
| ii 1g q4h | ✔ | + | + | ✘ | + | - | 89.7% | ||||||||||||||
| ci 4g q24h | ✔ | + | + | ✘ | - | + | 100.0% | ||||||||||||||
| ei 3g q6h | ✔ | + | + | ✘ | + | - | 93.8% | ||||||||||||||
| ci 6g q24h | ✘ | + | - | ✔ | - | - | 0.0% | ||||||||||||||
| Creatinine clearance 50 mL/min | |||||||||||||||||||||
| ii 0.5g q6h | ✘ | + | - | ✔ | - | - | 46.0% | ||||||||||||||
| ei 1g q8h | ✔ | + | + | ✘ | + | - | 99.6% | ||||||||||||||
Aside from one dosing regimen (ci 4g q24h for a CLCR of 190 mL/minutes), disagreements between table 5 and our simulations are underestimations of the fT>MIC.
A lot of these disagreements imply differences between the fT>MIC that we simulated and table 5 close to 10%.
We seemed to underestimate fT>MIC for higher creatinin clearances more than lower ones.
No kind of dosing regimen (continuous, extended or intermittent infusion) looked more susceptible to disagreements between our simulations and table 5.
For continuous infusions, we obtained fT>MIC equal to 0% or close or equal to 100%. However, in table 5, for a CLCR of 190 mg/L and a MIC of 16 mg/L, we notice that for he dosing regimen “Ci 6g q24h”, the target of fT > MIC >= 50% is atteined but not a fT > MIC of 100%. This may indicate that the article’s author didn’t perform calculate fT > MIC at steady state only. Instead, they may have included the loading dose in the time interval used to calculate fT > MIC.
We noticed that the authors simulated 7 days of treatment to create figure 2. We hypothesized that they also simulated 7 days of treatment and calculated fT > MIC on all the simulation duration in order to create table 5. We then calculated fT > MIC on seven days of treatment (results not shown here). This yielded results similar to the one obtained on 24 hours at steady state : we observed disagreements for the same dosing regimens, MIC and CLCR and fT > MIC equals or slightly inferiors. In particular, for a CLCR of 190 mg/L and a MIC of 16 mg/L, we still obtained a fT > MIC equal to 0%. We didn’t confirm our hypothesis.
We also calculated fT > MIC on the first 24 hours of treatment. However, the fT>MIC were underestimated compared to the results of the article and there were a lot more disagreements between our simulations and table 5. For a CLCR of 190 mg/L and a MIC of 16 mg/L, fT > MIC was only equal to 9%.
While it would seem that Figure 2 could also be used for validation of our model implementation, careful consideration of the plot makes us suspect that technical errors happened while preparing the figure. Therefore, no attempt at reproducing the figure will be made.
We think our mrgsolve model is an adequate representation of the published model since :
The slight discrepencies could be due to different simulation settings (e.g. different time points) and/or missing correlations between variability parameters which were not reported.