target_attainment.Rmd

---
title: Meropenem Target Attainment in Adults
---



# Introduction

- "Population Pharmacokinetic Analysis and Dosing Regimen Optimization of Meropenem in Adult Patients"
    - Li et al. J Clin Pharmacol 2006
- Meropenem is broad-spectrum carbapenem antibiotic
    - Efficacy related to time above MIC
- IV dosing every 8 hours by infusion or bolus
    - bolus over 3 to 5 minutes
    - infusion over 15 to 30 minutes
- Authors are interested in seeing if a longer infusion duration 
can increase time above MIC
- In this vignette, we'll try to look at probability of target attainment for 
different infusion times and different MIC values


# Packages
```{r, message = FALSE}
library(mrgsolve)
library(dmutate)
library(tidyverse)
library(magrittr)

a <- config::get()

set.seed(11000011)
```


# Load the `meropenem` model
```{r}
mod <- mread("meropenem", a$mero_project)
mod
```

- The model was obtained from DDMoRe model repository
  - http://repository.ddmore.foundation/model/DDMODEL00000213
  - We took the model code as-is
  
```{r}
blocks(mod, MAIN)
```

- The covariates are: WT, AGE, CLCR

# Read in the meropenem data set
```{r}
data <- 
  read_csv(a$mero_datafile,na='.')  %>% 
  mutate(CMT=1, DUR = round(AMT/RATE,1))
```

- Infusion durations of 0.5 and 3 hours
- 79 total subjects

```{r}
data %>% filter(EVID==1) %>% count(DUR)
```


What do we need?
 
- 1 g Q8h, duration 0.5, 1, 2, 3 hr
- Time > mic at ss (40 <= time <= 48)
- N=100 simulated sets of 79 patients (presumably the original population)


First, get the population set up

- From the publication, it it appears they simulated the studied patients
100 times
- We don't quite have all the information we need here.  For now, I'm going 
to re-sample 7900 patients from the original data set and simulate
some of the remaining covariates

- Re-sample our population

```{r}
ids <- 1:7900

id <- distinct(data, ID, WT, CLCR, AGE) %>% ungroup

set.seed(11020)
id <- sample_n(id, max(ids), replace=TRUE) %>% mutate(ID = 1:n())
```


- We need to simulate some covariates here
```{r}
count(data, CLCR)
```


- Simulate `SCR` and `SEX`

```{r}
cov1 <- covset(SCR[0.4,6.9] ~ rnorm(1,0.8), SEX ~ rbinomial(0.95))

id %<>% mutate_random(cov1)
```

- Then calculate `CLCR`

```{r}
id %<>% mutate(CLCR = 0.85^(SEX==1)*(140-AGE)*WT/(SCR*72))
```


Make a dosing data frame

- For this, the authors only consider the 1g Q8H dose

```{r}
durations <- c(0.5, 1, 2, 3)

data <- expand.ev(amt = 1000, ii = 8, addl = 5, 
                  ID = ids, dur = durations)

data <- mutate(data, 
               ID = rep(ids, times = length(durations)), 
               rate = amt/dur)

data <- left_join(data,id) %>% mutate(ID0 = ID, ID = 1:n())
```


- Create a `tgrid` object covering 40 to 48 hours
- It's usually worth doing this ... getting only the times that 
you really want to analyze

```{r}
des <- tgrid(0,8,0.1) + 5*8

stime(des)
```


# Simulate from our data set

```{r}
out <- 
  mod %>% 
  data_set(data) %>% carry_out(dur) %>%
  mrgsim(tgrid = des, obsonly = TRUE)
```


# Work ou the code to summarize for a single `MIC`

- Summarize for `MIC==4`
- Subtract 40 from time
- group by `ID` and `dur`
- Calculate fraction of the interval above the `MIC`

```{r}
mic <- 4
sum <- 
  out %>% 
  mutate(time = time - 40) %>%
  group_by(ID,dur) %>%
  mutate(TMIC = first(time[Y < mic & time > 0.5 ])) %>%
  mutate(fmic = TMIC/8, fmic = ifelse(is.na(fmic),1,fmic)) %>% 
  ungroup %>% 
  distinct(ID,dur,fmic) %>%
  group_by(dur) %>%
  summarise(PR = mean(fmic >= 0.4))
sum
```

# Write come code to analyze all `MIC` of interest

```{r}
mic <- c(0.064, 0.128, 0.256, 0.512, 
         1, 2, 4, 8, 16, 32, 64)
```


- Take the above code and wrap in a function
- Be sure to save the value for `mic` as well as the duration

```{r}
smry <- function(mic, out) {
  out %>% 
  mutate(time = time - 40) %>%
  group_by(ID,dur) %>%
  mutate(TMIC = first(time[Y < mic & time > 0.5 ])) %>%
  mutate(fmic = TMIC/8, fmic = ifelse(is.na(fmic),1,fmic)) %>% 
  ungroup %>% 
  distinct(ID,dur,fmic) %>%
  group_by(dur) %>%
  summarise(PR = mean(fmic >= 0.4)) %>%
  mutate(mic = mic)
}
```

## Summarize by different `MIC`
```{r}
sum <- parallel::mclapply(mic, mc.cores = 8, smry, out = out)
sum <- bind_rows(sum)
```

# Plot the results
```{r}
ggplot(data = sum, aes(factor(mic), PR, group = dur, shape = factor(dur))) + 
  geom_line() + geom_point()
```