Reset AUC compartment at each dose event

[mark0935git]

Dear Mrgsolve community,

I have a question on how to reset the AUC compartment values at each dose event.

I would like to calculate the cumulative drug exposure (AUC) during a dose interval and use this AUC value for subsequent PD model. I wish the AUC compartment to be reset at each dose event.

I have tried to use evid=3 but that flag resets the whole system rather than AUC compartment. Could you please instruct me which method I can use to reset only the AUC compartment?

Many thanks in advance!

There is the example codes for the PK and AUC sections.

...
[INIT] @annotated
CENT:   0   : central cmt (ug)
PERI:   0   : peripheral cmt (ug)
AUC:   0   :  cumulative AUC during a dose interval

[ODE]
dxdt_CENT = -(CL/V1)*CENT - (Q/V1)*CENT + (Q/V2)*PERI;
dxdt_PERI = (Q/V1)*CENT - (Q/V2)*PERI;
dxdt_AUC = CENT/V1;
....

[kylebaron]

Hi @mark0935git -

I think the NONMEM way would be to turn the compartment off and then on again. We can do it a little simpler in mrgsolve. See the suggestions below ... two approaches, one using the data and one using just the model.

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(mrgsolve)
#> 
#> Attaching package: 'mrgsolve'
#> The following object is masked from 'package:stats':
#> 
#>     filter

Reset the AUC compartment

In NONMEM, we’d turn the compartment off then on. With
mrgsolve, we can use evid 8 which replaces the amount in the
compartment with the amount of the “dose”.

code <- '
[PARAM] CL = 1, V1 = 20, Q = 5, V2 = 100

[INIT] @annotated
CENT:   0   : central cmt (ug)
PERI:   0   : peripheral cmt (ug)
AUC:    0   :  cumulative AUC during a dose interval

[ODE]
dxdt_CENT = -(CL/V1)*CENT - (Q/V1)*CENT + (Q/V2)*PERI;
dxdt_PERI = (Q/V1)*CENT - (Q/V2)*PERI;
dxdt_AUC = CENT/V1;

'

mod <- mcode("code", code, end = 120)
#> Building code ... done.

Just the dose (no reset)

dose <- ev(amt = 100, ii = 24, addl = 3) 

mrgsim(mod, dose) %>% plot()

Create reset events

reset <- ev(amt = 0, ii = 24, addl = 3, evid = 8, cmt = 3)

We combined these together

reset_dose <- c(reset,dose) %>% realize_addl()
reset_dose
#> Events:
#>   time amt ii addl cmt evid
#> 1    0   0  0    0   3    8
#> 2    0 100  0    0   1    1
#> 3   24   0  0    0   3    8
#> 4   24 100  0    0   1    1
#> 5   48   0  0    0   3    8
#> 6   48 100  0    0   1    1
#> 7   72   0  0    0   3    8
#> 8   72 100  0    0   1    1

We don’t need the reset at time 0; won’t hurt to leave it in

reset_dose <- filter(reset_dose, !(time==0 & evid==8))
reset_dose
#> Events:
#>   time amt ii addl cmt evid
#> 1    0 100  0    0   1    1
#> 2   24   0  0    0   3    8
#> 3   24 100  0    0   1    1
#> 4   48   0  0    0   3    8
#> 5   48 100  0    0   1    1
#> 6   72   0  0    0   3    8
#> 7   72 100  0    0   1    1

Simulate and plot

out1 <- mrgsim(mod, reset_dose) 
plot(out1)

Another model-based approach

Here, we don’t need to reset AUC; we just keep track of the value at
the last dose and calculate the difference.

code2 <- '
[PARAM] CL = 1, V1 = 20, Q = 5, V2 = 100

[INIT] @annotated
CENT:   0   : central cmt (ug)
PERI:   0   : peripheral cmt (ug)
AUC:    0   :  cumulative AUC during a dose interval

[ MAIN ] 
if(NEWIND <= 1) double last_auc = 0;

[ODE]
dxdt_CENT = -(CL/V1)*CENT - (Q/V1)*CENT + (Q/V2)*PERI;
dxdt_PERI = (Q/V1)*CENT - (Q/V2)*PERI;
dxdt_AUC = CENT/V1;

[ TABLE ] 
if(EVID==1) last_auc = AUC;
capture dAUC = AUC - last_auc;
'

mod2 <- mcode("code2", code2, end = 120)
#> Building code2 ... done.

out2 <- mrgsim(mod2, dose)

plot(out2, CENT+dAUC ~ time)

plot(out1, CENT+AUC ~ time)

library(ggplot2)
sim1 <- as_tibble(out1)
sim2 <- as_tibble(out2)

The two methods give the same result

ggplot() + 
  geom_line(data = sim1, aes(time,AUC)) + 
  geom_point(data = sim2, aes(time,dAUC), color = "firebrick") + 
  theme_bw()

^{Created on 2022-06-18 by the reprex package (v2.0.1)}

[mark0935git]

@kylebaron Really appreciate your thorough explanation! I will certainly use your example for the model!