testing AUC_inf against phoenix

Data was used under design of PP3 for Phar 747 where dose was IV bolus @ time 0

• Cl = 0.15 L/hr
• V = 7

sampling to 96 hours as:

[1]  0.00  0.10  0.25  0.50  0.75  1.00  2.00  4.00  6.00  8.00 10.00 12.00 24.00 48.00
[15] 72.00 96.00

library(ggplot2)
library(dplyr)

## Warning: package 'dplyr' was built under R version 3.0.3

## 
## 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(PKPDmisc)
library(knitr)

## Warning: package 'knitr' was built under R version 3.0.3

opts_chunk$set(message = F, warning = F, echo = FALSE, cache = TRUE)

Percent difference between R and Phoenix for AUC0_inf, AUC0-tlast, and Percent Extrapolated AUC

To evaluate how AUC_inf may be impacted by residual variability, time points from 0:500 by 10 were simulated using the same design as above.

AUC0-inf was compared using DV and IPRED values to AUC0-tlast where tlast was a simulated concentration greater than 0.26 concentration units (therefore should approximate the ‘true’ AUC0-inf)

Comparison for the impact of residual variability based on sampling was done by comparing if data was only sampled to 100 hours vs the rich sampling to BQL.

An example of the profiles for a single rep are shown below:

## Source: local data frame [6 x 7]
## Groups: X..repl
## 
##   X..repl id AUC0_inf extra_percent AUCall AUC0_200 AUC0_100
## 1       0  1     1520           1.0   1505     1481     1275
## 2       0  2     1329           0.9   1317     1288     1081
## 3       0  3     2313           0.4   2304     2304     2225
## 4       0  4     1817           0.9   1801     1675     1304
## 5       0  5     1187           1.1   1175     1175     1056
## 6       0  6     1684           1.1   1665     1597     1300

## Source: local data frame [6 x 6]
## Groups: X..repl
## 
##   X..repl id AUC0_inf extra_percent AUCall AUC0_100
## 1       0  1     1677          31.6   1275     1275
## 2       0  2     1326          22.6   1081     1081
## 3       0  3     2315           4.0   2225     2225
## 4       0  4     1564          20.0   1304     1304
## 5       0  5     1189          12.6   1056     1056
## 6       0  6     1688          29.9   1300     1300

## Source: local data frame [6 x 7]
## Groups: X..repl
## 
##   X..repl id AUC0_inf extra_percent AUCall AUC0_200 AUC0_100
## 1       0  1     1556           1.0   1541     1517     1311
## 2       0  2     1374           1.3   1356     1325     1115
## 3       0  3     2307           0.4   2299     2299     2221
## 4       0  4     1886           1.2   1863     1734     1351
## 5       0  5     1196           1.1   1184     1184     1072
## 6       0  6     1677           1.0   1660     1593     1303

## Source: local data frame [6 x 6]
## Groups: X..repl
## 
##   X..repl id AUC0_inf extra_percent AUCall AUC0_100
## 1       0  1     1677          31.6   1275     1275
## 2       0  2     1326          22.6   1081     1081
## 3       0  3     2315           4.0   2225     2225
## 4       0  4     1564          20.0   1304     1304
## 5       0  5     1189          12.6   1056     1056
## 6       0  6     1688          29.9   1300     1300

Compare AUC all (rich sampling to very low concentrations) to AUC0_inf using extrapolation

As you can see, using DV’s to calculate the kel for extrapolation under simulations results in a very wide range of % extrapolation due to noise when using even marginally rich sampling (every 10 units of time in this case) when sampling stopped at 100 hours.

More investigation is needed, however it seems that for simulation purposes, calculation of % extrapolated from the IPRED values still results in the overall AUC calculation similar to the “true” value