Results
Developing a population PK model using ChatGPT
Based on the literature review provided by ChatGPT, typical clearance
(CL) of FVIII in haemophilia A adults with a body weight of 70kg range
between 2.5-3.5 dL/h depending on various factors such as age and health
status. ChatGPT also suggested that the volume of distribution for a
70kg patient can range between 40-60 dL. We found that these PK
parameters are consistent with previous studies. For example, one study
by Björkman et al. found a CL of 2.3 dL/h and a V of 37.1 dL based on
adults of 70kg16. Another study by McEneny-king and
colleagues estimated a CL of 2.4-3.24 dL/h and a V of 35.4-46.2 dL ,
also based on adults of 70kg17. Based on the estimates
provided by ChatGPT and the previous literature, we chose 2.5 dL/h and
40 dL as typical PK estimates for CL and V, respectively. Next, ChatGPT
was queried to produce a population PK model in R using the previous
typical PK estimates. The R code that ChatGPT generated for the base
population PK model is displayed in figure 1. Not only does ChatGPT
produce functional code, it also gives an explanation of how it works.
Initially, the R code sets the initial FVIII level is zero, assuming no
dose was given. If a dose was administered, the initial FVIII level
(represented by C_0 in the code) as the dose divided by V.
We then iteratively asked ChatGPT to add components to the model.
ChatGPT understood how to normalize the PK parameters to body weight
using allometric scaling and how to include a baseline FVIII level (F0)
as a parameter, as shown in figure 2.