I have a PK model for which I would like to include a period (occasion) effect then explore whether a categorical covariate may also have an effect. There is overlap since the categorical covariate is related to the administration of a second drug, which was always in a different period than the administration of the drug alone. It is a small dataset. I have set it up as a 2-compartment IV bolus crossover model and included the period on both volumes and clearances. There are 3 periods because the second drug was also given alone in a separate period (so there were 3 treatments: drug A, drug B and drug A + drug B); since I am analyzing drug A, the data for drug B is not included and I have 2 treatments (1 and 3). When I set the treatment as a categorical covariate (but do not assign it to any parameter), the population covariate box plots appropriately show the 3 periods, but only show one treatment. I cannot figure out why.
Thank you,
BP
can you provide a screen shot and give more details how your time/occasion / period is setup ?
the interface never supported time varying covariates and kep the first value/occasion from each subject.
you can ask for a table that output all the data you need and then do the boxplots that you are after.
I am away from my PK computer, but the covariates are not varying with time. The data is setup similar to the screenshot of the Excel spreadsheet included. The variables assigned to the model are ID for ID, Time period for time, CObs for CObs, A1 for A1, reset for reset, period as an occasion covariate with 3 possible values (1, 2 and 3) and treatment as a categorical covariate with 2 possible values (1 and 3). Since it is a crossover model, the order input option is not selected in the run options and reset is selected in the input options. The data is setup so that all times for period 1 are between 0 and 480, for period 2 between 500 and 980 and for period 3 between 1000 and 1480, with a reset at 0, 500 or 1000. The box covariate plots (I am trying to see if treatment has an effect) only show treatment 1.
Here are some screenshots illustrating my issue: data setup, data mapping, covariate setup and Pop Covariate Box plot. Note that issue persists if I omit the period effect and only include the treatment covariate, and if I assign the treatment covariate to parameters, I get a value but the box plot still only shows one treatment.
This is showing treatment is time varying meaning the same patient ID received treatment 1 and 3 and the gui will show you only the treatment = 1 first occurence. if your model have IOV and or a time varying treamtment effect you need to output a table and do the boxplot separately. as the posthoc automatically generated only at time 0
