Hi - I am building a popPK model of pediatric data and have hit a snag with my covariate plots. Currently for the Base Model, under the Parameters/Structural tab, I have added 7 covariates using the “Add from unused columns” button. None of them are checked “Yes” or “+1” for inclusion in the model. Four of the covariates (weight, creatinine, albumin, and age) are continuous variables. The others are categorical and have been designated as such in the Covar. Type tab. When I execute the model, it runs fine but when I look at the Pop Covar Plots, I only get plots for nCl,age; nCl,wt; nV,age; and nV,weight. I do not get any plots for creatinine or albumin. Is there a reason for this? Thanks for any insight. Kevin
Update to the question above: When I look at the Output Data under EtaCov, all of the outut columns are populated except for creatinine (scre) and albumin (albe). ID ecmo wt pna scre race sex albe nV nCl 1 1 3.16 24 1 0 0.248178 0.276888 2 1 3.4 20 0 1 0.390051 -0.286574 3 1 7 165 1 1 -0.31641 1.29809 4 1 2.01 18 0 1 0.211571 -0.179859 Both of these covariates have values populated for every row in the source data and they are mapped in the setup for this model.
Sorry, the columns above did not transfer. Imagine a column with empty fields under scre and albe.
I have figured out the problem but do not have a great solution. For the covariate creatinine, I had multiple values per subject (e.g. Sub 1 was on study drug for 2 weeks and had daily creatinines drawn. All of those values were included in the dataset and imputed forward so that every row had a value for creatinine). NLME was not able to handle multiple different values per subject. When I created a mean creatinine value per subject (i.e. each subject had only 1 value for creatinine) creatinine was incorporated as a covariate without difficulty. The problem is that renal function changed over time for many of these children. Incorporating the mean creatinine for the entire dosing period loses this level of detail.
Kevin, As I understand it Albumin and Creatininine were measured more than once during the course of the experiment, and are currently set as Continuous. I’m not an expert in Population modelling but I think you might perhaps want to look at setting it as an OCC variable rather than simple Continuous/Categorical covariate; Certara | Drug Development Solutions However any covariates of type “occasion” will not appear in any covariate plot and I you’d have to create additional plots yourself. Could you perhaps attach a copy of the project to better understand your data and mappings etc? Simon
When there are time-varying covariates, in some versions (6.1, 6.2), NLME may not display the covariate plots for those covariates, or in other versions (6.3) will display the covariate plots with only the first value of the covariate for each subject. However, the engines will estimate parameters based on the correct time varying values of the covariates. It’s not that NLME cannot handle time varying covariates, it’s that we hadn’t designed for plots for time varying covariates. I simulated some data with time varying covariates and tested NLME to see if it would calculate the correct parameters and it did, so I know it’s working. Emily
In order to plot etas vs. covariates for time varying covariates, there would need to be an eta-covariate plot for each time point. If there are several time points, this would result in too many plots. Or the eta-covariate plots would need to be three-dimensional, with time on a third axis. Emily
What version are you using?
I am using WinNonlin v 6.2, NLME v 1.1.
Have you has a chance to try your project in the 6.3 release that came out at the end of March yet? did that help? Simon.