# Terminal BLQ Values and AUCinf Estimation **URL:** \ **Category:** Phoenix\ **Created:** [August 10, 2026, 4:22pm UTC](https://forum.certara.com/t/terminal-blq-values-and-aucinf-estimation/4419 "2026-08-10T16:22:33Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![Mohamed\_Dabour](https://avatars.discourse-cdn.com/v4/letter/m/439d5e/32.png) [@Mohamed\_Dabour](https://forum.certara.com/u/Mohamed_Dabour)\ **Post date:** [August 10, 2026, 4:22pm UTC](https://forum.certara.com/t/terminal-blq-values-and-aucinf-estimation/4419/1 "2026-08-10T16:22:33Z")
In a DDI study that included 27 subjects, our analyte was assessed with and without a co-administered drug for 0-120 h. The analyte was not detectable in any subject after 24 h and was detected in only 6–8 subjects at the 24-h time point and in 20 subjects at 12 h. All concentrations after 24 h through 120 h were BLOQ. I am trying to fit the data to assess the GMR for AUCinf, but I am challenged by the fact that, for most subjects, the percentage of extrapolated AUC exceeds 20%. Can the algorithm in Phoenix impute values based on the observed terminal phase? Please note that I have tried the imputation methods outlined in [this paper](https://www.tandfonline.com/doi/full/10.1080/19466315.2019.1701546). However, the methods described require at least some subjects to have observed values at these time points (24-120) to impute values for the other subjects. The elimination half-life of our analyte typically ranges from 10 to 20 hours. ---
**Author:** ![SimonDavis](https://avatars.discourse-cdn.com/v4/letter/s/d9b06d/32.png) [@SimonDavis](https://forum.certara.com/u/SimonDavis)\ **Post date:** [August 14, 2026, 6:20pm UTC](https://forum.certara.com/t/terminal-blq-values-and-aucinf-estimation/4419/2 "2026-08-14T18:20:19Z")
maybe try building a compartmental model instead if the data is very limited inthat region. You experiment with the BLQ option in the model too, since instead of completely **censoring** those data, model that it’s somewhere between 0 and the LLOQ. - In Maximum Likelihood models, the engine uses the **M3 method** to calculate the probability that the value falls below the LLOQ instead of using a single point value. [Certara NLME Engine Knowledgebase • Certara.RsNLME](https://certara.github.io/R-RsNLME/articles/NLME_KB.html#part-5-observation-and-error-models) [Handling of LLOQ values in Phoenix](https://www.youtube.com/watch?v=xX-yCO5Rzag) Alternatively Sparse NCA could be an option,an example is detailed inthe help. [Sparse\_sampling\_NCA\_example](https://onlinehelp.certara.com/phoenix/8.8/Phoenix_UserDocs/NCA_Object/Sparse_sampling_NCA_example.htm) Simon. ---
**Author:** ![SimonDavis](https://avatars.discourse-cdn.com/v4/letter/s/d9b06d/32.png) [@SimonDavis](https://forum.certara.com/u/SimonDavis)\ **Post date:** [August 14, 2026, 6:24pm UTC](https://forum.certara.com/t/terminal-blq-values-and-aucinf-estimation/4419/3 "2026-08-14T18:24:49Z")
also if you want to share some example data here as a CSV or PHXPROJ we could advise better, or send it in to [support@certara.com](mailto:support@certara.com) if it is confidential and we will respond privately from there, Simon.