Modeling variability in absorption and clearance (Auto induction)

Dear All, I am working on Efavirenz (NNRTI dosed at 600 mg/day oral) population pharmacokinetic data analysis (70 patients, 800 samples). Intense sampling data on day 1 and 14. Individual day 1 and 14 analysis shows an increased absorption in few patients plus increased clearance (auto induction) in some of these patients and other patients (who revealed no change in absorption). It is shown in earlier clinical studies that fed state results in increased bioavailability. I have no covariate information which may support this (available info Wt, Sex, total protein and TBR). Moreover the absorption profiles are very haphazard with troughs and peaks in the absorption phase, similar to what Bulitta et.al 2009, Pg 3462-3471 showed(Antimicrobial agents and chemotherapy). I have tried this Michaelis-Menton model in absorption with and without lag time ( run fails). NONMEM runs for one C.M. with no variability in absorption and clearance as an OBF value of 1498, 2CM 1390 (Vd is low 68 liters cv% 7 and Vd perip CV% 64 ). 2 CM with autoinduction model (for all patients)is OBV 1284 (run fails). Efavirenz preclinical trials shows that dose variability results in non linearity in absorption (Balani et.al 1998, vol 27, No.1), no clinical trials supporting this info and dose is 600 mg/day in my data. Auto induction was shown in clinical trials (zhu et.al). I have identified the patients who have clearance and absorption variability and created two columns in data sheet.Runs including the variability in absorption and clearance for some patients completely fails with Vd as low as 30 and perip as high as 400 liters. My questions to NM community are 1. How to handle this data if we dont have any covariate information for absorption and clearance variability. 2. Any references to model the data without crashing the runs Your suggestions are appreciated. Thank you very much in advance. Regards,