The present and future of withdrawal period calculations for milk in the European Union: dealing with data below the limit of quantification

Abstract

The assessment of withdrawal periods for milk is affected by the occurrence of data below the lower analytical quantification limit (BLQ data) and the resulting uncertainty. The current regulatory approach for dealing with BLQ residues is simple and easy: BLQ data (and missing data) are arbitrarily reassigned a value of one-half the LOQ before any calculation on the data with one of the three currently applicable methods. Here, we reconsider the determination of the withdrawal period of milk with data below the limit of quantification. Theoretical background on analytical limits and pharmacometric considerations will be established. Then, we analyze the uncertainty problems caused by the current approach and propose a calculation solution (maximum-likelihood estimation handling left-censored data) included in nonlinear mixed-effects modeling. Finally, we illustrate this issue using a case example.

Probabilistic Physiologically Based Pharmacokinetic Model for Penicillin G in Milk From Dairy Cows Following Intramammary or Intramuscular Administrations

“…Currently, different statistical methods, such as the "Time To Safe Concentration," the "Safe Concentration based on Linear Regression" and the "Safe Concentration Per Milking" methods, were used to calculate the withdrawal period for milk (Henri et al, 2017b). However, these methods have various limitations, including the assumption of linearity of the logtransformed tissue/milk depletion data, which may not be true for all drugs (Chevance et al, 2017;Henri et al, 2017b). The nonlinear mixed-effects pharmacokinetic modeling has many advantages (Martin-Jimenez and Riviere, 1998) as a pharmacostatistical approach to estimate the withdrawal period for milk .…”

The present and future of withdrawal period calculations for milk in the European Union: focus on heterogeneous, nonmonotonic data

“…1), the 468 observations were considered as above the limit of quantification. The case of data below the limit of quantification (BLQ data) is examined in a companion paper (Henri et al, 2017).…”

The lower limit of quantification in pharmacokinetic analyses

“…It is noteworthy that Option 2 has been recommended for determination of withholding times in residue studies, where its use causes a conservative error; arguably, this is reasonable in risk management. Methods for using BLLOQ data for withholding time determination for milk have been recommended (Henri, Jacques, Sanders, Chevance, & Laurentie, ).…”

“…arguably, this is reasonable in risk management. Methods for using BLLOQ data for withholding time determination for milk have been recommended (Henri, Jacques, Sanders, Chevance, & Laurentie, 2017).…”

“…Estimation of the parameters may be conducted using maximization of the likelihood (Tobin, 1958). This approach has been recommended for veterinary residue studies (Henri J 2017) and has been adopted as a regulatory option (Anonymous 2000b).…”