Optimization of sampling of injectable anticancer drug preparations and improvement of analytical control in a Centralized Preparation Unit

Introduction
Our unit prepares around 25.000 preparations a year whose less than 10.000 under a subcontract. Analytical control (AC) represents approximately 80% of the exhaustive control of our production. One of the main causes of nonconformities (NC) of AC results is the quality of sample and its representativeness of the controlled preparation.
The objective was to compare two sampling techniques in order to keep the most efficient method to guarantee a representative sample of the content of the final preparation.

Methods
A prospective comparative observational study was conducted between January and February 2019. The AC was carried out using the control system QcPrep+^® (Icônes Services) on two samples taken from the same preparation, using two different techniques. The current technique (P1) consists of a sample taken at the pocket delivery site (injection site of the drug) using a BD Vacutainer^® Luer Adapter and a BD Vacutainer^® One-Use Holder. The tested technique (P2) consists of a sample from the pocket injection site using a BD Eclipse^TM Safety Needle. The study included preparations whose active ingredient (AI) can be controlled analytically, whose final volume is greater than 100 mL and for which there is the most NC (absolute value of relative error > 15%) based on a 2018 NC review. The number of NC with the two sampling techniques was compared using a chi-square test of homogeneity (α=5%). The number of sting injuries was noted. A paired-sample t-test (α=5%) was used to compare the volume available for the QcPrep+^® (V) as well as the time it took to sample (T).

Results
431 preparations from 13 different injectable cancer drugs were controlled. 36 NC (8.4%) and 18 NC (4.2%) were observed for P1 and P2, respectively (p=.011). One of the AI represents 67% of NC. Removing this AI, 15 NC (4.3%) and 3 NC (0.9%) were observed for P1 and P2, respectively (p=.004). No sting injuries were reported. The average V is 0.89 mL and 0.95 mL for P1 and P2, respectively (p=.0002) and the average T is 11.5 s and 18.1 s for P1 and P2, respectively (p=.0001).

Discussion-conclusion
The number of NC was significantly reduced by a better representativeness of the content of the final preparation by sampling from a different site of the injection site of the drug and by a larger available volume. The safety of the manipulators is warranted. Despite a larger but not significant T for daily practice, the P2 technique was adopted within the unit. A study will be conducted to better understand reasons for the high percentage of analytical NC attributed to only one of the AI.

See the poster