Strategies to minimize variability and bias associated with manual pipetting in ligand binding assays to assure data quality of protein therapeutic quantification

https://doi.org/10.1016/j.jpba.2010.04.025Get rights and content

Abstract

Bioanalytical laboratories require accurate and precise pipetting to assure reproducible and accurate results for reliable data. Two areas where pipetting differences among analysts lead to poor reproducibility are long term stability testing and sample dilution. The purpose of this paper is to illustrate the problems with manual pipetting, describe an automation strategy to mitigate risks associated with manual pipetting, and provide recommendations on a control strategy that properly monitors samples requiring dilutions.

We determined differences among various manual pipetting techniques by analysts within a laboratory. To reduce variability in pipetting, a flexible modular liquid handling script was created on the Hamilton Microlab Star (HMS) to perform sample dilution, pre-treatment and plate loading. The script is capable of handling variable dilution factors. Additionally, two dilution controls were prepared and tested at concentrations of high and mid quality controls (QC). These same dilution controls were incorporated into both pre-study validation and in-study QCs to monitor dilution processing and assay performance.

Variability of manual pipetting among 11 analysts was more negatively biased with increasing dilution. Forward and reverse pipetting delivering different volumes contributed to the discordance. The dilutional bias with manual pipetting was eliminated using the liquid handler. Total error of dilution controls was less than 20%. The in-study pass rate was 100%.

Application of liquid handlers minimizes the variability and bias due to manual pipetting differences among analysts. The incorporation of dilution QCs serves a dual purpose to monitor the dilution process of the samples as well as the binding assay performance.

Introduction

Ligand binding assays have been used extensively to measure protein therapeutics. These methods are more variable than chromatographic methods that are used to quantify chemical entities [1]. The major contributing source of variation is pipetting technique, which is often performed manually. This problem is further amplified by the additional pipetting required for dilutions of most protein therapeutic study samples into the standard curve range of the binding assay. Variability in manual pipetting has not been investigated for its impact on ligand binding assay used to support protein therapeutic pharmacokinetic (PK) or toxicokinetic (TK) studies.

Publications from a pipette vendor have described optimal pipetting procedures, factors and techniques that influence the accuracy of pipetting [2], [3], [4]. The delivered volume can be significantly different due to environmental factors such as temperature [3] of the fluid being pipetted, the delivery method (forward or reverse pipetting) [4], and relative humidity in the ambient environment [2]. Therefore, operators using different methods of pipetting at different conditions could lead to different results.

Long term stability (LTS) testing has been challenging because multiple analysts and standard (STD) preparations are involved over time. To set up the LTS tests, often one analyst prepares stability test samples and at the designated time point another analyst prepares a fresh STD set to be used for the test of the previously prepared stability sample. Higher variability has been observed for the stability samples than that of QCs during pre-study validation. It is not uncommon that results at later test time points would be outside of the a priori established acceptance criteria. In some instances the observed values are higher than that of the baseline value for a protein therapeutic that is known to be stable. The method variance prohibits a clear interpretation on the analyte stability since the differences could be related to operator differences and not to changes in analyte stability.

Protein therapeutic quantification often requires very large dilutions. During an inter-laboratory comparison, we observed that higher concentration samples had greater variance. These results indicated that the dilution process could be a source of the increased variation. To address variability issues of the stability testing and sample dilution, we have developed an approach to reduce the variation associated with STD/QC preparations and sample dilution. We use automated liquid handlers for both processes to improve accuracy, precision and reproducibility. In addition, we have incorporated QC at concentrations above the STD curve range to reflect the study samples and determined their pre-study validation accuracy/precision total error, and use these QCs to monitor and accept/reject in-study runs.

Liquid handlers have been widely used for nearly three decades in the pharmaceutical industry in various aspects of drug development. However, it has been challenging to integrate them effectively for bioanalysis of protein therapeutics that support PK/TK studies. One of the challenges is the requirement of large and variable dilution factors depending on the dose and time point collected. In 2006 a paper described an in-house developed software program that fully automated the process for sample preparation using a Tecan Genesis integrated with Watson LIMS [5]. We asked the vendor to write a script for a Hamilton liquid handler and adopted a similar integration strategy by creating a Watson export file to be read by the liquid handler to prepare STDS, QCs and perform dilution at various dilution factors.

The purpose of this manuscript is to describe our findings on variability from manual pipetting, to illustrate the use of automated liquid handlers to successfully reduce this source of variation. We also present a revised monitoring process that accounts for both the sample dilution and binding portion of the immunoassay.

Section snippets

Materials and equipment

The following equipments were used: Spectra Max 340PC plate readers (Molecular Devices, Sunnyvale, CA), ELX-405 plate washers (Biotek, Winooski, VT), Titermix 100 plate shakers (Brinkmann, Westbury, NY), model 2005 incubators (VWR, West Chester, PA), Hamilton Microlab Star (Hamilton Robotics, Reno, NV), and Rainin pipettes (Oakland, CA).

All protein therapeutic standards and immunoassay reagents were produced and prepared by Amgen Inc. (Thousand Oaks, CA). Sera from non-human primates and humans

Long term stability test

LTS was tested over 3 months by two analysts using manual pipetting. Analyst 1 prepared the stability samples, assayed the baseline stability samples and two subsequent stability time point samples. Analyst 2 assayed the 3-month time point samples. Fig. 3 shows that the 3-month results biased high and were unacceptable, indicating stability problem. These results were repeated by the same analyst confirming the high bias. In order to rule out operator differences, analyst 1 also assayed the

Discussion

This paper illustrates that variation in manual pipetting can impact the final results for ligand binding assays. The processes where manual pipetting can have a great impact are STD and QC preparations and sample dilution. Implementation of a standardized manual pipetting program is not an easy task that would involve consensus-building, thorough training, change of habits, and continual monitoring. An obvious option to mitigate the risk of manual pipetting was to remove the human component by

Conclusion

Manual pipetting differences among individuals can lead to differences in final immunoassay results. Operator training within a laboratory to assure consistent pipetting technique would be one way to decrease data variability and bias. However, a better approach is to identify the critical processing points and remove the human error component by using automation. We have optimized the use of liquid handlers to reduce the inter-operator variation in STD/QC preparations and sample dilutions and

Acknowledgements

We would like to acknowledge contributions from Dom Calamba, Beth Johnson, Chris Macaraeg, Jessica Manlongat, Robert Ortiz, Vimal Patel, Ramak Pourvasei, Amir Sharifi, and Jennifer Tsoi.

References (9)

  • M. Kelley et al.

    Key elements of bioanalytical method validation for macromolecules

    AAPS J.

    (2007)
  • D. Rumery et al.

    Extreme pipetting IV: it's not the heat, it's the…pipettes perform exceptionally well in high humidity

    Pharm. Formulation Qual.

    (2008)
  • G. Rodrigue et al.

    Extreme pipetting II: the effect of fluid temperature on data integrity

    Pharm. Formulation Qual.

    (2007)
  • R. Curtis

    Minimizing liquid delivery risk pipettes as sources of error

    Am. Lab. News

    (2007)
There are more references available in the full text version of this article.

Cited by (0)

View full text