Effect of a statistical outlier in potency bioassays

Abstract

The USP<1032> guidelines recommend the screening of bioassay data for outliers prior to performing a relative potency (RP) analysis. The guidelines, however, do not offer advice on the size or type of outlier that should be removed prior to model fitting and calculation of RP. Computer simulation was used to investigate the consequences of ignoring the USP<1032> guidance to remove outliers. For biotherapeutics and vaccines, outliers in potency data may result in the false acceptance/rejection of a bad/good lot of drug product. Biological activity, measured through a potency bioassay, is considered a critical quality attribute in manufacturing. If the concentration-response potency curve of a test sample is deemed to be similar in shape to that of the reference standard, the curves are said to exhibit constant RP, an essential criterion for the interpretation of a RP. One or more outliers in the concentration-response data, however, may result in a failure to declare similarity or may yield a biased RP estimate. Concentration-response curves for test and reference were computer generated with constant RP from four-parameter logistic curves. Single outlier, multiple outlier, and whole-curve outlier scenarios were explored for their effects on the similarity testing and on the RP estimation. Though the simulations point to situations for which outlier removal is unnecessary, the results generally support the USP<1032> recommendation and illustrate the impact on the RP calculation when application of outlier removal procedures are discounted.

Comparisons of outlier tests for potency bioassays

“…When statistical outliers are present in the measured concentration‐response data for the batch and/or the reference standard, Sondag et al show that similarity testing may be compromised and RP estimates may be biased . These findings agree with the USP<1032> recommendations to screen for outliers before RP analysis .…”

“…Sondag et al define a single observation outlier as a single measurement excessively distant from the other values (Figure , left panel). A concentration point outlier refers to the set of replicate measurements from the same concentration level at which all the replicates exhibit a different behavior (in the same direction) from the values predicted by the true model (Figure , right panel) . Four outlier tests from the literature and one novel outlier test are detailed in this section.…”

“…As in Sondag et al, three types of outliers are examined: Single observation outlier: A single measurement is excessively distant from the other values. Concentration point outlier: All the replicates at one concentration level differ from the values at the other concentration points. Whole‐curve outlier: One of the curve replicates is affected by a manipulation error. It can affect the dilution factor, one of the asymptotes, or both asymptotes, inducing a vertical shift. …”

“…Although outlier detection may be performed in conjunction with a litany of concentration‐response curves, as in Sondag et al, the assumed mean model for potency bioassay concentration‐response data is the four‐parameter logistic (4PL) curve. The usual parametrization was proposed by Rodbard and Hutt: where x is the concentration and θ = ( yMin , yMax , EC 50 , H ) is the parameter vector.…”

“…Before estimating RP, similarity testing is performed to determine if the shapes of the two concentration‐response curves are equivalent. The similarity test may be performed by considering the statistical equivalence of reference and test parameters yMin , yMax , and H , as described in Sondag et al The exact form of our similarity test is shown in Section S1. When no outliers are present in the data, the type I error of our similarity test is 5%.…”