91 Potency Test Examples

The potency tests displayed here were collected from laboratories around the world.  All test names and other identifying information have been changed. Some of these potency tests were in the development stage, and some were screening assays for test samples containing indistinguishable bioactive material.  The graphs are of pools of 12, 20 or 30 assays, each assay having from 1-10 test samples.  Precision and residual outliers were automatically masked.  Many of the potency issues that are readily apparent in some of these pooled assay views, such as recurring hooks, lack of saturation plateaus, and widely divergent dilution curves, were not obvious when the assays were only viewed individually.  These issues, however, are quite evident in the many assay development graphs and tables in STATLIA MATRIX.  These tools can be used to optimize dilution doses and other factors to make your potency tests more stable and precise.  Examples of this optimization are shown in other case studies on this site (076EPotency Dilutions and 046NPotency Dilutions).  Contact the company for more information about any of these tests.  There are four graphs displayed here for each potency test:


The first graph shows the adjusted responses of the linked standard dilutions for each of the pooled assays, adjusted to remove label variation if a tracer was assayed.


The second graph shows the distribution of adjusted responses from each baseline and standard dilution in the pooled assays in a 3D histogram.


Residual Parallelism (also called Difference Testing) results in a measure of the similarity between the reference standard and unknown test sample curves, the weighted nonparallel RSSE. The third graph is a histogram of these nonparallel RSSEs from the unknown test samples in the pooled assays.  Unknown test sample results from critically failed standard or unknown curve regressions are excluded. The red line is plotted at a chi-square probability threshold of 0.01 for the nonparallel RSSEs. The cumulative percentage at each nonparallel RSSE level is plotted on the right axis in green.


Confidence Interval Parallelism (also called the Equivalence Method) is a measure of the lower and upper confidence limit ratio (or difference) deltas of the Unknown / Standard curve coefficients.  Confidence intervals were determined at 0.9 significance using the Monte Carlo method, which is more stable and precise than linear approximation or profile method estimates.  The fourth graph shows the upper (top) and lower (bottom) confidence limit ratio deltas plotted as green vertical lines for the 4PL a, b, and d coefficients of the pooled assays.  Gray lines are confidence limit delta ratios from critically failed standard or unknown curve regressions or critically failed parallelism determinations.  The upper and lower goalpost borders, plotted as blue vertical lines, were computed at the 99th percentile of the respective pooled assay ratio deltas after excluding failed deltas.