Reference Bridging for Bioassays: Techniques and Best Practice

Management of the reference standard over the life of a product is critical to ensuring consistent production of material. We always want to be able to draw a link between a new batch being produced and one that has demonstrated acceptable clinical properties, particularly safety and efficacy. That’s why reference bridging is so vital: it allows a new batch of reference standard to be swapped out, when, for example, the current reference batch is depleted. We’re going to investigate why reference bridging is required and introduce the statistical process behind a successful bridging exercise.

In the lifecycle of a product, there will come a time when the stock of reagent has simply run out. This is more likely to happen when small amounts of the reference were available in the first place, as may be the case for certain cell and gene therapy assays.

As we’ve discussed in extensive detail elsewhere, the comparison between a test sample and a reference standard lies at the heart of the concept of relative potency. By anchoring the potency of a test sample to the potency of a reference standard, we can mitigate the impact of the inherent variability of the bioassay system to measure the relative potency of the test sample with greater accuracy and precision than attempting to measure its absolute potency.

Importantly, therefore, the measured relative potency of a sample depends on the potency of the reference standard. If a reference standard of a different potency were used, the relative potency of the test sample would change also. Since batch release is usually dependent on the relative potency of a sample meeting specification limits, it is crucial to characterise the potency of a replacement reference standard so that future assay results remain comparable with the defined specification limits.

For example, consider Figure 1. Here, we see a test sample (orange) measured in the same assay as two potential reference standards, Reference 1 (pale green) and Reference 2 (dark green). The relative potency can be thought of graphically as related to the horizontal shift between a test sample curve and a reference curve on a plot. Specifically, the greater the rightward shift, the lower the relative potency of the test sample, as it takes a greater dose of the test sample to achieve a certain response.  We can see visually on the plot that there is a shift between the curves for the two references. This is an indication that there will be a difference in the test sample relative potency measured from the two references. And, indeed, we find:

Reference 1 is less potent than Reference 2, its curve is shifted to the right compared to that of Reference 2. In turn, this means that the rightward shift from the Reference 1 curve to the test sample curve is smaller than that from the Reference 2 curve to the test sample curve. Therefore, we measure the relative potency of the test sample to be greater when using Reference 1 than when using Reference 2.

This emphasises that, ideally, we would never change the reference standard. It is clear that a change from Reference 1 to Reference 2 would impact the results of the assay if the difference in reference potency isn’t properly accounted for. Since we are highly likely to face such changes, however, we must find methods to account for them.

In an ideal world, we would be able to use the same batch of reference for the full lifetime of an assay. In most cases, however, this is impossible – references run out, or reach their expiry dates. It is highly likely that a routine assay will face a change of reference at some point in its lifespan, particularly in the case of assays which use references which may be scarce, such as those for cell or gene therapy products.

In such cases, we can perform a reference bridging study. This is a study to quantify any differences between the old and new references. The key component of this is to understand how the potency of the new reference differs to the old.

Broadly, there are two ways one can approach a bridging study. One can seek to establish that the potency of the new reference is acceptably similar to that of the old reference, meaning that one can continue to use the new reference in the assay without the need to account for any differences in potency. This is known as an equivalence approach.

Alternatively, if the difference in potency is large enough that it would be inappropriate to consider the two references equivalent, one can choose to account for that difference – known as the offset – in future potency calculations. The goal here is to ensure that we have a sufficiently precise estimate of the offset.

Let’s take a closer look at the two approaches.

Ideally, we would like to establish that the new reference standard is equivalent to the outgoing standard. To do this, we first need to set equivalence limits. For some product, a sensible choice might be 90% to 111% of the potency of the old reference – this is close enough to 100% that we can reasonably claim equivalence. The two-sided 95% confidence interval on the geometric mean potency of the new reference must then fall entirely within those limits in order to be acceptable. Ideally, the limits would be justified to continue to support appropriate lot release decisions. That is, that the bounds are chosen such that the maximum allowable drift does not pose a risk of releasing ineffective and/or unsafe product.

A key component of any reference bridging study is the choice of sample size – how many measurements of the new reference potency are required before we have a reasonable chance of study success? This will depend on the chosen equivalence limits and the precision of the assay: the greater the variability of the assay, the more measurements will be required.

If we perform successive bridging exercises using an equivalence approach, we need to be aware that drift can accumulate. For example, we might accept a new reference as equivalent if its potency was 92% of the old reference. But what about when that reference runs out? Well, we might think we’re ok if the third reference is 92% of the potency of the second. But that means it’s 92% x 92% = 85% of the potency of the original reference. That falls outside the equivalence limits, so we should not consider the third reference as sufficiently equivalent for continued use in the assay with no correction.

Demonstrating equivalence may not be realistic if there is a large change in potency between the two references. Instead, we might choose to make a precise determination of the difference in potency between the old and new references so that it can be accounted for in future calculations.

The percentage comparison between the potencies of an old reference and its replacement is known as the offset. We can think of it as a measure of bias introduced into the assay results due to the difference in the potency of the references.

If the offset is 100%, our references would have identical potency, meaning we need not worry about any corrections.  If the offset is different than 100%, measured relative potencies will be biased with respect to results using the old reference. Specifically, if the offset is greater than 100% the relative potency will be lower, while if the offset is less than 100% the relative potency will be higher when the new reference is used.

For example, we can find the offset for the two reference samples we looked at previously – Reference 1 and Reference 2.

Reference 2 is, therefore, 140% as potent as Reference 1. In a real study, we would determine the offset using repeated measurements where we treat the new reference sample as a test sample.

Of course, estimation of the offset will carry some uncertainty: we do not expect to measure the same offset on each and every measurement. One way we can ensure that our measurement of the offset is sufficiently precise is to set an acceptance criterion based on the fold precision. This is the ratio of the upper and lower confidence limits on a measurement – here the offset. For example, let’s imagine we found confidence limits of (126%, 154%) on our offset of 140%. The fold precision on this measurement would be . Applying a criterion on the fold precision means we can ensure that the measurement of the offset we use moving forward is sufficiently precise. And, as in an equivalence-based study, we can use the fold precision with the variability of the assay to estimate an appropriate sample size for the study.

So, we have performed a bridging study and determined the offset. Hopefully the new reference has an identical potency as the old but, in the more likely scenario where there is a difference in potency, how do we account for the offset? This is important as we want to continue to use the specification limits determined for the original reference standard.

Thankfully, the solution is simple! When we measure the relative potency of a test sample with the new reference, we can multiply by the offset to map back to what we would have measured with the old reference. We can then compare this corrected value to our specification limits and determine whether the batch from which the sample was taken is acceptable.

Let’s return to our example. Imagine we’ve measured the same test sample with the Reference 2 and found the relative potency to once again be 0.49. This plot is shown in Figure 2.

and found the relative potency to once again be 0.49. This plot is shown below.

Recall that we determined the offset between Reference 1 and Reference 2 to be 140%. That means we can recover the sample potency relative to Reference 1 according to:

which is what we originally measured!

The above outlines the basic process by which one might go about replacing the reference in a routine assay. As always, however, there are some nuances which ought to be considered.

One detail to think about is whether the system suitability criteria for the assay continue to function as expected, particularly if any are based on the parameters of the model fit to the original reference. While the potency of the new reference can be accounted for, it is possible that the system suitability criteria might need to be revised to account for the change in reference.

Further, while one can correct for the offset, it is worth being aware of any accumulating error upon repeated replacements of the reference. No bridging study is perfect, and it is possible for a shift to build up inadvertently over time if the reference is replaced multiple times. One way to mitigate this is to keep an archive of old reference samples to compare with results using newer references and detect any changes.

This highlights the importance of pre-planning when it comes to reference bridging. A change of reference need not be a troublesome event in the life of an assay: there are well-established methods for bridging to the new reference which avoid a re-validation and allow the continued use of already established specification limits. But these require that the tools for implementing those methods – such as keeping an archive of reference material – are in place. A strategy for how any reference replacements will be handled should be discussed as early in the lifecycle of a routine assay as possible. And, ideally, those discussions should involve statistical input to ensure that you continue to get the most out of your assay.