Response Factor Variation and Uncertainty Factors in E&L Analysis

A Multi-detector Strategy for Reducing Response Factor Variation in Analysis of Extractables and Leachables (E&L)

Extractables and Leachables (E&L) commonly include antioxidants, slip agents, plasticizers, monomers, cross-linking and other agents. There are also a number of other kinds of common molecules that are very important, but these molecules pose distinct challenges because they are not intentionally added substances, and they are generally not commercially available as standards.

These unintentionally added molecules include oligomers, polymerization side products, process impurities, and degradation products. Organizations and researchers need to be able to quantify and identify these species, and the aforementioned lack of commercially available standards is actually the root of Response Factor Variation problems.

Understanding Response Factors and Limits of Detection

The Response Factor is the signal per unit concentration for an individual compound. Most detectors do not provide an equal response for different compounds at the same concentration, so Response Factor Variation is the norm rather than the exception.

It is also important to understand the Relative Response Factor (RRF) value. The Relative Response Factor can be defined as the slope of the calibration curve for a target compound, divided by the slope of the calibration curve for the surrogate standard. In practice, an RRF value of 0.58 would mean that any quantitation done with that surrogate standard will return a value that is only 58% of the true value.

RRF values are directly correlated to quantitative accuracy, and the LOD is also affected by Response Factor variation. A compound with a low RRF value will result in a poor LOD because its response is weak, so surrogate standards must be carefully chosen when establishing a LOD. Compounds should be at a range of RRF values, reflecting the potential diversity of any anticipated Extractables and Leachables (E&L).

Analytical Evaluation Thresholds

It is important to consider how low to go when performing an analytical evaluation of E&L, as well as establishing the correct LOD. The industry has defined the Analytical Evaluation Threshold (AET) to answer this question. This equation contains a Dose Based Threshold, as well as a series of parameters related to the device and the extraction process. It also contains an Uncertainty Factor (UF).

The purpose of the UF is to account for Response Factor variation. Typically, this equation is defined as UF equals one divided by one minus RSD, where RSD is the Relative Standard Deviation of the Response Factors for the distribution of all of the analytes being analyzed by the method.

Once an AET has been defined, the next step is to run a range of standards and determine the peak height or peak area associated with the AET concentration.

Problems with Response Factor Variation

Response Factor variation can lead to results displaying different peak heights for the same concentrations, resulting in compounds that are considered to be above or below the threshold depending upon which standards are used to set the AET level.

This problem is significant, and in 2018 a paper was published in The Journal of Pharmaceutical and Biomedical Analysis, which showed how the number of compounds considered to be above the AET changes as a function of which standard was used. This change was quite dramatic, ranging from a very low number of compounds to a very significant number of compounds based solely on which standard was applied to set the AET threshold.

The UF be used to compensate for this Response Factor variation by adjusting down the location of the AET line so that it is possible to capture all of the compounds which are at the AET concentration. This RF problem can be described as ‘AET under reporting’.

A universal method for analyzing E&L is needed, because it is not known in advance which kinds of compounds are going to be present in a sample. In other words, this is a screening analysis: users are not looking for a subset of compounds, just looking for anything that comes out of the device. During this process it should be noted that if the UF is not sufficient, it may not be sufficiently protective of the AET.

However, if a large UF is applied in order to compensate for a large Response Factor variation, this may extend beyond the sensitivity of the MS system, which could then require the use of concentration factors that could result in loss or degradation of extractables.

RF databases are very helpful here, as they allow the calculation of method-specific Uncertainty Factors – however, they do not reduce the Uncertainty Factor, they just allow its value to be calculated. This is important, but it is not a solution to this problem.

Relative Quantitation and Accuracy

When the RF of a compound is less than the RF of the standard, this does not provide a worst case estimate of the concentration – it overestimates the LOQ of the compound. When the RF of the compound is greater than the RF of the standard it is a worst case estimate, potentially resulting in a margin of safety less than one, which could cause the result to be rejected.

RF databases are useful here, in that they allow quantitative error correction for compounds in the database, but this will, of course, only work for compounds in the database. Unless all of the necessary compounds are present in the database, this is not a solution in itself. Additionally, UF factors do not compensate for quantitative inaccuracy - they are only used in the AET calculation.

Employing a Multi-Detector Approach to AET Evaluation and Quantitation

Using a Multi-detector Approach for AET evaluation and quantitation can help address these issues. This article describes two different chromatographic systems, commonly used for this purpose.

The first system is an LC system that contains a UV detector, a Charged Aerosol Detector, and a QTOF LCMS detector. This system is run in both positive and negative mode. There are in effect four detectors used in the LC system while the GC system uses two detectors: MS and a Flame Ionization Detector (FID). Using this system will provide two signals for each compound in GC, and four signals for each compound in LC; and using a combination of these signals will help mitigate these problems.

The signals on each detector are orthogonal, meaning that they are uncorrelated to one another. There may be a weak response on one detector and a strong response on a different detector for the same compound, because the principles of detection are distinct for the various detectors. It is very important to understand that this approach is not in any way combining these detectors. They are evaluated individually, and then the sum total of all results is used when evaluating at the AET.

For instance, if there is a weak response on any one detector, it is possible to compensate for this so long as there is a strong response on one of the other detectors. The only requirement for a compound to be included in the study is that the compound is captured by at least one detector.

Multi-Detection in Practice

One example application aimed to test the effectiveness of this strategy on a large dataset, analyzing a database made up of 217 unique compounds related to medical devices. These compounds had a very wide range of physiochemical properties, including a broad range of molecular weights from 93 to 1177 AMU, a wide range of log P values and a range of volatilities from boiling points ranging from 102 to greater than 600. DBE was from 0 to 20, meaning it had double bond equivalent values that ranged from no double bonds to a wide range of double bonds and the database included both acidic and basic species.

These compounds were analyzed on all six of the different detectors:

LCMS Detector in Positive Mode

An electrospray ionization, QTOF LCMS system was used. This system has some excellent attributes for identification, including high mass accuracy and excellent sensitivity down to the picogram range. This system has a polynomial response rather than a linear response, meaning that Response Factors will vary as a function of concentration. This system is also subject to matrix effects.

The RRF value was found to be in the range of 0 to 0.2, meaning that nearly 40 compounds in the database showed a response in this range. When comparing this to an average responding standard, the second most abundant outcome is to obtain a very strong response greater than 1.8. This has significant implications in terms of the AET and quantitative accuracy when using LCMS positive mode. The percent RSD is actually greater than 100%, meaning it is not possible to calculate a suitable UF factor protective of this distribution.

LCMS Detector in Negative Mode

LCMS negative mode returned a very similar distribution with a percent RSD that was also over 100%. However, in order to capture every compound, it is necessary that each compound be successfully detected on at least one detector above the AET. In this study, 54% of the compounds that were seen by LCMS were also seen by GCMS, meaning that there is the potential to capture those compounds by GCMS, as well as by the orthogonal detectors - the UV and CAD in LC.

UV Detector

Here, detection is based on the absorption of UV light. It is a highly linear detector, not subject to matrix effects and widely applicable to anything with a chromophore, offering nanogram level sensitivity. Distribution for the UV detector at 230 nanometers was found to be more Gaussian than that of the LCMS, and the RSD of this distribution is only 60%, meaning that it is possible to calculate a protective UF at 2.5 - much more reasonable than what was required for the LCMS detectors.

Charged Aerosol Detector

The Charged Aerosol Detector is a completely orthogonal detector, ideal for measuring any non-volatile species. It is curvilinear, which means that it has a fairly linear response. This detector is highly precise and widely applicable with nanogram level sensitivity, but it is affected by the mobile phase composition.

Distribution for this detector had the most abundant response between 0 and 0.2. This may seem poor, but here it is possible to readily predict when this outcome will occur because generally, volatile species cannot be successfully analyzed with the CAD. RSD for this distribution has a value of 65% and the UF value is calculated as 2.9 - again, a reasonable UF value.

GC Detectors

The electron ionization detector is an excellent detector for both identification and quantification. It is widely applicable, linear, and generally precise with nanogram level sensitivity. It has a percent RSD of 52%, meaning the UF is only 2.1. It is a robust detector from a Response Factor variation perspective.

In the few instances where the GC Detector does have a very poor response, 71% of the compounds detected by GC were also seen by at least one LC detector. Typically, this occurs when a compound is not very volatile, so those non-volatile species are better detected by LC - when the GC detector is weakest, the LC detector is strongest, allowing users to compensate for poor responding compounds.

Flame Ionization Detector

The Flame Ionization Detector is also completely orthogonal. It measures any carbon-containing species, has a linear response, is highly precise and is not generally subject to matrix effects. It is also widely applicable with picogram level sensitivity and a very wide linear dynamic range – overall an excellent quantitative detector. The Response Factor Distribution for FID returns a percent RSD of 54%, resulting in a relatively low UF of 2.2.

As the example above makes clear, there is no universal detector. Every detector has strengths and every detector has weaknesses, and some detectors will detect one compound better than another. The ideal approach is to use the best detector for each compound. It should be emphasized again that these detectors are not combined in any sense for data interpretation, beyond simply applying them all.

Evaluating AET

Evaluating AET and accurately quantifying compounds is best achieved using the combined strengths of all detectors. If the RRF value is greater than one, that means the response for that compound on that detector is going to be greater than that of the surrogate standard. If that is the case, then a UF of 1 is all that is needed to successfully register that compound as above the AET threshold.

This approach is still useful, even in instances where users may not have the full range of detectors listed above. Returning to the examples provided above, the database of extractables for different configurations of detectors at different UF values was plotted. This revealed that when applying a UF of 1, the highest extent of coverage obtainable is only 75%.

This does not seem to be sufficient, but by applying a UF of 2 with at least one orthogonal detector, in addition to LCMS and GCMS, it is possible to attain greater than 90% coverage of the database, no matter which orthogonal detector is applied. If on the other hand, a UF of 3 is applied, it is possible to obtain greater than 95% coverage using one orthogonal detector and greater than 97% coverage when using the full multi-detector solution.

Improving Quantitative Accuracy

The multi-detector solution can also be used to improve quantitative accuracy, and this is achieved by using a selection of appropriate surrogate standards.

For example, a common application uses two potential surrogate standards, Bisphenol A and Irganox 245 and two target compounds, Bisphenol A diglycidyl ether and Irgnox 1010. When quantitating Bisphenol A diglycidyl ether, the most chemically relevant standard would be Bisphenol A. Using this standard with UV detection will return a value that is 83% of the true value – a respectable level of quantitative accuracy, consistent with most guidance documents.

On the other hand, quantitating Irganox 1010 would likely involve the use of Irganox 245, which is the more similar chemistry and in this case, would return a value that is 92% of the true quantitative value. Substituting this for Bisphenol A for instance, would return a result that is only 32% of the true value. This is a good example of why appropriate surrogate standard selection is important.

In the example presented earlier, applying the appropriate surrogate standard led to 85% of the database compounds being identified within 20% of the true value and 91% were within 40% of the true value. Most importantly, none of the compounds were significantly underestimated.

Summary

There are definite advantages to targeted analysis in terms of quantitative accuracy and the specificity of a method. However, it is important to remember that extractable screening and leachable screening analyses are just that - screening analyses – and these should therefore be broadly applicable and able to capture unexpected compounds.

If a weak response is returned in any one detector, the best way to capture that compound is not through the use of UF, but rather to apply a better responding detector. This is because as long as every compound returns a sufficient response on at least one detector, that compound will be included in the study and captured.

As can be seen in the examples presented here, using a UF of 2 with the broadly constituted database of extractables, it was possible to attain 94% coverage using the full multi-detector solution. This was equivalent to applying a UF of 10 for LCMS and GCMS only; therefore mitigating the need for a large Uncertainty Factor while avoiding the application of unnecessary and large amounts of concentration. RRF databases are needed for proper standard selection and for determining the appropriate UF.

The standards of chemistry relevant to a device can qualify the method on an individual analysis and for an individual device. The use of multiple detectors enables the selection of a detector for quantitation, which has less RF variation, therefore improving quantitative accuracy. The importance of appropriate surrogate standard selection cannot be overstated.

When this approach was applied to the database of extractables in the examples here, 85% of the compounds were found to be within 20% of the true value and no compound was less than 60% of the accurate value. RF databases are important in this instance, but they are limited in that they are only useful for compounds contained in the database. Finally, it should be noted that UF values do not correct for quantitative inaccuracy.

This information has been sourced, reviewed and adapted from materials provided by Jordi Labs.

For more information on this source, please visit Jordi Labs.

Citations

Please use one of the following formats to cite this article in your essay, paper or report:

  • APA

    Jordi Labs. (2021, January 14). Response Factor Variation and Uncertainty Factors in E&L Analysis. AZoM. Retrieved on January 26, 2021 from https://www.azom.com/article.aspx?ArticleID=19891.

  • MLA

    Jordi Labs. "Response Factor Variation and Uncertainty Factors in E&L Analysis". AZoM. 26 January 2021. <https://www.azom.com/article.aspx?ArticleID=19891>.

  • Chicago

    Jordi Labs. "Response Factor Variation and Uncertainty Factors in E&L Analysis". AZoM. https://www.azom.com/article.aspx?ArticleID=19891. (accessed January 26, 2021).

  • Harvard

    Jordi Labs. 2021. Response Factor Variation and Uncertainty Factors in E&L Analysis. AZoM, viewed 26 January 2021, https://www.azom.com/article.aspx?ArticleID=19891.

Ask A Question

Do you have a question you'd like to ask regarding this article?

Leave your feedback
Submit