On Chemical Analysis – the truth

Chemical image

As I mentioned in my last article, some very interesting information came into the light at the ISO/TC 194 meeting in Berlin last year concerning variability in chemical analysis of medical devices. In that article I discussed repeatability, but this time I want to focus on Inter-laboratory consistency or Reproducibility.

I’m indebted to my ISO/TC 194 colleagues particularly: Anita Sawyer, Ted Heise and Bob Przygoda, and many others for the substance of what I present here. This article contains graphics that will require your scrutiny, so you might need a monitor, rather than a phone screen.

In chemical analysis: “Nothing is Free.”

(That’s the truth.)

You only get what you pay for. Just paying more won’t necessarily get you more (or a better result.) Your choices need to be informed. You need to know what you are paying for, and the law of diminishing returns indicates that the more you pay, the less you get (after some tipping point.)

“Paying” is not just throwing money at a supplier for a product or service. It is also investing time/effort/money to ensure you get what you need.

Reproducibility is difficult to achieve because every laboratory, indeed every instrument, is different. The following results are from “somewhat single variable” tests recently conducted by ISO/TC 194/WG 12 colleagues.

Two samples were extracted for 72 h, at both 37ºC and 50ºC, at both 6 cm2/mL and 12 cm2/mL, in Water, Ethanol/Water and Isopropanol. The extracts were run through a GC/MS by 4 separate laboratories (A-D). The vertical axis (comparison) is number of compounds identified by GC/MS.

For Sample A the comparison looks like this:

The first thing to notice is that Lab B apparently has a GC/MS that is most sensitive to the compounds in Sample A. The other Labs do not seem to be “sensibly contributing” to the discussion. (Not that “more” is necessarily “better!”)

You will also notice that the surface area:volume ratio is not as significant as might be expected.

The effect of temperature is where it gets interesting. It does make sense that hotter extraction results in more extractables. But the results falter somewhat for the hotter IPA extract. Not that I want to question the veracity of the data, but it does look a bit like Lab B has swapped the IPA results between 37ºC and 50ºC.

So now I reveal to you that Sample A is a PolyUrethane film. Does knowing that change anything about the interpretation of the data? Probably (hopefully!) not. Except that the IPA at 50ºC results from Lab B now seem like they really need to be > 35.

For Sample B the same comparison looks like this:

Again, Lab B is most sensitive. (And if anything, the contribution from the other Labs has decreased.)

And this time we can see that Sample B is not at all polar, which is true because the sample is High-Density PolyEthylene.

You might notice that the surface area:volume ratio is expected, given what we now know.
There isn’t any question about results being swapped this time.

The results from Lab B are mostly consistent between Samples A & B.

Takeaway # 1.

  • If you are mostly using one sort of material, then you can easily do this sort of a Lab comparison.
  • The comparison can be simpler. Say: 2 samples, 2 vehicles, 1 time/temperature, n Labs.
  • You could also make it more powerful by quantifying the compounds identified.

Two More Samples. Now let me show you the results for Sample C, (same conditions as above, except ratios are now g/mL).

Clearly, Lab D now has the GC/MS that is most sensitive to the compounds in the Sample. The other Labs are not sensibly contributing.

However, here are the results for Sample D:

Lab D still looks like they are in front. Except, these are reference materials for biocompatibility testing, and their compositions are very tightly controlled.

Samples C & D were obtained from the National Institute of Health Sciences (NIHS), Division of Medical Devices in Tokyo, Japan. Sample C (Y-4) was comprised of 57.8% PVC, 31.8% DEHP, 5.8% Genapol® X-080, and 4.6% EBSO [w/w]. Sample D (Y-1) was comprised of 61.3% PVC, 33.7% DEHP, and 4.9% EBSO [w/w].

Sample D contains 1 less substance than Sample C. There should be a very strong correlation between the GC/MS results for these two materials. Such a correlation is missing.

These results for Sample D look to be of much lower quality than those from Sample C, from the same Lab. Lab D does not appear to be very consistent.

Takeaway # 2.

  • Even if you are using many sorts of material, then you can easily do this sort of a Lab comparison.
  • You don’t need to use reference materials, just known materials. You can even make your own.

There’s more data available that I won’t share here, because it all paints the same picture.
LC/MS comparison showed one Lab was more sensitive for the Samples A & B, and another for Samples C & D, but the gaps were less clear.
ICP/MS comparison for Samples A &B was an unresolvable mess, but much clearer for C & D.

Takeaway # 3.

  • As extraction is clearly a variable, supply all the Labs with one controlled (reference) extract, as a baseline for the comparison.

Reproducibility is difficult to achieve because there are multiple factors that determine laboratory, and instrument, outcomes.

Here are some of those factors:
Variability from any processing of extract

  • Solvent reduction
  • Reconstitution
  • Storage until analysis (e.g., stability)

Variability in analytical methodology

  • Chromatographic column chemistry and configuration (e.g., film thickness, particle size, packing density)
  • Mobile phase (carrier/solvent choice, purity, flow rate, details of gradient)
  • Injection parameters (method, volume)
  • Plasma conditions (ICP)
  • Choice of detector (e.g., FID, UV-vis, MS, etc.)
  • Mass spectrometry approach (TIC, SIM, etc)
  • Inherent instrumental variability (e.g., laboratory temperature)
  • Operator

If it is a screening analysis (looking for unknown substances without a reference standard) it will also depend on the quality of matching to a chemical library for identification, can be affected by:

  • Robustness of the library
  • Algorithms used for matching
  • Analyst experience

Takeaway # 4.

  • Reproducibility is elusive.
  • So perhaps we should just give up on that concept.
  • Rather than pursuing the unachievable, each manufacturer might be better served to establish which lab to use for which material/s for which method/s.
  • Keep a tidy and compelling justification on file, to show why any results are dependable.

A clear lesson here is that a device manufacturer cannot simply throw a test article over the wall to a CRO and get back an Extractables & Leachables Analysis that is dependable. This is definitely going to be a “team play.” Not long ago, I was using three separate Labs for: VOC/SVOC leachables, VOCs desorbed off charcoal tubes and metals in solution.

Biocompatibility is a risk management exercise. Looks like this is going to be one of those risks we are going to have to manage.

Help button-Quick Call-Brandwood Biomedical

Looking to fly south for warmer weather? We have the skills in all aspects of Australian regulatory and reimbursement requirements and can assist you to prepare and file regulatory submissions to TGA. We can also hold Australian licenses on behalf of international manufacturers. Talk to us today about how to leverage international reviews to achieve TGA registration. Contact us to discuss your needs and how we can help. You can drop us an Email [email protected] or call 1 888-271-5063 (US toll free) ♦ 400-842 7017 (Beijing – toll free) ♦ +61 2 9906 2984 (Sydney)


Recent Posts






Translate »