﻿ Statistical Rationale of SST Requirements - Chromatography Forum

## Statistical Rationale of SST Requirements

Discussions about GC and other "gas phase" separation techniques.

### Statistical Rationale of SST Requirements

Hello,

When creating a new system suitability method, how does one go about determining the acceptance criteria of critical parameters like peak area percent RSD and tailing? I've looked in the USP, EP, and JP, and none seem to list specific requirements (e.g. the JP says "The allowable limit of "System repeatability" should be set at an appropriate level based on the data when suitability of the method for the evaluation of quality of the drug was verified, and the precision necessary for the quality test.")

Agilent, for example, requires a peak area RSD of 2% to pass their annual preventative maintenance. What process would they have gone through to ultimately justify this value?

Thanks

### Re: Statistical Rationale of SST Requirements

Hello.
EP 2.2.26. System suitability:
0,8 < As < 1,5 unless otherwise stated
max(RSD) = K*B*(n)^0.5/t unless otherwise stated
Best regards,
Dmitriy A. Perlow

### Re: Statistical Rationale of SST Requirements

Hello,

I've seen that equation in the three pharmacopoeias. Unless I'm interpreting it incorrectly it seems to apply only when you're using a chromatography method specifically to assay your drug substance. We're using GC headspace to quantify methanol concentration in the drug substance. The Residual Solvents section of the USP, 467, shows %RSDs of about 1 for methanol at the 3000ppm limit. However, we're working far below that (down to 10ppm), so I doubt a %RSD of 1 is feasible.

### Re: Statistical Rationale of SST Requirements

Yep, unless otherwise stated. In house method & validation can state other requirements too.
Best regards,
Dmitriy A. Perlow

### Re: Statistical Rationale of SST Requirements

Focusing on area RSD here, rather than try to use a "one size fits all" criterion (which often actually means "one size fits none", it may be better to take a step back:
The measurement system precision needs to be better than the precision of the process being evaluated or validated. That's what they mean by an "appropriate level".

I've always used a factor of 3-4 in RSD as a rule of thumb (i.e., if you need to control the analyte (whether its an API, impurity, solvent or whatever to +/- 10% RSD, then you need to measure with and RSD < 3%).

So the first thing you have to do is determine how tightly you need to control your product or process.
-- Tom Jupille
LC Resources / Separation Science Associates
tjupille@lcresources.com
+ 1 (925) 297-5374

### Re: Statistical Rationale of SST Requirements

Is the 3-4x factor considered valid solely by virtue of being more precise than what is called for in the method? It makes perfect sense logically, but I wonder how I'd go about potentially justifying it to an auditor (are there any specific mathematics behind it?).

As this is a residual solvents test we're just ensuring the methanol concentration is below the limit, rather than trying to attain a specific value. If any methanol is present, it requires peak area %RSD for the triplicate samples to be <10%. Could this be the basis for a RSD <3% requirement?

### Re: Statistical Rationale of SST Requirements

Bumping this thread because I'm working on this topic again.

I've read about people doing multiple six-injection replicates over a few days to determine typical operating conditions as the basis for system suitability criteria (peak area RSD, tailing, plates, resolution, etc.). What statistical operations are they performing specifically to derive the numbers they end up using?

As I said above, our residual solvent test calls for a peak area %RSD <10% IF any is found. Based on that, I find 3.0-3.5% for SST to be reasonable, but I'd prefer having some mathematical justification in the event an auditor starts digging. Our typical %RSD for injections is ~<2.0%.

### Re: Statistical Rationale of SST Requirements

Your question is partly answered in USP section <621>. A %RSD of 2.0% or less requires 5 injections (this is usually a high concentration like an assay), while a %RSD >2.0% requires 6 injections (typically a low concentration like a related substance or RS). For the RS I have set 5.0% as the acceptable %RSD maximum. Statistically a %RSD of 10.0% is the LOQ (based on the standard deviation mathematics of the linear equation).

### Re: Statistical Rationale of SST Requirements

ICH Q3 also requires (?) that no interfering peak is more than 0.05% wt/wt of the therapeutic daily dosage (TDD).

### Re: Statistical Rationale of SST Requirements

HPLC chemist wrote:
For the RS I have set 5.0% as the acceptable %RSD maximum. Statistically a %RSD of 10.0% is the LOQ (based on the standard deviation mathematics of the linear equation).

Can you expand on reporting an LOQ as a %RSD? I've never seen that before. Additionally, is the 5.0% an arbitrarily chosen value under the 10.0%?

### Re: Statistical Rationale of SST Requirements

Can you expand on reporting an LOQ as a %RSD? I've never seen that before. Additionally, is the 5.0% an arbitrarily chosen value under the 10.0%?
Actually, that flows from the definition of LOQ: the lowest level at which the analyte can be determined with the requisite precision and accuracy. So, if you need at least 5% RSD (see my earlier post on that), then LOQ is the lowest level where you get that precision. If 10% is good enough for your purposes, then use 10% as the cutoff.
-- Tom Jupille
LC Resources / Separation Science Associates
tjupille@lcresources.com
+ 1 (925) 297-5374

### Re: Statistical Rationale of SST Requirements

Hi Curio1,

vhttp://www.eoma.aoac.org/app_f.pdf

Tables A4 - A6.

I Agree with Tom Jupille, when s/n is 10, this reflects a %RSD of 10...usually set at the LOQ level. When s/n is 3, this reflects a %RSD of 33 or so...usually set at the LOD level.
MattM

### Who is online

In total there are 2 users online :: 0 registered, 0 hidden and 2 guests (based on users active over the past 5 minutes)
Most users ever online was 234 on Tue Feb 06, 2018 7:33 am

Users browsing this forum: No registered users and 2 guests