System suitability limits for "fronting" peaks

[Rob Burgess]

What would you say were acceptable sys. suitability limits for peaks that appear to be fronting?

Would a value of 0.7 be deemed acceptable do you think? To me, a value this "low" tends to suggest a underlying problem with kinetics of the chromatography. I'd be happier with a value set to about 0.9. Any thoughts / comments?

[tom jupille]

As far as I know, there is no spec on fronting. The easiest to justify to an auditor would simply be the "mirror image" of the tailing spec:

TF = 1.5 mirrors to TF = 0.75
TF = 2.0 mirrors to TF = 0.67

I'm using the USP tailing factor here (front-to-back divided by twice the front-to-center. measured at 5% of max peak height).

You're right about the implications for the chromatography (although I'd argue it's more closely related to equilibrium and overload than to kinetics), but from a regulatory perpective, the major impact is probably on resolution; asymmetric peaks result in more overlap than would be suggested by the resolution value. Fronting and tailing are equivalent in that respect.

One way to set a spec is to recalculate resolution on your close peak pair(s), but instead of using average baseline width, use the sum of the back half-width of the first peak and the front half-width of the second peak. This is easier to explain with a sketch (apologies for the bad art work!):

Compare that to the value you get using the standard definition of resolution. That should give you an idea of how much the tailing is hurting your "effective" resolution; you can extrapolate from there to estimate how much tailing is allowable.

All of that is probably more trouble than it's worth. If it were my problem, I'd just mirror the tailing spec and let it go at that.

[John]

I have always thought it strange that tailing and fronting are quantified by peak symmetry. A peak whose fronting and tailing were equally awful would sail through system suitability with a perfect value of 1.
J

[tom jupille]

Tailing Factor (or the Asymmetry Factor, for that matter) are essentially one-point measurements used to describe a complex distribution. The have only one virtue (in my not-so-humble opinion), and that is that they are easy to do with printout, pencil, and ruler. Pre-computer, that virtue was enough to enshrine them as standards.

Statisticians have much better measures of deviations from ideality (skewness and kurtosis come to mind), but these can only be calculated from the numeric distribution (i.e., the raw data), not from the displayed chromatogram.

As long as our crude single-point measurements are good enough to do the job, there is little pressure to change. [warning: I'm switching sides in the argument now]. To take your example of a peak that's equally ugly in the front and back side, so long is doesn't overlap with the adjacent peaks (i.e., so long as resolution is adequate), the ugliness doesn't really matter a whole lot.

[John]

But surely the front and tail ends of a compound's peak could be low enough to be nearing the limit of detection and, therefore, doubts of linearity could arise in chromatography, where a significant part of the peak area lies in these low level extremeties.
JWW

[tom jupille]

But surely the front and tail ends of a compound's peak could be low enough to be nearing the limit of detection

A couple of comments here.

First of all, you're right, which is why I think skewness and kurtosis (that's the one that measures symmetrical distortion) would provide more useful information than tailing factor.

Second, while you're right, that comment holds true for any distribution, including the Gaussian. The ends of the peak always go below the LOD, so we're talking about a matter of degree rather than a fundamental difference. So long as the shape of the distribution stays constant, with only the amplitude varying with mass on column, the response should be linear.

And don't call me Shirley.

[HW Mueller]

Could we agree that linearity depends on the degree of flatness, broadness of a peak? Probably all of us have seen a huge, but very broad peak, not appear when a bit less of the substance was injected.

[Peter Apps]

When a peak is missed by the integrator the amount injected was below the LOD, irrespective of peak shape. All else being equal, sharp peaks give lower LODs than broad ones, and the linear range of the calibration will extend to smaller quantities of analyte when peaks are sharp than when they are broad. Integration repeatability is nearly always better for sharp peaks, as long as the sampling frequency is high enough.

We probably can all agree that sharp and symmetrical is better than wide and tailed, but I doubt that I am the only one who falls for the seductive appeal of a good looking chromatogram when a scruffy one run quicker and with less optimisation would have done the job just as well !

Peter

[Rob Burgess]

Hi all, thanks for all your repsonses; much appreciated as per usual.

The reason I have erred on caution and stated a value of 0.9 (or higher) is a robustness standpoint. If a peak is fronting, at say around 0.7, then from my own previous experience it can quite often be the case that that such a peak is more liable to occasionly be split. Does this make sense and what are other chromatographers experience regarding this...

[danko]

Hi guys,
No one disagrees that asymmetry is a bad thing – whether it’s detection limit that’s bothering one or just like me, bad chromatography (inappropriate pH, overload, temperature and so on).
And yes, the Tailing Factor is only god enough as long as no fronting is going on.
I have “privatelyâ€

[Peter Apps]

Hi Danko

Peak heights are usually in mV, absorption units, or some arbitrary count, and peak width is in seconds, milliseconds or minutes (time in other words) so how can you divide height by width ?

Peter

[danko]

Hello Peter,

Oh that’s not a problem: You just normalize the one to the other – let’s say 1 mV corresponds to 1 sec. or something like that. This factor, together with the rest of the calculation, can be entered in a “custom fieldâ€

[tom jupille]

Rob, I think setting a fairly tight spec for fronting is a good idea if you have found that fronting provides warning of impending problems. System suitability "suggestions" from FDA/ICH/USP are minimum suggestions. It's up to us as chromatographers to make them as tight as necessary to ensure validity of results.

My take on the situation is that fronting is fairly rare and usually associated with inadequate buffering, solubility problems in the mobile phase, tendency of analytes to form aggregates, or protein conformational changes (partial reversible denaturation). Most of those can be fixed.

Hans and Peter are both right in their comments on LOD - a broad peak will have a higher LOD than a sharp one. However, there are many cases where LOD is not an issue (pharmaceutical potency or content uniformity, for example, where you only have to demonstrate linearity over a range bracketing the nominal concentration). As long as the calibration plot is linear (i.e., only peak amplitude changes with load) then the peak shape is of secondary concern. That said, I like to think of myself as a craftsman in chromatography , and skewed or squashed peaks offend my sense of aesthetics.

Danko, I think your triangle test can be useful, but is arbitrary, because the numbers depend entirely on how you do your normalization (I can make the number 1000 times bigger by simply setting 1 μV = 1 sec instead of 1 mV = 1 sec). Having said that, if you pick an appropriate normalization, I can see it being a useful system suitability criterion. I'd hate to have to justify it to an auditor, though.

[HW Mueller]

Aesthetics sounds good (that´s why I hate flat peaks), it usually guarantees better results, it is actually experience.
On Danko´s second method: I have a problem with comparing peaks obtained at different flow rates.

[Victor]

Tom,

I think your suggestions on the lines of that "as long as the calibration curve is linear, the peak shape is not too important" is somewhat simplistic. Fronting or tailing peaks are an indication of some problem in the chromatographic system. As you have mentioned, fronting or tailing peaks can result from overload. This means that the peak shape will vary as a function of the sample mass, and also therefore that the resolution will vary as a function of sample mass. If you are looking for an analyte in a well defined and known matrix (perhaps this is the case often in Pharma), this may not be a problem. However, if the matrix is not well known, then the analyte can suffer variable degrees of interference dependent on the concentration of analyte (or matrix compounds). Tailing peaks may be a worse case, because these may result from some partial contribution of irreversible adsorption processes, which may vary dependent on the sample composition. In all, bad peak shapes might contribute to poorer reproducibility of the chromatographic process.

I think the situation is analagous in some way to the idea of linear calibration graphs. It is perfectly possible to use curved calibration graphs as long as the system is studied and calibrated very carefully so that all the conditions which affect the graph are well understood. However, linear calibration graphs are so much easier and more robust and reliable.

Hmm, this is not a good analogy in this case because you have clearly stated the idea of linear calibration in your posts... However, I hope the point I am trying to make is clear....it's more than just aesthetics.