I just read a paper by V. R. Meyer from Chromatographia, 40, #1/2, pp 15-22 (1995) that indicated that to accurately assess the relative amount of a very small HPLC peak occurring on the leading edge of a large peak, peak height rather than peak area should be used. Is this the current conventional wisdom?

I have not too much confidence in peak height. I prefer peak area. However, I have at times been forced to use peak height.
My experience is that, to use height instead of area, you need to have a larger number of calibration points over a smaller dynamic range. For example, I have used 3 point calibration 100, 500, 1000ppm for peak area. I would suggest a minimum of 5 points over 100 to 500ppm or 9 to 10 points over 100 to 1000ppm range for peak height.
The peaks MUST be very thin for all standard and sample peaks if you are using height. I have observed that standard peaks are generally baseline resolved and very thin. Whereas sample peaks eluting on fronting or tailing larger peaks have a tendency to be wide.
I was recently asked to determine the concentration of a minor contaminant in a raw material. Using the method provided by the manufacturer, the contaminant appears as a small bump on the tail of a huge peak. I have spent the last 2 weeks changing columns and playing with mobile phase to separate the contaminant peak from the raw material peak. I was able to separate 6 homologues of the raw material but the anlyte of interest still eluted too close to the third homologue, C-12. Today, finally, I was able to find a column and gradient combination that worked. I would suggest trying to separate the peak of interest from the peak with which it co-elutes and use peak area. You might also consider SPE to separate the analyte of interest from the other component before sample injection. Both techniques are time consuming but result in better accuracy and precision. I would rather take the time and effort to produce one correct result that I am ready to defend with confidence, than quickly report erroneous values that I will have to explain later.

In J Chromatogr. B, 678, 137 (1996) we described an analysis of cortisol in plasma/serum in which we used peak hight for the integration. Even though we used 3 separation steps (one method with 3 columns, another with ultrafiltration + 2 columns)there was, in many patient´s samples, too complicated overlap for using area. Since two methods were used (we called this an intrinsic quality control) it was clearly indicated that integration by peak hight was better. Nevertheless, I agree that integration by peak area is to be preferred when the peak edges are defined.
The response was linear, so we did not use multi-point calibration. Incidentally, this study showed that RIA or ELISA of cortisol can yield extremely faulty values for some patient samples.

Meyer is an excellent chromatographer. I wouldn't call it conventional, but it does work.
As suggested by others, area is the only way to go with well defined peaks.

Best of luck.

I have encountered this integration problem over the years and the very best way to settle the issue is with your own data, see below. In general, when peaks are not well resolved, a mesurement of peak height will yield more precise integration than peak area, because the heights are less affected by errors in establishing the baseline from chromatogram to chromatogram than the area is affected. This is especially the case when a small component sits on the tail of a big one. One really needs to know their data system to set integration paraments to best handle this situration as small changes in skim parameters, or however you are handling this situation , will have a big effect on baseline for the small peak.

Data systems cheerfully report areas and heights, so quantitate with both for a while and see which is best in terms of presision.

As suggested above, the choice of height or area
for a small peak preceding a large one is best
made by generating data w both. Either way however
it’s a matter of peak resolution. This said,
intuitively peak height will be the choice for a
small peak which precedes a large peak (assume the
ongoing resolution of 2 peaks which coelute
completely to begin with. As resolution increases,
a point will be reached where the peak maximum of
the first peak has no contribution from the second
peak whereas at this same point the total area
suffers from second peak contribution). In
practice the peaks should be separated to allow a
max error of no greater than1% by either method of
calculation. Furthermore the ratio of larger to
smaller peak dictates the degree of resolution
required to meet the 1% specification. For
example, using peak height calculations; if the
ratio of the first /second peak is 0.1 the
resolution must be 3.7 sigma’s to meet the 1%
criteria and this is significantly less
separation
than that required by peak area. In summary then;
(1) you shouldn’t use either method of calculation
if you cant achieve sufficient resolution to
achieve +/- 1% (or whatever spec you you set) and
(2) there may be situations where you can meet
this goal only by using peak height

In reading over the preceding messages I think we may all, including myself, have missed something. How does the peak shape riding on the tail of the interfering species compare to the standard when run clean? Baseline resolved peaks are more likely to be single components. Riding peaks can be combination peaks. How you draw the baseline for a riding peak will affect your end result. I was given a project recently to determine the concentration of a residual raw material, less than 0.1%, in a sample of the concentrated end product, 40%, matrix. The calibration set provided well defined thin peaks. However, in samples it rode out on the tail of the main component's large flat wide peak. The thin, standard calibration set peaks, were the result of interaction between the stationary and mobile phases. The peaks from sample runs were the result of interactions between the two system phases and the interfering species. I have spent the last three weeks pulling peaks apart in my HPLC for samples supplied from 3 manufacturers competing for our contract. All of them use the same HPLC method I was originally provided. After developing a method for separating the indivual homologues for the main component, I was able to zoom in on the analyte of interst and achieve baseline separation between it and the main component peak's nearest homologue. Two suppliers were understating the concentration of the analyte of interest. The third, due to their different manufacturing process had other unknowns included in the peak they were assuming to be the analyte of interest. They actually had the lowest level of that component but a lot of other unknown materials.

Make sure your robustness/precision experiment supports the use of Peak Height. I have seen slight changes in temperature affecting peak height e.g. poor temp. control of the column heater cause by changes in the room temp. during the day/night.

While some interesting points have been introduced
the original query was concerned with a small peak
preceding a large one (case 1) - not a peak
occurring on the tailing edge of a large peak
(case 2). In my experience the small/first elution
pattern (case 1) is the situation in which peak
height calculations are most likely to be useful.
The reason being, in general, asymmetry values are
greater than 1: commonly encountered peaks tend to
tail to a greater extent than they front. The
result of using peak height calculations in the
small/ last situation (case 2) is that the data
processing system error in determining the base
line is exacerbated by now determining a sloping
baseline. While speculating is fun you really have
to base your decision from data you generate using
both methods of calculation

I agree with Bob Buglio.I have never known him to be wrong.

And this proves that you Bill, have never been
wrong. E mail me at
bbuglio@aol.com and tell me
what youve been up to and how things are going