Reading an HPLC purity report, line by line

A purity report is one page, but it rests on a method with a dozen adjustable parameters. The numbers in the middle of the page mean little without the header above them, and the header is the part most often skipped. What follows is a reading order, from the top of the page down, with the question each section answers.

1. The method header

The header should name the column, its length and internal diameter, the particle size and the bonded phase, the mobile phase composition, the gradient table where a gradient was run, the flow rate, the detection wavelength, the column temperature and the total run time. Each of these moves a peak in time or changes its width. A report missing them can be filed but not interpreted, and asking for the missing part is a reasonable first request.

Where the header is incomplete, ask for the trace and the method file rather than another summary, because a trace shows what a table leaves out. A supplier who can send both usually holds the rest of the record as well.

2. The main peak and its integration

Start with the retention time of the main peak, then look at how that peak was defined: where the baseline was placed at each end, whether the peak was integrated as one event, and where the integration stopped. An area is the output of a decision, not a property of the material. The report is only useful if the decision is visible on it.

3. Area percentage and its denominator

Area percentage is a ratio, not an absolute: the area of one peak divided by the total area included in the calculation. The denominator is fixed by the reporting threshold, the area below which small peaks are not counted. Move the threshold and the percentage moves with it. Read the threshold first, then the percentage, and read the two as one statement.

Two reports can carry the same headline figure while describing different pictures, because one counted a set of small peaks and the other excluded them. When figures from two lots are being compared, compare the thresholds before comparing the numbers.

4. Retention time across the sequence

Compare the main peak retention time across the whole sequence, not only within a single chromatogram. A slow drift of a few hundredths of a minute over a long run of samples is ordinary. A step change between two consecutive chromatograms points instead at a freshly prepared mobile phase, a column temperature that has not settled, or too little time allowed for equilibration. The report shows that a cause exists; it does not name it.

Retention time also provides a quick identity check on the page. When the main peak arrives at the same time as in the reference chromatogram, and the column and mobile phase match, the two traces are at least describing the same method.

5. Baseline, noise and the settings behind the trace

Look at baseline noise, normally reported as signal-to-noise over a stated region, and at the parameters that shaped the trace: slope sensitivity, peak width, smoothing window. Heavy smoothing quietens the baseline and flatters small peaks at the same time. The settings belong next to the result so that a second analyst can reproduce the same integration rather than a similar-looking one.

Noise figures are only comparable with the region they were measured over. A signal-to-noise value taken from a clean stretch at the end of the run describes something different from one taken in the first two minutes, where the solvent front still shapes the trace.

• A shoulder on the main peak, or a peak failing the stated symmetry criterion
• Additional peaks near the solvent front, where the method cannot resolve them
• A baseline that climbs through the run instead of returning to its starting level
• A main peak area that departs from the trend of the surrounding chromatograms
• A total area that differs sharply from the previous lot under the same method

6. When a figure sits outside the criterion

A value outside the stated limit is information, not a verdict. The useful next step is to separate the possibilities: a real difference in the sample, a method that has drifted, or an integration choice made differently from the last time. Each of the three produces the same kind of number, and the report alone cannot tell them apart. What it can do is show whether the rest of the sequence behaved consistently, which is often enough to narrow the field.

Read in three passes, the page goes quickly: header for context, peak and threshold for meaning, trend across the sequence for confidence. A report read that way takes about five minutes and answers questions that a single percentage never will.