Haematology
How to do a manual WBC differential count
A manual differential is counted from a stained blood film under oil immersion, moving in a battlement pattern through the monolayer and classifying each white cell until a set number is reached. One hundred cells is the routine count. The percentages are then reported, and may be converted to absolute counts.
What the count is, and when it is done
A differential white cell count reports what proportion of the white cells present belong to each class — neutrophils, lymphocytes, monocytes, eosinophils, basophils, and any immature or abnormal forms seen. Modern analysers produce a differential automatically on most samples. The manual count exists for the cases where the analyser flags something, cannot classify what it is seeing, or where morphology itself is the question.
That is worth stating plainly, because it frames everything below: the manual differential is the method used precisely when the answer matters and the machine would not do.
Where on the film to count
Not anywhere. A blood film is not uniform, and where you count changes what you count.
The count is made in the monolayer — the region where red cells lie in a single layer, touching but not overlapping. Too near the head of the film the cells are piled up and unreadable. Too near the tail and the distribution is distorted: larger cells, particularly monocytes and neutrophils, are dragged toward the edges and the feathered end, while lymphocytes are relatively over-represented in the middle. Counting in the wrong region does not add random noise; it adds a bias in a predictable direction, which is worse.
The battlement pattern, and why
Within the monolayer, fields are not examined at random or in a straight line. The standard is the battlement pattern: count two or three fields parallel to the edge of the film, then move several fields inward at right angles, then back again, tracing a crenellated path like the top of a castle wall.
The reason is the bias above. A straight track along the film samples one band of it and inherits whatever distortion that band carries. The battlement deliberately crosses between the edge and the centre repeatedly, so that over-represented and under-represented regions are sampled in turn and their biases partly cancel. The pattern is not ritual, it is a sampling correction.
How many cells: 50, 100, 200
One hundred cells is the routine count, and it is chosen because the arithmetic is trivial: with a denominator of 100, the count of each class is already its percentage. The number is a convenience as much as a standard.
| Cells counted | Used when | Note |
|---|---|---|
| 50 | The white count is very low and 100 cells cannot be found | Reported as a 50-cell count; the imprecision is larger and should be known |
| 100 | Routine | The count of each class is its percentage |
| 200 | Markedly raised white count, or rare cells to be characterised | Results halved to give percentages |
| 400 (2 × 200) | Reference method | CLSI H20-A2: 200 cells by each of two examiners |
The classical guidance is to begin with a hundred and extend to two hundred when circumstances require it — a rare population to pin down, or a raised white count where a small shift matters. The reference standard of two examiners counting two hundred cells each is rarely run outside method comparison work, for obvious reasons of workload.
The imprecision built into the method
This is the part that is easy to work with for years without internalising, and it is not a matter of care or skill. Classifying a sample of cells is a sampling problem, and sampling has irreducible error.
The counts follow a Poisson distribution, and the coefficient of variation is approximately 1 ÷ √n. At a hundred cells that is 10%, before anyone makes a single mistake. Not 10 percentage points — 10% of the figure — but the effect on an individual class is easier to see written out.
| Cells counted | CV | A result of 6% is consistent with | A result of 60% is consistent with |
|---|---|---|---|
| 50 | 14.1% | 0% to 12.6% | 46.4% to 73.6% |
| 100 | 10.0% | 1.3% to 10.7% | 50.4% to 69.6% |
| 200 | 7.1% | 2.7% to 9.3% | 53.2% to 66.8% |
| 400 | 5.0% | 3.7% to 8.3% | 55.2% to 64.8% |
Read the hundred-cell row. A monocyte figure reported as 6% is consistent with anything from about 1% to about 11%, purely from having counted a hundred cells rather than all of them. That is the method behaving correctly.
Two consequences follow, and both are practical:
- Small differences between two differentials are usually not differences. A move from 6% to 9% on successive hundred-cell counts is comfortably inside what sampling alone produces.
- Doubling the count does not halve the error. Because the error goes with √n, going from 100 to 400 cells — four times the work — halves it. That is the real reason the reference method is expensive and routine practice is not.
None of this is an argument against the manual differential. It is an argument for knowing what a percentage from it does and does not settle.
Keeping the tally without losing the field
The mechanical problem is small but real: your eyes are at the eyepieces and your hand is on the stage, so anything that requires looking away costs you your place in the field.
That is what the traditional multi-key laboratory tally counter solved — one key per class, found by touch, with a bell at a hundred. The same shape works on a keyboard: a key per class, a count that stops on its own at the target, and percentages calculated for you so no arithmetic happens at the end. What matters is that recording a cell takes no attention, because the attention belongs at the eyepiece.
What no counter does is the identification. Classifying the cell is the skilled part and the whole point; the counter holds the tally and nothing more. It carries no reference ranges, makes no comment on whether a result is expected, and is not a diagnostic aid. Reference intervals are the property of the reporting laboratory, and interpretation belongs to the clinician and the laboratory professionals responsible for the sample.
Frequently asked questions
How do you do a manual differential count?
Examine a stained film under oil immersion in the monolayer, move in a battlement pattern, and classify each white cell you meet until you reach the target count, usually 100. The counts are then reported as percentages, and may be converted to absolute counts using the total white cell count.
What is the battlement pattern and why is it used?
A crenellated path through the film: a few fields along the edge, several inward at right angles, then back out again. Cells are not evenly distributed on a smear, so a straight track inherits that bias. Crossing repeatedly between edge and centre lets the biases partly cancel.
How many cells should be counted in a differential?
One hundred is routine, because each count is then already a percentage. Fifty is used when the white count is too low to find a hundred cells, and two hundred when the white count is markedly raised or a rare population needs characterising. CLSI H20-A2 sets the reference method at 200 cells by each of two examiners.
Why count in the monolayer rather than anywhere on the film?
Because the film is not uniform. Near the head the cells overlap and cannot be read; near the tail and the edges, larger cells such as monocytes and neutrophils are over-represented and lymphocytes relatively under-represented. Counting outside the monolayer introduces a bias in a known direction rather than random noise.
How accurate is a 100-cell differential?
Less than most people expect. The coefficient of variation is roughly one over the square root of the number counted, so about 10% at a hundred cells before any human error. A result of 6% is consistent with roughly 1% to 11% on sampling alone.
Is a 200-cell count twice as accurate as a 100-cell count?
No. The error falls with the square root of the number counted, so doubling the count reduces the imprecision by about 29%, not by half. Quadrupling it to 400 cells is what halves it, which is why the reference method is rarely used routinely.
Does an online differential counter interpret the result?
No, and it should not. A counter holds the tally, stops at the target and works out the percentages. Identification of each cell is the operator's, reference intervals belong to the reporting laboratory, and interpretation belongs to the clinician and the laboratory professionals responsible for the sample.
Why do the automated and manual differentials disagree?
Partly because they measure differently, and partly because a manual count of a hundred cells carries around 10% imprecision by its nature, while an analyser classifies many thousands. A small disagreement is often within what sampling alone would produce.