Manual reticulocyte count: the Miller disc method, the calculation and the corrected count

The Miller disc formula

Reticulocyte % = reticulocytes in the large square × 100 ÷ (red cells in the small square × 9). A Miller disc is a glass reticle that sits in the microscope eyepiece and shows two squares, one nine times the area of the other. Every red cell in the small square is counted; only the reticulocytes are counted in the large one. Nine times the small-square count then stands for all the red cells in the large square, so the large square's reticulocytes can be expressed as a percentage without classifying every red cell.

Lim and colleagues, in the American Journal of Hematology (2016), write the general form: target cells in the large square ÷ all cells in the small square × a reticle factor, which is 100 divided by the area ratio. For a 1:9 reticle that factor is 100 ÷ 9 = 11.1, which is the same as the × 100 ÷ 9 in the formula above.

The method's advantage is time, and the evidence favours it. McKenzie and Gauger sent 50 specimens to six hospital laboratories and compared Miller disc counts, other ruled reticles and counts with no reticle against an automated instrument; they found the Miller disc "the most precise and accurate manual method", and counts made without one significantly higher (Clinical Laboratory Science, 1991).

Why 111 cells stand for 1,000

111 small-square red cells × 9 = 999, so stopping there is the Miller-disc equivalent of counting 1,000 red cells. The traditional manual count classifies 1,000 consecutive red cells and divides the reticulocytes found by 10 to get a percentage. With the disc you classify only the 111 cells in the small square and scan the rest of the large square for reticulocytes alone, which is the time saved.

Because 999 is not quite 1,000, the shortcut "reticulocytes ÷ 10" is very slightly low: 20 ÷ 10 = 2.00%, while 20 × 100 ÷ 999 = 2.002%. Both round to 2.0%, and a manual count cannot support a third figure anyway (see counting error). Where a procedure counts more, the arithmetic is the same: 222 small-square cells stand for 1,998 red cells, 333 for 2,997.

Count the reticulocytes inside the small square too

On a disc where the small square sits inside the large one, the ninefold area is the whole large square, small square included. A reticulocyte inside the small square is counted twice over in the procedure: once as a red cell in B, and once as a reticulocyte in A. Leaving it out of the reticulocyte count samples only eight ninths of the area and pulls the result about 11% low. The rule that settles it is simply that the area you count reticulocytes in must be the area your factor assumes.

Check your reticle's ratio before you trust the 9

The 9 is a property of the glass in your eyepiece, not of the method, and reticles differ. Lim and colleagues found laboratories in one research network using reticles of 1:4, 1:5 and 1:9, and one group using three different reticles while assuming they were the same. Their worked warning: reading a 1:5 reticle with a 1:4 factor overestimates by 25%. A reticle sold as a Miller disc by Microscope World describes its squares as related "in the proportion 1:10", and advises calibrating against a stage micrometer.

The check takes five minutes. Put a stage micrometer under the objective you count with, measure the side of each square, and square the ratio of the sides: sides of 0.30 mm and 0.10 mm in the specimen plane give (0.30 ÷ 0.10)² = 9. Lim and colleagues did the equivalent by photographing the field through the eyepiece and measuring the areas in ImageJ. Write the measured ratio into the procedure and use it in place of 9 if it differs.

The staining and counting routine

The count is made on a smear of blood stained supravitally, while the cells are alive, so the ribosomal RNA in young red cells precipitates as a blue network. New methylene blue was introduced as a reticulocyte stain by George Brecher in 1949, as Silpachai and colleagues' 1973 study of normal adults records, and it is the stain Hillman used in his 1969 study. The steps:

1. Stain. Mix EDTA blood with new methylene blue and leave it to stain before making the smears. The proportions and the time differ between laboratory procedures and stain makers, which is why pages on this subject disagree; follow your own procedure or the stain's instructions.
2. Make thin smears and let them dry. Hillman made coverslip smears, counterstained them and read them in pairs.
3. Choose the area. Count where the red cells lie in a single layer, not overlapping, the same monolayer used for a white cell differential.
4. Count with the disc under oil. In each field, count the red cells in the small square and the reticulocytes in the large square, then move to a fresh field.
5. Stop at 111, or at your procedure's multiple of it, and calculate.

Simple Online Counter is a free, browser-based set of counters that keeps its counts in your browser's own storage. In the multi counter, name one counter "red cells, square B" and one "reticulocytes, square A", and set a maximum of 111 on the first so it stops at the target. Undo takes back a double tap. The counter does not recognise reticulocytes; that judgement stays at the microscope.

What counts as a reticulocyte

A reticulocyte is a red cell without a nucleus that shows blue-stained reticulum, and the threshold for "shows" changes the count. Gaur and Sehgal, in a 2021 letter in the Pan African Medical Journal, give the International Council for Standardization in Haematology (ICSH) definition as a non-nucleated red cell with at least two blue-staining particles, or one particle linked to a filamentous thread.

Older work was looser. Silpachai, Pinthanond and Kulapongs counted any cell with "at least a single appropriately-stained inclusion" in their 1973 study of 100 Thai adults, and named the "lack of uniform criteria for identifying reticulocytes" as one reason published normal values varied so much. A one-dot rule counts the oldest reticulocytes that a two-dot rule leaves out, so it gives higher numbers from the same smear.

Not every blue dot is reticulum. Gaur and Sehgal list Pappenheimer bodies, Howell-Jolly bodies and Heinz bodies as inclusions that can be confused with mature reticulocytes.

Counting error, and how many cells to count

The error of a manual reticulocyte count is set by the number of reticulocytes counted, and at 1,000 red cells it is large. For a percentage, the WHO semen manual (5th edition, 2010, Box 2.5) gives the binomial standard error, √(p × (100 − p) ÷ N). For 2.0% in 1,000 cells that is 0.44 percentage points, 22% of the result. For small percentages it is almost exactly 1 ÷ √(reticulocytes counted):

The Miller disc adds one source of error of its own: the denominator rests on the 111 cells in the small square, which carry about 1 ÷ √111 = 9.5%. Combined with 20 reticulocytes, the error is √(1/20 + 1/111) = 24%, barely more than the 22% of a full 1,000-cell count, for far less work. That is the trade the disc makes. A hemocytometer count follows the same square-root rule.

The way to a tighter result is more cells. Hillman's study counted "at least 5000 consecutive red cells" per sample, by two observers (Journal of Clinical Investigation, 1969). In Lim and colleagues' study, two microscopists' 1:9 reticle counts on the same samples agreed only weakly (r = 0.39), one more reason a single 1,000-cell count of a low percentage is an estimate, not a measurement.

Worked example: two slides and the agreement check

A sample is counted on two smears with a 1:9 Miller disc, 111 small-square red cells on each.

1. Slide 1. 20 reticulocytes: 20 × 100 ÷ 999 = 2.0%.
2. Slide 2. 27 reticulocytes: 27 × 100 ÷ 999 = 2.7%.
3. Do they agree? With equal denominators the two reticulocyte counts can be compared the way the WHO manual (Box 2.10) compares replicate cell counts: the difference should be no more than 1.96 × √(sum). Here the sum is 47, so the limit is 1.96 × 6.9 = 13.4, and the difference is 7. They agree; a gap that large is what chance alone produces.
4. Pool them. 47 × 100 ÷ (222 × 9) = 4,700 ÷ 1,998 = 2.35%, reported as 2.4%.
5. Its error. √(1/47 + 1/222) = 16%, so about 1.6% to 3.1% at 95%.

Had the second slide shown 40 reticulocytes, the sum would be 60, the limit 1.96 × 7.7 = 15.2 and the difference 20: too far apart, which points to a smear or staining problem rather than chance, and both slides are recounted.

The corrected count, the RPI and the absolute count

The reticulocyte percentage rises in anaemia for two reasons that have nothing to do with more production: there are fewer red cells to divide by, and young cells are released early and stay reticulocytes longer. Two corrections deal with them, and both come from Hillman's 1969 study, in which healthy volunteers were made anaemic by phlebotomy and their marrow output was measured three ways.

The haematocrit correction. Hillman corrected each observed count for the haematocrit: observed % × observed haematocrit ÷ 45. Today this is usually called the corrected reticulocyte count. Hillman called it the "absolute reticulocyte count (%)", which is not what absolute reticulocyte count means now (see below).

The maturation correction, which gives the RPI. Hillman found the corrected count still tracked marrow output poorly, because reticulocytes took longer to lose their reticulum as anaemia deepened. He calculated the circulating maturation time:

Dividing the corrected count by the maturation time gives the reticulocyte production index (RPI). The single values in teaching tables are roundings of these ranges, and Hillman's volunteers were not bled below a haematocrit of about 25%, so any factor for a lower haematocrit is an extrapolation.

Worked through for a count of 4.0% at a haematocrit of 35%, with a red cell count of 3.9 × 10¹² per litre:

• Corrected count: 4.0 × 35 ÷ 45 = 3.1%.
• RPI: 3.11 ÷ 1.5 = 2.1, or 3.11 ÷ 2 = 1.6, depending on where in Hillman's 1.5 to 2 day range the factor is taken.
• Absolute reticulocyte count: 4.0% × 3.9 × 10¹² per litre = 156 × 10⁹ per litre, or 156,000 per µL. This is the modern meaning: reticulocytes per unit volume, which needs no haematocrit correction because it does not depend on dividing by the red cells.

The RPI range in this example straddles 2, the figure Gaur and Sehgal use: in anaemia, an index below 2 suggests an inadequate marrow response, and above 3 suggests red cell loss or destruction with a compensating marrow. A one-day-wide uncertainty in the maturation time moves the same patient across that line, which is reason to read the RPI as a rough guide, and why the interpretation belongs to the laboratory and the clinician.

Reference values belong to the laboratory

Published normal ranges differ with the method and the definition. Gaur and Sehgal give about 0.5 to 2.5% in adults and 2 to 6% in infants. Silpachai and colleagues, counting 100 healthy Thai adults with new methylene blue in 1973, found a mean of 0.37% in men with 95% limits of 0.1 to 0.8%, and higher values with brilliant cresyl blue, concluding that each method needs its own normal values. A manual count is compared with the range of the laboratory that made it, for the reasons the white cell ranges guide sets out.

Mistakes that change the answer

What this guide does not cover

This guide covers the arithmetic and the counting technique of a manual reticulocyte count for students and laboratory staff. It is educational, not medical advice: reference ranges, the correction factors in use and the interpretation of a result belong to the laboratory and the clinician. It does not cover automated reticulocyte counting, immature reticulocyte fractions or reticulocyte haemoglobin.

Two limits on the sources. CLSI's reticulocyte guideline, H44, is sold, and the ICSH documents were not available to read, so the ICSH definition is quoted as Gaur and Sehgal give it. The staining proportions and incubation time are left to your procedure because no procedure we could open and check states them.