Blood Smear Estimates
I shudder when I remember that, not terribly long ago, patients had appendectomies based on a manual differential. I remember a surgeon thinking aloud, “Only nine bands? We’ll hold off.” Hopefully, other variables were considered.
A 100-cell differential is notoriously inaccurate, never mind what a “band” is. A range can be estimated:
2 SD = 2 x SQRT((# cells x (100 - # cells)) / total cells counted)
Thus, a nine-cell band count in a 100-cell diff has a range of 3 to 15. Any copy of Clinical Diagnosis and Management by Laboratory Methods will have a table from 1960 based on a study of true variation for counts up to fifty percent. In this table, our band count has a range of 4 - 17. A savvy surgeon waiting for the band count to go over eleven percent might have taken the patient to the OR anyway.
It’s one reason some labs don’t report bands. But it also raises questions of accuracy. Other variables (hematocrit of the sample, wedge slide feathered edge, staining technique, humidity, human factors, etc.) suggest that these manual counts vary beyond statistics. Election polls are reported with a margin of error but not manual differentials. Hm.
Which brings me to white cell estimates. Since these do not simply count ratios but an absolute number, they can be significantly inaccurate. (Should we run a control with an estimate? I wonder.) Here’s our laboratory formula from a text on a shelf:
WBC Estimate = # cells in 10 fields (50x) where RBCs “slightly overlap” / 10 x 3
The College of American Pathologists recommends establishing an “estimation factor” comparing the number counted per field to an automated count, essentially calibrating a field of view (FOV). This still leaves a minor problem of standardizing the RBC field, which can be solved by counting RBCs on a radius, e.g. count 8 cells in a straight line from center to edge to estimate about 200 cells in the FOV.
So, I’m curious. How does your lab estimate WBC counts from peripheral smears?