Low Accuracy of Manual White Blood Cell Count in Amniotic Fluid
Elizabeth McMillen, Jessica Bautista, Anthony Sireci, Alexander Kratz, Brie Stotler. Columbia University, New York, NY
Background: Intraamniotic infection (IAI) is a common cause of preterm labor. Of the laboratory tests for IAI, amniotic fluid (AF) WBC count can be performed rapidly. An elevation of AF WBC over 50 cells/mm3 has been reported to be 36-64% sensitive and 72-95% specific for IAI. However, cell count on amniotic fluid is not routinely performed in the laboratory, nor is it offered at any of the major reference labs. At the request of clinicians, accuracy of manual AF WBC count was determined.
Design: Ten AF samples sent to the laboratory for routine AFP measurement were randomly selected for analysis. Two technologists performed cell counts on each sample using a hemocytometer. A total of 100 fields were counted. To assess accuracy of WBC counts at various concentrations, AF samples that had no WBCs were spiked with blood from CBC specimens with known WBC counts, producing samples with calculated WBC counts ranging from 11-195 WBC/µL. Cell counts were performed on the ten spiked samples, including a negative control. The WBC counts for all 20 specimens were then compared to the calculated expected value for WBC.
Results: No debris was reported. The initial WBC count on the non-spiked samples ranged from 0-4/mm³. Comparison between the technologists' manual WBC count on the spiked samples to the expected WBC count yielded R2 coefficients of 0.51 and 0.83. When technologist results were averaged together, the R2 coefficient between the manual count and the expected WBC count was 0.66. Percent agreement between the technologists was 80%, with an R2 coefficient of 0.83. The average CV was 21% with a range of 0-62% for each sample.
Conclusions: There was moderate correlation between the manual and the expected WBC in the spiked AF samples. The coefficient of determination from the averaged results was lower than usual acceptable criteria. Using a threshold of 50 cells/mm3 as the cutoff for infection, one sample would have erroneously been reported as less than 50, while the expected count was 54 cells/mm3. The accuracy of the WBC count may be increased by averaging the WBC counts reported by two technologists. The decision to perform amniotic fluid cell counts should involve a discussion with clinicians about the accuracy of the test and the potentially high CV associated with this test.
Category: Quality Assurance
Monday, March 19, 2012 1:00 PM
Poster Session II # 266, Monday Afternoon