Reproducible scoring of CFU-GM and BFU-E grown in collagen-based semisolid medium after a short (3 h) training

Colony counting remains an important source of variation in colony-forming unit-granulocyte-macrophage (CFU-GM) assays performed in methylcellulose or agar. We studied the reliability of colony scoring of CFU-GM assays carried out with collagen, a matrix that allows gel collection on glass slides and in situ cellular morphology. Fourteen slides were exchanged among laboratories, and two rounds of colony (CFU-GM and burst-forming units-erythrocyte [BFU-E]) counting were performed by 11 (first counting), then 8 (second counting) different laboratories, the majority of which had no previous experience of collagen gel cultures and reading. Two-way analysis of variance (ANOVA) of the first round of colony counting showed significant differences among centers in CFU-GM counts (p = 0.023) but not in BFU-E counts (p = 0.163). Coefficients of variation for the 14 slides ranged from 22% to 50% (median 28%) for CFU-GM counts and from 12% to 74% (median 23%) for BFU-E counts. After a 3 h session of collective colony reading attended by members of 8 laboratories, a second round of colony counting was performed. This time, ANOVA showed no significant difference among centers for CFU-GM (p = 0.533) and BFU-E (p = 0.328) counts, and coefficients of variation were significantly improved, with medians of 17% for CFU-GM counts and 20% for BFU-E counts. In addition, when data from the second round of readings were analyzed without the 2 centers counting consistently low (center 8) or consistently high (center 5), variance among centers was further improved for both CFU-GM (p = 0.798) and BFU-E (p = 0.619). In summary, this study shows for the first time that reproducible BFU-E and CFU-GM scoring can be achieved using collagen-based semisolid medium (now commercially available) as long as adequate training in colony identification is provided.