Abstract
Intraobserver and interobserver reproducibility of the FAB classification was assessed for two independent observers whose decisions are acted on for treatment of patients with acute leukemia in the Hamilton region. Intraobserver reproducibility was assessed for Wright-stained preparations that were examined independently on two consecutive occasions at least 2 weeks apart. A third reading was performed with Wright stain and cytochemical data, and the fourth reading was done with addition of immunophenotype data. Concordance was calculated using a statistic that corrects for chance-expected agreement (k), and a weighted statistic that takes into account the seriousness of disagreements was used. Samples were available for morphological and cytochemical assessment on 105 patients, and immunophenotype data were available on 93 specimens. Intraobserver concordance was 64.8% and 70.5% for observers A and B, respectively, with kappa values of .56 and .62. There were 37 discordant readings for observer A and 31 for observer B, with each observer discordant between lymphocytic:nonlymphocytic phenotypes in ten cases. Concordance between observers was 63% (k = .54) and 72% (k = .65) for each of two separate readings for Wright-stained preparations only. Reproducibility improved to 89% (k = .86) when cytochemistry was added. When immunophenotype information was provided in addition to Wright-stained and cytochemical preparations, the agreement was 99%. Lymphocytic:nonlymphocytic discordance between observers occurred on nine occasions when Wright- stained preparations only were available and four times when cytochemistry was added; it did not occur with immunophenotyping. The study suggests that immunophenotyping, when added to morphological assessment of acute leukemia, may contribute substantially to agreement between observers.
