Discriminant Functions Constructed for Early Prediction of Visceral Mycosis

Mayumi Mochizuki, Hikaru Kume, Hirokazu Tsuruta, Fumiya Tomonaga, Masahiko Okudaira

Research output: Contribution to journalArticlepeer-review


To construct a discriminant function for clinical use in predicting fungal infection in compromised hosts, we analyzed the determinants of leukopenia grading and used the evolving equations to estimate likelihood of fungal infection in patients of leukemia and lymphoma. From autopsy-confirmed male and female cases of random age of leukemia with visceral mycosis (21) and without visceral mycosis (19), clinicopathological data related to host immunity were collected and multivariate analysis and discriminant analysis were conducted on the raw data. Four predictor variables emerging from the analysis were employed to derive the following linear equation for estimating the presence of fungal infection in leukemia patients: Z1= - 100.527 - 0.00401 × (X2) - 0.01057 × (X4) + 0.05622 × (X6) + 2.61331 × (W1) where Z1 is the discriminant score, X2 the peripheral neutrophil count, X4 the peripheral monocyte count, X6 the number of consecutive days indicating a leukocyte count below 1000μl, and W1 the highest body temperature of the compromised host. If the Z1 score is greater than zero, the patient is classified as having mycosis. In autopsy-confirmed cases involving mycosis (21) and not involving visceral mycosis (19), we tested the discriminant function with clinicopathological data derived retrospectively from the patient records. In each case the underlying disease had been leukemia. The Z1 score correctly classified the visceral mycotic cases with a 100% hit ratio, and 89.5% of the non-mycotic cases (17). Applying the discriminant function clinically, we hypothesized that in patients of leukemia a Z1 score continuing above zero (Z1>0) for more than seven consecutive days would indicate visceral mycosis. Correct assessment in a prospective study was made in 39 out of 44 living patients (88.6%). With the same four predictor variables as those related to leukemia, a discriminant function was constructed for the screening of visceral mycosis in autopsy-confirmed cases of malignant lymphoma: Z2 = - 6.3247 - 0.00037 × (X2) - 0.00883 × (X4) + 0.20138 × (X6) + 0.14510× (W1) where Z2 is the discriminant score and a Z2 score greater than zero classifies the patient as having mycosis. In 20 cases of malignant lymphoma (7 with visceral mycosis, 13 without mycosis), the Z2 score provided correct assessment in 16 (80%). From these results, the discriminant functions derived from computations based on grade of leukopenia are clinically useful in early assessment of the presence of fungal infection in patients of leukemia and of malignant lymphoma.

Original languageEnglish
Pages (from-to)265-273
Number of pages9
JournalJapanese Journal of Medical Mycology
Issue number3
Publication statusPublished - 1993 Jul


  • discriminant score
  • early prediction
  • leukemia
  • malignant lymphoma
  • risk factor
  • risk index
  • visceral mycosis

ASJC Scopus subject areas

  • Microbiology
  • Infectious Diseases


Dive into the research topics of 'Discriminant Functions Constructed for Early Prediction of Visceral Mycosis'. Together they form a unique fingerprint.

Cite this