[Skip to Content]
[Skip to Content Landing]
June 21, 1985

Screening for HTLV-III Antibodies: The Relation Between Prevalence and Positive Predictive Value and Its Social Consequences

Author Affiliations

Massachusetts General Hospital Boston

JAMA. 1985;253(23):3395-3396. doi:10.1001/jama.1985.03350470047007

To the Editor.—  We are concerned that in the recent article by Weiss et al,1 some of the test operating characteristics reported may have been misleading. Specifically, the authors conclude that their screening test for human T-cell leukemia (lymphotropic) retrovirus type III (HTLV-III) antibodies (as a marker for the ability of a unit of blood to transmit acquired immunodeficiency syndrome [AIDS]) was "highly sensitive and specific" based on values calculated after "excluding borderline results." Neglecting these "borderline" results is inappropriate, as test results in the overlap zone between distributions for disease and nondisease actually define a test's sensitivity and specificity.2Sensitivity-specificity pairs for different cutoffs can be determined from data in the article's Table 1 (combining all patients with AIDS to define "disease" and Burlington blood donors to define "nondisease"). However, it may be more informative in examining a screening test to scrutinize the ratio of false-positives to