In Reply

In Reply.—Dr Breddin states “the presently available tests [for heparin-induced thrombocytopenia] have no or very little diagnostic value.” I disagree. Heparin-induced thrombocytopenia (HIT) antibody tests are among the most diagnostically useful assays in immunohematology.1,2 

Diagnostic usefulness of an assay can be represented by plotting sensitivity-specificity tradeoffs at various cutoffs between negative and positive assay results (operating characteristics). We have studied performance characteristics of the 2 major classes of assays, a platelet activation assay (platelet serotonin-release assay) and a PF4-dependent antigen assay, in 2 patient populations at high risk of HIT (postoperative orthopedic and cardiac surgery patients).1 The Figure shows the operating characteristics we found.

For comparison, curve A is shown (Figure), which represents a “useless” assay without diagnostic information. Indeed, such poor operating characteristics have been reported for certain assays of platelet-associated immunoglobulin G to diagnose immune thrombocytopenia.3 Dr Breddin's comments imply that this ought to be the profile of HIT antibody tests. However, curves B and C show the operating characteristics of the activation and antigen assays, respectively, for diagnosis of HIT in post–orthopedic surgery patients. Curves D and E show the corresponding curves in post–cardiac surgery patients.

The extent to which a given test result alters the physician's estimate of the pretest probability of HIT is known as the likelihood ratio, which is defined as Sensitivity/(1 − Specificity).4 Thus, for a post–cardiac surgery patient with a strong positive platelet-activation result (eg, 90% serotonin release), the likelihood ratio is about 20 (see the open circle on curve D, indicating 0.70/[1 − 0.965] = 20). Thus, if the physician had estimated a pretest probability of 50% (odds of 0.5:0.5), then this test result increases the posttest probability to more than 95% (0.5:0.5 × 20:1 = 20:1, or 95.2%). In contrast, the high sensitivity of this assay to detect clinically significant HIT antibodies (>95%) means that a negative test result lowers the posttest probability to less than 5%.

The diagnostic impact of such a strong positive result is even greater in post–orthopedic surgery patients, for whom the corresponding likelihood ratio is 90. As before, a negative test result essentially rules out HIT.

Although the antigen assay detects more clinically insignificant antibodies than the activation assay, it remains diagnostically useful. Likelihood ratios for a strong positive test (eg, optical density of 1.5) range from 10 to 40 for post–cardiac and post–orthopedic surgery patients, respectively. Also, its high sensitivity (>95%) means that a negative test generally rules out HIT.

Finally, Dr Breddin's inferences regarding test specificity (2%–30%) are incorrect. As clinically insignificant HIT antibodies are detected by antigen assay in 10% to 50% of post–orthopedic and post–cardiac surgery patients, respectively, the corresponding specificities range from 50% to 90%.1 (Specificity is defined as the proportion of individuals without HIT who have a negative test.) The corresponding specificity values for the activation assay range from 85% to 95%.1 If one considers that most patients with clinical HIT have quantitative test results far higher than the conventional cutoff between negative and positive,2 the practical specificity of these assays—as expressed by likelihood ratios—is even greater.