The CUE debate (continued): on surrogate tests and surrogate endpoints

To the Editor: Soloway’s letter in a previous issue of TRANSFUSION’ regarding confidential unit exclusion (CUE) argued that the performance of CUE should be judged by the same methodology and criteria as any other medical test, that is, the predictive value theory. We agree completely. We also agree that CUE can be considered a surrogate test that has the goal of detecting anti-human immunodeficiency virus 0 n e g a t i v e donors who harbor window-phase or silent infections but are compelled to misrepresent their risk activity during the overt donation process. Because of the difficulty in detecting such anti-HIV-negative HIV carriers and to evaluate the actual predictive value of CUE, proponents of CUE have relied on comparisons of relative anti-HIV prevalence rates among CUE and non-CUE donors to evaluate performance of the “test.”2 Thus, we are not only dealing with a surrogate test (CUE), we are employing a surrogate endpoint (anti-HIV positivity) to evaluate it. Therein lies the rub! We disagree strongly with Soloway’s statement, “To the extent that CUE’S positive predictive value [i.e., anti-HIV prevalence among CUE donors] exceeds HIV prevalence in the test population, CUE is mathematically superior to no test at all.” An analogy with the issue of HIV antigen screening illustrates the problem. Two major studies failed to fiid a single instance of HIV antigenemia among more than 1 million anti-HIV-negative donations and thereby concluded that the test would not identify antibody-negative HIV carriers at a rate warranting implementati~n.~.~ If the results of these studies had been evaluated with predictive value theory employing antiHIV positivity as a surrogate endpoint for performance, we might have concluded that the HIV antigen assay is indicated, because 11 percent of anti-HIV-positive donations were found to be HIV antigen positive (11% sensitivity), and 100 percent of donations confirmed as HIV antigen positive were anti-HIV positive (100% specificity, and, therefore, 100% positive predictive Indeed, by this analysis, the HIV antigen test appears to be a much better test than CUE, which has a sensitivity of only 6 percent and a specificity of only 0.1 to 1.6 percent.u Thus, the use of a surrogate endpoint can greatly exaggerate a test’s apparent usefulness, particularly if the surrogate endpoint (in this case, anti-HIV positivity) is much more common than the actual endpoint (silent HIV infections), and if the sensitivity of the test differs for the surrogate and actual endpoints. By analogy with surrogate tests for non-A,non-B hepatitis, the fact that a significant proportion of donors currently identified as anti-hepatitis C virus (anti-HCV) positive have elevated alanine amhotransferase (ALT) lmls and/or the presence of antibody to hepatitis B core antigen (anti-HBc) serves to substantiate earlier decisions to use these tests as surrogates before a specific test was available. However, these data have no direct relevance to the argument for continuation of ALT or anti-HBc screening now that the test for antibody to HCV has been implemented. The justification for continued donor ALT screening is evidence showing ALT elevation for months preceding anti-HCV stroconversion, as well as a demonstration of HCV RNA in a subset of donor samples with an elevated ALT level but a negative anti-HCV test. The argument for continuation of anti-HBc screening relates to hepatitis B virus (HBV), not HCV, and is based on cases of HBV transmission by hepatitis B surface antigen (HBsAg)-negative, antiHBc-positive units. In other words, once a test is available for specific detection of an agent, a surrogate test must stand on its own merit. If retention of CUE is warranted for political or legal reasons, so be it. But let us put an end to arguing its utility on the basis of anti-HIV prevalence. Even a comparison of the relative utility of alternative CUE mechanisms, as performed by Loiacono et al.,z is muddled by this approach. A blood center in a low-prevalence region with effective community education that dissuades seropositive persons from donating might have only two anti-HIV-positive donors a year, neither of whom used the center’s type of CUE (e.g., a callback system). Using anti-HIV-positive CUE donors as the endpoint, this Ctntcfs CUE apptars ineffeaive relative to the ballot method used by a center in a high HIV prevalence region a center that is viewed as an alternative testing site, with 37 percent of seropositive donors using CUE.I In our view, the risk of HIV infection from a CUE-negative unit is much greater in the latter center. Fine-tuning CUE to identify anti-HIV-positive donors is a self-fulfilling fallacy with no relevance to the safety of the blood supply.