Poor specificity of urinary cryptococcal antigen testing (Reply to Drain et al. Prevalence of cryptococcal antigenuria at initial HIV diagnosis in KwaZulu-Natal)

We read with interest the recent article by Drain et al. on screening for cryptococcal antigenuria in patients newly diagnosed with HIV infection[1]. As the authors state, HIV-associated cryptococcal meningitis exacts a staggering toll in sub-Saharan Africa, being the most common laboratory-confirmed etiology of meningitis in our previous cohort from Cape Town[2]. Screening for cryptococcal antigenemia (CRAG) in patients with depressed CD4⁺ T-cell counts and treating early asymptomatic infection with high-dose fluconazole may prevent development of meningitis. The World Health Organization (WHO) conditionally recommends this approach and the recently published REMSTART study provides the first randomized trial evidence for a mortality reduction with this screen-and-treat strategy[3]. Screening for cryptococcosis through urine sample collection would be ideal as a non-invasive approach easily implemented in resource-limited ambulatory settings without need for phlebotomy supplies. As the authors assert, urine CRAG screening has good sensitivity compared to serum antigen testing using latex agglutination or lateral flow assay (LFA) methods.

However, we have significant concerns about endorsing urine screening for cryptococcosis given poor LFA test specificity when used in point-of-care settings on fresh urine, which could result in a significant number of false positives, potential toxicities of unwarranted high-dose fluconazole, including teratogenicity, and drug-drug interactions with antituberulous and other medications[4]. Without confirmatory testing in blood, it is unknown whether false positive results contributed to the higher antigen-positive prevalence reported by Drain et al. in the group with CD4 cell counts of 100-200 cells/μL (5.5%) compared to most studies using plasma or serum samples[5-7]. In Tanzanian patients with CD4⁺ T-cell counts ≤ 200 cells/μL initiating antiretroviral therapy, using serum LFA CRAG screening as a reference, Magambo et al. showed fresh urine testing had a sensitivity of 100% but specificity of 73.8% with positive predictive value (PPV) of only 22.7% for cryptococcal infection. With modified LFA diluent, the sensitivity decreased to 90% and specificity remained sub-optimal at 91.5% with PPV of 42% [4]. We have noted similar results in Cape Town, where 50% of positive urine CRAG results were false positives using serum LFA as a reference standard (Longley N et al., 2015, submitted). Interestingly, freezing then thawing urine (as in the Drain et al. study) mitigated but did not eliminate the issue of false positives, and is not practical or conducive to point-of-care testing.

Unfortunately, another convenient method of screening for early cryptococcal infection by sampling saliva showed a disappointing sensitivity of only 27% in asymptomatic patients compared to a reference standard of serum/plasma LFA[8]. However, fingerstick (finger prick capillary blood) LFA CRAG testing performs well compared to CSF or serum/plasma LFA in patients with cryptococcal meningitis[9], and preliminary results from our ongoing studies suggest high agreement between fingerstick and serum testing in the context of screening. As such, serum, plasma, or possibly fingerstick LFA remain the best methods of screening for early cryptococcal infection to prevent meningitis in HIV-infected patient populations; and, as recommended by the WHO, CRAG screening is likely to be of most benefit in patients with CD4⁺ T-cell counts below 100 cells/μL.