PCR and culture for diagnosis of keratitis.

Background/aims: keratitis (AK) is a rare but sight-threatening infection. Molecular diagnosis of corneal scraping has improved the diagnosis of AK. Different molecular targets and conditions have been used in diagnosis thus far. In this study, we prospectively compared the performance of five PCR assays on corneal samples for the diagnosis of AK.

Methods: 1217 corneal scraping samples were obtained from patients, for whom an AK was suspected. Sample processing involved both molecular diagnostics and culture. PCR assays detected different regions of the nuclear small-subunit rRNA gene: three final point PCR assays using Nelson, ACARNA and JDP1-JDP2 pairs of primers, and two real-time PCR assays using Acant primer-probe. Human DNA and internal control were co-amplified in the real-time PCR assay to ensure scraping quality and the absence of inhibitors. In the absence of a gold standard, the performance of each test was evaluated using latent class analysis. Genotypes of isolates were also characterised.

Results: Estimated prevalence of AK was 1.32%. The sensitivity of diagnostic PCRs (73.3% to 86.7%) did not differ significantly from that of culture (66.7%), or according to the target sequence or the technology. Sensitivity could be increased to 93.8% or 100% by combining two or three assays, respectively. PCR specificity (99.3% to 100%) differed between the assays. T4 was the predominant genotype (84.6%).

Conclusions: Culture and a single PCR assay could lead to misdiagnosing AK. A combination of different PCR assays and improved sample quality could increase diagnosis sensitivity.