Microbiological culture versus 16S/18S rRNA gene PCR-sanger sequencing for infectious keratitis: a three-arm, diagnostic cross-sectional study

Background: To compare the diagnostic performance of microbiological culture and 16S/18S rRNA gene polymerase chain reaction (PCR)-Sanger sequencing for infectious keratitis (IK) and to analyse the effect of clinical disease severity on test performance and inter-test concordance.

Methods: This was a three-arm, diagnostic cross-sectional study. We included all eligible patients who presented with presumed bacterial/fungal keratitis to the Queen's Medical Centre, Nottingham, UK, between June 2021 and September 2022. All patients underwent simultaneous culture (either direct or indirect culture, or both) and 16S (pan-bacterial)/18S (pan-fungal) ribosomal RNA (rRNA) PCR-Sanger sequencing. The bacterial/fungal genus and species identified on culture were confirmed using matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry. Relevant clinical data were also collected to analyze for any potential clinico-microbiological correlation. Main outcome measures included the diagnostic yield, test accuracy (including sensitivity and specificity), and inter-test agreement [including percent agreement and Cohen's kappa (k)].

Results: A total of 81 patients (86 episodes of IK) were included in this study. All organisms identified were of bacterial origin. Diagnostic yields were similar among direct culture (52.3%), indirect culture (50.8%), and PCR (43.1%; p = 0.13). The addition of PCR enabled a positive diagnostic yield in 3 (9.7%) direct culture-negative cases. Based on composite reference standard, direct culture had the highest sensitivity (87.5%; 95% CI, 72.4-95.3%), followed by indirect culture (85.4%; 95% CI, 71.6-93.5%) and PCR (73.5%; 95% CI, 59.0-84.6%), with 100% specificity noted in all tests. Pairwise comparisons showed substantial agreement among the three tests (percent agreement = 81.8-86.2%, Cohen's k = 0.67-0.72). Clinico-microbiological correlation demonstrated higher culture-PCR concordance in cases with greater infection severity.

Conclusions: This study highlights a similar diagnostic performance of direct culture, indirect culture and 16S rRNA PCR for bacterial keratitis, with substantial inter-test concordance. PCR serves as a useful diagnostic adjuvant to culture, particularly in culture-negative cases or those with lesser disease severity (where culture-PCR concordance is lower).