LAUSR

Background PCR is more sensitive than immunofluorescence assay (IFA) for detection of Pneumocystis jirovecii. However, PCR cannot always distinguish infection from colonization. This study aimed to compare the performance of real-time PCR and IFA for diagnosis of P. jirovecii pneumonia (PJP) in a real-world clinical setting. Methods A retrospective cohort study was conducted at a 1,300-bed hospital between April 2017 and December 2018. Patients whose respiratory sample (bronchoalveolar lavage or sputum) were tested by both Pneumocystis PCR and IFA were included. Diagnosis of PJP was classified based on multicomponent criteria. Sensitivity, specificity, 95% confidence intervals (CI), and Cohen's kappa coefficient were calculated. Results There were 222 eligible patients. The sensitivity and specificity of PCR was 91.9% (95% CI, 84.0%–96.7%) and 89.7% (95% CI, 83.3%–94.3%), respectively. The sensitivity and specificity of IFA was 7.0% (95% CI, 2.6%–14.6%) and 99.2% (95% CI, 95.6%–100.0%), respectively. The percent agreement between PCR and IFA was 56.7% (Cohen's kappa -0.02). Among discordant PCR-positive and IFA-negative samples, 78% were collected after PJP treatment. Clinical management would have changed in 14% of patients using diagnostic information, mainly based on PCR results. Conclusions PCR is highly sensitive compared with IFA for detection of PJP. Combining clinical, and radiological features with PCR is useful for diagnosis of PJP, particularly when respiratory specimens cannot be promptly collected before initiation of PJP treatment.