Whether a detected virus or bacteria is a pathogen that may require treatment, or is merely a commensal ‘passenger’, remains confusing for many infections. This confusion is likely to increase with the wider use of multi-pathogen PCR.
To propose a new statistical procedure to analyse and present data from case-control studies clarifying the probability of causality.
We conducted a case-control study in US outpatient settings that enrolled patients aged 18 to 75 years with acute lower respiratory tract infection and controls without respiratory symptoms. Patients underwent multi-pathogen PCR testing. The positive etiologic predictive value was calculated to estimate the probability that each potential pathogen was the cause of symptoms. The outcome was illustrated using a modified forest plot and by classifying pathogens into five categories clarifying the probability for causality.
We enrolled 618 adult cases and 497 asymptomatic controls. The modified forest plot and the classification of risk for causality aimed to facilitate understanding. Pathogens likely to be causative when present included influenza A and B, SARS-CoV-2, rhinovirus, and parainfluenza viruses, while is almost always commensal. Broad confidence intervals for the positive etiologic predictive value made it difficult to draw conclusions for potential pathogens with low prevalence.
This pilot study shows that the proposed statistical approach is likely to be practical for analysing larger case-control studies or for a meta-analysis of multiple studies. This method may help when interpreting the results from multi-pathogen PCR.