Serial Procalcitonin Levels and Bacterial Etiology in Hospitalized Patients with Community-Acquired Pneumonia

Pierre Ankomah Massachusetts General Hospital Suzanne M McCluskey Massachusetts General Hospital Michael Abers Massachusetts General Hospital Benjamin Bearnot Massachusetts General Hospital Phillip Schuetz University of Basel, Kantonsspital Aarau Victor Chiappa Massachusetts General Hospital Jatin M Vyas Massachusetts General Hospital Michael K. Mansour (  mkmansour@mgh.harvard.edu ) Massachusetts General Hospital


Background
During management of pneumonia, clinicians consider and investigate potential microbial etiology in order to inform treatment decisions. However, the diagnostic tools remain limited for determination of bacterial etiology of pneumonia (1). The most commonly used diagnostic aid in this endeavor is sputum culture, which is fraught with multiple challenges including specimen quality and di culties in culturing/isolating organisms (2). Recent advances in nucleic acid detection may offer additional sensitivity, but these methods do not address issues of true invasive disease compared to colonization (3). To differentiate between commensal and invasive pathogens, serum biomarkers such as C-reactive protein (CRP) and procalcitonin (PCT) that may re ect the activation state of the host immune system have been investigated as supplementary diagnostic and prognostic aids in the clinical decision making process (4). PCT has shown promise as a means of distinguishing between viral and bacterial causes of pneumonia, as well as, via serial assessments, assisting in decisions regarding antibiotic course (4,5). This strategy of PCT-guided antibiotic treatment has the potential to reduce antibiotic exposure, though data are con icting about how this translates into clinical practice (6)(7)(8).
While the majority of studies investigating the clinical utility of PCT have been in the context of differentiating between clinical syndromes that require antibiotics and those that do not, a few studies have also assessed the association between PCT levels and bacterial etiologies of pneumonia (9)(10)(11)(12)(13). However, the pertinent comparisons in these studies were made between aggregate groups, such as "classic bacterial" versus "atypical" causes of pneumonia (9,13), gram positive versus gram negative organisms (10), and pneumococcal or non-pneumococcal pneumonia (11,12). Accordingly, though two studies (11,12) did report higher PCT levels for pneumococcal compared to non-pneumococcal pneumonia, no granular conclusions were reached regarding PCT values associated with speci c bacteria. Another crucial limitation of these studies is that they were generally limited to community-acquired pneumonia (CAP) due to relatively drug-susceptible organisms, with only a small proportion of cases being due to bacteria such as methicillin-resistant Staphylococcus aureus (MRSA) or Pseudomonas aeruginosa. In addition, with the exception of the study done by Tamura et al., (11), all the other studies assessed PCT values only during initial evaluation, and Tamura et al used the serial data to distinguish between disease severity and mortality risk in pneumococcal versus non-pneumococcal pneumonia.
Here, using data obtained during a prospective observational study to evaluate the utility of procalcitonin in clinical prognostication for patients hospitalized with pneumonia (14), we investigate the relationship between PCT measurements and bacterial etiology of pneumonia. Importantly, the population we examine comprises a large proportion of patients who had been hospitalized in an acute-care hospital within the preceding 90 days, or had received outpatient treatment, including intravenous therapy within the preceding 30 days (14) and thus had putatively elevated risk factors for pneumonia caused by multi-drug resistant organisms (MDRO) (15). We examine initial PCT values associated with different bacterial pathogens on admission, and assess serial daily PCT values to explore differences in trajectory of PCT during treatment/resolution of infections.

Methods
Study Design: We analyzed data from a prospective cohort study of adult patients admitted for at least one night to a tertiary care center in Boston, Massachusetts between March to September 2013, with symptoms and imaging ndings concerning for pneumonia (14). Chart review was performed by two independent Internal Medicine physicians to adjudicate the clinical diagnosis of pneumonia, with any discordance reviewed by a committee for a nal decision (14). A notable exclusion criterion was for patients with prior hospitalization within the preceding 14 days to eliminate potential complicating impact of lingering radiographic ndings from prior pneumonia or effect of previous antibiotic treatment on PCT levels. In addition, patients with cardiogenic shock, trauma, burns and ST-elevation myocardial infarction were excluded from the study as these conditions non-speci cally elevated PCT independent of bacterial infection.
Microbiological Diagnostics: Blood, sputum, pleural uid, bronchoalveolar lavage (BAL) cultures, respiratory viral PCRs and urine antigens for Streptococcus pneumoniae and Legionella pneumophila were obtained at the discretion of the primary clinical team. Organisms known to cause respiratory infections were classi ed as the causative agent(s) for pneumonia if at least one of the following criteria were met: (i) concurrent bacteremia, (ii) positive urine antigen test, (iii) organism isolated from appropriate sputum or endotracheal specimens (with <10 epithelial cells/hpf and >25 WBC/hpf), in at least "moderate quantity" (16) and/or (iv) organism isolated from pleural uid or BAL samples. Infections were classi ed as having mixed etiology if multiple pathogens met the above criteria.
Statistical Analysis: For comparisons of patient demographic and clinical characteristics, the p value for age was calculated using an ANOVA test, and p values for categorical variables were calculated from Fisher's exact test analyses. Comparisons of the relative proportions of different groups that met a composite severity endpoint of bacteremia, ICU admission or death were undertaken using Fisher's exact test analyses. Continuous data were summarized using median and interquartile ranges. PCT values of <0.05 ng/mL were coded as 0.05 ng/mL to provide non-zero values to facilitate statistical comparisons. A two-tailed Mann-Whitney-U test was utilized for two-group comparisons for continuous data, with p values less than 0.05 considered to indicate statistical signi cance. Analyses were undertaken using R version 4.0.2 (The R project for Statistical Computing).

Results
A clinical diagnosis of pneumonia was adjudicated in 322 out of 505 participants in the cohort study following blinded chart review. Bacterial etiology was determined in 64 (19.8%) of those cases, and 5 (1.6%) patients had viral pneumonias. Given the goal of this study, subsequent analyses focused solely on bacterial etiologies of pneumonia. Table 1 shows the bacteria identi ed, and for each, the number of patients who met a composite severity endpoint of bacteremia, admission to an Intensive Care Unit (ICU), or death.  Table S1). While there was no statistically signi cant difference between the rates at which pneumonia due to particular bacterial etiologies met the composite severity endpoint (p=0.79), patients who had speci c bacterial etiologies identi ed as the causative agent of their pneumonia were more likely to meet the severity endpoint than those who did not (p=3.5e-08)Subsequent analyses focused on comparing PCT levels for S. pneumoniae, H. in uenzae, MSSA, MRSA, P. aeruginosa, the Enterobacteriaceae (E. coli and S. marcescens) and mixed infections. This approach was chosen for two reasons: (i) to maximize comparisons between the largest groups, and (ii) to provide comparisons between microbial etiologies that prompt consideration for adjustment in therapy from typical CAP regimens. Demographic information for patients belonging to these groups is shown in Table 2. In this study, we compared serial PCT measurements in a population of hospitalized patients with community-acquired pneumonia caused by different bacteria. We observed higher values of PCT at the time of admission in pneumococcal infections compared to infections due to S. aureus (including MRSA), P. aeruginosa, H. in uenzae and to those involving multiple bacteria. By measuring PCT levels through the rst four days of hospitalization, we noted that differences in PCT values were most pronounced on day 2 of the hospitalization, with limited discriminatory utility from data collected at later time points.
Previous studies in patients with community-acquired pneumonia similarly demonstrated higher levels of PCT on admission for S. pneumoniae compared to non-pneumococcal causes (11,12). Our study examined the non-pneumococcal bacteria individually rather than in aggregate, and also included bacteria such as MRSA and P. aeruginosa that are more commonly associated with nosocomial pneumonias, yet we still observed similar results. Tamura et al. speculate that this effect is potentially because S. pneumoniae causes more systemic in ammation than the other etiologies of pneumonia, although further investigation is required to explore this hypothesis (11). Another notable result in our study is that infections with multiple bacterial pathogens identi ed in individual patients had the lowest PCT values. In fact, median PCT values associated with these mixed infections over the four days of our analysis were consistently less than 0.1 ng/mL, a cutoff value which in isolation would suggest a syndrome unlikely to represent bacterial pneumonia (17). The reason for this result is unclear and warrants further investigation. It does, however, prompt consideration that the in ammatory response to concurrent infection due to different bacteria, as measured by PCT levels, may not be synergistic.
In most clinical practice, admission values of PCT are the sole values used to assist clinical decision-making. It is worth noting, however, that serial PCT evaluations have been utilized in some settings to guide de-escalation of antibiotic therapy based on the expectation that PCT values fall by approximately 50% per day during resolution of infection (8). In our study, for MSSA, S. pneumoniae, and prominently for P. aeruginosa (for which the median day 1 PCT value was less than the aforementioned cutoff value of 0.1 ng/mL), there was a notable increase in PCT values from day 1 to day 2 of hospitalization. In addition, differences between PCT values associated with different bacteria were altogether greatest on day 2 of hospitalization. This prompts caution regarding the practice of making decisions about antimicrobial therapy solely on the basis of the admission PCT level. Our results suggest that serial measurement of PCT to include to include day 2 of hospitalization may be more appropriate to gain discriminatory information between different pathogens. We speculate that about 48 hours into a hospitalization may represent a period where in ammatory sequelae of infection are peaking, prior to antibiotic therapy reducing the infectious and in ammatory burden on the host, but this needs further experimental investigation.
Limitations of this study include the small sample number of cases attributable to speci c bacterial pathogens and thus a lack of statistical power to allow the derivation of PCT thresholds that discriminate between pathogens. In addition, given that microbial diagnostic samples were obtained at the discretion of the clinical team, there was a lack of testing standardization for all patients, though this is an approach that is more re ective of clinical reality. Also of note, viral infections, which are known to contribute a high proportion to cases of CAP (18), were under-represented in the larger study from which data was collected. While this may have been in part due to enrollment (from March to September) outside of the peak of respiratory viral season, the contribution of non-standardized sampling cannot be discounted.
Despite the aforementioned limitations, this is an exploratory and hypothesis-generating study that adds to the growing literature on the diagnostic utility of PCT values in pneumonia. Unlike previous studies which were limited to patients with CAP secondary to relatively drug-sensitive organisms, our study contains a large proportion of MDRO bacteria. In addition, our serial PCT data suggests the measurement of a single admission PCT provides limited information, and that serial values have more utility for the evaluation of pneumonia.

Conclusions
The identi cation of a bacterial etiology in pneumonia is often challenging. We examined the relationship between serial PCT measurements and bacterial etiology in patients with community-acquired pneumonia, including many with risk factors for MDRO infections. We found higher procalcitonin values for Streptococcus pneumoniae relative to other etiologies, a delayed rise for Pseudomonas over time, and consistently low PCT values for infections due to multiple bacteria. We also found that PCT levels obtained on day 2 of hospitalization were the most useful in distinguishing between different bacterial etiologies. The study was submitted to and approved by the Massachusetts General Hospital Institutional Review Board (Protocol no. 2012P001590), and the study was granted a waiver of informed consent. All administrative permissions required in order to access and use the data in the study were granted by the Massachusetts General Hospital Institutional Review Board.

Consent for publication
Not applicable Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.