Procalcitonin Levels Predict Bacteremia in Patients With Community-Acquired Pneumonia: A Prospective Cohort Trial FREE TO VIEW

Community-acquired pneumonia (CAP) is a common and potentially life-threatening disease that puts an enormous strain on health-care resources.¹ In the diagnostic workup of patients with CAP, identification of the causative organism to target a more favorable antibiotic therapy and to study local resistance patterns is of great value.^1,2 Current guidelines recommend routine drawing of two sets of pretreatment blood cultures in all hospitalized patients with CAP.³ Nevertheless, the benefit and cost-effectiveness of routine drawing of blood cultures is controversial mainly because of the low yield of positive blood cultures, which is in the range of 3% to 10% in nonselected hospital-admitted patients with CAP.^4-10 In order to limit blood culture sampling to high-risk patients for growth of bacteria, previous studies have evaluated clinical and laboratory predictors for blood culture positivity.^6,11-13 However, single parameters lack sensitivity, specificity, or both, which prevents its use in daily routine.^6,11 Some authors have developed clinical decision rules, with increased prognostic accuracy.^6,12,13 However, complexity of decision rules often restricts a widespread adoption. Hence, there is an unmet need to identify simple and accurate predictors for blood culture positivity in patients with CAP.

Procalcitonin (PCT) has emerged as a biomarker for bacterial infections because it correlates with the extent and severity of microbial invasion in different infections.^14-22 Previous trials have demonstrated that patients with bacteremic CAP have markedly increased initial PCT concentrations,^9,14,17,22 which makes PCT a promising candidate biomarker for prediction of blood culture positivity. Herein, we evaluated the prognostic accuracy of PCT as compared with other commonly used clinical and laboratory parameters in a large and well-defined derivation and validation cohort of patients with CAP.

The present study is a predefined substudy of the previously published Procalcitonin Guided Antibiotic Rherapy and Hospitalization in Patients with Lower Respiratory Tract Infections (ProHOSP) trial and evaluated data from 925 patients with radiologic-confirmed CAP between December 2006 and March 2008. A detailed description of the ProHOSP study has been published elsewhere.^14,23 In brief, consecutive patients with lower respiratory tract infection (LRTI) were included in six secondary- and tertiary-care hospitals in northern and central Switzerland. The aim of this randomized noninferiority trial was to compare two different treatment strategies using either a PCT algorithm or an algorithm based on current guidelines for conducting antibiotic therapy. The primary end point was the combined medical failure rate of patients. A study Web site provided information on the evidence-based management of all patients based on the most recent guidelines^3,24-27 and explicitly specified the need for radiograph confirmation of CAP and for sampling two sets of pretreatment blood cultures. A predefined secondary end point of this study was the evaluation of prognostic parameters for outcomes and blood culture positivity. Full ethical approval for this trial was obtained from all local ethical committees, and all patients gave written informed consent.

Inclusion criteria for patients were written informed consent, age ≥ 18 years, and hospital admission from the community or a nursing home for LRTI. Exclusion criteria were the inability to give written informed consent, insufficient German-language skills, active illegal IV drug use, previous hospitalization for LRTI within 14 days, severe immunosuppression other than use of corticosteroids, accompanying chronic infection or endocarditis, and most severe medical comorbidities where death was imminent.

LRTI was defined by the presence of at least one respiratory symptom (cough with and without sputum production, dyspnea, tachypnea, pleuritic pain) plus one auscultatory finding or one sign of infection (core body temperature > 38.0°C, shivers, WBC count > 10 g/L or < 4 g/L) independent of antibiotic pretreatment. Diagnosis of CAP was made if, in addition to the LRTI criteria, an underlying infiltrate on chest radiograph was present.²⁴

In all patients given a provisional diagnosis of CAP, two pairs of blood cultures for both aerobic and anaerobic conditions were collected before starting antibiotic therapy. Blood cultures were processed using an automated colorimetric detection system (BacT/ALERT; bioMerieux; Durham, NC) in three hospitals and an equivalent blood culture system (BACTEC; Becton-Dickinson; Cockeysville, MD) in the other three hospitals.²⁸ If blood culture bottles indicated bacterial growth, samples were Gram stained and subcultured. The correct identification of the pathogen was achieved according to routine laboratory procedures.

In accordance with a recently published study,⁶ a bacteremic episode was defined as growth of a typical organism for CAP in at least one of four collected blood cultures within the first 36 h of presentation to the ED. Episodes with growth of coagulase-negative staphylococci were assumed to be contaminants. Growth of Streptococcus species other than pneumococci (n = 4) and enterobacteriaeceae (n = 3), including Serratia marcescens, in the blood cultures of one patient were included in this analysis, even though a causal relationship with CAP was not clear.

In all patients on hospital admission, a thorough clinical examination was performed, and two prognostic scores, Pneumonia Severity Index (PSI) and CURB-65 (confusion, urea > 7 mmol/L, respiratory rate ≥ 30/min, low systolic [< 90 mm Hg] or diastolic [≤ 60 mm Hg] BP, age ≥ 65 years), were calculated.^29,30 For all patients, laboratory results were collected from the routine blood analysis, including markers of infection (PCT, C-reactive protein [CRP], and WBC), plasma sodium concentration, and blood urea nitrogen. CRP concentrations were determined by an enzyme immunoassay having a detection limit of < 5 mg/dL (EMIT; Merck Diagnostica; Zurich, Switzerland). PCT was measured using a time-resolved amplified cryptate emission technology assay (Kryptor PCT; Brahms AG; Hennigsdorf, Germany) with a functional assay sensitivity of 0.06 μg/L.

We used the first 50% (n = 463) of patients with CAP as the derivation set and the second 50% (n = 462) as the validation set based on the timely inclusion of patients. The validation set was not used until the analysis with the derivation set was complete, and it served as an independent test of the derived rule.

This report adheres to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for reporting observational studies.³¹ Discrete variables are expressed as counts (percentage) and continuous variables as medians and interquartile ranges (IQRs). Frequency comparison was done by χ² test. Two-group comparison Mann-Whitney U test was used. To assess the prognostic performance of different parameters to predict blood culture positivity, univariate and multivariate regression analyses adjusted for all significant parameters were used. Thereby, logarithmic transformation was performed to obtain normal distribution for skewed variables (ie, PCT concentrations), and outcomes were either positive blood cultures or negative blood cultures. Receiver operating characteristics were calculated with STATA 9.2 statistical software (Stata Corp; College Station, TX). All testing was two tailed, and P values < .05 were considered to indicate statistical significance.

The median age of the overall cohort of 925 patients with CAP was 73 years (IQR, 59-82 years), and 59% were men. The median PSI score was 91 (IQR, 66-115), and 51% of patients were in high-risk PSI classes IV and V. True-positive blood cultures were detected overall in 73 patients (43 of 463 in the derivation cohort and 30 of 462 in the validation cohort); thus, the overall rate of true-positive blood cultures was 7.9% (Fig 1). The following pathogens were detected: Streptococcus pneumoniae (n = 59), Escherichia coli (n = 3), Haemophilus influenzae (n = 2), Enterobacter cloacae (n = 2), Enterobacter aerogenes (n = 1), Streptococcus pyogenes (n = 1), Streptococcus acidominimus (n = 1), Streptococcus mitis (n = 1), Pseudomonas aeruginosa (n = 1), Staphylococcus aureus (n = 1), and Serratia marcescens (n = 1). Three patients had contaminated blood cultures, with coagulase-negative staphylococci in a single blood culture bottle each. Table 1 shows baseline characteristics on hospital admission of all patients (left column) and separated according to blood culture results. Patients with positive blood cultures were less frequently admitted to the hospital from nursing facilities and had less frequent antibiotic pretreatment and congestive heart failure, whereas renal dysfunction was more frequent. Laboratory analysis showed that CRP, blood urea nitrogen, and WBC counts were significantly higher in patients with positive blood cultures. In addition, patients with positive blood cultures had almost 15-fold higher PCT levels than those with negative cultures (5.8 μg/L vs 0.4 μg/L, respectively). There was no significant difference in PCT levels in patients with S pneumoniae (median PCT, 5.8 μg/L; IQR, 2.1-21.1 μg/L) compared with patients with other pathogens (median PCT, 6.3 μg/L; IQR, 3.3-11.4 μg/L) (P = .98). Although the majority of patients with positive blood cultures had increased PCT levels, one patient with a PCT level < 0.1 μg/L and 2 patients with a PCT level < 0.25 μg/L had growth of S pneumoniae in blood cultures. These three patients had increased PCT levels of > 0.25 μg/L in the follow-up PCT measurement. Patients with positive blood cultures were more frequently transferred to the hospital ICU, but mortality was similar.

In univariate analysis (Table 2), antibiotic pretreatment, congestive heart failure, and systolic BP were negative predictive factors for positive blood cultures. In contrast, renal dysfunction, heart rate, body temperature, blood urea nitrogen, WBC count, CRP serum levels, and PCT serum levels were positive predictors for positive blood cultures. Multivariate logistic regression analysis using all significant parameters from the univariate analysis showed that only antibiotic pretreatment (adjusted OR, 0.25; 95% CI, 0.08-0.76; P < .05) and PCT serum levels (adjusted OR, 3.72; 95% CI, 2.31-5.95; P < .001) were independent predictors for negative and positive blood cultures, respectively.

To assess the overall discriminatory ability of different parameters, we calculated receiver operating characteristic curves (Fig 2). The areas under the curve of PCT were similar in the derivation (0.83; 95% CI, 0.78-0.89) and validation (0.79; 95% CI, 0.72-0.88) cohorts (Table 3). With an area under the curve of 0.82 (95% CI, 0.78-0.87) in the overall population, PCT had the highest diagnostic accuracy to predict positive blood cultures than CRP and other clinical and laboratory predictors. At a PCT cutoff of 0.1 μg/L, sensitivity to predict positive blood cultures was 98% and 100% in the derivation and validation cohorts, respectively, and 99% in the overall cohort. Table 4 shows sensitivity, specificity, and positive and negative likelihood ratios for all parameters.

Blood culture positivity within different risk categories of PCT, CRP, and the two clinical risk scores (PSI and CURB-65) was assessed. Figure 3 shows the percentage of patients with positive (dark gray) and negative (light gray) blood cultures within risk categories. In patients with a PCT value < 0.1 μg/L and between 0.1 μg/L and 0.25 μg/L, only 0.9% (one of 117) and 0.9% (two of 224), respectively, had positive blood cultures, whereas 16.8% (61 of 364) had positive results with PCT levels > 1.0 μg/L. Patients with CRP < 20 mg/dL and between 20 mg/dL and 50 mg/dL, 3.7% and 5.9%, respectively, had positive culture results. Except for the highest risk category, the occurrence of positive cultures was balanced in the different risk classes of both clinical risk scores.

We calculated the potential cost-savings using different PCT cutoff values. We assumed costs of 145 US dollars (USD) for two sets of blood cultures per patient based on institutional data of the University Hospital in Basel. Thus, total costs for the whole CAP study cohort (925 patients × 145 USD) were estimated to be 134,125 USD. Similarly, we assumed costs of 30 USD per PCT measurement, resulting in total costs of 27,750 USD for the cohort. Estimated total costs for different PCT cutoffs are presented in Table 5. Using a cutoff value of 0.1 μg/L would reduce the total number of blood cultures by 12.6% (n = 117) while still identifying 99% (72 of 73) of positive blood cultures. Similarly, using a cutoff of 0.25 μg/L and 0.5 μg/L would result in 37% and 52%of blood cultures, respectively, while still identifying 96% and 88% of positive culture results, respectively.

In this large prospective multicenter study of patients with CAP admitted to the hospital from the ED, PCT proved to be the most reliable predictor of blood culture positivity. Depending on the cutoff applied, PCT levels of < 0.25 μg/L identified patients at very low risk for bacteremic episodes and, thus, helps to avoid unnecessary blood culture sampling, with resulting financial benefits. Conversely, increased PCT levels > 0.5 μg/L and especially > 1 μg/L may help to identify high-risk patients who would benefit from early and aggressive diagnostic workup and antibiotic therapy.

Routine sampling of blood cultures in patients with CAP have been a cornerstone for epidemiologic reasons, for monitoring antibiotic resistance patterns, and for better streamlining of antibiotic therapy in individual patients.^1,24,27 Nevertheless, the routine implementation of blood culture sampling in CAP has been questioned because of the low yield of true-positive results.^4-9 Consensus guidelines encourage a more rational approach to blood culture collection without, however, giving specific criteria.²⁴ Thus, there is an unmet need for accurate predictors of blood culture positivity that would increase pretest probability and, thus, the yield of blood cultures in patients with CAP. In this context, enormous attempts have been undertaken to correlate the levels of different biomarkers and mediators with the presence of bacteremia.^19,32 Herein, PCT is a promising biomarker as it correlates with the extent and severity of microbial infection because of the high specificity for bacterial etiology of LRTI and the superior clinical usefulness compared with other commonly used laboratory tests, namely CRP and WBC count.^15-22,32

This study validates a series of previous findings. In accordance with previous studies, we found that antibiotic pretreatment is an important factor to decrease the likelihood of positive blood culture findings.^6,33 In a large retrospective cohort study,⁶ different predictors for bacteremia in hospitalized patients with CAP were identified. Antibiotic pretreatment and 40°C < body temperature < 35°C were negative predictors, whereas liver disease, systolic BP < 90 mm Hg, heart rate ≥ 125 bpm, blood urea nitrogen ≥ 11 mmol/L, sodium < 130 mmol/L, and WBC count < 5,000/mm³ or > 20,000/mm³ were positive predictors. Using this decision support tool would result in 38% fewer blood cultures than in standard practice and still allow identification of 88% to 89% of patients with bacteremia.⁶ In our study, we validate these predictors; however, after including PCT in multivariate analysis, only antibiotic pretreatment and PCT remained independent predictors. Using the PCT cutoff of 0.25 μg/L indicated in our study would allow reduction of the amount of blood culture collection by a similar amount (37%) but would still result in a higher sensitivity of 96%.

Waterer and Wunderink⁷ reported that the yield of positive blood cultures in patients with CAP increases with increasing PSI. Thus, they concluded that blood cultures should be limited to high-risk patients in PSI classes IV and V. In the present study, we found no significant correlation between PSI class and the yield of positive blood cultures and only a weak association between the CURB-65 score and likelihood of positive blood cultures. The PSI is a mortality prediction tool and is mainly influenced by patient age.²⁹ Thus, young patients with bacteremic CAP usually have a lower PSI but still a high risk of blood culture positivity.

Based on the results of two retrospective studies reporting a reduced mortality in patients with CAP who received antibiotic treatment within 4 to 8 h, the time to first antibiotic dose recently has received significant attention from a quality-of-care perspective.²⁴ Within the ProHOSP trial, PCT was measured with a rapid sensitive assay and an assay time of approximately 20 min.¹⁴ The test was performed on site at the central laboratory of each participating hospital, and results were routinely available around the clock within 1 h and by the time results from routine chemistry were available. Thus, if PCT is measured on hospital admission of patients with suspected CAP, PCT can be used for the evaluation of the patients without putting them at risk because of time delays.

Some limitations merit consideration. First, patients with severe immunosuppression were excluded, limiting generalization. In the blood cultures of seven patients, we found some pathogens that are not typically found in CAP and, thus, might not be the true cause for the CAP diagnosis. Although we excluded patients with other sites of infection than the respiratory tract, it is not entirely possible to rule out the influence of concomitant infections. Exclusion of these patients, however, would not have altered our results (data not shown).

In conclusion, this study suggests that PCT is an accurate parameter for predicting bacteremia in patients with CAP. Based on the results of this study, we recommend that blood cultures be drawn from patients with CAP only when PCT levels are ≥ 0.25 μg/L because the likelihood for bacteremic CAP in patients with lower PCT levels is very low. In addition, blood cultures should be considered at lower PCT levels if antibiotics are prescribed based on validated overruling criteria, especially in areas with increased prevalence of resistant organisms. Obviously, the clinical picture of bacteremic infection is far too heterogeneous and complex to be reduced to a single cutoff of any surrogate marker. Different microbes might induce distinct responses, resulting in a variable upregulation of circulating PCT levels. Still, as demonstrated in this and other studies, the likelihood for bacterial CAP increases gradually with increasing serum levels of PCT, making PCT a putative indicator for blood culture positivity. Thus, used in the proper context of CAP, it may result in a substantial reduction in the numbers of cultures obtained, an optimized allocation of our limited health-care resources, and lower patient costs.

Author contributions:Dr Müller: performed the analyses, drafted the manuscript, amended and commented on the manuscript, and approved the final version.

Dr Christ-Crain: contributed to the study idea and its initiation, amending and commenting on the manuscript, and approving the final version.

Dr Bregenzer: contributed to amending and commenting on the manuscript and approving the final version.

Dr Krause: contributed to amending and commenting on the manuscript and approving the final version.

Dr Zimmerli: contributed to the study idea and its initiation, amending and commenting on the manuscript, and approving the final version.

Dr Mueller: contributed to the study idea and its initiation, performing the analyses, drafting the manuscript, amending and commenting on the manuscript, and approving the final version.

Dr Schuetz: contributed to the study idea and its initiation, performing the analyses, drafting the manuscript, amending and commenting on the manuscript, and approving the final version.

Financial/nonfinancial disclosures: The authors have reported to the CHEST the following conflicts of interest: Drs Christ-Crain, Mueller, and Shuetz received support from Brahms to attend meetings and fulfill speaking engagements. Dr Mueller has served as a consultant and received research support. Drs Müller, Bregenzer, Krause, and Zimmerli have reported that no potential conflicts of interest exist with any companies/organizations whose products or services may be discussed in this article.

Other contributions: We thank all local physicians, the nursing staff, and the patients who participated in this study. We especially thank the staff of the ED, medical clinics, and central laboratories of the University Hospital Basel; the Kantonsspitaeler Liestal, Aarau, Luzern, and Münsterlingen; and the Buergerspital Solothurn for their helpful assistance, patience, and technical support. We also thank the other members of the Data Safety and Monitoring Board, namely A. P. Perruchoud, S. Harbarth, and A. Azzola, and all members of the ProHOSP Study Group.

Role of sponsors: No commercial sponsor had any involvement in the design and conduct of this study, namely collection, management, analysis, and interpretation of the data and preparation, decision to submit, review, or approval of the manuscript.

The members of the ProHosp Study Group are Robert Thomann, MD; Claudine Falconnier, MD; Marcel Wolbers, PHD; Isabelle Widmer, MD; Stefanie Neidert, MD; Thomas Fricker, MD; Claudine Blum, MD; Ursula Schild, RN; Katharina Regez, RN; Rita Bossart, RN; Ronald Schoenenberger, MD; Christoph Henzen, MD; Claus Hoess, MD; Heiner C. Bucher, MD; Ayesha Chaudri, MD; Jeannine Haeuptle, MD; Roya Zarbosky, MD; Rico Fiumefreddo, MD; Melanie Wieland, RN; Charly Nusbaumer, MD; Andres Christ, MD; Roland Bingisser, MD; Kristian Schneider, RN; Christine Vincenzi, RN; Michael Kleinknecht, RN; Brigitte Walz, PhD; Verena Briner, MD; Dieter Conen, MD; Andreas Huber, MD; and Jody Staehelin, MD. Aarau: Chantal Bruehlhardt, RN; Ruth Luginbuehl, RN; Agnes Muehlemann, PhD; Ineke lambinon; Max Zueger, MD; D.Conen, MD; M.Wieland, RN; C. Nusbaumer, MD; C. Bruehlhardt, RN; R. Luginbuehl, RN; A. Huber, MD; B.Walz, RN; and M. Zueger, MD.