Rapid diagnosis of bloodstream infections (BSIs) in the emergency department (ED) is challenging, with turnaround times exceeding the timeline for rapid diagnosis. We studied the usefulness of procalcitonin as a marker of BSI in 367 adults admitted to our ED with symptoms of systemic infection. Serum samples obtained at the same time as blood cultures were available from 295 patients. Procalcitonin levels were compared with blood culture results and other clinical data obtained during the ED visit. Procalcitonin levels of less than 0.1 ng/mL were considered negative; all other levels were considered positive. In 16 patients, there was evidence of BSI by blood culture, and 12 (75%) of 16 patients had a procalcitonin level of more than 0.1 ng/mL. In 186 (63.1%) of 295 samples, procalcitonin values were less than 0.1 ng/mL, and all were culture negative. With a calculated threshold of 0.1475 ng/mL for procalcitonin, sensitivity and specificity for the procalcitonin assay were 75% and 79%, respectively. The positive predictive value was 17% and the negative predictive value 98% compared with blood cultures. Procalcitonin is a useful marker to rule out sepsis and systemic inflammation in the ED.
More than 500,000 bloodstream infections (BSIs) are reported annually in the United States and are associated with high mortality rates.1–5 Early diagnosis and implementation of “early goal-directed therapy” are essential components of the international Surviving Sepsis Campaign guidelines6–9 and result in reduced mortality. Similarly, differentiating sepsis from other noninfectious causes of systemic inflammation is often difficult because fever and leukocytosis have poor sensitivity and specificity in many clinical settings.8,10–15 Since sepsis cannot often be excluded, empiric antibiotic treatment is commonly prescribed in the acute care setting. Emergency departments (EDs) are often the first point of contact and care for many patients, and early diagnosis of bacteremia is extremely important but a diagnostic challenge.16,17 Blood cultures (BCs) are the “gold standard” for diagnosis of sepsis.3 However, test results (Gram-stain results, definitive organism identification, and antimicrobial susceptibility test results) are typically not available for 12 to 48 hours. This delay has prompted the development and evaluation of rapid and molecular tests for inflammatory and other markers of infection.14,18,19
An additional challenge in BCs that are obtained in the ED is that they have high rates of false-positive results due to organisms known to be skin contaminants (eg, coagulase-negative staphylococci). Patients admitted with febrile illness and positive BCs due to contaminant organisms undergo unnecessary diagnostic measures, hospitalization, and unwarranted antimicrobial therapy. The ability for more rapid diagnosis of sepsis and accurate assessment of disease severity and mortality risk at the ED admission combined with early initiation of appropriate treatment may have a great impact on outcome.
The observation that the serum procalcitonin level rises above normal values in patients with sepsis and other clinically significant bacterial infections was first described in 1993 by Assicot et al.20 The procalcitonin level is not elevated in viral infections. Recent studies, primarily conducted in the critical care setting, have evaluated procalcitonin as a diagnostic tool to define patients with sepsis and systemic inflammatory response syndrome (SIRS).21–25
Because procalcitonin analysis can be performed in less than 1 hour, we propose that procalcitonin would have substantial usefulness in the ED setting to assist evaluation of febrile patients at risk for bacteremia and sepsis. In this study, we evaluated the usefulness of procalcitonin as a diagnostic predictive marker of bacteremia and sepsis in the ED in an adult patient population.
Materials and Methods
Study Population and Review of Medical and Laboratory Records
Between January 2008 and March 2009, we evaluated medical records of patients seen in the ED at a university-affiliated, tertiary care medical center (Johns Hopkins Bayview Medical Center, Baltimore, MD), who had BCs obtained in the ED. This study was approved by the Johns Hopkins Institutional Review Board.
Adult patients (18 years or older) with symptoms of systemic infection and/or inflammation for whom a BC was obtained during the initial ED visit were enrolled. Laboratory records were screened for the presence of available, stored blood specimens (serum or EDTA blood) for additional procalcitonin analysis. Blood samples from serum-separator tubes were centrifuged, and an aliquot of the serum was removed and stored at −70°C for procalcitonin analysis. Clinical data were abstracted from the medical records and included age, sex, SIRS criteria (WBC count, body temperature, blood pressure, and heart rate), initial clinical diagnosis by International Classification of Diseases, Ninth Revision, Clinical Modification code, and BC results.
An episode of bacteremia or sepsis was defined as the recovery of any significant, pathogenic bacterial species in 1 or 2 sets of BCs (aerobic and anaerobic bottles) obtained in the ED. Organisms commonly considered as BC contaminants (eg, coagulase-negative staphylococci, aerobic and anaerobic diphtheroids, Micrococcus sp, Bacillus sp) were excluded from this definition.26 BCs were processed using the VersaTREK continuous BC monitoring system (TREK Diagnostics, Cleveland, OH). Bacteria from positive BCs were further identified using standard laboratory methods. Cultures that did not indicate growth within 5 days of incubation in the VersaTREK system were considered negative.
Procalcitonin Assay Analysis
Following the completion of requested laboratory tests and within 24 hours of sample collection, blood samples from ED patients were centrifuged at 4°C, and a portion of the serum was collected and stored at −70°C. Batched serum samples were analyzed for procalcitonin levels using the Kryptor chemiluminescence immunoassay (BRAHMS, Hennigsdorf, Germany). The assay measures procalcitonin levels in human serum or plasma using a sandwich immunoassay approach based on the principle of time-resolved amplified cryptate emission. The functional sensitivity of the assay is 0.06 ng/mL. All testing was performed according to the manufacturer’s specifications.27 The data for patients for whom serum samples or procalcitonin measurements were not available were excluded from further analysis.
We defined 3 categories for interpretation of procalcitonin values: 0.0 to 0.1 ng/mL, 0.1 to 1.0 ng/mL, and more than 1.0 ng/mL. The cutoff values were selected based on the analytic sensitivity of the assay (0.06 ng/mL) and previously published data.28–30 For the purpose of graphic display, the log method was chosen, and a “zero” procalcitonin value was set to be defined as 0.0001. Other clinical and laboratory parameters, eg, WBC count, were interpreted according to the standard and accepted SIRS criteria and categories.31 Categorical analysis (Fisher exact test) and continuous analysis (Wilcoxon signed rank test) were performed using the DiagnosisMed library for the R statistical package.32 The gold standard was defined as a BC result positive for a pathogen, excluding organisms commonly considered BC contaminants (eg, coagulase-negative staphylococci, aerobic and anaerobic diphtheroids, Micrococcus sp, Bacillus sp). Cultures positive for the latter organisms were considered as false-positives or “negative” for the purposes of statistical analysis.26 The optimal procalcitonin cutoff value was identified by maximizing the area under the receiver operating characteristic curve and checked against the optimal cutoff value for minimizing the cost of misclassification, assuming that the cost of misclassifying false-negative results takes on integers from 1 to 50.
A total of 367 patients (average age, 47.7 years; range, 18–94 years) were enrolled; the data for 72 patients (average age, 47.9 years; range, 22–94 years) were excluded from further analysis because a serum sample, collected within 24 hours of the BC, was not available for procalcitonin analysis. The demographic data for these patients and the percentage in distribution of BC results (positive-pathogen, positive-contaminant, negative) did not differ in comparison with those for all patients whose data were included in the analysis for whom procalcitonin and BC results were available.
The remaining 295 patients (average age, 47.5 years; range, 18–92 years) were classified according to the presence or absence of pathogenic organisms isolated in BCs and the level of procalcitonin detected in their serum samples Table 1. Statistically significant differences for procalcitonin levels in relation to patients’ BC results were observed (P = .0004); results are shown in Figure 1. Differences for procalcitonin values were statistically significant between patients with BCs positive for pathogens and patients with negative BCs (P = .00007) and for procalcitonin levels for patients with BCs positive for pathogens compared with patients with BCs positive for contaminants (P = .01). Procalcitonin levels did not differ between patients with negative BCs and patients with BCs positive for contaminants (P = .39).
When comparing clinical criteria for SIRS and procalcitonin results, we did not find any statistically significant correlation, irrespective of the number of SIRS criteria present in individual patients with positive or negative BC results Figure 2. Median procalcitonin levels did not differ between patients who had 0 or 1 SIRS criteria and patients who had 2 or 3 SIRS criteria (P = .66). Likewise, no significant difference was seen between patients with 0, 1, or 2 SIRS criteria compared with patients who had 3 SIRS criteria (P = .21). We also compared the presence of SIRS criteria in patients for whom procalcitonin values were available with SIRS criteria present in patients for whom no procalcitonin values were measured. Again, no statistically significant differences were observed.
The results of the procalcitonin test properties are illustrated in Figure 3. The procalcitonin cutoff value that maximizes the area under the receiver operating characteristic curve (AUC), relative to the gold standard of a BC positive for a pathogen, has been identified as 0.1475 ng/mL. At this level, the following statistical parameters were calculated: sensitivity, 75%; specificity, 78.9%; positive predictive value (PPV), 16.9%; negative predictive value (NPV), 98.2%; AUC, 0.79; and accuracy, 78.6%. With a threshold for the cutoff value of procalcitonin of 1.0 ng/mL in relation to positive BC results, the sensitivity was 37.5%, specificity was 92.8%, PPV was 23.1%, NPV was 96.3%, AUC was 0.65, and accuracy was 89.8. A threshold of 0.1 ng/mL for procalcitonin in relation to positive BC results resulted in the following statistical calculations: sensitivity, 75%; specificity, 70.6%; PPV, 12.8%; NPV, 96.3%; AUC, 0.73; and accuracy, 70.9%. These results illustrate a strong NPV for procalcitonin as a screening test for the presence of bacteremia in patients admitted to the ED; however, sensitivity and PPV remained low, even when previously described, validated cutoff values were used.
For 30 patients, there were unexpected results for procalcitonin in comparison with BC results Table 2. Of the 30 patients, 12 had already received antimicrobial therapy at the time of BC collection. Of the remaining 18 patients, 4 had BC results positive for a pathogen, but corresponding procalcitonin levels were less than 0.1 ng/mL. In 6 patients, there were positive BCs with a pathogen, but procalcitonin levels were between 0.1 and 1.0 ng/mL. Finally, 8 patients had procalcitonin values greater than 1.0 ng/mL, but their corresponding BCs were negative for growth. In these 8 patients, various clinical conditions (ie, inflammatory diseases or infections other than BSI) were identified that could explain increased levels of procalcitonin. In addition, all 8 patients in this group had a history of end-stage renal disease and hemodialysis.
Rapid detection of bacteremia facilitates early implementation of therapy and identifies patients at high risk for complications.33,34 Previous studies demonstrated that various clinical markers have poor sensitivity and specificity for predicting early bacteremia in febrile patients.12–16,33 Similarly, ruling out bacterial sepsis in febrile patients has substantial benefits, including reduction of hospitalization and antimicrobial use and facilitating clinician focus on alternative diagnostic pathways.34–36 In the present study, we found that procalcitonin is a promising candidate marker for rapid detection of BSI.
In this study of patients recruited from the ED setting, using BC result as the gold standard, we found that procalcitonin levels are dramatically different in patients with and without bacteremia, consistent with previous reports.20,21,34,35 Previous studies have investigated the usefulness of using procalcitonin as a positive predictive marker for BSI or sepsis with varying outcomes regarding its clinical suitability.37–39 In patients without bacteremia or systemic infections, procalcitonin levels are 0.1 ng/mL or less.28–30,40 With less than 0.1 ng/mL used as the cutoff, procalcitonin had a 96.3% NPV for ruling out bacteremia in the ED compared with BC as the gold standard. This high NPV is consistent with previous reports from multiple clinical settings. Liaudat et al41 studied procalcitonin as an early marker of sepsis in a hospitalized patient population (8% prevalence of bacteremia) and found a high NPV, depending on the cutoff value for the procalcitonin level (99% for 0.2 ng/mL and 95% for 0.5 ng/mL). In a study of 300 hospitalized febrile patients, Bossink et al14 reported a 90% NPV using a 0.5 ng/mL cutoff value for procalcitonin.
Elevated body temperature above 38.3°C is frequently used as a clinical predictor for bacteremia or sepsis (even though it has poor predictive value and may be a component of many different conditions and diseases).42–45 Studies have found that the rate of undiagnosed episodes of bacteremia or sepsis in febrile patients ranges from 15% to 50%.43–45
Based on the results in this study, we propose that a procalcitonin value of 0.1 ng/mL or less could be used to rule out bacteremia (NPV, 96.3%). We also found a highly significant correlation between procalcitonin levels and positive BC results. Previous studies have demonstrated that procalcitonin levels are correlated with the severity of sepsis and could potentially be used as a prognostic marker in patients with sepsis and severe sepsis.23,36,37 In our study, we observed differences in the numbers of BC sets turning positive with pathogens. Considering that 4 patients with positive BCs had procalcitonin levels of 0.1 ng/mL or less, we found that in these 4 patients a total of 11 BC sets were collected, yet only 4 sets, 1 for each patient, were positive (36%). The bacteremia in these patients was presumably of low magnitude or even transient, possibly resulting in low procalcitonin levels. On the other hand, in 6 patients with procalcitonin levels of 0.1 to 1.0 ng/mL, a total of 20 BC sets were collected and 9 sets were positive (45%); in patients with procalcitonin levels more than 1.0 ng/mL, 14 (82%) of 17 BC sets collected turned positive. We believe that our results support the previously described correlations of procalcitonin with severity of sepsis. In our study, patients with significant bacteremia and higher procalcitonin levels had significantly more BC sets positive for the pathogen detected, suggesting a more severe form of BSI.
We also observed a significant difference in the length of time to detection (LTD) for positive BCs in relation to procalcitonin levels; in the group of positive BCs with procalcitonin levels less than 0.1 ng/mL, the LTD was on average 57.4 hours, whereas in the group with procalcitonin levels more than 1.0 ng/mL, the mean LTD was 17.0 hours. Studies investigating the impact of LTD and results reporting on length of hospital stay and patient outcome support the importance of shorter LTD to improve patient care.46 Considering these data and shorter LTDs as an indicator of significant bacteremia, our study demonstrated that all significant BSIs were associated with higher levels of procalcitonin.
These observations help to discriminate between BC contaminants and true pathogens. Depending on country and region, false-positive BCs account for up to 50% of all BCs collected.47 BCs collected in the ED, which are often performed under suboptimal conditions, are frequently contaminated by common skin flora.48–50 False-positive BCs result in additional inpatient days, additional clinical evaluations and diagnostic tests, and unnecessary antimicrobial administration. Although algorithms have been proposed during past decades to reduce the burden of false-positive BCs,26,51,52 a consensus on a gold-standard approach has never been achieved. As such, there is a need for a rapid diagnostic tool to differentiate true-positive from false-positive BCs. In our study, procalcitonin levels in patients with negative BCs and false-positive BCs were similar, and we, therefore, conclude that the procalcitonin level could be used as a rapid tool to rule out bacteremia even in the setting of false-positive BCs. Considering the fact that some studies suggested that the number of BCs positive for the same organism can be used to determine the possibility of BC contamination,26 we believe that procalcitonin results can be used to further support such algorithms.
Because procalcitonin results are available long before BC results, procalcitonin results not only can be an important screening tool to rule out bacteremia but also will further assist in the retrospective assessment of BC results when perhaps only 1 or 2 sets of BCs are positive for organisms known to be potential contaminants. Furthermore, because the test requires only a small amount of serum, it could easily be implemented in the ED as an adjunct laboratory test to distinguish patients with bacteremia from those with a BC positive with a contaminant such as coagulase-negative staphylococci and as a predictor of severity of illness.53
As previously shown, many clinical observations, including SIRS criteria, in ED patients have poor predictive values for diagnosing bacteremia and/or sepsis at the bedside.12,13 The results in our study further support these observations because no statistically significant correlation was seen between procalcitonin values and various combinations of SIRS criteria, while procalcitonin and BC results showed significant correlations with regard to the presence of BSI or sepsis.
Our study has some limitations. The primary aim of this study was to correlate procalcitonin results with BC results, and it included only a limited set of additional clinical data (eg, standard SIRS criteria) in the analysis. Only a limited number of patients with culture-confirmed bacteremia were available for the purpose of this study. As expected, the majority of patients enrolled had negative BCs and were seen in the ED for reasons other than bacteremia or sepsis. We also identified a small subset of patients in our study with markedly elevated procalcitonin levels but negative BC results. It is interesting that all of these patients had a history of end-stage renal disease and hemodialysis but no other commonalities in clinical conditions. Further studies are necessary to better understand the role of procalcitonin as a marker of inflammation in such patient populations because significantly elevated procalcitonin levels in these patients may not consistently be associated with bacteremia and/or sepsis. Considering the performance characteristics of procalcitonin in our study, as well as other previously described studies, it appears to be most reasonable to use the procalcitonin level to rule out bacteremia rather than to predict the presence of a BSI. However, once a procalcitonin value above the defined cutoff has been identified, BCs could be collected in a controlled clinical setting using appropriate skin antisepsis to reduce the risk of contamination. Another factor to consider in our study setup was the possibility of patients receiving antibiotic treatment before BCs were obtained, hence resulting in “false-negative” BC results.
In this retrospective study, procalcitonin seems to be a useful marker to rule out bacteremia in ED patients in our hospital, using a cutoff value of 0.1 ng/mL or less for procalcitonin. BCs may not be necessary in this patient population. A procalcitonin value of 1.0 ng/mL or more was indicative of the presence of bacteremia, and BCs are needed to confirm and identify the presence of bacterial organisms. Additional studies may be necessary to further validate these proposed cutoff values for procalcitonin and the interpretation of procalcitonin values between 0.1 and 1.0 ng/mL. For patients whose procalcitonin values are between 0.1 and 1.0 ng/mL, BCs are still necessary to rule out bacteremia. If further confirmed, procalcitonin could be a very useful and rapid marker in the decision tree to rule out sepsis in ED patients.
We thank Lauren Powell for assistance in the clinical data collection process and medical chart review.
Upon completion of this activity you will be able to:
recognize the complexity and challenges of diagnosing sepsis in patients presenting to the emergency department.
describe the role of clinical and systemic inflammatory response syndrome (SIRS) criteria in the diagnosis of sepsis.
describe the relation between procalcitonin and clinical/SIRS criteria in the diagnosis of sepsis.
discuss the role of procalcitonin and proposed cutoff values for diagnosing sepsis in patients presenting to the emergency department.
The ASCP is accredited by the Accreditation Council for Continuing Medical Education to provide continuing medical education for physicians. The ASCP designates this educational activity for a maximum of 1 AMA PRA Category 1 Credit ™ per article. This activity qualifies as an American Board of Pathology Maintenance of Certification Part II Self-Assessment Module.
The authors of this article and the planning committee members and staff have no relevant financial relationships with commercial interests to disclose.
Questions appear on p 310. Exam is located at www.ascp.org/ajcpcme.