Determining the Clinical Utility of an Absolute Procalcitonin Value for Predicting a Positive Culture Result

ABSTRACT

Various procalcitonin ranges have been established to guide antimicrobial therapy; however, there are no data that establish whether the initial procalcitonin value can determine the likelihood of a positive culture result. This study aimed to establish if the initial procalcitonin value, on clinical presentation, has a positive predictive value for any positive culture result. This was a retrospective study of 813 medical intensive care unit patients. Data collected included patient demographics, procalcitonin assay results, sources of infection, culture results, and lengths of stay. Patients were excluded if they were immunocompromised. The primary outcome of this study was to determine a procalcitonin value that would predict any positive culture. Secondary outcomes included the sensitivity, specificity, positive predictive value, and negative predictive value for procalcitonin. After exclusions, a total of 519 patient charts were reviewed to determine the impact of the initial procalcitonin value on culture positivity. In our analyses, the receiver operating characteristic values were 0.62 for all cultures, 0.49 for pulmonary infections, 0.43 for urinary tract infections, and 0.78 for bacteremia. A procalcitonin value of 3.61 ng/ml was determined to be the threshold value for a positive blood culture result (prevalence, 4%). For bacteremia, the sensitivity of procalcitonin was 75%, the specificity was 72%, the positive predictive value was 20%, and the negative predictive value was 97%. Procalcitonin was a poor predictor of culture positivity. An initial procalcitonin value of less than 3.61 ng/ml may be useful in predicting whether bacteremia is absent. Procalcitonin should not be used as the only predictor for determining initiation of antibiotic therapy.

INTRODUCTION

Sepsis is a leading cause of death and disability in the United States (1, 2). Mortality from sepsis increases with each hour that anti-infective therapy is delayed (3). The impetus for promptly initiating appropriate broad-spectrum antimicrobial therapy must be weighed against the concern for reducing unnecessary antimicrobial therapy and potential harms, such as drug resistance, Clostridium difficile infections, and adverse drug effects (3). The lack of a specific test to assist clinicians in early decision-making regarding initiation and discontinuation of antimicrobial therapy for severe infection and presumed sepsis has led to expanded interest in inflammatory biomarkers, such as procalcitonin (PCT) (4–12).

Previous trials evaluating PCT have used value ranges to guide clinical decisions within the construct of initiating and discontinuing antimicrobial therapy using clinical criteria (4, 5, 7–10, 12–14). By using serial measurements and established cutoff values in intensive care units (ICUs), these studies have shown that PCT values guide antimicrobial therapy (13–18). However, no reports have described the utility of using initial PCT levels to guide upstream anti-infective therapy decision-making or the predictive ability of PCT values in the context of objective criteria of infection, such as culture positivity. Such data would provide clinicians with greater confidence in presuming the presence or absence of an infection when determining initial therapy. This study aimed to establish if a single initial PCT value can predict the result of a positive culture or the source of infection. The primary outcome of this study was to determine a PCT value that would predict any positive culture. Secondary outcomes included the sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of PCT.

RESULTS

Over the 23-month study period, 813 patients met the inclusion criteria (Fig. 1). Of these, 294 patients were excluded due to their immunocompromised status. The remaining 519 patients (48% female) were included for study analyses. The median age was 59 years (interquartile range [IQR], 48–71 years). Pulmonary infections were the most common (43%) among patients in the study.

FIG 1.

Flow chart depicting patient enrollment and culture results.

Of the different infection types analyzed (Fig. 2), bacteremia had the greatest area under the receiver operating characteristic (ROC) curve (AUC) value (0.78 [95% confidence interval (CI), 0.72–0.84]; P < 0.001) with a threshold PCT level of 3.61 ng/ml in determining culture positivity. Using our prevalence of bacteremia (4%), PCT testing had the following diagnostic characteristics: 75% sensitivity, 72% specificity, 20% PPV, and 97% NPV.

FIG 2.

ROC curve analysis: (a) all cultures; (b) bacteremia; (c) pulmonary infections; (d) urinary tract infections.

Patients with a positive culture from any source had a higher PCT level and acute physiology and chronic health evaluation II (APACHE II) score and a longer ICU/hospital length of stay (Table 1) than those with a negative culture. Although PCT levels performed poorly for predicting culture positivity in the setting of a pulmonary infection (AUC, 0.49 [95% CI, 0.40–0.56]; P = 0.76), respiratory culture positivity was associated with prolonged ICU/hospital stays and higher PCT levels (Table 2).

TABLE 1.

Comparison of culture-negative and culture-positive patient data (n = 519)

Patient characteristica	Value for group (median [IQR])b		P
	Culture negative	Culture positive
Age (yr)	61 (47–73)	57 (49–67)	0.25
ICU LOS (days)	4 (3–7)	6 (3–12)	<0.001
Hospital LOS (days)	8 (5–15)	13 (8–24)	<0.001
APACHE II score	20 (15–26)	23 (17–27)	0.01
PCT value (ng/ml)	0.74 (0.17–4.36)	2.12 (0.57–11.42)	<0.001

^aICU, intensive care unit; LOS, length of stay; APACHE II, acute physiology and chronic health evaluation II; PCT, procalcitonin.

^bIQR, interquartile range.

TABLE 2.

Comparison of culture-negative and culture-positive pulmonary patient data (n = 222)

Patient characteristica	Value for group (median [IQR])b		P
	Culture negative	Culture positive
Age (yr)	65 (50–78)	56 (50–67)	0.27
ICU LOS (days)	3 (2–7)	6 (3–15)	<0.001
Hospital LOS (days)	8 (5–15)	11 (7–21)	0.02
APACHE II score	19 (15–26)	23 (16–27)	0.09
PCT value (ng/ml)	0.74 (0.17–3.41)	2.07 (0.32–5.36)	0.03

^aICU, intensive care unit; LOS, length of stay; APACHE II, acute physiology and chronic health evaluation II; PCT, procalcitonin.

^bIQR, interquartile range.

DISCUSSION

This study highlights the potential role of initial PCT testing in predicting culture positivity from various sources of infection. The results suggest that PCT levels are not a useful biomarker for predicting culture positivity for infection from any source. However, in patients with bacteremia, a PCT value greater than 3.61 ng/ml appeared to be associated with blood culture positivity. The NPV (97%) for bacteremia would be useful in patients with a PCT value less than 3.61 ng/ml, which would be indicative of a low likelihood of a positive blood culture. This PCT level threshold might be an acceptable biomarker for future studies as a potential basis for withholding antibiotic treatment in patients with possible bacteremia until culture results are available, although this was not proven directly in our study. For infections from a pulmonary or urinary source (56% of infections studied in our population), PCT levels resulted in poor AUC ROC values (0.49 and 0.43, respectively) and therefore should not be used during the initial evaluation to determine initiation of antibiotic therapy for these types of infections.

In an era of escalating antimicrobial resistance and with a reported incidence of 30 to 50% inappropriate antimicrobial therapy (19–22), there is growing need to study biomarkers which might improve antimicrobial stewardship, especially during the initial patient care evaluation. A robust evidence-based approach to studying PCT in clinical care is currently lacking, as demonstrated by the grade 2C recommendation in the recent Surviving Sepsis Campaign guidelines (23). Two previous large clinical trials reported that measurement of serial PCT values resulted in fewer overall antibiotic days without a difference in number of deaths (13, 24). More recently, the Stop Antibiotics on Guidance of Prolactin Study (SAPS) confirmed that measuring PCT can help reduce the duration of antibiotic therapy, and this reduction is associated with a significant decrease in mortality (25). However, these trials have used PCT serially to determine duration of antimicrobial therapy. No studies preceding ours have examined the upstream utility of a single PCT measurement during the initial patient care evaluation in the context of culture positivity, which is a critical area for antimicrobial stewardship improvement.

Prior studies have compared PCT to other biomarkers, predominantly C-reactive protein (26–28). These studies have demonstrated a consensus that PCT may have a higher prognostic value for septic/bacteremic patients than does C-reactive protein (26–28). However, these other biomarkers are relatively inconsistent determinants for other sources of infection, with studies showing mixed results (26–28). Furthermore, no direct comparisons have been made between PCT levels and the clinical criteria used to determine the risk/severity of infections (e.g., systemic inflammatory response syndrome [SIRS] criteria). Our study serves to confirm prior evaluations regarding the utility of PCT in bacteremic patients and cautions against its use in determining initial antimicrobial therapy for other types of infection. Future areas of interest might be the study of biomarkers in comparison with rapid identification tests, such as those from BioFire (Salt Lake City, UT) and PNA FISH (Woburn, MA). Due to the costs associated with these rapid identification systems, their widespread use remains limited at this time.

There are several limitations of this study. In this retrospective design, our population demonstrated a variability in infectious diagnoses, with skewing toward pulmonary infections (43%) and a relatively smaller population of patients with bacteremia (4%). Additionally, although PCT levels were measured on the first day of admission to our tertiary care center, patients may have received antimicrobial therapy prior to arrival (i.e., in the emergency department or prior to transfer from an outside institution). Nevertheless, the purpose of our study was to examine PCT values within the context of arrival to our medical ICU regardless of antibiotic use in the emergency department or an outside institution. This limitation may have been mitigated by early obtainment of culture and PCT levels within the first 24 h of admission to our institution's ICU. It is important to note that PCT results are available within 24 h at our institution. Since culture results are not available until after 48 h, PCT values are available before culture results. The PCT level increases within the first 2 to 4 h of a triggering event, peaks within 12 to 24 h, and has a half-life that varies between 24 and 35 h (29–32). Therefore, in our cohort, early antibiotic use was less likely to impact PCT values in patients admitted by our institution's emergency department but may have had a greater impact on transferring patients, for whom the duration to ICU admission was longer.

Conclusion.

Initial PCT results were poor predictors of culture positivity. However, an initial PCT level less than 3.61 ng/ml might be acceptable for predicting whether bacteremia is absent. This is supported by a negative predictive value of 97%. Routine use of PCT level in conjunction with clinical judgment can be useful for bacteremic patients, but our study cautions against its use to determine initial antimicrobial therapy for other sources of infection, particularly pulmonary and urinary sources.

MATERIALS AND METHODS

Population.

Institutional review board approval was granted from Rush University Medical Center for this retrospective cohort study using an existing database. Data from April 2013 to December 2014 were collected for patients admitted to the medical ICU at Rush University Medical Center. Patients were included if they were greater than 18 years of age and had a suspected infection as their admitting diagnosis and an associated bacterial culture, and their first PCT value was obtained within 24 h upon admission to the medical ICU. Patients were excluded if they were immunocompromised, which was defined as receiving immunosuppressive therapy (6-mercaptopurine, prednisone equivalent to greater than 5 mg/day, antithymocyte globulin, azathioprine, basiliximab, cyclophosphamide, cyclosporine, calcineurin inhibitors, everolimus, methotrexate, mycophenolate, rituximab, or sirolimus) or having a diagnosis of HIV or active cancer.

Data collected included age, gender, acute physiology and chronic health evaluation II (APACHE-II) score, first PCT value, ICU length of stay (LOS), hospital LOS, suspected and confirmed source of infection, culture result, and type of infection.

Statistical analysis.

The ability of PCT to predict a positive culture was evaluated by using receiver operating characteristic (ROC) curve methodology. The best cutoff value of PCT for a positive culture result was determined according to Youden's index method. The areas under the ROC curves (AUC) were reported with 95% confidence intervals (CIs) in addition to prevalence, sensitivity, specificity, PPV, and NPV. Nonparametric statistical analyses (Mann-Whitney U test) were applied to compare culture-positive and culture-negative data among various baseline characteristics. Values were reported as medians with interquartile ranges (IQR). Statistical analyses were conducted using the Statistical Program for Social Science (SPSS version 23, Chicago, IL, USA). Statistical significance was defined as a P value of <0.05.