Determination of the Optimal Procalcitonin Threshold for Infection in Patients With Impaired Renal Function at a Community Hospital

Abstract

Background: Procalcitonin (PCT) is a biomarker that can help identify bacterial infections but can be difficult to interpret in the presence of renal dysfunction, which can elevate PCT even in the absence of infection. Objective: To determine the optimal PCT threshold to identify infection in patients with renal dysfunction and pneumonia or sepsis. Methods: A retrospective analysis was performed for inpatients with creatinine clearance of ≤60 mL/min and PCT level from 2018 to 2019. A pharmacist blinded to study outcomes classified patients as infected or noninfected based on predetermined criteria. Receiver operating characteristic curve analysis was performed to establish the optimal PCT threshold overall, as well as in subgroups of patients with chronic kidney disease (CKD), acute kidney injury (AKI), and end-stage renal disease (ESRD). Institutional review board approval was obtained. Results: A total of 198 patients were included in the study (99 infected, 99 noninfected). The optimal threshold in the AKI, CKD, and ESRD subgroups was determined to be 1.5 ng/mL, 0.1 ng/mL, and 1.75 ng/mL, respectively. Conclusion: The results of this study show that PCT thresholds were specific to type of renal dysfunction. These results differ from the traditionally accepted PCT threshold of 0.5 ng/mL for sepsis and 0.25 mg/mL for pneumonia. Future studies should confirm the appropriate PCT threshold in ESRD and CKD patient populations.

Introduction

Procalcitonin (PCT) is a peptide precursor of calcitonin that is produced in the C cells of the thyroid gland and can help determine if a patient has a bacterial or nonbacterial infection for a number of disease states.¹ For noninfected patients, PCT levels are typically <0.1 ng/mL. When a bacterial infection is present, the PCT levels in these patients can rise rapidly to >0.25 ng/mL.²

PCT is often used in guiding antibiotic therapy in patients with pneumonia and sepsis.² The 2016 Infectious Disease Society of America guidelines for hospital-acquired and ventilator-associated pneumonia provide a weak recommendation to utilize PCT plus clinical criteria to guide discontinuation of antibiotics.³ Similarly, the 2016 Surviving Sepsis Campaign suggests that PCT can support shortening duration of antibiotics in patients with sepsis.⁴ Current literature typically recommends using PCT thresholds of 0.25 ng/mL and 0.5 ng/mL for pneumonia and sepsis, respectively.^4-6

Interpreting PCT levels in patients with comorbid disease states can be difficult, considering different disease states can elevate PCT levels without a bacterial infection being present, such as massive stress, medullary thyroid carcinoma, small cell lung cancer, among others.² Additionally, and most pertinent to this article, PCT levels among patients with renal dysfunction are often elevated, even in the absence of infection. Interestingly, the half-life of PCT is very similar in patients with and without renal dysfunction. Patients without renal dysfunction have a PCT half-life of around 28.9 hours, while those with renal dysfunction have a PCT half-life of around 33 hours.⁷ Despite the similar half-live of PCT in patients with and without renal dysfunction, noninfected patients with renal dysfunction have been shown to have PCT levels much higher than the established threshold of <0.1 ng/mL in healthy individuals. For example, in one study of noninfected patients with CKD, patients had an average PCT level of 1.82 ng/mL.⁸ In patients with renal dysfunction, it can be hard to use PCT as an indicator of infection due to their baseline PCT already being higher than the traditional thresholds used for pneumonia and sepsis.

Currently, there are very few studies that have evaluated the optimal threshold of PCT for diagnosis of infection in patients with renal dysfunction. The purpose of this study is to determine the most appropriate PCT threshold value to identify infection in patients with impaired renal function and concomitant pneumonia or sepsis.

Patients and Methods

Patients and Setting

A retrospective chart review was conducted on patients who were admitted to a 344-bed community hospital in the southeastern United States. Patients were included if they had a creatinine clearance of ≤60 mL/min and a documented PCT level. Patients were identified via electronic report of all PCT levels ordered between the advent of PCT testing at the community hospital (June 1, 2018) and report generation date (April 7, 2019). Exclusion criteria included those with decompensated liver cirrhosis (Child-Pugh Score ≥7), medullary thyroid cancer, congestive heart failure (HFrEF or HFpEF), suspected ischemic stroke, lupus exacerbations, patients seen in the emergency department who were not admitted, resuscitated cardiac arrest, burns >30% of body surface area, small cell lung cancer, or pregnant patients.^9-12 If the same patient was admitted multiple times, only the patient’s first admission during the study period was analyzed. PCT was analyzed in-house via bioMerieux VIDAS BRAHMS PCT assay, which detects PCT at levels of 0.05 to 200 ng/mL. At the time of data collection, the hospital utilized an institutional protocol of PCT measurement for septic patients, which included a baseline level on initiation of sepsis order set, followed by daily PCT levels times 3 days. However, providers could also order PCT levels at their discretion.

Data

The maximum PCT level during the patient’s admission was recorded. Serum creatinine (SCr) was used to calculate creatinine clearance for the day corresponding to the recorded maximum PCT level. Creatinine clearance was assessed via Cockcroft-Gault; total body weight was used for patients weighing less than ideal body weight, ideal body weight was used for patients weighing >100% but <120% of their ideal body weight, and adjusted body weight was used for patients weighing >120% of their ideal body weight. Patients were classified in 3 predetermined subgroups: acute kidney injury (AKI), chronic kidney disease (CKD), or end-stage renal disease (ESRD). AKI was defined as increase in SCr by ≥0.3 mg/dL within 48 hours, increase in SCr to ≥1.5 times baseline that is known or presumed to have occurred within the prior 7 days, or urine volume <0.5 mL/kg/h for 6 hours.¹³ CKD was defined as glomerular filtration rate (GFR) <60 mL/min/1.73 m² at baseline, corroborated with provider documentation in the patient’s history and physical. ESRD was defined as GFR <15 mL/min/1.73 m².¹¹

Patients were classified either as infected or noninfected. The classification was made by a pharmacist blinded to the patients’ PCT levels. A patient was classified as “infected” if he or she met one of the following characteristics:

1. Two or more systemic inflammatory response syndrome criteria PLUS a confirmed or suspected source of infection as determined by positive culture results or provider progress note documenting infection source. Blood cultures with 1 of 2 sets positive for coagulase negative Staphylococcus aureus were not considered positive.¹⁴

2. Imaging of lung describing any of the following key terms: consolidation, opacity, opacification, ground glass, patchy, infiltrate, pneumonia, infection, hazy; PLUS positive respiratory culture OR at least 2 of the following clinical features—temperature >100.4 °F, white blood cells count >12 000, or <4000, noted sputum production from provider progress note, noted cough or wheezing from physician progress note.^3,15

Additional information collected included the following: age, gender, race, height, weight, white blood cell count, heart rate, respiratory rate, blood pressure, lactate, culture data, comorbid conditions (chronic obstructive pulmonary disease, diabetes mellitus, atrial fibrillation, hypertension, and hyperlipidemia), intensive care unit or non–intensive care unit location in the hospital, hemodialysis status, length of hospital stay, antibiotics received during stay, and SCr level.

The primary endpoint was to determine the PCT threshold with optimal specificity and sensitivity for diagnosing infection of sepsis or pneumonia in patients with renal dysfunction. Secondary endpoints included comparison of average and median PCT levels, length of hospital stay, antibiotic duration, and mortality rate in infected versus noninfected patients. A preplanned subgroup analysis was performed to determine the optimal threshold in patients with CKD, AKI, and ESRD separately.

Statistical Analysis

Receiver operating characteristic curve (ROC) analysis and Youden Index was performed to establish the optimal PCT threshold. Parametric and nonparametric continuous data were assessed using Student’s t test and Mann-Whitney U test, respectively. Categorical variables were assessed via χ² test. Microsoft Excel was used to perform the statistical analysis.

The study was conducted in compliance with the requirements of the study site’s institutional review board committee.

Results

Four hundred and ninety-six charts were reviewed, and a total of 198 patients were included in the study, with 99 in the infected group and 99 in the noninfected group. Two hundred and ninety-eight patients were excluded from the study, with the main reason for exclusion being lack of renal dysfunction. The breakdown of the patients between the 2 groups is shown in Supplemental Figure 1 (available online). Baseline patient demographics were similar between the 2 groups and are presented in Table 1.

Figure 1.

Receiver operating curve (ROC) analysis. Area under the ROC curve overall (0.736), AKI (0.797), CKD (0.683), and ESRD (0.714). AKI, acute kidney injury; CKD, chronic kidney disease; ESRD, end-stage renal disease.

Table 1.

Baseline Demographics^a.

	Infected group (n = 99)	Noninfected group (n = 99)	P
Average age, years (SD)	71.61 ± 14.04	72.98 ± 15.08	.511
Gender, % male	42%	37%	.468
Race, % African American	53%	52%	.887
Race, % Caucasian	47%	48%	.887
Average weight, kg (SD)	78.60 ± 23.46	76.42 ± 20.75	.489
Average height, in (SD)	66.54 ± 4.37	65.97 ± 4.38	.365
COPD, %	20%	14%	.300
Diabetes, %	42%	43%	.886
Atrial fibrillation, %	13%	14%	.841
Hypertension, %	79%	78%	.881
Hyperlipidemia, %	29%	29%	.938
AKI, %	49%	29%	.024
CKD, %	35%	54%	.055
ESRD, %	15%	17%	.724
Dialysis, %	15%	17%	.716
Hemodialysis, %	13%	16%	.577
Peritoneal dialysis, %	2%	1%	.563
Average serum creatinine, mg/dL (SD)	2.64 ± 2.74	2.59 ± 2.40	.884

Abbreviations: SD, standard deviation; COPD, chronic obstructive pulmonary disease; AKI, acute kidney injury; CKD, chronic kidney disease; ESRD, end-stage renal disease.

^aAll data are reported as either mean ± SD or n (%).

Based on the results from this study, the threshold with the optimal sensitivity and specificity for identification of infection was determined to be 1.0 ng/mL. The thresholds with the optimal sensitivity and specificity in the AKI, CKD, and ESRD subgroups were determined to be 1.5 ng/mL, 0.1 ng/mL, and 1.75 ng/mL, respectively. The sensitivities, specificities, negative predictive values, and positive predictive values are presented in Table 2. The areas under the curves (AUC) were calculated to be 0.763, 0.797, 0.683, and 0.714 for overall, AKI, CKD, and ESRD, respectively. The ROC curves for overall and respective subgroups are presented in Figure 1. Average and median PCT level, average length of hospital stay, average antibiotic duration, and mortality were all found to be significantly higher in the infected group versus the noninfected group. Secondary endpoints are shown in Table 2.

Table 2.

Primary and Secondary Outcomes^a.

Sensitivity, specificity, NPV, and PPV
	PCT thresholdb	Sensitivity	Specificity	NPV	PPV
Overall cohort (n = 198)	1.0 ng/mL	59%	76%	65%	71%
AKI subgroup (n = 78)	1.5 ng/mL	59%	86%	56%	88%
CKD subgroup (n = 89)	0.1 ng/mL	66%	64%	75%	55%
ESRD subgroup (n=32)	1.75 ng/mL	80%	65%	79%	67%
Median PCT levels and clinical outcomes
	Infected group (n = 99)		Noninfected group (n = 99)		P
Overall median PCT, ng/dL (IQR)	1.46 (0.18-13.29)		0.12 (0.05-0.97)		<.00001
AKI median PCT, ng/dL (IQR)	2.63 (0.54-21.45)		0.29 (0.05-1.03)		<.00001
CKD median PCT, ng/dL (IQR)	0.17 (0.05-1.35)		0.05 (0.05-0.21)		.0056
ESRD median PCT, ng/dL (IQR)	4.63 (1.93-14.11)		1.55 (0.65-3.4)		.02574
Average length of hospital stay, days (SD)	10.9 ± 11.1		5.9 ± 4.3		<.00001
Average antibiotic duration, days (SD)	9.1 ± 8.5		4.5 ± 3.8		<.00001
Mortalityc, %	9%		2%		.03

Abbreviations: NPV, negative predictive value; PPV, positive predictive value; PCT, procalcitonin; AKI, acute kidney injury; CKD, chronic kidney disease; ESRD, end-stage renal disease; IQR, interquartile range; SD, standard deviation.

^aSecondary outcomes, with the exception of mortality, are reported as either mean ± standard deviation or median (interquartile range).

^bPCT threshold identified in study with optimal sensitivity and specificity to identify infection, based on Youden index.

^cMortality is reported as percentage.

Discussion

Summary of Key Findings

PCT thresholds in this study were specific to type of renal dysfunction and were higher than the traditionally utilized PCT thresholds of 0.25 ng/mL for pneumonia and 0.5 ng/mL for sepsis. The higher threshold in our study was largely driven by the ESRD and AKI subgroups, which yielded optimal thresholds of 1.75 and 1.5 ng/mL, respectively, while CKD yielded a lower threshold of 0.1 ng/mL. Based on ROC analysis, calculated AUC for the overall threshold of 1.0 ng/mL was 0.763, suggesting a fair level of accuracy. The lowest AUC, 0.683, was for the CKD subgroup, which suggests a poor level of accuracy.

Procalcitonin Thresholds in Prior Literature

Other studies have identified a variety of different PCT thresholds for patients with renal dysfunction, ranging from 0.5 to 3.2 ng/mL. Table 3 below compares the thresholds and their associated sensitivity, specificity, and AUC, with our findings. Of note, however, prior literature often used different criteria for infection as well as renal dysfunction. Additionally, some literature did not report their method for determining a patient’s infection status. For example, in Park et al’s study, infection was defined as a clinical definable source of infection confirmed by microbiology tests and/or positive blood cultures.¹⁶ Lu et al and Lee et al did not define the criteria for diagnosis of infection.^1,17 Diagnosis of infection in El-Sayed et al was defined as presence of an organism on a culture that was clinically relevant. El-Sayed et al had the closest criteria for infection to this study by having a third party blinded to PCT levels determining whether the patient was considered as infected or noninfected.¹⁸ While these prior studies did analyze PCT thresholds in AKI, CKD, and ESRD patients, none of the prior literature examined these subgroups within the same study. Sensitivity, specificity, and AUC vary greatly in prior literature, with AUC ranging from 0.670 to 0.910, sensitivity from 53% to 87%, and specificity from 64% to 94%. Our results fell within the lower end of this range, largely driven by the low accuracy of PCT in our CKD sub-group. The large difference in findings, in the literature, suggests that PCT thresholds may depend on the unique patient population at a given institution, and a global PCT threshold may not be applicable in patients with renal dysfunction.

Table 3.

PCT Thresholds in Renal Dysfunction From Prior Literature.

	Number of patients	Definition of renal dysfunction	Threshold	Sensitivity	Specificity	AUC
Park et al16	493	eGFR <60 mL/min/1.73 m2	1.1 ng/mL	73%	89%	0.867
Lu et al1	803	N/A	0.5 ng/mL	68%	94%	0.82
El-Sayed et al18	473	eGFR ≤30 mL/min/1.73 m2	3.2 ng/mL	53%	75%	0.67
Lee et al17	41	ESRD on dialysis (either hemodialysis or peritoneal dialysis)	0.75 ng/mL	76.2%	80%	N/A
Nakamura et al19	806	KDIGO AKI guidelines	Stage 1: 0.74 ng/mL	83.8%	91.6%	0.910
			Stage 2: 0.59 ng/mL	75%	88.9%	0.867
			Stage 3: 4.07 ng/mL	87.2%	93.5%	0.946
Bowman et al (current study)	198	Overall: CrCl ≤60 mL/min calculated by Cockcroft-Gault	Overall: 1.0 ng/mL	59%	76%	0.736
		AKI: KDIGO AKI guidelines	AKI: 1.5 ng/mL	59%	86%	0.797
		CKD: GFR <60 mL/min/1.73 m2 at baseline, corroborated with provider documentation in the patient’s history and physical	CKD: 0.1 ng/mL	66%	64%	0.683
		ESRD: GFR <15 mL/min/1.73 m2	ESRD: 1.5 ng/mL	80%	65%	0.714

Abbreviations: PCT, procalcitonin; AUC, area under the curve; eGFR, estimated glomerular filtration rate; N/A, not applicable; ESRD, end-stage renal disease; KDIGO, Kidney Disease Improving Global Outcomes; AKI, acute kidney injury; CrCl, creatinine clearance; CKD, chronic kidney disease.

Average PCT Levels in Prior Literature

Prior literature has suggested that noninfected patients with renal dysfunction will have elevated PCT levels. PCT levels in patients with CKD have been shown to have mean values of 1.82 ng/mL or more without infection present. In noninfected patients undergoing peritoneal dialysis, the PCT level can be elevated to 0.40 ng/mL. In noninfected patients undergoing hemodialysis, the PCT level can be elevated up to 0.55 ng/mL.² Similarly, our study found elevated PCT levels in noninfected patients, with an average PCT of 2.47 ng/mL. However, when assessing the median to account for outliers, the median PCT level was just 0.12 ng/mL in noninfected patients. Nonetheless, noninfected patients in the subgroups with AKI and ESRD did demonstrate elevated median PCT levels of 0.29 and 1.55 ng/mL, suggesting again that the higher threshold found in our study was largely driven by the AKI and ESRD patient subgroups, rather than CKD. Interestingly, infected patients with CKD had a median PCT level of just 0.17 ng/mL, which is below even the traditionally accepted PCT thresholds for pneumonia of 0.25 ng/mL. Along with an AUC demonstrating poor accuracy in the CKD subgroup, we determined that PCT cannot be used to accurately rule out or rule in infection in the CKD population.

Strengths and Limitations

A strength of this study included blinded evaluation of the patients, while determining whether the patient was infected or noninfected. Nonetheless, we would have preferred to have additional entities confirm the determination of infection. Additionally, the study provided a real-world insight into PCT levels in patients at a community hospital in the Southeast. Limitations of the study include the range of PCT detection from 0.05 to 200 ng/mL; any value <0.05 ng/mL was recorded as 0.05 ng/mL and anything >200 ng/mL was recorded as 200 ng/mL when performing the statistical analysis. Additionally, while the hospital had a protocol in place for PCT measurement during the study (baseline level at suspected infection, daily levels times 3 days), providers could order levels at their discretion. We had a small number of patients in each subgroup, particularly ESRD, so the subgroup analyses should be assessed accordingly. As such, we did not categorize the AKI subgroup based on stage of AKI, as was done in a prior study,¹⁹ due to the concern of small sample size within each stage of AKI leading to potential uninterpretable data.

Future studies are needed on a larger scale to determine the utility of PCT as a biomarker in patients with renal dysfunction, particularly with such wide variability in PCT levels and thresholds reported across the literature. In particular, the CKD population should be evaluated, as our study found PCT to possess poor accuracy in identifying infection in this patient population.

Conclusion

The results of this study show that the PCT thresholds were specific to type of renal dysfunction and that no acceptable cutoff was found for the CKD subgroup. Our results differ from the currently accepted PCT threshold of 0.5 ng/mL for sepsis and 0.25 mg/mL for pneumonia used for the general population. Future studies should confirm the appropriate PCT thresholds in the AKI, ESRD, and CKD patient populations.