Risk Prediction With Procalcitonin and Clinical Rules in Community-Acquired Pneumonia
Article Outline
- Abstract
- Introduction
- Materials and Methods
- Results
- Limitations
- Discussion
- Acknowledgment
- Appendix
- References
- Copyright
Study objective
The Pneumonia Severity Index and CURB-65 predict outcomes in community-acquired pneumonia but have limitations. Procalcitonin, a biomarker of bacterial infection, may provide prognostic information in community-acquired pneumonia. Our objective is to describe the pattern of procalcitonin in community-acquired pneumonia and determine whether procalcitonin provides prognostic information beyond the Pneumonia Severity Index and CURB-65.
Methods
We conducted a multicenter prospective cohort study in 28 community and teaching emergency departments. Patients presenting with a clinical and radiographic diagnosis of community-acquired pneumonia were enrolled. We stratified procalcitonin levels a priori into 4 tiers: I: less than 0.1; II: greater than 0.1 to less than 0.25; III: greater than 0.25 to less than 0.5; and IV: greater than 0.5 ng/mL. Primary outcome was 30-day mortality.
Results
One thousand six hundred fifty-one patients formed the study cohort. Procalcitonin levels were broadly spread across tiers: 32.8% (I), 21.6% (II), 10.2% (III), and 35.4% (IV). Used alone, procalcitonin had modest test characteristics: specificity (35%), sensitivity (92%), positive likelihood ratio (1.41), and negative likelihood ratio (0.22). Adding procalcitonin to the Pneumonia Severity Index in all subjects minimally improved performance. Adding procalcitonin to low-risk Pneumonia Severity Index subjects (classes I to III) provided no additional information. However, subjects in procalcitonin tier I had low 30-day mortality, regardless of clinical risk, including those in higher risk classes (1.5% versus 1.6% for those in Pneumonia Severity Index classes I to III versus classes IV/V). Among high-risk Pneumonia Severity Index subjects (classes IV/V), one quarter (126/546) were in procalcitonin tier I, and the negative likelihood ratio of procalcitonin tier I was 0.09. Procalcitonin tier I was also associated with lower burden of other adverse outcomes. Similar results were observed with CURB-65 stratification.
Conclusion
Selective use of procalcitonin as an adjunct to existing rules may offer additional prognostic information in high-risk patients.
SEE EDITORIAL, P. 61.
Introduction
Background
Community-acquired pneumonia accounts for 1.3 million hospitalizations in the United States each year1 at a cost of $8.4 billion.2 It is the most common cause of severe sepsis3 and infection-related death.4 Key to the safe and efficient management of community-acquired pneumonia is the ability to reliably predict who will fare well or poorly. The Pneumonia Severity Index5 and CURB-65 (Confusion, Uremia, Respiratory rate, low blood pressure, age 65 years or older)6 are clinical rules that identify a subset of individuals at low risk of death who are candidates for outpatient care.7, 8 However, all remaining patients are classified as high risk, usually prompting hospital admission and parenteral antibiotics, even though a large proportion may do well.9 Thus, there has been considerable interest in the development of rapidly available biomarkers that might confer additional prognostic information.10
What is already known on this topic
Procalcitonin is a biomarker that appears to correlate with bacterial infection.
What question this study addressed
Does a procalcitonin level add prognostic information for pneumonia patients in conjunction with scoring systems such as the Pneumonia Severity Index or CURB-65?
What this study adds to our knowledge
Among 1,651 patients with community-acquired pneumonia in 28 US emergency departments, procalcitonin levels did not add prognostic information for most pneumonia patients. Among higher-risk groups by Pneumonia Severity Index score, low procalcitonin level predicted lower mortality.
How this might change clinical practice
Clinicians should continue using validated prognostic scoring systems for pneumonia. Low procalcitonin level could be considered as a factor for selected patients who would otherwise be considered high risk to be treated in a lower acuity setting.
Importance
Procalcitonin is a calcitonin precursor that is generally increased in bacterial infections but low in viral infections.11 Procalcitonin has good discrimination for bacterial infections and sepsis,12, 13, 14, 15 and 3 trials used low procalcitonin levels to withhold antibiotics in emergency department (ED) patients presenting with respiratory illnesses.16, 17, 18 However, 2 recent meta-analyses concluded that procalcitonin could not reliably differentiate sepsis from noninfectious inflammation in critically ill patients19 and had only moderate diagnostic performance for identifying bacteremia in ED patients.20 Furthermore, the prognostic value of procalcitonin measurement beyond existing prediction rules is unclear. Masia et al21 observed that patients with high Pneumonia Severity Index scores had higher procalcitonin levels and that higher concentrations were associated with mortality and complications,21 but Beovic et al23 found no association between procalcitonin and Pneumonia Severity Index score.22 These single center studies were limited by small sample sizes and used older procalcitonin assays with low sensitivity.23
Goals of This Investigation
Our goal was to determine the prognostic utility of a newer, high-sensitivity procalcitonin assay for 30-day mortality and assess its value beyond established clinical prediction rules. We tested this assay within a multicenter, prospective cohort of patients presenting to the ED with a clinical and radiographic diagnosis of community-acquired pneumonia. We hypothesized that an early singular procalcitonin measurement would aid risk assessment beyond that available from the Pneumonia Severity Index and CURB-65.
Materials and Methods
Study Design and Setting
We conducted a multicenter, prospective, cohort study of patients presenting to the EDs of 28 teaching and nonteaching hospitals in southwestern Pennsylvania, Connecticut, southern Michigan, and western Tennessee between November 2001 and November 2003 (Genetic and Inflammatory Markers of Sepsis study). A specific aim of the Genetic and Inflammatory Markers of Sepsis study was to develop and validate risk prediction tools according to information available early in the course of disease. As part of this aim, we sought to determine the prognostic utility of procalcitonin for 30-day mortality.
Selection of Participants
Eligible subjects were older than 18 years and had a clinical and radiologic diagnosis of pneumonia according to Fine et al.5 We excluded those transferred from another hospital, discharged from a hospital within the previous 10 days, with an episode of pneumonia within the past 30 days, receiving chronic mechanical ventilation, with cystic fibrosis, with active pulmonary tuberculosis, with a known positive HIV antibody titer, having alcoholism with evidence of end-organ damage, admitted for palliative care, enrolled previously in the Genetic and Inflammatory Markers of Sepsis Study, incarcerated, and who were pregnant. We obtained informed consent from the subject or proxy. The institutional review boards of the University of Pittsburgh and all participating sites approved the study.
Data Collection and Processing
We gathered baseline and sequential clinical information by structured patient or proxy interviews, bedside assessment by study nurses, and structured medical record reviews. Median time from ED admission to day 1 blood sample collection was 1.3 hours. We did not obtain day 1 samples from patients presenting after 11 pm or on weekends and holidays for logistic reasons. Study personnel collected blood sample into pyrogen-free vials containing heparin and separated plasma by centrifugation within 1 hour. Plasma was frozen and shipped on dry ice to our central laboratory in Pittsburgh. We tracked clinical data and blood samples with unique anonymized identification numbers, merging data only after assay completion. We observed strict data confidentiality and audited clinical data and assays for accuracy, including random chart audits, repeated blood assays, and computer flags for inconsistencies.
Methods of Measurement
We measured procalcitonin with a time-resolved, amplified cryptate emission assay (Kryptor PCT; Brahms, Hennigsdorf, Germany). The assay has a functional assay sensitivity of 0.06 ng/mL, and at 0.1 ng/mL the coefficient of variation is 10% to 15%.24 Study nurses ascertained deaths in hospital. Postdischarge deaths were ascertained by telephone and National Death Index search. We enrolled subjects between November 2001 and November 2003, locked clinical data in 2004, completed assays in 2005, and petitioned complete National Death Index data when they became available in 2006.
We prospectively assessed severity of illness with the Pneumonia Severity Index.5 We calculated CURB-65 retrospectively with altered mental status or a new change in Glasgow Coma Scale score as proxy measures for confusion.25 According to previous studies, we stratified procalcitonin into 4 tiers: tier I: less than 0.1; tier II: greater than 0.1 to less than 0.25; tier III: greater than 0.25 to less than 0.5; and tier IV: greater than 0.5 ng/mL.16, 17, 18 We defined a clinically significant positive culture result according to published guidelines and previous literature.26, 27, 28, 29, 30, 31, 32, 33 For example, single blood cultures that yielded coagulase-negative staphylococci and sputum cultures that yielded normal oral flora were not counted as clinically significant. We defined severe sepsis as infection plus acute organ dysfunction, following international consensus criteria.34 Acute organ dysfunction was defined as a new Sequential Organ Failure Assessment score of greater than 2 in any of 6 organ systems.35
Outcome Measures
Our primary outcome was 30-day mortality, the traditional endpoint used for clinical prediction rules in community-acquired pneumonia, including the Pneumonia Severity Index and CURB-65. Secondary outcomes included 90-day mortality, length of stay, and ICU admission.
Primary Data Analysis
We generated descriptive data, comparing initial presentation and outcome measures across procalcitonin tiers. To test for trends across procalcitonin tiers for ordinal variables, we used the Jonckheere-Terpstra trend test. To test for differences across procalcitonin tiers for continuous variables, we used the Kruskal-Wallis test.
To understand the prognostic utility of procalcitonin and the 2 clinical prediction rules as standalone tests, we first generated Kaplan-Meier plots of 30-day mortality by category (tier, class, or group). We then generated test characteristics (sensitivity, specificity, negative and positive likelihood ratios) for each test, dichotomized into low and high risk. A positive likelihood ratio greater than 10 and a negative likelihood ratio less than 0.1 may be considered to provide strong evidence to rule in or rule out the condition of interest.36 We defined low risk as classes I to III for Pneumonia Severity Index, group 1 for CURB-65, and tier I (<0.1 ng/mL) for procalcitonin, according to previous criteria.5, 18, 37
To understand the value of adding procalcitonin to the Pneumonia Severity Index, we first generated 30-day mortality Kaplan-Meier plots, stratified by procalcitonin tier within each Pneumonia Severity Index class. Second, we assessed the change in test characteristics for a logistic regression model in which procalcitonin was added to a model with Pneumonia Severity Index alone. This approach assumes both “tests” (Pneumonia Severity Index and procalcitonin) are performed in all patients. We derived and validated models with a 3:1 random split of the overall cohort, fitting with all possible combinations of Pneumonia Severity Index, procalcitonin, and the corresponding interaction term. These models were also fit with the Pneumonia Severity Index and procalcitonin treated as categorical variables. Third, because one could selectively order procalcitonin measurement, depending on the clinical risk assessment, we assessed the test characteristics of procalcitonin within the clinical low- and high-risk strata. We used the same approach to determine the prognostic value of adding procalcitonin to CURB-65. Analyses were performed with SAS 9.1 (SAS Institute, Inc., Cary, NC).
Sensitivity Analyses
To complete sensitivity analyses, we explored the effects of reclassifying Pneumonia Severity Index classes I to III subjects as high risk if they were hypoxic5 and excluding patients who were subsequently ruled out for community-acquired pneumonia during their hospitalization. Because of the large number of variables for the analysis stratifying and comparing the baseline presentation of high-risk patients by procalcitonin tier I versus tiers II to IV, we chose a significance level of P<.01. For all other analyses, we assumed significance at P<.05.
Results
Characteristics of Study Subjects
Of the 2,320 subjects enrolled in the Genetic and Inflammatory Markers of Sepsis study, 1,651 (71.2%) had a day 1 procalcitonin test and formed the study cohort. Pneumonia Severity Index was measured in the ED in 1,384 (83.8%) of the 1,651 study cohort subjects (Figure 1). The study cohort was predominantly white, underlying disease was common, approximately half were identified as high risk by the Pneumonia Severity Index and CURB-65 clinical risk rules, and most were admitted to the hospital. Few subjects identified in the ED as having community-acquired pneumonia were subsequently “ruled out for community-acquired pneumonia” by the inpatient clinical team (Table 1). Severe sepsis developed in one quarter, and 30-day mortality was 6.4% overall (Table 2) and 15.0% and 11.3% in those labeled as high risk by Pneumonia Severity Index and CURB-65, respectively (Table E1, Table E2; available online at http://www.annemergmed.com).
Figure 1.
Flow diagram of study. Gen IMS, Genetic and Inflammatory Markers of Sepsis study.
Table 1. Study cohort demographic and clinical characteristics (n=1,651).
| Variable | Value |
|---|---|
| Age, y, mean (SD) | 65.0 (18.5) |
| Sex, male, No. (%) | 860(52) |
| Race, No. (%) | |
| White | 1,336(81) |
| Black | 227(14) |
| Other | 88(5) |
| Previous antibiotics, No. (%) | 271(16) |
| Pre-enrollment duration of symptoms, mean (SD), median | 4.8(14.5),3.0 |
| Charlson comorbidity index | |
| Mean (SD), median | 1.69(2.12),1.0 |
| Index >0, No. (%) | 1,094(66%) |
| PSI⁎ | |
| Mean (SD), median | 82.0(34.3),80.5 |
| Class, No. (%) | |
| I, II | 541(39) |
| III | 297(21) |
| IV | 419(30) |
| V | 127(9) |
| CURB-65 | |
| Mean (SD), median | 1.57(1.19),1.0 |
| Group, No. (%) | |
| 1 | 826(50) |
| 2 | 421(25) |
| 3 | 404(24) |
| ED discharges, No. (%) | 265(16) |
| Ruled out for CAP, No. (%) | 94(6) |
⁎PSI was measured in the ED in 1,384 (83.8%) subjects. There were no significant differences between subjects who did and did not have a PSI measured. |
Table 2. Initial presentation and outcomes, by procalcitonin tier.
| No. (%) | P Value⁎ | All | Procalcitonin Tier (ng/mL) | |||
|---|---|---|---|---|---|---|
| I (<0.1) | II (≥0.1, <0.25) | III (≥0.25, <0.5) | IV (≥0.5) | |||
| Initial presentation | 1,651(100) | 542(32.8) | 356(21.6) | 169(10.2) | 584(35.4) | |
| SIRS criteria, No. (%) | ||||||
| SIRS by WBC | .0018 | 1,475(96) | 449(93) | 317(96) | 157(97) | 552(97) |
| SIRS by temperature | <.0001 | 766(47) | 155(29) | 169(48) | 87(52) | 355(61) |
| SIRS by respiratory | .1292 | 1,605(97) | 532(99) | 341(96) | 166(99) | 566(97) |
| SIRS by pulse rate | <.0001 | 1,138(76) | 312(67) | 239(73) | 116(74) | 471(85) |
| Total number of SIRS criteria met, mean (SD) | <.0001 | 3.02(0.87) | 2.67(0.85) | 2.99(0.87) | 3.11(0.82) | 3.33(0.78) |
| Pneumonia Severity Index | ||||||
| Mean (SD) | <.0001 | 82.0(34.3) | 71.2(30.9) | 83.9(33.7) | 86.8(32.9) | 89.7(35.5) |
| Class, No. (%) | ||||||
| I, II | <.0001 | 541(39) | 224(48) | 109(37) | 52(35) | 156(33) |
| III | 297(21) | 112(24) | 66(22) | 28(19) | 91(19) | |
| IV | 419(30) | 110(24) | 91(31) | 55(37) | 163(34) | |
| V | 127(9) | 16(3) | 29(10) | 13(9) | 69(14) | |
| CURB-65 | ||||||
| Mean (SD), median | <.0001 | 1.57(1.19), | 1.16(1.06), | 1.56(1.15), | 1.73(1.19), | 1.91(1.22), |
| 1.0 | 1.0 | 2.0 | 2.0 | 2.0 | ||
| Group, No. (%) | ||||||
| 1 | <.0001 | 826(50) | 361(67) | 172(48) | 70(41) | 223(38) |
| 2 | 421(25) | 111(20) | 99(28) | 48(28) | 163(28) | |
| 3 | 404(24) | 70(13) | 85(24) | 51(30) | 198(34) | |
| Previous antibiotics, No. (%) | .0707 | 271(16) | 103(19) | 64(18) | 23(14) | 81(14) |
| Outcomes, No. (%) | ||||||
| 30-Day mortality | <.0001 | 106(6.4) | 8(1.5) | 30(8.4) | 16(9.5) | 52(8.9) |
| 90-Day mortality | <.0001 | 161(9.8) | 24(4.4) | 41(11.5) | 26(15.4) | 70(12.0) |
| Hospital LOS (admitted patients), | <.0001 | 6.2(5.3), | 5.1(4.4), | 6.3(5.4), | 6.9(5.2), | 7.0(5.9), |
| mean (SD), median | 5.0 | 4.0 | 5.0 | 5.0 | 5.5 | |
| ICU admission | <.0001 | 215(13) | 42(8) | 48(13) | 20(12) | 105(18) |
| ED discharges | <.0001 | 265(16) | 137(25) | 62(17) | 16(9) | 50(9) |
| Ruled out for CAP | .040 | 94(6) | 38(7) | 24(7) | 6(4) | 26(4) |
| Clinically significant culture | <.0001 | 215(13) | 34(6) | 29(8) | 22(13) | 130(22) |
| Severe sepsis | <.0001 | 423(26) | 91(17) | 99(28) | 51(30) | 182(31) |
| Severe sepsis on day 1 | <.0001 | 202(12) | 33(6) | 43(12) | 26(15) | 100(17) |
| Mechanical ventilation | .0002 | 89(5) | 14(3) | 23(6) | 6(4) | 46(8) |
⁎Test for trend across procalcitonin tiers, with the Jonckheere-Terpstra test for trend for ordinal variables, and test for differences across procalcitonin tiers with the Kruskal-Wallis test for continuous variables. |
Table E1. Procalcitonin, Pneumonia Severity Index, mortality, and severe sepsis.
| Procalcitonin Tier (ng/mL) | No. | 30-Day Mortality (%) | 90-Day Mortality (%) | Severe Sepsis (%) | Severe Sepsis on Day 1 (%) |
|---|---|---|---|---|---|
| PSI classes I, II | |||||
| <0.1 | 224 | 2(1) | 3(1) | 15(7) | 4(2) |
| ≥0.1, <0.25 | 109 | 0 | 0 | 11(10) | 5(5) |
| ≥0.25, <0.5 | 52 | 2(4) | 2(4) | 9(17) | 4(8) |
| ≥0.5 | 156 | 1(1) | 1(1) | 19(12) | 10(6) |
| Total | 541 | 5(1) | 6(1) | 54(10) | 23(4) |
| PSI class III | |||||
| <0.1 | 112 | 3(3) | 8(7) | 11(10) | 3(3) |
| ≥0.1, <0.25 | 66 | 2(3) | 3(5) | 20(30) | 6(9) |
| ≥0.25, <0.5 | 28 | 0 | 1(4) | 4(14) | 0 |
| ≥0.5 | 91 | 2(2) | 3(3) | 22(24) | 9(10) |
| Total | 297 | 7(2) | 15(5) | 57(19) | 18(6) |
| PSI class IV | |||||
| <0.1 | 110 | 2(2) | 11(10) | 42(38) | 13(12) |
| ≥0.1, <0.25 | 91 | 13(14) | 15(16) | 33(36) | 14(15) |
| ≥0.25, <0.5 | 55 | 10(18) | 16(29) | 25(45) | 15(27) |
| ≥0.5 | 163 | 19(12) | 31(19) | 62(38) | 34(21) |
| Total | 419 | 44(11) | 73(17) | 162(39) | 76(18) |
| PSI class V | |||||
| <0.1 | 16 | 0 | 0 | 8(50) | 8(50) |
| ≥0.1, <0.25 | 29 | 11(38) | 16(55) | 19(66) | 9(31) |
| ≥0.25, <0.5 | 13 | 4(31) | 5(38) | 5(38) | 3(23) |
| ≥0.5 | 69 | 23(33) | 27(39) | 44(64) | 29(42) |
| Total | 127 | 38(30) | 48(38) | 76(60) | 49(39) |
Table E2. Procalcitonin, CURB-65, mortality and severe sepsis.
| Procalcitonin Tier (ng/mL) | No. | 30-Day Mortality (%) | 90-Day Mortality (%) | Severe Sepsis (%) | Severe Sepsis on Day 1 (%) |
|---|---|---|---|---|---|
| CURB-65 group 1 | |||||
| <0.1 | 361 | 4(1) | 8(2) | 35(10) | 12(3) |
| ≥0.1, <0.25 | 172 | 3(2) | 4(2) | 28(16) | 11(6) |
| ≥0.25, <0.5 | 70 | 2(3) | 2(3) | 16(23) | 5(7) |
| ≥0.5 | 223 | 4(2) | 6(3) | 39(17) | 16(7) |
| Total | 826 | 13(2) | 20(2) | 118(14) | 44(5) |
| CURB-65 group 2 | |||||
| <0.1 | 111 | 4(4) | 9(8) | 33(30) | 11(10) |
| ≥0.1, <0.25 | 99 | 8(8) | 13(13) | 26(26) | 11(11) |
| ≥0.25, <0.5 | 48 | 5(10) | 11(23) | 16(33) | 9(19) |
| ≥0.5 | 163 | 10(6) | 17(10) | 46(28) | 17(10) |
| Total | 421 | 27(6) | 50(12) | 121(29) | 48(11) |
| CURB-65 group 3 | |||||
| <0.1 | 70 | 0 | 7(10) | 23(33) | 10(14) |
| ≥0.1, <0.25 | 85 | 19(22) | 24(28) | 45(53) | 21(25) |
| ≥0.25, <0.5 | 51 | 9(18) | 13(25) | 19(37) | 12(24) |
| ≥0.5 | 198 | 38(19) | 47(24) | 97(49) | 67(34) |
| Total | 404 | 66(16) | 91(23) | 184(46) | 110(27) |
The mean procalcitonin level at presentation was 3.4 ng/mL, but levels were broadly spread (SD 16.5), such that 542 subjects (32.8%) were in tier I, 356 (21.6%) in tier II, 169 (10.2%) in tier III, and 584 (35.4%) in tier IV. Higher procalcitonin tiers were associated with more clinical signs of infection and a worse course and outcome (Table 2). For example, subjects in the lowest tier had the lowest severity scores, lowest likelihood of developing clinically significant cultures, and lowest rates of severe sepsis, mechanical ventilation, ICU admission, and death. Lowest procalcitonin tier patients were also the most likely to be discharged from the ED and, if admitted, to be subsequently ruled out for community-acquired pneumonia (Table 2). Among subjects who developed severe sepsis, those in the lowest procalcitonin tier were less likely to develop central nervous system (5.3% versus 24.2%; P=.0001), cardiovascular system (4.0% versus 11.0%; P=.08), and greater than 3 organ (1.3% versus 16.5%; P=.0001) system dysfunctions and had more organ failure–free days in the first 30 days (25.5 versus 18.5; P<.0001) compared with those in procalcitonin tiers II to IV.
Main Results
As standalone tests, the Pneumonia Severity Index, CURB-65, and procalcitonin had similar characteristics. For each, 30-day mortality generally increased with increasing category (Figure 2, Figure 3; Table 2). In addition, after dichotomizing of subjects as “high” or ”low” risk, each test alone had moderate specificity (35% to 64%), high sensitivity (87% to 92%), low positive likelihood ratios (1.41 to 2.43), and modest negative likelihood ratios (0.20 to 0.23) (Table 3). Simply adding the procalcitonin test result to the Pneumonia Severity Index score in all subjects led to only minimal improvement in test performance. In the derivation cohort, sensitivity increased from 71.8% to 76.1%, specificity from 77.3% to 79.8%, and area under the receiver operating curve from 0.83 to 0.85. Similarly, in the validation cohort, sensitivity and specificity changed from 78.0% to 78.3% and from 75.2% to 78.9%. Results were similar when procalcitonin level was added to CURB-65 score.
Figure 2.
Kaplan-Meier survival curves, by Pneumonia Severity Index class and procalcitonin tier. In Pneumonia Severity Index classes I to III, mortality was low, and stratification by procalcitonin tier did not provide additional information. In Pneumonia Severity Index classes IV/V, patients with a procalcitonin level less than 0.1 ng/mL had the lowest 30-day mortality. PCT, Procalcitonin.
Figure 3.
Kaplan-Meier survival curves, by CURB-65 group and procalcitonin tier. In CURB-65 group 1 patients, mortality was low, and stratification by procalcitonin tier did not provide additional information. In CURB-65 groups 2/3, patients with a procalcitonin level less than 0.1 ng/mL had the lowest 30-day mortality.
Table 3. Test characteristics of alternative risk-assessment strategies for 30-day mortality.
| Risk-Assessment Strategy | No. | Positive Likelihood Ratio (95% CI) | Negative Likelihood Ratio (95% CI) | Specificity | Sensitivity |
|---|---|---|---|---|---|
| Using each test alone in all patients | |||||
| PSI⁎ | 1,384 | 2.43(2.18–2.70) | 0.20(0.12–0.34) | 0.64 | 0.87 |
| CURB-65† | 1,651 | 1.85(1.70–2.02) | 0.23(0.14–0.39) | 0.53 | 0.88 |
| Procalcitonin‡ | 1,651 | 1.41(1.32–1.51) | 0.22(0.11–0.43) | 0.35 | 0.92 |
| Using procalcitonin in low-risk patients‡ | |||||
| PSI classes I–III | 838 | 0.97(0.60–1.58) | 1.04(0.53–2.04) | 0.40 | 0.58 |
| CURB-65 group 1 | 826 | 1.23(0.85–1.78) | 0.70(0.31–1.59) | 0.44 | 0.69 |
| Using procalcitonin in high-risk patients‡ | |||||
| PSI classes IV/V | 546 | 1.33(1.25–1.42) | 0.09(0.02–0.36) | 0.27 | 0.98 |
| CURB-65 group 2/3 | 825 | 1.26(1.19–1.34) | 0.18(0.07–0.47) | 0.24 | 0.96 |
⁎PSI classes IV/V vs classes I-III. |
†CURB-65 group 2/3 vs group 1. |
‡Procalcitonin level >0.1 ng/mL vs procalcitonin level <0.1 ng/mL. |
Mortality at 30 days was low among subjects identified as low risk by the clinical prediction rules (1.4% for Pneumonia Severity Index classes I to III and 1.6% for CURB-65 group 1), and there was no additional prognostic advantage when results were stratified by procalcitonin tier (Figure 2, Figure 3; Table 3). However, among those identified as high risk by Pneumonia Severity Index score, 23.1% (126/546) had a procalcitonin level in tier I. In this subgroup, only 2 subjects died by day 30, yielding a mortality rate of 1.6%, similar to that of the low-risk subjects. In contrast, mortality was higher in the Pneumonia Severity Index class IV/V subjects who had a procalcitonin tier greater than I (19.0% [80/420]; P<.0001) (Figure 2). The negative likelihood ratio for a low procalcitonin level within clinically high-risk subjects was 0.09 (Table 3). Results were similar with CURB-65: 21.9% (181/825) of CURB-65 group 2/3 subjects had a procalcitonin level in tier I, and mortality was 2.2% (4/181) versus 13.8% (89/644) for subjects with procalcitonin levels in tier I versus tiers II to IV (P<.0001), yielding a negative likelihood ratio for a low procalcitonin of 0.18 (Figure 3 and Table 3).
Sensitivity Analyses
We explored possible reasons for the large mortality difference by procalcitonin tier among subjects identified clinically as high risk. Some clinical signs were worse for those with higher procalcitonin tiers in comparison with those in tier I, but differences were generally modest, and many other aspects of clinical presentation were similar. For example, those in procalcitonin tiers II to IV were more likely to have renal impairment, altered mental status, fever, and leukocytosis but no more likely to be hypoxic, tachypneic, tachycardic, hypotensive, or thrombocytopenic or have abnormal blood glucose or sodium levels. There were also no differences in age, race, sex, Charlson comorbidity level, functional status, duration of symptoms, or domicile (nursing home versus home).
Within clinically high-risk subjects, those in procalcitonin tier I also had lower 90-day mortality (Pneumonia Severity Index 9% versus 26%, P<.001; CURB-65 9% versus 19%, P=.0004) and, in those identified as high risk by CURB-65, shorter length of stay (6.7 days versus 7.9 days; P=.002) and reduced likelihood of ICU admission (13% versus 21%; P=.01) compared with those in procalcitonin tiers II to IV. Neither excluding subjects who ruled out for community-acquired pneumonia nor reclassifying hypoxic Pneumonia Severity Index class I to III subjects as “high risk” significantly changed our results.
Limitations
We chose 30-day mortality as the primary outcome, following the methodology of the original Pneumonia Severity Index and CURB-65 studies. Other outcomes are important and not necessarily correlated with mortality. Nevertheless, subjects with a low procalcitonin level appeared to have lower rates of many other outcomes, including mortality measured at different points (Table E1, Table E2; available online at http://www.annemergmed.com). Second, although procalcitonin tier I was consistently associated with a low mortality, a “dose response” was not observed across procalcitonin tiers II to IV, with similar 30-day mortality rates across these tiers (range 8.4% to 9.5%) (Table 2). Thus, only the lowest range of procalcitonin levels is associated with low mortality. Third, community-acquired pneumonia is a clinical diagnosis with inherent subjectivity. A recent acute bronchitis review noted that a procalcitonin level less than 0.1 ng/mL may be able to safely discriminate between acute bronchitis and community-acquired pneumonia but that more data were needed,38 raising the possibility that some of our cohort with a low procalcitonin level, although diagnosed clinically and radiographically with community-acquired pneumonia by their treating physicians, may not actually have had pneumonia. However, all Genetic and Inflammatory Markers of Sepsis patients met the community-acquired pneumonia criteria of Fine et al,5 an expected low percentage of patients were later deemed not to have community-acquired pneumonia, this percentage did not markedly vary by procalcitonin tier (range 4% to 7%), and eliminating these patients from our analyses had minimal effect. Last, as with the Pneumonia Severity Index, CURB-65, or any other clinical decision adjunct, procalcitonin level must be interpreted in the context of the individual patient, and clinical judgment is always necessary.
Discussion
More than a century ago, it was observed that “… the result of a laboratory test should have, in a given case, the same value as a cardinal symptom or an approved clinical sign … many … forget this, and fail to correlate the laboratory findings with the clinical findings.”39 We also observed that used alone, procalcitonin performed similarly to existing clinical prediction rules but that indiscriminately adding procalcitonin to all patients, regardless of clinical risk category, provided little additional information. However, clinical prediction rules have 2 important limitations: physicians may misapply or not remember them, and within a given risk category there can be a significant range in outcome. We therefore sought to determine whether procalcitonin could address these concerns, first as a standalone test and then as a layer on top of clinical risk assessment.
Our main finding was that in a large, contemporaneous cohort of patients diagnosed in the ED with community-acquired pneumonia, patients with a procalcitonin level less than 0.1 ng/mL had a low 30-day mortality rate, even in patients defined as high risk by established clinical risk-prediction rules. Thus, adding procalcitonin to the assessment of high clinical risk patients significantly improved the ability to rule out the likelihood of death. There are, however, important caveats to these observations.
First, although a procalcitonin level less than 0.1 ng/mL in high-risk subjects had a low likelihood ratio for death, the relatively wide 95% confidence interval merits caution. Second, a good outcome for a high-risk subject could either be because the risk prediction tool was inadequately discriminant or because the ensuing care averted an adverse outcome. Thus, a retrospective identification of a rule or test with potentially valuable test characteristics should be followed up by prospective assessment of its effect on clinical decisionmaking and outcomes. The potential of the Pneumonia Severity Index was not fully understood until Marrie et al40 and Yealy et al37 demonstrated that it could help physicians safely withhold hospital admission and intravenous antibiotics. We similarly recommend that procalcitonin as an adjunct to clinical tools be tested prospectively before wider use.
Our primary goal was to determine how procalcitonin might enhance existing community-acquired pneumonia prediction rules and decisionmaking. We recognize that physicians often do not explicitly calculate the Pneumonia Severity Index in daily clinical practice. However, the same factors that comprise Pneumonia Severity Index and other prediction rules also go into the bedside clinical judgment many physicians use to guide their decisions. Our results therefore suggest that procalcitonin may aid decisionmaking in high-risk patients, defined explicitly or implicitly with the Pneumonia Severity Index or a similar tool. Most important, we again emphasize that procalcitonin level should never be used in isolation to make clinical decisions and does not replace physician assessment.
Current community-acquired pneumonia guidelines recommend that clinically high-risk patients be hospitalized and that ICU admission be considered for patients in the highest risk categories.7, 41, 42 Of interest, Marrie and Huang43 also observed that many clinically high-risk patients might be safely treated at home. Our data suggest that procalcitonin may aid in identifying Pneumonia Severity Index/CURB 65 high-risk patients who will rarely experience mortality and other complications. Thus, there could be considerable benefits, both in terms of conserved resources and antibiotic management, if one could better stratify high-risk patients. Prospective studies are needed to determine whether procalcitonin can improve physician management decisions and outcomes in high-risk patients.
However, 2 clinically important questions merit emphasis. First, is a low procalcitonin level in a high-risk patient clinically obvious and only a “costly surrogate measure of health status”?44 We believe not, because these patients did not appear different from other high-risk patients across a wide range of baseline variables, suggesting that a low procalcitonin level is not particularly obvious at the bedside. Second, why did those patients with high clinical risk scores, yet low procalcitonin levels, do so well? We found that a low procalcitonin level was associated with shorter length of stay, lower proportions of mechanical ventilation and ICU admission, and a more benign severe sepsis phenotype. This observation suggests that although community-acquired pneumonia patients with high clinical risk scores often already have severe sepsis at ED presentation or develop it later, a low procalcitonin level portends a less severe course, potentially explaining the associated low mortality.
Our observational study design does not allow us to address whether procalcitonin can be useful in guiding antibiotic treatment, according to ability to determine bacterial infection. Instead, our work focused on procalcitonin level and links to adverse outcomes in those already diagnosed with community-acquired pneumonia. We recognize, however, that these 2 domains, although theoretically distinct, are practically intermingled. For example, patients with community-acquired pneumonia who have positive blood culture results tend to fare worse, and low procalcitonin level is associated with both low positive culture rates and low mortality.
We conclude that in a large, multicenter, community-acquired pneumonia cohort, patients in the lowest procalcitonin tier (<0.1 ng/mL) were at a low risk of death, regardless of clinical risk. Used indiscriminately, procalcitonin level provided little additional information over Pneumonia Severity Index score and CURB-65 risk assessment. However, a selective 2-tiered approach of first performing a clinical risk assessment and then obtaining procalcitonin level only in those judged to be high risk offers potentially important value.
We thank the patients for their participation in this study. We are also indebted to the investigators, coordinators, and clinical, laboratory, and research personnel at each of the participating hospitals for their efforts. The work was performed at the CRISMA Laboratory, Department of Critical Care Medicine, University of Pittsburgh, Pittsburgh, PA, and the participating sites.
Appendix
In addition to the authors, the following individuals and institutions participated in the Genetic and Inflammatory Markers of Sepsis (GenIMS) study:
Coinvestigators
University of Pittsburgh, Pittsburgh: Mark S. Roberts, MD; Mitchell P. Fink, MD; Michael J. Fine, MD; Kelly A. Wood, MD, MHS; Kenneth Kalassian, MD; Russell L. Delude, PhD; Haichao Wang, PhD; Thomas Auble, PhD; Andrew Schaefer, PhD; Michael R. Pinsky, MD; Eleanor Finegold, PhD; Greg Cooper, MD, PhD; Vincent Arena, PhD; Alex Krichevsky, DVM, PhD
Norwalk Hospital, Norwalk: Jonathan M. Fine, MD
New York University, New York: Kevin J. Tracey, MD, FACS
Site Investigators
PennsylvaniaJefferson Hospital/SHHS, Pittsburgh: Christopher Dooley, MD; Jacqueline Anderson, RN; David Laman, MD; Mahapareh Mostoufi, MD
Mercy Hospital, Pittsburgh: Bruce A. MacLeod, MD; Susan Rolniak, RN, CRNP; Pat Riley, MD; Ken Greer, MD; Dennis Borochovitz, MD
Sewickley Valley Hospital, Sewickley: Frank Gaudio, MD; Greg Maggi, RN; Mindy Hufnagel
St. Clair Memorial Hospital, Pittsburgh: Christopher DeLuca, MD; Arlene Grogan, RN; Stephen Basheda, DO; Martha Clark, MD
UPMC Braddock, Braddock: Richard Heath, MD; Sandra Casey, RN; Emily Yee, MD; Charles Krifcher, MD
UPMC Horizon Health System (Greenville), Greenville and UPMC Horizon Health System (Shenango), Ferrell: Jeffrey Moldovan, DO; Amy Pagano, MSN; Janet Moldovan, RN, MSN
UPMC Lee Regional, Johnstown: Sandy Ergas, MD; Karen Betcher, RN; Ed Rocker
UPMC McKeesport, McKeesport: Rani K. Kumar, MD; Linda Hewitt, RN; Linda Knestaut, RN; Rahut Chaudhry, MD; Ray Probst
UPMC Passavant, Pittsburgh: William G. Kristan, MD; Carol Hewlett, RN; Linda Campbell, RN; Mark Provenzano, MD; Joe Kuzma
UPMC Presbyterian, Pittsburgh: Theodore R. Delbridge, MD, MPH; Mary Ann Murcek, RN; Alan Wells, MD
UPMC Shadyside, Pittsburgh: Jerold Solot, DO; Joel Weinberg, MD; Linda Waddell, RN; Michael Becich, MD
UPMC South Side, Pittsburgh: Joel Rosenbloom, DO; Carol Joyce, RN; C. Vaugh Strimlan, MD; Sukamal Khasnabis, MD
UPMC St. Margaret, Pittsburgh: James Nicholas, MD; Ann Morris, RN; Nancy Gorsha, RN; Mitchell Patti, MD; Jagjit Singh, MD; Meredith Naples
The Western Pennsylvania Hospital, Pittsburgh: Thomas P. Campbell, MD; Diana Morrow, RN; Paul C. Fiehler, MD; Stanley Geyer, MD; Bud Jozwiak
ConnecticutBridgeport Hospital, Bridgeport: Michael Werdmann, MD; Kathy Nunnink, RN, BSN, CCRN; Herbert Scherzer, MD; Larry H. Bernstein, MD; Raymond Haddad, MD
Hartford Hospital, Hartford: Robert Grant, MD; Caryn St. Clair, RN; Beverly Reynholds, RN, Roseanne Papa, RN
Milford Hospital, Milford: Jay Walshon, MD; Tina Null, RN; Lloyd Friedman, MD; Suri Pappu; Will Jones
New Britain General, New Britain: Louis Graff, MD; Michael McNamme, MD; Barry Jacobs, MD; Stephen Wolf, MD; David Buono, MD; Nancy Bennett, RN
Norwalk Hospital, Norwalk: Jonathan Fine, MD; Michael Carius, MD; Saraswathi Nair, MD; Christine Belden, RN; Leonard Scinto
St. Mary's Hospital, Waterbury: Steve Holland, MD; Eleanor Flynn, RN; Edie Kalan, RN; Rose Riordan
Yale University Hospital, New Haven: Linda C. Degutis, DrPH; Nancy Olson, RN, MS; Mark Siegel, MD; Peter Jatlow, MD
MichiganHenry Ford Health System, Detroit: Emanuel P. Rivers, MD, MPH; Kant Shah, MD
Detroit Receiving/Wayne State, Detroit: Robert Welch, MD; Denise Waselewsky, RN; James Kruse, MD; Kristen Bilicki
Sinai-Grace, Detroit: Robert Dunne, MD; Denise Waselewsky, RN; Kristen Bilicki
TennesseeMethodist Healthcare, Memphis: Richard Wunderink, MD; Carol Jones, RN, BSN; Lori Kessler, RN
OthersUniversity of Pittsburgh, Pittsburgh: Tammy L. Young; Gerard Nau, MD; Malik Rahim, MD; Michael Coughlin, MS, PhD; Margaret V. Bowman, RN; Tricia A. Powell, RN; LeeAnn Mandich, RN; Heather Sterling, RN; Karen Greenwald, RN; Tracy Hoteck, RN; Susan Hebda, RRT; Mark Cohen; Marcia McCaw, RN; Heather Woods, RN; Michael Martino, MD; Melinda Carter; Angela Darnley; Angel Shaufl; Jodi Gigler; Javier Martinez, PhD; Xia Tang, MSIS
University of Pennsylvania, Philadelphia: Jason Christie, MD, PhD
North Shore-LIJ Research Institute, Manhasset: Annette Lee, PhD
Yale University School of Medicine, New Haven: Courtney McDonald, BS
References
- . 2004 National Hospital Discharge Survey. Adv Data. 2006;371:1–19
- The cost of treating community-acquired pneumonia. Clin Ther. 1998;20:820–837
- Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med. 2001;29:1303–1310
- Hospitalized community-acquired pneumonia in the elderly: age- and sex-related patterns of care and outcome in the United States. Am J Respir Crit Care Med. 2002;165:766–772
- A prediction rule to identify low-risk patients with community-acquired pneumonia. N Engl J Med. 1997;336:243–250
- Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study. Thorax. 2003;58:377–382
- . Clinical policy for the management and risk stratification of community-acquired pneumonia in adults in the emergency department. Ann Emerg Med. 2001;38:107–113
- Effect of increasing the intensity of implementing pneumonia guidelines: a randomized, controlled trial. Ann Intern Med. 2005;143:881–894
- Severe community-acquired pneumonia: use of intensive care and evaluation of the American Thoracic Society and British Thoracic Society criteria. Am J Respir Crit Care Med. 2002;166:717–723
- . Protein biomarker discovery and validation: the long and uncertain path to clinical utility. Nat Biotechnol. 2006;24:971–983
- . Procalcitonin in bacterial infections—hype, hope, more or less?. Swiss Med Wkly. 2005;135:451–460
- Comparison of procalcitonin and C-reactive protein as markers of sepsis. Crit Care Med. 2003;31:1737–1741
- Serum procalcitonin in children with suspected sepsis: a comparison with C-reactive protein and neutrophil count. Pediatr Crit Care Med. 2003;4:190–195
- Bacteremic elder emergency department patients: procalcitonin and white count. Acad Emerg Med. 2004;11:393–396
- Procalcitonin as a diagnostic test for sepsis in critically ill adults and after surgery or trauma: a systematic review and meta-analysis. Crit Care Med. 2006;34:1996–2003
- Procalcitonin-guidance of antibiotic therapy in community-acquired pneumonia: a randomized trial. Am J Respir Crit Care Med. 2006;174:84–93
- Antibiotic treatment of exacerbations of COPD: a randomized, controlled trial comparing procalcitonin-guidance with standard therapy. Chest. 2007;131:9–19
- Effect of procalcitonin-guided treatment on antibiotic use and outcome in lower respiratory tract infections: cluster-randomised, single-blinded intervention trial. Lancet. 2004;363:600–607
- Accuracy of procalcitonin for sepsis diagnosis in critically ill patients: systematic review and meta-analysis. Lancet Infect Dis. 2007;7:210–217
- Procalcitonin test in the diagnosis of bacteremia: a meta-analysis. Ann Emerg Med. 2007;50:34–41
- Usefulness of procalcitonin levels in community-acquired pneumonia according to the Patients Outcome Research Team Pneumonia Severity Index. Chest. 2005;128:2223–2229
- Serum procalcitonin levels in patients with mild community-acquired pneumonia. Clin Microbiol Infect. 2005;11:1050–1051
- The future diagnostic role of procalcitonin levels: the need for improved sensitivity. Clin Infect Dis. 2003;36:823–824
- Multicenter evaluation of a new immunoassay for procalcitonin measurement on the Kryptor System. Clin Chem Lab Med. 2004;42:440–449
- Prospective comparison of three validated prediction rules for prognosis in community-acquired pneumonia. Am J Med. 2005;118:384–392
- Minimizing the workup of blood culture contaminants: implementation and evaluation of a laboratory-based algorithm. J Clin Microbiol. 2002;40:2437–2444
- . Blood culture contamination: persisting problems and partial progress. J Clin Microbiol. 2003;41:2275–2278
- Predicting bacteremia in patients with community-acquired pneumonia. Am J Respir Crit Care Med. 2004;169:342–347
- . ACP Best Practice No 167: the laboratory diagnosis of urinary tract infection. J Clin Pathol. 2001;54:911–919
- In: Mayhall C, Horan TC, Gaynes R editor. Hospital Epidemiology and Infection Control. Philadelphia, PA: Lippincott Williams & Wilkins; 2004;p. 1659–1702
- Practice guidelines for the management of community-acquired pneumonia in adults (Infectious Diseases Society of America). Clin Infect Dis. 2000;31:347–382
- Guidelines for the management of adults with community-acquired pneumonia (Diagnosis, assessment of severity, antimicrobial therapy, and prevention). Am J Respir Crit Care Med. 2001;163:1730–1754
- . Pathogenic and indigenous microorganisms of humans. In: Murray PR editors. Manual of Clinical Microbiology. Washington, DC: American Society for Microbiology Press; 2003;p. 44–54
- 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions conference. Intensive Care Med. 2003;29:530–538
- Sepsis in European intensive care units: results of the SOAP study. Crit Care Med. 2006;34:344–353
- . Diagnostic tests 4: likelihood ratios. BMJ. 2004;329:168–169
- Effect of increasing the intensity of implementing pneumonia guidelines: a randomized, controlled trial. Ann Intern Med. 2005;143:881–894
- . Acute bronchitis. N Engl J Med. 2006;355:2125–2130
- . JAMA 100 years ago: the laboratory in diagnosis. JAMA. 2007;297:538
- A controlled trial of a critical pathway for the treatment of community acquired pneumonia (Community-Acquired Pneumonia Intervention Trial Assessing Levofloxacin). JAMA. 2000;283:749–755
- . 2004 Update of BTS pneumonia guidelines: what's new?. Thorax. 2004;59:364–366
- Infectious Diseases Society of America/American Thoracic Society consensus guidelines on the management of community-acquired pneumonia in adults. Clin Infect Dis. 2007;44(suppl 2):S27–S72
- . Admission is not always necessary for patients with community-acquired pneumonia in risk classes IV and V diagnosed in the emergency room. Can Respir J. 2007;14:212–216
- . Prognostic value of B-type natriuretic peptide in unstable coronary artery disease. JAMA. 2006;295:1895–1896
Supervising editors: Gregory J. Moran, MD; Judd E. Hollander, MD
Dr. Moran and Dr. Hollander were the supervising editors on this article. Dr. Yealy did not participate in the editorial review or decision to publish this article.
Author contributions: DTH, LAW, JAK, DMY, and DCA conceived and designed the study, analyzed and interpreted the data, and provided important critical revisions of the article. DTH, DMY and DCA drafted the article. DTH, JAK, DMY, and DCA provided final approval. LAW and LK provided statistical expertise. LAW also contributed to drafting of the article, and LK contributed to analysis and interpretation of the data. JAK and DCA obtained funding for the study. JAK, MM, and DCA provided administrative, technical, and logistic support. MM also contributed to analysis and interpretation of the data and article revision. DTH takes responsibility for the paper as a whole.
Funding and support: By Annals policy, all authors are required to disclose any and all commercial, financial, and other relationships in any way related to the subject of this article, that might create any potential conflict of interest. See the Manuscript Submission Agreement in this issue for examples of specific conflicts covered by this statement. Funding received from: NIGMS R01 GM061992. DTH, LAW, LK, DCA have provided consulting work to BRAHMS Diagnostica, the manufacturer of the procalcitonin assay used in this study. DTH has received one speaking honorarium and travel expenses for presentation at a satellite symposium of the September 2006 European Respiratory Society meeting in Munich, Germany. BRAHMS Diagnostica covered sample shipping and procalcitonin assay costs; no direct monetary grant support was provided.
Publication dates: Available online March 17, 2008.
Reprints not available from the authors.
PII: S0196-0644(08)00030-9
doi:10.1016/j.annemergmed.2008.01.003
© 2008 American College of Emergency Physicians. Published by Elsevier Inc. All rights reserved.
Refers to article:
- Risk Stratification of Community-Acquired Pneumonia: What Does All of This Mean? , 04 April 2008
