Leaders in Pharmaceutical Business Intelligence (LPBI) Group

The role of biomarkers in the diagnosis of sepsis and patient management

Author: L. H. Bernstein, MD, FCAP

Key words: Systemic Inflammatory Response Syndrome (SIRS), sepsis, septic shock, mean arterial blood pressure (MAP), procalcitonin, combinatorial analysis, effect size

Highlights:
1.   The systemic inflammatory response syndrome (SIRS) is an acute response to trauma, burn, or infectious injury characterized by fever, hemodynamic and respiratory changes, and metabolic changes, not all of which are consistently present.

2.   The SIRS reaction involves hormonally driven changes in liver glycogen reserves, triggering of  lipolysis, lean body proteolysis, and reprioritization of hepatic protein synthesis with up-regulation of synthesis of acute phase proteins and down-regulation of albumin and important circulating transport proteins.

3.   Understanding of the processes leads to the identification of biomarkers for identification of sepsis and severe, moderate or early SIRS, which also can hasten treatment and recovery.

4.   The SIRS reaction unabated leads to a recurring cycle with hemodynamic collapse from septic shock, indistinguishable from cardiogenic shock, and death.

Abbreviations: Systemic Inflammatory Response Syndrome (SIRS), procalcitonin (PCT), C-reactive protein (CRP), mean arterial blood pressure (MAP), Intensive care unit (ICU), total white blood cell count (WBC), transthyretin(TTR)

Summary
By focusing on early and accurate diagnosis of infection in patients suspected of SIRS antibiotic overuse and its associated morbidity and mortality may be avoided, and therapeutic targets may be identified. This discussion investigates the performance of diagnostic algorithms and biomarkers for sepsis in patients presenting with leukocytosis and other findings. Suspected patients are usually identified by WBC above 12,000/mL  PCT level , SIRS and other criteria, such as serum biomarkers of sepsis. In this writer’s study of 435 patients, procalcitonin alone was a superior marker for sepsis. In patients with sepsis there was a marked increase in PCT (p = 0.0001). PCT was increased in patients requiring ICU admission, heart rate and blood pressure monitoring, and assisted ventilation.(p = 0.0001). Means for SIRS/non-SIRS were: CRP 802/404 mg/L; PCT 20.6/7.5 ng/mL; TTR 87.8/125 mg/L. A comprehensive overview of at least a decade of work is provided.

Introduction

Sepsis is a costly diagnosis in hospitalized patients and carries a high financial risk as a comorbidity and a payment penalty for failure to diagnose in a timely manner (1-4) under the severity of illness CMS reimbursement guidelines as a patient safety hazard, with pneumonia and sepsis among the ten most costly among hospital admissions. A polypeptide identical to a prohormone of calcitonin, procalcitonin, was initially described as a potential marker of bacterial disease by Assicot et al.(5). Procalcitonin is almost undetectable under physiological conditions (pg/ml range), but rises to very high values in response to bacteraemia or fungaemia, and appears to be related to the severity of infection.(6)  Sequential measurements in patients with bacteraemia have shown a rapid fall within 48 hours of antibiotic administration.

Is it an unlikely candidate sepsis biomarker? It is a precursor of calcitonin, and under ordinary circumstances can be accounted for by its production in the thyroid. It is widely produced extrathyroidally with sepsis. The main reason for such interest is the response seems to be uncoupled to the systemic inflammatory response, as CRP is.  What is it a response to? What is the role it has in primary cellular immune response? There is no question that it is found in very low level without serious infection, so that a low level would remove the necessity for a blood culture. In other words it predicts absence of sepsis. Is it fortuitous or serendipity?

On the other hand, we now refer to bacterial and abacterial sepsis. So there are a significant proportion of sepsis diagnoses for which cultures are negative. You might decide that the method of taking cultures x3 at different times and from different sites (not universally practiced) might account for the negative cultures. It was reported in the New York Times a week ago that a child came to the emergency room in a NY hospital with a simple sports injury abrasing the skin surface, was sent home, returned, and died in 24 hours. It’s not such an easy call, despite the result.

There have been numerous studies of the early recognition of sepsis and related diseases in patients related to admission and intensive care. The consensus  guideline has used a SIRS criteria (7, 8).  The SIRS criteria (7,8)  include temperature ^o C. >= 38^o C., heart rate > 90 beats per minute, respiratory rate >20 breaths per minute or PaCO₂   < 32 mm of Hg, WBC > 12,000/mL or < 4,000/mL, or > 10% band forms. For the diagnosis of sepsis (2 or more SIRS criteria and confirmed or suspected infection) the SIRS criteria are insufficient with a very high false positive rate, so that there are medical institutions that require three.  The observation of fever, tachypnea, and tachycardia may well not be present early, and the observation of leukocytosis with or without band neutrophilia are unreliable.  The use of a plasma lactic acid measurement has been added, but it is basically a reflection of decreased splanchnic circulation to the superior mesenteric artery and liver bed.  A better case has been made for absolute neutophilia (9), with an adjustment for high lymphocyte counts in children in the first six months, and for the identification of increased immature myeloid precursors (myelocytes and metamyelocytes). The C-reactive protein, a long established acute phase protein, is not predictive at levels under 520 mg/L because it is elevated in a variety of chronic inflammatory conditions.  Moreover, it is infrequently used in adult medical practice for that reason, except for the high sensitivity CRP, which has been shown to be relevant to treatment of coronary heart disease by the Jupiter study (10), but also has raised the question of overuse of the statins with the undesirable side effect of skeletal muscle loss.  We can explore the validity of over ten years of studies in Europe and US, showing a debated benefit from using the procalcitonin biomarker (PCT, Brahms)(11-28).

The question arises whether there is a difference in PCT response between Gm- and Gm+, and there is, but it doesn’t matter. We can’t be certain that this reaction is  produced by vascular endothelium. I could identify at least one case where I knew the patient had no culture taken and died in cardiovascular shock. At that stage of decompensation what would make one think that it was other than pump failure? But in the evolution of cardiogenic shock, the perfusion of the superior mesentery artery could well be decompensated with the release of bacteria of intestinal organs. Quite a bit of work has been done in surgical research that indicates that there is bacterial migration to regional nodes, but it isn’t problematic unless there is a systemic insult.

Then there needs to be an explanation of events related to gene regulation and cell signaling, that is a blank screen. The cost of the test has been a factor limiting the use, which I think is not the key issue.

A multivariable study with 435 patients to determine whether procalcitonin alone is a superior marker for sepsis established a difference between patients with and without sepsis , as the increase in PCT with sepsis was significant at p = 0.0001.  The PCT was also increased in patients who were admitted to the ICU, requiring assisted ventilation and monitoring of heart rate and blood pressure, significant at p = 0.0001.  The highest values occur in septic shock, even though mild elevations are seen with localized infection.  A PCT level distinguishes between four subgroups of the population: no infection, local infection, sepsis, and severe sepsis and septic shock. The PCT is very low with no infection and rises to levels above 5 with moderate to severe sepsis.  Table 1 compares the selected biomarker results for patients in ICU and not in ICU by the t test with adjustment for unequal variances.  The salient points are noted. PCT, not CRP, is significant for a difference between SIRS positive and SIRS negative patients in ICU (p=0.0001 vs 0.166), with a mean CRP of 495 mg/L in non ICU patients.  The result is expected because the CRP is elevated in other conditions without sepsis, and the patients in the ICU and outside the ICU may have a variety of inflammatory conditions.  There is a difference in PCT between ICU and non-ICU patients without SIRS (p = 0.0001), indicating that PCT is identifying infection, if not sepsis, in this cohort of patients.  Patients in the ICU with SIRS (vs without) have a mean PCT of 35.2 vs 3.7 ng/ml, a clinically as well as statistically significant increase in PCT (Mann-Whitney, p = 0.0001). On the other hand, patients without SIRS in the ICU also have higher CRP [1] and PCT [2] than those not in ICU (1, p = 0.032; 2, p = 0.004), but the CRP difference just reaches statistical significance while the unexpected PCT increase is a big effect.  The means obtained for SIRS/non-SIRS are: CRP [1] 802/404 mg/L; PCT [2] 20.6/7.5 ng/ml; TTR [3] 87.8/125 mg/L. So we see an increase of CRP and PCT with SIRS, as expected, and a decrease in TTR to below 100 mg/L as an obligatory response in relationship to the severity of the inflammatory state. The patients with SIRS and those with sepsis, respectively, had significant elevations of CRP [1] and PCT[2], and decrease of TTR[3] by Mann-Whitney test      (SIRS, p = 0.006, CRP [1]; 0.0001, PCT [2]; 0.0001 TTR[3]; sepsis, p = 0.010, CRP [1]; 0.014, PCT [2]; 0.004, TTR [3]).

Finally, we are particularly interested in combinations of variables for predicting infection and the stage of infection using PCT, MAP (mean arterial blood pressure), WBC.  The model fit looks good based on the chi-squared statistics.   The relationship between the predictor variables and SIRS or ICU is consistent with the large increases of PCT described.  In exploring this strong association between the chosen predictors for SIRS and ICU – PCT, WBC, MAP and TTR (CRP not used) demonstrated that each can be used in a classification matrix. (Table 2)   One might create a 4-letter code using M (MAP)_T (TTR)_W (WBC)_S (SIRS) to test for PCT effectiveness in the same way that one would test PCT against the clinical diagnosis of sepsis, but excluding the variable that we are interested in proving.  PCT can then be added as the fifth variable. The data was classified according to mean arterial blood pressure (MAP), TTR, white cell count (WBC), and SIRS as described and the PCT was kept out of the classification.  The PCT, MAP and SIRS differences were all significant at 0.0001, and the TTR had a p = 0.021 in the Kruskall Wallis analysis by ranks. A similar, but simpler classification did not include either PCT or MAP, resulting in three classes. The PCT increased within the subgroups of WTSIRS (2.5, 25, 50, 75 and 97.5 percentiles. Further, the subgroup of patients classified by MAP, TTR, WBC and SIRS, had a dramatic decline in the MAP by subclass, the last of which is comparable to septic shock.   We constructed a latent class model using the ordinals of PCT, MAP and WBC scaled to intervals. The combinatorial classes formed by the predictors have a defined contribution (percent) of each predictor.  The R², L² and C² are shown.

Degrees of freedom (df)            282                  p-value

L-squared (L²)                         217.6904         1.00

X-squared                                235.4391         0.98

Cressie-Read                           209.1484         1.00

^aBIC (based on LL)                  634.1041

^bAIC (based on LL)                 494.7921

^aBayes Information Criterion, ^bAkaike’s Information Criterion

Model for Dependent

R²

Class1             0.7961

Class2             0.9136

Class3             0.9782

Class4             0.6410

Overall            0.9319

The chi-squared with a p-value close to 1 indicates that the partitioning of the combinatorial groups is excellent, and the model is not underfit or overfit.   A second cluster LCA model using SIRS as covariate also gives a good fit. The p-values are acceptable.

Chi-squared Statistics

Degrees of freedom (df)                        258                  p-value

L-squared (L²)                                     248.9633         0.65

X-squared                                            230.6279         0.89

Cressie-Read                                       222.0675         0.95

BIC (based on L²)                                -1238.4560

AIC (based on L²)                                -267.0367

Even though CRP has validity for identifying SIRS in our exploratory study, the specificity and its usefulness at a cutoff at 52 mg/dL raised significant concerns with respect to with respect to distinguishing SIRS with and without infection.  Keep in mind that large variances in CRP exist between the subgroups, which is problematic because of the CRP elevation in patients with metabolic syndrome and with chronic inflammatory diseases, but not necessarily sepsis. TTR is also decreased, as CRP is elevated, in patients with inflammatory conditions without sepsis.

The data demonstrated a clear relationship between high PCT and both SIRS and the ICU placement of the patient. We also showed the elevation of PCT in concert with severity of infection, as determined by medical record review.  Patients with sepsis are transferred to the ICU for management. These patients are on antibiotics and they are monitored for drop in mean arterial blood pressure over the first 72 hours, as a drop in MAP below 72 indicates septic shock. The progression from SIRS to septic shock goes from single organ (lung, kidney) to double organ, to multiple organ dysfunction syndrome is accompanied by increased lactic acid, tachycardia, decreased MAP, and finally, increase in serum bilirubin.  We would expect and find elevations of PCT with pneumonia, deep wounds, and pyelonephritis as well as sepsis, but these patients would be expected to meet the SIRS criteria.  Our study was not designed to demonstrate how PCT might detect sepsis in older patients with negative SIRS criteria in the ICU who are on ventilator assistance and who may not develop a febrile response. One has to consider that this is related to PCT elevation being partly independent of the SIRS response.

We then showed the same relationship to biomarker combinations that classify the data in accordance with the severity of infection. The WBC count is the weakest classification variable, but it is in the SIRS criteria and it is of greatest value with a finding of immature neutrophils. However, the band count is now considered part of the mature neutrophil count so that the absolute neutrophil count and the immature granulocytes measured by flow cytometry (myelocytes and metamyelocytes) are superior to the tWBC and bands (  ).  The MAP is valuable for identifying the patients at highest risk of circulatory collapse with a drop in the MAP and transfer to ICU (on pressors). The drop in MAP within 72 hours is important for making decisions about Activated protein C (Xygris), which is a critical decision.  APC has been removed from the market.  We included the TTR, which decreases in patients with SIRS in proportion to the severity of illness.  If we compare these predictors in patients with a diagnosis of sepsis, CRP, PCT (and TTR) all have significant differences between ICU and not ICU admission at p < 0.05. The elevation of white cell count and percent neutrophils, unexpectedly has no separatory power. This would not be the case for a postoperative patient who had an increase in WBC that was previously not elevated, or a failure to drop the WBC given antibiotics. On the other hand, for the combination of SIRS and elevated white cell count, the CRP, PCT and TTR all give good separation between ICU and non-ICU patients, an improvement over SIRS positive status alone. This reflects the non specificity with a high false positive rate of the SIRS criteria when neither of the two predictors is granulocytosis.  We asked whether we can combine the key predictor variable into combinatorial classes that would improve the accuracy of identification. CRP was not of interest with respect to this analysis as the poor specificity could only confound matters, and the test offers no additional information than PCT provides.   A more formal classification can then be designed using existing multivariate protocols, but keeping in mind that a mixed model or one using a log-linear model is preferred.

Our study is unique in at least two respects: It is the first to clearly demonstrate a graded response of PCT to severity of infectious challenge.  It is also a rare study to use a classification method to study the effect of combinations of variables associated with SIRS and infection on PCT elevation (11-13). We have created a graded classification by using feature extraction and forming an N-length class based on established information principles (Kullback entropy and Akaike Information Criteria).  The features used to classify the data were PCT, MAP, TTR, WBC and a SIRS score of at least two. The data was also classified without using PCT in order to establish validity for PCT independently of a classification using the feature being evaluated.The classifications have been extensively examined using one-way ANOVA, Kruskal Wallis analysis by ranks,  and latent class analyses. The Kruskal-Wallis analysis by ranks is a powerful tool for comparing the PCT levels within and across feature classes when the variance of PCT values across groups does not have the same distribution and the assumption of normal distribution doesn’t hold. Latent class analysis is a group of methods that can be used to classify data when the data is converted to ordinal classes under the assumption that there is a hidden or “latent” underlying classification. In defining latent classes we assume the criterion of “conditional independence,” so that, each variable is statistically independent of every other variable within each latent class, In our case, latent classes correspond to probabilities for and against the presence of a distinct medical syndrome. In this case we call attention to the classes representing no infection, soft tissue infection, sepsis, and severe sepsis or septic shock.  In reality, the variables used to form the classes are not independent, but they are used as ordinal variables and can be ranked in order of importance in forming a classification. There is a paper of some interest with regard to classification and prediction of severe sepsis that finds PCT, lactic acid and aminoterminal pro B-type natriuretic peptide correlating with sepsis severity scores (12) and another using PCT in combination with waveform analysis (13).  Classification and prediction has been of particular interest to the corresponding investigator for laboratory procedure validation for more than 10 years (20-22).  We provide valuable references to this approach (23-27).

In all respects, we find agreement with a large number of studies, including those recently cited (5-19). PCT has been shown to be useful for discriminating systemic inflammatory response, infection and sepsis (5,6) and identifying a high mortality subgroup with sepsis (15), for identifying sepsis at the onset with either gram negative or gram positive bacteremia (10), for evaluating time to treatment effectiveness (16), and for identifying sepsis in the febrile neutropenic patient (18).

What is it in our findings that makes the results compelling?  CRP has been used for many years, although more widely in European countries than in US until the high sensitivity assay association with risk of coronary artery disease generated a new interest in this test.  CRP increase is associated with SIRS and its production is driven by the cytokines TNFa and Il6.  The liver reprioritises the production of acute phase reactants (CRP, a1 acid glycoprotein [orosomucoid], a1 antitrypsin and transferrin with down-regulation of production of albumin, TTR, transcortin, and other serum transport proteins. However, if this happens in the extreme condition, it also is the case in the more chronic conditions, such as, rheumatoid arthritis and autoimmune diseases, and is a feature of the metabolic syndrome, characterized by type 2 diabetes, obesity, and insulin resistance with or without dyslipidemia. Consequently, CRP is quite high associated with a variety of non-septic conditions.  The emergence of PCT as a sepsis biomarker is somewhat serendipidous.  The protein is the precursor of the peptide calcitonin, produced in the thyroid.  Sepsis appears to upregulate the extrathyroidal production of procalcitonin by vascular endothelium by a mechanism not understood.  PCT can be elevated with non-septic shock, such a cardiogenic shock, but it is easy to comprehend that secondary sepsis could occur by loss of gut integrity and bacterial translocation in such a state.

The importance of pneumonia and pyelonephritis as a precursor of sepsis, and the rapid progression of sepsis in the hospitalised ICU patient can’t be taken lightly.  The identification of sepsis in the emergency room is problematic because the clinical presentation is variable and not so specific.  The SIRS criteria – tachycardia, tachypnea, fever, and leukocytosis – are all common presentations in patients without sepsis.   On the one hand, more accurate identification of the patients with sepsis should eliminate many patients who are screened for sepsis and subsequently have negative culture, and should identify patients who require prompt antimicrobial treatment at an early stage.

The progression from SIRS to sepsis to organ failure and septic shock should also be monitored effectively, which depends on the level of PCT reflecting the level of systemic disorder.  This would permit the best identification of patients who are improving and those who might require activated protein C intervention.  The progression from SIRS to sepsis and beyond is a complex process involving repeated events in the course of the acute illness.  This gives even more justification for combining PCT with other measures to manage the septic state.

Conclusion

We have demonstrated a clear relationship between high PCT and both SIRS and the ICU placement of the patient.  In addition, we have shown the strength of classifying these patients using PCT in combination with MAP, TTR, WBC and SIRS.

References

1. Martin MS, Mannino DM, et al. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med. 2003; 348:1546-1554.

2. NationalCenter for Health Statistics. National vital statistics reports. 2004; 53:9.

3. Angus DC, Linde-Zwirble WT, et al. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med. 2001; 29:1303-1310.

4.  Smith BS, Schroeder WS; Tataronis GR.Hospital reimbursement for adult Patients with severe sepsis treated with drotrecogin alfa (activated). Hospital pharmacy 2005; 40:146-153.

5.  Assicot M, Gendrel D, Cersin H, Raymond J, et al. High serum procalcitonin concentrations in patients withsepsis and infection.  Lancet 1993: 341: 515-18.

6.  Delveraux I, Andre M, Columbier M, Albuisson E, et al. Can procalcitonin measurement help in differentiating between bacterial and other kinds of inflammatory processes?  Ann Rheum Dis 2003; 62: 337-40.

7. Bone R, Balk R, Cerra F, Dellinger R, Fein W, et. Definitions of sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. Amer Coll of Chest Physicians/Soc of Critical Care Med. Chest 1992; 101(6): 1644-655.

8.Dellinger R, Phillip R, Levy MM, Carlet JM, et al. Surviving Sepsis Campaign: Internatl Guidelines for Management of Severe Sepsis and Septic Shock:2008. Crit Care Med 2008; 36(1): 296-377.

9. Bernstein LH and Rucinski R. The relationship between granulocyte maturation and the septic state measurement of granulocyte maturation may improve the early diagnosis of the septic state.  Clin Chem Lab Med 2011;         DOI 10.1515/CCLM.2011.xxx

10. Ridker PM, Danielson E, Fonseca FAH, Genest J, Gotto AM, et al.  Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein.  NEJM 2008; 359: 2195-2207.

11. Jensen JU, Heslet L,; Jensen TH, Espersen K, et al. Procalcitonin increase in early identification of critically ill patients at high risk of mortality. Crit Care Med 2006; 34(10):1-7.

12.  Luzzani A, Polati E, Dorizzi R, Rungatscher A, et al.  Comparison of procalcitonin and C-reactive protein as markers of sepsis. Critical Care Medicine 2003:31(6):1737-1741.

13.  Reith HB, Mittelkotter U, Wagner R, Thiede A. Procalcitonin (PCT) in patients with abdominal sepsis. Intensive Care Med 2000; 26 suppl2: S165-9.

14.  Rau B, Steinbach G, Baumgart K, Gansauge F, et al. The clinical value of procalcitonin in the prediction of infected necrosis in acute pancreatitis. Intensive Care Med 2000; 26, suppl2. S159-64.

15.  Cheval C, Timsit JF, Garrouste-Orgeas M, Assicot M, et al. Procalcitonin (PCT) is useful in predicting the bacterial origin of an acute circulatory failure in critically ill patients. Intensive Care Med 2000; 26, suppl2 S153-8.

16.  Brunkhorst FM, Wegscheider K, Forycki ZF, Brunkhorst R. Procalcitonin for early diagnosis and differentiation of SIRS, sepsis, severe sepsis and septic shock. Intensive Care Med 2000; 26, suppl2 S148-52.

17.  Becker KL, Snider R, Nylen ES. Procalcitonin assay in systemic inflammation, infection, and sepsis: clinical utility and limitations. Crit Care Med 2008; 36(3): 941-52.

18.  Charles PE, Ladoire S, Aho S, Quenot JP, Doise JM, et al. Serum procalcitonin elevation in critically ill patients at the onset of bacteremia caused by either Gram negative or Gram positive bacteria. BMC Infectious Diseases 2008, 8:38 (http://www.biomedcentral.com/1471-2334/8/38)

19.  Wanner GA, Keel M, Steckholzer U, Beier W, Stocker R, Ertel W.  Relationship between procalcitonin plasma levels and severity of injury, sepsis, organ failure, and mortality in injured patients. Critical Care Medicine. 2000; 28(4): 950-957.

20.   Harbath S, Holeckova K, Froidevaux C, Pittet D, et al. Diagnostic Value of Procalcitonin, Interleukin-6, and Interleukin-8 in Critically Ill Patients Admitted with Suspected Sepsis. Amer J Respir Crit Care Med 2001;164: 396-402.

21.   Simon L, Gauvin F, 2 Devendra K. Amre DK,Saint-Louis P, Lacroix J. Serum Procalcitonin and C-Reactive Protein Levels as Markers of Bacterial Infection: A Systematic Review and Meta-analysis. CID 2004: 39:206-217.

22.   Selberg O, Hecker H, Martin M, Klos A Bautsch, et al.  Discrimination of sepsis and systemic inflammatory response syndrome by determination of circulating plasma concentrations of procalcitonin, protein complement 3a, and interleukin-6. Critical Care Medicine. 2000; 28(8): 2793-2798.

23.   Phua J, Koay ES, Lee KH. Lactate, procalcitonin, and amino-terminal pro-B-type natriuretic peptide versus cytokine measurements and clinical severity scores for prognostication in septic shock. Shock 2008; 29(3):328-33.

24.   Zakariah AN ; Cozzi SM ; Van Nuffelen M ; Clausi CM ; Pradier O ; Vincent JL. Combination of biphasic transmittance waveform with blood procalcitonin levels for diagnosis of sepsis in acutely ill patients. Crit Care Med.  2008; 36(5): 1507-12

25.   Viallon A, Guyomarch S, Marjollet O, Berger C, et al. Can Emergency physicians identify a high mortality subgroup of patients with sepsis: role of procalcitonin? Eur J Emerg Med 2008; 15(1):26-33.

26.   Bobre V, Harbarth S, Graf JD, Rohner P, Pugin J. Use of procalcitonin to shorten antibiotic treatment duration in septic patients: a randomized trial. Am J Respir Crit Care Med 2008;177 (5):498-505.

27.   Indino P, Lemarchand P, Bady P, de Torrente A, et al. Prospective study on procalcitonin and other systemic infection markers in patients with leukocytosis. Int J Infect Dis 2008;12 (3):319-24.

The role of biomarkers in the diagnosis of sepsis and patient management
Author: L. H. Bernstein, MD
Key words: Systemic Inflammatory Response Syndrome (SIRS), sepsis, septic shock, mean arterial blood pressure (MAP), procalcitonin, combinatorial analysis, effect size
Highlights:
1. The systemic inflammatory response syndrome (SIRS) is an acute response to trauma, burn, or infectious injury characterized by fever, hemodynamic and respiratory changes, and metabolic changes, not all of which are consistently present.
2. The SIRS reaction involves hormonally driven changes in liver glycogen reserves, triggering of lipolysis, lean body proteolysis, and reprioritization of hepatic protein synthesis with up-regulation of synthesis of acute phase proteins and down-regulation of albumin and important circulating transport proteins.
3. Understanding of the processes leads to the identification of biomarkers for identification of sepsis and severe, moderate or early SIRS, which also can hasten treatment and recovery.
4. The SIRS reaction unabated leads to a recurring cycle with hemodynamic collapse from septic shock, indistinguishable from cardiogenic shock, and death.

Abbreviations: Systemic Inflammatory Response Syndrome (SIRS), procalcitonin (PCT), C-reactive protein (CRP), mean arterial blood pressure (MAP), Intensive care unit (ICU), total white blood cell count (WBC), transthyretin(TTR)

Summary
By focusing on early and accurate diagnosis of infection in patients suspected of SIRS antibiotic overuse and its associated morbidity and mortality may be avoided, and therapeutic targets may be identified. This discussion investigates the performance of diagnostic algorithms and biomarkers for sepsis in patients presenting with leukocytosis and other findings. Suspected patients are usually identified by WBC above 12,000/L PCT level , SIRS and other criteria, such as serum biomarkers of sepsis. In this writer’s study of 435 patients, procalcitonin alone was a superior marker for sepsis. In patients with sepsis there was a marked increase in PCT (p = 0.0001). PCT was increased in patients requiring ICU admission, heart rate and blood pressure monitoring, and assisted ventilation.(p = 0.0001). Means for SIRS/non-SIRS were: CRP 802/404 mg/L; PCT 20.6/7.5 ng/mL; TTR 87.8/125 mg/L. A comprehensive overview of at least a decade of work is provided.

Introduction
Sepsis is a costly diagnosis in hospitalized patients and carries a high financial risk as a comorbidity and a payment penalty for failure to diagnose in a timely manner (1-4) under the severity of illness CMS reimbursement guidelines as a patient safety hazard, with pneumonia and sepsis among the ten most costly among hospital admissions. A polypeptide identical to a prohormone of calcitonin, procalcitonin, was initially described as a potential marker of bacterial disease by Assicot et al.(5). Procalcitonin is almost undetectable under physiological conditions (pg/ml range), but rises to very high values in response to bacteraemia or fungaemia, and appears to be related to the severity of infection.(6) Sequential measurements in patients with bacteraemia have shown a rapid fall within 48 hours of antibiotic administration. There have been numerous studies of the early recognition of sepsis and related diseases in patients related to admission and intensive care. The consensus
guideline has used a SIRS criteria (7, 8). The SIRS criteria (7,8) include temperature </= 36o C. >= 38o C., heart rate > 90 beats per minute, respiratory rate >20 breaths per minute or PaCO2 < 32 mm of Hg, WBC > 12,000/L or < 4,000/L, or > 10% band forms. For the diagnosis of sepsis (2 or more SIRS criteria and confirmed or suspected infection) the SIRS criteria are insufficient with a very high false positive rate, so that there are medical institutions that require three. The observation of fever, tachypnea, and tachycardia may well not be present early, and the observation of leukocytosis with or without band neutrophilia are unreliable. The use of a plasma lactic acid measurement has been added, but it is basically a reflection of decreased splanchnic circulation to the superior mesenteric artery and liver bed. A better case has been made for absolute neutophilia (9), with an adjustment for high lymphocyte counts in children in the first six months, and for the identification of increased immature myeloid precursors (myelocytes and metamyelocytes). The C-reactive protein, a long established acute phase protein, is not predictive at levels under 520 mg/L because it is elevated in a variety of chronic inflammatory conditions. Moreover, it is infrequently used in adult medical practice for that reason, except for the high sensitivity CRP, which has been shown to be relevant to treatment of coronary heart disease by the Jupiter study (10), but also has raised the question of overuse of the statins with the undesirable side effect of skeletal muscle loss. We can explore the validity of over ten years of studies in Europe and US, showing a debated benefit from using the procalcitonin biomarker (PCT, Brahms)(11-28).
Results
A multivariable study with 435 patients to determine whether procalcitonin alone is a superior marker for sepsis established a difference between patients with and without sepsis , as the increase in PCT with sepsis was significant at p = 0.0001. The PCT was also increased in patients who were admitted to the ICU, requiring assisted ventilation and monitoring of heart rate and blood pressure, significant at p = 0.0001. The highest values occur in septic shock, even though mild elevations are seen with localized infection. A PCT level distinguishes between four subgroups of the population: no infection, local infection, sepsis, and severe sepsis and septic shock. The PCT is very low with no infection and rises to levels above 5 with moderate to severe sepsis. Patients in the ICU with SIRS (vs without) have a mean PCT of 35.2 vs 3.7 ng/ml, a clinically as well as statistically significant increase in PCT (Mann-Whitney, p = 0.0001). On the other hand, patients without SIRS in the ICU also have higher CRP [1] and PCT [2] than those not in ICU (1, p = 0.032; 2, p = 0.004), but the CRP difference just reaches statistical significance while the unexpected PCT increase is a big effect. The means obtained for SIRS/non-SIRS are: CRP [1] 802/404 mg/L; PCT [2] 20.6/7.5 ng/ml; TTR [3] 87.8/125 mg/L. So we see an increase of CRP and PCT with SIRS, as expected, and a decrease in TTR to below 100 mg/L as an obligatory response in relationship to the severity of the inflammatory state. The patients with SIRS and those with sepsis, respectively, had significant elevations of CRP [1] and PCT[2], and decrease of TTR[3] by Mann-Whitney test (SIRS, p = 0.006, CRP [1]; 0.0001, PCT [2]; 0.0001 TTR[3]; sepsis, p = 0.010, CRP [1]; 0.014, PCT [2]; 0.004, TTR [3]).
Finally, we are particularly interested in combinations of variables for predicting infection and the stage of infection using PCT, MAP (mean arterial blood pressure), WBC. The model fit looks good based on the chi-squared statistics.

Degrees of freedom (df) 282 p-value
L-squared (L²)       217.69       1.00
X-squared                235.44       0.98
Cressie-Read           209.15       1.00

aBIC (based on LL)      634
bAIC (based on LL)     495
aBayes Information Criterion, bAkaike’s Information Criterion

Model for Dependent
Class1    Class2       Class3    Class4     Overall
R² 0.7961    0.9136    0.9782   0.6410   0.9319

The chi-squared with a p-value close to 1 indicates that the partitioning of the combinatorial groups is excellent, and the model is not underfit or overfit. A second cluster LCA model using SIRS as covariate also gives a good fit. The p-values are acceptable.
Chi-squared Statistics
Degrees of freedom (df) 258      p-value
L-squared (L²)              248.96      0.65
X-squared                       230.62      0.89
Cressie-Read                  222.07     0.95
BIC (based on L²)       -1238.46
AIC (based on L²)        -267.04
Even though CRP has validity for identifying SIRS in our exploratory study, the specificity and its usefulness at a cutoff at 52 mg/dL raised significant concerns with respect to with respect to distinguishing SIRS with and without infection. Keep in mind that large variances in CRP exist between the subgroups, which is problematic because of the CRP elevation in patients with metabolic syndrome and with chronic inflammatory diseases, but not necessarily sepsis. TTR is also decreased, as CRP is elevated, in patients with inflammatory conditions without sepsis.
The data demonstrated a clear relationship between high PCT and both SIRS and the ICU placement of the patient. We also showed the elevation of PCT in concert with severity of infection, as determined by medical record review PWe would expect and find elevations of PCT with pneumonia, deep wounds, and pyelonephritis as well as sepsis, but these patients would be expected to meet the SIRS criteria. Our study was not designed to demonstrate how PCT might detect sepsis in older patients with negative SIRS criteria in the ICU who are on ventilator assistance and who may not develop a febrile response. One has to consider that this is related to PCT elevation being partly independent of the SIRS response.
We then showed the same relationship to biomarker combinations that classify the data in accordance with the severity of infection. The WBC count is the weakest classification variable, but it is in the SIRS criteria and it is of greatest value with a finding of immature neutrophils. However, the band count is now considered part of the mature neutrophil count so that the absolute neutrophil count and the immature granulocytes measured by flow cytometry (myelocytes and metamyelocytes) are superior to the tWBC and bands ( ). The MAP is valuable for identifying the patients at highest risk of circulatory collapse with a drop in the MAP and transfer to ICU (on pressors). The drop in MAP within 72 hours is important for making decisions about Activated protein C (Xygris), which is a critical decision. APC has been removed from the market. We included the TTR, which decreases in patients with SIRS in proportion to the severity of illness. We asked whether we can combine the key predictor variable into combinatorial classes that would improve the accuracy of identification. CRP was not of interest with respect to this analysis as the poor specificity could only confound matters, and the test offers no additional information than PCT provides.
Our study is unique in at least two respects: It is the first to clearly demonstrate a graded response of PCT to severity of infectious challenge. It is also a rare study to use a classification method to study the effect of combinations of variables associated with SIRS and infection on PCT elevation (11-13). We have created a graded classification by using feature extraction and forming an N-length class based on established information principles (Kullback entropy and Akaike Information Criteria). The features used to classify the data were PCT, MAP, TTR, WBC and a SIRS score of at least two. The data was also classified without using PCT in order to establish validity for PCT independently of a classification using the feature being evaluated.The classifications have been extensively examined using one-way ANOVA, Kruskal Wallis analysis by ranks, and latent class analyses. The Kruskal-Wallis analysis by ranks is a powerful tool for comparing the PCT levels within and across feature classes when the variance of PCT values across groups does not have the same distribution and the assumption of normal distribution doesn’t hold. Latent class analysis is a group of methods that can be used to classify data when the data is converted to ordinal classes under the assumption that there is a hidden or “latent” underlying classification. In defining latent classes we assume the criterion of “conditional independence,” so that, each variable is statistically independent of every other variable within each latent class, In our case, latent classes correspond to probabilities for and against the presence of a distinct medical syndrome. In this case we call attention to the classes representing no infection, soft tissue infection, sepsis, and severe sepsis or septic shock. In reality, the variables used to form the classes are not independent, but they are used as ordinal variables and can be ranked in order of importance in forming a classification. There is a paper of some interest with regard to classification and prediction of severe sepsis that finds PCT, lactic acid and aminoterminal pro B-type natriuretic peptide correlating with sepsis severity scores (12) and another using PCT in combination with waveform analysis (13). Classification and prediction has been of particular interest to the corresponding investigator for laboratory procedure validation for more than 10 years (20-22). We provide valuable references to this approach (23-27).
In all respects, we find agreement with a large number of studies, including those recently cited (5-19). PCT has been shown to be useful for discriminating systemic inflammatory response, infection and sepsis (5,6) and identifying a high mortality subgroup with sepsis (15), for identifying sepsis at the onset with either gram negative or gram positive bacteremia (10), for evaluating time to treatment effectiveness (16), and for identifying sepsis in the febrile neutropenic patient (18).
We have demonstrated a clear relationship between high PCT and both SIRS and the ICU placement of the patient. In addition, we have shown the strength of classifying these patients using PCT in combination with MAP, TTR, WBC and SIRS.

References

1. Martin MS, Mannino DM, et al. The epidemiology of sepsis in the United States from 1979 through 2000. N Engl J Med. 2003; 348:1546-1554.
2. National Center for Health Statistics. National vital statistics reports. 2004; 53:9.
3. Angus DC, Linde-Zwirble WT, et al. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med. 2001; 29:1303-1310.
4. Smith BS, Schroeder WS; Tataronis GR. Hospital reimbursement for adult Patients with severe sepsis treated with drotrecogin alfa (activated). Hospital pharmacy 2005; 40:146-153.
5. Assicot M, Gendrel D, Cersin H, Raymond J, et al. High serum procalcitonin concentrations in patients withsepsis and infection. Lancet 1993: 341: 515-18.
6. Delveraux I, Andre M, Columbier M, Albuisson E, et al. Can procalcitonin measurement help in differentiating between bacterial and other kinds of inflammatory processes? Ann Rheum Dis 2003; 62: 337-40.
7. Bone R, Balk R, Cerra F, Dellinger R, Fein W, et. Definitions of sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. Amer Coll of Chest Physicians/Soc of Critical Care Med. Chest 1992; 101(6): 1644-655.
8.Dellinger R, Phillip R, Levy MM, Carlet JM, et al. Surviving Sepsis Campaign: Internatl Guidelines for Management of Severe Sepsis and Septic Shock:2008. Crit Care Med 2008; 36(1): 296-377.
9. Bernstein LH and Rucinski R. The relationship between granulocyte maturation and the septic state
measurement of granulocyte maturation may improve the early diagnosis of the septic state. Clin Chem Lab Med 2011; DOI 10.1515/CCLM.2011.xxx
10. Ridker PM, Danielson E, Fonseca FAH, Genest J, Gotto AM, et al. Rosuvastatin to prevent vascular events in men and women with elevated C-reactive protein. NEJM 2008; 359: 2195-2207.
11. Jensen JU, Heslet L,; Jensen TH, Espersen K, et al. Procalcitonin increase in early identification of critically ill patients at high risk of mortality. Crit Care Med 2006; 34(10):1-7.
12. Luzzani A, Polati E, Dorizzi R, Rungatscher A, et al. Comparison of procalcitonin and C-reactive protein as markers of sepsis. Critical Care Medicine 2003:31(6):1737-1741.
13. 2008, 8:38 (http://www.biomedcentral.com/1471-2334/8/38)
19. Wanner GA, Keel M, Steckholzer U, Beier W, Stocker R, Ertel W. Relationship between procalcitonin plasma levels and severity of injury, sepsis, organ failure, and mortality in injured patients. Critical Care Medicine. 2000; 28(4): 950-957.
20. Harbath S, Holeckova K, Froidevaux C, Pittet D, et al. Diagnostic Value of Procalcitonin, Interleukin-6, and Interleukin-8 in Critically Ill Patients Admitted with Suspected Sepsis. Amer J Respir Crit Care Med 2001;164: 396-402.
21. Simon L, Gauvin F, 2 Devendra K. Amre DK,Saint-Louis P, Lacroix J. Serum Procalcitonin and C-Reactive Protein Levels as Markers of Bacterial Infection: A Systematic Review and Meta-analysis.
CID 2004: 39:206-217.
22. Selberg O, Hecker H, Martin M, Klos A Bautsch, et al. Discrimination of sepsis and systemic inflammatory response syndrome by determination of circulating plasma concentrations of procalcitonin, protein complement 3a, and interleukin-6. Critical Care Medicine. 2000; 28(8): 2793-2798.
23. Phua J, Koay ES, Lee KH. Lactate, procalcitonin, and amino-terminal pro-B-type natriuretic peptide versus cytokine measurements and clinical severity scores for prognostication in septic shock. Shock 2008; 29(3):328-33.
24. Zakariah AN ; Cozzi SM ; Van Nuffelen M ; Clausi CM ; Pradier O ; Vincent JL. Combination of biphasic transmittance waveform with blood procalcitonin levels for diagnosis of sepsis in acutely ill patients. Crit Care Med. 2008; 36(5): 1507-12
25. Viallon A, Guyomarch S, Marjollet O, Berger C, et al. Can Emergency physicians identify a high mortality subgroup of patients with sepsis: role of procalcitonin? Eur J Emerg Med 2008; 15(1):26-33.
26. Bobre V, Harbarth S, Graf JD, Rohner P, Pugin J. Use of procalcitonin to shorten antibiotic treatment duration in septic patients: a randomized trial. Am J Respir Crit Care Med 2008;177 (5):498-505.
27. Indino P, Lemarchand P, Bady P, de Torrente A, et al. Prospective study on procalcitonin and other systemic infection markers in patients with leukocytosis. Int J Infect Dis 2008;12 (3):319-24.
28. Massaro KS, Costa SF, Leone C, Chamone DA. Procalcitonin (PCT) and C-reactive Protein (CRP) as severe systemic infection markers in febrile neutropenic adults. BMC Infect Dis 2007; 22(7): 137 (http://www.biomedcentral.com/1471-2334-7-137).
29. Vermunt J and Magidson J. Latent Gold 4.5. http://www.statistical.innovations.com

Table 1. Comparison of key sepsis biomarkers by means between ICU and non-ICU, SIRS positive status, and sepsis status.

Marker	Mean		p
SIRS Positive	ICU vs not
CRP (mg/L)	780	495	0.166
PCT (ng/ml)	35.2	3.7	0.0001
SIRS negative	ICU vs not
PCT (ng/ml)			0.0001
Sepsis Diagnosis	ICU vs not
CRP (mg/L)	817	521	0.010
PCT (ng/ml)			0.014
WBC			0.921
% neutrophils			0.452
SIRS with WBC	ICU vs not
CRP (mg/L)	802	404	0.006
PCT (ng/ml)	20.6	7.5	0.0001

Table 2. Comparison of PCT, MAP, TTR and WBC means and variation in SIRS and ICU placement.

variable-by-variablea	Fisher’s Exact testb	t-testc
PCT*SIRS	0.0001	0.0210
PCT*ICU	0.0001	0.0001
MAP*SIRS		0.0130
MAP*ICU	0.0001	0.0001
WBC*SIRS	0.0001	0.0001
WBC*ICU		NS
TTR*SIRS	0.0001	0.0001
TTR*ICU		0.0001

^atwo variables, ^b,cp-value

Archival Table

Table 3.  Classes and 4-variable contributions for 4-class LCA cluster model.

Cluster Size

0.3646             0.2738             0.2209             0.1408

Indicators

PCT0425         (PCT: < 0.4, 0.41-2.5, > 2.5)

0          0.6393             0.8894             0.1278             0.0000

1          0.3190             0.1076             0.3752             0.0013

2          0.0355             0.0029             0.2459             0.0285

3          0.0062             0.0001             0.2511             0.9702

MAP758088    (MAP: > 88, 81-87, 75-80, < 80)

0          0.5234             0.6186             0.3709             0.4228

1          0.1861             0.1794             0.1771             0.1828

2          0.1065             0.0838             0.1361             0.1272

3          0.1841             0.1181             0.3160             0.2673

WBC1217       (WBC: < 12, 12-17, > 17)

0          0.3325             0.1772             0.0002             0.3084

1          0.4152             0.3948             0.0216             0.4171

2          0.2523             0.4280             0.9782             0.2745

TTR7105         (TTR: > 10.5, 7.0-10.5, < 7)

0          0.2457             0.9805             0.2159             0.3920

1          0.3352             0.0192             0.3279             0.3391

2          0.4191             0.0003             0.4562             0.2688

            Table 1. Comparison of key sepsis biomarkers by means between ICU and non-ICU, SIRS positive status, and sepsis status.

Marker	Mean		p
SIRS Positive	ICU vs not
CRP (mg/L)	780	495	0.166
PCT (ng/ml)	35.2	3.7	0.0001
SIRS negative	ICU vs not
PCT (ng/ml)			0.0001
Sepsis Diagnosis	ICU vs not
CRP (mg/L)	817	521	0.010
PCT (ng/ml)			0.014
WBC			0.921
% neutrophils			0.452
SIRS with WBC	ICU vs not
CRP (mg/L)	802	404	0.006
PCT (ng/ml)	20.6	7.5	0.0001

Table 2. Comparison of PCT, MAP, TTR and WBC means and variation in SIRS and ICU placement.

variable-by-variablea	Fisher’s Exact testb	t-testc
PCT*SIRS	0.0001	0.0210
PCT*ICU	0.0001	0.0001
MAP*SIRS		0.0130
MAP*ICU	0.0001	0.0001
WBC*SIRS	0.0001	0.0001
WBC*ICU		NS
TTR*SIRS	0.0001	0.0001
TTR*ICU		0.0001

^atwo variables, ^b,cp-value

Archival Table

Table 3.  Classes and 4-variable contributions for 4-class LCA cluster model.

Cluster Size

0.3646             0.2738             0.2209             0.1408

Indicators

PCT0425         (PCT: < 0.4, 0.41-2.5, > 2.5)

0          0.6393             0.8894             0.1278             0.0000

1          0.3190             0.1076             0.3752             0.0013

2          0.0355             0.0029             0.2459             0.0285

3          0.0062             0.0001             0.2511             0.9702

MAP758088    (MAP: > 88, 81-87, 75-80, < 80)

0          0.5234             0.6186             0.3709             0.4228

1          0.1861             0.1794             0.1771             0.1828

2          0.1065             0.0838             0.1361             0.1272

3          0.1841             0.1181             0.3160             0.2673

WBC1217       (WBC: < 12, 12-17, > 17)

0          0.3325             0.1772             0.0002             0.3084

1          0.4152             0.3948             0.0216             0.4171

2          0.2523             0.4280             0.9782             0.2745

TTR7105         (TTR: > 10.5, 7.0-10.5, < 7)

0          0.2457             0.9805             0.2159             0.3920

1          0.3352             0.0192             0.3279             0.3391

2          0.4191             0.0003             0.4562             0.2688