Efficacy of National Nosocomial Infection Surveillance score, acute-phase proteins, and interleukin-6 for predicting postoperative infections following major gastrointestinal surgery

ABSTRACT CONTEXT AND OBJECTIVE: Postoperative infections should be detected earlier. We investigated the efficacy of the National Nosocomial Infection Surveillance (NNIS) score, interleukin-6 (IL-6) and various acute-phase proteins for predicting postoperative infections. DESIGN AND SETTING: Case series study at the Júlio Müller University Hospital. METHODS: Thirty-two patients who underwent major gastrointestinal procedures between June 2004 and February 2005 were studied. The NNIS score and the evolution of serum IL-6 and various acute-phase proteins (C-reactive protein [CRP], albumin, prealbumin and transferrin) were correlated with postoperative infections and length of hospital stay (LOS). RESULTS: NNIS > 1 (p = 0.01) and low preoperative albumin (p = 0.02) significantly correlated with infection. IL-6 and CRP increased significantly more in patients with infections. Multivariate analysis showed greater risk of infection when NNIS > 1 (odds ratio, OR = 10.66; 95% confidence interval, CI: 1.1-102.0; p = 0.04); preoperative albumin < 3 g/dl (OR = 8.77; 95% CI: 1.13-67.86; p = 0.03); CRP > 30 mg/l on the second postoperative day (OR = 8.27; 95% CI: 1.05-64.79; p = 0.03) and > 12 mg/l on the fifth postoperative day (OR = 25.92; 95% CI: 2.17-332.71; p < 0.01); and IL-6 > 25 pg/ml on the fifth postoperative day (OR = 15.46; 95% CI: 1.19-230.30; p = 0.03). Longer LOS was associated with cancer, transferrin, IL-6 and albumin (p < 0.05). CONCLUSIONS: NNIS, albumin, CRP and IL-6 may be useful as predictive markers for postoperative infections. For predicting LOS, malignant condition, transferrin, albumin and IL-6 are useful.


INTRODUCTION
Infections play an important role in postoperative mortality rates. 1 However, the main cause of mortality in such patients is multiple organ system dysfunction (MOSD) as a result of an exacerbated systemic infl ammatory response, with or without infection. 1,2reoperative identifi cation of patients at risk of postoperative complications may have an impact on surgical management and enable treatment modifi cations so as to reduce morbid-mortality.][5][6][7][8] Assessment of nutritional status and immune response may have prognostic value for such patients.It has been demonstrated that prior sensitization of the immune system may result in an inappropriate response to injury or infection.This sensitization is easily shown through the detection of components of the acute-phase response in the serum.There is evidence that exacerbation of the acute response is associated with a more aggressive inflammatory response. 9Thus, rather than diagnosing a nutritional defi cit, it might be more important to assess its impact on the regulation of the immune system.This assessment can be carried out by detecting acute-phase proteins and cytokines.
There seems to be an association between preoperatively detected acute-phase response and the development of postoperative MOSD. 1,2Therefore, patients presenting acute-phase response during the preoperative period may be at risk.Studies focusing on the cause and functional characteristics of this preoperative or early postoperative sensitization could put forward therapeutic options for avoiding postoperative complications.Such assessments would enable individual identification of patients at risk, thereby restricting more vigorous assessments to a smaller number of patients.
Since the 1980s, several physiological scoring systems have been introduced for assessing patients with sepsis and MOSD, including the National Nosocomial Infections Surveillance (NNIS) score. 10This score assesses the intrinsic postoperative risk of infections at the site concerned and can predict infectious morbidity at other sites, such as pneumonia and urinary infections, and also overall mortality. 10The assessment of patients using these physiological scores makes it possible to estimate the risk for groups of patients, while identifi cation of immune system activation by detecting infl ammatory response markers allows individual assessment.The use of scoring systems such as NNIS 11 allows stratifi cation of the infection risk and, when this is added to analysis of infl ammation markers during the postoperative period, they may be capable of selecting patients who are at greater risk of infectious morbidity and mortality, at lower cost and with greater effi cacy.
On the other hand, there are still no ideal tests or markers for early prediction and identifi cation of patients who are prone to septic complications.It is known that critically ill patients present an exacerbated systemic infl ammatory response and that the start of such responses may precede the onset of clinical manifestations.Thus, the adoption of earlier strategies and therapeutic measures may benefi t these patients.Among the various mediators of the systemic infl ammatory response, C-reactive protein (CRP) has wide clinical application for confi rming the existence of an infl ammatory process, through its reasonable sensitivity and low cost.Recently, several studies have reported that the sensitivity and specifi city of CRP in differentiating systemic infl ammatory response syndrome (SIRS) and sepsis is around 60 to 90%. 12 CRP is the prototype acute-phase protein, and its concentration rises sharply in response to various infl ammatory stimuli, and particularly in response to interleukin-6 (IL-6). 13revious studies have suggested that IL-6 serves as both a marker and a mediator for the severity of sepsis.Although IL-6 has both pro-and anti-infl ammatory characteristics, when at high levels it correlates with the extent of the acute phase response and thus it has been considered to be an alarm indicator for the magnitude of the infl ammatory response. 14,15][18] However, there are few studies on this issue, and no previous study has investigated the effi cacy of the NNIS score and infl ammatory markers at the same time.Furthermore, the defi nition of an ideal score for individual preoperative prediction of septic complications and the creation of methods for early sepsis detection have not yet been drawn up, and efforts should continue to be made towards reaching these objectives.

OBJECTIVE
The aim of this study was to investigate the effi cacy of the NNIS score, IL-6 and various acute-phase proteins (CRP, albumin, prealbumin and transferrin) for predicting septic postoperative complications in patients who have undergone major surgical procedures in the digestive tract.

METHODS
This study was carried out after gaining approval from the Research Ethics Committee of the Júlio Müller University Hospital (case number 146/CEP/HUJM/2004).Patients were included in the study after giving their informed consent.

Patients Patients
This was a case series study involving 32 consecutive patients, conducted at the Department of Surgery of the Júlio Müller University Hospital of the Federal University of Mato Grosso, Cuiabá, Brazil, between June 2004 and February 2005.The eligibility criteria for the study were that the subjects could be patients of either sex undergoing major elective surgery of the gastrointestinal tract with at least one intestinal anastomosis performed during the operation.The individual NNIS scores and the evolution of serum IL-6 and various acute-phase proteins (CRP, albumin, prealbumin and transferrin) were assessed in order to investigate associations with the oc-currence of septic complications and length of hospital stay (LOS).

Independent and outcome Independent and outcome variables variables
The following clinical variables were recorded for all patients: which is a score ranging from 0 to 3 based on the presence (1 point) or absence (0 point) of three independent statuses: ASA score greater than two, operation classifi ed as dirty, and duration of the operation longer than expected (more than three hours for esophagus, stomach, small intestine and colon operations, and more than four hours for hepatic, biliary and pancreatic procedures); 20 and presence of infection, sepsis or septic shock according to the consensus of the American College of Chest Physicians/Society of Critical Care Medicine (ACCP/SCCM). 21fectious morbidity and LOS were considered to be the main endpoints of the study.The NNIS score was categorized as high (scores 2 and 3) and low (scores 0 and 1).Serum IL-6, CRP, albumin, prealbumin and transferrin were assayed at the time of inducing anesthesia and on the second and fi fth postoperative days.

Defi nitions Defi nitions
The following defi nitions were adopted: 1. Infection of the surgical incision site: a) purulent drainage through the incision with or without laboratory confirmation; or b) isolation of organism from the wound culture or incision tissue culture and concomitant presence of at least one of the following infection signs and symptoms: pain or hypersensitivity, edema, hyperemia, or fever.2. Intra-abdominal infection: a) purulent drainage through the drain; or b) isolation of organism obtained from abdominal cavity tissue or secretion; or c) abscess found during reoperation, histopathological examination or on image scan. 20. Urinary tract infection: a) urine culture with 100,000 or more colony-forming units (CFU)/ml, with one or at most two bacterial species; or b) when two of the following parameters were present: fever, urination urgency, increased urination frequency, dysuria or suprapubic pain, in addition to one of the following parameters: pyuria, presence of nitrites in the urinary sediment, or positive bacterioscopy. 22. Pneumonia: presence of new or progressive infi ltrate, consolidation, cavitation or pleural effusion in chest X-ray, in addition to clinical manifestation. 22. SIRS (systemic infl ammatory response syndrome) and sepsis were defi ned according to the ACCP/SCCM 21 consensus.SIRS was defined as a condition that included two or more of the following: temperature > 38º C or < 36º C; heart beats > 90/min; respiratory frequency > 20/min or PaCO 2 < 32 mmHg and white cells > 12,000/mm 3 or < 4,000 mm 3 , or > 10% young cells.Sepsis was defi ned if SIRS was associated with infection.

Statistical method Statistical method
The results were initially assessed by univariate analysis using either logistic or linear regression analysis, depending on the outcome variable, in an attempt to identify potential associations with infectious morbidity or LOS.Repeated-measurement analysis of variance (ANOVA) was used to analyze the evolution of all the acute-phase proteins and IL-6 for the presence or absence of infection using LOS as a covariable.Either linear or logistic multivariate regression analysis was used to determine the strength of association between independent variables and the outcome variables.Continuous variables (IL-6 and acute-phase proteins) were categorized to enter the logistic regression analysis.The cutoff used was the mean from the results observed among patients without infection, plus one standard deviation.The backward stepwise model was used for the regression construction.To evaluate whether the model fi t the data, we used the goodness-of-fi t test described by Hosmer and Lemeshow.Finally, validity parameters (sensitivity, specifi city, positive and negative predictive values, and accuracy) were calculated for the variables that reached signifi cance in the multivariate analysis.All the tests were performed using the Statistical Package for the Social Sciences (SPSS) for Windows 11.0.The results were expressed as means, odds ratios (OR) or β, and the respective 95% confi dence intervals (95% CI), as appropriate.The statistical signifi cance was set at 5% probability level (p < 0.05).

RESULTS
The clinical characteristics of the patients studied and the procedures carried out on them are shown in Tables 1 and 2, respectively.Sixteen patients (50%) were considered malnourished.Malignancies were found in 11 (34.4%)patients.Eight (25%) patients needed additional procedures (n = 11) du ring the planned operation, to resolve intercurrences.All patients were included and thus there was no sample loss.
The mortality rate was 6.3% (two cases), due to infection-related MOSD.Infections occurred in 16 patients (50%) who presented a total of 27 infection sites (Table 2).The most frequent infection site was the incision wound

Acute-phase proteins and IL-6 Acute-phase proteins and IL-6
The descriptive data for the evolution of all the acute-phase proteins and IL-6, according to the presence or absence of infection and the LOS, can be seen in Table 4 and Figure 1.

C-reactive protein C-reactive protein
There was no preoperative association between CRP levels and infection: OR = 1.50; 95% CI: 0.27-8.13;p = 0.64.There was a signifi cant rise in CRP levels after the operation in all cases (p < 0.01).This increase in CRP was signifi cantly greater in the patients with infection than without infection over time: second postoperative day, OR = 17.37, 95% CI: 2.43-124.18;fi fth postoperative day, OR = 26.00,95% CI: 2.98-226.97;p < 0.01.This was seen even when LOS was considered as a covariable (Table 4).On the fi fth postoperative day, CRP levels decreased more signifi cantly (p < 0.01) in patients without infection than others (Figure 1A).

Albumin Albumin
Infection was approximately nine times more frequent in patients with preoperative albumin below 3 g/dl (OR = 9.10; 95% CI: 1.38-59.62;p = 0.02).The albumin level diminished over time (p < 0.01), with a signifi cantly greater decrease in patients with infections than in others (p = 0.04).However, this difference did not correlate with infection when LOS was used as a covariable, either using repeated-measurement ANOVA or using logistic regression analysis: second postoperative day, OR = 0.92, 95% CI: 0.18-4.82,p = 0.98; and fi fth postoperative day, OR = 1.18, 95% CI: 0.16-8.58,p = 0.87.These data are shown in Table 4 and Figure 1B.

Transferrin Transferrin
At basal conditions, there was no difference in transferrin levels between those who acquired postoperative infections and those who did not (OR = 4.06; 95% CI: 0.63-26.13;p = 0.14).After the operation, there was a signifi cant decrease in transferrin levels in the two groups (p = 0.01), but without any difference between the patients with or without posto perative infections (Figure 1D).When LOS was considered as a covariable, the comparisons between groups and within groups were not signifi cantly different (Table 4).
The correlation between the preoperative transferrin level and LOS was marginally signifi cant (OR = -0.34;95% CI: -0.18 to 0.00; p = 0.07).On both the second and fi fth postoperative days, the transferrin levels negatively correlated with LOS.In this context, the mean increases in transferrin of 10 units on the second postoperative day (OR = -0.48;95% CI: -0.44 to -0.07; p < 0.01) and on the fi fth postoperative day (OR = -0.44;95% CI: -0.24 to -0.02; p = 0.01) decreased the LOS by approximately fi ve and four days, respectively.

Interleukin-6 Interleukin-6
At basal conditions, similar serum IL-6 levels were found in the two groups (OR = 4.33; 95% CI: 0.42-44.42;p = 0.21).The serum IL-6 levels increased in both groups on the second postoperative day (p < 0.001) and decreased thereafter.However, it was only in patients without infection that the IL-6 values returned to basal levels on day 5. Infection occurred around 8 to 20 times more * p < 0.01 versus preoperative period; † p < 0.01 versus without infection.
The validity parameters for these variables to predict postoperative infection can be seen in Table 6.The best prediction was found with CRP greater than 12 mg/l on the fi fth postoperative day.Although showing low sensitivity, CRP greater than 30 mg/l on the second postoperative day was associated with both 100% specifi city and 100% positive predictive value.

DISCUSSION
The fi ndings from this study have confi rmed that both the acute-phase proteins and IL-6 become modifi ed after the operative trauma.As a rule, there is a rapid increase in CRP and IL-6 and, conversely, a decrease in all other acute-phase proteins following the operation.The results also showed that these modifi cations differ according to whether patients acquire postoperative infection or not.][25][26][27][28] Our data suggest that the NNIS index and preoperative albumin level are important predictive factors for perioperative infection.Indeed, except for albumin, no other biochemical marker measured during the operation was predictive of infection in the postoperative period in the light of multivariate analysis.However, both NNIS and  preoperative albumin had low sensitivity for predicting postoperative infections.These fi ndings are important, because no other previous study has correlated NNIS scores with acute-phase proteins or cytokines.In the early postoperative period, IL-6 and most of the acute-phase proteins were correlated with postoperative infection by means of univariate analysis.However, when the data were analyzed using multivariate logistic regression, CRP on the second day and again CRP and IL-6 on the fi fth day were the only predictive factors for postoperative infection.When these results were seen in the light of the accuracy and predictive value indices, CRP again reaffi rmed its position as a good predictor of infection on both the second and the fi fth postoperative days.Taking all this in account, the data suggest that CRP might be the most important predictive factor for postoperative infection.However, these results should be considered with caution, because the power analysis for some parameters, such as for transferrin, was very low.Nevertheless, the power analysis for both CRP and IL-6 was greater than 90%.][29] Both high NNIS scores and the presence of malignant disease were clinical factors signifi cantly associated with longer LOS by univariate analysis.However, in the multivariate analyses, only malignant conditions were associated with prolonged hospitalization.The nutritional state was only marginally predictive in this analysis.Again, because of the small number of cases in this study, these data should be considered with caution, because previous studies have demonstrated the value of nutritional status assessment in predicting complications and LOS. 9,24,25The acute-phase proteins and IL-6 did not correlate with longer LOS either during the operation or during the early postoperative course.However, the data from the fi fth postoperative day strongly correlated IL-6 and albumin with longer LOS.In this context, maintenance of a low level of albumin or a high level of IL-6 by the fi fth postoperative day was found to be a predictive factor for longer LOS.In the same way, persistence of low serum levels of CRP, prealbumin or transferrin was prognostic for longer hospitalization.The multivariate analysis showed, however, that transferrin on the second postoperative day and both albumin and IL-6 on the fi fth postoperative day were the most powerful variables associated with LOS.These fi ndings suggest that it is possible to create a protocol to show which are the best markers for predicting LOS, according to the postoperative day.Thus, these results call for further investigations on this issue.
Another important finding from this study was the confi rmation that the immuneinfl ammatory response evaluation may be of more importance than simple diagnosis of the nutritional status, for evaluating the risk of postoperative morbidity.][8]17,23,26 Patients who are candidates for major operations on the abdominal cavity deserve attention because various factors, including previous bacterial infections, malnutrition and malignancies, for instance, may primarily trigger an immune system response. 23Patients who are more prone to complications are also more reactive to operative trauma than o thers with a normal immune response.In this context, a smaller bacterial contamination may provoke an intense immune-infl ammatory response with greater release of cytokines and greater acute-phase protein response.The common pathway for such cases is their predisposition to postoperative infections, intense SIRS and MOSD. 23

CONCLUSIONS
In summary, the overall results from this study have clearly shown that the acute-phase response does exist and is greater in cases of postoperative infections.Moreover, our data suggest that changes detected early in the evolution of either acute-phase proteins or IL-6 may alert the surgeon to the development of insidious infection that may not have been clinically evident until then.This is crucial and could change the management for many patients.Out of all the variables studied, CRP was the most important acute-phase protein for predicting postoperative infection.Another possible conclusion is that NNIS scores are an important instrument for predicting posto perative infections.Although NNIS scores, in association with other acute-phase proteins such as CRP, albumin and prealbumin, and with IL-6, were not the most effective tool, they may enable more accurate prognoses for postoperative infection.With regard to predicting longer LOS, it was found that malignant condition, transferrin, albumin and IL-6 were generally useful.

Figure 1 .
Figure 1.Evolution of various acute-phase proteins (A-D) and interleukin-6 (E) in patients with postoperative infections (solid lines) or without postoperative infections (dotted lines).

Table 1 .
Demographic and clinical characteristics of the 32 patients at baseline evaluation, before the gastrointestinal tract operation

Table 2 .
Gastrointestinal operations performed and infectious morbidity incidence according to infection site

Table 3 .
Univariate analysis by logistic and linear regression of preoperative and intraoperative variables, according to presence of postoperative infection or length of hospital stay (LOS) CI = confi dence interval; ASA = American Society of Anesthesiologists score, NNIS = National Nosocomial InfectionSurveillance score.

Table 4 .
Evolution of the acute-phase proteins and interleukin-6 (IL-6) in operated patients with and without infection.Data are expressed as means and 95% confi dence intervals, with repeated-measurement analysis of variance (ANOVA) using length of stay as a covariable POD = postoperative day.

Table 6 .
Validity parameters for perioperative variables with statistical signifi cance from multivariate analysis, for predicting postoperative infection

Table 5 .
Multivariate analysis by logistic and linear regression of perioperative variables, with statistical signifi cance according to the presence of postoperative infection or length of hospital stay Sao Paulo Med J. 2007;125(1):34-41.