ABSTRACT
Aim
The aim of this study was to evaluate the factors affecting the prognosis of sepsis in patients admitted to the intensive care unit (ICU).
Materials and Methods
We retrospectively included all adult patients admitted to the ICU, who were diagnosed with sepsis according to the Sepsis 3 criteria between September 2013 and February 2021. Demographic, clinical data and laboratory results were recorded.
Results
Of the 245 patients in the ICU, 100 (40.8%) died during the 30-day follow-up. In univariate logistic regression analysis, Sequential Organ Failure Assessment (SOFA) score, vasopressor need, immunosuppressive treatment, neutropenic fever, hematological malignancy, pneumonia, urinary tract infection, lactate, ferritin, lactate dehydrogenase (LDH), and albumin levels were found to be independently associated with mortality. When evaluated in terms of prognostic significance in ROC curve, the optimal cutoff values for 30-day mortality were 7 for SOFA score (AUC=0.713, p<0.001), 1518 µg/L for ferritin (AUC=0.732, p<0.001), 324 U/L for LDH (AUC=0.593, p=0.035), and 2.9 g/dL for albumin (AUC=0.632, p=0.001), in mortality. Using these values, in multivariate logistic regression analysis, we determined that a SOFA score >7 [odds ratio (OR): 95% confidence interval (CI): 9.66 (1.16-80.82), p=0.036], history of immunosuppressive treatment [OR 95% CI: 12.41 (1.45-106.17), p=0.021], and ferritin levels >1518 µg/L [OR 95% CI: 9.46 (1.36-65.79), p=0.023] were independent risk factors for 30-day mortality.
Conclusion
In our single-center study, serum ferritin level was determined to be a valuable prognostic biomarker in patients with sepsis.
INTRODUCTION
Sepsis is characterized by an excessive immune response to an infectious agent, leading to multiple organ failure with a high mortality rate1. Sepsis is a major concern in public health due to its widespread impact, with over 19 million individuals diagnosed each year globally2. Despite advances in supportive care and intensive care technologies, the mortality rate in septic shock remains approximately 40%3. Sepsis involves the release of high levels of proinflammatory cytokines due to microbial agents, accompanied by the release of anti-inflammatory cytokines. The initial hyperinflammatory phase can lead to early mortality4. Numerous studies have been conducted on various laboratory parameters to evaluate the severity of the increased inflammatory response and to predict mortality. Mortality predictor parameters can help identify patients who, in addition to supportive therapies, may benefit from potentially effective additional treatments by predicting early mortality. Recently, in addition to commonly used parameters such as C-reactive protein (CRP), procalcitonin, neutrophil-to-lymphocyte ratio (NLR), and platelet-to-lymphocyte ratio (PLR), studies have been conducted on numerous new markers such as N-terminal pro b-type natriuretic peptide (NT-proBNP), ferritin, and presepsin. Some of these studies have yielded significant results for these parameters5-10. However, none of these parameters have been accepted as additional mortality predictive markers in scoring systems such as Sequential Organ Failure Assessment (SOFA) and Acute Physiology and Chronic Health Evaluation (APACHE) according to international guidelines11. In this study, we aimed to evaluate the capability of certain inflammatory markers to predict prognosis independently of general patient characteristics and parameters already associated with disease severity and mortality, such as the SOFA score and serum lactate levels.
MATERIALS AND METHODS
Patient Selection and Study Design
This retrospective study was conducted in a single-center internal medicine intensive care unit (ICU) between September 2013 and February 2021. The Institutional Ethical Review Board of Ege University Hospital approved the study (decision no: 21-6.1T/54, date: 10.04.2016). This study was conducted in accordance with good clinical practice guidelines and adhered to the principles of the Declaration of Helsinki. Patients or their relatives provided written informed consent. We included all adult patients (≥18 years old) diagnosed with sepsis. The diagnosis of sepsis and septic shock was made according to the Sepsis-3 criteria12.
Demographic characteristics and clinical features were extracted from patients’ medical records. In addition, the following laboratory parameters were obtained on admission and 48 h after hospitalization: neutrophil count, lymphocyte count, platelet count, CRP at admission and 48 h, procalcitonin, albumin, ferritin, NT-proBNP, troponin-T, lactate, and LDH. The primary endpoint of the study was 30-day mortality. Secondary endpoints included factors associated with prognosis.
Statistical Analysis
Descriptive statistics were used to summarize the data. For continuous (numerical) variables, depending on the distribution, either mean±standard deviation or median, minimum, and maximum values were presented. Categorical variables are summarized as counts and percentages. The normality of the numerical variables was assessed using the Shapiro-Wilk, Kolmogorov-Smirnov, and Anderson-Darling tests. For comparisons between two independent groups, the Independent Samples t-test was used when numerical variables were normally distributed, and the Mann-Whitney U test was used when they were not. Risk factors affecting 30-day mortality were investigated using univariate and multivariate logistic regression models. To identify the ideal cutoff level to evaluate 30-day mortality, receiver operating characteristic curve (ROC), in which the Youden J index was considered in determining the threshold value, was performed. Statistical analyses were performed by Jamovi project (2020), Jamovi (version 1.8.4.0) [computer software], and JASP with a significance level set at 0.05 (p value).
RESULTS
During the study period, 245 patients diagnosed with sepsis were followed up in the ICU. The 30 day-mortality rate was 40.8%. Demographic information, comorbid conditions, and infection sites, laboratory data along with the differences between the survivors and non-survivors are presented in Table 1. Within comorbidities, vasculitis, hematologic malignancy, and the use of immunosuppressive therapy were significantly more frequent in the non-survivor group (p values 0.009, <0.001, 0.004 respectively). Among the sources of infection, pneumonia was more commonly observed in the non-survivor group (p=0.001), whereas urinary tract infections were more frequent in the survivor group (p=0.003). Neutropenic fever, high SOFA scores, and the need for vasopressors were more common in the non-survivor group (p values <0.001, <0.001, 0.004 respectively). Among baseline data, albumin, LDH, lactate, and ferritin levels at admission as well as the changes in CRP and LDH levels at the 48 h of hospitalization compared with their admission values were found to be statistically different between the non-survivor and survivor groups (p values 0.002, 0.034, 0.005, <0.001, 0.002, 0.012, respectively).
Logistic regression analysis was performed to determine the parameters that were significantly different between the survivor and non-survivor groups. In the evaluation based on 30-day mortality outcomes, univariate regression analysis identified SOFA score, use of immunosuppressive therapy, neutropenic fever, pneumonia, urinary tract infection, lactate, albumin, ferritin, and LDH as statistically significant factors. CRP D2-D0, LDH D2-D0, NLR, PLR, and NPAR were not found to be statistically significant in univariate analysis. Because the SOFA score and lactate levels are already established parameters used in determining the severity of sepsis and in diagnosing septic shock, model-1 was formed incorporating these along with other patient characteristics found to be associated with mortality. Ferritin, LDH, and albumin were added to the existing model to perform a multivariate regression analysis. After the modeling yielded inconclusive results, ROC curve was conducted on the relevant parameters (Figure 1). The cutoff values obtained from the ROC curve are shown in Table 2. Using the cutoff values obtained from the ROC curve, logistic regression analysis was repeated (Table 3). Initially, the univariate logistic regression model revealed that the following factors individually had a significant impact on 30-day mortality: a SOFA score above 7, the need for vasopressors at admission, administration of immunosuppressive therapy, the presence of neutropenic fever, hematologic malignancy, pneumonia, urinary tract infection, an albumin level below 2.9 g/dL, an LDH level over 324 U/L, and a ferritin level exceeding 1.518 µg/L. When examining the results of the multivariate logistic regression analysis, it was observed that a SOFA score above 7 [odds ratio (OR): 95% confidence interval (CI): 9.66 (1.16-80.82) p=0,036], history of immunosuppressive therapy [OR 95% CI: 12.41 (1.45-106.17) p=0.021], and a ferritin level over 1.518 µg/L [OR 95% CI: 9.46 (1.36-65.79) p=0.023] were independent risk factors for 30-day mortality (p=0.036, p=0.021, and p=0.023).
DISCUSSION
In the current study, which included patients with sepsis, the 30-day mortality was 40.8%. High SOFA score (>7), history of immunosuppressive therapy, and ferritin levels >1.518 µg/L were significantly associated with 30-day mortality.
Ferritin is a protein composed of heavy and light chain structures that are responsible for iron binding and storage. It prevents the free circulation of iron in the body, thereby protecting proteins, lipids, and DNA structures from potential iron toxicity. Serum ferritin is also known as an acute phase reactant, which is regulated by proinflammatory cytokines. An increase in ferritin levels is often observed in inflammatory processes following the stimulation of heme oxygenase-1. This rise in ferritin levels plays a protective role aimed at preventing oxidative damage that occurs during inflammatory processes13-15. In adults, hyperferritinemia can be seen in various conditions, including hemophagocytic lymphohistiocytosis, in patients undergoing hemodialysis, having hemochromatosis, receiving frequent transfusions, having liver failure, antiphospholipid antibody syndrome, adult-onset still’s disease, and patients with sepsis. Each of these conditions has unique mechanisms and implications related to elevated ferritin levels, emphasizing the importance of ferritin as a marker in diverse clinical scenarios16. High ferritin levels are increasingly being recognized as valuable biomarkers for the prognosis of several conditions, including cancer, connective tissue diseases, and notably, Coronavirus disease-2019 infection17, 18.
Studies on the relationship between ferritin levels and mortality in sepsis have particularly been conducted in the pediatric patient group. In a study conducted on pediatric patients, it was found that a ferritin level greater than 500 µg/L significantly increased the mortality risk by 3.2 times6. In another study conducted with pediatric patients, it was found that a ferritin level above 3.000 µg/L was associated with a 4.32-fold increase in the risk of mortality19. Studies conducted on elderly inpatients with ferritin levels above 1.000 µg/L have demonstrated the prognostic significance of high ferritin levels in sepsis and their importance in the recognition of malignancies20. In another study conducted on sepsis patients, ferritin levels exceeding 4.420 µg/L g/ml were found to be associated with a diagnosis of macrophage activation-like syndrome and related to 28-day mortality. Furthermore, a reduction of less than 15% in ferritin levels within three days compared to the initial value has been found to be associated with 28-day mortality21. In a recent study among adult septic patients, high ferritin levels were found to be associated with mortality. In this study, the cutoff value for ferritin was determined to be 591 µg/L22. In our study, the cutoff value for ferritin was found to be 1.518 µg/L. The different cutoff values obtained in these studies may be attributable to variations in the comorbid conditions of the patients. As previously mentioned, patients with malignancies, those undergoing hemodialysis, and those receiving frequent transfusions tend to have higher ferritin levels compared to other patients16.
In addition to ferritin, univariate analysis in our study also found that low serum albumin levels and high LDH levels were predictors for mortality. In studies conducted on patients with abdominal sepsis, albumin levels below 2.9 g/dL were associated with high SOFA and APACHE scores, although no direct relationship with mortality was established23. In a prospective study investigating the relationship between low albumin levels and 28-day mortality, albumin levels below 2.9 g/dL were identified as an independent risk factor24. In our study, similar to other studies, the cutoff value for albumin was determined to be 2.9 g/dL in the ROC curve; however, in the multivariate analysis, it was not identified as an independent risk factor. In previous studies, LDH levels exceeding the upper limit of local laboratory standards were reported as to be associated with mortality25. In our study, the ROC curve identified a cutoff value of 324 U/L for LDH. However, LDH was not determined to be an independent predictive factor in the multivariate analysis.
Study Limitations
The present study has some limitations. First, this was a retrospective and observational study. Second, it is a single-center study and has limited generalizability. Serial determination of biomarkers will be more useful than single measurements. However, changes were not evaluated in this study.
CONCLUSION
In conclusion, many biomarkers have been studied in the prognosis of sepsis. How to guide the therapy is still a question for the clinicians. The findings of our study suggest that ferritin can be a useful bedside prognostic biomarker with clinical evaluation.