Thrombocytopenia as an important determinant of poor prognosis in patients with pyogenic liver abscess: a retrospective case series

Background Thrombocytopenia and poor prognosis in severe conditions are associated. However, the clinical significance of thrombocytopenia in pyogenic liver abscess (PLA) has not been evaluated. Objective To evaluate the association between thrombocytopenia and the prognosis of patients with PLA. Methods A consecutive case series of 458 adult patients with PLA hospitalized at Tongji Hospital (Wuhan, China) between October 2011 and June 2021 was included in this cross-sectional analysis. Patient data were compared between the thrombocytopenia and non-thrombocytopenia groups. Multivariate logistic regression, receiver operating characteristic (ROC) curve and propensity score -matched analyses (PSM) were performed. Results Of the 458 patients with PLA, 94 (20.5%) developed thrombocytopenia, 19 (4.1%) developed septic shock, 14 (3.1%) were admitted to the ICU, and 15 (3.3%) died during hospitalization. Thrombocytopenia was independently associated with shock (95%CI = 3.529–57.944, P < 0.001), ICU admission (95%CI = 1.286–25.733, P = 0.022), and mortality (95%CI = 1.947–34.223, P = 0.004) in multivariate regression analysis. ROC analysis showed that thrombocytopenia may be an identified marker of shock [area under the ROC curve (AUC), 0.8119; cut-off, 92.50; P < 0.0001], ICU admission (AUC, 0.7484; cut-off, 82.50; P < 0.0015), and mortality (AUC, 0.7827; cut-off, 122.50; P < 0.002). These findings remained consistent across 86 pairs of patients analyzed for PSM analyses. Conclusions Thrombocytopenia is an independent risk factor for poor prognosis in PLA and patients may be more prone to adverse outcomes.


Introduction
Pyogenic liver abscess (PLA) is an infectious disease caused by a microbial infection that leads to liver necrosis, septic shock, and other serious consequences (1). The disease has an acute onset and complex condition, and missing the best treatment time may lead to various complications that seriously harm the health and quality of life of PLA patients (2).
The incidence of PLA is higher in Asian countries, which can reach 12-18 cases per 100,000 people per year, with a mortality rate of approximately 2%-31% (3,4).Therefore, prognostic markers should be identified to provide more aggressive and timely resuscitation for patients and plan effective treatment for patients with PLA to improve prognosis.
Thrombocytopenia is the most common hemostatic disorder in critically ill patients, with a prevalence of approximately 40%-53% (5). Previous studies have reported that platelet counts (PLT) regulate inflammation by controlling tissue integrity, and protecting against infection (6). Hence, PLT is considered to be valuable in predicting some disease outcomes, such as sepsis (7), COVID-19 (8,9), and cancer (10). Zhou (11) demonstrated that patients with severe thrombocytopenia had more blood transfusions, more days requiring advanced support, and worse outcomes than those in the normal group. Orak et al. (12). retrospectively studied 330 patients diagnosed with sepsis in the emergency department and found that the PLTs were lower in patients who died than in the survivors. What's more, platelets have been used as prognostic markers for other hepatobiliary duct diseases, such as cholecystitis/ cholangitis, which are closely associated with PLA (13)(14)(15). PLA is an inflammatory disease, and we hypothesized that thrombocytopenia may be also associated with its prognosis, which is worthy of our in-depth and detailed exploration and research.
To date, many studies have demonstrated a number of risk factors related to the prognosis of liver abscesses, such as malnutrition, pleural effusion, fever, multiple organ dysfunction syndrome (MODS), presence of gas formation, the size of abscess, and microbiology (16)(17)(18)(19)(20). However, few studies have explored the direct association between thrombocytopenia and PLA prognosis. If PLTs, as a common laboratory indicator that is easy to detect, could help to recognize patients with PLA at highrisk for adverse outcomes, it would have even greater clinical benefits. This study was conducted to explore the relationship between thrombocytopenia and PLA prognosis to provide a new and convenient prognostic marker of PLA.

Study design and participant
This was a retrospective collection of a consecutive case series of 458 patients with PLA admitted to Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology, between October 2011 and June 2021.
Patients with PLA were identified by retrospectively searching the medical record code (search term "pyogenic liver abscess", ICD 10 code = K750) in the medical record department. The inclusion criteria for analysis were as follows: (1) diagnosis of PLA according to the following diagnostic criteria: a. imaging examination found liver abscess lesion (either magnetic resonance imaging (MRI), computed tomography (CT), or ultrasound (US); b. the patient had fever, chills, liver percussion pain, and other clinical manifestations and signs; c. positive bacterial culture; d. the lesions subsided after antibiotic treatment; e. liver biopsy or surgical pathology confirmed; a was essential for diagnosis, and b-e were non-essential diagnostic criteria; (2) age ≥18 years; (3) complete key laboratory results and imaging data. The following were the exclusion criteria: (1) incomplete clinical data; (2) amebic liver abscess or parasitic liver abscess; and (3) underlying conditions at risk for thrombocytopenia (hematologic malignancies, chemotherapy, cirrhosis, and chronic heart failure). The management of PLA was in accordance to the Expert consensus on multidisciplinary management of intra-abdominal infections (21) and Quality Improvement Guidelines for Percutaneous Drainage/Aspiration of Abscess and Fluid Collections (22). Typically, the duration of parenteral antibiotic treatment of liver abscesses is between 4 and 6 weeks. Interval imaging was typically performed after a specific period of antibiotic treatment. The interval between imaging studies also varied depending on the patient's clinical progress, response to therapy, and the physician's judgment.
The study was approved by the ethics committee of Tongji Hospital of Huazhong University of Science and Technology and conducted in accordance with the Declaration of Helsinki (TJ-IRB20221240). This study was exempt from informed consent because of its retrospective design.

Data collection
Data collected from the electronic medical record system in hospital included patients' demographic characteristics (age and sex), comorbidities (diabetes mellitus, hypertension, chronic respiratory disease, cardiovascular disease, gastrointestinal and hepatobiliary diseases, and malignancy), clinical symptoms and signs (fever, nausea and vomiting, abdominal distension or pain, diarrhea, fatigue), vital signs on admission (temperature, heart rate, respiration rate, and blood pressure), lab tests upon admission [blood routine test, liver function, renal function, serum levels of inflammatory markers, coagulation function, serum lipid parameters, random blood glucose, troponin I, n-terminal pro-brain natriuretic peptide(NT-proBNP), and other related indexes], bacterial cultures, treatments, imaging findings, serious complications, and adverse outcomes including septic shock, acute renal failure, acute hepatic failure, heart failure, myocardial infarction, pulmonary edema, pulmonary infection, acute respiratory distress syndrome (ARDS), pleural effusion, ICU occupancy, and death.

Definitions
Thrombocytopenia was defined as a PLT less than 125 × 10 9 /L (laboratory reference range 125-350 × 10 9 /L). Patients with liver abscess were categorized into the thrombocytopenia and nonthrombocytopenia groups based on PLTs less than or above 125 × 10 9 /L, respectively. Septic shock was defined as acute circulatory failure with uncorrectable hypotension unexplained by other causes, despite sufficient fluid resuscitation (23). Acute renal failure was diagnosed when serum creatinine above 176 µmol/L, or an absolute increase was greater than 44 µmol/L (24). Acute hepatic injury was defined by WHO diagnostic criteria as alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL) where any one of these markers was more than 1.25 times the reference value upper limit. Heart failure was defined in accordance with the guidelines of the Heart Failure Association of the European Society of Cardiology (25). Myocardial infarction was defined as a serum level of hypersensitive cardiac troponin I (hs-cTnI) >34 pg/ml (26). Pulmonary edema is mainly diagnosed on the basis of pulmonary imaging findings. ARDS was defined according to The Berlin Definition of Acute Respiratory Distress Syndrome (2012) (27).

Statistical analysis
The mean ± standard deviation was used to express the continuous variable of normal distribution, and the median (quartile distance) was used to express the continuous variable of skewness distribution. Categorical variables were expressed as frequencies and percentages (%). For comparison between the two groups (thrombocytopenia group vs. non-thrombocytopenia group), continuous variables were analyzed by Student's t-test or Mann-Whitney U test, and categorical variables were analyzed by Fisher's exact or chi-squared test. Univariate logistic regression analysis was performed to ascertain the possible risk factors that showed a relationship with shock, ICU admission, and mortality. Variables without collinearity were selected for the multivariable analysis to assess the association between thrombocytopenia and PLA, taking into consideration previous research findings (28), clinical implications, and significant variables identified in the univariate logistic regression analysis. Finally, age, gender, hemoglobin (Hb), TBIL, creatinine, white blood cell, ARDS, presence of gas and pleural effusion were included in the multivariate analysis. The distribution of isolated microorganisms of the total or two groups (thrombocytopenia group vs. nonthrombocytopenia group) is presented in Figure 1. Receiver operating characteristic curve (ROC) was plotted to evaluate the discriminatory performance for adverse outcomes (shock, ICU admission, and mortality) of PLA according to the value of the area under the ROC curve (AUC). To adjust for additional confounding factors, a propensity score matching analysis was performed. Standardized mean differences and t-tests were used to ensure baseline demographics were comparable between the two groups. After matching, the standardized mean differencece of all matching factors in the t-test was not statistically significant. The matched variables included: age, sex, fever, nausea and vomiting, abdominal distension or pain, diarrhea, fatigue and muscle pain, palpitation, cough and sputum, dizziness or headache, SBP, DBP, heartrates, diabetes mellitus, hypertension, cardiovascular disease, malignancy, liver and gallbladder stones, viral hepatitis, fatty liver disease, presence of gas, hemoglobin, total bilirubin, direct bilirubin, ALT, AST, Profiles of isolated microorganisms in patients with pyogenic liver abscess.
Patients were categorized, based on PLT, into either the thrombocytopenia group (n = 94) or non-thrombocytopenia group (n = 364). Compared with the non-thrombocytopenia group, the thrombocytopenia group was more likely to report nausea, vomiting, fatigue, muscle pain, palpitations, and dyspnea and had lower blood pressure and higher heart rates for vital signs at admission (P < 0.05). However, no significant differences in size of abscess, presence of gas, or multiloculation were observed between the two groups.

Baseline laboratory parameters of patients with PLA
The thrombocytopenia group showed significantly decreased levels of lymphocyte, albumin, total cholesterol, high density lipoprotein-cholesterol (HDL-C), low density lipoproteincholesterol (LDL-C), and fibrinogen, as well as increased levels of inflammatory markers like C-reactive protein, hepatic function indicators (TBIL, direct bilirubin, ALT, and AST), kidney function indicators (urea nitrogen and creatinine), coagulation function indicators (PT and D-D dimer), cardiac parameters (NT-proBNP and cTnI), and others, such as random blood glucose, compared with the non-thrombocytopenia group ( Table 2, P < 0.05).

Profiles of isolated microorganisms of patients with PLA
Of the 458 patients with PLA, 194 (42.4%) showed positive bacterial culture results. Twenty-six (12.37%) of the cases were multiple bacterial infections. Among them, 22 cases were cultured with two kinds of bacteria, and 4 cases were cultured with three or more kinds of bacteria. Additionally, there were 4 cases of bacterial infection with fungal infection. As shown in Figure 1, Klebsiella pneumoniae was the most common pathogenic microorganism in both the thrombocytopenia and non-thrombocytopenia groups, accounting for 61.86% of all patients, followed by Escherichia coli (17.01%). Streptococcus (7.73%), Staphylococcus (5.15%), and Enterococcus (4.64%) accounted for a certain proportion. There was no difference in the bacterial culture composition ratio between the thrombocytopenia and non-thrombocytopenia groups (P = 0.318).

Treatment and clinical outcomes of patients with PLA
As shown in Table 3, 128 patients (27.9%) received conservative treatment with antibiotics alone, 278 patients (60.7%) received antibiotics with abscess puncture and drainage, and 33 patients (7.2%) received antibiotics with surgery. In these areas, there were no statistical differences between the thrombocytopenia and non-thrombocytopenia groups regarding the treatment received. Albumin infusion (P = 0.002), glucocorticoid use (P < 0.001), and antiviral drug use (P = 0.010) were significantly higher in the patients with thrombocytopenia than in those without thrombocytopenia.
Furthermore, we compared baseline data of liver abscess patients grouped by death, ICU admission, and shock, and the results are presented in the Supplementary (Tables S1-S3).

Propensity score-matched analysis
The PSM analysis resulted in a total of 86 pairs of study participants. Due to the limited number of subjects, several analyses were constrained. Therefore, we combined three outcome events (shock, ICU admission, and death) as "adverse outcomes" for analysis. It is noteworthy that the baseline features were well balanced between the two groups, as indicated in Table 1. Consistent with findings in the overall population, the PSM analysis demonstrated an independent association between thrombocytopenia and adverse outcomes in patients with PLA.

Prognostic value of PLT
As shown in Figure 2, ROC analysis was performed to evaluate the relationship between PLT and PLA prognosis. For the discriminative ability of shock in patients with PLA, the AUC value of the PLT was 0.8119 (95% CI = 0.6773-0.9465, P < 0.0001) (Figure 2A). The optimal cut-off (the value of PLT when the Youden index reaches the maximum) was 92.5 × 10 9 /L) with a corresponding sensitivity and specificity of 89.07% and 73.68%, respectively. The AUC value of PLT as an identified marker of ICU admission was 0.7484 (95% CI = 0.5958-0.9010, P = 0.0015) ( Figure 2B), and the cut-off value was 82.50 × 10 9 /L with a corresponding sensitivity and specificity of 90.54% and 64.29%, respectively. The AUC of PLT for the discriminative ability of mortality was 0.7827 (95% CI = 0.6539-0.9295, P = 0.002) ( Figure 2C) and the cut-off value was 122.50 × 10 9 /L with a corresponding sensitivity and specificity of 81.53% and 73.33%, respectively. In addition, in the PSM analysis, the AUC value of PLT was 0.7209 (95%CI = 0.5887-0.8531, P = 0.0011) as an indicator of adverse outcomes in PLA patients ( Figure 2D). The cut-off value was 82.5 × 109/L, the sensitivity was 79.47%, and the specificity was 66.67%. Bold value means statistically significant (P < 0.05). "Adverse oucomes" included shock, ICU admission, and death. PSM, propensity score-matched; CI: confidence interval; ALT, alamine aminotransferase; PT, prothrombin time; CRP, C-reactive protein; PCT, procalcitonin; ARDS, acute respiratory distress syndrome.

Discussion
In the study involving 458 patients with PLA, thrombocytopenia was found in 20% of cases. Patients with thrombocytopenia were more likely to have deranged (including inflammatory markers and hepatic and renal function indicators were deranged), increased risk of complications, higher ICU admission rates, and elevated mortality. ROC analysis indicated that PLT, an inexpensive and easily obtained biomarker, may be a good candidate to meaningfully identify the poor prognosis of PLA.
Platelets are small anucleate circulating cells that are increasingly recognized as key effector cells that modulate host responses to inflammation and infections (29,30). Platelets play a key role in innate immunity by interacting with other immune cells, through multiple receptors on their surface, such as Tolllike receptor 4 (31). Thrombocytopenia is a common complication of infection-related diseases. Platelets can interact with pathogens through a variety of platelet damage-related molecular pattern receptors, with the most common causative agent being viruses (28%), followed by bacteria (28%), and fungi (15%) (31). Current studies indicate that the causes of infection or sepsis-associated thrombocytopenia are complex, not fully elucidated, and may involve several factors, for example, a decrease in platelet production. The liver is the most vital site for the synthesis of thrombopoietin (TPO), a crucial plateletstimulating factor that modulates platelet production (31). Liver injury due to PLA may result in an absolute or relative deficiency in TPO levels in these patients, leading to thrombocytopenia. In addition, various inflammatory cytokines can cause thrombocytopenia by destroying stem cells in the bone marrow (32, 33). This is followed by increased platelet destruction and consumption. Infections can stimulate platelet activation and aggregation through various pathways leading to thrombocytopenia (34, 35). Bacteria, such as Escherichia coli, Staphylococcus aureus, and Streptococcus pneumonia, can directly stimulate platelet activation and platelet-leukocyte aggregation by binding to and activating platelet receptors such Frontiers in Surgery as toll-like receptors or by participating plasma proteins related pathways (36). The release of diverse antimicrobial peptides can give rise to cell destruction and tissue damage. Endothelial cell injury and its release of inflammatory factors can also induce platelet activation and aggregation, increase platelet consumption or thrombosis (31,(37)(38)(39), and lead to thrombocytopenia. Studies have shown that the release or upregulation of sialidase during infection leads to the hydrolysis of sialic acid, a natural sugar acid that protects platelets from destruction (40)(41)(42), thereby causing thrombocytopenia. Additionally, treatmentrelated drug induction has been shown to cause thrombocytopenia. For example, vancomycin (43), linezolid, cephalosporin, and chloroquine phosphate (44,45) can produce antiplatelet antibodies that induce increased platelet destruction. Finally, thrombocytopenia was found to be associated with a lower protein concentration and higher fluid balance, which may suggest a hemodilution effect caused by heavy fluid rehydration during treatment (46). The pathophysiological mechanism of adverse outcomes in thrombocytopenia remains unclear. Possible reasons for this are as follows. Platelets fulfill vital functions in microbial host defense, angiogenesis and tissue remodeling, as well as wound healing (45-47). The higher incidence of adverse outcomes in the thrombocytopenia group can be directly explained by the decreased antimicrobial defenses of thrombocytopenia and changes in platelet function. The interaction between platelets, white blood cells, and the endothelium can result in endothelial dysfunction, leading to inflammation and thrombosis. This process is primarily mediated by platelets, which serve as the fundamental components driving this pathological cascade (7). The specific process is that platelets can be activated by microbial components or inflammatory mediators, leading to their interaction with neutrophils (48). This interaction exacerbates systemic inflammatory responses, triggers the release of inflammatory cytokines, stimulates endothelial cells, and dysregulates host defense responses by inhibiting or activating relevant signaling pathways (49). Additionally, platelets may play a chief part in the pathophysiology of disseminated intravascular coagulation (DIC) and MODS, together with the activation of endothelial cells and leukocytes (7). Although thrombocytopenia is often referred to as a condition of hypocoagulation, the reality is complex, and patients are at a potential risk for bleeding and thrombosis (50). Although different types of infection and definitions of thrombocytopenia influenced the outcome analysis, PLTs may be used to indicate infection severity.
As a suppurative infection of the liver parenchyma, PLA may be caused by a wide variety of bacteria, including biliary tract, portal vein, blood-borne or cryptogenic, and adjacent structure infections (51). Klebsiella pneumoniae was identified as the predominant bacterium in patients with PLA in our study, which aligns with findings from previous studies (52,53). It has been reported that multibacteremia occurs in 14%-55% of PLA cases (54), a proportion slightly lower than what we observed in our study. Our study found no significant difference in bacterial culture results between the thrombocytopenia and nonthrombocytopenia groups, which was coincident with the findings of Johansson, et al. (36). However, some studies have demonstrated that thrombocytopenia is more common in gramnegative bacterial infections (55) and its duration is longer in gram-negative bacterial or fungal infections than in grampositive bacteria (56). Unfortunately, this study lacked data on patients' platelet dynamics and did not draw similar conclusions. Owing to the differences in the distribution of microorganisms in previous studies, prevalence and duration of thrombocytopenia based on different bacterial infections remain unclear.
As mentioned above, bacterial components can directly activate platelets and release sialidase, thereby causing platelet destruction. During the deterioration of PLA, a large number of inflammatory cytokines are released, and endothelial damage occurs, triggering the activation of the coagulation cascade, which in turn result in platelet activation (7). And activated platelets lead to the production of thrombin by promoting the release of procoagulant factors, which further aggravates the development of the disease and leads to septic shock, liver and kidney function injury, and even multiple organ dysfunction in severe cases. Thrombocytopenia has been proved to be closely related to the increase of ICU occupancy rate, length of stay and mortality in critically-ill patients (57). A prospective study (58) also showed that changes in platelet mitochondrial function occurred early in patients with septic shock and were independently associated with the development of organ failure. Because thrombocytopenia induces organ damage, including renal failure, acute lung injury, and septic shock, it may indirectly lead to death (7,57). Similarly, our findings showed that thrombocytopenia patients with PLA had nearly 7 times the risk of death, 12 times the risk of shock, and 11 times the risk of ICU transfer compared to patients without thrombocytopenia. These findings suggest that thrombocytopenia may be associated with severe infection and poor prognosis.
Additionally, previous research has indicated that the size of the abscess, gas production, and multiloculation are closely associated with the prognosis of PLA (19,20). However, in our study, we did not observe significant differences in these variables between the two groups. Only a positive correlation between gas production and mortality was identified through univariate regression analysis. This discrepancy may be attributed to several factors, such as our relatively small sample size, missing data, confounding factors, or the heterogeneity of different ethnic groups.
In recent years, many advances have been made in the treatment of PLA. Percutaneous aspiration/drainage under US or CT guidance plus long-term antibiotic therapy has replaced traditional surgical drainage as the cornerstone of treatment (22). As in this study, most patients received antibiotics combined with puncture and drainage (60.7%). The prognosis of PLA is influenced by early diagnosis, as delayed detection can lead to complications such as liver and kidney failure, as well as respiratory failure, resulting in poor outcomes (2). However, thanks to the continuous advancements in treatment approaches, the mortality rate associated with PLA has significantly declined in recent years. Li et al. 10.3389/fsurg.2023.1192523

Frontiers in Surgery
The study's limitations are as follows: (1) This is a retrospective study and lacks follow-up of patient parameters over time to directly establish absolute causality based on the results of this study. (2) Although we continuously recruited participants at our hospital and our participants were representative of local inpatients, they were not representative of the entire regional population of patients with PLA. (3) The etiological specimens were obtained from the blood or fester of patients, and different specimens for detection may have led to certain errors. (4) Data deficiencies include no details on the duration of antibiotic use, interval imaging, and a small sample size of patients reporting abscess size. (5) The data collected in this study were all indicators obtained on hospital admission without dynamic observation. In the future, a large-scale prospective study should be conducted to validate our conclusions.

Conclusion
Thrombocytopenia, an independent risk factor, was significantly associated with septic shock, ICU admission, and mortality in patients with PLA. PLT, a rapid and easy clinical laboratory index, can assist risk assessment and hierarchical management in patients with liver abscess and help identify those patients with a poor prognosis early, important for timely and appropriate intervention to improve poor outcomes. Larger sample sizes and prospective studies are required to confirm this conclusion.

Data availability statement
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.

Ethic statement
The Ethics Committee of Tongji Hospital of Huazhong University of Science and Technology had approved this study, which was exempt from informed consent because of its retrospective design (TJ-IRB20221240).

Author contributions
ZlL, LL and SzL designed the study. SzL, ShL and MH collected data. SzL, ShL and LL performed statistical analyses and generated figures and tables. ShL, LL and SzL wrote the manuscript. TY, ShL and SH read and approved the final manuscript. All authors have approved the final version of the manuscript.