Prognostic value of dynamic procalcitonin (PCT-Δ) for adverse outcomes in patients with pyogenic liver abscess: a retrospective cohort study

Abstract

Objective

To characterize the clinical profile of pyogenic liver abscess (PLA) complicated by sepsis and investigate the prognostic value of dynamic Procalcitonin (PCT-Δ, peak-to-trough difference) for adverse outcomes (primary: death, multiple organ dysfunction syndrome [MODS]; secondary: hospital stay > 14 days) in PLA patients, and to analyze its non-linear association with adverse outcomes.

Methods

This retrospective cohort study enrolled 170 consecutive PLA patients at a tertiary hospital in Guangzhou (January 2020–December 2024). Sepsis was defined per Sepsis-3 criteria (infection + acute organ dysfunction, SOFA score increment ≥ 2 or absolute score > 2) and was assessed throughout hospitalization. Clinical/laboratory data, including PCT levels at admission, peak, and trough during hospitalization, were collected. Restricted cubic spline (RCS) analysis with 3 knots explored non-linear associations between PCT-Δ and adverse outcomes. Receiver operating characteristic (ROC) curves with 1000-iteration bootstrap resampling and calibration curves evaluated the prognostic performance of PCT-Δ, with multivariable logistic regression models adjusted for sex, age, diabetes, biliary disease, malignancy, and abscess size.

Results

Of 170 patients, 72 (42.35%) developed sepsis during hospitalization. The cohort had a median age of 61 years (IQR 51.0–70.0) with male predominance (67.60%). PCT-Δ (median 30.03 vs. 1.28 ng/mL, P < 0.001) was significantly higher in PLA patients with adverse outcomes. RCS analysis revealed a significant non-linear relationship between PCT-Δ and adverse outcomes (P-overall < 0.001, P-nonlinear = 0.002): the risk of adverse outcomes rose rapidly when PCT-Δ was 0–50 ng/mL and slowed above 50 ng/mL. ROC analysis showed robust prognostic performance: Model 1 (PCT-Δ alone) AUC = 0.833 (95% CI: 0.771–0.895); Model 2 (adjusted for covariates) AUC = 0.843 (95% CI: 0.784–0.902). Bootstrap validation yielded a corrected C-statistic of 0.834 (AUC SD = 0.0423), with calibration curves showing excellent agreement between predicted and observed probabilities.

Conclusion

Sepsis complicating PLA is associated with poor outcomes. PCT-Δ is significantly non-linearly associated with adverse outcomes in PLA patients and has good prognostic value for this population. This dynamic biomarker may assist clinicians in risk stratification of PLA patients, particularly those with complicated sepsis. However, the application of PCT-Δ is limited by its time window of measurement, and prospective multicenter validation is needed to confirm its clinical utility.

Clinical trial number

Not applicable.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-026-13086-z.

Introduction

Pyogenic liver abscess (PLA) is a serious intra-abdominal infection characterized by the formation of a pus-filled cavity within the liver parenchyma [1]. It remains a significant cause of morbidity and mortality worldwide, with an estimated incidence ranging from 1.1 to 3.6 per 100,000 person-years, showing a rising trend in recent decades [2]. Despite advances in diagnostic imaging and the widespread adoption of image-guided percutaneous drainage combined with appropriate antibiotic therapy, PLA continues to be associated with severe complications such as sepsis, septic shock, metastatic infections, and multiple organ dysfunction syndrome (MODS) [2–4]. Consequently, early identification of patients at high risk for adverse outcomes is crucial for optimizing clinical management and improving prognosis.

In the context of infectious diseases, biomarkers play a pivotal role in diagnosis, monitoring therapeutic response, and risk stratification. Procalcitonin (PCT) is a well-recognized biomarker for bacterial infections and sepsis, with faster kinetics and higher specificity than C-reactive protein (CRP) [5, 6]. A single measurement of PCT at admission has been widely used to diagnose bacterial infections in PLA patients, but recent evidence suggests that dynamic changes in PCT levels (PCT kinetics) may provide more comprehensive prognostic information than static measurements [7–9].

In sepsis and other severe infections, a failure of PCT levels to decrease significantly after the initiation of appropriate treatment has been consistently associated with poor outcomes, including persistent organ failure and increased mortality [8]. This dynamic assessment may be particularly relevant in complex infections like PLA, where the initial clinical picture can be misleading, and the response to drainage and antibiotics is paramount. However, the specific prognostic value of dynamic PCT (PCT-Δ) in patients with PLA remains underexplored. Most existing studies on PLA prognosis have focused on static clinical scores (e.g., APACHE II, MELD), microbiological findings, or radiological characteristics.

This study therefore aims to investigate the prognostic value of PCT-Δ (peak-to-trough difference during hospitalization) for adverse outcomes in PLA patients, and to explore its non-linear association with adverse outcomes using restricted cubic spline analysis. We hypothesize that PCT-Δ is significantly associated with adverse outcomes in PLA patients, and that this dynamic biomarker can serve as a potential prognostic marker for risk stratification of PLA patients.

Methods

Study design

This study employed a retrospective chart review design, analyzing all hospitalized patients diagnosed with PLA at a tertiary medical center in Guangzhou, Guangdong, between January 2020 and December 2024.

Diagnostic criteria for PLA: (1) Clinical manifestations of infection (fever, chills, right upper quadrant pain); (2) Imaging confirmation of hepatic abscess cavities via abdominal ultrasonography, computed tomography, or magnetic resonance imaging; (3) Laboratory evidence (leukocytosis, neutrophilia, abnormal liver function tests); (4) Positive blood or abscess cultures (or resolution of lesions following empirical antibiotic therapy in cases without microbiological confirmation) [10].

Sepsis definition and assessment: Sepsis was defined only per Sepsis-3 criteria (infection + acute organ dysfunction, SOFA score increment ≥ 2 from baseline or absolute score > 2 for patients without documented baseline) [11]. Sepsis was assessed throughout hospitalization, and the first time the patient met the Sepsis-3 criteria was defined as the time of sepsis onset. For patients without documented baseline SOFA scores, the first SOFA assessment at admission was used as the baseline.

Inclusion criteria: (1) Age ≥ 18 years; (2) Fulfillment of clinical and radiological diagnostic criteria for PLA; (3) Complete clinical and laboratory data, including at least three PCT measurements (admission, peak, trough) during hospitalization. Exclusion criteria: (1) Incomplete medical records; (2) Non-pyogenic liver abscesses (amebic, fungal, parasitic); (3) Missing key data (PCT measurements, outcome data); (4) Death or discharge within 24 h of admission.

Adverse outcomes were stratified as: primary (death, MODS) and secondary (hospital stay > 14 days). PCT-Δ was defined as peak-to-trough in-hospital PCT difference (PCT-max – PCT-min), with measurements recorded at admission (PCT-admit), peak (PCT-max), and trough (PCT-min) during hospitalization.

Data collection

This study systematically documented a comprehensive array of variables, including: (1) Patient demographics (sex, age, body mass index); (2) Comorbid conditions (diabetes mellitus, hypertension, biliary disorders, malignancies); (3) Clinical presentation and physical examination findings; (4) Laboratory parameters (hematological indices, PCT levels, hepatorenal function markers, microbiological culture results); (5) Radiological assessments (abscess location, number, size); (6) Interventions (percutaneous drainage, surgical resection, antibiotic therapy); (7) Clinical outcomes (death, MODS, hospital stay).

PCT levels were measured as part of routine clinical monitoring during hospitalization, with measurements recorded at three distinct time points: admission (PCT-admit), peak in-hospital value (PCT-max), and trough in-hospital value (PCT-min). PCT-Δ was defined as the peak-to-trough difference in PCT levels during hospitalization (PCT-Δ = PCT-max – PCT-min). All laboratory measurements were performed in the clinical laboratory of Sun Yat-sen Memorial Hospital with standard methods.

Statistical analysis

Continuous variables were expressed as mean ± SD (normal distribution) or median (IQR) (non-normal distribution), with group comparisons using independent samples t-tests or Mann-Whitney U tests. Categorical variables were expressed as n (%), with group comparisons using Chi-square or Fisher’s exact tests.

Stratified comparisons were performed between the sepsis and non-sepsis groups for baseline characteristics, laboratory parameters (including PCT and PCT-Δ), and clinical outcomes, to explore clinical features associated with sepsis. Two logistic regression models were constructed to analyze the association between PCT-Δ and adverse outcomes in patients with PLA: Model 1 (unadjusted, including PCT-Δ alone); Model 2 (adjusted for sex, age, diabetes mellitus, biliary tract disease, malignancy, and abscess size). The covariates were selected based on clinical relevance and previous literature on PLA prognosis. Biliary tract disease and malignancy are well-established clinical factors associated with the development of PLA, while abscess size is a key radiological indicator reflecting the severity of PLA. All of these factors have been verified to be associated with adverse clinical outcomes in patients with PLA [12].

Restricted cubic spline (RCS) analysis: RCS analysis with 3 knots (10th, 50th, 90th percentiles of PCT-Δ) was used to assess the non-linear dose-response relationship between PCT-Δ and adverse outcomes. Wald χ² tests were used to evaluate departures from linearity. ROC curve analysis and validation: Receiver operating characteristic (ROC) curves were used to evaluate the prognostic performance of PCT-Δ, with the area under the curve (AUC) as the primary index. Internal validation was performed via 1000-iteration non-parametric bootstrap resampling, and the corrected C-statistic, AUC standard deviation (SD), accuracy, and Cohen’s Kappa were calculated. Calibration curves were used to evaluate the agreement between predicted and observed probabilities of adverse outcomes. Decision curve analysis (DCA): DCA was used to evaluate the clinical utility of the PCT-Δ prognostic model, by comparing the net benefit of the model with “treat all” and “treat none” strategies across different threshold probabilities.

Complete case analysis was used for all statistical analyses (missing data < 5%). All analyses were performed using R 4.3.0 (rms, ggplot2, tidyverse, pROC, rmda packages). A two-sided P < 0.05 was considered statistically significant.

Results

Demographic data and comorbidities

A total of 170 patients with confirmed PLA were included in the final analysis, with no loss to follow-up. All patients had complete PCT measurements and outcome data, forming the study population for all statistical analyses.We summarized the demographic features, symptoms, and underlying conditions of patients stratified into sepsis (n = 72) and non-sepsis (n = 98) groups (Table 1). The cohort comprised 115 males (67.60%) and 55 females (32.40%), with a median age of 61 years (IQR 51.0–70.0). The most common comorbidity was diabetes mellitus (38.00%), followed by biliary tract disease (30.40%). Nausea/vomiting (69.4% vs. 20.4%, P < 0.001) was significantly more common in PLA patients with sepsis than in those without sepsis. No significant differences were found between the two groups in terms of BMI, smoking history, alcohol consumption, or malignancy (all P > 0.05).

Table 1.

Demographics, clinical presentation, and comorbidities in sepsis versus non-sepsis patients with PLA

Characteristic	Overall (n = 170)	Sepsis (n = 72)	non-sepsis (n = 98)	Test Statistic	P value
Age [years, M (P25, P75)]	61(51.00, 70.00)	64(55.50, 73.75)	58(50.00, 68.25)	U = 2727.500	0.012
Gender [n (%)]				χ2 = 2.030	0.154
Male [n (%)]	115 (67.60)	53(73.60)	62(63.30)
Female [n (%)]	55 (32.40)	19(26.40)	36(36.70)
BMI (kg/m2, ± SD)	23.40 ± 3.29	23.44 ± 3.25	23.37 ± 3.32	t = 0.132	0.895
Smoking history [n (%)]	27 (15.80)	10(13.90)	17(17.30)	χ2 = 0.372	0.542
Alcohol consumption [n (%)]	11(6.40)	2(2.80)	9 (9.20)	χ2 = 2.814	0.093
Comorbidities [n (%)]				χ2 = 6.874	0.442
Diabetes mellitus [n (%)]	65(38.00)	32(44.40)	33(33.70)	χ2 = 3.634	0.163
Hypertension [n (%)]	49(28.70)	23(31.90)	26(26.50)	χ2 = 0.593	0.441
Biliary tract disease [n (%)]	52(30.40)	24(33.30)	28(28.60)	χ2 = 0.443	0.506
Malignancy [n (%)]	20(11.70)	10(13.90)	10(10.20)	χ2 = 0.543	0.461
Clinical manifestations [n (%)]
Fever, chills [n (%)]	137(80.10)	59(81.90)	78(79.60)	χ2 = 0.147	0.702
Nausea, vomiting [n (%)]	70(40.90)	50(69.40)	20(20.40)	χ2 = 41.204	< 0.001
Abdominal pain [n (%)]	60(35.10)	20(27.80)	40(40.80)	χ2 = 3.090	0.079
Physical signs [n (%)]
Right upper quadrant tenderness [n (%)]	74(43.30)	28(18.90)	46(46.90)	χ2 = 1.094	0.296
Liver percussion pain [n (%)]	36(21.10)	15(20.80)	21(21.40)	χ2 = 0.009	0.925

BMI : Body Mass Index, weight (kg) / [height (m)]²

Imaging and laboratory results

Imaging findings showed no significant differences in abscess location, number, or size between PLA patients with and without sepsis (all P > 0.05). The right hepatic lobe was the most common location (70.60%), and large-sized abscesses (50–100 mm) predominated (58.90%) in Table 1 (Supplementary Materials).

Laboratory results showed that PCT-related indices (PCT-admit, PCT-max, PCT-min, PCT-Δ) were significantly higher in PLA patients with sepsis than in those without sepsis (all P < 0.001). PCT-Δ was the most discriminatory index, with a median of 30.03 ng/mL in sepsis patients and 1.28 ng/mL in non-sepsis patients (P < 0.001). In addition, white blood cell count (WBC-max), neutrophil percentage (NE%), total bilirubin (TBil), and peak body temperature (T-max) were significantly higher in sepsis patients, while platelet count (PLT) and albumin (ALB) were significantly lower (all P < 0.05).

Complications and treatments

PLA patients with sepsis had a significantly higher incidence of MODS (9.70% vs. 0%, P = 0.001) and in-hospital death (5.60% vs. 0%, P = 0.031) than non-sepsis patients (Table 2, Supplementary Materials). The most common treatment for PLA was antibiotics combined with percutaneous catheter drainage (58.80%), which was used more frequently in sepsis patients (68.10%) than in non-sepsis patients (52.00%). No statistically significant differences were found in the duration of catheter indwelling or days to catheter removal post-discharge between the two groups (all P > 0.05).

Table 2.

Association between PCT-Δ and Prognosis (Logistic regression)

Characteristic	Model 1					Model 2
	N	Event N	OR	95% CI	p-value	N	Event N	OR	95% CI	p-value
PCT-Δ (continuous)	170	72	1.04	1.02, 1.06	< 0.001	170	72	1.04	1.02, 1.06	< 0.001
PCT-Δ
Q1	43	5	—	—		43	5	—	—
Q2	42	8	1.79	0.53, 5.99	0.346	42	8	1.84	0.52, 6.47	0.342
Q3	42	24	10.13	3.32, 30.90	< 0.001	42	24	12.71	3.84, 42.11	< 0.001
Q4	43	35	33.25	9.93, 111.28	< 0.001	43	35	42.03	11.40, 154.94	< 0.001
P for trend					< 0.001					< 0.001

Abbreviations: CI = Confidence Interval, OR = Odds Ratio; Model 1 : no covariates were adjusted ; Model 2 : adjusted for Sex, Age, Diabetes mellitus, Biliary tract disease, Tumour disease, and Abscess classification

Etiological profile and antimicrobial susceptibility

Blood cultures were performed in 144 patients, with a positive rate of 31.25% (45/144), which was significantly higher in sepsis patients (44.44%, 32/72) than in non-sepsis patients (13.27%, 13/98) (P < 0.001). Pus cultures were performed in 113 patients, with a positive rate of 75.22% (85/113), with no significant difference between the two groups (P = 0.782). Klebsiella pneumoniae (KP) was the predominant isolate in both blood and pus cultures, accounting for 22.22% and 55.75% of positive specimens, respectively (Table 3, Supplementary Materials).

Pathological findings of the patients

Eight study participants (4.7%) ultimately required surgical resection; notably, all were non-sepsis and received formal histopathological assessment postoperatively. The clinical phenotype of this subset was characterized by fever and abdominal pain. Operative indication included: inadequate clinical response to percutaneous drainage therapy in three cases, coexisting cholelithiasis in two patients, and failure of conservative therapy to achieve disease control in the remaining three patients. Comprehensive pathological review of all eight surgical specimens uniformly demonstrated inflammatory cell infiltration, findings entirely concordant with the established histomorphological patterns of hepatic abscess (Fig. 1, Supplementary Materials).

Fig. 1.

ROC curve of the PCT-Δ model for predicting prognosis. Note: Model 1: no covariates were adjusted; Model 2: adjusted for Sex, Age, Diabetes mellitus, Biliary tract disease, Tumour disease, and abscess classification

ROC curves and model validation

ROC curve analysis showed that PCT-Δ had good prognostic performance for adverse outcomes in PLA patients: Model 1 (PCT-Δ alone) had an AUC of 0.833 (95% CI: 0.771–0.895), and Model 2 (adjusted for covariates) had an AUC of 0.843 (95% CI: 0.784–0.902) (Fig. 1). Bootstrap validation (1000 iterations) yielded a corrected C-statistic of 0.834 (AUC SD = 0.0423), an accuracy of 0.726 (SD = 0.0460), and a Cohen’s Kappa of 0.413 (SD = 0.0912), indicating stable prognostic performance of the model (Fig. 2, Supplementary Materials). Calibration curves showed excellent agreement between predicted and observed probabilities of adverse outcomes, with a deviation of < 5% in the predicted probability range of 0.2–0.7 (covering 76.5% of patients)(Fig. 3, Supplementary Materials).

Fig. 2.

Non-linear association between procalcitonin delta (PCT-Δ) and prognosis in patients with pyogenic liver abscess: a restricted cubic spline analysis. Note: x-axis = PCT-Δ (ng/mL); left y-axis=odds ratio (OR) for adverse outcomes (reference OR = 1); right y-axis = PCT-Δ distribution frequency (blue histogram); red curve=association trend; pink shaded area = 95% CI

Non-linear association between PCT-Δ and adverse outcomes

RCS analysis revealed a significant non-linear relationship between PCT-Δ and adverse outcomes in PLA patients (P-overall < 0.001, P-nonlinear < 0.001). The risk of adverse outcomes rose rapidly when PCT-Δ was in the range of 0–50 ng/mL (OR increased from 1.0 to 4.2), and the rate of risk increase slowed significantly when PCT-Δ exceeded 50 ng/mL (OR > 3.5). A histogram showed that 89.4% of PLA patients had a PCT-Δ in the range of 0–50 ng/mL, which is consistent with clinical practice (Fig. 2).

Logistic regression analysis showed that PCT-Δ was significantly associated with adverse outcomes in PLA patients: each 1 ng/mL increase in PCT-Δ was associated with a 4% higher risk of adverse outcomes (Model 1: OR = 1.04, 95%CI:1.02–1.06, P < 0.001; Model 2: OR = 1.04, 95%CI:1.02–1.06, P < 0.001). Stratified by quartiles, patients in Q3 (5.61–30.03 ng/mL) and Q4 (> 30.03 ng/mL) had a significantly higher risk of adverse outcomes than those in Q1 (0-1.28 ng/mL) (Q4 adjusted OR = 42.03, P < 0.001),Although the odds ratio of adverse outcomes was extremely high in the highest PCT-Δ quartile (Q4), the wide 95% confidence interval indicated a certain degree of instability in the effect size estimation of this subgroup. In contrast, the results of the continuous PCT-Δ model (OR = 1.04, 95%CI:1.02–1.06) were more stable with a narrow and non-overlapping confidence interval, which could more objectively reflect the dose-response relationship between PCT-Δ and adverse outcomes in PLA patients (Table 2).

Discussion

The present study found that 42.35% of PLA patients developed sepsis during hospitalization, which is consistent with previous studies on the severity of PLA. The cohort had a median age of 61 years with male predominance, and diabetes mellitus was the most common comorbidity, which is in line with the demographic characteristics of PLA patients reported in Asian populations [13, 14]. Nausea/vomiting and pulmonary infection were significantly more common in sepsis patients, which may be due to sepsis-induced intestinal mucosal injury and hematogenous spread of infection, respectively. These findings suggest that PLA patients with nausea/vomiting or pulmonary infection should be closely monitored for the development of sepsis.

Existing studies have demonstrated that a reduction of procalcitonin (PCT) by less than 15% within the first 72 h (0–72 h) or 20% between 24 and 72 h (24–72 h) independently predicts 30-day mortality in patients with severe sepsis and septic shock admitted to the intermediate care unit [15]. In the latest research, the neutrophil-to-lymphocyte ratio (NLR), dynamic changes in platelet counts, and the Carmeli score have shown important clinical utility for the early diagnosis and prognostic stratification of sepsis, underscoring the critical role of inflammatory biomarkers in this setting [16]. However, none of these studies have further investigated whether dynamic changes in inflammatory biomarkers exhibit a non-linear association with the prognosis of sepsis. Static PCT measurement involves determining the serum concentration at a single, predefined time point, most commonly at the time of hospital or intensive care unit (ICU) admission. This approach provides a valuable baseline snapshot of the severity of the underlying systemic inflammatory response to bacterial infection. High static PCT levels upon presentation are robustly associated with an increased likelihood of bacterial etiology, the presence of sepsis or septic shock, and a higher risk of mortality. For instance, in a study on postoperative intracranial infection, PCT demonstrated an AUC of 0.895 for diagnosis, highlighting its discriminative power at a single point [17]. Static values are often incorporated into diagnostic algorithms for sepsis and are used to guide initial decisions, such as the initiation of antibiotics. However, static PCT measurements have inherent limitations for prognostication. A single value cannot distinguish between a patient with a severe infection who is responding well to therapy and a patient with an equally severe infection who is failing to respond. Both may present with similarly high initial PCT levels. The prognostic information is thus incomplete.

A significant decline in PCT (e.g., a relative decrease of more than 80% from peak or a decrease below a specific threshold like 0.5 ng/mL) is strongly correlated with a favorable prognosis and survival. Conversely, a failure of PCT to decrease, or a rising trend, is an ominous sign, often heralding treatment failure, the development of complications, or impending mortality. This dynamic information is arguably more prognostically valuable than the initial static value alone. It allows for a more nuanced assessment, enabling clinicians to identify patients who are not responding as expected despite initial presentation, and potentially intervene earlier. Studies in sepsis populations underscore this, where the prognostic value of PCT in emergency patients is well-established, and serial measurements provide superior risk stratification compared to admission values alone [18].

The core finding of this study is that PCT-Δ is significantly non-linearly associated with adverse outcomes in PLA patients and has good prognostic value for this population. It should be noted that the extremely high OR value in the Q4 subgroup was accompanied by a wide 95% confidence interval, and thus the results need to be interpreted cautiously. In contrast, the continuous variable model showed that each 1 ng/mL increase in PCT-Δ was associated with a 4% higher risk of adverse outcomes (OR = 1.04, 95%CI:1.02–1.06), which had better statistical stability and provided a more reasonable interpretation of the association between PCT-Δ and adverse outcomes. The prognostic utility of PCT-Δ, evidenced by an area under the curve (AUC) of 0.833, aligns with a growing body of research emphasizing the superiority of dynamic, longitudinal biomarkers over static, single-time-point measurements for risk stratification in critical illnesses.In traumatic brain injury, for instance, time-weighted average lactate significantly outperformed admission values in predicting ICU mortality, highlighting that the trajectory of a biomarker often carries more prognostic information than its isolated value [19]. PCT-Δ reflects the dynamic change of PCT levels during hospitalization, which is a comprehensive indicator of the host’s inflammatory response and therapeutic response. Compared with a single PCT measurement at admission, PCT-Δ can better reflect the entire course of the disease and the effect of treatment, which is the main reason for its higher prognostic value.

Our findings demonstrated that KP represented the foremost gram-negative pathogen implicated in PLA development, regardless of specimen source - peripheral blood cultures or pus isolates. This organism was recovered from 25.28% to 29.20% of sepsis patients, respectively, followed by EC. Among gram-positive isolates, streptococcal species predominated, with enterococci comprising the secondary group. These microbiological patterns corroborate prior observations from a retrospective analysis of 105 hepatic abscess cases at Shanghai Renji Hospital [20]. The predominance may reflect heightened intestinal carriage of K1/K2 serotype KP among ethnic Chinese populations, making this infection more prevalent in Asian demographics [21, 22]. In patients presenting with extrahepatic spread, cerebritis developed in two cases and endophthalmitis in one, all attributable to hypervirulent KP strains; prompt pathogen identification and therapeutic intervention yielded symptomatic resolution and favorable outcomes. The elevated incidence of metastatic infection in sepsis cohorts aligns with earlier epidemiologic data [13]. Inflammatory markers, PCT, WBC, NE, and NE%, showed statistically significant elevation in sepsis patients versus non-sepsis controls. While recent retrospective work suggests that dynamic trajectories of PCT, CRP, and lactate better predict 30-day mortality in sepsis [23].

Thrombocytopenia represented a conspicuous finding among sepsis patients, originating from infection-triggered inflammatory activation that enhances immune-mediated platelet clearance. PLA-associated hepatic impairment concomitantly diminishes clotting factor synthesis, further perturbing platelet function and survival [24]. Thrombocytopenia furthermore portends adverse prognosis in sepsis [25], among our 72 sepsis patients, four succumbed - two with platelet counts of 15 × 10⁹/L and one with 16 × 10⁹/L - highlighting the crucial need for early recognition and treatment of sepsis.

Our study also provides directions for future research. First, a direct comparison between PCT-Δ and other dynamic indices for sepsis (e.g., lactate clearance, SOFA score trends) is needed to clarify the relative incremental prognostic value of PCT-Δ. Second, integrating PCT-Δ with clinical scores (e.g., qSOFA) or imaging characteristics of PLA into a multimodal prediction tool may improve the prognostic accuracy for adverse outcomes [25, 26]. In addition, machine learning techniques can be further explored to capture the complex time-dependent relationship between PCT kinetics and clinical outcomes in PLA patients [27].

This study has several inherent limitations. First, the single-center retrospective design may lead to potential selection bias and information bias; therefore, validation by multicenter prospective cohort studies is still needed. Second, the relatively small sample size prevented sufficient adjustment for all confounding factors and limited the performance of stratified analyses in the restricted cubic spline model. Finally, given that PCT-Δ was calculated based on peak and trough PCT levels during hospitalization, this dynamic biomarker reflects the infection severity and therapeutic response in patients with PLA during hospitalization. It was neither designed nor validated to serve as an early pre-sepsis predictor for identifying patients at risk of sepsis before the occurrence of organ dysfunction. Accordingly, its clinical application is mainly focused on in-hospital risk stratification rather than early warning of sepsis in PLA patients.

Conclusion

Sepsis complicating pyogenic liver abscess is associated with poor clinical outcomes. Procalcitonin variation (PCT-Δ) exhibits strong prognostic value for adverse outcomes (death, MODS, hospital stay > 14 days) in PLA patients, with a significant non-linear relationship identified via 3-knot restricted cubic spline analysis. PCT-Δ, as a dynamic biomarker integrating infection severity and treatment response, can assist clinicians in risk stratification and timely therapeutic adjustment, particularly in emergency settings. Prospective multicenter studies are required to validate these findings and establish the role of PCT-Δ-guided management in improving clinical outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CRP

C-reactive protein

PCT

Procalcitonin

PCT-Δ

Peak-to-trough procalcitonin difference

PLA

Pyogenic liver abscess

RCS

Restricted cubic spline

ROC

Receiver operating characteristic curve

SOFA

Sequential Organ Failure Assessment

MODS

Multiple organ dysfunction syndrome

ESBL

Extended-spectrum β-lactamase

KP

Klebsiella pneumoniae

EC

Escherichia coli

ALT

Alanine aminotransferase

AST

Aspartate aminotransferase

TBil

Total bilirubin

ALB

Albumin

WBC

White blood cell

PLT

Platelet count

CI

Confidence interval

OR

Odds ratio

Author contributions

CXL, YJ and HXL contributed equally to this work. CXL conceived the study, conducted the formal analysis, and wrote the original draft. YJ and HXL were responsible for data curation, methodology, and validation. CXL performed statistical analysis and visualization. LL collected resources and conducted investigation. LL, ZGH and YT curated data and performed literature review. ZGH and YT assisted in investigation and project administration. ZGH and YT supervised the study and reviewed and edited the manuscript. All authors read and approved the final manuscript.

Funding

None.

Data availability

The datasets generated and/or analyzed during the study are not publicly available due to hospital privacy policies but are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Medical Ethics Committee of Sun Yat-sen Memorial Hospital, Sun Yat-sen University (Approval No.: SYSKY-2026-081-01). Due to the retrospective design using de-identified patient records, informed consent was waived by the Ethics Committee. All methods were performed in compliance with relevant institutional and national ethical regulations.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.