Association between platelet-white cell ratio and all-cause mortality in critically ill patients with atrial fibrillation: a retrospective cohort study

Abstract

Background

Atrial fibrillation is common in intensive care units and portends poor prognosis. The platelet-to-white blood cell ratio is an emerging inflammatory-coagulatory biomarker with prognostic value in cardiovascular diseases. This study examines the association between platelet-white cell ratio (PWR) and all-cause mortality in this high-risk population.

Methods

In this retrospective cohort study, clinical data were obtained from the MIMIC-IV database. We included adult intensive care unit (ICU) patients with a diagnosis of atrial fibrillation. The PWR was evaluated as the primary exposure, analysed both as a continuous measure and by quartiles. The primary endpoint was 28-day all-cause mortality, with secondary endpoints comprising mortality at 90 days, 180 days, and 1 year. Multivariable Cox models were used to evaluate the association between PWR and mortality, with adjustment for established confounders including demographics, comorbidity, and illness severity. Non-linearity was examined using restricted cubic splines, and subgroup analyses were conducted to evaluate the consistency of the association.

Results

The analysis included 2,401 patients. After full adjustment, a unit increase in PWR was independently associated with a 2% decrease in the hazard of 28-day mortality (HR 0.98, 95% CI 0.97–0.99, p < 0.001). When analysed by quartiles, a significant inverse dose-response relationship was observed (P for trend < 0.001). Compared to the lowest quartile (Q1), patients in the highest PWR quartile (Q4) exhibited a substantially reduced hazard of death at 28 days (adjusted HR 0.55, 95% CI 0.43–0.72, p < 0.001). This inverse association remained statistically significant for 90-day and 180-day mortality but was absent at 1 year. The relationship was linear across the range of PWR values for 28-day (P for non-linearity = 0.213), 90-day (P for non-linearity = 0.476), and 180-day mortality (P for non-linearity = 0.377). The results were robust across all predefined subgroups, with no evidence of significant effect modification.

Conclusion

Among critically ill patients with atrial fibrillation, a higher platelet-to-white blood cell ratio at ICU admission is inversely and linearly associated with short- and intermediate-term mortality, demonstrating a clear dose-response relationship. As an easily derived biomarker, PWR has potential utility for enhancing risk assessment in this clinical setting.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12872-026-05730-y.

Introduction

Atrial Fibrillation (AF) represents the most prevalent sustained cardiac arrhythmia in clinical practice, with a progressively increasing epidemiological burden [1]. According to the Global Burden of Disease Study, the number of AF patients worldwide reached 333 million in 2010, with incidence demonstrating a marked upward trend with advancing age [2, 3]. AF is strongly associated with severe complications, including stroke, heart failure, and cognitive impairment, significantly elevating all-cause mortality and imposing substantial healthcare costs [4]. AF is particularly prevalent in the Intensive Care Unit (ICU), accounting for approximately 10%–20% of hospitalized patients [5]. These individuals are often accompanied by adverse clinical outcomes including prolonged hospital stay and elevated mortality risk [6–8]. Given global population aging and the rising burden of comorbidities, the impact of AF is expected to escalate further. It is critical to underscore its prognostic determinants.

The onset and progression of AF in critical illness involve complex pathophysiological mechanisms, including the interaction of multiple factors such as electrical remodeling, structural remodeling, and systemic inflammation [9, 10]. In critical conditions such as sepsis, major surgery, and acute coronary syndrome, sympathetic nerve excitation, oxidative stress and the release of pro-inflammatory cytokines (e.g., Interleukin-6, tumor necrosis factor-α) exacerbate atrial electrical instability, thereby precipitating or aggravating AF [11, 12]. Studies have shown that inflammation plays a key role in the initiation, maintenance, and poor prognosis of AF: activated white blood cells release proteases and reactive oxygen species, directly damaging cardiomyocytes and promoting atrial interstitial fibrosis [13]. Simultaneously, endothelial dysfunction and abnormal platelet activation collectively facilitate intra-atrial thrombosis [14, 15]. On the one hand, hemodynamic disturbances caused by AF activate endothelial cells and promote platelet adhesion, aggregation, and activation [16, 17]. On the other hand, sustained inflammatory responses induce platelets to release mediators such as thromboxane, further amplifying inflammation and accelerating thrombus formation [18–20]. Consequently, identifying biomarkers capable of integrating inflammatory and coagulation status holds significant clinical value for prognosis in critically ill AF patients.

In recent years, the Platelet-to-White Blood Cell Ratio (PWR), as an emerging integrated biomarker of inflammatory and coagulation function, has shown promising potential in prognostic assessment of various cardiovascular diseases and critical illnesses [21, 22]. The biological rationale for PWR stems from the coordinated regulation of white blood cell and platelet counts during the inflammatory resolution process: successful resolution of inflammation is characterized by an exponential decay in white blood cell count followed by a delayed linear increase in platelet count [23, 24]. PWR dynamically captures this pattern, potentially serving as an indicator of inflammatory resolution status. Systematic validation studies have demonstrated that in acute inflammatory conditions such as COVID-19, acute heart failure, myocardial infarction, and stroke, PWR exhibits a stronger association with short-term mortality compared to conventional biomarkers including Neutrophil-to-Lymphocyte Ratio (NLR), Platelet-to-Lymphocyte Ratio (PLR), and Red Blood Cell Distribution Width (RDW) [25–27]. However, research on the association between PWR and the prognosis of critically ill patients with AF has not been systematically verified and elucidated. The study aims to validate the prognostic value of PWR, thereby providing evidence for early risk stratification and formulation of intervention strategies in critically ill AF patients.

Methods

Data source

This retrospective cohort study utilized data from the MIMIC-IV database (version 3.1). This publicly available repository contains de-identified electronic health records from patients admitted to the Beth Israel Deaconess Medical Center between 2008 and 2019. The Institutional Review Board (IRB) of the center approved the use of this database for research and waived the requirement for individual informed consent. Author Cui passed the online training courses and exams (certification number: 59921922). The modular structure of MIMIC-IV, which supports linkage with external data, enabled the evaluation of out-of-hospital mortality.

Study population

Figure 1 delineates the patient selection process. From 78,437 patients with atrial fibrillation in the MIMIC-IV database (ICD-10 and ICD-9 diagnostic codes are provided in Supplementary Table S1), 51,089 non-ICU admissions were excluded, yielding 27,348 ICU patients. Subsequent exclusions were applied as follows: 8,545 patients not admitted to the ICU for the first time, 1,395 patients not hospitalized for the first time, 2,683 patients with an ICU stay of less than 24 h, and 90 patients who survived less than 24 h after admission, resulting in 14,635 eligible patients. We further excluded 2,389 patients with pre-existing sepsis to minimize the substantial confounding effect of sepsis on platelet and white cell counts, thereby allowing for a clearer examination of the association between PWR and mortality attributable to atrial fibrillation [28–32]. An additional 7,955 patients with missing troponin data were excluded. Additionally, patients with pre-existing hematological diseases (n = 1,755), including idiopathic thrombocytopenic purpura (n = 1,367), lymphoma (n = 152), leukemia (n = 123), multiple myeloma (n = 59), and myelodysplastic syndromes (n = 54), were excluded. After excluding 135 patients with missing white blood cell (n = 132) or platelet (n = 3) data, a final cohort of 2,401 patients was included in the analysis and categorized into four groups.

Fig. 1.

Flow chart of patient selection. MIMIC IV: Medical Information Mart in Intensive Care IV

Data extraction

Baseline data were extracted from the first 24 h of ICU admission. Demographics (age, gender, race) and vital signs (heart rate, respiratory rate, SBP, DBP, MBP, SpO₂) were collected. Disease severity scores (APACHE II, SOFA, APS III, OASIS, SAPS II, Charlson Comorbidity Index) and the CHA₂DS₂-VA thromboembolic risk score were calculated. Laboratory parameters, including complete blood counts, electrolytes, renal and hepatic function markers, and coagulation profiles, were analysed. For variables with repeated measurements, the average value was used. Assessed comorbidities included myocardial infarction, heart failure, hypertension, diabetes, stroke, COPD, liver cirrhosis, CKD, cancer, and hyperlipidemia. We also documented concomitant medications (β-blockers, ACEIs/ARBs, anticoagulants) and clinical interventions (invasive mechanical ventilation).

Study endpoints

The primary endpoint of this study was all-cause mortality at 28 days. Secondary endpoints included all-cause mortality at 90 days, 180 days, and 1 year.

Statistical analysis

Baseline characteristics were summarized using descriptive statistics. Continuous variables are presented as mean ± standard deviation (SD) or median (interquartile range [IQR]), based on their distribution [33]. Categorical variables are expressed as numbers (percentages). Between-group comparisons were made using the Kruskal-Wallis or Pearson’s chi-squared test, as appropriate. Variables with missing data exceeding 20% were excluded from the imputation process. For the remaining covariates with missingness within the 20% threshold, we addressed the missing data through multiple imputation [34, 35]; the proportion of missing data per variable is detailed in Supplementary Table S2.

Multivariable Cox proportional hazards models were employed as the primary analysis to assess the independent association between predictors and mortality. We constructed three models to calculate hazard ratios (HRs) with 95% confidence intervals (CIs) for all-cause mortality at 28 days, 90 days, 180 days, and 1 year. The selection of covariates for adjustment was based on clinical relevance and their established role as prognostic factors in the literature [36]. To examine the dose-response relationship, we fitted restricted cubic spline (RCS) curves. Survival rates were plotted using Kaplan-Meier curves and compared with the log-rank test. Subgroup analyses were performed across pre-specified strata, including age, sex, race, and comorbidities (e.g., hypertension, heart failure, diabetes, CKD), as well as the CHA₂DS₂-VA score. Multicollinearity among covariates in the multivariable Cox models was assessed using variance inflation factors (VIF). A VIF < 5 was considered indicative of no substantial multicollinearity. A two-tailed p < 0.05 defined statistical significance.

To evaluate the robustness of our primary findings and assess the potential impact of reverse causation—where early post-ICU mortality could influence the measurement of PWR at admission—we conducted two pre-specified sensitivity analyses: (1) excluding patients who died within 48 h of ICU admission, and (2) excluding those who died within 72 h. In each of these analytical cohorts, we repeated all primary analyses, including Cox proportional-hazards regression models (to assess the association of continuous and quartile-based PWR with 28-day all-cause mortality) along with the corresponding Kaplan–Meier survival curves and RCS plots.

To assess whether the prognostic value of PWR varied over time and to address potential confounding by early deaths and acute-phase disturbances, we performed landmark analyses anchored at ICU discharge, day 3, and day 7. Among patients who survived to each landmark, the association between PWR and subsequent 1-year mortality was evaluated using the same multivariable Cox models as in the primary analysis. PWR was measured at the respective landmark time point.

All the analyses were performed with the statistical software packages R (http://www.R-project.org, The R Foundation), IBM SPSS Statistics 27, and Free Statistics software versions 1.9.

Results

Cohort characteristics

Baseline characteristics of the study population, stratified by platelet-white cell ratio (PWR) quartiles, are presented in Table 1. Among the 2,401 enrolled patients, the overall mean age was 77.4 ± 11.6 years, and 1,383 (57.6%) were male. Significant differences were observed across PWR quartiles for a majority of demographic, clinical, and laboratory variables.

Table 1.

Baseline Characteristics of the Study Population Stratified by Platelet-White Cell Ratio Quartiles

Variables	Overall (n = 2401)		Q1 (n = 600)	Q2 (n = 600)	Q3 (n = 600)	Q4 (n = 601)	P value
Demographics
Age, year		77.4 ± 11.6	77.0 ± 11.3	76.9 ± 11.9	77.2 ± 11.8	78.4 ± 11.2	0.087
Gender, Male, n (%)		1383 (57.6)	373 (62.2)	358 (59.7)	333 (55.5)	319 (53.1)	0.006
Race							0.008
White, n (%)		1686 (70.2)	408 (68)	422 (70.3)	443 (73.8)	413 (68.7)
Black, n (%)		149 (6.2)	28 (4.7)	37 (6.2)	31 (5.2)	53 (8.8)
Other, n (%)		566 (23.6)	164 (27.3)	141 (23.5)	126 (21)	135 (22.5)
Vital signs
Heart rate, bpm		84.0 ± 17.3	86.4 ± 17.9	83.3 ± 16.3	83.3 ± 17.1	82.9 ± 17.6	0.001
Respiratory rate, bpm		19.9 ± 3.6	20.1 ± 3.8	20.0 ± 3.6	19.8 ± 3.4	19.8 ± 3.6	0.37
SBP, mmHg		117.1 ± 17.9	113.8 ± 16.6	116.7 ± 17.2	118.7 ± 19.0	119.0 ± 18.4	< 0.001
DBP, mmHg		64.3 ± 16.4	63.2 ± 20.1	64.1 ± 11.3	65.2 ± 17.2	64.6 ± 15.7	0.184
MBP, mmHg		77.2 ± 11.7	75.4 ± 11.5	77.5 ± 11.2	78.1 ± 12.5	77.7 ± 11.6	< 0.001
SpO2, %, Mean ± SD		96.8 ± 6.4	96.8 ± 2.3	96.5 ± 2.0	97.2 ± 12.2	96.7 ± 2.3	0.384
Disease Severity Score
APACHE II score		22.7 ± 6.9	25.2 ± 7.4	22.3 ± 6.8	21.4 ± 6.3	22.0 ± 6.4	< 0.001
SOFA score		4.0 (2.0, 7.0)	6.0 (4.0, 9.0)	4.0 (3.0, 6.0)	4.0 (2.0, 6.0)	3.0 (2.0, 5.0)	< 0.001
APSIII score		48.0 (37.0, 59.0)	53.0 (40.0, 68.0)	47.0 (37.0, 59.0)	45.0 (35.0, 55.0)	46.0 (36.0, 56.0)	< 0.001
OASIS score		33.8 ± 8.4	36.6 ± 8.8	33.7 ± 8.2	32.4 ± 8.1	32.5 ± 7.9	< 0.001
SAPSII score		41.5 ± 12.5	46.5 ± 14.1	40.9 ± 11.9	39.5 ± 11.5	39.1 ± 11.1	< 0.001
Thromboembolic risk
CHA2DS2-VA score		3.7 ± 1.4	3.6 ± 1.4	3.7 ± 1.4	3.7 ± 1.4	3.8 ± 1.4	0.118
Laboratory Examination
WBC, K/uL		11.8 ± 5.2	16.1 ± 6.1	12.3 ± 4.0	10.5 ± 3.5	8.2 ± 3.1	< 0.001
RBC, m/uL		3.6 ± 0.7	3.6 ± 0.7	3.7 ± 0.7	3.7 ± 0.7	3.5 ± 0.7	< 0.001
Hemoglobin, g/dL		10.8 ± 2.1	10.9 ± 2.2	11.1 ± 2.1	10.9 ± 2.1	10.3 ± 1.9	< 0.001
Platelet, K/uL		216.9 ± 90.6	162.8 ± 61.2	199.5 ± 65.1	227.6 ± 76.7	277.8 ± 109.3	< 0.001
Hematocrit, %		33.1 ± 6.2	33.1 ± 6.5	33.4 ± 6.2	32.2 ± 7.1	32.0 ± 5.8	< 0.001
Serum creatinine, mg/dL		1.3 (0.9, 1.9)	1.4 (1.0, 2.1)	1.2 (0.9, 1.9)	1.2 (0.9, 1.8)	1.3 (0.9, 2.0)	0.01
eGFR, mL/min/1.73m2		48.4 (29.8, 69.9)	45.4 (27.2, 68.3)	50.7 (30.5, 70.3)	49.8 (31.5, 71.7)	48.2 (30.0, 69.5)	0.07
BUN, mg/dL		28.0 (19.0, 44.3)	29.6 (20.0, 45.0)	28.0 (18.5, 45.5)	25.5 (17.5, 42.0)	28.0 (19.5, 45.7)	0.005
Glucose, mg/dL		149.2 ± 58.8	161.0 ± 66.1	154.0 ± 55.2	145.7 ± 56.0	135.9 ± 54.6	< 0.001
Sodium, mEq/L		138.3 ± 4.9	138.5 ± 4.8	138.3 ± 4.6	138.5 ± 5.1	138.0 ± 4.9	0.261
Chloride, mEq/L		102.7 ± 6.2	103.6 ± 6.0	102.8 ± 6.0	102.4 ± 6.5	101.9 ± 6.1	< 0.001
Potassium, mEq/L		4.3 ± 0.6	4.3 ± 0.6	4.3 ± 0.5	4.2 ± 0.5	4.3 ± 0.6	< 0.001
Total Calcium, mg/dL		8.5 ± 0.7	8.3 ± 0.7	8.5 ± 0.7	8.6 ± 0.7	8.6 ± 0.7	< 0.001
Anion gap, mEq/L		15.1 ± 3.7	15.7 ± 4.4	15.0 ± 3.5	14.8 ± 3.5	14.9 ± 3.4	< 0.001
PT, sec		14.8 (12.9, 18.9)	15.2 (13.3, 19.4)	14.5 (12.6, 18.2)	14.4 (12.8, 18.2)	14.9 (13.0, 19.2)	0.001
APTT, sec		34.9 (28.9, 54.9)	35.6 (29.2, 57.9)	35.0 (28.7, 55.8)	34.5 (28.7, 55.2)	34.4 (29.0, 50.9)	0.513
INR		1.3 (1.2, 1.7)	1.4 (1.2, 1.8)	1.3 (1.1, 1.7)	1.3 (1.2, 1.7)	1.4 (1.2, 1.8)	0.002
CKMB, ng/mL		6.3 (3.5, 15.0)	8.0 (4.0, 24.1)	7.0 (4.0, 19.1)	6.0 (3.0, 13.0)	5.0 (3.0, 9.3)	< 0.001
Troponin T, ng/mL		0.1 (0.0, 0.5)	0.1 (0.0, 0.7)	0.1 (0.0, 0.7)	0.1 (0.0, 0.5)	0.1 (0.0, 0.3)	< 0.001
Comorbidities
Myocardial Infarct, n (%)		587 (24.4)	162 (27)	165 (27.5)	150 (25)	110 (18.3)	< 0.001
Congestive Heart Failure, n (%)		1529 (63.7)	368 (61.3)	374 (62.3)	392 (65.3)	395 (65.7)	0.296
Hypertension, n (%)		851 (35.4)	196 (32.7)	220 (36.7)	220 (36.7)	215 (35.8)	0.418
Diabetes Mellitus, n (%)		936 (39.0)	233 (38.8)	230 (38.3)	227 (37.8)	246 (40.9)	0.006
Ischemic Stroke, n (%)		289 (12.0)	65 (10.8)	73 (12.2)	72 (12)	79 (13.1)	0.676
COPD, n (%)		257 (10.7)	70 (11.7)	60 (10)	68 (11.3)	59 (9.8)	0.651
Liver Cirrhosis, n (%)		75 ( 3.1)	29 (4.8)	17 (2.8)	12 (2)	17 (2.8)	0.034
CKD, n (%)		829 (34.5)	199 (33.2)	205 (34.2)	209 (34.8)	216 (35.9)	0.782
Malignant Cancer, n (%)		267 (21.3)	77 (26.2)	61 (19.4)	66 (20.6)	63 (19.5)	0.133
Hyperlipidemia, n (%)		1183 (49.3)	307 (51.2)	288 (48)	317 (52.8)	271 (45.1)	0.037
Charlson Comorbidity Index		7.0 ± 2.6	6.8 ± 2.5	6.7 ± 2.6	7.0 ± 2.5	7.3 ± 2.6	0.003
Concomitant Medication
β-blocker, n (%)		1743 (72.6)	423 (70.5)	434 (72.3)	452 (75.3)	434 (72.2)	0.302
ACEIs/ARBs, n (%)		423 (17.6)	102 (17)	105 (17.5)	108 (18)	108 (18)	0.965
Anticoagulant drugs, n (%)		2155 (89.8)	536 (89.3)	532 (88.7)	548 (91.3)	539 (89.7)	0.472
Invasive Procedures
Invasive mechanical ventilation, n (%)		840 (35.0)	288 (48)	216 (36)	189 (31.5)	147 (24.5)	< 0.001
Length Of Stay (LOS)
LOS in hospital		8.0 (5.1, 12.9)	9.0 (5.7, 13.6)	7.4 (4.8, 12.2)	8.0 (5.3, 13.2)	7.8 (5.0, 12.4)	0.003
LOS in ICU		2.9 (1.8, 5.0)	3.3 (2.0, 6.0)	2.9 (1.8, 4.9)	3.0 (1.9, 5.2)	2.5 (1.7, 4.0)	< 0.001
Outcomes
28-day mortality, n (%)		512 (21.3)	161 (26.8)	143 (23.8)	114 (19)	94 (15.6)	< 0.001
90-day mortality, n (%)		720 (30.0)	204 (34)	189 (31.5)	170 (28.3)	157 (26.1)	0.016
180-day mortality, n (%)		842 (35.1)	229 (38.2)	209 (34.8)	205 (34.2)	199 (33.1)	0.288
1-year mortality, n (%)		1003 (41.8)	254 (42.3)	247 (41.2)	245 (40.8)	257 (42.8)	0.89

PWR platelet-white cell ratio, SBP systolic blood pressure, DBP diastolic blood pressure, MBP mean blood pressure, ACEIs angiotensin-converting enzyme inhibitors, ARBs angiotensin receptor blockers, CKD chronic kidney disease, BUN blood urea nitrogen, PT prothrombin time, APTT activated partial thromboplastin time, INR international normalized ratio

PWR: Q1(0.99, 13.73), Q2(13.73, 18.83), Q3(18.83,25.47), Q4(25.47,112.76)

Collectively, patients in the lower PWR quartiles (Q1 and Q2) presented with significantly greater disease severity at ICU admission compared to those in higher quartiles. This was objectively demonstrated by: (1) higher organ dysfunction scores, including SOFA (Q1: 6.0 [4.0, 9.0] vs. Q4: 3.0 [2.0, 5.0], p < 0.001) and APS III (Q1: 53.0 [40.0, 68.0] vs. Q4: 46.0 [36.0, 56.0], p < 0.001); (2) a higher requirement for invasive mechanical ventilation (Q1: 48.0% vs. Q4: 24.5%, p < 0.001), with a clear decreasing trend from Q1 to Q4; and (3) more pronounced laboratory abnormalities, including pronounced leukocytosis, lower platelet counts, and elevated markers of myocardial injury (CK-MB and Troponin T, all p < 0.001).

This profile of increased baseline illness severity was consistent with the observed short-term mortality outcomes. A clear gradient was evident, with 28-day mortality being substantially higher in the lowest PWR quartile (Q1: 26.8%) compared to the highest (Q4: 15.6%, p < 0.001). This trend persisted for 90-day mortality (Q1: 34.0% vs. Q4: 26.1%, p = 0.016). However, the differences in long-term mortality (180-day and 365-day) across quartiles were not statistically significant (p = 0.288 and p = 0.890, respectively).

Association of PWR and mortality

The results of the multivariable Cox regression analyses for primary and secondary outcomes are presented in Table 2. In the fully adjusted model (Model III), which accounted for a comprehensive set of demographic, clinical, laboratory, and treatment covariates, a lower PWR remained independently associated with an increased risk of short- and intermediate-term mortality. Collinearity diagnostics confirmed no substantial multicollinearity among the covariates in the fully adjusted model (all VIF < 5; see Supplementary Table S3).

Table 2.

Association between PWR and mortality in critically ill patients with atrial fibrillation

Categories	Model I		Model II		Model III
	HR (95%CI)	P-value	HR (95%CI)	P-value	HR (95%CI)	P-value
Primary Outcome
28-day mortality
PWR (continuous)	0.98 (0.97 ~ 0.99)	< 0.001	0.98 (0.97 ~ 0.99)	< 0.001	0.98 (0.97 ~ 0.99)	< 0.001
PWR (quartiles)
Quartile 2:1	0.86 (0.68 ~ 1.07)	0.177	0.89 (0.71 ~ 1.12)	0.334	0.93 (0.74 ~ 1.16)	0.509
Quartile 3:1	0.66 (0.52 ~ 0.83)	0.001	0.67 (0.53 ~ 0.86)	0.001	0.72 (0.57 ~ 0.92)	0.009
Quartile 4:1	0.53 (0.41 ~ 0.68)	< 0.001	0.52 (0.4 ~ 0.67)	< 0.001	0.55 (0.43 ~ 0.72)	< 0.001
P for trend	0.81 (0.75 ~ 0.87)	< 0.001	0.8 (0.74 ~ 0.87)	< 0.001	0.82 (0.76 ~ 0.89)	< 0.001
Secondary outcomes
90-day mortality
PWR (continuous)	0.99 (0.98 ~ 0.99)	< 0.001	0.98 (0.98 ~ 0.99)	< 0.001	0.98 (0.98 ~ 0.99)	< 0.001
PWR (quartiles)
Quartile 2:1	0.89 (0.73 ~ 1.09)	0.255	0.94 (0.77 ~ 1.14)	0.526	0.97 (0.8 ~ 1.19)	0.802
Quartile 3:1	0.77 (0.63 ~ 0.94)	0.011	0.79 (0.64 ~ 0.97)	0.022	0.85 (0.69 ~ 1.04)	0.12
Quartile 4:1	0.69 (0.56 ~ 0.85)	< 0.001	0.68 (0.55 ~ 0.84)	< 0.001	0.71 (0.57 ~ 0.88)	0.002
P for trend	0.88 (0.83 ~ 0.94)	< 0.001	0.88 (0.82 ~ 0.94)	< 0.001	0.89 (0.83 ~ 0.95)	0.001
180-day mortality
PWR (continuous)	0.99 (0.98 ~ 1)	0.007	0.99 (0.98 ~ 1)	0.002	0.99 (0.98 ~ 1)	0.009
PWR (quartiles)
Quartile 2:1	0.88 (0.73 ~ 1.06)	0.172	0.92 (0.76 ~ 1.11)	0.375	0.96 (0.79 ~ 1.16)	0.653
Quartile 3:1	0.83 (0.69 ~ 1)	0.048	0.84 (0.69 ~ 1.02)	0.073	0.9 (0.74 ~ 1.09)	0.271
Quartile 4:1	0.78 (0.65 ~ 0.95)	0.011	0.77 (0.64 ~ 0.93)	0.008	0.81 (0.66 ~ 0.98)	0.031
P for trend	0.92 (0.87 ~ 0.98)	0.009	0.92 (0.86 ~ 0.97)	0.005	0.93 (0.88 ~ 0.99)	0.024
1-year mortality
PWR (continuous)	1 (0.99 ~ 1)	0.189	0.99 (0.99 ~ 1)	0.09	1 (0.99 ~ 1)	0.171
PWR (quartiles)
Quartile 2:1	0.94 (0.79 ~ 1.11)	0.458	0.98 (0.82 ~ 1.17)	0.828	1.02 (0.86 ~ 1.22)	0.819
Quartile 3:1	0.9 (0.75 ~ 1.07)	0.218	0.91 (0.76 ~ 1.08)	0.284	0.96 (0.8 ~ 1.15)	0.644
Quartile 4:1	0.92 (0.77 ~ 1.09)	0.349	0.91 (0.76 ~ 1.09)	0.293	0.94 (0.78 ~ 1.12)	0.476
P for trend	0.97 (0.92 ~ 1.03)	0.303	0.96 (0.91 ~ 1.02)	0.208	0.97 (0.92 ~ 1.03)	0.369

Model I: did not adjust any variables

Model II: adjusted for age, gender, race, HR, MBP

Model III: adjusted for age, gender, race, HR, MBP, eGFR, INR, APACHE II score, Troponin T, hypertension, congestive heart failure, myocardial infarct, diabetes mellitus, ischemic stroke, COPD, CKD, liver cirrhosis, malignant cancer, anticoagulant drugs

PWR: Q1(0.99, 13.73), Q2(13.73, 18.83), Q3(18.83,25.47), Q4(25.47,112.76) 

For the primary outcome of 28-day mortality, a significant inverse association was observed. In the fully adjusted model (Model III), each unit increase in PWR as a continuous variable was associated with a 2% reduction in the hazard of death (adjusted HR 0.98, 95% CI 0.97–0.99, p < 0.001). When analysed by quartiles, a significant dose-response relationship was evident (P for trend < 0.001). Specifically, patients in the highest PWR quartile (Q4) had a substantially lower risk of 28-day mortality compared to those in the lowest quartile (Q1) (adjusted HR 0.55, 95% CI 0.43–0.72, p < 0.001).

This protective association persisted for the secondary outcomes of 90-day and 180-day mortality, though the effect estimates were slightly attenuated. For 90-day mortality, the hazard ratio for the highest versus lowest quartile was 0.71 (95% CI 0.57–0.88, p = 0.002), with a significant trend across quartiles (P for trend < 0.001). For 180-day mortality, the corresponding hazard ratio was 0.81 (95% CI 0.66–0.98, p = 0.031), and the trend remained statistically significant (P for trend = 0.024).

However, the association between PWR and 1-year mortality was not statistically significant in the fully adjusted model, either when analysed as a continuous variable (adjusted HR 1.00, 95% CI 0.99–1.00, p = 0.171) or across quartiles (P for trend = 0.369). The effect estimates for the quartile comparisons at one year were all close to the null and non-significant.

The consistency of the results across the three adjustment models underscores the robustness of the observed inverse association between PWR and short-term mortality risk.

Kaplan-Meier survival curves illustrating the association between PWR quartiles and all-cause mortality at different time points are presented in Fig. 2. A significant and graded positive association was observed between higher PWR levels and improved survival probability in the short term. Specifically, the log-rank test indicated statistically significant differences across PWR quartiles for 28-day (p < 0.0001, Fig. 3A) and 90-day (p = 0.0025, Fig. 3B) mortality. Visually, a clear separation of the curves demonstrates that patients in the highest PWR quartile (Q4) consistently experienced the best survival, while those in the lowest quartile (Q1) had the worst prognosis.

Fig. 2.

Kaplan-Meier survival curves stratified by PWR quartiles. Kaplan-Meier estimates for (A) 28-day, (B) 90-day, (C)180-day, and (D) 1-year mortality are shown. The cohort was divided by PWR quartiles: Q1(0.99, 13.73), Q2(13.73, 18.83), Q3(18.83,25.47), Q4(25.47,112.76)

Fig. 3.

Restricted cubic spline curves for the association between PWR and 28-day mortality. The association was modeled using restricted cubic splines and adjusted for age, gender, race, HR, MBP, eGFR, INR, APACHE II score, Troponin T, hypertension, congestive heart failure, myocardial infarct, diabetes mellitus, ischemic stroke, COPD, CKD, liver cirrhosis, malignant cancer, anticoagulant drugs

However, this significant association attenuated over longer follow-up periods. For 180-day mortality (Fig. 3C), the survival curves converged, and the difference across quartiles was no longer statistically significant (p = 0.065). Finally, for 1-year mortality (Fig. 3D), there was no significant difference in survival probabilities among the four PWR quartiles (p = 0.65).

The association between the PWR and the hazard of mortality was linear for both 28-day (P for non-linearity = 0.213) and 90-day outcomes (P for non-linearity = 0.476; Supplementary Figure S1), exhibiting a consistent inverse relationship where higher PWR levels were associated with a lower risk of death. A similar linear inverse trend was observed for 180-day mortality (P for overall association = 0.02; P for non-linearity = 0.377; Supplementary Figure S2). In contrast, the association between PWR and 1-year mortality was not statistically significant (P for overall association = 0.492; P for non-linearity = 0.778; Supplementary Figure S3).

Subgroup analyses

As shown in Fig. 4, the association remained generally consistent across nearly all subgroups, with no significant effect modification observed (except for one interaction term with a p-value < 0.05). This supports the robustness and reliability of the relationship across diverse patient demographics and clinical comorbidities.

Fig. 4.

Subgroup Analysis of the Association Between PWR and 28-Day Mortality in Critically Ill Patients with Atrial Fibrillation. The forest plot displays hazard ratios (HRs) and 95% confidence intervals derived from multivariable Cox proportional hazards models

Sensitivity analyses

After excluding patients who died within 48 h (n = 2,331) and 72 h (n = 2,277) of ICU admission, the 28-day mortality rates were 19.4% and 17.0%, respectively. In both cohorts, the inverse association between PWR and 28-day mortality remained robust, with adjusted HRs for continuous PWR consistently around 0.98 (all P < 0.001). The protective effect of the highest versus lowest PWR quartile also persisted (e.g., fully adjusted Model III: HR 0.62 [95% CI 0.47–0.82] for 48-h exclusion; HR 0.66 [95% CI 0.49–0.88] for 72-h exclusion; both P < 0.01), and significant inverse trends across quartiles were maintained (all P for trend ≤ 0.002; Supplementary Tables S3–S4).

Corresponding Kaplan–Meier curves and RCS plots consistently confirmed these patterns (Supplementary Figs. 4–7 S).

Landmark analysis for long-term mortality

A total of 2,135 patients (34.6%) survived to ICU discharge and were included in the landmark analysis (Supplementary Table S6). In this cohort, PWR at ICU discharge was not associated with 1-year mortality (adjusted HR: 1.00, 95% CI: 1.00–1.01, P = 0.372). Consistent null associations were observed in sensitivity analyses using day 3 (n = 2,331; HR: 1.00, 95% CI: 0.99–1.00, P = 0.612) and day 7 (n = 2,109; HR: 1.01, 95% CI: 1.00–1.01, P = 0.096) as alternative landmarks. These findings suggest that PWR, whether measured at ICU admission or at each landmark time point, does not provide independent prognostic information for long-term mortality in critically ill patients with AF.

Discussion

Based on the MIMIC-IV database, this study systematically investigated the prognostic value of the PWR in critically ill patients with AF for the first time. Through the analysis of 2401 patients with AF in ICU, we found that PWR was negatively associated with all-cause mortality at 28, 90, and 180 days, with a significant dose-response relationship. Specifically, in the fully adjusted model, each unit increase in PWR was associated with a 2% reduction in 28-day mortality risk (adjusted HR 0.98, 95% CI 0.97–0.99, p < 0.001). However, this association lost statistical significance at 1 year, suggesting that PWR primarily reflects short-to medium-term prognosis. Further analysis revealed a linear relationship between PWR and mortality at 28 and 90 days. Moreover, this association remained consistent across multiple clinically relevant subgroups, indicating the significant value of PWR for early risk stratification in critically ill AF patients.

The occurrence and persistence of AF depend on atrial electrical and structural remodeling, with inflammation serving as the core mechanism driving these pathological processes [9]. Previous studies have confirmed that white blood cells (particularly neutrophils and monocytes) directly promote cardiomyocyte apoptosis, interstitial fibrosis, and electrical conduction heterogeneity by releasing pro-inflammatory cytokines and reactive oxygen species, thereby contributing to the formation of AF [37–39], AF patients exhibit a hypercoagulable state. Platelets are not only involved in coagulation but also amplify inflammatory responses by releasing mediators such as platelet-derived growth factor (PDGF) [40, 41]. Meanwhile, persistent inflammation can activate endothelial cells and platelets, promoting microthrombosis and atrial endothelial dysfunction, which further exacerbates AF progression and thromboembolic risk [42, 43]. PWR captures the dynamic balance between inflammatory activity and coagulation status. During the acute phase of inflammation, white blood cell counts rise sharply, while platelet counts may decrease due to consumptive loss or myelosuppression, leading to a reduction in PWR. As inflammation subsides, white blood cell counts gradually normalize, and platelet counts increase reactively, resulting in a subsequent rise in PWR [24, 44], a low PWR may represent a persistent inflammatory state with inadequate activation of resolution mechanisms, which is closely associated with poor prognosis in AF patients.

This study confirms a significant inverse correlation between PWR and all-cause mortality at 28, 90, and 180 days in critically ill AF patients, indicating that PWR predominantly influences short-to-medium-term clinical outcomes. We hypothesize that PWR’s predictive capacity for short-term mortality risk may stem from its integrated reflection of inflammation-coagulation axis imbalance. In this pathological process, immune cells and platelets interact via formations such as platelet-leukocyte aggregates (PLAs), collectively promoting thromboinflammation, characterized by a significant increase in white blood cells accompanied by thrombocytopenia [45–47]. A low PWR suggests a persistent and uncontrolled inflammatory response coupled with coagulopathy. This imbalance can further lead to microcirculatory dysfunction, endothelial injury, and multi-organ failure, thereby significantly increasing short-term mortality risk [48]. Our baseline data align with this mechanism: patients in the lowest PWR quartile (Q1) had significantly higher disease severity, evidenced by higher organ dysfunction scores, including SOFA (Q1: 6.0 [4.0, 9.0] vs. Q4: 3.0 [2.0, 5.0], p < 0.001) and APS III (Q1: 53.0 [40.0, 68.0] vs. Q4: 46.0 [36.0, 56.0], p < 0.001), as well as a higher requirement for invasive mechanical ventilation (Q1: 48.0% vs. Q4: 24.5%). Our findings are consistent with research by Foy et al., which indicated that in various critical illness settings, PWR outperforms traditional inflammatory markers like NLR and PLR in predicting short-term mortality, and its dynamic changes correlate with the inflammatory resolution process [25]. A nationwide cohort study by Lyu et al. also reported an independent inverse association between PWR and all-cause mortality, further supporting the reliability of our conclusion [49]. Notably, the association between PWR and mortality in this study was no longer statistically significant at 1 year (Q4 vs. Q1: adjusted HR 0.94, 95% CI 0.78–1.12, p = 0.476). This may be because long-term survival in critically ill patients is predominantly driven by AF-related chronic complications (e.g., heart failure, Ischemic stroke) and underlying diseases (e.g., renal dysfunction, malignancy) [50, 51]. Secondly, successful ICU supportive care may reverse acute-phase inflammation and coagulation disorders, diminishing the representativeness of PWR for long-term prognosis. Similar time-dependent associations have been reported for other acute-phase biomarkers such as procalcitonin and C-reactive protein [52–54]. This limitation highlights the importance of distinguishing between acute-phase and chronic-phase influencing factors in the prognostic assessment of critically ill patients. In contrast to its consistent association with short- and medium-term mortality, PWR was not predictive of 1-year mortality among patients who survived to ICU discharge. This time-dependent prognostic pattern was formally confirmed by landmark analyses, suggesting that the prognostic utility of PWR may be primarily driven by its reflection of acute physiological derangement during the critical phase, rather than by chronic disease burden.

The robustness of our primary finding was supported by subgroup analyses, which showed a consistently reduced mortality risk with higher PWR across diverse demographics and clinical profiles. A significant interaction by race (P = 0.002) was observed, indicating a potential heterogeneity in the effect size of PWR. Although the direction of the association was uniformly protective, the magnitude of benefit may differ among racial groups, possibly reflecting variations in underlying inflammatory burden, genetic predispositions, or social determinants of health. Future studies should validate and explore the drivers of this heterogeneity. For all other subgroups—including those defined by key demographics and comorbidities—no significant interactions were observed, suggesting the universal prognostic value of PWR across patients with different demographic characteristics and comorbidities. This robustness is particularly critical in heart failure and ischemic stroke. Heart failure and ischemic stroke are common and serious complications of atrial fibrillation, which are closely associated with poor prognosis in AF and can increase the risk of death independently or cooperatively [55–58]. The results of this study are consistent with previous evidence. Ye et al. reported a significant association between low PWR and short-term mortality in patients with acute heart failure [59]. PWR has also been confirmed to be associated with prognosis in stroke populations [60]. Furthermore, this study conducted a stratified analysis based on the CHA₂DS₂-VAS score-a comprehensive assessment tool encompassing stroke risk, vascular disease, age, and other factors [61, 62]. The results showed that the negative correlation between PWR and mortality remained consistent in subgroups with scores < 2 (low-to-moderate risk) and ≥ 2 (high risk), with no significant effect modification. This indicates that the predictive value of PWR is equally applicable to critically ill AF patients with different stroke risks, further supporting the reliability and universality of this biomarker in complex clinical settings. Importantly, the association between PWR and mortality remained robust after rigorous adjustment for APACHE II score, indicating that PWR provides prognostic information independent of, and incremental to, this comprehensive measure of illness severity.

We restricted the cohort to patients with their first ICU admission and first hospitalization, in order to preserve the statistical independence of observations and minimize unmeasured confounding, which are essential for valid causal inference between PWR exposure and mortality outcomes. First, inclusion of multiple admissions from the same patient would violate the statistical assumption of independence of observations, potentially leading to underestimated standard errors and inflated type I error rates. Second, focusing on the first recorded episode enabled a clear temporal sequence to be established, with baseline PWR measured at a well‑defined starting point linked to a non-overlapping follow-up period for mortality—a critical requirement for valid causal inference. Third, patients with prior ICU stays have inherently different baseline risks, treatment histories, and potential for drug resistance or chronic organ dysfunction. Restricting to the first episode therefore enabled us to investigate the de novo association between PWR and mortality in a more homogeneous population, reducing bias from unmeasured prior healthcare exposures and disease trajectories.

This study is the first to systematically evaluate the prognostic value of PWR in ICU-admitted AF patients, addressing a significant clinical issue in critical cardiovascular care. By incorporating multiple time-point outcomes (28 days, 90 days, 180 days, and 1 year), the study clarifies the time dependency of PWR’s predictive effect, providing multi-dimensional evidence for clinical precise risk stratification. In subject selection, we rigorously excluded patients with prior hematologic diseases and sepsis, minimizing relevant confounding factors and significantly enhancing the internal validity of our results. Moreover, subgroup analyses confirmed that the inverse association between PWR and 28-day mortality persisted consistently across subgroups defined by age, sex, ethnicity, and key comorbidities, further substantiating the robustness of PWR’s prognostic value.

These findings suggest that PWR can serve as an easily obtainable, integrative biomarker at ICU admission for the early identification of AF patients at higher risk of mortality. The simple calculation and low cost of PWR facilitate its clinical adoption, providing a practical tool for early risk stratification and the formulation of individualized intervention strategies, particularly in settings where traditional scoring systems are complex or data are incomplete.

Study limitations

This study has several limitations. First, as a retrospective observational analysis, residual confounding cannot be fully ruled out despite rigorous adjustments. This concern is particularly relevant for key laboratory variables (e.g., markers of inflammation and organ dysfunction) that were excluded due to high rates of missing data. We reason that these are likely positive confounders, and their omission may bias the estimated effect of PWR toward the null. Second, to ensure internal validity, we excluded patients with sepsis, which limits the direct generalizability of our findings to non-septic critically ill patients with AF. The prognostic role of PWR in septic AF patients requires future study. Third, our restriction to first ICU admissions, while methodologically necessary for statistical independence, may exclude sicker patients with recurrent disease, potentially attenuating the observed associations and limiting generalizability to this subgroup. Fourth, the single-center design may affect generalizability, and validation in multi-center cohorts is needed. Finally, PWR was assessed only at ICU admission; its dynamic changes and their prognostic implications remain to be investigated.

Conclusions

Among critically ill patients with atrial fibrillation, a higher platelet-to-white blood cell ratio at ICU admission is inversely and linearly associated with short- and intermediate-term mortality, demonstrating a clear dose-response relationship. As an easily derived biomarker, PWR has potential utility for enhancing risk assessment in this clinical setting.

Authors’ contributions

XB: conceptualization, data curation, formal analysis, investigation, methodology, project administration, software, validation, visualization, writing - original draft, writing - review \& editing.ZR: data curation, formal analysis, methodology, software, writing - review \& editing.XJ: data curation, methodology, writing - review \& editing.CM: methodology, formal analysis, resources, supervision, writing - review \& editing.WJ: methodology, resources, supervision, writing - original draft, writing - review \& editing.CXJ: conceptualization, investigation, project administration, resources, validation, supervision, writing - original draft, writing - review \& editing.All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Eighth Chengdu Pharmaceutical Association Pharmaceutical Research Project (Grant No. Cdyxky8010).

Data availability

This study was a retrospective cohort study with data from the MIMIC-IV database. Clinical trial number: not applicable. The datasets used in the present study are available from the first author and corresponding authors on reasonable request.

Declarations

Ethics approval and consent to participate

The Institutional Review Board at the Beth Israel Deaconess Medical Center granted a waiver of informed consent and approved the sharing of the research resource. Author Cui passed the online training courses and exams (certification number: 59921922). This study was conducted in accordance with the Helsinki Declaration. All authors have read and approved the final manuscript. As a result, neither consent for participation, nor ethical approval for this work were required.

Consent for publication

Not required as the manuscript does not contain individuals’ data.

Competing interests

The authors declare no competing interests.