Association between the advanced lung cancer inflammation index and all-cause mortality in critically ill patients with atrial fibrillation

Abstract

Evidence indicates that the advanced lung cancer inflammation index (ALI) is a prognostic tool for patient outcomes across various diseases. Nevertheless, its application in atrial fibrillation (AF) patients remains underexplored. This study aimed to evaluate the prognostic significance of ALI in critically ill AF patients. AF patients from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database were enrolled. ALI was evaluated using body mass index, serum albumin levels, and neutrophil-to-lymphocyte ratio. Patients were stratified into quartiles based on log₂-transformed ALI. Clinical endpoints were 30- and 90-day all-cause mortality (ACM). The relationship between ALI and patient outcomes was assessed using Cox proportional hazards models and Kaplan–Meier curves. Restricted cubic splines (RCS) were employed to investigate the nonlinear association between ALI and outcomes. Subgroup and sensitivity analyses were performed to verify the stability. Receiver operating characteristic curves evaluated ALI’s predictive efficacy. Multivariable Cox analysis revealed that a higher log2-ALI was linked to a reduced risk of 30-day ACM in AF patients (HR, 0.85; 95% CI: 0.79–0.92; P < .001). Compared with quartile Q1, quartiles Q2 (HR, 0.75; 95% CI: 0.59–0.96), Q3 (HR, 0.58; 95% CI: 0.44–0.78), and Q4 (HR, 0.52; 95% CI: 0.38–0.71) were associated with lower 30-day ACM risk. Similar results were observed for 90-day ACM. Kaplan–Meier curves showed that patients with higher log2-ALI levels had better survival rates at both 30 days and 90 days. RCS indicates a nonlinear association between log2-ALI and 30-day and 90-day ACM, revealing a threshold effect. When the indicator is below 5.39, higher log2-ALI levels are strongly linked to a lower ACM in AF patients, whereas values above 5.39 correlate with an increased ACM. These findings were consistent across most subgroups. The receiver operating characteristic curve demonstrates that ALI outperforms single nutritional or inflammatory markers in predicting prognosis in AF patients. Its predictive accuracy is further improved when combined with the sequential Organ Failure Assessment score. Elevated log₂-ALI is independently associated with reduced ACM risk in AF patients. Maintaining ALI within an optimal range via targeted interventions is crucial for reducing short-to-medium-term mortality, highlighting ALI’s potential as a valuable prognostic biomarker for risk assessment and individualized management.

1. Introduction

Atrial fibrillation (AF) represents a frequently diagnosed cardiac arrhythmia in clinical practice, particularly in intensive care units (ICU).^[1] Its incidence increases significantly with age and is especially prevalent among the elderly. According to the most recent findings from the Global Burden of Disease research, the worldwide prevalence of AF is approximately 59.7 million people.^[2,3] AF not only severely impacts patients’ quality of life, causing symptoms such as palpitations, fatigue, and dizziness, but is also closely associated with multiple serious complications. Most critically, it heightens the likelihood of heart failure and stroke, which significantly increase patients’ all-cause mortality (ACM).^[4] Patients with AF face a risk of ACM that is nearly double that of individuals without AF.^[5] This highlights the critical necessity for the early identification, thorough risk assessment, and effective clinical management of AF, which are essential challenges in cardiovascular disease care.

The advanced lung cancer inflammation index (ALI), originally derived from research on patients with advanced lung cancer, is a comprehensive indicator of inflammation and nutritional status. Its primary clinical purpose is to facilitate prognostic evaluation in oncology patients.^[6–8] Mounting evidence highlights the critical role of systemic inflammatory and nutritional status in determining disease progression and patient outcomes across various conditions.^[9–12] In the field of cardiovascular disease, inflammatory responses are strongly linked to pathological processes, including atherosclerosis, myocardial injury, and heart failure. Malnutrition also impacts cardiac function and the body’s immune capacity.^[13,14] Therefore, it is reasonable to speculate that ALI could be linked to the outcome of patients with AF. Current studies on ALI in the field of cardiovascular diseases mainly focus on specific conditions such as hypertension, heart failure, coronary heart disease, and myocardial infarction, and its value in prognostic assessment has been demonstrated.^[15–19] Nevertheless, direct evidence supporting the application of ALI for prognostic risk stratification in patients with AF remains insufficient, and systematic investigations into the association between ALI and mortality among AF patients are still lacking.

This retrospective cohort study was conducted using data from the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database. It aimed to systematically explore the association between ALI and ACM in critically ill patients with AF, to provide new evidence and references for the clinical management of AF.

2. Methods

2.1. Study population

For this investigation, data were extracted from MIMIC-IV (version 3.0), a publicly available critical care database. This resource contains detailed, de-identified medical records for ICU patients admitted to the Beth Israel Deaconess Medical Center in Boston between 2008 and 2022. Featuring rich data dimensions, a large sample size, and high authenticity and reliability, MIMIC-IV has been widely adopted for clinical research and prognostic analysis in critical care medicine. The MIMIC-IV database complies with the principles of the Declaration of Helsinki and has been approved by the Institutional Review Board of Beth Israel Deaconess Medical Center (Approval No.: 2001P-001699/14). All personally identifiable information of patients in the database has been de-identified and replaced with random identifiers, so the database qualifies for Institutional Review Board exemption. Accordingly, additional patient-informed consent and repeated ethical review are not required for studies based on this database. One of the authors of this study, Yingshan Lin, has successfully registered and completed the required courses to obtain permission to use this database (License ID: 66450944).

This study initially enrolled patients diagnosed with AF based on International Classification of Diseases (ICD) codes, specifically ICD-9: 42731; ICD-10: I48, I480, I481, I4811, I4819, I482, I4820, I4821, I489, and I4891(Supplementary Table S1, Supplemental Digital Content 1). Exclusion criteria are as follows: age under 18 years; patients with AF who are not initially ICU admissions; an ICU length of stay shorter than 24 hours; and patients who do not have the essential laboratory measures (height, weight, albumin, neutrophil count, lymphocyte count) needed to calculate ALI. Patients with ALI values falling within the 0 to 1st and 99 to 100th percentiles were also excluded. Finally, the study cohort included 1404 patients. Given the skewed distribution of ALI scores, a base-2 logarithmic conversion was used in the regression modeling to ensure the accuracy and validity of subsequent statistical analyses. Participants were then grouped into 4 categories according to log2-ALI quartile (Fig. 1).

Figure 1.

Population selection for the research.

2.2. Exposed variables and outcomes

The ALI is determined using the following formula: ALI = body mass index (BMI) × Albumin/neutrophil-to-lymphocyte ratio (NLR). In this equation, BMI is obtained by dividing weight (in kilograms) by the square of height (in meters). Albumin represents the concentration of serum albumin, measured in grams per deciliter, while NLR denotes the ratio of the absolute neutrophil count to the absolute lymphocyte count. The main outcomes of this research were 30-day and 90-day ACM following ICU admission.

2.3. Variable extraction

Data extraction was carried out using Navicat Premium (Version 16.0) software in conjunction with Structured Query Language. The following information was extracted from the MIMIC-IV database: demographic data: height, weight, age, gender, race; Laboratory parameters and vital signs: albumin, absolute neutrophil count, absolute lymphocyte count, white blood cells, hemoglobin, red cell distribution width, red blood cells, platelets, prothrombin time, international normalized ratio (INR), chloride, bicarbonate, potassium, calcium, sodium, creatinine, blood urea nitrogen (BUN), glucose, heart rate, mean arterial pressure, respiratory rate, temperature, and pulse oximetry oxygen saturation(SpO₂); score for illness severity: sequential organ failure assessment (SOFA); comorbidities: myocardial infarction, chronic kidney disease, malignant cancer, congestive heart failure (CHF), peripheral vascular disease (PVD), diabetes, hypertension, hyperlipidemia, stroke, chronic obstructive pulmonary disease; and treatment: Vasopressin, statins, beta-blockers, angiotensin converting enzyme Inhibitors/ Angiotensin II receptor blockers, digoxin, amiodarone, aspirin, warfarin, clopidogrel, heparin, mechanical ventilation, continuous renal replacement therapy. All laboratory parameters and vital signs data, including those required for ALI calculation (BMI, serum albumin, neutrophil count, and lymphocyte count), were obtained from the first recorded measurement within the first 24 hours of ICU admission. Covariates with missing data exceeding 20% were omitted from the statistical analysis. In this study, covariates like temperature, prothrombin time, INR, heart rate, and respiratory rate had missing values below 5%, and for these variables, the missing data were imputed using the mean value (Supplementary Table S2, Supplemental Digital Content 2).

2.4. Statistical analysis

The analysis and processing of data were performed using R version 4.3.3, alongside Zstats 1.0 (www.zstats.net). A P-value of <.05 was deemed statistically significant. Data following a normal distribution were presented as mean ± standard deviation and analyzed with either Student t test or analysis of variance. For data that did not follow a normal distribution, values were reported as median (interquartile range) and analyzed using the Mann–Whitney U test or Kruskal–Wallis test. Categorical variables were expressed as percentages and evaluated with the chi-square test or Fisher’s exact test.

Univariate and multivariate Cox regression models were used to assess the association between ALI and 30-day and 90-day ACM. Regression models were constructed through stepwise adjustment of covariates, correcting for confounding factors with P-values < .05 in univariate analysis and probable confounders determined using our team’s clinical experience. Crude model: unadjusted; Model 1: adjusted for age, sex, and race; Model 2: heart rate, mean arterial pressure, respiratory rate, temperature, SPO₂, white blood cell, red cell distribution width, INR, bicarbonate, potassium, creatinine, BUN, SOFA, malignant cancer, stroke, chronic obstructive pulmonary disease, vasopressin, statins, beta-blockers, angiotensin converting enzyme Inhibitors/ Angiotensin II receptor blockers, aspirin, warfarin, heparin, and continuous renal replacement therapy were further adjusted. Survival analysis was conducted using the Kaplan–Meier curves, while group differences were evaluated with the log-rank test. Restricted cubic splines (RCS) analysis was performed to explore potential nonlinear relationships between ALI and outcomes. Upon detecting nonlinearity, a recursive algorithm was employed to identify the inflexion point. The threshold effect was further estimated using a segmented Cox regression model. Subsequently, subgroup analyses were performed based on gender, age, and comorbidities. Sensitivity analyses were performed by excluding malignant cancer patients and patients with missing covariates to further validate the reliability of our findings. The ability of ALI and other markers, including BMI, albumin, NLR, and SOFA, to predict 30-day and 90-day ACM was evaluated using receiver operating characteristic analysis.

3. Result

3.1. Characteristics of the study population

Table 1 summarizes the baseline characteristics of 1404 critically ill patients with AF, distributed across log2-ALI quartiles. Males comprised 61% of the cohort, and the mean age was 72.26 ± 12.54 years. Subjects in the highest log2-ALI quartile exhibited higher values for weight, BMI, SPO₂, albumin, lymphocyte, ALI, hemoglobin, red blood cell, bicarbonate, and calcium, along with lower values for heart rate, respiratory rate, neutrophil, NLR, white blood cell, red cell distribution width, prothrombin time, INR, creatinine, BUN, glucose, SOFA score, and lower chronic kidney disease prevalence. In comparison to subjects in the Q1 (lowest log2-ALI quartile) group, those in the Q4 (highest log2-ALI quartile) group had lower 30-day ACM (44.44% vs 34.47% vs 22.51% vs 18.80%, P < .001) and 90-day ACM (52.14% vs 44.73% vs 29.63% vs 23.08%, P < .001).

Table 1.

Baseline characteristics of patients based on the quartiles of the log2-ALI at ICU admission.

Variables	Total (n = 1404)	Q1 (n = 351)	Q2 (n = 351)	Q3 (n = 351)	Q4 (n = 351)	P
Demographics
Age, yr	72.26 ± 12.54	73.56 ± 12.20	73.82 ± 11.85	71.25 ± 13.05	70.43 ± 12.72	<.001
Male gender, n (%)	858 (61.11)	217 (61.82)	220 (62.68)	205 (58.40)	216 (61.54)	.672
Race, n (%)						.009
white	889 (63.32)	229 (65.24)	221 (62.96)	226 (64.39)	213 (60.68)
black	71 (5.06)	15 (4.27)	13 (3.70)	11 (3.13)	32 (9.12)
other	444 (31.62)	107 (30.48)	117 (33.33)	114 (32.48)	106 (30.20)
Weight, Kg	86.46 ± 26.42	78.79 ± 21.33	83.97 ± 22.11	89.32 ± 27.76	93.75 ± 30.92	<.001
Height, Cm	169.87 ± 10.70	169.64 ± 10.39	169.69 ± 10.34	169.44 ± 11.15	170.71 ± 10.90	.396
BMI, Kg/m2	29.85 ± 8.36	27.28 ± 6.70	29.06 ± 6.88	30.98 ± 8.71	32.09 ± 9.92	<.001
Vital signs
HR, bpm	93.31 ± 23.48	97.74 ± 24.68	92.49 ± 23.05	94.08 ± 22.17	88.92 ± 23.18	<.001
MAP, mm Hg	82.40 ± 19.30	81.44 ± 20.12	82.16 ± 19.96	81.25 ± 18.81	84.77 ± 18.12	.058
RR, bpm	20.60 ± 6.09	21.80 ± 6.57	20.81 ± 5.98	20.63 ± 6.07	19.17 ± 5.41	<.001
Temperature, °C	36.67 ± 0.86	36.60 ± 0.86	36.73 ± 0.88	36.70 ± 0.90	36.63 ± 0.80	.175
SpO2,%	95.81 ± 5.14	94.95 ± 5.51	95.83 ± 5.20	95.69 ± 5.47	96.77 ± 4.09	<.001
Laboratory data
Albumin, g/dL	3.09 ± 0.65	2.77 ± 0.58	3.00 ± 0.61	3.16 ± 0.62	3.43 ± 0.60	<.001
Neutrophil, × 109/L	9.80 (6.35, 14.44)	14.40 (10.39, 19.45)	11.10 (7.70, 15.00)	8.98 (6.39, 12.31)	5.99 (4.21, 8.91)	<.001
Lymphocyte, × 109/L	0.94 (0.56, 1.46)	0.46 (0.30, 0.72)	0.80 (0.57, 1.16)	1.11 (0.81, 1.58)	1.66 (1.12, 2.45)	<.001
NLR	10.16 (5.67, 19.17)	30.45 (22.35, 43.99)	13.24 (10.50, 16.95)	7.89 (6.33, 9.75)	3.78 (2.70, 5.11)	<.001
ALI	8.55 (4.28, 17.07)	2.46 (1.66, 3.38)	6.37 (5.13, 7.42)	11.94 (9.91, 14.28)	26.65 (20.08, 37.93)	<.001
WBC, ×109/L	11.70 (8.10, 16.83)	16.00 (11.25, 21.85)	12.90 (9.50, 17.20)	10.90 (8.10, 14.40)	8.20 (6.30, 11.75)	<.001
Hemoglobin, g/dL	10.93 ± 2.38	10.73 ± 2.29	10.70 ± 2.32	10.99 ± 2.38	11.29 ± 2.50	.003
RDW, %	15.49 ± 2.52	15.76 ± 2.50	15.73 ± 2.53	15.29 ± 2.36	15.17 ± 2.62	.002
RBC, ×109/L	3.65 ± 0.82	3.58 ± 0.79	3.59 ± 0.82	3.66 ± 0.81	3.75 ± 0.84	.025
Platelet, ×109/L	190.00 (136.75, 262.25)	188.00 (131.50, 268.50)	188.00 (134.50, 258.50)	200.00 (143.50, 279.00)	184.00 (136.50, 250.50)	.164
PT, s	15.30 (13.10, 20.00)	15.70 (13.80, 20.80)	15.80 (13.40, 20.30)	15.10 (13.20, 20.00)	14.30 (12.60, 18.75)	<.001
INR	1.40 (1.20, 1.80)	1.40 (1.20, 1.90)	1.40 (1.20, 1.90)	1.40 (1.20, 1.80)	1.30 (1.10, 1.70)	<.001
Chloride, mmol/L	102.55 ± 7.23	102.88 ± 7.67	102.46 ± 7.81	102.11 ± 6.72	102.74 ± 6.65	.499
Bicarbonate, mEq/L	22.05 ± 5.34	21.00 ± 5.76	21.62 ± 5.62	22.87 ± 5.02	22.70 ± 4.67	<.001
Potassium, mmol/L	4.33 ± 0.82	4.37 ± 0.83	4.34 ± 0.83	4.28 ± 0.78	4.32 ± 0.84	.576
Calcium, mg/dL	8.29 ± 0.93	8.02 ± 0.97	8.27 ± 0.93	8.37 ± 0.84	8.51 ± 0.91	<.001
Sodium, mmol/L	138.08 ± 5.93	137.92 ± 6.33	138.18 ± 6.65	137.89 ± 5.46	138.33 ± 5.18	.723
Creatinine, mg/dL	1.30 (0.90, 2.10)	1.50 (0.90, 2.55)	1.40 (0.90, 2.20)	1.20 (0.90, 1.75)	1.10 (0.90, 1.70)	<.001
BUN, mg/dL	27.00 (18.00, 45.00)	36.00 (21.00, 55.00)	29.00 (19.00, 47.50)	25.00 (17.00, 36.50)	23.00 (16.00, 34.00)	<.001
Glucose, mEq/L	135.00 (107.00, 179.00)	145.00 (109.50, 188.50)	138.00 (109.00, 183.00)	139.00 (110.00, 177.50)	124.00 (102.00, 170.00)	.003
Disease severity score
SOFA	6.00 (4.00, 9.00)	7.00 (5.00, 10.00)	7.00 (4.00, 10.00)	6.00 (3.00, 9.00)	5.00 (3.00, 8.00)	<.001
Complication, n (%)
Myocardial infarct	415 (29.56)	100 (28.49)	106 (30.20)	111 (31.62)	98 (27.92)	.698
CKD	423 (30.13)	114 (32.48)	122 (34.76)	98 (27.92)	89 (25.36)	.028
Malignant cancer	194 (13.82)	62 (17.66)	43 (12.25)	45 (12.82)	44 (12.54)	.119
CHF	735 (52.35)	180 (51.28)	185 (52.71)	203 (57.83)	167 (47.58)	.055
PVD	219 (15.60)	50 (14.25)	66 (18.80)	55 (15.67)	48 (13.68)	.239
Diabetes	488 (34.76)	112 (31.91)	123 (35.04)	126 (35.90)	127 (36.18)	.618
Hypertension	1037 (73.86)	257 (73.22)	258 (73.50)	260 (74.07)	262 (74.64)	.975
Hyperlipidemia	591 (42.09)	150 (42.74)	129 (36.75)	152 (43.30)	160 (45.58)	.105
Stroke	221 (15.74)	45 (12.82)	57 (16.24)	51 (14.53)	68 (19.37)	.102
COPD	146 (10.40)	40 (11.40)	39 (11.11)	33 (9.40)	34 (9.69)	.770
Medication, n (%)
Vasopressin	304 (21.65)	99 (28.21)	96 (27.35)	58 (16.52)	51 (14.53)	<.001
Statins	704 (50.14)	157 (44.73)	163 (46.44)	175 (49.86)	209 (59.54)	<.001
Beta blocker	913 (65.03)	211 (60.11)	226 (64.39)	243 (69.23)	233 (66.38)	.079
ACEI/ARB	294 (20.94)	65 (18.52)	60 (17.09)	91 (25.93)	78 (22.22)	.019
Digitalis	160 (11.40)	47 (13.39)	34 (9.69)	42 (11.97)	37 (10.54)	.429
Amiodarone	481 (34.26)	115 (32.76)	126 (35.90)	122 (34.76)	118 (33.62)	.833
Aspirin	648 (46.15)	131 (37.32)	159 (45.30)	168 (47.86)	190 (54.13)	<.001
Warfarin	227 (16.17)	45 (12.82)	51 (14.53)	73 (20.80)	58 (16.52)	.027
Clopidogrel	127 (9.05)	26 (7.41)	33 (9.40)	39 (11.11)	29 (8.26)	0.350
Heparin	1132 (80.63)	307 (87.46)	289 (82.34)	277 (78.92)	259 (73.79)	<.001
Intervention, n (%)
Ventilation	1265 (90.10)	326 (92.88)	322 (91.74)	311 (88.60)	306 (87.18)	.040
CRRT	216 (15.38)	68 (19.37)	69 (19.66)	40 (11.40)	39 (11.11)	<.001
Outcome
30-d ICU mortality	422 (30.06)	156 (44.44)	121 (34.47)	79 (22.51)	66 (18.80)	<.001
90-d ICU mortality	525 (37.39)	183 (52.14)	157 (44.73)	104 (29.63)	81 (23.08)	<.001

Data are means ± SD, median (interquartile range), or n (%).

ACEI = angiotensin converting enzyme inhibitors, ALI = advanced lung cancer inflammation index, ARB = Angiotensin II receptor blockers, BMI = body mass index, BUN = blood urea nitrogen, CHF = congestive heart failure, CKD = chronic kidney disease, COPD = chronic obstructive pulmonary disease, CRRT = continuous renal replacement therapy, HR = heart rate, INR = international normalized ratio, MAP = mean arterial pressure, NLR = neutrophil-to-lymphocyte ratio, PT = prothrombin time, PVD = peripheral vascular disease, RBC = red blood cell, RDW = red cell distribution width, RR = respiratory rate, SOFA = sequential organ failure assessment, SPO₂ = pulse blood oxygen saturation, WBC = white blood cell.

3.2. Kaplan–Meier survival curve

The 4 log2-ALI quartile groups had significantly different 30- and 90-day ACM, according to The 4 log2-ALI quartile groups had significantly different 30- and 90-day ACM, according to Kaplan–Meier survival analysis (Fig. 2). Compared with patients with higher log2-ALI levels, the Q1 group exhibited the lowest 30- and 90-day survival rates (log-rank P < .001).

Figure 2.

Kaplan–Meier survival curves for mortality at 30 days (A) and 90 days (B) of log2-ALI groups after ICU admission.

3.3. Association between log2-ALI and outcomes

Supplementary Table S3, Supplemental Digital Content 3, presents the results of the univariate Cox proportional hazards model. The multivariate Cox proportional hazards model demonstrated a significant relationship between log2-ALI and 30-day ACM in patients with AF. In the fully adjusted model (Model 2), treating log2-ALI as a continuous variable, each one-unit increase in log2-ALI corresponded to a 15% decrease in 30-day ACM (HR = 0.85; 95% CI: 0.79–0.92; P < .001). When log2-ALI was categorized into quartiles, the risk of mortality in AF patients was reduced by 25%, 42%, and 48% in the Q2 (HR, 0.75; 95% CI: 0.59–0.96), Q3 (HR, 0.58; 95% CI: 0.44–0.78), and Q4 (HR, 0.52; 95% CI: 0.38–0.71) groups, respectively, compared to the Q1 group (Table 2).

Table 2.

Association between log2-ALI and mortality (Multivariate cox regression analysis).

Variables	Non-adjusted HR (95% CI) P-Value	Model 1 HR (95% CI) P-Value	Model 2 HR (95% CI) P-Value
30-d all-cause mortality
log2-ALI (continuous)	0.76 (0.71–0.81) < .001	0.76 (0.71–0.82) < .001	0.85 (0.79–0.92) < .001
log2-ALI (quartiles)
Q1	ref	ref	ref
Q2	0.73 (0.58–0.93) .010	0.73 (0.57–0.92) .008	0.75 (0.59–0.96) .024
Q3	0.44 (0.34–0.58) < .001	0.45 (0.34–0.59) < .001	0.58 (0.44–0.78) < .001
Q4	0.36 (0.27–0.47) < .001	0.37 (0.27–0.49) < .001	0.52 (0.38–0.71) < .001
P for trend	<.001	<.001	<.001
90-d all-cause mortality
log2-ALI (continuous)	0.76 (0.72–0.81) < .001	0.77 (0.72–0.82) < .001	0.86 (0.81–0.92) < .001
log2-ALI (quartiles)
Q1	ref	ref	ref
Q2	0.79 (0.64–0.98) .034	0.79 (0.64–0.98) .029	0.84 (0.67–1.05) .117
Q3	0.47 (0.37–0.60) <.001	0.49 (0.38–0.62) < .001	0.64 (0.50–0.83) < .001
Q4	0.35 (0.27–0.46) < .001	0.37 (0.28–0.48) < .001	0.52 (0.40–0.70) < .001
P for trend	<.001	<.001	<.001

Model 1: Age, Gender, Race.

Model 2: Age, Gender, Race, Heart Rate, Mean Arterial Pressure, Respiratory Rate, Temperature, SPO2, White Blood Cell, Red Cell Distribution Width, INR, Bicarbonate, Potassium, Creatinine, BUN, SOFA, Malignant Cancer, Stroke, COPD, Vasopressin, Statin, Beta-blockers, ACEI/ARB, Aspirin, Warfarin, Heparin, CRRT.

ACEI = angiotensin converting enzyme inhibitors, ALI = advanced lung cancer inflammation index, ARB = Angiotensin II receptor blockers, BUN = blood urea nitrogen, CI =confidence interval, COPD = chronic obstructive pulmonary disease, CRRT = Continuous Renal Replacement Therapy, HR = hazard ratio, INR = international normalized ratio, SOFA = sequential organ failure assessment, SPO2 = pulse blood oxygen saturation.

A similar relationship was observed for 90-day ACM. In Model 2, each one-unit rise in log2-ALI, treated as a continuous variable, was associated with a 14% reduction in 90-day ACM (HR = 0.86; 95% CI: 0.81–0.92; P < .001). When categorized into quartiles, the Q2 (HR, 0.84; 95% CI: 0.67–1.05), Q3 (HR, 0.64; 95% CI: 0.50–0.83), and Q4 (HR, 0.52; 95% CI: 0.40–0.70) groups all showed varying levels of decrease compared to the Q1 group.

3.4. The detection of nonlinear relationships

Multivariate adjustment of the RCS using covariates from Model 2 demonstrated a nonlinear association between log2-ALI and ACM at 30- and 90 days (Fig. 3, nonlinear P < .001). Specifically, log2-ALI values under 5.39 were linked to a reduced risk of 30-day ACM (HR, 0.81; 95% CI: 0.75–0.87), while values exceeding 5.39 corresponded to an elevated risk of ACM (HR, 2.43; 95% CI: 1.46–4.04). A comparable association was noted for the 90-day ACM (Table 3).

Figure 3.

Multivariable-adjusted RCS regression showing the nonlinear relationship between log2-ALI and 30-day (A) and 90-day mortality (B).

Table 3.

Threshold effect analysis of the relationship between log2-ALI and all-cause mortality.

Threshold of log2-ALI	30-d all-cause mortality	P-value
	HR (95%CI)
< 5.39	0.81 (0.75, 0.87)	<.001
≥ 5.39	2.43 (1.46, 4.04)	<.001
Likelihood ratio test		<.001
Threshold of log2-ALI	90-d all-cause mortality	P-value
	HR (95%CI)
< 5.39	0.83 (0.77, 0.89)	<.001
≥ 5.39	2.01 (1.23, 3.30)	.006
Likelihood ratio test		.003

HRs were adjusted for age, gender, race, heart rate, mean arterial pressure, respiratory rate, Temperature, SPO2, white blood cell, red cell distribution width, INR, bicarbonate, potassium, creatinine, BUN, SOFA, malignant cancer, stroke, COPD, vasopressin, statin, beta-blockers, ACEI/ARB, aspirin, warfarin, heparin, CRRT.

ACEI = angiotensin converting enzyme inhibitors, ALI = advanced lung cancer inflammation index, ARB = Angiotensin II receptor blockers, BUN = blood urea nitrogen, CI = confidence interval, COPD = chronic obstructive pulmonary disease, CRRT = Continuous Renal Replacement Therapy, HR = hazard ratio, INR = international normalized ratio, SOFA = sequential organ failure assessment, SPO2 = pulse blood oxygen saturation.

3.5. Subgroup analysis and sensitivity analysis

In order to assess the reliability of our findings and investigate potential subgroup variations, we performed subgroup analyses (Fig. 4). Subgroup analysis and interaction tests indicate that higher log2-ALI is linked to reduced 30-day and 90-day ACM, with this association remaining consistent across most subgroups. Nonetheless, a significant interaction was identified within the myocardial infarction subgroup for 30-day ACM (interaction P = .044).

Figure 4.

Forest plot of subgroup analysis of the relationship between log2-ALI and all-cause mortality at 30 days (A) and 90 days (B) in patients with AF. AF = atrial fibrillation.

Subsequently, after excluding participants with malignant cancer, 1210 individuals remained, and the data were reanalyzed using Cox regression. In Model 2, when log2-ALI was divided into quartiles, the Q4 group demonstrated a 40% reduction in 30-day mortality (HR = 0.60, P = .003), and 90-day mortality decreased by 42% (HR = 0.58, P < .001) compared to the Q1 group (Supplementary Table S4, Supplemental Digital Content 4).

Following the exclusion of participants with missing covariates, 1299 individuals were included in the analysis. In Model 2, when log2-ALI was divided into quartiles, the Q4 group demonstrated a 51% reduction in 30-day mortality (HR = 0.49, P < .001), with a 50% reduction in 90-day mortality (HR = 0.50, P < .001) compared to those in the Q1 group (Supplementary Table S5, Supplemental Digital Content 5). In general, the outcomes of the sensitivity analysis are robust.

3.6. Receiver operating characteristic curve analysis

Receiver operating characteristic analysis was employed to assess the ability of ALI, BMI, albumin, NLR, and SOFA to predict mortality. ALI demonstrated superior predictive ability for both 30-day and 90-day ACM compared to BMI, albumin, and NLR but was inferior to SOFA for 30-day prediction (Fig. 5; Table 4). However, when combined with SOFA, ALI significantly improved the predictive capability of SOFA (0.697 vs 0.643, P < .001, and 0.674 vs 0.647, P < .001).

Figure 5.

Receiver operating characteristic curves for ALI to predict all-cause mortality at 30 days (A) and 90 days (B). ALI = Advanced Lung Cancer Inflammation Index.

Table 4.

Predictive performance of inflammatory and nutritional indicators for mortality.

Model	30-d mortality		90-d mortality
	AUC	95% CIs	AUC	95% CIs
ALI	0.643	0.618, 0.668	0.647	0.621, 0.672
BMI	0.505	0.479, 0.532	0.547	0.520, 0.573
ALB	0.607	0.581, 0.633	0.608	0.582, 0.634
NLR	0.629	0.603, 0.654	0.622	0.596, 0.648
SOFA	0.661	0.635, 0.686	0.628	0.602, 0.654
ALI + SOFA	0.697	0.672, 0.720	0.674	0.649, 0.699

ALB = albumin, ALI = advanced lung cancer inflammation index, AUC = area under the curve, BMI = body mass index, CI = confidence interval, NLR = neutrophil-lymphocyte ratio, SOFA = sequential organ failure assessment.

4. Discussion

We observed that higher log2-ALI was independently associated with a reduced risk of 30-day and 90-day ACM in patients with AF, with this association remaining significant even after adjusting for multiple confounding factors. Log2-ALI and short- and medium-term mortality had a nonlinear association with a threshold saturation effect, according to RCS analysis. Specifically, a log₂-ALI value below 5.39 was inversely associated with mortality risk. Conversely, a log2-ALI exceeding 5.39 was associated with an elevated risk of mortality. This association was particularly pronounced in the 30-day mortality risk among the non-myocardial infarction cohorts. Furthermore, the ALI demonstrated enhanced predictive performance for patient outcomes relative to other commonly utilized markers of nutrition and inflammation.

AF represents a common cardiac arrhythmia characterized by rapid and disorganized electrical activity within the atria. This abnormal electrical activity leads to ineffective atrial contractions, which result in compromised cardiac function. Inflammatory mediators, including interleukin-6, C-reactive protein, and NLR, play significant roles in the onset and progression of AF.^[20–22] Inflammation-mediated atrial electrophysiological and structural remodeling are important factors in the onset of AF. Conversely, the active state of AF can trigger inflammatory responses.^[23–26] In addition to inflammation, nutrition is another area of focus. Budzyński et al.^[27] demonstrated that nutritional assessments, including the NRS-2002 score, independently predict in-hospital mortality and 30-day readmission among patients who were diagnosed with AF. Czapla et al.^[28] explored the link between nutritional status and the duration of hospitalization among patients with AF. Malnutrition was identified as a risk factor for extended hospital stays in this group, along with reduced serum levels of potassium, sodium, and low- and high-density lipoprotein. Systematic nutritional assessment, appropriate dietary patterns, and targeted micronutrient supplementation are essential for lowering the risk of AF and enhancing patient outcomes.

As a composite indicator, ALI integrates BMI, serum albumin, and NLR to offer a thorough assessment of inflammatory status and nutritional condition. ALI was originally designed to evaluate survival outcomes in patients with lung cancer. Subsequent research has expanded its application to predict prognostic outcomes for multiple malignancies and chronic diseases.^[29–33] In the realm of cardiovascular disorders, ALI has also demonstrated its potential prognostic significance. For instance, a cohort study involving subjects with coronary artery disease who received percutaneous coronary intervention revealed that an ALI level <334.96 independently predicted major unfavorable cardiovascular outcomes, with a 5-year area under the curve of 0.749.^[30] Tu et al^[15] also confirmed using NHANES data from over 20,000 hypertensive adults and found that those in the highest ALI tertile experienced a 52% reduction in the risk of cardiovascular mortality in comparison to those in the lowest tertile. Fan et al.^[34] Further incorporated ALI into the nomogram, increasing the area under the curve for predicting coronary heart disease and coronary artery calcification to 0.739 and 0.728, respectively. This research broadens the use of ALI in cardiovascular disease, marking the initial exploration of its connection to ACM in patients with AF. The findings highlight a nonlinear relationship along with a threshold effect. This indicates that a comprehensive evaluation of both inflammation and nutritional status is crucial when evaluating the prognosis of AF patients, with ALI potentially serving as an effective integrated indicator.

The results presented by this composite index integrating BMI, albumin, and NLR may be attributed to multiple underlying mechanisms. First, NLR reflects systemic inflammatory status: neutrophilia indicates acute inflammatory activation, while lymphopenia suggests suppressed immune regulation.^[35] Previous research has demonstrated a strong link between NLR and the onset, progression, and postoperative recurrence of AF, exhibiting predictive value across different populations.^[36–39] During the pathogenesis and perpetuation of AF, chronic inflammation continuously releases pivotal pro-inflammatory mediators, including interleukin-6, tumor necrosis factor-α, and C-reactive protein, which directly contribute to AF-specific atrial structural remodeling, atrial fibrosis, and electrical remodeling.^[23] Inflammatory signals markedly activate atrial fibroblasts and promote excessive collagen deposition as well as extracellular matrix remodeling, resulting in atrial interstitial fibrosis, atrial dilatation, and impaired systolic function. Meanwhile, inflammation can disrupt the regulation of myocardial calcium homeostasis and the function of various ion channels, leading to abnormalities in action potential duration, increased conduction heterogeneity, and the formation of reentry circuits, which ultimately constitute the electrophysiological basis for the onset and persistence of AF.^[40,41] Inflammation induces excessive reactive oxygen species production and activates signaling pathways such as nuclear factor-κB and mitogen-activated protein kinases. It also triggers inflammasome activation, further exacerbating mitochondrial dysfunction and cardiomyocyte apoptosis and accelerating the progression of atrial remodeling.^[42] Furthermore, elevated NLR is strongly associated with left atrial thrombosis, a mechanism potentially involving inflammation-mediated upregulation of procoagulant factors (such as tissue factor and fibrinogen) and downregulation of anticoagulant factors and fibrinolytic activity, thereby exacerbating the tendency toward thrombosis. Concurrently, inflammatory mediators can enhance platelet reactivity, further promoting the development of a prothrombotic state.^[43–45] Elevated ALI levels reflect a relatively controlled inflammatory environment, consistent with improved prognosis.

Second, hypoalbuminemia is a classic indicator of the severityof malnutrition and chronic disease. According to a dose-response meta-analysis, albumin levels are significantly associated with AF risk, exhibiting an inverse relationship. Specifically, a 10 g/L increase in serum albumin is associated with a 36% reduction in the risk of developing AF.^[46] Genetic evidence supports a protective causal effect of serum albumin in AF. Chen et al.^[47] employed Mendelian randomization to demonstrate that genetically predicted higher serum albumin levels were strongly linked to a reduced risk of AF. Low serum albumin levels may play a significant role in cardiovascular disease, primarily due to their influence on the regulation of colloid osmotic pressure, as well as their anti-inflammatory, antioxidant, and antithrombotic effects.^[48] Finally, BMI is a widely recognized measure for evaluating nutritional status and body shape, with a value ≥ 30 kg/m² generally indicating obesity. Obesity is well-established as a key risk factor for cardiovascular diseases, with a crucial role in forecasting both ACM and cardiovascular events.^[49] However, in certain clinical settings, patients with AF who have a higher BMI tend to exhibit lower mortality rates, an observation frequently described as the obesity paradox.^[50,51] At present, the link between obesity and prognosis in AF patients remains inconclusive. Nevertheless, existing evidence suggests that there is not a simple linear association between BMI and the prognosis of AF patients. A U-shaped correlation between BMI and the risk of AF recurrence was identified in a large-scale registry analysis of those recovering following AF ablation.^[52] The study indicated that not only did obesity (BMI ≥ 30 kg/m²) correlate with raised recurrence likelihood (HR = 1.78, 95% CI: 1.17–2.72), but also underweight status (BMI < 18.5 kg/m²) significantly increased recurrence likelihood (HR = 1.85, 95% CI: 1.12–3.08). Research by Ntel et al^[53] similarly indicates that a BMI exceeding 25 kg/m² raises the risk of hospitalization and death, while a BMI lower than 18 kg/m² also elevates mortality risk. Furthermore, in individuals with AF, being overweight or obese increases the risk of ischemic stroke, thrombosis, or mortality.^[54] Obesity mediates the onset and progression of AF through multiple known pathways, including inflammation, neurohormonal activation, the influence of cardiac fat stores, and alterations in cardiac structure and function.^[55,56] Our findings align with the notion that, within the composite ALI framework, the protective effect of BMI is likely confined to an “optimal window.” When the aggregate ALI remains low, indicative of pronounced inflammation and poor nutritional reserve, an elevated BMI may furnish essential metabolic support. Once the ALI surpasses the inflection point, however, the pathological burden intrinsic to obesity itself becomes dominant, neutralizing its earlier protective influence.

Subgroup analysis showed a significant interaction in the myocardial infarction subgroup, where non-myocardial infarction patients exhibited a stronger association between ALI and ACM. An acute myocardial infarction triggers a rapid, explosive systemic inflammatory response and a myocardial necrosis process. Short-term prognosis is primarily influenced by coronary reperfusion status, infarct size, and acute complications such as cardiogenic shock and fatal arrhythmias.^[57–59]

From a clinical perspective, the ALI index serves as a simple and effective tool for prognostic evaluation in patients with AF. All parameters required for its calculation are obtainable from routine clinical examinations, eliminating additional testing costs and complicated operational procedures, which render it highly feasible for widespread clinical application. Receiver operating characteristic curve analyses in this study verified that ALI outperformed individual nutritional or inflammatory biomarkers in predicting the prognosis of AF patients. Moreover, its predictive performance was markedly improved when combined with the SOFA score. These findings indicate that ALI can not only provide independent prognostic information but also exert synergistic effects with other indicators, thereby optimizing the risk stratification for mortality among critically ill AF patients. Composed of BMI, albumin, and NLR, ALI directly reflects the balance between nutritional reserve and inflammatory burden, providing a valuable reference for early targeted intervention and individualized nutritional management. A low ALI value is generally accompanied by decreased albumin, reduced BMI, and elevated NLR, indicating malnutrition and persistent inflammation. Accordingly, clinicians can formulate personalized nutritional and therapeutic strategies, including albumin supplementation, enteral or parenteral nutrition support, and anti-inflammatory treatment. Conversely, an excessively high ALI, mainly attributed to elevated BMI or high albumin levels, suggests the need to address obesity-related cardiovascular burden and implement weight management and metabolic interventions. In summary, ALI assessment facilitates clinical practice in multiple aspects, including risk stratification, early decision-making for interventions, and guidance for nutritional support and inflammation control. The management of AF should not merely focus on pharmacological therapy but also incorporate inflammatory and nutritional status into overall treatment strategies. Concurrent monitoring of ALI and its components, combined with comprehensive interventions, may contribute to better long-term management of AF patients. Further prospective studies are required to validate these potential benefits and confirm whether ALI-guided interventions can translate into improved clinical outcomes.

However, there are several limitations in this research. First, although we identified a correlation between ALI and ACM in AF patients due to its retrospective observational design, a direct causative relationship cannot be deduced. Second, we selected BMI and albumin as markers to assess overall nutritional status. Despite their utility, both metrics are subject to limitations related to their sensitivity and specificity. Additionally, NLR could be influenced by unconsidered inflammatory factors. Third, we only utilized baseline ALI data obtained upon ICU admission. This static measurement approach fails to capture the dynamic evolution of ALI during hospitalization, which may hold greater prognostic value. Fourth, the diagnosis of AF in this study was solely based on ICD codes, without verification using electrocardiography (ECG) or Holter monitoring. This diagnostic approach may introduce misclassification bias. For instance, some atypical cases might be missed, or patients without AF could be incorrectly enrolled due to coding errors. In future studies, stricter diagnostic criteria for AF will be adopted, combining ECG, Holter findings, and clinical medical records to further enhance the reliability and generalizability of the results. Finally, our research utilizes data from the MIMIC-IV database. However, further validation is needed to confirm the generalizability of these findings.

5. Conclusion

The present study revealed that log₂-ALI was independently and significantly correlated with short- and medium-term ACM in critically ill patients with AF. As a readily available and low-cost composite indicator reflecting inflammation and nutritional status, ALI has the potential to serve as a reference for the comprehensive assessment of inflammatory and nutritional conditions in this patient population and is conducive to prognostic risk stratification and clinical management. As a retrospective observational analysis based on the MIMIC-IV database, this study cannot establish causal relationships. The generalizability and clinical applicability of the findings are limited. Large-sample prospective studies and external cohort validation are warranted in future research to further clarify its clinical value.

Acknowledgments

YL and RL contributed equally to this article. LZ and YL contributed equally to this work and share senior authorship.

The authors sincerely thank all institutions and investigators responsible for curating and providing publicly accessible databases used in the present study.

Author contributions

Data curation: Yingshan Lin, Renzhe Lin, Duo Yang, Shujun Ye.

Formal analysis: Yingshan Lin.

Writing – review & editing: Yingshan Lin.

Writing – original draft: Renzhe Lin, Duo Yang, Shujun Ye.

Project administration: Longsheng Zhang.

Supervision: Longsheng Zhang.

Validation: Longsheng Zhang, Yuxin Lin.

Software: Yuxin Lin.