Introduction

Coronary artery bypass grafting (CABG) remains the gold-standard treatment for multivessel coronary artery disease, particularly in patients with diabetes mellitus, by restoring normal coronary blood flow and significantly improving long-term survival. Globally, approximately 470,000 CABG and valve replacement procedures are performed annually, establishing CABG as one of the most frequently performed cardiovascular interventions1,2,3. However, recent high-quality studies have demonstrated that female patients undergoing CABG experience higher short- and long-term mortality compared with male patients4,5,6,7, underscoring the need for more proactive management in women8. It is therefore essential to recognize women as a distinct patient subgroup and to establish simple, clinically applicable risk indicators tailored to this population. Such tools may facilitate the development of targeted clinical strategies and improve early postoperative outcomes9.

Inflammation plays a central role in atherosclerosis, and inflammatory markers may be valuable for risk assessment and management of CABG patients10,11. Among these markers, fibrinogen—a major acute phase reactant and coagulation factor - has emerged as a potential prognostic indicator in cardiovascular disease12. Fibrinogen not only mediates platelet aggregation13, but also contributes significantly to the development and progression of atheromatous plaques14. Previous studies have shown that abnormal fibrinogen levels are associated with an increased risk of mortality in several cardiovascular diseases, including heart failure15, myocardial infarction16, coronary heart disease17, end-stage renal disease18, even in asymptomatic individuals19. Research in CABG has identified preoperative fibrinogen levels as important predictors of mortality20, hospital length of stay21 and graft failure22. Nevertheless, factors such as cardiopulmonary bypass (CPB), aprotinin administration, hemofiltration, and surgical duration can influence fibrinogen levels, suggesting that early postoperative fibrinogen measurement may offer superior prognostic value compared with preoperative assessment23.

To date, no large-scale study has systematically evaluated the association between postoperative fibrinogen levels and one-year mortality specifically among female CABG patients. Using the Medical Information Mart for Intensive Care (MIMIC-IV) database, this study aims to elucidate the relationship between fibrinogen levels and 1-year mortality in women undergoing CABG. By analyzing a robust real-world cohort, we seek to determine whether fibrinogen acts as an independent risk factor for mortality in this high-risk population, while adjusting for potential confounders.

Method

Study design and ethics

This study employed a retrospective cohort design using deidentified patient-level data from intensive care unit (ICU) admissions. We accessed the MIMIC-IV database, an enhanced iteration of MIMIC-III that includes updated data and reconstructed tables containing clinical information for over 190,000 patients and 450,000 hospitalisations at Beth Israel Deaconess Medical Center (BIDMC) in Boston, MA, USA24. Access to this data was granted following completion of the National Institutes of Health Human Research Participant Protection course and successful completion of the Collaborative Institutional Training Initiative (CITI) review. The study adhered to the tenets of the 1975 Declaration of Helsinki. The Medical Information Mart for Intensive Care is publicly available and has received Institutional Review Board (IRB) approval from BIDMC(2001-P-001699/14) and the Massachusetts Institute of Technology (0403000206).

Selection of participants

The analysis included 6318 patients aged 18 years or older who underwent CABG and were admitted to the ICU for the first time. Exclusion criteria comprised: (1) repeat ICU admissions; (2) lack of fibrinogen data within 24 h of admission; and (3) male patients. After applying these criteria, the final cohort comprised 1097 patients (Fig. 1).

Fig. 1
figure 1

Flowchart of the study cohort selection criteria.

Full size image

Covariates extraction

Initial measurements taken within the first 24 h of ICU admission served as the parameters for this study. To extract the relevant patient information, we utilized SQL alongside Navicat software (version 15). The variables analyzed were categorized as follows: (1) Demographics: age and BMI; (2) Vital Signs: arterial systolic pressure (ABPS), arterial diastolic pressure (ABPD), temperature, and heart rate; (3) Blood Biochemical Markers: white blood cell (WBC) count, red blood cell (RBC) count, platelet count, sodium, potassium, total calcium, chloride, glucose, anion gap, lactate, urea nitrogen, and creatinine; and (4) Clinical Scores: Sequential Organ Failure Assessment (SOFA), Acute Physiology Score III (APS III), Simplified Acute Physiology Score II (SAPS II), Glasgow Coma Scale (GCS), and Charlson Comorbidity Index (CHARLSON).

Expose and outcome

The primary exposure variable was early fibrinogen measurement, defined as the first fibrinogen level within 24 h postoperatively, a method validated in previous studies15. Patients were divided into six groups based on serum fibrinogen levels: Q1 (< 155 mg/dL), Q2 (155–178 mg/dL), Q3 (178–201 mg/dL), Q4 (201–230 mg/dL), Q5 (230–267 mg/dL) and Q6 (>267 mg/dL). This method ensures that the sample size of each group is relatively balanced. It also facilitates comparison of clinical outcomes among patients with different fibrinogen levels, thereby minimizing bias. For the MIMIC-IV data, the primary outcome was one-year mortality, with secondary outcomes including sepsis and length of hospital stay. Follow-up began at the time of patient admission, and mortality data were obtained from the US Social Security Death Index.

Statistical analyses

Continuous variables are presented as mean ± standard deviation or median (interquartile range), as appropriate, and categorical variables are presented as proportions. Statistical analyses included Student’s t-test, analysis of variance and Mann-Whitney U test, where appropriate. Categorical variables were compared using the χ² test.

Cox regression analyses used multivariate models to assess the predictive role of fibrinogen on mortality, considering fibrinogen levels as both categorical and continuous variables. Confounders were selected on the basis of statistical significance and clinical relevance. Five Cox regression models were developed, with results reported as hazard ratios and 95% confidence intervals (CIs). Selection of covariates for the final model was based on clinical significance, previous studies, and degree of correlation with the exposure. We check the models for multicollinearity, and we did not find multicollinearity between the variables in models. Model 1 included no covariates. Model 2 adjusted for demographic variables, including age and BMI. In Model 3, eleven additional covariates related to basic vital signs were incorporated, such as heart rate, ABPS, ABPD, and temperature. Model 4 further included eleven blood biochemical indicators based on Model 3, encompassing WBC count, RBC count, platelet count, sodium, potassium, total calcium, chloride, glucose, anion gap, lactate, urea nitrogen, and creatinine. Finally, Model 5 added eleven clinical scores based on Model 4, including the SOFA, APS III, SAPS II, GCS, and CHARLSON. Fibrinogen levels were further evaluated as a continuous variable using RCS to investigate the dose-response relationship with long-term mortality risk. Non-linear associations were assessed using a recursive algorithm to identify inflection points for fibrinogen and one-year mortality. A two-stage Cox proportional hazards model was used to analyse the relationship before and after the identified inflection point.

Sensitivity analyses were performed to assess the robustness of the results. In addition to the primary outcomes, secondary outcomes were analysed using regression models. Stratified analyses examined whether the effect of fibrinogen differed between subgroups defined by BMI, SOFA, APS III, SAPS II and CHARLSON scores. Multiplicative interactions were tested using interaction terms and the likelihood ratio test. Missing data were handled using predictive mean adjustment with the Multivariate Imputation via Chained Equations method in R. Analyses of imputed datasets confirmed the stability of the results25.

All analyses were conducted using R (version 4.1.1) and Free Statistics software (version 1.8). A two-tailed P < 0.05 was considered statistically significant, unless otherwise specified. The STROBE checklist was adhered to for reporting purposes.

Results

Subject characteristics

Table 1 summarizes the baseline characteristics of the 1,097 participants categorized into six quartiles based on fibrinogen levels. The mean age of the cohort was 70.8 years, with significant age variations across the quartiles (p = 0.002); notably, patients in Q1 had a mean age of 73.0 ± 10.4 years, while those in Q6 were younger, averaging 68.5 ± 10.8 years. BMI differed significantly among the groups (p < 0.001), with Q6 exhibiting the highest mean BMI of 33.9 ± 7.4. Although vital signs, including heart rate, ABPS, and ABPD, did not show significant differences across quartiles, several blood biochemical markers did. These included WBC) counts (p < 0.001), RBC counts (p = 0.001), platelet counts (p < 0.001), sodium levels (p < 0.001), potassium levels (p < 0.001), calcium (p = 0.023), chloride (p < 0.001), lactate (p < 0.001), urea nitrogen (p < 0.001), and creatinine (p < 0.001). The SOFA score also varied significantly across quartiles (p = 0.001), reflecting differences in organ dysfunction. Regarding clinical outcomes, the length of hospital stay was significantly longer in Q6 (12.2 ± 9.9 days) compared to other quartiles (p < 0.001). The one-year mortality rate varied markedly, with 9.6% in Q1 compared to 3.3% in Q5, indicating a significant link between fibrinogen levels and mortality (p = 0.013). Furthermore, the incidence of sepsis was notably higher in Q1 (64%) than in Q6 (39.1%) (p < 0.001).

Table 1 Baseline characteristics of participants.
Full size table

Relationship between fibrinogen and clinical outcomes in female patients with CABG

Table 2 details the results from the multivariable Cox regression analyses that evaluated the relationship between fibrinogen levels and one-year mortality. In Model 1, the hazard ratio (HR) for the continuous fibrinogen variable was not significant (HR: 1.02, p = 0.095); however, it became significant in Model 2 (HR: 1.03, p = 0.028) and Model 3 (HR: 1.02, p = 0.037), suggesting that higher fibrinogen levels may be associated with an increased mortality risk. This association diminished in Models 4 and 5, where the HRs were 1.02 (p = 0.308) and 1.01 (p = 0.563), respectively. When considering fibrinogen as a categorical variable, patients in the lowest quartile (Q1, ≤ 155 mg/dL) exhibited a significantly elevated risk of mortality, with HRs ranging from 2.72 (p = 0.026) in Model 1 to 2.91 (p = 0.026) in Model 5. In contrast, patients in the second (Q2, 155–178 mg/dL) and third quartiles (Q3, 178–201 mg/dL) did not show significant mortality risks compared to the reference group (Q4, 201–230 mg/dL). Notably, patients in the highest quartile (Q6, > 267 mg/dL) demonstrated a significantly increased mortality risk, with HRs of 2.71 (p = 0.025) in Model 1 and 4.30 (p = 0.003) in Model 5. Despite these findings, the trend across quartiles did not reach statistical significance (p for trend > 0.05), indicating a nonlinear relationship between fibrinogen levels and mortality.

Dose–response relationships

The analysis of RCS revealed a nonlinear, U-shaped relationship between fibrinogen levels and mortality across one year (Fig. 2). To investigate this association further, Cox proportional hazards models and two-piecewise Cox models were employed, yielding significant results (P for log-likelihood ratio < 0.05 in both models) (Table 3). An inflection point was identified at 174 mg/dL. Below this threshold, fibrinogen levels were inversely related to mortality (HR 0.969, 95% CI 0.947–0.991), whereas levels above this threshold correlated with increased mortality risk (HR 1.0057, 95% CI 1.0001–1.0114). SFigure 1 shows the Kaplan-Meier curve for patients in the different fibrinogen groups. A too-high or too-low fibrinogen was significantly associated with risk of 90-day mortality (P = 0.013 by the logrank test).

Fig. 2
figure 2

Dose–response relationships between fibrinogen with 1-year mortality rate. Solid and dashed lines represent the predicted value and 95% confidence intervals. Adjusted for demographic variables (age, BMI), Basic vital signs (heart rate, ABPS, ABPD, Temperature), Blood biochemical indicators (WBC, RBC, platelet count, sodium, potassium, calcium total, chloride, glucose, anion gap, lactate, urea nitrogen, creatinine) SOFA, APS Ⅲ, SAPSⅡ, GCS, HARLSON Abbreviations: the unit of fibrinogen is mmol/L, BMI: body mass index; ABPS: Arterial systolic pressure; ABPD: Arterial diastolic pressure; WBC, white blood cell count; RBC, red blood cell count; sofa, Sequential Organ Failure Assessment; APS Ⅲ: Acute Physiology and Chronic Health Evaluation Ⅲ; SAPSⅡ: simplified acute physiology score ii, GCS, Glasgow Coma Scale; CHARLSON, Charlson comorbidity index, AKI: acute kidney injury, CRRT, Continuous Renal Replacement Therapy; CI: confidence interval; HR: hazard ratio; Ref: reference

Full size image
Table 2 Multivariable cox regression to assess the association of fibrinogen with mortality
Full size table
Table 3 Threshold effect analysis of relationship of fibrinogen with one year mortality rate.
Full size table

Subgroup

Figure 3 illustrates the results from the stratified analysis of the relationship between serum fibrinogen levels and one-year mortality, showing adjusted hazard ratios (HRs) across various subgroups. We investigated fibrinogen’s predictive capacity regarding mortality outcomes across different population segments defined by BMI, SOFA, APS III, SAPS II and Charlson scores. No statistically significant interactions between the variables were identified, and the relationship between fibrinogen levels and mortality was consistent in direction. The stratified analysis results indicated a similar association of fibrinogen levels across most subpopulations.

Fig. 3
figure 3

Stratified analysis of the association of association of serum fibrinogen with one-year mortality. Adjusted HRs are shown. Interaction was tested adjusted. Subgroups with and adjusted P-interaction < 0.05 were considered effect modifiers on the association of serum fibrinogen with one-year mortality.

Full size image

Sensitivity analysis

Several sensitivity analyses were conducted to evaluate the robustness of our findings. First, Table S1 summarizes the results from the multivariable Cox regression analyses concerning the association between fibrinogen levels and sepsis risk. The continuous fibrinogen variable consistently exhibited a significant inverse relationship with sepsis across all models, with HRs ranging from 0.95 to 0.96 and p-values below 0.003, suggesting that higher fibrinogen levels are associated with a reduced sepsis risk. When evaluated as a categorical variable, patients in the second quartile (Q2, 155–178 mg/dL) displayed a notably lower risk of sepsis compared to the reference group (Q1, ≤ 155 mg/dL), with HRs decreasing from 0.63 (p = 0.032) in Model 1 to 0.55 (p = 0.011) in Model 5. Patients in the third quartile (Q3, 178–201 mg/dL) also showed a substantial reduction in sepsis risk, with HRs ranging from 0.51 (p = 0.002) to 0.50 (p = 0.002) across models. The fourth quartile (Q4, 201–230 mg/dL) demonstrated a significant decrease in sepsis risk as well, with HRs ranging from 0.40 (p < 0.001) to 0.37 (p < 0.001). Furthermore, patients in the fifth (Q5, 230–267 mg/dL) and sixth quartiles (Q6, > 267 mg/dL) exhibited similar trends, with HRs of 0.39 (p < 0.001) and 0.36 (p < 0.001) for Q5, and 0.32 (p < 0.001) and 0.31 (p < 0.001) for Q6, respectively. The trend across quartiles exhibited statistical significance (p for trend < 0.001), reinforcing the association between higher fibrinogen levels and a decreased risk of sepsis.

Second, Table S2 displays the results of multivariable linear regression analyses assessing the association between fibrinogen levels and hospital length of stay. The continuous fibrinogen variable consistently demonstrated a significant positive correlation with length of stay across all models, with β coefficients ranging from 0.15 to 0.20 and p-values less than 0.001, indicating that higher fibrinogen levels correlate with longer hospital stays. Analyzing the data categorically, patients in the highest quartile (Q6, > 267 mg/dL) showed a substantial increase in length of stay, with β coefficients of 2.95 (p < 0.001) in Model 1 and 3.19 (p < 0.001) in Model 2, suggesting a significant impact of elevated fibrinogen levels on hospital duration. In contrast, patients in lower quartiles (Q1 to Q5) did not reveal significant differences in length of stay compared to the reference group (Q1, ≤ 155 mg/dL), with p-values exceeding 0.05 across all models. The trend across quartiles displayed statistical significance (p for trend < 0.001), reinforcing the relationship between higher fibrinogen levels and longer hospital stays. Lastly, multiple imputation analyses addressed missing data, and logistic regression on imputed datasets confirmed the stability of the results (Table S3).

Discussion

This study is, to our knowledge, the first retrospective investigation examining the association between fibrinogen levels and all-cause mortality in female patients undergoing CABG. Our Cox regression analysis reveals that both low and high fibrinogen levels correlate with an elevated risk of long-term mortality. Further validation through RCS analysis demonstrated a U-shaped relationship between fibrinogen levels and one-year mortality, suggesting that exceeding a certain fibrinogen threshold is associated with a higher likelihood of mortality. These results underscore the prognostic significance of fibrinogen as a biomarker for adverse outcomes in this high-risk population, particularly among women, who have historically been underrepresented in cardiovascular surgical research. Our findings are consistent with previous evidence linking fibrinogen to the pathophysiology of postoperative complications while extending existing knowledge by emphasizing sex-specific associations that may inform tailored risk stratification and therapeutic strategies.

In recent years, the gender disparities in outcomes following CABG have received increasing attention. A meta-analysis of 84 observational studies involving nearly one million patients highlighted that women (n = 224,340) face a significantly higher risk of operative mortality and major adverse cardiac events compared to men26. Despite improvements in CABG outcomes, this gender gap has persisted over time27,28. A large-scale study with over one million patients, including more than 300,000 women, confirmed that operative mortality remains disproportionately higher in women, with no observed reduction in this excess risk over the past decade5. Differences in anatomical, biological, and functional characteristics, including a greater burden of comorbidities and older age, are associated with increased mortality and likely contribute to this disparity29,30. Furthermore, the pathophysiology of ischemic heart disease varies significantly between men and women31,32. Given these factors, it is essential to consider women as a distinct group and to further explore the determinants of CABG outcomes specific to this population5,6,7. Addressing these unique challenges is vital for enhancing postoperative outcomes for female CABG patients.

Fibrinogen, a crucial component of blood coagulation and an inflammatory marker, has been recognized as a key predictor of mortality, hospital length of stay, and graft failure in CABG patients20,21,22,33. Considering that CABG is among the most frequently performed cardiac surgeries, fibrinogen levels are likely to influence outcomes in female patients. However, research in this area remains limited, and our study aims to fill this gap. We are the first to investigate the relationship between serum fibrinogen levels and long-term mortality in women undergoing CABG, employing RCS functions to analyze the dose-response relationship. Our findings reveal a U-shaped association between serum fibrinogen levels and HRs for all-cause mortality, indicating that both elevated and reduced fibrinogen levels are linked to an increased risk of mortality. This highlights the importance of managing serum fibrinogen levels in the clinical care of female CABG patients.

Previous studies have explored various perioperative inflammatory markers, such as the Systemic Inflammatory Response Index (SIRI)34 and triglyceride-glucose index35, in relation to long-term survival after CABG. However, these markers can be costly, time-consuming, and difficult to obtain, often lacking sensitivity to outcomes. In contrast, our findings provide a novel perspective on the relationship between fibrinogen levels and CABG outcomes. As a readily accessible and inexpensive marker derived from routine tests, fibrinogen can be easily measured in most CABG patients. Given its significant association with outcomes, healthcare professionals can utilize fibrinogen levels to monitor coagulation and anticoagulant status in female CABG patients across various clinical settings.

While the precise mechanisms by which fibrinogen increases the risk of cardiovascular events are not fully elucidated, several potential explanations exist for the observed association between elevated fibrinogen levels and mortality risk. One possibility is that systemic inflammation stimulates the production and release of fibrinogen36. Inflammation is a key factor in the formation of atherosclerotic plaques10. Additionally, fibrinogen influences endothelial integrity and vascular permeability by binding to receptor ligands, recruiting leukocytes to the vessel wall, and inducing platelet aggregation and activation37. It also promotes smooth muscle cell proliferation and monocyte chemotaxis38. Elevated fibrinogen levels may lead to denser fibrin clots and impaired fibrinolysis, further exacerbating atherosclerosis39. As a marker of both inflammation and coagulation, fibrinogen may contribute to the increased all-cause mortality associated with coronary heart disease (CHD)40. Following CABG, multiple molecular pathways remain persistently activated, resulting in elevated inflammation, hemostasis, and oxidative stress41. Higher fibrinogen levels have been associated with worse postoperative pulmonary function decline42. While inflammation is essential for clearing infection and debris post-surgery, prolonged inflammation can result in tissue damage43.

This study benefits from national data, a large sample size, multiple outcomes, long-term follow-up, and robust statistical analyses. To our knowledge, it is the first to assess optimal early postoperative fibrinogen levels specifically for female patients undergoing CABG, thereby enhancing the understanding of managing this patient population. However, several limitations must be acknowledged. First, the retrospective, single-center design utilizing observational data from the MIMIC-IV database limits the ability to establish causality. Additionally, despite controlling for available confounders, the model may still be influenced by unmeasured factors. Third, while the MIMIC database provides ICU data, details on long-term medication use (e.g., antiplatelet agents, statins) and postoperative rehabilitation were not available, which may impact mortality risk. Finally, external validation in multi-center cohorts is necessary, particularly given the underrepresentation of racial and ethnic minorities in the MIMIC population.

Conclusions

In summary, our study reveals a U-shaped association between serum fibrinogen levels and mortality, indicating that both low and high fibrinogen levels are linked to increased overall mortality among female patients undergoing CABG. Future research should focus on assessing whether interventions aimed at lowering fibrinogen levels can enhance outcomes in high-risk female subgroups.