Mediation of postoperative length of stay by major adverse cardiovascular events in elderly patients underwent major thoracic and abdominal surgery receiving peripheral nerve blocks

Abstract

Background

The association between preoperative peripheral nerve block (PNB), major adverse cardiovascular events (MACE), and postoperative length of hospital stay (LOS) in elderly patients who underwent major thoracic and abdominal surgery remains unclear. This study aims to explore the potential mediating effect of MACE on the association between preoperative PNB and postoperative LOS using a statistical mediation framework.

Methods

In this retrospective cohort study, perioperative data were collected from elderly patients (aged over 65 years) who underwent major thoracic and abdominal surgery. Mediation analysis was employed to examine the relationships between PNB, MACE, and postoperative LOS.

Results

A total of 1915 patients were included in the analysis, with 68.7% (1316/1915) receiving preoperative PNB. Compared to patients who did not receive PNB, those who did had a significantly lower incidence of MACE (P < 0.001) and a shorter postoperative LOS (P < 0.001). The adjusted total and direct associations of PNB with postoperative LOS were − 0.809 days (95% confidence interval [CI], -1.236 to -0.390; P < 0.001) and − 0.661 days (95% CI, -1.077 to -0.250; P = 0.003), respectively. A statistically significant indirect association via MACE was observed (adjusted β=-0.149 days; 95% CI, -0.271 to -0.060; P < 0.001), indicating that 18.1% (95% CI, 6.7% to 41.0%) of the total association was statistically attributable to the indirect pathway through MACE under the model assumptions. A sensitivity analysis excluding postoperative covariates yielded consistent results (proportion mediated: 25.3%).

Conclusions

Our findings suggest that the observed association between preoperative PNB and reduced postoperative LOS in elderly patients following major thoracic and abdominal surgery may be partly explained by a statistically significant indirect pathway through a reduction in MACE, potentially accounting for approximately 18% of the total effect. These findings are hypothesis-generating and represent statistical associations rather than demonstrated causal mechanisms.

Trial registration

ChiCTR2400087610; https://www.chictr.org.cn.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12871-026-03879-9.

Introduction

Despite significant advancements in surgical techniques that have led to a substantial reduction in mortality rates following major thoracic and abdominal surgeries, the prevalence of both minor and severe postoperative complications remains a critical issue [1, 2]. These complications encompass a wide range of adverse events, including major adverse cardiovascular events (MACE) [3] and pulmonary complications [4]. Such outcomes not only compromise patient health and quality of life but also impose considerable economic burdens on healthcare systems and patients alike [5–7]. Additionally, these complications are associated with prolonged length of hospital stay (LOS) and increased mortality rates, further highlighting the need for effective preventive strategies [4, 8].

Efforts to mitigate negative postoperative outcomes are crucial, as they can significantly improve patient recovery and reduce the overall cost of care [6]. Contemporary surgical research has increasingly focused on understanding the multifaceted nature of postoperative complications and developing evidence-based preventive measures [2]. Evidence indicates that various perioperative factors are linked to extended LOS, including age, preoperative hemoglobin levels, intraoperative fluid volumes, anesthetic techniques, and the number of operative levels [9–11]. Among these, the choice of anesthetic technique, postoperative analgesia, and physical therapy seem to be the preventive measures [12].

Peripheral nerve block (PNB) has emerged as a promising component of anesthetic management, particularly in terms of improving postoperative pain control. Studies have shown that PNB can provide superior analgesia compared to traditional methods, potentially reducing the incidence of postoperative complications [4, 13]. This improved pain management may lead to faster recovery, reduced opioid use, and lower rates of pulmonary and cardiovascular complications. Consequently, there is growing interest in exploring the relationship between the use of PNB and postoperative outcomes, including LOS.

However, the potential mediating role of postoperative complications, such as MACE, in the relationship between preoperative PNB and postoperative LOS remains unclear. To address this knowledge gap, this study aims to explore the potential mediating effect of MACE in the statistical association between the use of preoperative PNB and postoperative LOS. It is hypothesized that by reducing the incidence of MACE, PNB may indirectly contribute to shorter hospital stays.

Methods

Study design and ethics

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Clinical Research Ethics Committee of Peking University First Hospital (Approval No. 2024 [224]). The Ethics Committee waived the requirement for written informed consent, as data were retrospectively collected from electronic medical records without any additional data collection or patient contact. Patient privacy was strictly protected throughout the study. The trial was a priori registered with the Chinese Clinical Trial Registry (chictr.org.cn, ChiCTR2400087610) on July 31, 2024.

Patient recruitment

Patients were identified and screened through the electronic medical records system at Peking University First Hospital (Beijing, China) from November 1, 2021, to July 31, 2023. Inclusion criteria were older patients aged 65 years or older who underwent major thoracic and abdominal non-cardiac surgery lasting at least two hours. Exclusion criteria encompassed: (1) chronic opioid dependence or use of painkillers for more than three months; (2) surgeries without routine PNB or involving epidural analgesia, organ transplantation, or secondary surgery during hospitalization; (3) pre-existing neurological conditions such as epilepsy, Parkinson’s syndrome, or myasthenia gravis; (4) severe liver insufficiency (Child-Pugh grade C), or American Society of Anesthesiologists (ASA) classification IV or higher; and (5) incomplete data.

Anesthesia and perioperative care

Patients received either general anesthesia alone or general anesthesia combined with PNB. Intraoperative monitoring included electrocardiogram, pulse oximetry, intra-arterial blood pressure, sevoflurane concentrations, expired carbon dioxide, bispectral index (BIS), nasopharyngeal temperature, and urine output. General anesthesia was induced with propofol/etomidate, sufentanil, and rocuronium, and maintained with propofol infusion and remifentanil infusion and sufentanil injection, with or without sevoflurane inhalation, and dosages were adjusted appropriately by an anesthesiologist according to clinical experience. Anesthesia depth was targeted to maintain a BIS between 40 and 60. Mechanical ventilation was provided with an oxygen-air mixture, and fluid therapy was administered per standard protocols. Intraoperative mean arterial pressure was maintained within a 30% range of the patient’s baseline values using vasoactive agents when necessary.

PNB included thoracic paravertebral block (TPVB), transversus abdominis plane block (TAPB), erector spinae plane block (ESPB), and quadratus lumborum block (QLB), and were completed by giving 0.375% ropivacaine under the guidance of a GE ultrasound system using a linear or curvilinear array probe by a designated anesthesiologist before surgery. Types of PNB and injectate volume were usually determined based on the specific surgical procedure, patient’s individual situation and intention, and visualization of injectate spread by ultrasound, e.g., unilateral TPVB (20 mL 0.375% ropivacaine) for patients undergoing pulmonary lobectomy and bilateral TAPB (40 mL 0.375% ropivacaine) for those undergoing gastrointestinal surgery.

Post-surgery, patients were transferred to the post-anesthesia care unit or intensive care unit (ICU) based on their clinical and hemodynamic status. Patient-controlled intravenous analgesia was provided with morphine or sufentanil, and supplemental analgesics such as NSAIDs, oxycodone, or other analgesics were administered as needed.

Data collection and outcomes

Data were retrospectively collected from the electronic medical records system at the study center. Demographic data included age, sex, and body mass index (BMI). Preoperative data encompassed surgical diagnosis, comorbidities (Charlson Comorbidity Index [14] and Revised Cardiac Risk Index [15]), laboratory test results, smoking and drinking history, and ASA classification. Intraoperative data included the type and site of PNB, type and duration of surgery, type and duration of anesthesia, and fluid input and output. Postoperative data included ICU admission, surgery-related complications, the incidence of MACE, and postoperative LOS. MACE comprised cardiogenic and cerebrovascular death, acute myocardial infarction, congestive heart failure, new-onset atrial fibrillation, and stroke [7, 16]. All MACE events were defined as those occurring during the postoperative in-hospital period, from the day of surgery until discharge, and were identified from discharge diagnoses recorded in the electronic medical record system.

Statistical analysis

Patients were categorized into two groups based on whether they received a PNB before surgery. Continuous variables were compared using the Student’s t-test or Mann-Whitney U test, while categorical variables were analyzed using the chi-squared or Fisher’s exact tests. Missing data were not imputed.

Perioperative variables were initially assessed using univariate analysis to identify potential associations with postoperative LOS. Variables with P < 0.20 in univariate analyses, or those deemed clinically significant, were included in a multivariable linear regression model. PNB and MACE were always included in the model. Additionally, perioperative risk factors for MACE, including the relationship between PNB and MACE, were analyzed using logistic regression.

The mediation analysis was conducted using a counterfactual potential-outcomes framework, implemented via the “mediation” R package [17]. The primary objective was to examine how the total statistical association of PNB with postoperative LOS (path c) could be partitioned into an indirect effect (path a and b, operating through MACE) and a direct effect (path c′; Fig. 1). The indirect effect was defined as the change in expected postoperative LOS when PNB was fixed, but MACE varied as if PNB status were altered. The direct effect was defined as the change in expected postoperative LOS when PNB varied, but MACE was held constant. MACE was modeled using logistic regression with a logit link, with PNB as an independent variable. Postoperative LOS was modeled using a generalized linear model (GLM) with a Gamma distribution and log link, with both PNB and MACE as independent variables; this distributional choice is appropriate for LOS as a positive, right-skewed continuous variable, and model adequacy was confirmed by deviance residual diagnostics (Supplemental Figure S1). The total, direct, and indirect associations were quantified using this package with 5,000 bootstrap iterations. As the “mediate” function estimates effects as differences in model-predicted values on the original response scale, all reported mediation estimates are expressed in days.

Fig. 1.

Mediation Analysis of preoperative Peripheral Nerve Block (PNB) on postoperative Length of Hospital Stay (LOS). The total effect of preoperative PNB on postoperative LOS is denoted by path c. The direct effect of preoperative PNB on postoperative LOS, after controlling for the incidence of postoperative Major Adverse Cardiac Events (MACE), is represented by path c′. The effect of preoperative PNB on the incidence of postoperative MACE is indicated by path a The effect of postoperative MACE on postoperative LOS, after controlling for preoperative PNB, is denoted by path b The indirect effect of preoperative PNB on postoperative LOS, mediated through postoperative MACE, is calculated as the product of paths a and b (a×b)

Post-adjustment for confounders was performed to refine the coefficients for total, direct, and indirect effects. Confounders were selected based on: (1) independent predictors of postoperative LOS identified in the multivariable linear regression model, (2) imbalanced variables between patients with and without PNB, and (3) clinically significant factors such as age. Collinear variables were addressed by selecting one representative variable. Accordingly, two sets of confounder models were specified. The primary adjusted model (Adjusted ^a) included age, sex, BMI, CCI, haemoglobin, alanine transaminase, albumin, glucose, high-sensitivity cardiac troponin I, brain natriuretic peptide, site of surgery, type of surgery, duration of surgery, total fluid input, ICU admission, reoperation, sepsis, and unplanned ICU transfer. The four postoperative variables (ICU admission, reoperation, sepsis, and unplanned ICU transfer) were retained in the primary model because they are among the strongest independent predictors of LOS in our multivariable analysis; however, we acknowledge that they may also represent downstream consequences of PNB or MACE, and thus act as potential colliders on the causal pathway to LOS. To address this concern, a pre-specified sensitivity analysis (Adjusted ^b) was conducted using the same model but excluding these four variables, retaining only pre- and intraoperative covariates (Supplemental Figure S2).

To verify the robustness of our primary findings and minimize potential selection bias, we conducted a sensitivity analysis using propensity score matching. We performed a propensity score matching in which patients with PNB were matched with those without PNB in a 1:1 ratio based on the following variables: age, sex, BMI, coronary heart disease, chronic kidney disease, previous stroke, Revised Cardiac Risk Index, CCI, ASA, site of surgery, type of surgery, and duration of surgery.

To examine whether the mediating role of MACE differed according to surgical approach, pre-specified subgroup mediation analyses were conducted stratified by type of surgery (minimally invasive surgery vs. open surgery). Within each subgroup, the same primary adjusted and sensitivity confounder models were applied as described above.

Statistical significance was set at P < 0.05 for two-sided tests. All analyses were conducted using R 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria, 2019).

Power calculation

Given the fixed sample size of 1,915 patients, a post hoc power analysis for the mediation effect was performed using the R package power mediation [18]. Under the observed effect parameters (mediator SD = 0.216; regression coefficient for the mediator = 4.909; residual SD = 7.253; predictor–mediator correlation = -0.144; α = 0.05), the study achieved a statistical power of 0.99. Although post hoc power analysis has limited utility for retrospective studies, these results indicate that the sample was adequately powered for the primary mediation analysis.

Results

Patients

From November 1, 2021, to July 31, 2023, a total of 46,682 patients were screened. Of these, 8,278 were excluded due to undergoing non-general anesthesia surgeries or having incomplete anesthesia information. Among the remaining 38,404 patients who underwent surgeries with general anesthesia, 21,401 were excluded because they had emergency surgeries or non-thoracoabdominal procedures. Of the 17,003 patients left, 354 were excluded for undergoing cardiac surgery, 10,317 were excluded for being under 65 years of age, and 4,263 were excluded due to a surgery duration of less than two hours. Subsequently, 2069 patients were further excluded for the following reasons: 35 for not receiving routine PNB, 11 for receiving combined epidural analgesia, 2 for kidney transplantation, 6 for secondary surgery during hospitalization, 4 for preoperative epilepsy, 22 for preoperative Parkinson’s syndrome, 9 for preoperative myasthenia gravis, 59 for ASA class IV, and 6 for incomplete data. Ultimately, 1915 patients were included in the study, with 1316 (68.7%) receiving PNB and 599 (31.3%) not receiving PNB (Fig. 2).

Fig. 2.

Flow diagram of the study

Patients receiving PNB were found to have a lower proportion of men (P = 0.010), a lower CCI score (P = 0.035), a reduced level of high-sensitivity cardiac troponin I (P = 0.011), and a lower level of brain natriuretic peptide (P = 0.034) compared to those not receiving PNB. Significant differences between the two groups were also observed in terms of site (P = 0.045) and type (P = 0.036) of surgery. Additionally, patients receiving PNB had less colloid fluid input (P = 0.003), more urine output (P = 0.020), and greater total fluid input (P = 0.045). They also exhibited a lower rate of ICU admission (P = 0.002), a reduced incidence of MACE (P < 0.001), and a shorter postoperative LOS (P < 0.001) (Table 1).

Table 1.

Baseline and perioperative data

	Without PNB (n = 599)	With PNB (n = 1316)	P value
Baseline data
Age (years)	72 ± 6	71 ± 5	0.168
Male sex	417 (69.6%)	835 (63.5%)	0.010
BMI (kg·m− 2)	24.1 ± 3.3	24.3 ± 3.2	0.431
Preoperative comorbidities
Hypertension	319 (53.3%)	696 (52.9%)	0.920
Diabetes mellitus	143 (23.9%)	306 (23.3%)	0.811
Hyperlipemia	10 (1.7%)	28 (2.1%)	0.624
Coronary heart disease	91 (15.2%)	189 (14.4%)	0.684
PCI	37 (6.2%)	67 (5.1%)	0.388
CABG	4 (0.7%)	4 (0.3%)	0.446
Arrhythmia a	37 (6.2%)	94 (7.1%)	0.497
Chronic kidney diseases	15 (2.5%)	22 (1.7%)	0.295
Dialysis	1 (0.2%)	2 (0.2%)	> 0.999
Previous stroke	75 (12.6%)	129 (9.8%)	0.088
Hemiplegia	6 (1.0%)	8 (0.6%)	0.517
COPD	50 (8.4%)	78 (5.9%)	0.062
Asthma	4 (0.7%)	20 (1.5%)	0.183
Hyperthyroidism	1 (0.2%)	4 (0.3%)	0.951
Hypothyroidism	3 (0.5%)	10 (0.8%)	0.734
Chronic smoking	35 (5.8%)	83 (6.3%)	0.773
Regular drinking	4 (0.7%)	6 (0.5%)	0.799
CCI (point)	2.5 ± 0.8	2.5 ± 0.7	0.035
Revised CRI (point)	1.5 ± 0.7	1.5 ± 0.7	0.217
ASA classification			0.083
I	3 (0.5%)	3 (0.2%)
II	287 (47.9%)	697 (53.0%)
III	309 (51.6%)	616 (46.8%)
Laboratory tests
Hemoglobin (g/L)	131 ± 20	131 ± 19	0.783
Platelet (⋅109/L)	219 ± 72	220 ± 71	0.763
AST (IU/L)	22 ± 37	20 ± 25	0.177
ALT (IU/L)	25 ± 27	24 ± 19	0.160
Albumin (g/L)	41 ± 6	42 ± 6	0.193
Creatinine (mmol/L)	89 ± 47	87 ± 50	0.412
Glucose (mmol/L)	6.2 ± 1.7	6.1 ± 1.7	0.746
hs-cTnI (ng/ml)	6.0 ± 9.5	5.2 ± 5.6	0.011
BNP (pg/mL)	73 ± 75	65 ± 66	0.034
PT (s)	11.3 ± 0.9	11.3 ± 1.3	0.582
APTT (s)	30.6 ± 3.3	30.4 ± 3.3	0.118
Intraoperative data
Site of surgery			0.045
Esophageal	8 (1.3%)	24 (1.8%)
Pulmonary	131 (21.9%)	375 (28.5%)
Gastrointestinal	250 (41.7%)	503 (38.2%)
Hepatopancreatobiliary	23 (3.8%)	39 (3.0%)
Renal/nephroureteral	92 (15.4%)	196 (14.9%)
Prostatic	91 (15.2%)	176 (13.4%)
Total pelvic	4 (0.7%)	3 (0.2%)
Type of surgery			0.036
Thoraco-/laparoscopic	511 (85.3%)	1169 (88.8%)
Open thoracic/abdominal	88 (14.7%)	147 (11.2%)
Duration of surgery (min)	188 (149, 242)	182 (150, 234)	0.274
Duration of anesthesia (min)	273 (229, 331)	266 (225, 322)	0.063
Fluid balance
Crystalloid (mL)	1600 (1300, 2100)	1600 (1350, 2100)	0.513
Colloid (mL)	500 (0, 500)	500 (0, 500)	0.003
Estimated blood loss (mL)	50 (30, 100)	50 (20, 100)	0.055
Blood transfusion b	70 (11.7%)	120 (9.1%)	0.097
Urine output (mL)	300 (100, 625)	400 (150, 700)	0.020
Total fluid output (mL)	450 (200, 810)	500 (200, 820)	0.373
Total fluid input (mL)	2100 (1600, 2600)	2100 (1600, 2600)	0.045
Postoperative data
ICU admission	181 (30.2%)	308 (23.4%)	0.002
Reoperation c	6 (1.0%)	16 (1.2%)	0.860
Lower limbs venous thrombosis d	4 (0.7%)	4 (0.3%)	0.446
Pulmonary embolism e	0 (0%)	3 (0.2%)	0.585
Sepsis f	0 (0%)	3 (0.2%)	0.585
Unplanned ICU transfers	9 (1.5%)	10 (0.8%)	0.204
MACE	57 (9.5%)	37 (2.8%)	< 0.001
Cardiovascular death g	1 (0.2%)	2 (0.2%)	> 0.999
Myocardial infarction h	2 (0.3%)	2 (0.2%)	0.788
Congestive heart failure i	13 (2.2%)	10 (0.8%)	0.009
New-onset atrial fibrillation j	32 (5.3%)	24 (1.8%)	< 0.001
Stroke k	9 (1.5%)	3 (0.2%)	0.003
Length of hospital stay (days)	7 (5, 11)	7 (5, 9)	< 0.001

Data are mean ± SD, n (%), or median (interquartile range). P values in bold indicate < 0.05

PNB peripheral nerve block, BMI body mass index, PCI percutaneous coronary intervention, CABG coronary artery bypass grafting, COPD chronic obstructive pulmonary disease, CCI Charlson comorbidity Index, Revised CRI revised cardiac risk Index, ASA American society of anesthesiologists, AST aspartate transaminase, ALT alanine transaminase, hs-cTnI high-sensitivity cardiac troponin I, BNP brain natriuretic peptide, PT prothrombin time, APTT activated partial thromboplastin time, ICU intensive care unit, MACE major cardiovascular adverse events

^a Include atrial fibrillation, frequent (> 6 beats/min) or multifocal ventricular premature beat, paroxysmal supraventricular tachycardia, second/third-degree atrioventricular block, and sick sinus syndrome

^b Include allogeneic blood and autologous blood transfusion

^c Reoperation was performed due to postoperative anastomotic fistula, poor wound healing, or bleeding

^d Confirmed by lower limbs venous ultrasonography

^e A new blood clot or thrombus within the pulmonary arterial system and confirmed by scintigraphy and CT angiography

^f Two or more criteria of systemic inflammatory response syndrome, with known infection and new-onset dysfunction of at least one organ/system

^g A death caused by cardio/cerebrovascular factors

^h Confirmed by ischemic symptoms, electrocardiogram, and blood tests of myocardial markers

ⁱ Diagnosed by the presence of new-onset orthopnea with evidence of fluid retention (i.e., elevated jugular venous pressure, evidence of pulmonary edema, and/or peripheral edema) and an elevated plasma brain natriuretic peptide of above 400 pg/mL

^j Confirmed by 12-lead electrocardiogram and necessitated medical treatment and/or cardioversion

^k Persisted new focal neurologic deficit and confirmed by neurologic imaging

Risk factors of postoperative LOS and MACE

Univariate analysis was conducted on all baseline and perioperative variables to identify those associated with postoperative LOS and MACE, while accounting for variable collinearity. A total of 31 variables with P < 0.20 or those deemed clinically significant were included in a multivariable linear regression model (Table 2 and Supplemental Table S1). After adjusting for confounding factors, both PNB (β=-0.78; 95% CI, -1.291 to -0.269; P = 0.003) and MACE (β = 2.125; 95% CI, 0.983 to 3.267; P < 0.001) were independently associated with the postoperative LOS (Table 2). Additionally, logistic regression analysis revealed that PNB (OR = 0.256; 95% CI, 0.159 to 0.407; P < 0.001) was independently associated with MACE (Supplemental table S2).

Table 2.

Associations between the baseline/perioperative variables and postoperative LOS

Variables	Univariate analysis a		Multivariable analysis
	β (95% CI)	P value	β (95% CI)	P value
PNB	-1.185 (-1.893, -0.478)	0.001	-0.780 (-1.291, -0.269)	0.003
Age (years)	0.084 (0.023, 0.146)	0.007	-0.042 (-0.091, 0.007)	0.096
Male sex	0.607 (-0.085, 1.298)	0.086	0.276 (-0.289, 0.841)	0.339
BMI (kg·m-2)	-0.334 (-0.434, -0.233)	< 0.001	-0.197 (-0.273, -0.120)	< 0.001
Hypertension	-0.453 (-1.112, 0.206)	0.178	-0.052 (-0.552, 0.447)	0.838
CABG	3.674 (-1.425, 8.774)	0.158	0.992 (-2.751, 4.735)	0.603
Arrhythmia b	1.381 (0.079, 2.683)	0.038	0.552 (-0.418, 1.522)	0.265
COPD	1.476 (0.160, 2.792)	0.028	-0.229 (-1.298, 0.840)	0.675
Chronic smoking	0.923 (-0.445, 2.291)	0.186	0.740 (-0.275, 1.756)	0.153
Regular drinking	3.025 (-1.539, 7.589)	0.194	2.360 (-0.992, 5.711)	0.168
CCI (point)	0.330 (-0.108, 0.767)	0.140	-0.026 (-0.519, 0.466)	0.917
Revised CRI (point)	0.324 (-0.155, 0.802)	0.185	-0.296 (-0.837, 0.246)	0.285
ASA classification
I	Ref.		Ref.
II	2.559 (-3.321, 8.439)	0.394	0.523 (-3.688, 4.734)	0.808
III	3.677 (-2.204, 9.558)	0.220	0.475 (-3.763, 4.713)	0.826
Hemoglobin (g/L)	-0.064 (-0.080, -0.047)	< 0.001	-0.018 (-0.033, -0.002)	0.026
Platelet (⋅109/L)	0.005 (0.000, 0.009)	0.044	0.001 (-0.002, 0.005)	0.449
ALT (IU/L)	0.067 (0.053, 0.082)	< 0.001	0.032 (0.021, 0.043)	< 0.001
Albumin (g/L)	-0.124 (-0.182, -0.067)	< 0.001	0.063 (0.017, 0.109)	0.008
Glucose (mmol/L)	0.163 (-0.027, 0.353)	0.093	0.157 (0.011, 0.303)	0.036
BNP (pg/mL)	0.012 (0.007, 0.017)	< 0.001	0.003 (-0.001, 0.007)	0.145
PT (s)	0.686 (0.408, 0.964)	< 0.001	0.157 (-0.048, 0.362)	0.133
Site of surgery
Esophageal	Ref.		Ref.
Pulmonary	-10.979 (-13.339, -8.619)	< 0.001	-4.586 (-6.576, -2.595)	< 0.001
Gastrointestinal	-5.306 (-7.643, -2.969)	< 0.001	-1.097 (-3.052, 0.859)	0.272
Hepatopancreatobiliary	-4.352 (-7.170, -1.534)	0.003	-0.685 (-2.992, 1.622)	0.561
Renal/nephroureteral	-11.361 (-13.774, -8.948)	< 0.001	-4.79 (-6.866, -2.714)	< 0.001
Prostatic	-11.914 (-14.336, -9.492)	< 0.001	-4.503 (-6.627, -2.378)	< 0.001
Total pelvic	-1.237 (-6.639, 4.166)	0.654	-0.087 (-4.44, 4.267)	0.969
Type of surgery (Thoraco-/laparoscopic vs. Open)	-4.594 (-5.576, -3.612)	< 0.001	-1.293 (-2.079, -0.506)	0.001
Duration of surgery (min)	0.037 (0.034, 0.041)	< 0.001	0.024 (0.020, 0.028)	< 0.001
Total fluid output (mL)	0.003 (0.002, 0.003)	< 0.001	0.000 (0.000, 0.001)	0.848
ICU admission	2.441 (1.695, 3.188)	< 0.001	0.112 (-0.53, 0.754)	0.732
Reoperation c	31.820 (29.081, 34.559)	< 0.001	27.782 (25.527, 30.036)	< 0.001
Lower limbs venous thrombosis d	5.181 (0.084, 10.277)	0.046	-1.271 (-4.905, 2.363)	0.493
Pulmonary embolism e	-1.927 (-10.248, 6.393)	0.650	1.898 (-4.038, 7.834)	0.531
Sepsis f	30.457 (22.249, 38.665)	< 0.001	24.355 (17.87, 30.84)	< 0.001
Unplanned ICU transfers	18.274 (15.057, 21.492)	< 0.001	4.157 (1.446, 6.867)	0.003
MACE g	5.173 (3.668, 6.679)	< 0.001	2.125 (0.983, 3.267)	< 0.001

P values in bold indicate < 0.05

CI confidence interval, PNB peripheral nerve block, BMI body mass index, CABG coronary artery bypass grafting, COPD chronic obstructive pulmonary disease, CCI Charlson comorbidity index, Revised CRI revised cardiac risk index, ASA American society of anaesthesiologists, ALT alanine transaminase, BNP brain natriuretic peptide, PT prothrombin time, ICU intensive care unit, MACE major cardiovascular adverse events

^a Factors with P < 0.20 in univariable analyses or were considered clinically important were included in the multivariable generalized linear model

^b Include atrial fibrillation, frequent (> 6 beats/min) or multifocal ventricular premature beat, paroxysmal supraventricular tachycardia, second/third-degree atrioventricular block, and sick sinus syndrome

^c Reoperation was performed due to postoperative anastomotic fistula, poor wound healing, or bleeding

^d Confirmed by lower limbs venous ultrasonography

^e A new blood clot or thrombus within the pulmonary arterial system and confirmed by scintigraphy and CT angiography

^f Two or more criteria of systemic inflammatory response syndrome, with known infection and new-onset dysfunction of at least one organ/system

^g Include cardiogenic and cerebrovascular death, acute myocardial infarction, congestive heart failure, new-onset atrial fibrillation, and stroke

Mediation analysis

Following adjustment for potential confounders including age, sex, BMI, CCI, hemoglobin, albumin, alanine transaminase, high-sensitivity cardiac troponin I, brain natriuretic peptide, glucose, site of surgery, type of surgery, duration of surgery, total fluid input, ICU admission, reoperation, sepsis, and unplanned ICU transfer, the total and direct associations of PNB with postoperative LOS were quantified. In the primary adjusted analysis, the total association was β=-0.809 days (95% CI, -1.236 to -0.390; P < 0.001) and the direct association was β=-0.661 days (95% CI, -1.077 to -0.250; P = 0.003) (Table 3).

Table 3.

Total, direct, and indirect association of PNB with postoperative LOS mediated via MACE

Model	Total association		Direct association		Indirect association via MACE		Proportion mediated
	β (95% CI)	P value	β (95% CI)	P value	β (95% CI)	P value	% (95% CI)
Unadjusted	-1.225 (-1.987, -0.510)	0.002	-0.891 (-1.637, -0.200)	0.014	-0.333 (-0.580, -0.150)	< 0.001	27.0 (11.5, 65.0)
Adjusted a	-0.809 (-1.236, -0.390)	< 0.001	-0.661 (-1.077, -0.250)	0.003	-0.149 (-0.271, -0.060)	< 0.001	18.1 (6.7, 41.0)
Adjusted b	-0.791 (-1.306, -0.300)	< 0.001	-0.590 (-1.087, -0.110)	0.016	-0.201 (-0.359, -0.080)	< 0.001	25.3 (10.2, 68.0)

CI confidence interval. P values in bold indicate < 0.05

^a Adjusted for age, sex, BMI, CCI, haemoglobin, alanine transaminase, albumin, glucose, high-sensitivity cardiac troponin I, brain natriuretic peptide, site of surgery (pulmonary/gastrointestinal/hepatopancreatobiliary/renal/nephroureteral/prostatic/total pelvic vs. esophageal), type of surgery (thoraco-/laparoscopic vs. open), duration of surgery, total fluid input, ICU admission, reoperation, sepsis, and unplanned ICU transfer

^b Adjusted for age, sex, BMI, CCI, haemoglobin, alanine transaminase, albumin, glucose, high-sensitivity cardiac troponin I, brain natriuretic peptide, site of surgery (pulmonary/gastrointestinal/hepatopancreatobiliary/renal/nephroureteral/prostatic/total pelvic vs. esophageal), type of surgery (thoraco-/laparoscopic vs. open), duration of surgery, total fluid input

A statistically significant indirect association via MACE was observed, with an adjusted β of -0.149 days (95% CI, -0.271 to -0.060; P < 0.001), indicating that 18.1% (95% CI, 6.7% to 41.0%) of the total association between PNB and postoperative LOS was statistically attributable to the indirect pathway through MACE under the model assumptions. In the sensitivity analysis excluding the four postoperative covariates, the indirect association via MACE remained statistically significant (β=-0.201 days; 95% CI, -0.359 to -0.080; P < 0.001; proportion mediated: 25.3%, 95% CI, 10.2% to 68.0%), supporting the robustness of the primary findings (Table 3).

To verify the robustness of our findings and mitigate potential selection bias, a sensitivity analysis was conducted using a propensity score-matched cohort (1:1 ratio). The baseline and intraoperative characteristics were balanced between the PNB and non-PNB groups (Supplemental Table S3). In this matched cohort, the total association between PNB and reduced postoperative LOS remained statistically significant (adjusted β = -0.978 days; 95% CI, -1.440 to -0.520; P < 0.001), as did the direct association (adjusted β = -0.867 days; 95% CI, -1.329 to -0.410; P < 0.001). Consistent with the primary analysis, a significant indirect association via MACE was also observed in the matched cohort. The indirect association via MACE was statistically significant (adjusted β = -0.111 days; 95% CI, -0.239 to -0.020; P = 0.012), accounting for 11.0% (95% CI, 1.7% to 28.0%) of the total effect (Supplemental Table S4).

Subgroup mediation analyses stratified by surgical approach are presented in Supplemental Table S5. In the minimally invasive surgery subgroup (N = 1,680), the indirect association via MACE remained statistically significant in both the primary adjusted model (Adjusted ^a: β=-0.157 days; 95% CI, -0.294 to -0.060; P < 0.001; proportion mediated: 19.6%) and the sensitivity analysis excluding postoperative covariates (Adjusted ^b: β=-0.200 days; 95% CI, -0.355 to -0.080; P < 0.001; proportion mediated: 23.8%), consistent with the overall cohort findings. In the open surgery subgroup (N = 235), neither the total association nor the indirect association via MACE reached statistical significance in the adjusted models, likely attributable to the substantially smaller sample size and correspondingly reduced statistical power.

Discussion

Our study found evidence suggesting a statistically significant indirect association between preoperative PNB and reduced postoperative LOS that appeared to operate through a reduction in MACE in older patients undergoing major thoracic and abdominal surgery. Given the retrospective observational design and inherent limitations of mediation analysis with observational data, these findings are consistent with a partial mediating role of MACE but should not be interpreted as establishing a causal pathway.

Postoperative complications have been consistently linked with prolonged LOS [4, 8]. MACE, although relatively infrequent, is a severe complication following non-cardiac surgery and represents a leading cause of significant perioperative morbidity and mortality, with reported incidence rates varying between 1% and 7% [7]. Given their severity, MACE warrants close attention in perioperative management. Notably, angina was not included in our definition of MACE due to the retrospective nature of the study, whereas stroke was included, aligning with the majority of clinical research [19–22].

PNB is increasingly recommended for improving postoperative recovery and outcomes [13, 23]. Evidence indicates that PNB can reduce perioperative cardiopulmonary complications [4, 13] and contribute to a shorter hospital stay [4]. A meta-analysis involving 122 studies found that PNB was associated with a lower incidence of respiratory failure and cardiac complications following total hip or knee arthroplasty [13]. Similarly, a retrospective cohort study revealed that thoracic paravertebral blockade combined with general anesthesia was linked to reduced rates of postoperative atrial fibrillation and a shorter LOS compared to general anesthesia alone [4]. However, there is limited research on the mediating role of postoperative complications in the relationship between PNB and postoperative LOS.

Our findings corroborate previous studies, showing a significant difference in both MACE incidence and postoperative LOS between patients who received PNB and those who did not, among older individuals undergoing major thoracic and abdominal surgery. Notably, the indirect association via MACE was statistically significant, with MACE statistically explaining 18.1% of the total association between PNB and postoperative LOS under the primary adjusted model assumptions. Meta-analyses and large observational studies have reported associations between regional anesthesia and analgesia and reduced perioperative cardiovascular complications across various surgical populations [4, 13, 24]. In a meta-analysis of 122 studies, PNB use was associated with lower risks of cardiac complications after total hip and knee arthroplasties [13]. In a retrospective cohort study involving 13,966 patients, paravertebral blockade combined with general anesthesia was associated with lower rates of new-onset atrial fibrillation after thoracoscopic lung surgery [4]. Our results were generally consistent with the above studies, i.e., PNB use was associated with fewer MACE.

The clinical magnitude of the observed associations is modest and warrants cautious interpretation. The total adjusted association corresponded to a reduction of less than one day in LOS (β = -0.809 days), and the indirect component via MACE accounted for only 18.1% of this total association (β = -0.149 days). These effect sizes, though statistically robust in this large sample, are of uncertain clinical significance. A reduction of approximately 0.15 days (roughly 3.6 h) attributable to the MACE pathway specifically is unlikely to be clinically meaningful in isolation. Furthermore, the majority (81.9%) of the total statistical association operated through pathways other than MACE, including unmeasured factors such as analgesia quality, opioid consumption, and early mobilization. Accordingly, our findings should not be interpreted as establishing MACE reduction as a clinically dominant mechanism by which PNB influences hospital stay.

Poorly managed postoperative pain can lead to adverse outcomes, such as myocardial ischemia, atrial fibrillation, tachycardia, and stroke [25, 26]. These complications, including MACE, are associated with prolonged LOS [3, 4, 8]. Therefore, the observed indirect association may partly reflect the improved analgesia provided by PNB, although this mechanistic interpretation remains speculative in the absence of direct pain score or opioid consumption data. It is also important to note that MACE statistically explained only 18.1% (primary analysis) to 25.3% (sensitivity analysis excluding postoperative covariates) of the total association between PNB and postoperative LOS. The consistency of results across both adjusted models supports the robustness of this finding and suggests that the indirect pathway via MACE is present regardless of whether postoperative covariates — which may act as potential colliders — are included in the model. This indicates that the majority of the association operates through other pathways that were not directly measurable in the current dataset, including superior analgesia, reduced opioid consumption, and enhanced early mobilization.

Our study has several limitations. Firstly, the retrospective observational design introduces potential selection bias. Secondly, systematic postoperative troponin surveillance was not performed as a standard institutional protocol during the study period; troponin measurements were obtained clinically only when a cardiac event was suspected. This may have led to underdiagnosis of myocardial injury after non-cardiac surgery (MINS). However, MINS was not included in the definition of MACE in this study. Thirdly, as a retrospective observational study, the mediation analysis relies on the assumption of sequential ignorability — that is, no unmeasured confounding between the exposure and mediator, the mediator and outcome, or the exposure and outcome. This assumption is unlikely to be fully satisfied in the present dataset, as the absence of key perioperative variables including intraoperative hemodynamic instability, vasopressor use, opioid consumption, postoperative pain scores, quality-of-life assessments using validated instruments (e.g., EQ-5D, SF-36, QoR-15), and adherence to enhanced recovery protocols, which are central to the proposed mechanism. Accordingly, all mediation estimates should be interpreted as statistical associations under model assumptions, and not as evidence of a causal biological mechanism. Therefore, we explicitly caution against interpreting our indirect effect as a causal mechanism. Future prospective studies should incorporate standardized, time-point-specific pain assessments and validated patient-reported outcome measures to more comprehensively evaluate the clinical benefits of PNB. Finally, although the subgroup mediation analysis in the open surgery group yielded point estimates consistent in direction with the overall cohort, statistical significance was not reached, likely due to the limited sample size (N = 235) in this subgroup.

Conclusion

In summary, our findings suggest that the association between preoperative PNB and shorter postoperative LOS in elderly patients following major thoracic and abdominal surgery may be partly explained by a statistically significant indirect pathway through a reduction in MACE, potentially accounting for approximately 18% to 25% of the total association. It is essential to emphasize that these hypothesis-generating findings are subject to the limitations of observational study and represent statistical associations rather than demonstrated causal mechanisms. Accordingly, all mediation estimates should be interpreted as statistical associations under model assumptions of no unmeasured confounding (sequential ignorability), and not as evidence of a causal biological mechanism. Nonetheless, they suggest that both PNB and MACE merit attention as components of perioperative management strategies aimed at minimizing postoperative LOS, pending confirmation in prospective studies.

Authors’ contributions

Zhen Zhang, Chen Li: data collection and manuscript draft. Zhen-Zhen Xu, Jia-Hui Ma: conception and design, data analysis, manuscript revision, and final approval.

Funding

The study was funded by the National Science and Technology Major Project of the Ministry of Science and Technology of China (Grant No. 2023ZD0512900), the Capital’s Funds for Health Improvement and Research (Grant No. 2026-4G-40717), the National Natural Science Foundation of China (Grant No. 82401497) and the Beijing High-Level Innovation and Entrepreneurship Talent Support Program-Healthcare Platform young backbone talent projects (Grant No. G202532203). The funders had no role in study design, data acquisition, analysis, interpretation of results, or in the writing and submission of the report.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Clinical Research Ethics Committee of Peking University First Hospital (Approval No. 2024 [224]). The Clinical Research Ethics Committee of Peking University First Hospital waived the requirement for written informed consent, as data were retrospectively collected from electronic medical records without any additional data collection or patient contact. Patient privacy was strictly protected throughout the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.