The effect of depression comorbid with alcohol use disorder on postoperative morbidity of patients undergoing resection of upper gastrointestinal and pancreatic tumors: a study based on a national inpatient sample

Abstract

Background

Patients undergoing resection of upper gastrointestinal and pancreatic tumors are at considerable risk for postoperative morbidity. The co-occurrence of alcohol use disorder (AUD) and depression may synergistically impair recovery. This study aimed to investigate the effect of depression comorbid with alcohol use disorder (Dep-AUD) on postoperative morbidity of patients undergoing resection of upper gastrointestinal and pancreatic tumors.

Methods

Data from the National Inpatient Sample (2010–2019) were analyzed. Patients undergoing resection of upper gastrointestinal and pancreatic tumors were identified using ICD codes, and divided into Dep-AUD, Non-Dep-AUD, depression, and AUD groups. Multivariable logistic regression was used to investigate the effect of Dep-AUD, Non-Dep-AUD, depression, and AUD on postoperative morbidity.

Results

Among 60,783 patients, 637 (1.05%) had Dep-AUD. After covariate adjustment, Dep-AUD was independently associated with increased risks of delirium (aOR 2.27), wound dehiscence (aOR 2.16), and invasive respiratory support (aOR 1.47), alongside the longest median stays and highest costs. A significant supra-additive interaction was observed for wound dehiscence, particularly in esophagectomy and gastrectomy. For other complications like delirium, the elevated risk primarily reflected the additive accumulation of independent effects rather than true biological synergy.

Conclusion

Dep-AUD significantly exacerbates postoperative complications — notably wound dehiscence, delirium, and respiratory failure — and increases resource utilization after upper gastrointestinal and pancreatic resections. These procedure-specific findings emphasize the urgent need for systematic preoperative screening and tailored multidisciplinary optimization strategies to mitigate surgical risks in this vulnerable population.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12893-026-03995-w.

Introduction

Upper gastrointestinal (UGI) and pancreatic cancers pose a significant global health burden, characterized by high mortality rates and complex management strategies [1, 2]. These malignancies, encompassing cancers of the esophagus, stomach, duodenum, and pancreas, are often diagnosed at advanced stages, necessitating surgical resection as a primary treatment modality [3]. While advancements in surgical techniques and perioperative care have improved outcomes, postoperative complications remain a significant challenge, impacting patient morbidity, mortality, and healthcare costs [4]. The etiology of UGI and pancreatic cancers is multifactorial, involving complex interactions between genetic predisposition, environmental factors, and lifestyle choices [5]. Established risk factors include chronic Helicobacter pylori infection (gastric cancer) [6], tobacco smoking (esophageal, gastric, and pancreatic cancer) [7, 8], dietary patterns high in processed foods and low in fruits and vegetables (gastric and esophageal cancer) [9], obesity (esophageal and pancreatic cancer) [10], and a family history of cancer [11].

Alcohol use disorder (AUD), particularly chronic heavy use, is also a well-recognized and modifiable risk factor for several UGI and pancreatic cancers, including esophageal squamous cell carcinoma, gastric cancer, and pancreatic adenocarcinoma [12, 13]. Alcohol’s carcinogenic effects are attributed to several mechanisms, including the production of acetaldehyde (a toxic metabolite), increased oxidative stress, impaired DNA repair, and altered folate metabolism [14]. Furthermore, alcohol consumption can exacerbate existing risk factors, such as H. pylori infection, by promoting gastric inflammation and increasing the risk of dysplasia [15].

Beyond its direct carcinogenic effects, alcohol use often co-occurs with other adverse lifestyle factors and mental health conditions, such as depression, creating a complex clinical picture that can significantly impact surgical outcomes. Depression, a prevalent and often underdiagnosed condition in cancer patients, is associated with impaired immune function, increased inflammation, and poorer adherence to treatment regimens [16, 17]. The interplay between depression comorbid with alcohol use disorder (Dep-AUD) can be particularly detrimental in the context of major surgery. Alcohol-induced hepatic dysfunction, immune suppression, and impaired wound healing, combined with depression-related neuroendocrine dysregulation and reduced self-care, can synergistically increase the risk of postoperative complications, such as infections, anastomotic leaks, delirium, and prolonged hospital stay [18].

Despite the significant impact of Dep-AUD on surgical outcomes, there is a lack of comprehensive studies examining their combined effects on postoperative complications in patients undergoing upper gastrointestinal and pancreatic tumor resection surgery [19]. This study aims to explore the impact of comorbid depression with alcohol use disorder on patients undergoing upper gastrointestinal and pancreatic tumor resection surgery using the National Inpatient Sample (NIS) data. The study analyzed the patient’s morbidity, demographic characteristics, medical insurance type, Charlson Comorbidity Index (CCI), length of stay (LOS), total medical expenses, comorbidities, and perioperative complications to evaluate the impact of this comorbidity on surgical prognosis comprehensively. The focus will be on the incidence of Dep-AUD, and further explore its potential impact on patients’ postoperative outcomes.

Materials and methods

Data sources

Data were extracted from the 2010–2019 National Inpatient Sample (NIS) [20], which is part of the Healthcare Cost and Utilization Project (HCUP) funded by the Agency for Healthcare Research and Quality. The NIS represents the largest publicly available all-payer inpatient database in the United States, containing approximately 8 million hospital stays annually from about 1,050 hospitals across 44 states. This sample captures roughly 20% of all US hospital discharges and represents approximately 90% of non-federal academic medical centers nationwide [21].

Data were retrieved from the database, encompassing patient demographic details (age, sex, race, pay type, hospital bed size, etc.), overall hospitalization expenses, medical diagnoses (including Depression comorbid with alcohol use disorder), as well as procedure codes as per the International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM and ICD-10-CM). This observational study made use of anonymized data that is publicly accessible, therefore it was considered exempt. The NIS is a weighted sample drawn from the HCUP State Inpatient Databases by a complex, single-cluster survey design.

Study cohort

We queried the Nationwide Inpatient Sample (NIS) to identify patients aged 18 years or older with a primary diagnosis of upper gastrointestinal malignancies (including esophageal cancer, gastric cancer, duodenal cancer, and pancreatic cancer). The ICD-9-CM and ICD-10-CM diagnostic codes for these cancers are provided in Supplementary Table S5. Among these cancer patients, we further selected those who underwent corresponding surgical resection: esophagectomy, gastrectomy, duodenectomy, and pancreatectomy. The ICD-9-PCS and ICD-10-PCS surgical codes for each procedure are also detailed in Supplementary Table S5. To maintain consistency in perioperative risk profiles for pancreatectomy, we focused on high-complexity resections involving gastrointestinal reconstruction, specifically including pancreaticoduodenectomy and total pancreatectomy. This study specifically selected patients undergoing esophagectomy, gastrectomy, duodenectomy, and pancreatectomy based on methodological considerations. First, these organs constitute the anatomically continuous upper gastrointestinal tract, and their malignancies share well-established epidemiological associations with alcohol consumption, providing a unified etiological basis for investigating the effects of alcohol use disorder (AUD) on postoperative outcomes. Second, all selected procedures represent high-complexity surgeries requiring extensive gastrointestinal reconstruction (e.g., esophago-enteric anastomosis, pancreaticojejunostomy), resulting in comparable physiological stress and similar risk profiles for severe complications (anastomotic leak, pancreatic fistula, organ dysfunction). The ICD-9-PCS and ICD-10-PCS surgical codes for each included procedure are detailed in Supplementary Table 1. A total of 65,151 patients undergoing upper gastrointestinal and pancreatic tumor resection surgery were identified based on the procedure codes from the International Classification of Disease Clinical Modification version 9 and 10 (ICD-9-PCS and ICD-10-PCS).Patients were excluded if they had missing data on certain variables, including age (n = 5), missing sex data (n = 31), primary expected payer (n = 93), elective admission (n = 114), total charge data (n = 1,025), length of stay (LOS) (n = 2), age < 18 years (n = 67), and other race (n = 3,031). After these exclusions, 60,783 patients were included in the final analysis. Among them, 637 had both depression and AUD (Dep–AUD). The remaining 60,146 patients comprised the non-comorbid group (Non–Dep–AUD), which included 50,542 patients with neither disorder, 4,912 with depression only, and 4,692 with AUD only. Accordingly, the total numbers of patients with depression and AUD were 5,549 and 5,329, respectively, as shown in Table 1. The detail information was shown in Fig. 1.

Table 1.

Characteristic of the study cohorts undergoing upper gastrointestinal and pancreatic tumor resection surgery (2010–2019)

Characteristic	Study cohorts
	Dep–AUDa n = 637	Non–Dep–AUDb n = 60,146	P value (Dep–AUD VS None)	Depression n = 5,549	P value (Depression VS None)	AUD n = 5,329	P value (AUD VS None)
Gender (%)			0.934		< 0.001		< 0.001
Male	390(61.22)	36,723(61.06)		2883(51.96)		4005(75.15)
Female	247(38.78)	23,423(38.94)		2666(48.04)		1324(24.85)
Age (Mean ± SD)c	66.68 ± 12.043	61.27 ± 9.341		61.33 ± 10.332		66.17 ± 11.382
Race (%)			< 0.001		< 0.001		< 0.001
White patients	501(78.65)	41,561(69.10)		4385(79.02)		3659(68.66)
Black patients	75(11.77)	7618(12.67)		478(8.61)		927(17.40)
Hispanic patients	38(5.97)	5843(9.71)		419(7.55)		409(7.67)
Asian or pacifc islander patients	6(0.94)	2770(4.61)		94(1.69)		123(2.31)
Native American	7(1.10)	273(0.45)		21(0.38)		42(0.79)
Other	10(1.57)	2081(3.46)		152(2.74)		169(3.17)
Pay type (%)			< 0.001		< 0.001		< 0.001
Medicare	283(44.43)	33,595(55.86)		3234(58.28)		2176(40.83)
Medicaid	127(19.94)	5540(9.21)		544(9.80)		1039(19.50)
Private insurance	190(29.83)	17,988(29.98)		1536(27.68)		1693(31.77)
Self–pay	15(2.35)	1443(2.40)		89(1.60)		227(4.26)
No charge	0(0.00)	162(0.27)		6(0.11)		25(0.47)
Other	22(3.45)	1418(2.36)		140(2.52)		169(3.17)
Hospital bed size (%)			0.180		0.413		0.269
Small	51(8.02)	5995(9.99)		524(9.46)		497(9.35)
Medium	148(23.27)	12,871(21.45)		1197(21.61)		1160(21.83)
Large	437(68.71)	41,139(68.56)		3818(68.93)		3656(68.81)
Location of hospital (%)			0.169		0.609		< 0.001
Rural	22(35.46)	1552(2.59)		138(2.49)		191(3.60)
Urban	614(96.54)	58,453(97.41)		5401(97.51)		5122(96.41)
Region of hospital (%)			< 0.001		< 0.001		< 0.001
Northeast	126(19.78)	13,975(23.24)		1245(22.44)		1106(20.75)
Midwest or North Central	164(25.75)	12,057(20.05)		1279(23.05)		1175(20.05)
South	251(39.40)	22,788(37.89)		2045(36.85)		2171(40.74)
West	96(15.07)	11,326(18.83)		980(17.66)		877(16.46)
Hospital teaching status (%)			0.181		0.926		< 0.001
Teaching	510(80.19)	49,342(82.23)		4551(82.16)		4222(79.47)
Non–teaching	126(19.81)	10,663(17.77)		988(17.84)		1091(20.53)
Admission type (%)
Elective	370(58.08)	34,803(57.86)	0.911	3024(54.50)	< 0.001	3070(57.61)
Non–elective	267(41.92)	25,342(42.14)		2525(45.50.)		2259(42.39)
Died during hospitalization (%)	20(3.14)	2125(3.53)	0.596	136(2.45)	< 0.001	206(3.87)	0.163
Cancer diagnosis(%)
Esophageal cancer	172(27.00)	465(73.00)	< 0.001	1123(20.24)	0.002	1469(27.57)	< 0.001
Gastric cancer	174(23.32)	463(72.68)	< 0.001	1667(30.22)	< 0.001	1772(36.37)	< 0.001
Duodenal cancer	17(2.67)	620(97.33)	0.035	207(3.73)	0.016	158(2.96)	< 0.001
Pancreatic cancer	282(44.27)	265(55.73)	0.278	2609(47.02)	< 0.001	1995(37.44)	< 0.001
Surgical diagnosis(%)
Esophagectomy	211(33.12)	426(66.88)	< 0.001	1542(22.79)	0.019	1695(31.81)	< 0.001
Gastrectomy	230(36.11)	407(63.89)	< 0.001	2274(40.98)	< 0.001	22,219(41.64)	< 0.001
Duodenectomy	180(28.26)	457(71.74)	0.747	1528(27.54)	0.790	1395(26.18)	0.010
Pancreatectomy	257(40.35)	380(59.65)	0.459	2332(42.03)	< 0.001	1890(35.47)	< 0.001
Pre–existing comorbidity(%)
Congestive heart failure	44(6.91)	4390(7.30)	0.705	446(8.04)	0.026	304(5.70)	< 0.001
Chronic pulmonary disease	230(36.11)	10,202(16.96)	< 0.001	1387(25.00)	< 0.001	1517(28.47)	< 0.001
Coagulopathy	53(8.32)	4919(8.18)	0.897	474(8.54)	0.302	457(8.58)	0.270
Diabetes, uncomplicated	130(20.41)	12,184(20.26)	0.925	1145(20.63)	0.465	891(16.72)	< 0.001
Diabetes with chronic complications	42(6.59)	6289(10.46)	0.001	775(14.97)	< 0.001	279(5.24)	< 0.001
Hypertension (combine uncomplicated and complicated)	391(61.38)	36,507(60.70)	0.725	3672(66.17)	< 0.001	2934(55.93)	< 0.001
Liver disease	104(16.33)	4636(7.71)	< 0.001	564(10.16)	< 0.001	660(12.39)	< 0.001
Metastatic cancer	216(33.91)	19,854(33.01)	0.631	1758(31.68)	0.026	1710(32.09)	0.130
Obesity	61(9.58)	6836(11.37)	0.157	845(15.23)	< 0.001	404(7.58)	< 0.001
Peripheral vascular disorders	49(7.69)	3434(5.71)	0.032	344(6.20)	0.115	371(6.96)	< 0.001
Renal failure	31(4.87)	5132(8.53)	0.001	504(9.08)	0.099	246(4.12)	< 0.001
Solid tumor without metastasis	220(34.54)	30,542(50.78)	< 0.001	2973(53.58)	< 0.001	1717(32.22)	< 0.001
Valyular disease	23(3.61)	2693(4.48)	0.292	262(4.72)	0.338	166(3.12)	< 0.001
Weight loss	232(36.42)	16,962(20.04)	< 0.001	1697(30.58)	< 0.001	1743(32.71)	< 0.001

^a Dep–AUD: participants with both depression and AUD

^bNon–Dep–AUD: participants without comorbid depression and alcohol use disorder, including those with neither diagnosis, depression only, or AUD only

^c SD: standard difference

Fig. 1.

The flow chart of included patients with upper gastrointestinal and pancreatic tumor resection surgery

For each cohort, we assessed the patient and hospital-specific factors that could potentially impact postoperative morbidity. Inherent patient characteristics encompassed gender, age, racial background, payment method, Hospital characteristics included hospital bed size, hospital teaching status, and hospital location.

Data collection

After upper gastrointestinal and pancreatic tumor resection surgery, perioperative adverse events were searched from the database. Based on the ICD-9/10 codes (detailed in Supplementary Table S5), we identified the following outcomes: stroke, wound dehiscence, postoperative infection, postoperative hemorrhage, postoperative delirium, invasive respiratory support.To minimize the risk of misclassification, postoperative delirium was defined using specific ICD-10-CM codes (F05, R41.0) and ICD-9-CM codes (293.0, 293.1, 292.81). Critically, we explicitly excluded codes that could represent alcohol withdrawal syndromes, metabolic encephalopathy, or organic cognitive disorders (e.g., ICD-9-CM 293.8, 293.9, and 780.0x) to ensure a more specific and homogenous definition of delirium in this high-risk population. Healthcare utilization was quantified through analysis of both total hospitalization costs and LOS. Extended LOS and elevated costs were operationally defined as values exceeding the 75th percentile of the study population’s distribution.

Statistical analysis

Multivariable logistic regression analyses were performed to calculate Odds Ratios (ORs) and 95% Confidence Intervals (CIs) for postoperative outcomes, adjusting for demographic characteristics and comorbidities.

Assessment of Synergistic Interaction: To formally test the synergistic effect of comorbid Depression and AUD, we employed both a multiplicative interaction term and an additive interaction model. First, a multiplicative interaction term (Depression ×AUD) was included in the logistic regression model. The statistical significance of this multiplicative interaction was assessed using the Wald test, and a P value < 0.05 was considered statistically significant. Second, to quantify the biological interaction on the additive scale, we calculated the Relative Excess Risk due to Interaction (RERI) using the following formula: RERI=OR₁₁−OR₁₀−OR₀₁+1.

Where: OR₁₁= Odds Ratio for the Dep‑AUD group (exposed to both factors),

OR₁₀= Odds Ratio for the Depression‑only group.

OR₀₁= Odds Ratio for the AUD-only group.

OR ₀₀ = Reference group (Non-Dep-AUD, set to 1).

In addition, we computed the Attributable Proportion (AP) and the Synergy Index (S) to further characterize additive interaction:

A multiplication interaction with P < 0.05 is considered statistically significant, and RERI is used to measure the absolute magnitude (or degree) of this interaction effect [22]. The definitions and interpretations of these measures follow standard epidemiological practice [23].Four-Category Comparison: To address direct comparisons, we constructed an omnibus four-category regression model (Dep‑AUD vs. Depression-only vs. AUD-only vs. Non-Dep-AUD) to simultaneously evaluate the adjusted risk differences between groups.

No ethical approval was required because the study used a publicly accessible database. All statistical analyses were conducted using SPSS version 26 (IBM Corp., Armonk, NY).

Results

Demographics in four groups

This study ultimately included 60,783 patients who underwent upper gastrointestinal and pancreatic tumor resection, including 5,549 in the depression-only group, 5,329 in the alcohol use disorder-only (AUD-only) group, 637 in the depression comorbid with alcohol use disorder (Dep-AUD) group, 60,146 were diagnosed with neither depression nor AUD.

As shown in Table 1, the average age of patients in the Dep-AUD group was 66.68 ± 12.04 years, which was slightly higher than that of the pure depression group (61.33 ± 10.33 years) and the control group, and also higher than that of the pure AUD group (66.17 ± 11.38 years). In terms of gender distribution, the proportion of males in the Dep-AUD group was 61.22%, significantly higher than that in the pure depression group (51.96%), but lower than that in the pure AUD group (75.15%). Regarding racial composition, white patients accounted for the highest proportion in the Dep-AUD group (78.65%).

Regarding economic status, the proportion of patients in the Dep-AUD group using Medicaid (an insurance for low-income individuals) was as high as 19.94%, significantly higher than that of the control group (9.21%) and the group with pure depression (9.80%). This suggests that this comorbid group may face more severe socioeconomic pressures.

In terms of comorbidities, the incidence rates of hypertension (61.38%), chronic lung disease (36.11%), and weight loss (36.42%) in the Dep-AUD group were between those in the pure depression group and the pure AUD group, but significantly higher than those in the control group. It is worth noting that the prevalence of liver disease in the Dep-AUD group (16.33%) was significantly higher than that in the pure depression group (10.16%) and the control group (7.71%), which may be related to the direct damage to the liver caused by alcohol intake.

Overall postoperative outcomes

Healthcare utilization

The median length of stay (LOS) for the Dep-AUD group was 9.00 days (IQR 6.00-15.00), compared to 8.00 days for the control group. The median total hospital charges for this cohort were $102,998.00 (IQR $58,044.00-$169,398.50) (Table 2).

Table 2.

Healthcare utilization after upper gastrointestinal and pancreatic oncologic resections in Non–Dep–AUD, Dep–AUD, AUD, and depression patients

	Dep–AUD n = 637	Non–Dep–AUD n = 60,146	AUD n = 5,329	Depression n = 5,549
Length of stay (day)	9.00 (6.00, 15.00)	8.00 (5.00, 13.00)	9.00 (6.00, 14.00)	8.00 (6.00, 13.00)
Total charges ($) (median (IQR)	102,998.00(58,044.00–169,398.50)	94,358.50(55,453.50–161,611.00)	98,108.00(58,236.50–167,206.00)	94,293.00(54,272.50–164,665.50)

Overall cohort

In the overall cohort of patients undergoing upper gastrointestinal cancer resections (including esophagectomy, gastrectomy, duodenectomy, and pancreatectomy), the association between depression and/or AUD and postoperative outcomes was assessed using multivariable logistic regression, with patients having neither depression nor AUD serving as the reference group (see Table 3).

Table 3.

Association of depression and alcohol use disorder with postoperative outcomes in all cancer resection surgeries

Outcome	Dep only OR (95% CI)	AUD only OR (95% CI)	Both OR (95% CI)	P- interaction	RERI	AP	S
Stroke	0.796 (0.549,1.154)	1.021 (0.688,1.516)	0.534 (0.132,2.160)	0.580	-0.283	-0.530	2.546
Wound Dehiscence	1.016 (0.763,1.354)	0.910 (0.682,1.213)	2.156 (1.311,3.544)	0.008	1.230	0.571	-15.622
Postoperative Infection	1.091 (0.928,1.284)	1.095 (0.933,1.284)	0.949 (0.616,1.463)	0.349	-0.237	-0.250	-0.274
Postoperative Hemorrhage	0.883 (0.673,1.159)	1.560 (1.265,1.925)	1.418 (0.811,2.482)	0.930	-0.025	-0.018	0.944
Postoperative Shock	0.676 (0.453,1.010)	1.409 (1.048,1.894)	0.374 (0.092,1.510)	0.213	-0.711	-1.901	-7.365
Postoperative Fistula	0.896 (0.536,1.499)	1.599 (1.076,2.378)	1.212 (0.384,3.828)	0.800	-0.283	-0.233	0.428
Postoperative Gastrointestinal Complications	1.022 (0.908,1.149)	0.909 (0.804,1.027)	1.161 (0.871,1.548)	0.181	0.230	0.198	-2.333
Postoperative Pulmonary Complications	0.869 (0.765,0.987)	1.700 (1.535,1.881)	1.293 (0.981,1.704)	0.406	-0.276	-0.213	0.515
Dysphonia	1.339 (0.956,1.877)	1.519 (1.138,2.028)	1.607 (0.788,3.278)	0.572	-0.251	-0.156	0.707
Postoperative Delirium	1.846 (1.573, 2.166)	1.205 (0.979,1.484)	2.271 (1.516,3.402)	0.932	0.220	0.097	1.209
Invasive Respiratory Support	0.992 (0.888,1.107)	1.637 (1.488,1.801)	1.472 (1.153,1.879)	0.491	-0.157	-0.107	0.750
Atrial Fibrillation and Flutter	1.044 (0.957,1.138)	0.930 (0.842,1.026)	1.185 (0.934,1.503)	0.143	0.211	0.178	-7.115
Blood Transfusion	1.015 (0.939,1.098)	1.403 (1.299,1.514)	1.307 (1.071,1.595)	0.449	-0.111	-0.085	0.734
Urinary Tract Infection	1.136 (1.017,1.268)	1.248 (1.092,1.425)	1.288 (0.941,1.763)	0.595	-0.096	-0.075	0.750
Postoperative Ileus	0.927 (0.767,1.121)	0.857 (0.702,1.046)	0.843 (0.501,1.416)	0.843	0.059	0.070	0.727
Gastrointestinal Bleeding	0.899 (0.788,1.025)	0.994 (0.869,1.137)	0.740 (0.494,1.109)	0.402	-0.153	-0.207	2.430
Extended LOS	1.097 (1.022,1.178)	1.285 (1.197,1.380)	1.346 (1.127,1.608)	0.647	-0.036	-0.027	0.906
Increased Cost	1.049 (0.979,1.125)	0.992 (0.924,1.065)	1.023 (0.854,1.227)	0.871	-0.018	-0.018	0.561
Died During Hospitalization	0.645 (0.531,0.782)	1.111 (0.945,1.306)	0.880 (0.557,1.389)	0.433	0.124	0.141	0.492

Data are presented as odds ratios (ORs) with 95% confidence intervals (CIs). The reference group is patients with neither depression nor alcohol use disorder (AUD). P for interaction tests the statistical significance of the multiplicative interaction. A P value < 0.05 indicates a significant multiplicative interaction. RERI is used to measure the absolute magnitude of the additive interaction

The depression-only group was associated with significantly lower odds of in-hospital mortality (OR = 0.645, 95% CI: 0.531–0.782) and postoperative pulmonary complications (OR = 0.869, 0.765–0.987). In contrast, depression alone was associated with increased odds of postoperative delirium (OR = 1.846, 1.573–2.166), urinary tract infection (OR = 1.136, 1.017–1.268), and extended length of hospital stay (LOS) (OR = 1.097, 1.022–1.178).

The AUD-only group was associated with increased odds of numerous complications, including postoperative hemorrhage (OR = 1.560, 1.265–1.925), postoperative shock (OR = 1.409, 1.048–1.894), postoperative fistula (OR = 1.599, 1.076–2.378), postoperative pulmonary complications (OR = 1.700, 1.535–1.881), dysphonia (OR = 1.519, 1.138–2.028), invasive respiratory support (OR = 1.637,1.488–1.801), blood transfusion (OR = 1.403, 1.299–1.514), urinary tract infection (OR = 1.248, 1.092–1.425), and extended LOS (OR = 1.285, 1.197–1.380).

For the combined exposure (both depression and AUD), a significant multiplicative interaction was observed only for wound dehiscence (P for interaction = 0.008). No significant multiplicative interaction was detected for postoperative delirium (P for interaction = 0.932) or any other outcomes, suggesting that the elevated risks observed for these complications primarily reflect the additive effects of the two conditions rather than true synergy. Regarding additive interaction, positive synergy (RERI > 0, AP > 0) suggesting excess risk beyond the sum of individual effects was observed for wound dehiscence (RERI = 1.230, AP = 0.571), postoperative delirium (RERI = 0.220, AP = 0.097), and gastrointestinal complications (RERI = 0.230, AP = 0.198).

Subgroup analysis by surgical procedure

Stratified analyses by surgical type demonstrated marked heterogeneity in the effects of depression and AUD across procedures (Tables S1-S4).

Among patients undergoing esophagectomy (Table S1), significant multiplicative interactions were observed for wound dehiscence (P for interaction = 0.043). Positive additive interactions (RERI > 0, AP > 0) were observed for wound dehiscence (RERI = 1.606, AP = 0.656), postoperative hemorrhage (RERI = 1.174, AP = 0.398), postoperative delirium (RERI = 0.874, AP = 0.347), and gastrointestinal complications (RERI = 0.466, AP = 0.313). Depression alone was associated with lower odds of postoperative pulmonary complications (OR = 0.757, 0.608–0.942) and lower in-hospital mortality (OR = 0.678, 0.485–0.947). AUD alone was associated with increased odds of invasive respiratory support (OR = 1.743, 1.500–2.026) and dysphonia (OR = 1.618, 1.160–2.255).

In patients undergoing gastrectomy (Table S2), a significant multiplicative interaction was observed for wound dehiscence (P for interaction = 0.011). Positive additive interactions were identified for wound dehiscence (RERI = 1.921, AP = 0.685), gastrointestinal complications (RERI = 0.337, AP = 0.272), atrial fibrillation and flutter (RERI = 0.424, AP = 0.272), and urinary tract infection (RERI = 0.129, AP = 0.095). Depression alone was associated with increased odds of extended LOS (OR = 1.133, 1.017–1.262) and increased cost (OR = 1.115, 1.003–1.239). AUD alone was associated with increased odds of invasive respiratory support (OR = 1.561, 1.352–1.801), blood transfusion (OR = 1.436, 1.282–1.608), and extended LOS (OR = 1.425, 1.281–1.584).

For patients undergoing duodenectomy (Table S3), a significant multiplicative interaction was observed for urinary tract infection (P for interaction = 0.046). Positive additive interactions included postoperative hemorrhage (RERI = 0.514, AP = 0.328), postoperative delirium (RERI = 0.522, AP = 0.234), blood transfusion (RERI = 0.241, AP = 0.154), and extended LOS (RERI = 0.139, AP = 0.098). Due to small sample sizes and low event rates in the combined exposure stratum, certain outcome estimates (e.g., for stroke) were mathematically unstable and are denoted as “—” in the tables.

In patients undergoing pancreatectomy (Table S4), comorbid Dep-AUD was significantly associated with an increased risk of postoperative delirium (OR 2.615, 1.398–4.889) and blood transfusion requirement (OR 1.813, 1.351–2.435). Notably, the risk of postoperative delirium in the Dep-AUD group exceeded that of either condition alone (depression only: OR 1.836, 1.431–2.357; AUD only: OR 1.061, 0.724–1.554). Although the multiplication effect is not statistically significant (P for interaction = 0.444), positive additive interaction was observed for postoperative delirium (RERI = 0.718, AP = 0.275, S = 1.800), indicating that the combined effect of Dep-AUD on delirium risk exceeded the sum of their individual effects. Similarly, blood transfusion showed positive additive interaction (RERI = 0.340, AP = 0.188), despite the absence of significant multiplicative interaction (P = 0.258).

AUD group was significantly associated with postoperative hemorrhage (OR 1.526, 1.133–2.054) and postoperative pulmonary complications (OR 1.847, 1.533–2.225), whereas depression only showed no significant association with these outcomes. Both depression alone (OR 1.128, 1.009–1.262) and AUD group (OR 1.252, 1.107–1.415) were independently associated with extended length of stay, with the comorbid group demonstrating the highest risk (OR1.353,1.015–1.804). Interestingly, depression alone was associated with a reduced risk of in-hospital mortality (OR 0.545,0.369–0.805).

Discussion

This nationwide analysis demonstrates that the comorbid state of depression and alcohol use disorder (Dep-AUD) exerts procedure-dependent effects on postoperative morbidity following resections for upper gastrointestinal and pancreatic tumors. While a pooled analysis provides an overarching framework, the stratified analyses by surgical subtype reveal marked heterogeneity, underscoring that the impact of Dep-AUD is not monolithic but is instead modulated by the unique physiological stress and anatomical intricacies of each procedure [24]. For postoperative delirium and most other outcomes, multiplicative interactions were not significant, although positive additive interactions (RERI > 0) were observed in some analyses. Only wound dehiscence demonstrated a significant multiplicative interaction in the overall cohort (P < 0.05), with RERI = 1.23 and AP = 0.57.

Comorbid depression and AUD increase wound dehiscence risk in esophageal and gastric surgery

Wound dehiscence is the complication that can best reflect the synergistic effect of mental and behavioral factors rather than a single exposure effect in this study. In the overall cohort of patients, Dep-AUD significantly increased the risk of wound dehiscence, and the interaction effect still existed in the subgroup of esophagectomy and gastrectomy. It is worth noting that neither depression nor AUD alone showed independent association with wound dehiscence in the above subgroups, which further highlighted the specificity of the synergy between the two. The biological mechanism of this discovery has been fully confirmed. Recent meta-analysis showed that depression can delay wound healing and increase the risk of wound complications through excessive activation of hypothalamic pituitary adrenal (HPA) axis [25], persistent systemic inflammatory response, immunosuppression, and behavioral factors such as sleep disorders, malnutrition, reduced activity and postoperative care compliance [26]. Chronic alcohol use acts as a secondary insult: preoperative drinking is strongly associated with elevated risks of postoperative complications, infections, and wound-related adverse events. The underlying mechanisms involve ethanol-mediated disruption of immune signaling pathways and impaired tissue repair processes [27]. This observed synergy in wound dehiscence during esophagectomy and gastrectomy is consistent with the convergence of impaired healing biology and the inherently high wound or anastomotic tension characteristic of these two procedures. However, this pattern was not seen in duodenectomy or pancreatectomy, suggesting that the same mental-behavioral risk may produce different postoperative outcomes depending on the surgical site and its technical complexity. This operation specific risk regulation mode emphasizes that the psychological vulnerability at the patient level and the physiological stress characteristics at the operation level should be included in the perioperative risk assessment [28].

Respiratory complications and invasive respiratory support were most consistently associated with AUD, especially in esophagectomy, gastrectomy, and duodenectomy, while depression only played a less stable role. This is biologically plausible because esophagectomy is a highly morbid procedure in which pulmonary complications are particularly frequent and are a major driver of ICU admission and prolonged hospitalization; reviews of esophagectomy consistently describe pneumonia, ARDS, and acute lung injury as key postoperative threats [29, 30]. Alcohol may compound respiratory risk through immune suppression and aspiration vulnerability, factors particularly relevant in upper gastrointestinal surgery where postoperative pulmonary complications are already common [31, 32]. In contrast, the respiratory effect of depression is less direct and may depend more on indirect pathways such as reduced self-care, altered stress responses, and coexisting frailty rather than on an intrinsic pulmonary mechanism [33].

Postoperative delirium

Postoperative delirium emerged as the most consistently depression-related complication across surgical subgroups, with the comorbid Dep-AUD state conferring the highest delirium burden through additive rather than multiplicative mechanisms. This finding aligns with prior evidence demonstrating that preoperative depression independently elevates delirium risk after major surgery, and that alcohol use disorder is similarly associated with postoperative delirium [34]. Mechanistically, depression is linked to systemic inflammation, autonomic imbalance, and hypothalamic–pituitary–adrenal (HPA) axis dysregulation, all of which can lower the threshold for acute postoperative neurocognitive dysfunction [35]. Alcohol use disorder may contribute through withdrawal physiology, neurotransmitter instability, and baseline cognitive vulnerability. The heightened delirium signal in duodenectomy and pancreatectomy likely reflects the greater physiologic stress, prolonged operative times, higher transfusion requirements, and intensive postoperative monitoring that characterize these procedures—all well-established contributors to delirium in major abdominal surgery [36]. Our delirium definition excluded alcohol withdrawal and metabolic encephalopathy. This strengthens outcome specificity, suggesting that the excess risk in the Dep-AUD group is unlikely to be an artifact of withdrawal-related coding, but rather reflects a genuine synergistic neurocognitive vulnerability driven by the interplay of depression-related neuroinflammation and alcohol-induced neuroadaptation .

Alcohol-associated coagulopathy and perioperative hemorrhage in upper GI surgery

Blood transfusion and postoperative hemorrhage were linked more strongly to AUD than to depression, and this was most obvious in esophagectomy, gastrectomy, duodenectomy, and pancreatectomy. This is consistent with the known hematologic effects of alcohol, including platelet dysfunction, prolonged bleeding tendency, and coagulation abnormalities, which can increase perioperative blood loss and transfusion requirements [37]. In esophagectomy, this may be worsened by the technical complexity of thoracoabdominal dissection and the already high baseline risk of postoperative morbidity; in pancreatic surgery, the alcohol-related bleeding risk appears especially strong, likely because these operations involve extensive vascular dissection and substantial blood loss [38]. Depression, in contrast, is more plausibly linked to transfusion indirectly through poorer nutritional status and inflammatory dysregulation than through a direct bleeding mechanism [39].

Cardiopulmonary and resource utilization implications

The consistent association between AUD and postoperative pulmonary complications across all surgical subtypes (overall OR 1.700, esophagectomy OR 1.524, gastrectomy OR 1.826, duodenectomy OR 1.861, pancreatectomy OR 1.847) aligns with recognized mechanisms including alcohol-induced immunosuppression, aspiration risk, and the pro-inflammatory state of alcohol withdrawal syndrome [40]. A recent meta-analysis of over 18 million surgical patients confirmed that preoperative AUD significantly increases respiratory complications (RR 2.59), infections, hospital stay, and in-hospital mortality [41]. Significant comorbid interaction effects for invasive respiratory support in the overall cohort (OR 1.472) and gastrectomy (OR 1.673) suggest that depression-related respiratory muscle weakness may compound AUD-associated pulmonary vulnerability [42].

The combined effect of Dep-AUD on blood transfusion was strongest in duodenectomy and pancreatectomy, while longer hospital stays were seen across duodenectomy, gastrectomy, and the overall cohort. These patterns reflect how complications like delirium, wound problems, and withdrawal management add to the burden of recovery. A systematic review and meta-analysis confirmed that AUD independently increases postoperative transfusion risk, especially when comorbid with depression [41], and a meta-analysis demonstrated that postoperative delirium significantly prolongs hospitalization [43]. Prolonged length of stay was the most universal outcome, appearing across all subgroups, suggesting that depression and AUD worsen recovery even when they do not manifest as a single dominant organ-specific complication. This aligns with evidence that wound complications, respiratory failure, delirium, and transfusion all lengthen hospitalization and increase ICU-level care needs.

While inverse associations were observed for depression alone and AUD alone in some subgroups, these likely reflect selection bias or unmeasured confounding rather than true biological protection, as noted in our limitations. For instance, depression alone was associated with lower odds of postoperative pulmonary complications in esophagectomy (OR = 0.757, 0.608–0.942) and reduced in-hospital mortality in both the overall cohort (OR = 0.645, 0.531–0.782) and pancreatectomy (OR = 0.545, 0.369–0.805). Similarly, AUD alone showed apparent protective associations for gastrointestinal complications and atrial fibrillation following pancreatectomy. These counterintuitive findings likely reflect healthy-user effects, coding artifacts, or unmeasured confounding, and should not be interpreted as causal protective effects. Similarly, the isolated significant multiplicative interaction for urinary tract infection in duodenectomy was an unexpected finding and, given the small sample size and sparse events in this subgroup, should be interpreted with caution rather than as a robust procedure-specific signal.

Clinical and policy implications

The interaction effects observed for wound dehiscence, particularly in esophagectomy and gastrectomy, support considering preoperative mental health and substance use screening in patients undergoing upper gastrointestinal cancer surgery. Tools such as the Patient Health Questionnaire-9 for depression and the Alcohol Use Disorders Identification Test for AUD may help identify patients who could benefit from closer perioperative monitoring and targeted prevention strategies.

The variation in interaction effects across procedures also suggests that risk mitigation approaches should be tailored to surgical context. For esophagectomy and gastrectomy, where wound-related interactions were most pronounced, preoperative nutritional support and alcohol abstinence may be particularly relevant. For pancreatectomy and duodenectomy, where delirium showed the strongest association with comorbid depression and AUD, enhanced postoperative cognitive monitoring and early mobilization may warrant greater attention.

Strengths and limitations

This study has several advantages. We used a large national database, adjusted for multiple confounders, tested for additive interaction using RERI/AP/SI, and analyzed each operation type separately. However, there are some limitations. First, the NIS database uses billing codes, which may miss some psychiatric diagnoses and complications. Second, while we excluded alcohol withdrawal and metabolic encephalopathy from our delirium definition, some misclassification may still remain. Third, we did not have detailed data on depression severity, how much patients drank, or what medications they used. Fourth, some subgroups had few events, making interaction estimates unstable. For example, the synergy index for urinary tract infection in duodenectomy produced an implausible protective estimate with a very wide confidence interval. Some Dep-AUD strata had zero events, so ORs could not be calculated. These subgroup findings should be treated as exploratory. Fifth, some subgroups showed unexpected inverse associations for AUD alone and depression alone. These likely reflect selection bias, confounding, or coding issues, and should not be seen as true protective effects. Finally, our analysis only shows associations between variables and cannot establish cause and effect. Taken together, the subgroup results are suggestive rather than conclusive, and should be validated in future research.

Conclusion

In summary, depression and AUD were associated with different but overlapping postoperative risks after upper gastrointestinal cancer surgery, and their coexistence conferred the highest risk, especially for wound dehiscence, delirium, blood transfusion, invasive respiratory support, and prolonged LOS. The strength and direction of these associations varied by surgical site, underscoring the importance of procedure-specific perioperative risk stratification and targeted prehabilitation.

Abbreviations

AHRQ

Agency for Healthcare Research and Quality

aOR

Adjusted odds ratio

AP

Attributable Proportion

ARDS

Acute respiratory distress syndrome

AUD

Alcohol use disorder

CCI

Charlson Comorbidity Index

CI

Confidence Interval

Dep-AUD

Depression comorbid with alcohol use disorder

HCUP

Healthcare Cost and Utilization Project

HPA

Hypothalamic-pituitary-adrenal

ICD

International Classification of Diseases

ICU

Intensive care unit

IQR

Interquartile range

IRB

Institutional review board

LOS

Length of stay

NIS

National Inpatient Sample

OR

Odds Ratio

RERI

Relative Excess Risk due to Interaction

S

Synergy Index

UGI

Upper gastrointestinal

Authors' contributions

XZF wrote the first draft of the paper and analyzed the data. GHP, MCL and ZY interpreted the data and critically revised the manuscript. WJ and XL conceptualized this study, interpreted the data, and critically revised the manuscript. All authors approved the final submitted manuscript.

Funding

This work was supported by the Ganzhou City “Technology+National Regional Medical Center” Joint Plan Project [No: 2022–YB1319].

Data availability

Study utilizes data from the Nationwide Inpatient Sample (NIS), which is a component of the Healthcare Cost and Utilization Project (HCUP) managed by the Agency for Healthcare Research and Quality (AHRQ). The NIS is a large, publicly accessible inpatient database in the United States. Access to the database is available through the HCUP information portal: https://www.hcup-us.ahrq.gov/db/nation/nis/nisdbdocumentation.jsp. Due to restrictions, individual-level or aggregated datasets cannot be shared or distributed by the authors.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.