Outcomes of pancreatoduodenectomy in underlying liver cirrhosis: a single institution experience and literature review

Abstract

Introduction

Pancreatoduodenectomy (PD) is a complex procedure associated with up to 20–30% morbidity and 1–2% mortality. In the present study, we aimed to evaluate the outcomes of PD in patients with cirrhotic liver and assessed predictors of perioperative morbidity and mortality.

Materials and methods

This is a retrospective study from a prospectively maintained database. Amongst the patients who underwent PD between January 2013 till June 2024, only patients who had underlying cirrhotic liver were included in the study.

Results

A total of 24 patients were included. On binary logistic regression history of pancreatitis in the preoperative period (OD-25.8), stent block (OD-64.9), cholangitis (OD-273), and preoperative platelets less than 1.37(OD-40), preoperative INR more than 1.31(OD-40) and platelets count on POD1 less than 1.23 lakhs (OD-40) were associated with mortality.

Conclusion

Patients with clinically significant portal hypertension with thrombocytopenia and a deranged coagulation profile are associated with a high risk of mortality.

Introduction

Pancreatoduodenectomy (PD) is a complex procedure performed for benign & malignant tumors of the periampullary region [1]. This complex procedure is associated with up to 20–30% morbidity and 1–2% mortality when performed at high-volume centres or by experienced surgeons [1, 2]. Multiple factors affect the morbidity and prognosis after PD, which include patient-related factors, tumour-related factors, the risk status of the pancreatic stump for anastomosis, and surgeon experience [1]. Traditionally, cirrhosis was considered to be a contraindication for performing PD or any other major gastrointestinal surgery as the associated mortality was very high, ranging from 17.5 to 38% [3, 4]. A hyperdynamic circulation with increased cardiac output and decreased systemic vascular resistance is present in the setting of cirrhosis. Decreased baseline hepatic perfusion causes the cirrhotic liver to be more vulnerable to hypoxemia and hypotension during surgery [4]. Hepatic blood flow may also be reduced in the operating room due to certain anesthetic agents, intraoperative hemorrhage, vasoactive drugs, and even traction on the abdominal viscera [5, 6]. Hence, patients with cirrhosis are at risk of liver decompensation due to the stress of both anesthesia and surgery. Patients with cirrhosis undergoing major surgery have a significantly increased risk of complications which include bleeding due to portal hypertension and coagulopathy, liver decompensation, and sepsis [1]. However, with improvements in strategies of preoperative optimization, minimized intraoperative blood loss, and intensive postoperative care, major surgeries can be performed in selected cirrhotic patients [7]. As resection remains the only chance of cure in periampullary malignancies, selected cases even with cirrhosis can be subjected to surgery. Determining the risk-benefit ratio in these patients is challenging [7].

In the present study, we aimed to evaluate the outcomes of PD in patients with cirrhotic liver and looked at the predictors of perioperative morbidity and mortality.

Materials and methods

This is a retrospective study from a prospectively maintained database. All patients who underwent PD between January 2013 till June 2024 were screened. Amongst them, only patients who had underlying liver cirrhosis diagnosed pre-operatively or intraoperatively were included in the study. Ethics committee approval was not required, given the retrospective nature of the study. Data collection was done as described in the declaration of Helsinki [8].

All patients were evaluated with a triple-phase contrast-enhanced computed tomography of the thorax, abdomen, and pelvis. Preoperative diagnostic workup for all patients included clinical assessment, and laboratory investigations (complete blood counts, liver functions, HCV and HBV markers, renal function tests, liver function tests, and tumor markers, such as CEA and CA19-9). Preoperative endoscopic retrograde cholangiopancreatography (ERCP) and biliary drainage were performed in patients with serum total bilirubin levels exceeding 10 mg/dL, in those presenting with cholangitis, or when surgery was deferred for optimisation of nutritional status or comorbidities. In patients with liver cirrhosis and obstructive jaundice, the threshold for preoperative biliary drainage was lowered to 5 mg/dL, considering their limited hepatic reserve and increased risk of postoperative decompensation. The diagnosis of cirrhosis was based on characteristic imaging features on a CT scan or ultrasound. Liver biopsies were performed in some cases. Clinically significant portal hypertension was indirectly defined based on the presence of oesophageal varices detected by endoscopy, or splenomegaly (major diameter > 12 cm) with thrombocytopenia. In the present study, we had taken only Child A patients for surgery; Child B and C are contraindication for major oncological resections as per our institutional protocol.

PD was performed with a standardized approach as described previously in our publications [2, 9]. The complications were defined as per the International Study Group for Pancreatic Surgery (ISGPS) classification [10–14] as well as the Clavien Dindo (CD) classification. A score more than 3 in CD classification was defined as a major complication/significant morbidity [15]. Mortality was defined as death within 90 days of surgery. Primary outcome measures studied were 90-day mortality and major complication rate. Statistical analyses were performed using the Statistical Product and Service Solutions (SPSS), version 25.0, for Windows (SPSS Inc., Chicago, IL, USA). Descriptive statistics were calculated for continuous variables and reported as the mean ± SD. Categorical variables were described using frequency distributions. A receiver operator characteristic (ROC) curve was constructed to identify the cut-off value of continuous variables that significantly predicted mortality or morbidity. Bivariate logistic regression was used to determine the odds ratio in predicting morbidity and mortality. A p-value of less than 0.05 was considered as significant.

Results

Out of 1594 PDs performed during the period, 24 patients had underlying cirrhosis. Baseline characteristics of the participating patients are illustrated in Table 1. The median age was 54(31–76) years with a higher proportion of males 18(75%). The median BMI was 24.6(16.8–31.25.8.25) kg/m². Jaundice was observed in 19 (79.2%) patients during initial presentation, and all of them underwent biliary drainage. Features of cholangitis were seen in 3 (12.5%) patients. The median serum bilirubin at presentation was 8.75(1.38–36.38) mg/dL. The pre-surgery median serum bilirubin was 0.84 (range of 0.38–6.71) mg/dL. Diagnosis of cirrhosis was known in 22 (91.7%) patients before surgery, while 2 patients were intraoperatively detected to have a cirrhotic liver. Preoperative upper GI endoscopy was performed in all 22 patients with a preoperative diagnosis of cirrhosis, and esophageal varices were detected in only 1 (4.2%) patient. This patient was a known case of hepatitis B on anti-viral treatment and had grade 2 oesophageal varices. Postoperatively, the patient had an SSI, which was managed conservatively. There were no derangements in platelet count and coagulation profile during the perioperative period.

Table 1.

Clinical characteristics of patients with cirrhosis undergoing PD (n = 24)

Characteristics		Frequency
Age		54(31–76) years
Sex	Male	18(75%)
	Female	6(25%)
Pre-Operative Haemoglobin (median)		12.2(7–15.4.4)
Preoperative platelets (median)		2.4(0.88–5.29)
Preoperative INR (median)		1.01(0.89–1.45)
BMI (median)		24.6(16.8–31.25.8.25) kg/m2
Bilirubin at presentation (median) (in mg/dL)		8.75(1.38–36.38)
Preoperative bilirubin (median) (in mg/dL)		0.84(0.38–6.71)
Preoperative albumin (median)		3.98(3.3–4.4)
Pre-operative pancreatitis		1(4.2%)
Pre-operative cholangitis		3(12.5%)
Pre-operative history of stent block		2(8.3%)
Cause of cirrhosis	Hepatitis B	9(37.5%)
	Hepatitis C	5(20.8%)
	Alcohol related	2(8.3%)
	NASH	8(33.3%)
Comorbidities	DM	7(29.2%)
	HTN	7(29.2%)
	Cardiac	2(8.3%)
	Pulmonary	2(8.3%)
Primary	Pancreas	5(20.8%)
	Non-pancreas	19(79.2%)
Pancreas consistency	Soft	10(41.7%)
	Non-soft	14(58.3%)
Duct size	< 3 mm	4(16.7%)
	> 3 mm	20(83.3%)
Duration of surgery (median)		420(240–700) minutes
Blood loss (median)		1225(500–7300) ml
Complications	CR-POPF (Grade B and C)	9(37.5%)
	Ascites	8(33.3%)
	Bile leak	1(4.2%)
	Chyle leak	4(16.7%)
	PPH	4(16.7%)
	DGE (Grade B and C)	4(16.7%)
	SSI	3(12.5%)
	Re-exploration	3(12.5%)
Hospital stay (median)		15.5(8–54) days
Readmission		2(8.3%)
Significant morbidity (> 3 CD)		12(50%)
90-day mortality		3(12.5%)
Final histology	Adenocarcinoma	17(70.8%)
	Others	7(29.2%)
Margin	R + resection	3(12.5%)
Nodal yield (median)		18(10–38)

INR International Normalised Ratio

BMI Body mass index

NASH Non-alcoholic steatohepatitis

DM Diabetes Mellitus

HTN Hypertension

CR-POPF Clinically relevant postoperative pancreatic fistula

PPH Post-pancreatectomy haemorrhage

DGE Delayed Gastric Emptying

SSI Surgical site infection

CD Clavien Dindo

None of the patients had splenomegaly. Preoperative thrombocytopenia (< 1.5 lakhs) was seen in 5 (20.8%) patients.

The median duration of surgery was 420 (240–700) minutes, and the median blood loss was 1225 (500–7300) mL. There was significant morbidity with CD > 3 of 50% and mortality of (3/24)12.5%. Post-surgery, 8 (33.3%) patients developed ascites, and 4 (16.7%) developed post-pancreatectomy haemorrhage (PPH). Amongst the patients with PPH, 2 patients had hematemesis, one was managed conservatively but had respiratory failure and ascites due to liver decompensation and ultimately died, and the other patient required endoscopic variceal banding and was discharged after a hospital stay of 27 days. One patient had bleeding from superior rectal vessels which was angioembolised. The same patient had a bile leak which was managed with PTBD and SEMS placement but ultimately, he could not be salvaged. The 4th patient with PPH had bleeding from a branch of the superior mesenteric artery which was embolised, but even this patient succumbed.

Clinically relevant post-operative pancreatic fistula (CR-POPF) was seen in 9 (37.5%) patients, with grade B in 6 patients, and grade C in 3 patients. All 6 patients with POPF grade B had peripancreatic collections, which were drained with pigtail insertion. One patient amongst them had developed hematemesis due to variceal bleeding, which was managed with endoscopic banding. Amongst the 3 patients with POPF C, one patient underwent re-exploration twice, once for lavage of purulent contamination on POD 16 and subsequently on POD 26 for grade C PPH, ultimately resulting in mortality. Another patient with grade C POPF also had a bile leak from the hepaticojejunostomy site, which was managed with PTBD insertion with SEMS placement, but this patient also could not be salvaged. This patient also had postoperative bleeding from the superior rectal artery, which was managed with angioembolisation of the culprit vessel. The third patient with POPF C had undergone surgical re-exploration on POD 12 with peritoneal lavage for purulent contamination and was salvaged.

Factors predicting mortality

An ROC analysis was performed to determine the cutoff for various continuous variables predicting mortality, as depicted in Table 2. Preoperative platelet counts, INR and platelet count on day 1 were found to be significant predictors of mortality. The cut-off for preoperative platelets was 1.37 lakhs/ccm with a sensitivity of 66.7% and specificity of 95.2%. The cut-off for preoperative INR was 1.31 with a sensitivity of 66.7% and specificity of 95.2%. The cut-off for post-surgery day 1 platelets was 1.23 with a sensitivity of 66.7% and specificity of 95.2%. Figure 1 shows the ROC curves for preoperative platelets and INR.

Table 2.

ROC assessment for various continuous variables for its effect on mortality

Factors	AUC	95% CI	P value
Age	0.508	0.129–0.887	0.965
Preoperative-Hemoglobin	0.778	0.549–1.0.549.0	0.127
Preoperative-Platelets	0.937	0.812–1.0.812.0	0.016	1.37 lakhs/ccm with sensitivity of 66.7% and specificity of 95.2%.
Preoperative-INR	0.921	0.806–1.0.806.0	0.021	1.31 with sensitivity of 66.7% and specificity of 95.2%.
Post surgery day 1 platelets	0.889	0.691–1.0.691.0	0.032	1.23 with sensitivity of 66.7% and specificity of 95.2%.
BMI	0.722	0.44–1.0.44.0	0.222
Preoperative Bilirubin	0.651	0.446–0.856	0.407
Preoperative Albumin	0.698	0.445–0.952	0.275
Blood loss	0.810	0.637–0.982	0.089
Surgery duration	0.825	0.647–1.0.647.0	0.074

INR International Normalised Ratio

BMI Body mass index

Fig. 1.

ROC curve for Preoperative platelets, INR, and POD1 platelets

Binary logistic regression was performed to calculate the odds ratio (Table 3). History of pancreatitis in the preoperative period, stent block, cholangitis, and preoperative platelets less than 1.37, preoperative INR more than 1.31, and platelets count on POD1 less than 1.23 lakhs were associated with mortality. Since the sample size was limited, multivariate analysis was not performed.

Table 3.

Univariate analysis with binary logistic regression for factors affecting mortality

		OR (95%CI)	p-value
Sex	Male	1	0.45
	Female	0.34 (0.002–4.31)
Site of primary	non-pancreas	1	0.56
	pancreas	0.43 (0.003–5.57)
Preoperative pancreatitis	no	1	0.04
	yes	25.8 (1.08- 4235.11)
History of stent block	NO	1	0.004
	YES	64.9 (3.43 −10944.97)
History of cholangitis	NO	1	0.0001
	YES	273.0 (10.41–90512.21)
Jaundice at presentation	no	1	0.97
	yes	1.06 (0.06–158.91)
Smoking	NO	1	0.85
	YES	0.75(0.005–10.95)
Alcohol	NO	1	0.56
	YES	0.42(0.003–5.57)
Diabetes mellitus	NO	1	0.86
	YES	1.250(0.095–15.5)
Pancreatic duct size	< 3 mm	1	0.42
	> 3 mm	0.333(0.023–4.92)
Preoperative-Platelets	< 1.37	40(1.749–914.787.749.787)	0.01
	> 1.37	1
Preoperative-INR	< 1.31	1
	> 1.31	40(1.749–914.787.749.787)	0.02
POD1 platelets	< 1.23	0.025(0.001–0.572)
	> 1.23	1	0.02

INR International Normalised Ratio

POD Post-operative day

OR Odds ratio

CI Confidence interval

Factors affecting morbidity

A ROC assessment was performed for continuous variables as depicted in Table 4. Binary logistic regression was performed to assess factors affecting morbidity and no factor was detected to have any significant impact on morbidity [Table 5].

Table 4.

ROC assessment for various continuous variables for its effect on morbidity

Factors	AUC	95% CI	P value
Age	0.729	0.521–0.937	0.057
Preoperative-Hemoglobin	0.688	0.465–0.910	0.119
Preoperative-Platelets	0.663	0.435–0.892	0.175
Preoperative-INR	0.354	0.123–0.585	0.225
Post surgery day 1 platelets	0.722	0.501–0.944	0.065
BMI	0.538	0.301–0.766	0.751
Preoperative Bilirubin	0.365	0.128–0.601	0.260
Preoperative Albumin	0.549	0.308–0.790	0.686
Blood loss	0.670	0.445–0.896	0.157
Surgery duration	0.559	0.324–0.794	0.624

INR International Normalised Ratio

BMI Body mass index

AUC Area under the curve

CI Confidence interval

Table 5.

Univariate analysis with binary logistic regression for factors affecting morbidity

		OR (95% CI)	p-value
Sex	Male	1
	Female	1.0(0.158–6.34)	1.000
Site of primary	non-pancreas	1
	pancreas	1.66(0.225–12.35)	0.617
Preoperative pancreatitis	no	1	0.45
	yes	3.26 (0.15–495.94)
History of stent block	NO	1	0.19
	YES	6.05 (2.36–875.65)
History of cholangitis	NO	1	0.084
	YES	9.47 (0.76- 1339.95)	0.452
Jaundice at presentation	no
	yes	0.90(0.04–28.13)	0.944
Smoking	NO	1.000
	YES	2.20(0.172–4.44)	0.544
Alcohol	NO	1
	YES	0.60(0.08–23.95)	0.617
Diabetes mellitus	NO	1
	YES	3.57(0.53–18.08)	0.190
Pancreatic duct size	< 3 mm	1
	> 3 mm	1.00(0.11–1.11)	1.000

OR Odds ratio

CI Confidence interval

Discussion

Traditionally, cirrhosis was considered to be a contraindication for major gastrointestinal surgeries with a high rate of postoperative mortality, ranging from 17.5 to 38% [3, 4]. The intraoperative and postoperative course remains challenging in terms of an increased likelihood of bleeding due to coagulopathy, collateral bleeding due to portal hypertension, encephalopathy, post-operative ascites, liver decompensation, and increased risk of anastomotic dehiscence due to poor healing, and resultant local and systemic sepsis responsible for increased morbidity and mortality.

Pancreaticoduodenectomy remains one of the most complex gastrointestinal surgeries and is associated with up to 20–30% morbidity and 1–3% mortality [1, 2]. The postoperative outcomes are affected even more when PD is performed in patients with a background of cirrhosis. The associated mortality in this setting can be up to 30% [16]. This forms a big dilemma in decision-making between considering the risks of a PD in cirrhosis versus mortality associated with the primary malignancy and its effect on the patient’s quality of life. One of the initial reports by Warnick et al. reported outcomes in elective PD and compared with non-cirrhotic patients. PD was found to have a significantly higher incidence of complications (47% vs. 22%; P = 0.035) and reoperations (34% vs. 12%; P = 0.03) [17]. Another study by Nakeeb et al. reported higher blood loss and transfusion requirements in cirrhosis. They also reported higher POPF, PPH, hospital stay, and wound complications in them [1]. In a recent case series by Ray et al. from India, they reported outcomes in 4 patients, of them morbidity was 50%, and mortality was seen in 1 (25%) patient and another patient died after 3 months due to liver decompensation [16].

The majority of the studies have reported poor outcomes after PD in cirrhotic patients. Hence it is crucial to have a meticulous case selection while offering PD to patients where morbidity and mortality can be kept to the minimum. We studied the factors predicting mortality and we found that preoperative thrombocytopenia (less than 1.37 lakhs/mm2), INR >1.31 and thrombocytopenia (< 1.23 lakhs/mm2) on POD1 were significantly associated with mortality. Other factors like preoperative history of pancreatitis, cholangitis, and stent block were also significantly associated with mortality. When evaluating postoperative morbidity, no specific factor showed a significant association. In our study, we defined significant morbidity as Clavien–Dindo grade >3. This category included a wide range of complications such as postoperative pancreatic fistula (POPF), bile leak, ascites, chyle leak, and delayed gastric emptying (DGE). The diverse nature of these complications likely explains why no single factor emerged as a predictor of morbidity. The resulting finding of preoperative thrombocytopenia and raised INR are indirectly suggestive of underlying clinically significant portal hypertension and severity of cirrhosis; which is responsible for poor outcomes in these patients. Nakeeb et al. had identified portal hypertension as a significant risk in patients who underwent PD [1]. At our institute, all patients who are known cirrhotic undergo evaluation with endoscopy to rule out the presence of varices which indirectly suggest clinically significant portal hypertension.

In the past, portal hypertension was considered a contraindication for abdominal surgery in all cirrhotic patients due to the risk of perioperative and postoperative complications. Assessment of portal hypertension is an important component of preoperative risk stratification in patients with cirrhosis undergoing non-hepatic abdominal surgery. Hepatic venous pressure gradient (HVPG) remains the gold standard for directly estimating portal pressure. Previous studies have demonstrated that an HVPG value >16 mmHg is associated with a significantly increased risk of postoperative complications and mortality following extrahepatic surgery, with values ≥ 20 mmHg identifying patients at particularly high risk [18]. However, the invasive nature of HVPG measurement limits its routine use before surgery. The Baveno VII consensus recommends the use of non-invasive surrogates, such as liver stiffness measurement (LSM) by elastography or FibroScan, to more accurately estimate the degree of portal hypertension and predict clinically significant outcomes [19]. In addition, composite risk models, such as the VOCAL-Penn score, offer an objective quantification of surgical risk and outperform traditional scores, like MELD or Child–Pugh [20]. Lahat et al. reported that preoperative TIPS was able to significantly reduce the hepatic venous pressure gradient (HVPG) in patients with portal hypertension planned for non-hepatic surgery [21]. However, it must be noted that preoperative TIPS would reduce complications that may arise due to severe portal hypertension, but would not eliminate the risks due to existing cirrhosis. Zamorano et al. reported outcomes of PD in cirrhotic patients in 16 patients and reported 0 mortality and 37.5% morbidity rates. Preoperative TIPS was performed in 2 patients with favorable outcomes [22]. However, a recent meta-analysis concluded that liver cirrhosis of any severity with or without portal hypertension is associated with 4 4-fold increase in mortality and a 2-fold increase in morbidity [23].

A comparison with historical data is depicted in Table 6.

Table 6.

Literature review and comparison with the present study

Publication, Year, Country	Study design	Number of patients	Morbidity	Mortality
Warnick et al., 2011, Germany [17]	Retrospective	N = 32(2 Child B)	47% morbidity	3% Mortality in child A and 100% in child B
Nakeeb et al., 2013, Egypt [1]	Retrospective	N = 67 (4 Child B)	20% morbidity	9.5% (Child A) vs. 50% in Child B
Regimbeau et al., 2015, France [24]	Retrospective multicentre	N = 35(11 Child B)	79% morbidity and 100% for child B,	Mortality 4%(Child A)vs.66% (Child B)
Busquets et al., 2016, Spain [25]	Retrospective	N = 15, 0 child B	73% morbidity	0% mortality
Ray et al., 2021, India [16]	Retrospective	N = 4, 1 Child B	50% morbidity	25% mortality, 100% in Child B
Butler et al. , 2019, USA [26]	Metaanalysis	N = 36, Child A-30, Child B-5, Child C-1	Morbidity Child A-63.3% Child B/C-50%	Mortality Child A-16.6% Child B-16.6%
Zamorano et al., 2023, USA [22]	Retrospective	N = 16	Morbidity-37.5%	Mortality Nil
Nevarez et al. , 2024, USA [27]	Retrospective national database study	N = 203	Longer hospital stay. Morbidity not mentioned.	11.3% mortality
Hajibandeh et al., 2024, UK [23]	Metaanalysis	11 studies included, N = 1001	LC of any severity with or without portal hypertension results in at least a fourfold increase in mortality and a twofold increase in morbidity after pancreatoduodenectomy
Shrikhande et al. [4]	Retrospective	N = 4 who underwent Whipple procedure	Mortality was 1(25%).
Present study, 2024, India	Retrospective study	N = 24,All Child A	50% Morbidity and 12.5% mortality

Limitations of the study include its retrospective nature and a small sample size. However, this study gives a strong message to look into variables pointing towards severity of portal hypertension while planning a major surgical resection like PD in cirrhotic patients.

Conclusion

Performing PD in the background of liver cirrhosis is challenging and extensive preoperative evaluation is crucial in achieving favorable outcomes. Patients with clinically significant portal hypertension are associated with a high risk of surgical mortality. This could be considered as a subset of high-risk population and warrants further evaluation/optimisation before planning surgery.

Author contributions

Conceptualization; KN, PK, AC Data curation; KN, MSB, Formal analysis; KN, MSB; Methodology; MSB, VC, SVS Roles/Writing - original draft; KN, MSB Writing - review & editing: VO, AR, VC, SVS.

Funding

Open access funding provided by Department of Atomic Energy. The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

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

Declarations

Ethics approval

This is an observational study hence no ethical approval is required.

Competing interests

The authors declare no competing interests.