Pancreatic texture is influenced by pancreatic fibrosis and fat content [1]. A hard pancreas results from increased fibrosis and reduced fat infiltration [1]. In pancreatic surgeries such as pancreaticoduodenectomy, a soft pancreas is associated with a higher incidence of complications, including pancreatic fistulas or leaks, because it has inferior suture retention and is more susceptible to ischemia [2,3]. When a soft pancreas is encountered intraoperatively, management strategies vary widely. Some advocate for a completion pancreatectomy, while others dispute the success of alternative pancreatic anastomoses including pancreaticogastrostomy or externalization of the pancreatic duct [4-7]. For patients undergoing pancreaticoduodenectomy, the reported rate of postoperative pancreatic fistulas (POPFs) is as high as 27%, significantly contributing to morbidity and mortality [8].

Some centers use imaging-based predictors of pancreatic texture; however, pancreatic texture is typically determined intraoperatively through palpation [9,10]. Factors such as pancreatic texture, duct size, and blood loss contribute to the development of POPF risk scores; however, these scores are only applicable postoperatively and do not aid in preoperative decision-making [11,12]. Certain patient factors, such as a soft pancreatic texture and the development of POPF, are associated with diabetes. Patients with diabetes tend to have less fatty pancreatic textures and more fibrosis, suggesting that diabetes may offer protection against POPF development [1]. Conversely, hypertension is associated with increased POPF rates, which is not well understood; it is unclear whether hypertension itself or the use of antihypertensives contributes to poor anastomotic healing [1]. To date, few studies have investigated whether specific preoperative patient characteristics, beyond imaging findings, are associated with a soft pancreas.

The association between soft pancreatic texture and POPF is well established, and calculators like the Fistula Risk Score (FRS) exist for this risk. Our goal was to examine all postoperative complications resulting from a soft pancreas. Moreover, we aimed to identify patient characteristics that could predict soft pancreatic texture preoperatively, as identifying such patients before surgery could significantly enhance risk stratification and improve surgical outcomes. Using these factors, we developed and validated a clinically applicable risk score to preoperatively identify patients likely to have soft or hard pancreatic textures. While we cannot yet control for intraoperative pancreatic texture, we hope our findings may enable preoperative stratification of risks for postoperative complications.

Data source

Data was collected from the 2016 to 2021 American College of Surgeons National Surgical Quality Improvement (ACS NSQIP) database, which captures information from the preoperative period up to 30 days postoperatively. Our query focused exclusively on patients undergoing pancreaticoduodenectomy. Data was gathered across 874 hospitals in Canada and the United States. The study was exempt from ethics approval. ACS NSQIP is considered a publicly available database with de-identified patient data. As such, and given the retrospective nature of this study without patient identifiers, this study was exempt from formal institutional ethics review.

Study design and outcomes

A retrospective cohort study analyzed prospectively collected data from the ACS NSQIP database. Our primary aim was to assess the impact of pancreatic texture on postoperative outcomes following pancreaticoduodenectomy. Our secondary outcomes investigated whether preoperative patient factors could predict a soft pancreatic texture.

Patient population and variables definitions

Our study included 17,706 patients who underwent pancreaticoduodenectomy between 2016 and 2021 in Canada or the United States, categorized by pancreatic texture into hard/intermediate and soft groups.

We compared patient characteristics between the two cohorts, including age, sex, body mass index (BMI), functional status, and American Society of Anesthesiology (ASA) category. Additionally, we compared patient comorbidities such as severe chronic obstructive pulmonary disorder (COPD), congestive heart failure, hypertension, diabetes, smoking, chronic steroid use, bleeding disorders, hypoalbuminemia, and dialysis dependence. We also examined the effects of preoperative pancreatic stenting and neoadjuvant therapy (chemotherapy or radiation therapy), the reasons for surgery, and in cases of malignancy, the type and location of neoplasm. Two preoperative variables, sepsis and hypoalbuminemia, were compared to assess clinical stability before surgery. Finally, we looked at operative time and length of hospital stay.

To assess postoperative morbidity and mortality, we compared 30-day postoperative outcomes, examining serious complications such as surgical and deep wound infections, wound dehiscence, pneumonia, intubation, pulmonary embolism, acute kidney injury, urinary tract infections, cerebral vascular events, cardiac events, bleeding, and sepsis. Additionally, we evaluated rates of reoperation, chronic hospital stays exceeding 30 days, readmission, and development of POPF. Statistically and clinically significant variables were highlighted in our analysis. Demographic and outcome definitions were based on NSQIP participant user data files [13]. Data on POPF and serious postoperative complications were also collected. POPF was defined according to the 2005 International Study Group on Pancreatic Surgery definition and classification, including Grade A (biochemical leak only), Grade B (leak requiring percutaneous or endoscopic drain placement), and Grade C (leak necessitating reoperation) [14].

Statistical analysis

Statistical analysis was conducted using STATA 17 statistical software (STATA Corp LP). Discrete variables are presented as absolute values (%) and were analyzed using the chi-squared test. Continuous variables are reported as weighted means with standard deviations and analyzed using the ANOVA test. A p-value<0.05 was considered significant.

Two multivariable models were constructed to assess the association between pancreatic texture and serious complications and mortality. An additional model investigated preoperative patient factors associated with a soft pancreas. These models involved univariate analysis of both clinically relevant variables and those with p<0.10, leading to a preliminary main effects logistic regression model. The Wald test determined significant variables (p<0.05) influencing the models. Collinearity was assessed using the variance inflation factor, with values greater than 10 prompting exclusion from the models. Model accuracy was evaluated using the receiver operating characteristic (ROC) and Brier score (BS).

Our secondary analysis developed a tool, termed the preoperative soft pancreas risk score (PSPRS), to predict pancreatic texture using preoperatively identifiable factors. Our study population was equally divided into derivation and validation cohorts. In the derivation cohort, we identified patients with a soft pancreas and compared them with those having an intermediate/hard pancreas. A multivariable analysis that evaluated factors associated with hard pancreatic texture excluded variables unidentifiable preoperatively, such as the minimally invasive surgical approach. The model aimed to predict hard pancreatic texture, achieving a scoring system with primarily positive values (i.e. +1). The coefficients' ratios from the derivation model were applied to develop the scoring model. The scoring threshold was set at three standard deviations above the mean score of the low-risk group, enabling the model to identify patients with a hard pancreas with high specificity. The PSPRS was then used on the validation cohort to assess the score's accuracy in identifying pancreatic texture. We calculated the area under the receiver operating characteristic curve (AUROC) for both the derivation and validation datasets. Additionally, we analyzed the area under the curve of the receiver operating characteristic curves (AUC-ROC) to determine the optimal PSPRS cut-off value for detecting a hard pancreas. A non-parametric estimation of the AUC-ROC, using Bamber and Hanley confidence intervals, helped determine optimal cut-off levels employing the Liu method [15].

Patient demographics

Our analysis included 17,706 patients, with 9,686 (54.7%) classified in the hard/intermediate pancreas cohort and 8,020 (45.3%) in the soft pancreas cohort. Demographic similarities were noted between the groups. The mean age was 65.9 ± 10.6 years in the hard pancreas group and 65.0 ± 12.3 years in the soft pancreas group; this difference was statistically significant (p < 0.001) but clinically negligible. In the soft pancreas cohort, there were more females (48.3% vs 45.4%; p < 0.001) and fewer males (51.8% vs 54.6%; p < 0.001) compared to the hard pancreas cohort. The soft pancreas group also had a significantly higher mean BMI (28.0 ± 6.0 vs 26.9 ± 5.5; p < 0.001) and more ASA II patients, but fewer ASA III or IV patients, than the hard pancreas group (Table 1).

Table 1 . Patient characteristics

	Hard pancreas (n = 9,686)	Soft pancreas (n = 8,020)	p-value
Age (yr)	65.9 ± 10.6	65.0 ± 12.3	< 0.001
Sex			< 0.001
Female	4,399 (45.4)	3,870 (48.3)
Male	5,287 (54.6)	4,150 (51.8)
BMI (kg/m2)	26.9 ± 5.5	28.0 ± 6.0	< 0.001
ASA			< 0.001
1	21 (0.2)	32 (0.4)
2	1,572 (16.2)	1,686 (21.0)
3	7,340 (75.8)	5,750 (71.8)
4	752 (7.8)	544 (6.8)
5	0 (0)	1 (0.0)
COPD	375 (3.9)	321 (4.0)	0.655
CHF	64 (0.7)	59 (0.7)	0.550
Hypertension	5,249 (54.2)	4,150 (51.8)	0.001
Diabetes			< 0.001
Insulin-dependent	1,704 (17.6)	707 (8.8)
Non-insulin dependent	1,374 (14.2)	1,036 (12.9)
Smoker	1,755 (18.1)	1,113 (13.9)	< 0.001
Dialysis	29 (0.3)	28 (0.4)	0.561
Chronic steroid use	292 (3.0)	272 (3.4)	0.155
Bleeding disorder	294 (3.0)	194 (2.4)	0.013
Hypoalbuminemia	2,119 (21.9)	1,427 (17.8)	< 0.001
Neoadjuvant	3,511 (36.2)	1,276 (15.9)	< 0.001
Preoperative Sepsis			0.598
SIRS	81 (0.8)	75 (0.9)
Sepsis	38 (0.4)	26 (0.3)
Functional health status prior to Surgery			0.034
Independent	9,595 (99.1)	7,951 (99.1)
Partly dependent	82 (0.9)	52 (0.7)
Totally dependent	3 (0.0)	11 (0.1)
Preoperative biliary stent			< 0.001
No stent at time of surgery	3,373 (34.8)	4,087 (51.0)
Endoscopic stent	5,519 (57.0)	3,207 (40.0)
Percutaneous stent	315 (3.3)	167 (2.1)
Stent of other or unknown type	479 (5.0)	559 (7.0)
Diagnosis			< 0.001
PDAC	6,693 (69.1)	2,653 (33.1)
Pancreatitis	397 (4.1)	81 (1.0)
Carcinoid & neuroendocrine tumor	480 (5.0)	1,083 (13.5)
Duodenal and ampullary lesions	877 (9.1)	1,955 (24.4)
Cyst	77 (0.8)	143 (1.8)
Cholangiocarcinoma	337 (3.5)	632 (7.9)
Other	825 (8.5)	1,473 (18.4)

Values are presented as mean ± standard deviation or number (%).

BMI, body mass index; ASA, American Society of Anesthesiologists; COPD, chronic obstructive pulmonary disease; SIRS, systemic inflammatory response syndrome; PDAC, pancreatic ductal adenocarcinoma.

Patients in the soft pancreas group were less likely to require hypertension medication (51.8% vs. 54.2%; p = 0.001) or have diabetes (21.7% vs. 31.8%; p < 0.001), specifically insulin-dependent diabetes (8.8% vs. 17.6%; p < 0.001). They were also less likely to be smokers (13.9% vs. 18.1%; p < 0.001), have hypoalbuminemia at the time of surgery (18.4% vs. 22.8%; p <0.001), or have had a preoperative pancreatic stent (49.1% vs. 65.3%; p < 0.001). Additionally, the diagnoses and reasons for surgery differed significantly between the two cohorts. Patients in the soft pancreas group were less likely to have pancreatic ductal adenocarcinoma (PDAC) (33.1% vs. 69.1%; p < 0.001) or pancreatitis (1.0% vs. 4.1%; p < 0.001), but more likely to have carcinoid and neuroendocrine tumors (13.5% vs. 5.0%; p < 0.001), duodenal and ampullary lesions (24.4% vs. 9.1%; p < 0.001), and cholangiocarcinoma (7.9% vs. 3.5%; p < 0.001) compared to the hard pancreas cohort.

Bivariate analysis of postoperative outcomes comparing patients with a hard pancreas to those with a soft pancreas

Patients in the soft pancreas group experienced a shorter mean operative time (370.1 ± 117.3 vs. 388.3 ± 134.4 minutes; p < 0.001) than those in the hard pancreas group. However, the soft pancreas group had longer hospital stays from the day of surgery to discharge (10.1 ± 6.4 vs. 8.8 ± 5.4 days; p < 0 .001).

Overall, the soft pancreas group exhibited a higher rate of serious complications compared to the hard pancreas group (56.5% vs. 42.2%; p < 0.001). This group experienced a greater incidence of pneumonia, reintubation, pulmonary embolism, acute kidney injury, urinary tract infection, cardiac arrest, sepsis, prolonged stay, and readmission (Table 2). Remarkably, the rate of POPF was significantly higher in the soft pancreas group (32.4% vs. 10.3%; p < 0.001), with 69.2% of fistulas classified as Grade B. Conversely, the hard pancreas group exhibited a higher incidence of bleeding requiring transfusion (19.3% vs. 14.1%; p < 0.001). A higher rate of organ space infection was noted in the soft pancreas group (24.9% vs. 11.3%; p < 0.001), but there were no significant differences in rates of superficial surgical site infection, deep incisional infection, or wound dehiscence between the two groups. Notably, there was no difference in mortality rates.

Table 2 . Outcomes thirty days postoperatively

	Hard Pancreas (n = 9,686)	Soft Pancreas (n = 8,020)	p-value
Operative time (min)	388.3 ± 134.4	370.1 ± 117.3	< 0.001
Length of stay (days)a)	8.8 ± 5.4	10.1 ± 6.4	< 0.001
Superficial surgical site infection	582 (6.0)	523 (6.5)	0.160
Deep incisional infection	67 (0.7)	67 (0.8)	0.272
Organ space infection	1,090 (11.3)	2,000 (24.9)	< 0.001
Wound dehiscence	103 (1.1)	91 (1.1)	0.650
Pneumonia	278 (2.9)	329 (4.1)	< 0.001
Reintubation	226 (2.3)	327 (4.1)	< 0.001
Pulmonary embolism	84 (0.9)	132 (1.7)	< 0.001
Acute kidney injury	70 (0.7)	115 (1.4)	< 0.001
Urinary tract infection	183 (1.9)	221 (2.8)	< 0.001
Cerebral vascular accidents	30 (0.3)	16 (0.2)	0.152
Cardiac arrest	97 (1.0)	115 (1.4)	0.008
Myocardial infarction	113 (1.2)	90 (1.1)	0.782
Bleed, requiring transfusion	1,865 (19.3)	1,128 (14.1)	< 0.001
DVT/thrombophlebitis	253 (2.6)	256 (3.2)	0.021
Sepsis	732 (7.6)	814 (10.2)	< 0.001
Septic shock	198 (2.0)	310 (3.9)	< 0.001
Unplanned reoperation	406 (4.2)	499 (6.2)	< 0.001
Length of stay > 30 days	196 (2.0)	340 (4.2)	< 0.001
Readmission	1,453 (15.0)	1,661 (20.7)	< 0.001
POPF	995 (10.3)	2,576 (32.4)	< 0.001
Grade A	284 (3.0)	651 (8.2)	< 0.001
Grade B	639 (6.6)	1,783 (22.2)
Grade C	72 (0.8)	142 (1.8)
Serious complication	4,088 (42.2)	4,529 (56.5)	< 0.001
Death	143 (1.5)	143 (1.8)	0.107

Values are presented as mean ± standard deviation or number (%).

DVT, deep vein thrombosis; POPF, postoperative pancreatic fistula.

^a)From day of surgery to discharge.

Multivariable logistic regression evaluating predictors of serious postoperative complications, mortality, and soft pancreas

Multivariable logistic regression revealed that a soft pancreas was independently associated with serious complications, even after adjusting for demographic factors (odds ratio [OR]: 1.60, confidence interval [CI]: 1.49–1.71, p < 0.001). Other significant predictors of serious complications (Table 3) included preoperative dialysis dependence (OR: 3.05, CI: 1.54–6.04, p = 0.001), congestive heart failure (OR: 1.85, CI: 1.24–2.78, p = 0.003), preoperative sepsis (OR: 2.12, CI: 1.56–2.87, p < 0.001), partially dependent functional status (OR: 1.72, CI: 1.17–2.52, p = 0.006), and COPD (OR: 1.59, CI: 1.35–1.88, p < 0.001). Additional factors linked to serious complications included hypoalbuminemia, steroid use, bleeding, hypertension, diabetes, tumor invasion, and diagnosis as outlined in Table 3. Female sex (OR: 0.90, CI: 0.84–0.96, p = 0.001) was associated with a reduced risk of postoperative complications. PDAC was independently associated with a decreased risk of serious complications (OR: 0.78, CI: 0.70–0.86, p < 0.001). The multivariable model to predict serious complications had an ROC of 0.622 and BS of 0.239.

Table 3 . Multivariable logistic regression evaluating predictors of serious complications

Risk factor	Odds ratio	95% confidence interval	p-value
Soft pancreas (compared to hard)	1.60	1.49–1.71	< 0.001
Age	1.00	0.99–1.00	0.862
BMI	1.02	1.01–1.02	< 0.001
Female sex	0.90	0.84–0.96	0.001
COPD	1.59	1.35–1.88	< 0.001
CHF	1.85	1.24–2.78	0.003
HTN	1.17	1.09–1.25	< 0.001
Diabetes
Insulin-dependent	0.97	0.88–1.07	0.543
Non-insulin dependent	1.02	0.93–1.12	0.678
Active smoker	0.88	0.81–0.96	0.005
Dialysis	3.05	1.54–6.04	0.001
Steroid	1.33	1.11–1.60	0.002
Bleeding disorder	1.31	1.08–1.58	0.007
Preoperative sepsis	2.12	1.56–2.87	< 0.001
Partially dependent functional status (compared to independent)	1.72	1.17–2.52	0.006
Neoadjuvant therapy	1.02	0.94–1.10	0.706
Invasion	1.14	1.06–1.22	< 0.001
Hypoalbuminemia	1.45	1.34–1.56	< 0.001
Diagnosis
PDAC	0.78	0.70–0.86	< 0.001
Pancreatitis	1.07	0.86–1.32	0.536
Carcinoid & neuroendocrine tumor	0.98	0.86–1.12	0.786
Duodenal and ampullary lesions	1.07	0.95–1.21	0.263
Cyst	1.07	0.80–1.42	0.659
Cholangiocarcinoma	1.19	1.01–1.40	0.033

BMI, body mass index; COPD, chronic obstructive pulmonary disease; CHF, congestive heart failure; HTN, hypertension; PDAC, pancreatic ductal adenocarcinoma.

Multivariable modeling showed that a soft pancreas was not independently associated with mortality (Supplementary Table 1). The most significant predictors of mortality included congestive heart failure (OR: 2.43, CI: 1.04–5.71, p = 0.041), preoperative bleeding disorders (OR: 2.02, CI: 1.20–3.42, p = 0.008), COPD (OR: 1.74, CI: 1.09–2.77, p = 0.041), and hypoalbuminemia (OR: 1.44, CI: 1.10–1.90, p = 0.009). PDAC was independently associated with a reduced mortality risk (OR: 0.66, CI: 0.45–0.97, p = 0.033) as indicated in Supplementary Table 1. The multivariable model for mortality prediction had an ROC of 0.667 and BS of 0.016.

Scoring model analysis predicting hard pancreas and protecting against soft pancreas

A multivariable logistic regression was conducted to assess preoperative demographics potentially predictive of a soft pancreas (Supplementary Table 2). Female sex (OR: 1.14, CI: 1.06–1.22, p < 0.001) and increased BMI (OR: 1.12, CI: 1.09–1.16, p < 0.001) were independently associated with a soft pancreas. Among diagnostic groups, duodenal and ampullary lesions (OR: 1.48, CI: 1.30–1.69, p < 0.001) and cholangiocarcinoma (OR: 1.48, CI: 1.24–1.76, p < 0.001) were independently linked to a soft pancreas, while PDAC (OR: 0.32, CI: 0.28–0.35, p < 0.001) and pancreatitis (OR: 0.12, CI: 0.09–0.16, p < 0.001) showed a reduced likelihood of a soft pancreas. Additionally, neoadjuvant treatment, insulin-dependent diabetes, non-insulin-dependent diabetes, active smoking, hypoalbuminemia, and preoperative biliary stenting were significantly associated with a reduced likelihood of a soft pancreas.

Using the findings from the multivariable logistic models in our validation cohort (ROC = 0.750), we generated a risk calculator to preoperatively predict a patient having a hard pancreas (or, being protected against having a soft pancreas). Point assignments for the PSPRS are displayed in Table 4. Reevaluating our model in the validation cohort confirmed similar factors associated with a hard pancreas (ROC = 0.750, Supplementary Fig. 1). Upon implementation of the PSPRS in our validation cohort, ROC analysis of an optimal cut-off found that a score of ≥6 is 68.9% specific and 71.7% sensitive for predicting a hard pancreas, with 41.7% of patients achieving this score. Conversely, lower scores increased the sensitivity of the PSPRS, while higher scores enhanced the specificity (Table 5).

Table 4 . Risk calculator for having a hard pancreas (protecting against soft pancreas)

Risk factor	Point
BMIa)	–1
Female sex	–1
Hypoalbuminemia (< 40 g/L)	+2
Preoperative biliary stent	+5
Neoadjuvant therapy	+4
Active smoker	+2
Insulin-dependent diabetes	+6
Non-insulin dependent diabetes	+2
Invasion	+1
PDAC (reason for surgery)	+11
Pancreatitis (reason for surgery)	+23
Duodenal tumor (reason for surgery)	–4

BMI, body mass index; PDAC, pancreatic ductal adenocarcinoma.

^a)1 point subtracted from total score for every 5 points on the BMI scale. Example, a BMI of 35 kg/m² would subtract 7 points from the total score.

Table 5 . Sensitivity and specificity of various preoperative soft pancreas risk scores for determining a hard pancreatic texture evaluated in the validation cohort of this study

Score cut-off	Sensitivity (%)	Specificity (%)	Proportion of patients with this score (%)
≥–9	99.5	4.0	81.5
≥–5	94.6	23.7	72.1
≥0	82.9	51.7	65.1
≥2	79.2	59.1	59.7
≥4	75.9	64.1	49.7
≥6a)	71.7	68.9	41.7
≥8	65.6	73.3	29.5

Receiver operating characteristic: 0.750.

^a)Optimal cut-off value determined by the Liu method.

This retrospective study analyzed postoperative outcomes from 17,706 patients who underwent a pancreaticoduodenectomy between 2016 and 2021 in Canada and the United States. As previously shown, we determined that a soft pancreas is independently associated with serious complications. We identified several protective factors that mitigate the likelihood of having a soft pancreas and can assist surgeons in preoperative risk stratification. These factors were utilized to generate a clinical preoperative risk calculator that accurately predicts pancreatic texture and may assist in counseling patients.

Significant progress has been made to enhance pancreatic imaging techniques and develop radiologic protocols for assessing pancreatic texture. Promising advances include the development of the MINIMAP protocol for MRI assessment of pancreatic texture in chronic pancreatitis [16]. However, these studies require larger sample sizes to correlate radiologic findings with histology, and to consider patient factors such as age, sex, BMI, and ethnicity [16]. Expanding sample sizes is also crucial for training AI models on texture and fibrosis, since these parameters can vary widely even within an individual pancreas [16].

These predictive scoring models typically focus on POPF risk assessment and build upon the FRS [17]. Although POPF is the most common postoperative complication in a soft pancreas, our findings indicate that a soft pancreas is associated with a high risk of other postoperative complications beyond POPF. The soft pancreatic texture is linked to increased leaks, reduced suture-holding capacity, and a higher risk of infection. A scoring system that preoperatively identifies these patients, and one that is patient-centered and considers individual patient demographics, would enhance FRS and enable more comprehensive preoperative counseling by addressing all potential complications—especially respiratory and cardiac.

In our study, patients with a soft pancreas had significantly higher rates of most postoperative complications, particularly a three times higher rate of POPF, and at least a two times higher rate of acute kidney injury, deep organ space infection, septic shock, and prolonged length of stay over 30 days. These complications may be due to higher secretion of pancreatic juices and proteolytic enzymes from a soft pancreas, increased ischemia secondary to suture injury as noted by previous studies, and the technical challenges of suture placement [2,3,18,19]. These POPF-related complications likely contribute to other postoperative complications in patients with soft pancreas, including higher rates of pulmonary complications, deep vein thrombosis, urinary tract infection, cardiac arrest, reoperation, and readmission. Although the link between soft pancreas and complications is well established, few studies have associated a hard pancreas with increased bleeding as demonstrated by our data. This association might be due to higher rates of preoperative stenting (indicative of biliary obstruction) and neoadjuvant therapy in patients with hard pancreas, potentially predisposing them to fibrosis and complicating dissection [20,21]. It may also be attributed to the diagnosis or tumor location itself, as there were almost three times more PDAC cases in the hard pancreas group than in the soft pancreas cohort. Future studies investigating these correlations would be valuable. Additional independent factors contributing to serious postoperative complications included dialysis, preoperative sepsis, and congestive heart failure, with heart failure being the strongest independent predictor of mortality. Although these factors do not predict a soft pancreas, their strong link with serious complications is crucial in counseling patients and emphasizes the need for a preoperative cardiology and anesthesia assessment and the optimization of cardiac risk factors.

The ability to preoperatively identify patients at risk for a soft pancreas can guide decision-making not only on the type of resection performed but also on the specific anastomosis used and can provide an overall risk assessment. The two main predictors of a soft pancreas in our study were female sex and increased BMI. Thus, it is unsurprising that several studies have demonstrated an association between adiposity and POPF [22-24]. Contrastingly, factors such as smoking, diabetes, preoperative biliary stenting, and neoadjuvant therapy decreased the likelihood of a soft pancreas; these factors have all previously been linked to pancreatic fibrosis [20,21,25-27]. The association between female sex and soft pancreas appears to contradict the findings of Mungroop et al. [28], which associated male sex with POPF in the adapted FRS. In the adapted FRS, where the calculator accounts for pancreatic texture, male sex was independently associated with fistula formation. However, in our PSPRS calculator, where each variable is assessed independently, female sex was identified as a strong predictor of soft pancreatic texture. This may be related to the smaller main pancreatic duct diameter in female patients [29], which has previously been associated with a soft pancreatic texture [30], but it is difficult to determine from the NSQIP database alone. In addition to these patient factors, certain diagnoses and tumor locations were also significant. PDAC and pancreatitis were strongly associated with a hard pancreas, whereas duodenal and ampullary lesions, as well as cholangiocarcinoma, were associated with a soft pancreas.

Considering all this, the PSPRS was able to discriminate between patients likely to have a hard pancreas and those likely to have a soft pancreas with reasonable confidence. The PSPRS cut-off evaluation demonstrated that a score ≥6 would categorize >40% of preoperative patients with a hard pancreas at 68.9% specificity and 71.7% sensitivity. The specificity of PSPRS exceeds that of accepted FRS and adapted FRS scoring systems (ranging from 34% to 51.5%), although its sensitivity is lower than comparable calculators (ranging from 87% to 100%) [31]. Neither these accepted calculators nor PSPRS alone should dictate surgical practice. However, we hope PSPRS will contribute to preoperative discussions on potential risks. For those patients with a low PSPRS, a preoperative discussion should emphasize the higher risks of serious postoperative complications including longer lengths of stay, organ space infection, reintubation, acute kidney injury, sepsis, POPF, and readmission or reoperation. These patients should also be advised about the potential risk of converting to a total pancreatectomy intraoperatively if the pancreatic texture does not permit a safe anastomosis [4,32-35]. The calculator could also contribute to improved surgical outcomes. For example, those patients with low PSPRS and PDAC, particularly those who fall into a higher surgical risk category due to age or comorbidities, could be considered for neoadjuvant therapy to induce pancreatic fibrosis and reduce the risk of POPF [36,37]. Furthermore, by predicting pancreatic texture, PSPRS could also aid in identifying patients at risk of postoperative leaks and in determining those who might benefit from prophylactic somatostatin analogs, especially preoperative octreotide, which has been shown to reduce secretion of pancreatic enzymes and promote healing at the anastomosis [38].

The outcomes of this study should be considered in the context of its limitations. This study was constrained by its retrospective nature, and our analysis involved considering patient factors individually in relation to pancreatic texture and postoperative morbidity. The NSQIP aggregates data from a wide geographic area and numerous centers, lacking detailed data on local practices, treatments, and demographic differences that might affect patient outcomes and surgical approaches. Moreover, the differentiation between soft or hard pancreatic texture remains subjective, dependent on the surgeon's judgement and not founded on histopathology. Importantly, the size of the pancreatic duct was excluded from the PSPRS. Although present in the NSQIP database, its intraoperative measurement renders it less useful for preoperative risk assessment. However, future iterations of the calculator could be enhanced by incorporating radiological criteria. Additionally, the calculator's robustness could be further improved by external validation and using a prospectively collected dataset. Furthermore, neoadjuvant therapy, identified as protective against a soft pancreas, was generically described as including chemotherapy and radiation. Variations might exist between these treatments. Further research is crucial to determine if specific combinations of factors contribute to a soft pancreatic texture.

Despite these limitations, our findings indicate high rates of postoperative complications in patients with soft pancreatic texture, and that having a soft pancreas independently correlates with serious complications, irrespective of POPF. Our analyses confirm that female gender, higher BMI, and cholangiocarcinoma are independently linked to a soft pancreas, whereas neoadjuvant therapy, diabetes, preoperative biliary stenting, smoking, and diagnoses of PDAC or pancreatitis are associated with a hard pancreas. Utilizing these findings, we have developed a clinically relevant risk scoring calculator to predict hard pancreatic texture or identify patients less likely to have a soft pancreas. We aim for these results to aid in preoperative risk assessment, patient counseling, and surgical planning.

Supplementary data related to this article can be found at https://doi.org/10.14701/ahbps.24-172.

American College of Surgeons National Surgical Quality Improvement Program and the hospitals involved in the ACS NSQIP are sources of the data utilized in this study; they have neither verified nor bear responsibility for the statistical validity of the data analysis or the conclusions drawn by the authors. This research was exempt from ethics review.

None.

No potential conflict of interest relevant to this article was reported.

Conceptualization: ZC, KV, BA, KD, DB, JS. Data curation: KV. Methodology: ZC, KV, BA, KD, DB, JS. Writing - original draft: ZC, KV. Writing - review & editing: ZC, KV, BA, KD, DB, JS.