Impact of visceral adiposity on severity of acute pancreatitis: a propensity score-matched analysis

We analyzed 1662 consecutive hospitalized patients with AP between August 1st, 2010 and August 31st, 2017 at Ningbo First Hospital, China. This study was approved by the Human Ethics Committee. The clinical study registration number is NCT03482921. Severe AP (SAP) and moderately severe AP (MSAP) were defined in 111 patients according to the Revised Atlanta Classification of AP [18] (Fig. 1). Exclusion criteria for the study included age < 18 years, missing data in the electronic medical record, and history of AP. Nine unqualified patients were excluded. SPSS R plug-in (SPSS R Essentials) was applied for matching [19]. We used the SPSS “PS Matching” feature to perform propensity score-matched analysis. Matching factors include age, sex, and BMI. SAP, and MSAP patients were matched 1:2 in a multivariable logistic analysis using stepwise regression based on a greedy matching algorithm with a caliper of 0.2 times the standard deviation (SD) of the logit. After applying 1:2 propensity score matching, 102 eligible patients were matched to 204 patients with mild AP (MAP). In total, 306 hospitalized patients were enrolled in this study.

AP severity was defined using the Revised Atlanta Classification criteria. The clinical classification defined 3 grades of severity: MAP, MSAP, and SAP.

Systemic complication was defined as a score ≥ 2 in one of three assessed organ systems (respiratory, cardiovascular, or renal) according to the modified Marshall scoring system [20].

Local complications of AP were divided into: acute peri-pancreatic fluid collection (APFC), acute necrotic collection (ANC), pancreatic pseudocyst, and walled-off necrosis (WON). Diagnosis of local complication on unenhanced computed tomography (CT) was evaluated by two experienced radiologists who were blinded to detailed information about the cases. When necessary, follow-up contrast-enhanced CT (CECT) and magnetic resonance imaging (MRI) are performed based on the clinical status of patients, including persistence or recurrence of abdominal pain, increasing organ dysfunction, and development of clinical signs of sepsis. Once patients are diagnosed with APFC, ANC, WON, or pseudocyst in the follow-up imaging examinations, they are considered to present with local complications.

Creatinine, calcium, and albumin levels obtained within the first 24 h were collected for each patient. Additional variables included intensive care unit (ICU) requirement, length of hospital stay, and in-hospital mortality.

The APACHE-II, BISAP, and SIRS scores were calculated within the first 24 h of admission, and Ranson’s score was recorded within 48 h of admission.

A dual 16-slice CT scanner (SIMENS SOMATOM Sensation; Siemens, Germany) was used for entire / upper abdominal CT at 250 mA; tube voltage, 120 kV; data collection thickness 5 mm; reconstruction thickness 1 mm; and reconstruction interval 1 mm. CT scans were performed within 24 h after patient admission. Entire abdomen CT was performed to 165 patients. Upper abdomen CT was performed to 141 patients. Two consecutive, axial images on unenhanced CT at the L3/4 intervertebral space were retrospectively reviewed by two radiologists who were blinded to patient information. We calculated fat and muscle cross-sectional areas, which have been validated as the best proxies of SAT and VAT independent of age [21]. Skeletal muscle areas, including the psoas, paraspinal, and the abdominal wall muscles excluding the intra-abdominal visceral muscles, were also measured (Fig. 2). We used the Photoshop “magic wand” to outline and measure regions of similar signal intensity based on pixel counts. We manually outlined the margin of fat deposition manually in fat-tissue regions with poor contrast, which may be a more accurate way to measure compartment volumes. The Hounsfield units (HU) range for adipose tissue was − 190 to − 30 HU [22]. The mean value of the two images was calculated. The abdominal muscular wall distinguishing between VAT and SAT was traced automatically and adjusted manually [23, 24]. The VAT/SMT ratio was calculated by dividing VAT by SMT. All CT scans were analyzed using a commercially available software program (Photoshop CS6, Adobe Systems, CA) [25].

In this study, every MSAP and SAP patient was propensity score matched with another MAP patient at a ratio of 1:2. Finally, 306 hospitalized patients were enrolled for further analysis. Among the 306 patients, causative factors of AP included gallstone-related (142; 46.4%), idiopathic (91; 29.7%), hypertriglyceridemia (48; 15.7%), alcohol abuse (22; 7.2%), and others (3;1.0%). The mean age of the patients was 50.6 ± 18.7 years. The mean estimated BMI was 24.4 ± 4.5 kg/m². The mean SAT value was 131.1 ± 23.1 cm². The mean VAT value was 136.1 ± 27.3 cm². The mean SMT value was 138.2 ± 25.8 cm².

Age, sex, and BMI distribution were not statistically different among groups after propensity matching. There was a significant association between AP severity and individual fat parameters, except for SAT (Table 1). VAT and the VAT/SMT ratio were significantly higher in the MSAP and SAP groups compared to the MAP group (both p < 0.001). The mean SMT values decreased as AP severity increased (p < 0.001).

Based on the VAT value at the L3/4 intervertebral space, patients were separated into five groups. ICU transfer, AP severity, and systemic complications (APACHE-II scores ≥8, BISAP score ≥ 3, SIRS score ≥ 2) were significantly related to VAT values in the AP patients (Table 2).

Patients were also separated into five groups according to the VAT/SMT ratio. ICU transfer, AP severity, systemic complications, and prognostic scores (APACHE-II scores ≥8, Ranson’s score ≥ 3, BISAP score ≥ 3, SIRS score ≥ 2) had significant relationship with the VAT/SMT ratio (Table 3).

Further analyses were conducted for creatinine, calcium, and albumin levels using an adjusted model (Table 4). The multivariate adjusted hazard ratios (HRs) for VAT and the VAT/SMT ratio in the relationship between body parameters and AP mortality were 1.042 (95% confidence interval (CI), 1.019–1.066) and 7.820 (95% CI, 1.978–30.917), respectively. The VAT/SMT ratio was positively related to the incidence of AP mortality (p < 0.001).

The AUCs were measured to evaluate individual fat parameters of VAT and the VAT/SMT ratio for predicting the incidence of SAP and MSAP, ICU stay, and the incidence of WON (Fig. 3).

VAT had the highest area under the curve of ROC (0.943, 95% CI, 0.909–0.976) for predicting the incidence of SAP and MSAP, among independent predictors. The AUC of the VAT/SMT ratio was also significantly higher compared to the predictive scoring systems (0.896, 95% CI, 0.856–0.936). The optimal cut-off value of VAT for the prognosis of AP was 145.0. The sensitivity and specificity of this cut-off value were 90.2 and 93.1%, respectively. The optimal cut-off point of the VAT/SMT ratio for the prognosis of AP was 0.706. The sensitivity and specificity of this cut-off value were 80.4 and 90.2%, respectively. The cut-off values for VAT and the VAT/SMT ratio can be used to predict the incidence of SAP and MSAP.

We also demonstrated that VAT and the VAT/SMT ratio had high sensitivity and specificity for predicting the incidence of ICU stay and WON compared to other predictive scoring systems. The AUC of VAT for predicting the incidence of ICU stay was 0.900 (95% CI, 0.851–0.949). The second highest AUC of ROC was the VAT/SMT ratio (0.865, 95% CI, 0.810–0.921). The AUC of the VAT/SMT ratio was the highest predictor of the incidence WON (0.745, 95% CI, 0.628–0.863).