Sensitivity of Biochemical and Imaging Findings for the Diagnosis of Acute Pancreatitis in Children

Abstract

Objective:

To determine diagnostic sensitivity of serum biomarkers and imaging in diagnosis of acute pancreatitis (AP) in children.

Study design:

Cross-sectional analysis of prospective registry data for children (age <21-years) whose first documented attack of AP occurred between March 2013 - October 2016 at a single institution, tertiary care center. Main outcome was sensitivity of serum biomarkers and of imaging modalities, measured via descriptive statistics.

Results:

112 children met criteria for AP, 57 (51%) were male with a median age of 13.4 years (interquartile range: 9.3-15.8 years). Serum Amylase and lipase levels were obtained in 85 (76%) and 112 (100%) patients, respectively. Imaging was performed in 98 (88%) patients with abdominal ultrasound (US) performed in 84 (75%) and CT and/or MRI performed in 46 (41%) patients.

Fifty-three (47%) patients met all three diagnostic criteria (clinical, biochemical and imaging) for AP. Laboratory testing had a 5.4% false negative rate for AP. Serum lipase alone and amylase alone were 95% (95% CI: 89-98%) and 39% (95% CI: 28-50%) sensitive for AP, respectively. Imaging (any modality) was 61% sensitive (95% CI: 51-71%) for AP with a 34% false negative rate. US alone was 52% (95% CI: 41-63%) sensitive for AP and CT/MRI was 78% (95% CI: 63-89%) sensitive. Combinations of diagnostic criteria performed no better than laboratory testing alone.

Conclusions:

The majority of children coming to medical attention with their first documented occurrence of AP have characteristic symptoms. Serum lipase is highly sensitive for the diagnosis of AP, and serum amylase is moderately sensitive. Imaging, particularly US, is only moderately sensitive and cross-sectional imaging provides higher sensitivity for diagnosing AP.

Acute Pancreatitis is increasingly being recognized in children(1, 2). With an incidence of up to 13/100,000 per year and a median hospitalization time of 5 to 8.5 days, the economic burden associated with this diagnosis is significant(1–4). On average, the median estimated cost per admission for an episode of pediatric AP is $22,663, with costs reaching $43,000 for those admitted to the intensive care unit (ICU)(5, 6). Average total cost of admissions for pediatric AP approximate $200 million per year(7). Despite both the growing recognition of this disease and the cost of care, there remains little data on optimal diagnostic approaches for pediatric AP. To date, data and clinical practice are largely extrapolated from adult studies. This is problematic for multiple reasons, including the differing etiologies and risk factors for AP between adults and children: gallstones and alcohol being the predominant causes in adult, and biliary, metabolic, systemic, hereditary and anatomic abnormalities predominating in children(7, 8).

Presently, diagnosis of AP in children is made when a patient meets two out of three criteria: 1) biochemical (serum amylase or serum lipase ≥3× the upper limit of normal), 2) imaging (findings of pancreatic edema, fat stranding or peripancreatic fluid collection) and/or 3) clinical symptoms consistent with pancreatitis(9). Each of these diagnostic elements are understudied in children and mainly derived from retrospective data, with a broad range of results reported in the literature. For example, the sensitivity of serum lipase has been reported to be 77 to 100% and amylase sensitivity is 52 to 54% in pediatric studies(3, 4, 10). In adult studies, the specificity has ranged from 85-99% for serum lipase, and ~70% for amylase(11–13). Ultrasound has been shown to be 24 to 86% sensitive for AP and CT has been shown to be 47 to 75% sensitive, with few data regarding the diagnostic accuracy of MRI in pediatric AP (3, 4, 7, 10, 14). The purpose of our study was to leverage a novel prospectively collected database of children presenting with a first attack of AP to assess the sensitivity of each of the diagnostic criteria for AP in children.

Methods:

This was a cross-sectional analysis of patients enrolled between March 2013 and October 2016 in a prospectively collected, institutional review board approved database of children < 21 years of age presenting to Cincinnati Children’s Hospital Medical Center (Cincinnati, Ohio) with their first documented episode of AP. Diagnosis of AP was based on the presence of 2 or more of the following criteria with testing acquired per the treating provider’s discretion (9) : 1) Pain consistent with a pancreatic origin or surrogate markers of vomiting, fussiness, and feeding intolerance, which may be more common signs/symptoms of AP in infants and toddlers (4); 2) Serum amylase or lipase level ≥3 times the upper limit of normal; 3) Imaging (ultrasound [US], computed tomography [CT], or magnetic resonance imaging [MRI]) with findings of AP including: pancreatic edema, pancreatic hypoenhancement, peri-pancreatic edema or fluid collection, and frank pancreatic or peripancreatic necrosis.

Severe acute pancreatitis was defined as per the pediatric AP classification. (15)

A subset of the patients presented here have previously been included in prior publications, with different study aims and hypotheses(16, 17).

Clinical, laboratory and demographic data were extracted from the database at the time of presentation and included: age, sex and body mass index (BMI). Additional compiled data included risk factor(s) for AP: concurrent illness, genetic risk factors, trauma, gallstones and medications at the time of AP diagnosis. A patient’s AP was considered idiopathic if no specific cause or risk factor was identified. All imaging obtained within +/− 1 week of AP diagnosis was retrospectively blindly reviewed for this study by a board-certified pediatric radiologist to confirm and document findings of AP. All available data were analyzed.

Statistical analyses:

Because the database included only patients with confirmed AP, the primary outcome of interest for the current study was sensitivity for AP diagnosis. Specificity could not be calculated due to the absence of negative cases. Sensitivity was calculated for individual diagnostic elements (eg, serum lipase and amylase levels, and imaging by ultrasound, cross-sectional imaging (CT and MRI)) as well as for combinations of diagnostic elements.

The sensitivity analysis was carried out via a binomial distribution model, where exact confidence intervals were computed using the Clopper-Pearson method. For categorical and continuous factors respectively, Fisher exact and Kruskal-Wallis tests were used for inference testing. The association between SAP as an outcome and serum lipase and amylase measurements was analyzed via a logistic regression model. All analyses were performed using SAS version 9.4 (Cary, NC).

Results:

During the study period, 112 patients were enrolled into the registry. The patient cohort included 57 (51%) males, with a median age of 13.4 years (IQR: 9.3-15.8) (Table 1); 84% of patients were white. Fourteen patients (12.5%) had SAP.

Table 1 -.

Population Characteristics Including Risk Factors for AP (n=112)

Age at AP diagnosis (median, IQR)	13.4 years (9.3-15.8)
Duration of follow-up (median, IQR)	20.8 months (10.0-32.8)
Sex (male)	n=57 (51%)
Race
White/Caucasian	n=94 (84%)
Black/African-American	n=9 (8%)
Other	n=9 (8%)
Body mass index percentile (median, IQR)	60.0 (24.1 – 92.1)
Risk Factor for AP c
None (idiopathic)	n=34 (30%)
Medication	n=25 (22%)
Biliary or gallstone	n=21 (19%)
Viral/systemic infection	n=19 (17%)
Trauma	n=9 (8%)
Genetic predisposition	n=8 (7%)
Obstructive predisposition	n=3 (3%)
Metabolic predisposition	n=2 (2%)

^aData presented as median (IQR) or n (%)

^bAP=acute pancreatitis

^cn=8 patients had multiple risk factors

Presence or absence of symptoms was recorded for all patients and serum lipase level was collected in all patients. Serum amylase was collected in 85 patients (76%). Imaging of any type (US, CT, MRI) was performed within 1 week of presentation in 98 patients (88%).

All but one patient had signs or symptoms of AP. The patient without characteristic signs or symptoms met AP criteria with an elevated serum lipase level (but normal serum amylase level) and AP findings identified by imaging performed as part of an evaluation for acute hepatitis.

Fifty-three patients (47%) met 3/3 diagnostic criteria for AP. The remaining 59 patients (53%) met 2/3 diagnostic criteria for AP. There were no significant differences in age, sex or BMI between patients who met 2/3 diagnostic criteria versus 3/3 criteria (Table 2; available at www.jpeds.com).

Table 2–

Comparison of Patient Demographics Depending on Presenting Symptoms

Criteria Met for Diagnosis of AP	n	Median Age (IQR)	Sex	BMIa
Laboratory testing and imaging (regardless of clinical)	54	14.2 years (9.87-16.2)	34 (63%) male 20 (37%) female	Underweight: 7 (13%) Normal: 29 (54%) Overweight: 4 (7%) Obese: 14 (26%)
Laboratory testing and clinicalb (no or negative imaging)	52	12.6 years (9.09-15.4)	21 (40%) male 31 (60%) female	Underweight: 5 (10%) Normal: 27 (52%) Overweight: 6 (12%) Obese: 5 (10%)
Imaging and clinical (i.e., no testing or negative laboratory tests)	6	12.6 years (11.5-13.1)	2 (33%) male 4 (67%) female	Underweight: 0 (0%) Normal: 4 (67%) Overweight: 2 (33%) Obese: 0 (0%)
p-value		0.294	0.053	0.292

^aNo BMI calculated for patients < 2 years or > 20 years of age.

^bBMI data not available for 52 pts as height not obtained at time of attack

^cBMI=body mass index

AP Risk Factors

The single most common identified risk factor for AP was medication (22%) followed by gallstones (19%). No specific risk factor for AP (i.e., idiopathic pancreatitis) was identified in 30% of the patients. Eight patients (7.1%) had more than one identified risk factor for AP. Three patients had combined biliary and genetic risk, and five patients had combination of viral/systemic, medication, or metabolic risk factors.

Laboratory markers

The frequency with which pancreatic enzyme levels were obtained and the frequency with which they were positive or negative is detailed in Figure 1. No patient had an elevated level of serum amylase without a corresponding elevated lipase. Of patients with both a serum amylase and serum lipase measured, five patients (4.5%) had neither elevated above three times the upper limit of normal, but did meet diagnostic criteria for AP based on the remaining two criteria of symptoms and imaging. Of the patients who had serum lipase but not serum amylase measured, one patient had a normal lipase level, bringing the total false negative rate for serum lipase to 5.4% (95% CI: 2.0–11.3%) (Table 3; available at www.jpeds.com). Serum lipase alone was 95% sensitive for AP (95% CI: 89-98%). Serum amylase alone was 39% sensitive for AP (95% CI: 28–50%), with a false negative rate of 61% (95% CI: 50-72%).

Figure 1.

Table 3–

Sensitivity of Individual Diagnostic Criteria and Combinations of Diagnostic Criteria for AP

Criteria	Sensitivity ratio	Sensitivity	95% CI (Clopper-Pearson)
Single Criterion
Clinical	111/112	0.99	(0.95, 1.00)
Laboratory	106/111	0.95	(0.90, 0.99)
Imaging	60/98	0.61	(0.51, 0.71)
Combination of Criteria
Any 2 out of 3 (clinical, laboratory, imaging)	112/112	1.00	(0.97, 1.00)
Laboratory and clinical	105/111	0.95	(0.89, 0.98)
Clinical and imaging	59/98	0.60	(0.50, 0.70)
Laboratory and imaging	54/97	0.56	(0.45, 0.66)

^aAP=acute pancreatitis

For the patients who had both serum amylase and lipase measured at presentation, logistic regression (log-scale) showed both lipase and amylase were positively, but not statistically significantly, correlated to severe AP with odds ratios of 1.28 (95% CI: 0.13-12.66, p=0.83) and 1.60 (95% CI: 0.13-19.17, P = .71), respectively.

Imaging findings

Eighty-four patients (75%) had an US performed, and CT or MRI was performed in 46 (41%) patients (Figure 2). The frequency of imaging findings identified by each imaging modality is detailed in Table 4. The most frequent finding across all three modalities was pancreatic edema (Figure 3; available at www.jpeds.com), followed by peripancreatic edema and acute peripancreatic fluid collection.

Figure 2.

Table 4 –

Imaging Findings of AP Within 1 Week of Attack

Finding	CT	MRI	Ultrasound
Pancreatic edema	56.8% (42/74)	75.0% (27/36)	88.2% (15/17)
Peripancreatic edema	45.9% (34/74)	72.2% (26/36)	76.5% (13/17)
Acute peripancreatic fluid collection	22.2% (16/72)	52.8% (19/36)	47.1% (8/17)
Pancreatic necrosis		12.5% (4/32)	7.7% (1/13)
Peripancreatic necrosis		3.2% (1/31)	0%
Hemorrhage		3.3% (1/30)	0%
Pancreatic duct dilation	20.5% (15/73)	6.7% (2/30)	29.4% (5/17)

^aNot all findings could be assessed on every exam, accounting for denominators less than the total

Figure 3.

Imaging (of any type) within 1 week of presentation for subsequently confirmed AP was only 61% sensitive (95% CI: 51-71%) for AP (Table 3). Thirty-eight (34%) patients had pain and positive laboratory findings with normal imaging, a 34% false negative rate for imaging. Ultrasound alone was 52% (95% CI: 41%-63%) sensitive for AP and CT or MRI was 78% (95% CI: 63%-89%) sensitive. CT or MRI identified AP in seven of 12 patients (58%) for whom testing was performed following a negative US. CT/MRI and ultrasound were concordant in 25 patients (78% of the 32 patients for whom both studies were performed).

Combinations of diagnostic criteria

Diagnostic sensitivity of combinations of accepted diagnostic criteria are detailed in Table 3. Of note, combinations of criteria performed no better than individual criteria alone.

Discussion:

In this cross-sectional study, based on a population enrolled in a prospective clinical database, we sought to define the sensitivity of each of the accepted diagnostic criteria (characteristic symptoms, laboratory values and imaging) for the diagnosis of AP in pediatrics. Our results demonstrate that serum lipase is a more sensitive biomarker for AP in children than serum amylase, a finding that is consistent with the broader literature, but here confirmed from a prospectively collected registry(3, 12, 18, 19). Relative to serum amylase, serum lipase elevations are less time-dependent during the disease process of AP due to its delayed peak and longer duration of elevation, whereas serum amylase peaks and normalizes more rapidly (8, 12, 20). Furthermore, serum amylase may be elevated on the basis of non-pancreatic pathology, more frequently than lipase(7, 21, 22). Although serum lipase is a highly sensitive biomarker for AP in children, elevated pancreatic enzymes may not be present in all cases of AP. From our patient cohort, 5.4% of patients did not meet diagnostic criteria by lipase elevation and/or amylase elevation but met criteria based on signs/symptoms and imaging findings supportive of AP.

Although our results indicate a relatively high sensitivity for cross-sectional imaging, there was likely selection bias of patients undergoing CT or MRI. Collectively, imaging had a 34% false negative rate for AP. This moderate diagnostic performance is consistent with prior reports. Specifically, US has been shown to be 62 to 67% sensitive in diagnosing AP in adults, while in pediatric patients, pancreatic findings suggestive of AP are present in roughly 30 to 50% of cases (3, 4, 14, 18). CT findings of AP are present in pediatric patients approximately 47 to 75% of the time (4, 7, 14). Although US traditionally has been the initial imaging modality of choice due to wide availability, ease of use, absence of sedation needs or ionizing radiation exposure, our data as well as those of prior studies suggest that there are significant diagnostic limitations for this modality and call into question the utilization of US as the first imaging modality of choice for suspected AP in children. Further prospective studies are needed to define the optimal imaging paradigm for the child with suspected AP.

Our study has several limitations. First, although enrollment in the registry from which this population is drawn was prospective, there was no standardization of the management of the enrolled patient population. Because of this, both amylase and lipase were not measured in all patients, imaging was not consistently obtained as part of a diagnostic algorithm, and the utilized imaging modality was not standardized. We have the leverage of standardizing management at our institution, through an admission order set that is used hospital wide and hence we see that imaging was highly available within admissions of AP. However, orders even from the order set remain provider dependent and studies were not obtained in all patients systematically, which limits our ability to compare biochemical and imaging markers, albeit, significant differences were identified. Second, because this study is based on a population of patients enrolled on the basis of meeting established diagnostic criteria for AP, there is an inherent assumption in our study that these criteria captured all patients with AP during the time period of interest. It is feasible that occurrences of AP may have been missed if they did not meet diagnostic criteria. Third, because the study population was derived from a prospective registry of patients with confirmed AP, we are unable to report the specificity of the diagnostic elements. Additionally, we only evaluated the first episode of AP, and did not investigate the utility of laboratory testing or imaging in acute recurrent pancreatitis or chronic pancreatitis, as these populations may present differently. Finally, it is possible that our evaluation of imaging within one week of presentation missed detection of milder AP episodes, especially if obtained at either extreme of the 7-day period. This may have contributed to the 34% false negative rate.

We conclude that serum amylase has a limited utility in the diagnosis of pediatric AP and imaging paradigms for the diagnosis of pediatric AP need to be re-evaluated.

Abbreviations:

BMI

Body Mass Index

CT

Computed Tomography

MRI

Magnetic Resonance Imaging

US

Ultrasound