Ultrasound has limited diagnostic utility in children with acute lymphoblastic leukemia developing pancreatitis

Abstract

Purpose:

Acute pancreatitis (AP) due to chemotherapy-induced pancreatic injury is a common side effect of treatment for acute lymphoblastic leukemia (ALL), the most common childhood malignancy. The American College of Radiology recommends ultrasound (US) for initial imaging of AP in all populations to assess for ductal obstruction. However, US may be insensitive to diagnose and assess chemotherapy-associated AP.

Methods & Materials:

The institutional review board approved this retrospective study. Patients with ALL and AP were identified from protocol databases, using Common Terminology Criteria for Adverse Events (CTCAE) version 3. Chemotherapy dosing, amylase/lipase levels, clinical symptoms, and US/Computed Tomography (CT) reports within 10 days of diagnosis were recorded. All CT images were reviewed for revised Atlanta classification and CT Severity Index.

Results:

69 patients, aged 2–21 years, experienced 88 episodes of AP, undergoing 98 US’s and 44 CT’s. 72 events (82%) occurred within 30 days of asparaginase administration. 69 episodes (78%) were initially diagnosed by the presence of abdominal pain and pancreatic enzyme elevation. Overall sensitivities for AP detection were 47% using US and 98% for CT. US sensitivity was greatest in CTCAE grade 4 (86%) and necrotizing pancreatitis (67%).

Conclusions:

Most cases of AP in children with ALL can be diagnosed with clinical history and labs. US has limited sensitivity in detecting pancreatitis in this population. Imaging to diagnose AP in this patient population could be limited to clinically equivocal cases.

INTRODUCTION:

Acute pancreatitis (AP) is a known complication of treatment with asparaginase, steroids, and 6-mercaptopurine for acute lymphoblastic leukemia (ALL) [1]. An estimated 3–18% of patients with ALL receiving asparaginase therapy will develop AP during treatment, usually within the first year of asparaginase-intensive phases. This is thought to be due to parenchymal toxicity from chemotherapy [2–4].

Clinical diagnosis of AP requires two out of three of the following: characteristic abdominal pain, serum amylase or lipase three times the upper limit of normal, and/or imaging findings characteristic of AP[6, 7].

While current recommendations from the American College of Radiology (ACR) suggest ultrasound (US) as the initial imaging modality in first episodes of AP in both adults and children, this recommendation is based on the premise that most cases of AP are secondary to biliary ductal obstruction, as opposed to the direct pancreatic injury believed to cause chemotherapy-induced AP [6]. Additionally, US has many known limitations, including operator experience, frequent lack of visualization of the entire pancreas related to overlying bowel gas or body habitus, and variability in the appearance of a normal and abnormal pancreas. As such we believe that US may be insensitive in the diagnosis of ALL-associated AP.

Furthermore, we believe that the best initial imaging modality to identify early AP related to direct treatment toxicity is one that offers a more thorough examination of the entire pancreas and adjacent soft tissues with greater reproducibility and diagnostic accuracy, such as computed tomography (CT) or magnetic resonance imaging (MRI).

In a retrospective cohort of children with ALL and known episode(s) of AP, we evaluated the sensitivity of US in demonstrating AP findings. Then we compared the sensitivity of US to CT imaging when available.

METHODS:

Approval for waiver of consent was obtained from the St Jude Children’s Research Hospital Institutional Review Board for this retrospective study.

Cases of pancreatitis in children treated for ALL that were diagnosed between October 26, 2007 and September 10, 2018 were identified through review of institutional databases. The Total Therapy XVI (T16, NCT00549848) and XVII (T17, NCT03117751) protocols include children with ALL diagnosed at ages 0–18 years and 1–18 years, respectively. Both protocols administered pegaspargase (PEGylated L-asparaginase) to all patients unless allergic reaction required transition to alternative asparaginase formulations.

Diagnosis of AP based on the Revised Atlanta Classification/International Study Group of Pediatric Pancreatitis: In Search for a Cure (INSPPIRE) consortium criteria was confirmed by manual chart review, and demographic and clinical data were collected [6, 7].

Clinical data were extracted from the medical record including demographics, body mass index (BMI), complications by Common Terminology Criteria for Adverse Events (CTCAE) version 3 grade of AP, clinical symptoms, chemotherapy, and dates of events. Amylase and lipase levels were collected from the medical record. Statistical analyses were performed using Stata version 16.1 (College Station, TX) and Graphpad Prism version 8.4.3 (San Diego, CA).

Defining the diagnostic timing

A “true” date and time of diagnosis was identified to determine how the patient first met criteria for the diagnosis of AP (i.e. met 2 of 3 criteria including characteristic pain, amylase (total) or lipase elevation, and imaging evidence of pancreatitis), the timing of the episode of AP with respect to asparaginase administration, and the duration from onset of pain until diagnosis. The date and time when abdominal pain was first documented was extracted from clinical notes. Based on documented lab values, the first date of elevated serum amylase or lipase levels greater than 3 times the upper limit of normal for each lab was recorded. The date and time of the report from the first positive imaging study was recorded. The date and time of the first two positive components of diagnosis were designated as the “true” diagnostic period.

Radiological Review

The first two US studies that attempted visualization of the pancreas and the first abdominal CT performed within 10 days of the true date of diagnosis were selected for review. In cases without an abdominal CT, an available chest CT that included the pancreas was designated as the corresponding CT. A board-eligible radiology resident blinded to the original reports reviewed all US and CT images.

Ultrasound Imaging Review:

In addition to recording the original interpretation/report, each US was assessed for the extent of the pancreas visualized (full, partial, or completely obscured), diagnostic quality (diagnostic or non-diagnostic), characteristic imaging findings of AP (pancreatic enlargement, peripancreatic inflammation, peripancreatic free fluid, and ascites), and whether a diagnosis of AP would be given in a typical clinical setting. The pancreas was measured in the head, neck, body, and tail (if each was visualized) on all US images and in all CT cases. Enlargement was defined as greater than two standard deviations above normal for sex and age per Trout, et al. (ref). A study was deemed of diagnostic quality only if one of the following conditions was met: the entirety of the pancreas could be viewed with certainty and was normal OR the visualized portion of the pancreas demonstrated sufficient findings to confidently diagnose AP in a typical clinical setting (FIGURE 1).

FIGURE 1.

The entirety of the pancreas must be viewed on all imaging modalities because pancreatitis can be focal as in this example. (A) Transverse ultrasound image through the pancreas demonstrating a normal sized pancreas without surrounding fluid or peripancreatic inflammation (white arrow). (B) Corresponding CT through the pancreatic head at the same level demonstrates a normal CT appearance (black arrow, center). (C) However, a more inferiorly located slice through the pancreatic tail on the same CT study demonstrates focal pancreatic parenchymal enlargement with regions on non-enhancement and trace peripancreatic free fluid, consistent with necrotizing pancreatitis with an acute necrotic collection, CTSI 5 (black arrow, right).

Computed Tomography Imaging Review:

In addition to recording the original interpretation/report, each CT was reviewed for diagnostic quality and characteristic imaging findings of AP (focal or diffuse parenchymal enlargement, changes in pancreatic density or enhancement, indistinct pancreatic margins, peripancreatic fat stranding, hemorrhage, pancreatic emphysema, and peripancreatic fluid collections). Diagnostic quality was determined based on identical criteria used for US exams. Additionally, all CT exams were assigned a categorization based on the Revised Atlanta Classification[6, 9] and scored by the CT Severity Index (CTSI; Table S4) [10]. The CT scans were performed with dose modulation, iterative reconstruction, and/or weight-based dosing when appropriate to reduce the radiation dose to as low as reasonably achievable. Radiation dosing data for each CT was recorded.

Adjudication of Imaging Discrepancies:

After all US and CT studies were reviewed, the results were compared with the original imaging reports. All discrepancies in the diagnosis of AP between the radiology trainee’s interpretation and the original radiology report from a board-certified pediatric radiologist were adjudicated by re-review of the imaging by a single board-certified pediatric radiologist (CEM) who was blinded to the original radiology report, clinical record, and trainee interpretation.

RESULTS

Incidence of AP & demographics

Ninety-six AE reports for AP were identified in 71 patients. Of those, 8 episodes were excluded because incomplete data prevented confirmation of 2 or 3 criteria for AP diagnosis [6]. The analyzed cohort consisted of 69 patients with 88 episodes of AP. Regarding repeat episodes, 54 patients had a single episode, 12 patients had 2 episodes, 2 patients had 3 episodes, and 1 patient experienced 4 episodes while enrolled on protocol. Patients were predominantly male (64%) and Caucasian (77%), with a median age at diagnosis of pancreatitis of 10 years (range 2–21). Most episodes were documented as CTCAE grade 2 (49%) or 3 (40%) (Table 1).

TABLE 1.

Demographics by patient and episode. n_P: number of patients, n_E: number of episodes, SD: standard deviation

Characteristics	Results
Patients, Episodes	nP(%), nE(%)
Male	44 (64%), 58 (66%)
Female	25 (36%), 30 (34%)
Race/Ethnicity
Caucasian	53 (77%), 69 (78%)
African American	12 (18%), 14 (16%)
Asian	3 (4%), 4 (5%)
Other	1 (1%), 1 (1%)
Age at Episode
Mean, yr (±SD)	10.1 (±4.8)
≤ 5 years	18 (20%)
>5 years but ≤ 10 years	28 (32%)
> 10 years but ≤ 15 years	31 (35%)
> 15 years	11 (13%)
CTCAE v 3 of Episode5
Grade 1	3 (4%)
Grade 2	43 (49%)
Grade 3	35 (40%)
Grade 4	6 (7%)
Pancreatic Enzyme Levels at ...	Mean, Median (Range) in units/L
Amylase (normal < 91 units/L)
Diagnosis	343,269 (60 – 1,334)
1st Ultrasound	306, 232 (21 – 1,334)
2nd Ultrasound	232, 160 (19 – 809)
CT	300, 217 (16 – 1,334)
Lipase (normal < 60 units/L)
Diagnosis	1176, 820 (36 – 7,345)
1st Ultrasound	773, 461 (30 – 7,345)
2nd Ultrasound	465, 207 (27 – 2,492)
CT	844, 457 (9 – 7,345)

One hundred forty-three imaging studies (98 US, 44 CT, and 1 MRI) were performed in 70 of the 88 episodes of AP (Figure 2). Both US and CT were available for 43% (38/88) of episodes, obtained a median of 1 day from the onset of pain (interquartile range 1–3 days). Thirty-four of 88 episodes (39%) of episodes underwent 2 or more US’s. Because only 1 MRI was performed, this imaging modality was excluded from our analysis.

FIGURE 2.

Consort diagram of episode inclusion and associated imaging studies. (AP acute pancreatitis, US ultrasound, CT computed tomography)

Type of Asparaginase Administered and Temporal Association of AP with Administration

Prior to the onset of AP, patients received PEGylated L-asparaginase (n = 73 episodes), Erwinia asparaginase (n = 12 episodes), or native L-asparaginase (n = 2 episodes). In one episode, no asparaginase had been administered(1/88, 1.1% of total episodes). The onset of AP was within 30 days of asparaginase administration in 72 of the 87 episodes (83%).Onset of pancreatitis within 30 days of asparaginase administration was higher for first episodes (88%; 58/66 episodes) compared to recurrent episodes (66%; 14/21 episodes, p=0.04).

How Patients Meet Criteria for AP

Nearly all (86/88, 98%, 95% confidence interval [95%CI] 92–99.7%) episodes had documentation of abdominal pain prior to lab or imaging results. In 80/88 episodes (91%, 95%CI 83–96%), the diagnosis of AP could have been made by elevated serum amylase or lipase concentrations within 24 hours of abdominal pain documentation without the use of imaging.

Imaging was often ordered prior to or at the same time as blood chemistries, although the diagnosis was first made in the majority of episodes by elevated enzymes with abdominal pain (69/88, 78%, 95%CI 68–86%). Less frequently, the diagnosis was made by imaging results and abdominal pain prior to lab results (17/88, 19%). Out of the 69 cases that were diagnosed with labs and abdominal pain, 11 episodes had imaging prior to diagnosis, and all of those studies were either false negative US examinations for the diagnosis of AP (9/11) or US was performed to assess for an alternate diagnosis (e.g. appendicitis or colitis) and imaging of the pancreas was not obtained (2/11).

Average Time to Diagnosis After Onset of Abdominal Pain

The mean time from the onset of abdominal pain to diagnosis for all episodes was 1.2 days (TABLE S1). Regardless of the severity or method of diagnosis, most episodes were diagnosed within 24 hours (58/86, 66%) or 48 hours (73/86, 83%). When the diagnosis was made with pancreatic enzymes prior to imaging results (n=58, 67%) the average time from onset of abdominal pain to diagnosis was 1.02 days. Conversely, when a non-diagnostic US was performed prior to blood chemistry results (n=11, 13%) the diagnosis was made an average of 2 days from the onset of pain (p=0.87).

Regardless of the method by which the patient met diagnostic criteria, there was an inverse relationship between the clinical severity of AP and the average number of days from onset of abdominal pain until diagnosis: CTCAE grades 1–4 taking 3, 1.2, 1.1, and 0.8 days, respectively (p=0.16).

Sensitivity of US and CT

Of the 98 US studies, 47% (46/98, 95% CI 37–57%) demonstrated sufficient imaging characteristics to diagnose AP (FIGURE 3). This sensitivity was unchanged when comparing initial versus follow-up US studies. The sensitivity of US was highest for CTAEC grade 4 AP (6/7, 86%). No ultrasounds were diagnostic for grade 1 AP (0/3, 0%). Sensitivities were similar for grade 2 AP (22/50, 44%) and grade 3 AP (18/38, 47%). Sensitivity of US did not vary if it preceded (sensitivity 41%, 13/32) or followed diagnostic CT (50%, 12/24).

FIGURE 3.

Flowchart of imaging sensitivity. (AP acute pancreatitis, US ultrasound, CT computed tomography, TP true positive, FN false negative)

The remainder of the US studies (52/98, 53%) demonstrated incomplete visualization of the pancreas without sufficient findings to diagnose AP or a completely obscured pancreas. Considering all of the US studies, the pancreas was fully visualized in none of the cases (0/98, 0%), partially visualized in 86 studies (86/98, 88%), and non-visualized in 12 studies (12/98, 12%). The percent of cases with non-visualization of the pancreas was similar across CTCAE grades 2–4 AP: 10% (5/45), 13% (5/33), and 14% (1/6), respectively.

Every abdominal CT (or chest CT when applicable) was considered diagnostic, demonstrating the entirety of a normal-appearing pancreas or enough of the pancreas to diagnose AP. Forty-three of 44 CTs (98%, 95%CI 88–100%) were positive for AP (p<0.001 vs. US). The only false negative CT was performed as a follow-up study after diagnosis with pancreatic enzymes and characteristic abdominal pain.

Severity of Episodes of AP Based on CTCAE, CT, and US

Based on the revised Atlanta classification, 29 out of the 44 CTs (66%) demonstrated interstitial edematous pancreatitis (IEP) and 14 of the studies (32%) demonstrated necrotizing pancreatitis (NP). One of 44 CTs (2%) demonstrated a normal pancreas (CTCAE grade 2). CTSI scores increased with increasing CTCAE grades (TABLE 2; p=0.058).

TABLE 2.

Average CTSI score, percentage of CT’s obtained that demonstrate necrotizing pancreatitis, and the ultrasound sensitivity for each CTCAE grade. n_CT: total number of CT’s obtained per grade, n (%): total number of CT’s that demonstrate necrotizing pancreatitis per grade and corresponding percentage of all CT’s obtained in that grade that demonstrate necrotizing pancreatitis

CTCAEv3 Grade	Average CTSI	CT’s with NP n(%)	US Sensitivity (%)
Grade 1 (nCT=2)	2.5	0 (0%)	0%
Grade 2 (nCT=17)	4.3	4 (24%)	44%
Grade 3 (nCT=20)	4.8	7 (35%)	47%
Grade 4 (nCT=5)	7.6	3 (60%)	86%

Of those episodes in which a CT demonstrated AP and comparison US imaging was performed, the overall US sensitivity was 45% (25/56), and the sensitivity by CTCAE grade was 0% (0/3) for grade 1, 36% (8/22) for grade 2, 46% (11/24) for grade 3, and 86% (6/7) for grade 4. Furthermore, the US sensitivity was greatest when structural injury, such as necrosis and peripancreatic fluid collections, were present, with US sensitivity of 80% for CTSI score 7–10 (severe) versus 46% for CTSI 0–3 (mild) (FIGURE 4).

FIGURE 4.

Three transverse US images at the level of the pancreas with corresponding axial CECT images, showing (A) non-diagnostic US with partially visualized, normal appearing pancreas & NP/ANC on CT performed 1 day earlier (CTSI 8, white arrows), (B) diagnostic US with cystic pancreas & NP/ANC on CT performed 1 week later (CTSI 10, white arrows), & (C) non-diagnostic US with non-visualized pancreas & IEP on CT performed 1 day later (CSTI 4, white arrows).

Association of AP and imaging findings with BMI and Age

Consistent with prior findings, obesity limited US performance in our patients. The median patient BMI was 18.8 kg/m2 (range of 13–59 kg/m2) at the time of AP. Based on BMI percentile ranges for age, the patients were underweight in 3 episodes (3/88, 3%), normal weight in 51 episodes (51/88, 58%), overweight in 15 episodes (15/88, 17%), and obese in 19 episodes (19/88, 22%) (TABLE S2). Rates of diagnostic US were similar for patients in the underweight, healthy weight, and overweight categories (~50%) compared to 33% in obese patients (p=0.25, TABLE S3).

US sensitivity did not differ by age: 33% in those greater than 15 years (95%CI 10–65%), 43% for ≤5 years (95%CI 23–66%), 56% >5-≤10 years (95%CI 35–75%), and 47% >10 years - ≤15 years (95%CI 30–65%). In multivariable analysis including age and obesity, neither was associated with US sensitivity (p=0.47 and 0.1, respectively).

Association between Type/Number of US Findings and Likelihood of Diagnosis

We assessed 4 commonly reported sonographic signs of AP (pancreatic enlargement, peripancreatic inflammation, peripancreatic free fluid, and ascites) and found the more US findings of pancreatitis that were present, the more likely the study was diagnostic of AP (3.2 out of 4 findings on average vs. 1.5 out of 4 findings in non-diagnostic studies). Findings on US present in the studies evaluated are shown in table S4.

When examining only those studies in which the pancreas was partially visualized, ascites was the most common US finding, present in 62 studies (62/86, 72%). Pancreatic enlargement was the second most common finding present in 59 studies (59/86, 68%). Ascites and pancreatic enlargement were only moderately associated with the studies being considered diagnostic for AP (37/62, 60% and 43/59, 73%, respectively). Peripancreatic inflammation and peripancreatic free fluid were the least common findings, present in only 42/86 (49%) and 38/86 (44%) studies, respectively.

CT Radiation Dose Data

The total average volume computed tomography dose index (CTDI_vol) across all ages was 7mGy with a range of 1 to 25 mGy. Averages varied over age ranges as follows: 3.1 mGy for ages ≤5 years, 5.4 mGy for ages 5–10 years, 10 mGy for ages 10–15 years, and 12 mGy for ages >15 years.

DISCUSSION

Our results suggest that ultrasound has limited diagnostic utility in the diagnosis of acute pancreatitis in children with ALL. In our review of 88 episodes of pancreatitis in patients receiving modern ALL therapy, most cases (69/88, 78%) of AP were diagnosed without imaging using only characteristic abdominal pain and elevated pancreatic enzymes. In our cohort, 91% of patients could be diagnosed without imaging, including 86% on the day of presentation and 5% within the subsequent 24 hours.

In our cohort of ALL patients with documented AP, US had a lower sensitivity than CT (47% versus 98%, p<0.001). These findings underscore the limited value of US in the diagnosis of asparaginase related AP.

Because the differential diagnosis of abdominal pain in pediatric ALL patients is broad, a specific algorithm for work-up of AP is challenging. In patients with abdominal pain potentially consistent with AP, we suggest first obtaining serum amylase and lipase levels. For most patients, this will be sufficient for diagnosis (Figure S3). In patients with enzyme levels less than 3x the upper limit of normal, abdominal ultrasound to evaluate the pancreas and other potential causes of abdominal pain (e.g. colitis, cholecystitis, appendicitis) can be obtained. Further evaluation should be directed by the history and clinical circumstances. If pancreatic enzymes remain <3x the upper limit of normal after 48 hours and pancreatitis remains a consideration, a contrast enhanced CT or MRI of the abdomen should be obtained. If at any point in the acute setting urgent imaging is necessary for management, a contrast-enhanced CT or MRI of the abdomen should be performed, as necrotizing pancreatitis with or without superimposed infection is not well differentiated from interstitial edematous pancreatitis with US [11]. Additionally, contrast enhanced abdominal MRI or CT will identify complications of AP such as pseudocyst formation, development of walled off necrosis, splenic vein thrombosis, and arterial pseudoaneurysm formation.

Limitations of our study include ascertainment bias due to the use of an adverse events database to identify patients with known episodes of pancreatitis. The lack of availability of US and CT on all patients reduces the precision of sensitivity estimates and the ability to compare these two methods; however, this reflects clinical practice at both our own and external institutions. While some cases of AP may not have been detected, these cases never presented to clinical attention and therefore were likely to be mild.

This study is also limited in part by our institutional clinical practices. Currently, US is used to rule out numerous complications of oncologic treatment, including colitis, as well as common sources of pediatric abdominal pain, including appendicitis, cholecystitis, and gallstones. This reliance on US lends itself to a tendency to use US as a method to rule out any abdominal pathology rather than CT. Furthermore, MRI is rarely used to assess for pancreatitis at our institution (only 1/88 episodes included comparison MRI), so the sensitivity of MRI could not be assessed in comparison with the other modalities.

CONCLUSION

Our study demonstrates the limited sensitivity of US in patients with ALL being evaluated for potential pancreatitis. In children being treated for ALL, US cannot be relied on to diagnose acute pancreatitis. In our population, most cases were diagnosed with clinical history and serum enzymes. Alternative imaging should be considered in clinically equivocal cases in which there is a strong suspicion for acute pancreatitis and the diagnosis cannot be made by history and serum enzymes alone.

Supplementary Material

fig S3

FIGURE S3 Flow-chart of suggested algorithm for diagnosing acute pancreatitis in pediatric patients with ALL presenting with abdominal pain with clinical suspicion for pancreatitis. For the majority of patients with this presentation, AP can be diagnosed by elevated amylase/lipase levels. If serum levels are normal, we suggest repeating in 24 hours at which point 91% of patients in this retrospective cohort could be diagnosed with AP without the need for imaging. Alternative etiologies for abdominal pain should be considered, in which case ultrasound may be helpful (e.g. colitis). If after 24 hours with persistent clinical concern for AP and non-elevated enzymes, CT or MRI should be considered. Of course, a patient’s individual clinical circumstances should dictate the need for imaging at any stage.

fig S2

FIGURE S2 Box plot of the number of US findings in a diagnostic US and a non-diagnostic US.

fig S1

FIGURE S1 Duration of time from asparaginase administration until diagnosis of AP.

Abbreviations Table

ACR

American College of Radiology

ALL

acute lymphoblastic leukemia

AP

acute pancreatitis

BMI

body mass index

CT

computed tomography

CTCAE

Common Terminology Criteria for Adverse Events

CTDIvol

computed tomography dose index

CTSI

computed tomography severity index

MRI

magnetic resonance imaging

US

ultrasound