We conducted a retrospective, descriptive study of adult patients who received ICI therapy at The University of Texas MD Anderson Cancer Center and developed grade ≥ 3 elevation of the serum lipase according to Common Terminology Criteria for Adverse Events (CTCAE) v4.03 from April 2011 through April 2018. Grade 3 is defined as lipase elevation of > 2 times the upper limit of normal (60 U/L in our laboratory). The study was approved by the Institutional Review Board at MD Anderson. Included patients met all of the following criteria: (1) aged 18 years or older, (2) received ICI therapy, and (3) developed grade ≥ 3 elevation of serum lipase during or after ICI therapy within 6 months. After a comprehensive medical chart review, we excluded patients who developed acute pancreatitis due to other reasons, such as endoscopic retrograde cholangiopancreatography, gallstone or biliary obstruction, alcohol, and other non-ICI drugs that have known risk of pancreatic toxicity.

Collected data consisted of patient demographics, comorbidities, oncology history, ICI regimen, and non-pancreatic irAEs. Patient demographics consisted of age, sex, and race/ethnicity. Smoking history, alcohol consumption, type II diabetes mellitus, drug allergy, prior history of pancreatitis, and pancreatic metastasis were considered and recorded as baseline risk factors for pancreatic injury. Comorbidities included hypertension, diabetes mellitus, hyperlipidemia, congestive heart failure, and myocardial infarction. Baseline laboratory tests were reported for all patients before ICI therapy.

Oncologic details related to the type, site, and stage of cancer, as well as interruption of ICI therapy, were recorded. Malignancies were categorized as melanoma, solid tumors (including genitourinary, lung, gastrointestinal, head and neck, and other solid tumors), and hematologic malignancies. The stage of melanoma and solid tumors was assessed according to the American Joint Committee on Cancer Staging System 7th edition [29]. The stage of hematologic malignancies was not reported. ICI agents used were ipilimumab, nivolumab, pembrolizumab, atezolizumab, durvalumab, avelumab, and tremelimumab. ICI agents were categorized as (1) CTLA-4 monotherapy, (2) PD-1/L1 monotherapy, or (3) combination of both. The mean duration of ICI therapy and the median number of ICI doses were recorded. Reasons for ICI therapy discontinuation, permanent or temporary, were also documented.

Baseline lipase and amylase values were assessed for all patients to check for pancreatic injury prior to ICI therapy. We collected the peak value of serum lipase and amylase, as well as computed tomography (CT) findings of the pancreas (such as peripancreatic fat stranding, pancreatic enlargement with heterogeneous enhancement, and segmental hypoenhancement). We defined ICIPI as elevation of lipase with and without clinical presentation. Typical clinical presentation of acute pancreatitis was epigastric pain aggravated by leaning forward or radiating to the back, with or without associated nausea and vomiting. Other clinical signs and symptoms of ICIPI included fever, diarrhea, dyspnea, and hemodynamic instability (as a result of ICIPI or other reasons). Diarrhea was recorded as a part of ICIPI clinical presentation if it was not diagnosed previously as a symptom of immune-mediated colitis and was not of infectious cause. Thereafter, patients were divided into two groups on the basis of the presence of typical acute pancreatitis symptoms. Duration of clinical symptoms was measured from symptom onset to improvement.

Data relating to the treatment of pancreatic injury included use and amount of intravenous (IV) fluid, corticosteroids (either for pancreatitis only or concurrently for other reasons), and duration of treatment. Hospitalization to treat elevated serum lipase with IV fluids or/and steroids was also documented. Short-term adverse outcomes of ICIPI were reported as pancreatic pseudocyst and need for hospitalization (along with duration of hospitalization). Long-term adverse outcomes of pancreatitis included diabetes, chronic pancreatitis features on imaging (such as pancreatic atrophy), and recurrence of lipase elevation within a mean follow-up duration of 16 months.

The distribution of continuous variables was summarized by means and standard deviations or medians and interquartile ranges. The distribution of categorical variables was summarized using frequencies and percentages. Continuous variables were compared between groups using the Wilcoxon rank-sum test. The Fisher exact test was conducted to evaluate associations between categorical variables. Univariate logistic regression analysis was conducted to assess for the risk of long-term adverse outcomes of ICIPI. Kaplan-Meier curves and the log-rank test were used to estimate and compare unadjusted overall survival time distributions. A multivariable Cox proportional hazards model was used to evaluate the ability of independent covariates to predict overall survival. All statistical evaluations were two-sided, and a P value of ≤0.05 was considered statistically significant. Statistical analysis was carried out using the SPSS Statistics software program (version 24.0; IBM Corporation, Armonk, NY).

Among 5,762 patients who received ICI therapy during the period studied, 2,279 patients had lipase levels tested; 627 patients received anti-CTLA-4 monotherapy, 1434 received PD-1/L1 monotherapy, and 218 received combination therapy. In the CTLA-4 monotherapy group, 12 patients (2%) developed grade ≥ 3 serum lipase elevation that was deemed related to ICI therapy. Among patients who received PD-1/L1 monotherapy, 53 (4%) had ICIPI. In the combination therapy group, 17 (8%) developed ICIPI. Thus, our cohort included 82 patients. Baseline clinical characteristics of the patients are shown in Table 1. In our cohort, most patients (66%) were male with a mean age of 57 years. Melanoma was the most common malignancy in our cohort (37%). The median number of ICI doses was 4 (interquartile range 1–25). Other non-pancreatic irAEs reported at the time of ICIPI onset included entercolitis in 27 patients (33%), hepatic injury in 17 patients (21%), dermatologic events in 13 patients (16%), and endocrine events in 7 patients (9%).Tab1

Clinical characteristics of patients in our cohort (n = 82)

The median time from ICI initiation to peak lipase elevation was longer in patients who received PD-1/L1 monotherapy than in patients who received CTLA-4–based regimens (median 146 days for PD-1/L1 compared with 69 days for CTLA-4 monotherapy and 110 days for combination therapy, P = 0.033; Additional file 1: Table S1). Clinical symptoms of pancreatitis were epigastric pain in 32 patients (39%), nausea and vomiting in 23 (28%), fever in 7 (9%), and diarrhea in 16 (20%). Twenty-four patients had more than one of these symptoms. Forty five patients had completely asymptomatic presentation. Three of the patients who had completely asymptomatic presentation had CT findings suggestive of ICIPI with lipase elevation. No differences were observed in the clinical characteristics of patients according to ICI type. Typical pancreatitis presentation was observed in 32 patients (39%) and the other 50 (61%) did not have typical pancreatitis symptoms (Table 2). According to CTCAE grades for pancreatitis, 41 patients (50%) in our cohort had grade 2 pancreatitis (enzyme elevation or radiologic findings only) and the rest had grade 3 pancreatitis (pain, vomiting, medical intervention indicated). Fever occurred more frequently in patients with typical symptoms of pancreatitis than in those with asymptomatic lipase elevation (P = 0.013). Patients with a prior history of pancreatitis had an increased risk of ICIPI with clinical symptoms (P = 0.013). Although not statistically significant, CT findings suggestive of pancreatitis were more prevalent in patients who had symptoms of pancreatitis than in those who did not (P = 0.055). Patients with clinical symptoms of pancreatitis developed higher mean peak values of serum lipase than did patients without symptoms (3227 U/L for clinical symptoms compared with 1989 U/L for no symptoms, P = 0.032). There was no statistically significant difference in the peak values of serum amylase between the two groups (378 U/L for clinical symptoms compared with 270 U/L for no symptoms, P = 0.082).Tab2

Clinical characteristics and outcomes stratified by clinical symptoms

ICI therapy was more often discontinued in patients who developed clinical symptoms than in patients without symptoms of pancreatitis (P = 0.012). Thirty two patients received IV fluids for ICIPI; most were normal saline (n = 28), followed by ½ normal saline (n = 3) and lactated ringer (n = 1). IV fluids were given more frequently in patients who developed clinical symptoms of pancreatitis than in patients who did not (P < 0.001). Patients who had clinical symptoms of pancreatitis received steroids and required hospitalization more frequently than patients who had no symptoms (P = 0.008 for steroid; P < 0.001 for hospitalization).

Eleven patients (13%) had pancreatic metastasis at the time of ICI initiation. Mean baseline lipase values were similar between patients who had pancreatic metastasis and those who did not (146 for those with metastasis and 162 for those with no metastasis, P = 0.769). We did not find any statistically significant differences in regards to epigastric pain (P = 0.325) and peak lipase value (P = 0.827) between patients who had pancreatic metastasis and those who did not. Also, there was no statistically significant difference in the prevalence of CT findings of pancreatitis (P = 0.238) and the treatment administered for pancreatitis (IV fluid, P = 0.325; steroids, P = 0.347) between patients who had pancreatic metastasis and those who did not. No significant associations were found between pancreatic metastasis and short-term or long-term adverse outcomes of ICIPI.

Among the 64 patients who had CT scan evaluation, findings of pancreatitis were found in 11 patients (17%). The most common features suggestive of pancreatitis were segmental hypoenhancement, stranding in peripancreatic fat, and pancreatic enlargement with heterogeneous enhancement. Three of the patients who had CT findings of pancreatitis had also pancreatic metastasis, and the other eight did not. Additional file 2: Figure S1 shows features of pancreatitis with and without pancreatic metastasis.

In regards to short-term clinical outcomes of ICIPI, three patients (4%) developed a pancreatic pseudocyst and 18 (22%) required hospitalization with a mean duration of 5 days. The associations of steroids and IV fluids with short-term outcomes of ICIPI are summarized in Table 3. The use of steroids and IV fluids did not shorten the time from peak lipase value to improvement to grade 1 or below (P = 0.711 for steroids and P = 0.850 for IV fluids), the duration of symptoms (P = 0.824 for steroids and P = 0.650 for IV fluids), or the duration of hospitalization (P = 0.762 for steroids and P = 0.166 for IV fluids). The administration of steroids or IV fluids was not associated with ICI therapy discontinuation (P = 1.000 for steroids and P = 0.113 for IV fluids). Upon stratifying patients by the grade of lipase elevation, we found that steroid therapy did not affect the short-term outcomes in patients with grade 3 compared with grade 4 lipase elevation (Additional file 1: Table S2). Similarly, IV fluid administration was not associated with different short-term outcomes in patients with grade 3 or grade 4 lipase elevation (Additional file 1: Table S3).Tab3

Short-term clinical outcomes of pancreatitis by treatment for immune checkpoint inhibitor-induced pancreatic injury

Long-term adverse outcomes of ICIPI were chronic pancreatitis in three patients (4%) and diabetes in six (7%); five of them required insulin and the sixth required metformin. Four of the patients who developed consequent diabetes did not receive steroids for ICIPI. Additionally, 4 of the 35 patients (11%) who resumed ICI therapy after ICIPI had recurrence of lipase elevation, one symptomatic and three asymptomatic. Patients who developed adverse outcomes of ICIPI received longer-duration ICI therapy than patients who did not (mean 412 days for adverse outcomes and 200 days for no adverse outcomes, P = 0.006; Table 4). IV fluids were more frequently given to patients without subsequent long-term adverse outcomes of ICIPI than patients with adverse outcomes (P = 0.044).Tab4

Clinical characteristics of patients who had adverse outcomes of immune checkpoint inhibitor-induced pancreatic injury (chronic pancreatitis, diabetes, and recurrence)

Results of logistic regression analysis of long-term adverse outcomes are shown in Table 5. Smoking history (odds ratio 4.35; 95% confidence interval 1.06–17.81, P = 0.041) and hyperlipidemia (odd ratio 6.15; 95% confidence interval 1.24–30.55, P = 0.026) were associated with an increased risk of long-term adverse outcomes of ICIPI. Patients who received IV fluids had a favorable ICIPI long-term outcomes profile (odds ratio 0.21; 95% confidence interval 0.06–0.79, P = 0.022). No significant associations were reported between long-term adverse outcomes of ICIPI and ICI type, symptoms of pancreatitis, abnormal CT findings, peak lipase and amylase values, and use of steroids.Tab5

Univariate logistic regression analysis for long-term pancreatic adverse outcomes

Among the 82 patients, 18 received IV fluids without steroids, and none of these patients developed long-term adverse outcomes from ICIPI. Fourteen patients received both IV fluids and steroids, and among these patients, only one (7%) developed long-term adverse outcomes of ICIPI. On the other hand, four of the 17 patients (24%) who received steroids without IV fluids developed long-term adverse outcomes from ICIPI. Similarly, six of the 33 patients (18%) who received neither IV fluids nor steroids developed long-term adverse outcomes (Additional file 3: Figure S2A).

Only one (3%) of the 32 patients who received IV fluids developed long-term adverse outcomes of ICIPI, whereas 10 (20%) of the 50 patients who did not receive IV fluids developed long-term adverse outcomes. None of the patients who discontinued ICI therapy after ICIPI and received IV fluids had long-term adverse outcomes (Fig. 1). By contrast, 24% of the patients who continued ICI therapy and did not receive IV fluids had long-term adverse outcomes. Overall, patients who had no symptoms of pancreatitis and received no IV fluids had the highest rate of long-term adverse outcomes of ICIPI (22%) among all patients (Fig. 2). A total of 21 patients had no typical symptoms of pancreatitis, continued ICI therapy after ICIPI, and did not receive IV fluids. Of these, 6 (29%) had long-term adverse outcomes. Long-term outcomes of ICIPI stratified by the duration of follow-up is shown in Additional file 4: Figure S2B.Fig1Fig. 1

Long-term adverse outcomes of immune checkpoint inhibitor-induced pancreatic injury by treatment for pancreatitis

Kaplan-Meier curves showed that patients who resumed ICI therapy after ICIPI onset had slightly longer overall survival than did patients who did not, although this was not statistically significant (P = 0.0559; Additional file 5: Figure S3). Patients with long-term adverse outcomes of ICIPI had significantly longer overall survival duration than did patients without long-term adverse outcomes (P = 0.0295; Additional file 6: Figure S4). Overall survival rate did not differ between patients who received steroids and those who did not (P = 0.4088; Additional file 7: Figure S5) or between those who had symptoms of pancreatitis and those who did not (P = 0.1436; Additional file 8: Figure S6). The multivariable Cox model showed no significant associations between any of the clinical features and overall survival duration (Additional file 1: Table S4).