Clinical and Genetic Risk Factors for Acute Pancreatitis in Patients With Acute Lymphoblastic Leukemia

Abstract

Purpose: Acute pancreatitis is one of the common causes of asparaginase intolerance. The mechanism is unknown, and genetic predisposition to asparaginase-induced pancreatitis has not been previously identified.

Methods: To determine clinical risk factors for asparaginase-induced pancreatitis, we studied a cohort of 5,185 children and young adults with acute lymphoblastic leukemia, including 117 (2.3%) who were diagnosed with at least one episode of acute pancreatitis during therapy. A genome-wide association study was performed in the cohort and in an independent case-control group of 213 patients to identify genetic risk factors.

Results: Risk factors associated with pancreatitis included genetically defined Native American ancestry (P < .001), older age (P < .001), and higher cumulative dose of asparaginase (P < .001). No common variants reached genome-wide significance in the genome-wide association study, but a rare nonsense variant rs199695765 in CPA2, encoding carboxypeptidase A2, was highly associated with pancreatitis (hazard ratio, 587; 95% CI, 66.8 to 5166; P = 9.0 × 10(-9)). A gene-level analysis showed an excess of additional CPA2 variants in patients who did versus those who did not develop pancreatitis (P = .001). Sixteen CPA2 single-nucleotide polymorphisms were associated (P < .05) with pancreatitis, and 13 of 24 patients who carried at least one of these variants developed pancreatitis. Biologic functions that were overrepresented by common variants modestly associated with pancreatitis included purine metabolism and cytoskeleton regulation.

Conclusion: Older age, higher exposure to asparaginase, and higher Native American ancestry were independent risk factors for pancreatitis in patients with acute lymphoblastic leukemia. Those who inherit a nonsense rare variant in the CPA2 gene had a markedly increased risk of asparaginase-induced pancreatitis.

Fig 1.

Study design and genotyping platforms of the initial genome-wide association study (5,398 patients) that aimed to identify the top-ranked single-nucleotide polymorphisms (SNPs) and genes from the cohort, and the SNPs or genes significantly associated with pancreatitis (P < .05) in the case-control group. The number of patients and number of cases (in brackets) in each protocol are shown.

Fig 2.

Incidence of pancreatitis differed by protocols. (A) Cumulative incidence estimates of pancreatitis in the cohort (N = 5,185) accounting for competing risks and time at risk by different protocols. Time at risk was censored at therapy completion; failures other than pancreatitis, such as induction failure, relapse, secondary malignancy, or death, are considered as competing events. The inset table shows the number of patients at risk at 0, 1, 2, and 2.5 years into therapy. (B) The relationship between the crude rates of pancreatitis and protocol-planned dose of asparaginase (ASP), and (C) the relationship between the crude rates of pancreatitis and protocol-planned treatment duration of ASP in this study (solid circles) and in some recently reported protocols^-,,,, (open circles). These rates do not account for competing events or time at risk and are not incidences. The individual points show the average protocol-planned dose or treatment duration of native Escherichia coli–ASP across all treatment arms for each protocol. PEG-ASP 2,500 U/m² every other week is considered equivalent to native E coli-ASP at 25,000 U/m² once per week for 2 weeks. The linear regression lines, R², and P values are shown. PEG-ASP, pegylated Escherichia coli–asparaginase.

Fig 3.

Nonsense single-nucleotide polymorphism (SNP) rs199695765 in CPA2 associated with pancreatitis. (A) Protein structure of proCPA2 (without the 16–amino acid signal peptide) with the nonsense SNP rs199695765 (arrow) located in the propeptide domain. The unaffected region of p.Q51X variant (magenta), the affected propeptide (yellow), and the active enzyme moiety (blue) are shown. (B) Cumulative incidence of pancreatitis in patients in the cohort or in the case-control group who carried any CPA2 nonsense or missense SNPs (variant) was higher than that in patients who are classified as wild type (ie, they carried no or only synonymous or noncoding SNPs). *P = .001 using the Fine-Gray competing risks regression after adjusting for clinical characteristics.