Immune checkpoint inhibitor associated diarrhoea

Abstract

A man in his 80s was undergoing immunotherapy with pembrolizumab, an anti-PD-1 monoclonal antibody, following his diagnosis of adenocarcinoma of primary lung origin. 24 weeks into treatment, the patient reported experiencing loose stools associated with malaise and poor appetite but no further symptoms. This progressed in frequency and a clinical diagnosis of grade 2 immune checkpoint inhibitor colitis was made. Management with oral prednisolone was commenced but symptoms persisted. Common enteric infections had been ruled out, as were coeliac disease and hyperthyroidism. Flexible sigmoidoscopy and colonoscopy results were not in keeping with colitis, having revealed normal looking mucosa. Following this, a faecal elastase level was found to be low. A diagnosis of pembrolizumab-induced pancreatic exocrine insufficiency was made, and stool frequency and consistency swiftly improved following the use of pancreatic enzyme replacement therapy.

Background

Immune checkpoint inhibitors (ICIs) have radically changed cancer therapy since the authorisation of ipilimumab in 2011, the first antibody blocking an immune checkpoint (CTLA4). Approval of additional agents followed and interest in this new class of cancer therapy accelerated due to their ability to produce long-lasting responses in numerous malignant tumours that had been previously difficult to treat.¹ Alongside their benefits come a wide range of adverse effects, referred to as ‘immune-related adverse events’ (irAEs). This case provides an example of some of these side effects and highlights the role cognitive bias can play in reaching the correct diagnosis.

Case presentation

A man in his 80s, with a background of mild chronic obstructive pulmonary disease, hypertension and peripheral neuropathy, presented to the emergency department with shortness of breath and calf swelling. A CT pulmonary angiogram revealed bilateral pulmonary emboli and a lower lobe mass, which was diagnosed as an adenocarcinoma of primary lung origin. Histology of tissue taken at an endobronchial ultrasound-guided biopsy found this lesion to be strongly positive for programmed death-ligand 1 (PD-L1). Immunotherapy with pembrolizumab, an anti-PD-1 monoclonal antibody, was started.

Further CT imaging to assess the tumour response showed improvement; however, a week after the patient’s second cycle of pembrolizumab he began to experience widespread muscle and joint discomfort. A brief course of ibuprofen was trialled. On review before his third cycle, the patient reported this did not fully resolve the symptoms and hence a weaning regime of low dose oral prednisolone was started, and the pembrolizumab delayed. After a good response to corticosteroids, treatment with pembrolizumab was resumed.

During a routine review on the day of his sixth cycle of pembrolizumab, 24 weeks into treatment, the patient reported loose stools occurring two times a day over the previous week. This progressed over the next 4 weeks to loose watery stools three to four times per day. This was associated with a general feeling of malaise and poor appetite, but no fevers, abdominal pain, nor mucous or blood per rectum. A working clinical diagnosis of grade 2 ICI colitis was made, defined as an increase of four to six stools per day over baseline. His oral prednisolone dose was increased to 60 mg once daily, and ambulatory medical review was arranged. After senior medical review, the patient was discharged with an increased prednisolone dose of 60 mg once per day as per regional irAE guidelines.

When symptoms persisted despite corticosteroid therapy, the patient was referred to gastroenterology for review.

Investigations

Blood laboratory results at 28 weeks into pembrolizumab treatment at the point of ambulatory care review were as follows: haemoglobin 129 g/L (130–180 g/L), white blood cell count 7.6×10⁹/L (3.50–11.00×10⁹/L), C reactive protein 6.4 mg/L (<5.0 mg/L), sodium 139 mmol/L (133–146 mmol/L), potassium 4 mmol/L (3.5–5.3 mmol/L), urea 6.1 mmol/L (2.5–7.8 mmol/L), serum creatinine 84 µmol/L (59–104 µmol/L), bilirubin 5 µmol/L (<21 µmol/L), alkaline phosphatase 50 U/L (30–130 U/L), alanine transferase 23 U/L (<42 U/L) and thyroid stimulating hormone 1.11 mU/L (0.27–4.2 mU/L). Tissue transglutaminase antibodies were not detected.

A faecal microbiology panel was negative for Clostridium difficile, Cryptosporidium, Giardia, Salmonella, Shigella, Campylobacter, verotoxin-producing Escherichia coli and norovirus genogroup 1 and 2. Microscopic examination identified no ova, cysts or parasites.

Flexible sigmoidoscopy was performed to interrogate the diagnosis of ICI colitis. Normal mucosa was seen, along with five macroscopically benign-appearing polyps in the sigmoid and rectum. A follow-up colonoscopy was arranged. This showed pancolonic diverticular disease and the afore-mentioned polyps, but normal appearing mucosa. Random colonic biopsies showed no histological abnormality.

A faecal elastase was checked 7 months after the last cycle of pembrolizumab was given and found to be 112 µg/g (>200 µg/g).

A CT of the thorax and abdomen 2 months prior to commencing pembrolizumab commented on the normal appearance of the pancreas. A progress CT performed after seven cycles of pembrolizumab commented on an atrophic pancreas. However, these images were reviewed retrospectively by a subspecialist gastrointestinal radiologist for purposes of this report, and it was felt that the pancreas seen in the CT imaging done prior to immunotherapy did in fact appear atrophic, with little change in the subsequent postimmunotherapy scan.

Differential diagnosis

Diarrhoea while being treated with ICIs is common and has a wide differential diagnosis. Transient, self-resolving loose stool alone can be a side effect of treatment, but vigilance must always be kept for a diagnosis of ICI colitis.

Common enteric infections must be ruled out, and in immunocompromised patients opportunistic infections such as Cryptosporidium and cytomegalovirus should be considered.

As in patients not being administered immunotherapy, hyperthyroidism, coeliac disease and diarrhoea-predominant irritable bowel syndrome (IBS-D) should be excluded. Immunotherapy may increase the risk of these diagnoses, with ICI-induced endocrinopathies being well documented² and reported cases of ICI-induced coeliac disease found in the literature.³ The not inconsiderable psychological toll of cancer treatment may be expected to contribute to a disorders of gut–brain interaction, such as IBS-D.

Pancreatic exocrine insufficiency (PEI) is an often overlooked cause of chronic diarrhoea, and loose stool is often accompanied by unintentional weight loss with stools that are described to be pale, greasy and foul-smelling. Diagnosis is clinical, supported by a low faecal elastase.

Treatment

High-dose oral prednisolone 60 mg once daily was initially trialled along local guidelines with the initial diagnosis of grade 2 ICI colitis. Pembrolizumab was appropriately put on hold. Codeine was also used throughout this period to try to reduce stool frequency.

On initial gastroenterology review, a trial of loperamide was suggested to partial effect. A 2-week course of metronidazole was also prescribed in case of small bowel bacterial overgrowth. Probiotic drinks were also suggested.

Following the faecal elastase result suggestive of moderate PEI, the patient was started on pancreatic enzyme replacement therapy (PERT). Rapid clinical improvement was seen within days, with stool frequency reducing from six to three times per day, and stools becoming formed again.

Outcome and follow-up

Following the improvement of the patient’s symptoms with PERT in the absence of pre-existing PEI, a working diagnosis of pembrolizumab-induced PEI was made.

At clinic review 4 months following the diagnosis, the patient had well-controlled bowel symptoms and needed no dose adjustment to his PERT. His lung disease burden was stable on CT with no further pembrolizumab required.

Discussion

Pembrolizumab is a type of antiprogrammed death receptor-1 (anti-PD-1) monoclonal antibody, and more broadly, an ICI. PD-1 is a receptor most commonly expressed on T cells and functions as an inhibitory regulator of T-cell activation. PD-1 has two binding ligands: PD-L1 and PD-L2. Via interaction with PD-1, the immune response can be regulated by carefully balancing T-cell activity.⁴ This interaction is exploited by cancer cells. PD-L1 is often found on tumour cells as a mechanism of reducing antitumour activity by creating a microenvironment more suited for tumour cell survival.⁵ Anti-PD-1 monoclonal antibodies such as pembrolizumab therefore function by antagonising the PD-1/PD-L1 axis, thus promoting the immune response against tumour cells.⁶

ICIs have been associated with various irAEs. Data for pembrolizumab specifically is less prevalent; data for anti-PD-1/PD-L1 related adverse effects is often referred to instead. This includes skin-related adverse effects in ~34% of patients, thyroid disorders in 5%–10% and hepatitis in 5%–10%. Data regarding gastrointestinal toxicity are scarcer, but more established for CTLA4 use, in which diarrhoea (27%–54%) and colitis (8%–22%) are common. Pneumonitis, cardiac, neurological, ocular and renal toxicity have also been noted.⁷

There are many guidelines for the investigation and management of irAEs including those from the European Society for Medical Oncology.⁷ Withdrawal of the offending treatment and rapid initiation of steroid treatment is indicated in most cases including colitis, hepatitis, pneumonitis, rheumatological reactions, dermatological reactions and neuromuscular toxicity.

In the case described in our article, the diagnosis of PEI is clear, with a low faecal elastase and good clinical response to PERT. Although the patient had radiological evidence of an atrophic pancreas prior to immunotherapy, this was an incidental finding with no associated symptoms. The use of pembrolizumab may have therefore triggered symptoms of PEI in an already susceptible patient. It should be noted that the CT reports which had incorrectly suggested that the pancreatic atrophy following immunotherapy was new do not appear to have influenced the decision to start PERT, with no reference to them made when PEI was diagnosed.

There are four other documented cases of PEI presenting following treatment with pembrolizumab. Prasanna et al ⁸ report of a case of isolated PEI following pembrolizumab treatment for metastatic melanoma. The patient had developed increased stool frequency associated with urgency but no bloody diarrhoea nor abdominal pain. The patient was empirically started on a dose of oral prednisolone for possible ICI colitis despite a normal sigmoidoscopy which revealed no macroscopic or microscopic evidence of colitis. Colonoscopy and imaging with CT and MRI were done as the patient did not respond to the treatment. No significant findings could explain the patient’s symptoms, and it was not until faecal elastase levels were found to be low that PERT was commenced and the patient’s symptoms improved.

A second case⁹ reports a patient on cycle 1 of pembrolizumab for metastatic melanoma presenting with unintentional weight loss and increased stool frequency, but no pain or nausea. The patient had previously received three cycles of ipilimumab (anti-CTLA4 antibody), with a 3-week break before starting pembrolizumab. Amylase and lipase were normal in this patient, but faecal elastase was low. A CT scan showed an atrophic pancreas. A diagnosis of autoimmune PEI was made and the patient regained his lost weight following pancrelipase commencement.

A third case¹⁰ reports a patient who experienced steatorrhoea, weight loss and fat-soluble vitamin deficiencies within a year of starting pembrolizumab therapy. This was initially managed as irAE colitis with steroid therapy, but symptoms continued. Faecal elastase levels were noted to be undetectable, which, alongside the patient’s symptoms, led to a diagnosis of pembrolizumab-induced PEI. A colonoscopy later revealed normal appearing mucosa. The diarrhoea resolved following the use of PERT.

A fourth case¹¹ describes a woman who had finished 2 years of pembrolizumab therapy. 8 weeks following this, she presented with pancreatitis, deemed to be ICI induced after more common causes had been excluded. Although the pancreatitis resolved spontaneously, the patient continued to experience abdominal pain following meals, as well as loose stools and weight loss. Faecal elastase level was found to be low, and the patient’s symptoms resolved with introduction of pancreatic enzymes.

The British Society of Gastroenterology guidelines for the investigation of chronic diarrhoea in adults suggest faecal elastase testing when fat malabsorption is suspected.¹² Although there was no suspicion of the latter in our case, nor in three of the four aforementioned cases, the commencement of necessary treatment was made possible in all instances through the investigation of faecal elastase levels.

ICI colitis is a potentially life-threatening side effect of pembrolizumab, which requires prompt recognition and management. In this patient’s case, the appropriate management was commenced in response to the patient’s symptoms, as per local guidelines. It is important to note, however, that the widely adopted Common Terminology Criteria for Adverse Events has been shown to be inadequate to diagnose and assess the severity of ICI colitis in isolation.¹³ This case highlights that the threshold should be low for pursuing investigations, including endoscopy, to exclude other causes of diarrhoea in patients being administered ICIs.

Cognitive bias may have played a small role in delaying the correct diagnosis. For example, ‘satisfaction of search’ refers to the phenomenon whereby further investigations or thinking about alternative diagnoses ceases, as the first plausible solution has been found. This initial diagnosis may be influenced by confirmation bias, which refers to seeking out and interpreting information that confirms a clinician’s preconceived diagnosis. Cognitive errors are thought to be responsible for up to 75% of errors in internal medicine, with cognitive biases contributing to this. Metacognition, an insight into one’s own thinking, has been shown to help combat this.¹⁴

Learning points.

• Pembrolizumab-induced pancreatic exocrine insufficiency is a rare complication of pembrolizumab therapy. It is easy to diagnose and easy to treat. The role of immune checkpoint inhibitors is growing and clinicians in all fields are more likely to come across patients experiencing immune-related adverse events. Hence, an awareness of their clinical presentations, how to investigate and manage them, and when to involve other specialities, is vital.

• Although the patient had no classic symptoms suggestive of fat malabsorption (other than the diarrhoea itself), faecal elastase testing is a relatively easy investigation that should at least be considered in patients presenting with chronic diarrhoea once more routine blood results and stool tests have been conducted.

• Cognitive biases can impact patient care and an awareness of the types that exist is an important part of combatting them. Metacognition can also help. An example of this is asking oneself questions such as ‘what else could this be?’.

Ethics statements

Patient consent for publication

Consent obtained directly from patient(s).