Immune checkpoint inhibitor induced hypophysitis: a specific disease of corticotrophs?

Introduction

The immune checkpoint inhibitors (ICIs) are a group of relatively new biological agents that have revolutionised therapy for a group of cancers, including melanoma, renal and some lung tumours. A side effect of these treatments is a high incidence of organ-specific autoimmune diseases. Consequently, we have seen a rise in immune-mediated endocrine disease (1, 2, 3, 4). Thyroid dysfunction is common (5). Beta cell destruction with a disease similar to type 1 diabetes is much less common (6, 7). In addition, this group of patients has an increased frequency of pituitary disease – hypophysitis with a much smaller group presenting with primary adrenal disease. Unlike other causes of hypophysitis, the literature suggests that this condition is confined to the anterior pituitary and primarily involves glucocorticoid deficiency (8, 9, 10, 11, 12, 13, 14). Published evidence of pituitary disease is restricted to case reports with a small number of case series.

There are three main groups of ICIs in use: cytotoxic T-lymphocyte-associated protein 4 inhibitors (CTLA-4), programmed cell death protein-1 inhibitors (PD-1), and programmed cell death ligand inhibitors (PDL-1). The use of these agents is evolving. The three groups may be used individually or in combination (15). At present, it is not possible to accurately define the risk of developing hypophysitis for each individual medication or combination.

In this study, we identified individuals presenting with clinical features of glucocorticoid deficiency, who were confirmed to have pituitary disease through formal pituitary testing, and subsequently underwent pituitary imaging with dedicated MRI. The cohort was followed annually with repeat pituitary assessments. The aim of the study was to define functional and anatomical pituitary disease at the time of presentation following ICI therapy and to describe any changes in pituitary function over time.

Methods

We conducted a retrospective observational audit of patients on ICI therapy in our centre between January 2013 and September 2023. We reviewed all the patients on ICI therapy under the care of the oncology department at University Hospital Plymouth, a secondary care hospital with 1000 beds serving a population of 500,000 people. From this group, we identified all individuals referred to the endocrinology department with a suspected diagnosis of adrenal insufficiency. Patients were established on adrenal steroid replacement and subsequently underwent formal pituitary testing. The majority of people underwent short Synacthen tests and, at the same time, a 09:00 h assessment of the other anterior pituitary axes. A small number were assessed using the insulin stress test; if necessary, a non-stimulated 09:00 h cortisol was accepted if the cortisol was significantly low (<100 nmol/L). Individuals were asked to withhold their morning dose of hydrocortisone until after the assessment. No other glucocorticoids were being taken at the time of the test. Tests were performed in the endocrine department by the specialist endocrine team and were transferred immediately to the analysing laboratory. Prior to performing the test, the patient would be contacted to confirm the appointment. At the time, the patient would be advised to omit steroids on the morning of the test. On the day of the test, if the patient had taken any inhalers or applied any steroid creams or eye drops, the test was cancelled and re-arranged for another date. Hypersensitivity reactions to Synacthen may occur rarely. After establishing if any specific instructions were discussed at the endocrine multi-disciplinary team meeting, the procedure was explained to the patient, ensuring they were aware of potential side effects. The patient’s verbal consent for the procedure was obtained, with anaphylaxis shock equipment and drugs ready in case of an allergic response to Synacthen. A 09:00 h timed basal blood sample to measure cortisol was first taken along with samples for additional tests if required. A 09:00 h Synacthen 250 mcg injection was then administered deep into the deltoid muscle or IV via a cannula, and 30-min and 60-min samples post-Synacthen injection were collected for stimulated cortisol level measurements.

An increase in cortisol by at least 200 nmol/L to at least 440 nmol/L at 30 min or 60 min would be used to exclude cortisol insufficiency. Individuals were questioned about the possibility of arginine vasopressin (AVP) deficiency (diabetes insipidus), although formal testing for this was not required in any individual. Individuals were included in the cohort if they had endocrine or radiological evidence of pituitary disease consistent with hypophysitis and/or a low ACTH measurement. A cohort of 37 people with evidence of pituitary failure was included in the analysis.

The clinical notes of this group were then studied to establish the timing of the onset of symptoms and the clinical presentation. For the majority, it is not possible to give an exact date for the onset of disease as symptoms are, for most people, insidious. For example, headache was present in a considerable proportion of the cohort, but it is not clear if this represented the onset of pituitary inflammation or was secondary to dehydration and hypotension associated with established cortisol lack. The month of onset can be given with some confidence but the timing of endocrine testing and MRI imaging in relation to the onset of disease cannot be given with any accuracy.

Brain MRI imaging was performed on 32 of the individuals. For a further five, the MRI imaging did not provide dedicated pituitary slices and was not suitable for inclusion.

The study was registered with the audit department at Plymouth Hospitals NHS Trust CA_2022-23-285. Ethical review and consent were not required for this study.

Results

Between January 2013 and September 2023, a total of 954 people received ICI therapy. These medications were used to treat a variety of cancer types and were combined with various other chemotherapeutic agents.

Melanoma was the most common cancer treated. In that time, 41 people developed adrenal insufficiency. Two people had raised ACTH but had evidence of other adrenal disease (adrenal metastases). Two people had evidence of tumours within the pituitary. The remaining 37 individuals had evidence of pituitary disease which had developed after treatment with ICIs. This gives a prevalence of pituitary hypophysitis leading to hypocortisolism of 3.9% for the cohort following immunotherapy. The majority of these patients were confirmed to have low ACTH. No cases of primary autoimmune adrenal disease following immunotherapy were found in this cohort. Since the development of pituitary disease, the cohort has been followed for a median of 35 months, with a range of 6–68 months. Ten people have died during follow-up.

Table 1 describes the cancer diagnoses and checkpoint inhibitor treatment for the 37 individuals diagnosed with hypophysitis. The median age of the population that developed hypophysitis was 65 years. Thirty per cent of affected patients were female, while the remaining 70% were male. Sixty-five per cent of the total patients affected were treated for a primary diagnosis of malignant melanoma involving various sites of the body. One additional person had received treatment at another hospital with dual agents and later moved to our area, where they were followed up. This individual was not included in the analysis. The majority of patients receiving ICI (86.5%) had metastatic disease. Fifty-seven per cent patients were steroid naive at the time of testing. Thirty-five per cent patients received steroids prior to testing for cortisol insufficiency. Of these, 84 per cent received steroids for metastatic disease, eight per cent for drug-induced side effects and eight per cent for symptoms suggesting acute hypophysitis. For these patients, a Synacthen test was done after their tapering steroid regimen was complete, and steroid was stopped or tapered to physiological doses. In one case with suspected acute hypophysitis features, the morning glucocorticoid dose was delayed on the morning of the short Synacthen test until after the blood samples were collected.

Twenty-seven per cent patients received single-agent therapy, while 73 per cent had received dual-agent therapy. Of those receiving single agents, seven received a PD-1 inhibitor agent, one received a PDL-1 inhibitor and two patients received a CTLA-4 inhibitor. The patients receiving dual-agent therapy (27) received a combination of a PD-1 and a CTLA-4 inhibitor. Of the 37 patients who received treatment at UHP, the average number of cycles of therapy received was 17.6. Twenty-five patients’ therapy details were available. The average duration of therapy these patients received was 4.34 months.

Table 2 describes the clinical presentation of the cohort, the timing of the onset of symptoms, and symptoms at presentation. Although a small number of patients developed an acute illness with a clearly defined start, for most people, the onset of pituitary disease was insidious. Fifty-one per cent of patients were admitted to the hospital as a consequence of hypophysitis, with one person admitted to intensive care due to the sudden onset of severe headache, confusion, hypotension and acute kidney injury. This was an unusual presentation, with all other patients managed in the general ward or as outpatients. Metastases were the commonest (85%) reason for commencing steroids for the patients who received glucocorticoids. Of the remaining 15%, one patient received glucocorticoid therapy for drug-induced hepatitis, while another had symptoms suggestive of hypophysitis, which resulted in commencing glucocorticoid therapy. The commonest symptoms were fatigue (62%), nausea (30%) and headache (27%). Fatigue (69%), hypotension (39%) and nausea (23%) were the commonest symptoms among the proportion that were on glucocorticoid therapy prior to testing for cortisol insufficiency. As these are common symptoms in this group with a diagnosis of cancer and on a variety of other medications, the diagnosis of pituitary disease was not always straightforward. A proportion were already administered glucocorticoid therapy as mentioned above, and in some cases, the diagnosis was not clear until the glucocorticoids were withdrawn. It is therefore not possible to be extremely specific about the timing of the development of hypophysitis in relation to starting immunotherapy. New onset of confusion (11%) or vomiting (11%), development of hypotensive symptoms (22%) and the finding of hyponatraemia (14%) were less common and almost always led to hospital admission. For this smaller group, the diagnosis was easier.

It is only possible to estimate the onset of symptoms in terms of weeks following the start of treatment. The shortest period was 3 weeks after starting treatment, while the longest interval was 13 months, with a median interval of 4 months. The timing of the onset of the disease did not relate to the severity of illness, with no difference in the timing of onset for the group admitted to hospital compared with those not admitted.

Table 3 displays the pituitary endocrinology assessed by the endocrine service following referral from oncology. Data are then shown for 1, 2 and >3 years follow-up for the cohort. Full pituitary assessment annually was not considered appropriate for a number of individuals. Long-term follow-up data for the complete pituitary profile are therefore not available for the complete cohort. At the initial presentation, low cortisol was universal. A stimulation test was performed for the majority, but where the 09:00 cortisol was <50 nmol/L, a non-stimulated value was accepted. ACTH was measured simultaneously with the cortisol for 28 individuals and was <10 ng/L in all cases. Following repeated testing, there was no recovery in cortisol or ACTH for any individual. The complete cohort that remains alive continues to remain on glucocorticoid replacement. Formal testing of cortisol production was available for ten individuals at >3 years, with no evidence of improvement. Despite a significant effect on the pituitary–adrenal axis, there was less evidence of permanent impairment of other pituitary axes. Primary hypothyroidism was common. This either preceded the development of pituitary insufficiency or developed at the same time. Three people had evidence of primary hyperthyroidism before developing hypothyroidism. One out of these three patients had positive TPO antibodies, suggesting hashitoxicosis. The anti-TSH receptor antibody test was negative. Secondary hypothyroidism, with low TSH and low free thyroxine, was less common. One individual showed improvement in secondary hypothyroidism at 1 year, and one further individual showed development of secondary hypothyroidism at 1 year. Prolactin was mildly elevated in a small number of people at presentation and fell over time. Prolactin was not measured serially in those with elevated levels as the people did not survive. Formal stimulation tests for growth hormone were not performed, but IGF-1 was available as a surrogate and was normal in 26/26 people. Hypogonadotrophic hypogonadism was present in six women – all were post-menopausal, and this was not treated. Two men were found to have slightly low testosterone at presentation with inappropriately normal gonadotrophins. Total testosterone concentrations were 5.1 and 6.2 nmol/L. They did not start testosterone replacement. There were no cases where AVP deficiency was suspected. Copeptin was not measured. In summary, cortisol and ACTH (where measured) were universally abnormal. Other permanent anterior pituitary hormone defects were unusual. Hypophysitis associated with immunotherapy appears to specifically target the corticotrophs, with no evidence of recovery over time.

Table 4 describes the imaging characteristics for the cohort. As the exact timing of the onset of disease is difficult to define, we can only estimate the timing of imaging in relation to the time of onset of disease. For those people admitted to the hospital, imaging was performed during the admission. For those managed out of hospital, imaging may have been delayed, with the longest delay being 27 weeks after the onset of symptoms. There were two main MRI pituitary imaging protocols used. The pituitary protocol involves fine-slice, small field-of-view images with coronal and sagittal views for both T1 and T2, followed by a delayed post-contrast T1 sequence. As there were also concerns about the presence of brain metastases in this cohort, an alternative protocol was used: magnetisation prepared rapid gradient echo (MP RAGE) provides fine-slice, three-dimensional (3D) imaging of the whole brain, including the pituitary, and allows for 3D reconstruction in any plane. A mixture of these two protocols was used. The radiological changes associated with immune-mediated hypophysitis are subtle. Pituitary gland or pituitary stalk enlargement was present in one-third of cases on the initial scan. All of these cases received dual-agent therapy, comprising a PD-1 inhibitor and a CTLA-4 inhibitor. None of the cases showed pituitary enlargement impinging on the optic apparatus. In those that had repeat scans at an interval of 6–12 months, the enlargement had resolved. This is relevant to those who may have had their initial scans sometime after the onset of the disease, as it may have been too late to demonstrate this abnormality. There was a specific abnormality seen in 7 of 27 scans that would appear to be a particular feature of immune-mediated hypophysitis. Figures 1 and 2 show examples of transient and unusual linear abnormalities within the anterior part of the pituitary gland that were not present on repeat imaging. These appear to be confined to the anterior aspect of the pituitary gland and appear as lines or striations within the pituitary. These were not visible on any of the scans performed more than 3 months after the likely onset of the disease.

View Full Size

(A, B) Post-contrast T1 coronal and sagittal slices showing linear abnormalities present in the anterior aspect of the pituitary that were not present on repeat imaging at 6 months. Admitted to the hospital with severe headache, nausea, fatigue and hyponatraemia 12 weeks after starting Ipilimumab and Nivolumab. Pituitary function was normal apart from low cortisol and ACTH.

Citation: Endocrine Connections 13, 11; 10.1530/EC-24-0223

• Download Figure
• Download figure as PowerPoint slide

View Full Size

(A, B) Post contrast T1 coronal and sagittal slices showing linear abnormalities within the anterior part of the pituitary. The patient presented with persistent headache, having had treatment for ocular melanoma, including starting Ipilimumab 8 weeks earlier. Endocrinology showed raised prolactin at 1034 mIU/L, hypogonadotrophic hypogonadism and low ACTH and low cortisol. Imaging and endocrinology were not repeated as she died 5 months later.

Citation: Endocrine Connections 13, 11; 10.1530/EC-24-0223

• Download Figure
• Download figure as PowerPoint slide

For most of this cohort, checkpoint inhibitor therapy was temporarily stopped while the individuals remained unwell, then restarted. In those with more severe symptoms requiring admission, the symptoms mostly resolved in weeks or days. It is not clear how much of this was related to starting glucocorticoid replacement. Detailed symptom questionnaires have not been performed during follow-up, but clinic letters have not documented the continuation of the presenting symptoms. Confusion and vomiting resolved for all patients. Ten of 37 have died during follow-up as a direct consequence of, or complications from their primary cancer diagnosis. There have been no adverse events related to glucocorticoid replacement.

Discussion

We followed a cohort of patients who developed pituitary disease within a short period of commencing ICI cancer treatment. The term used for this is immune-related hypophysitis (IR hypophysitis). We have shown that this treatment results in a specific defect in pituitary function that is confined to the permanent destruction of pituitary corticotrophs. This conclusion is supported by detailed pituitary endocrinology at diagnosis and subsequent follow-up. There was a subset of patients who received glucocorticoid therapy prior to referral to the endocrinology department. For these patients, testing was done after their tapering steroid regimen was complete and had stopped or was at a physiological replacement dose. In these cases, the morning glucocorticoid dose was delayed on the day of the Synacthen test until after blood samples were collected. This was to reduce the possibility of results being influenced by the use of glucocorticoids prior to testing their cortisol axes. We appreciate that some of this could potentially lead to inappropriate assessment of the numbers of patients with adrenal insufficiency or the actual severity of it. Serial testing of the cohort, however, did not show recovery of their cortisol axes with time. The imaging abnormalities seen in a subgroup also support the conclusion that this immune attack is confined to a subtype of pituitary cells rather than the whole gland. Although initial endocrinology evaluation showed a variety of other defects, these resolved with time and would seem to relate to a bystander effect on the rest of the pituitary rather than a direct attack on those cell types that then resolved.

The incidence of pituitary disease has been estimated in a number of cohorts (2, 10, 11, 13). Most of these cohorts do not have detailed measures of pituitary function and conclude that there is an abnormality in either the pituitary or adrenal. Our estimate of 3.9% of cases developing IR hypophysitis is in broad agreement with previous publications although comparison is difficult as case ascertainment and the diagnostic tests used vary considerably between studies. Some studies have reported much higher incidences of hypophysitis, but these have included secondary hypothyroidism and hypogonadotropic hypogonadism at the time of acute illness. These may have been the pituitary response to acute illness rather than autoimmune disease. We found no cases of primary autoimmune adrenal disease in this cohort. Previous publications would suggest that this is correct. In most cases where adrenal disease has been diagnosed, ACTH was not reported. Primary pituitary disease may therefore have been misdiagnosed as adrenal disease.

Symptoms at presentation reported in this cohort commonly include fatigue, headache, nausea, vomiting and symptoms related to hypotension. Previous publications strongly support this, although some have also reported visual symptoms (10, 11, 13). It is not clear whether these symptoms are entirely due to cortisol deficiency or partly due to a rapid change in pituitary size. In this cohort, there was no relation between the degree of pituitary enlargement and the severity of the headache. We would not expect visual problems in this cohort, as none of the imaging showed pituitary enlargement great enough to reach the optic chiasm. Previous publications have reported the onset of hypophysitis symptoms as days since starting immunotherapy, but this is difficult to access as the onset of symptoms is often slow (16). Our data suggest that, for the majority of people, IR hypophysitis develops a short time after starting treatment, with a median time of 4 months after commencing treatment. This is slightly longer than reported in other studies (10, 11, 13). However, this was not universal, one individual presented with hypophysitis 5 years after starting checkpoint inhibitor therapy. For the majority of patients, there was significant improvement in symptoms with glucocorticoid replacement, although this group is complex, with a variety of other medical problems that make symptom interpretation difficult.

The endocrine pituitary assessment and follow-up performed in this cohort is more detailed than most previous publications. The conclusion of most other publications is that ICIs cause permanent glucocorticoid deficiency. Our data support this view and strongly suggest that this is due to a specific disease of corticotrophs. We have not assessed changes in thyroid function for the whole cohort of 954 individuals, but it is clear that thyroid disease following checkpoint inhibitor therapy is the most common endocrine abnormality in this group. Previous studies clearly support this view. Half of this cohort had developed primary thyroid disease either preceding or simultaneous with the development of IR hypophysitis. The universal finding in this cohort is loss of cortisol response with low ACTH where measured. The cohort was followed with serial measurements of cortisol for a median of 3.5 years and a maximum of 6.5 years. There has been no recovery in cortisol production in any subject. Other pituitary dysfunction at presentation was variable with some cases of mildly raised prolactin or hypogonadotropic hypogonadism but these were transient.

Radiological changes associated with IR hypophysitis are variable and subtle (16, 17, 18, 19, 20). There are a variety of types of hypophysitis previously described. These range from lymphocytic hypophysitis to granulomatous hypophysitis to necrotising hypophysitis. These conditions mostly have distinctive radiological characteristics. IR hypophysitis, although perhaps closest to lymphocytic disease in terms of pathophysiology, seems to be a distinct disease as the radiological features are not the same. The most common finding in this cohort was a normal pituitary appearance. Where the pituitary or pituitary stalk was enlarged, this was not significant. The absence of pituitary imaging abnormalities, therefore, does not argue against the diagnosis of IR hypophysitis. When seen, the linear abnormalities in the anterior pituitary in these cases are quite distinctive. There are some previous publications that show this abnormality but, at the current time, this is not universally recognised as a feature of this condition. The timing of scanning in relation to the onset of disease and the timing of the images in relation to when contrast is given may be important and may explain why the pattern is not seen more often. With the numbers we have, we were not able to show significant differences between drug classes or single compared with multiple drugs. Immunohistology descriptions of the placement of the corticotroph cells within the pituitary suggest that they are mainly distributed in the anterior aspect and middle of the pituitary, with growth hormone-producing cells mostly positioned laterally and prolactin-producing cells posteriorly (21). If this immune-mediated disease was specifically targeted at corticotrophs, then the expected imaging characteristics might be confined to the area with the greatest densities of these cells. The fact that the imaging appearances resolve on subsequent scans suggests that these are directly related to IR hypophysitis.

In conclusion, these data show that IR hypophysitis is a common complication of ICI therapy. This results in an imaging abnormality within the areas of the pituitary that are richest in corticotrophs. The endocrine consequence of this is a permanent defect in ACTH production and, therefore, cortisol production.

Declaration of interest

The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the research reported.

Funding

This research did not receive any specific grant from any funding agency in the public, commercial or not-for-profit sector.

Author contribution statement

All authors had an equal contribution in this research.