To clarify the clinical features of endocrine-irAEs, we conducted a prospective study to analyze irAEs in patients treated with ICIs since November 2, 2015 (UMIN000019024). All patients with NSCLC or MM who started ICI treatment between November 2, 2015 and August 30, 2019 at Nagoya University Hospital were included in this study (figure 1). Patients who had more than two malignancies were excluded. Written informed consent was obtained from all patients. Ipilimumab was administered at 3 mg/kg every 3 weeks for 4 cycles. Nivolumab was administered at 3 mg/kg or 240 mg every 2 weeks, with the exception of some patients with MM, who were treated with 2 mg/kg every 3 weeks. Pembrolizumab was administered at 2 mg/kg or 200 mg every 3 weeks. Atezolizumab was administered at 1200 mg every 3 weeks. All ICI treatment was continued until disease progressed, death, or unacceptable severe adverse events occurred, or if patients withdrew consent for treatment.

To examine endocrine-irAEs, ACTH, cortisol, free T3 (FT3), free T4 (FT4), thyroid-stimulating hormone (TSH), and blood glucose levels were assessed at baseline and every 6 weeks after the first administration of ICI for 24 weeks as previously described.19 After the initial 24 weeks, pituitary hormones were measured if clinically needed until the visits stopped. Each endocrine-irAE was defined according to the Japan Endocrine Society clinical guidelines for endocrine-irAEs.12 The cut-off for the diagnosis of ACTH deficiency was decreased peak serum cortisol value (<18 µg/dL) and impaired responses of ACTH (<2-fold of baseline) in corticotropin-releasing hormone loading tests (intravenous injection with 100 µg human corticorelin),20 which were performed in the morning. The frequency of hyponatremia, defined as a serum sodium concentration <135 mEq/L, at the onset of pituitary-irAE was compared with that for patients who did not develop pituitary-irAE. In the latter group, the frequency was evaluated around the median time of onset of pituitary-irAE induced by each ICI (ipilimumab, 71 days; nivolumab, 171 days; pembrolizumab, 127 days). OS was determined for participants in this study until death from any cause. Patients with MM were excluded from the OS analysis if they were treated with ICIs as an adjuvant therapy (figure 1). Non-endocrine-irAEs, as reported in the guideline,13 were diagnosed by attending physicians and subjected to analysis if the grade was higher than 1. All irAEs were monitored and graded using CTCAE 5.0 criteria. Fifty-eight patients treated with nivolumab and 78 patients treated with pembrolizumab were included in our previous study that analyzed the incidence of thyroid dysfunction.19 21

Continuous variables of patient characteristics (eg, age, days to diagnosis) are expressed as means±SD or median and IQR. Differences among continuous variables were tested for significance with the two-sample t-test. Values of nominal variables were compared using Fisher’s exact test. OS was analyzed using the Kaplan-Meier method, expressed as median and 95% CI and compared using a log-rank test. All statistical tests were two sided, and significance was defined as a p value of <0.05. All statistical analyses were performed with IBM SPSS Statistics V.25.

A total of 174 patients with either NSCLC or MM who started ipilimumab, nivolumab, pembrolizumab, or atezolizumab therapy were enrolled in this study (figure 1). Clinical variables for patients with NSCLC and MM are presented in table 1. IrAEs occurred in 34/108 patients (31.5%) with NSCLC and 38/66 patients (57.6%) with MM (table 1). Endocrine-irAEs including pituitary-irAE, thyroid-irAE, and T1DM occurred in 13/108 (12.0%) patients with NSCLC and 20/66 (30.3%) patients with MM. The number of each type of irAE observed, including pituitary-irAE, thyroid-irAE, T1DM, interstitial pneumonitis (lung-irAE), skin toxicities (skin-irAE), and gastrointestinal toxicities (GI-irAE), is shown in table 1. No cases were treated with a reduced dose of ICI due to the development of irAEs.

Pituitary-irAE occurred in 6/25 patients (24.0%) during treatment with ipilimumab, 10/167 patients (6.0%) during treatment with nivolumab or pembrolizumab. Pituitary-irAE did not occur during treatment with atezolizumab (0/9). Pituitary-irAE induced by aPD-1 Ab, nivolumab, or pembrolizumab was observed in 6/59 patients with MM and 4/108 patients with NSCLC. There was no difference in the frequency of pituitary-irAE induced by aPD-1 Ab between patients with MM and those with NSCLC (10.2% (6/59) in patients with MM vs 3.7% (4/108) in patients with NSCLC, p=0.168). Among the 16 patients who developed pituitary-irAE, three demonstrated clinical characteristics of hypophysitis such as pituitary enlargement and deficiencies in multiple anterior pituitary hormones including ACTH (online supplementary table S1). These three patients presented with headache as an initial symptom that could be caused by enlargement of the pituitary glands at irAE onset. A representative MRI showing an enlargement of the pituitary gland at hypophysitis onset is shown in figure 2A. This enlargement was completely resolved at 6 months after onset (figure 2B) and did not change during a 1-year follow-up (data not shown). The other 13 patients developed isolated ACTH deficiency (IAD), but none exhibited enlarged pituitary glands at IAD onset (online supplementary table S1 and figure 2C). Re-evaluation of pituitary gland size 6 months and/or 1 year after IAD onset for 11 of these patients showed no change in size for seven patients (representative MRI image, figure 2D) and a slightly decreased size for the other four patients. In terms of clinical characteristics of patients who developed hypophysitis and IAD, the usage rate of aCTLA-4 Ab and incidence of pituitary enlargement were significantly higher for patients who developed hypophysitis than those who had IAD (table 2). The time to diagnosis of pituitary-irAE after the first ICI administration was significantly shorter in patients with hypophysitis compared with those with IAD (table 2). There were no significant differences in the other clinical variables we examined (eg, malignancy type, age, sex, and history of prior immunotherapy) (table 2). ACTH deficiency was observed in all cases with pituitary-irAE (online supplementary table S2), and these patients were treated with physiological doses of hydrocortisone (10–20 mg/day). ICI was re-administered with the same dose initially used after general conditions were stabilized by hormone replacement therapy. During the observation period of this study, no patients discontinued hormone replacement therapies (hydrocortisone and/or levothyroxine). The frequency of hyponatremia was significantly higher in patients who developed pituitary-irAE than those who did not (6/16 (37.5%) vs 10/120 (8.3%), p<0.01).

Thyroid-irAE occurred in 1/25 patients (4.0%) during treatment with ipilimumab and in 18/167 patients (10.8%) during treatment with nivolumab or pembrolizumab. The incidence of thyroid-irAE tended to be higher in patients treated with aPD-1 Ab than those who received aCTLA-4 Ab, although the difference was not significant (18/167 (10.8%) vs 1/25 (4.0%), p=0.477). The number of patients who developed destructive thyroiditis, hypothyroidism, and hyperthyroidism was 13, 5, and 1, respectively. All five patients who developed hypothyroidism and 12/13 who developed destructive thyroiditis required hormone therapy with levothyroxine.

OS was significantly longer for both patients with NSCLC and MM who developed any irAE than those without irAEs (NSCLC: not reached vs 378 (95% CI 220 to 536) days, p<0.01, figure 3A, online supplementary table S3; MM: 534 (95% CI 245 to 823) vs 189 (95% CI 52 to 326) days, p<0.05, figure 3B, online supplementary table S3). To examine the contribution of endocrine-irAEs to increased OS, 13 patients with NSCLC and 14 patients with MM who developed endocrine-irAEs were included in OS analysis. OS was significantly longer for patients with NSCLC who developed endocrine-irAEs than those who did not (not reached vs 401 days (95% CI 240 to 562), p<0.01, figure 3C, online supplementary table S3). In patients with MM, there were no significant differences in OS between patients who developed endocrine-irAEs and those who did not (649 days (95% CI 367 to 931) vs 342 days (95% CI 131 to 553), p=0.107, figure 3D, online supplementary table S3).

Four patients with NSCLC and nine patients with MM with pituitary-irAE were included in the analysis of OS. OS was significantly longer for both patients with NSCLC and MM who developed pituitary-irAE than those who did not (NSCLC: not reached vs 441 days (95% CI not calculated), p<0.05, figure 3E, online supplementary table S3; MM: 885 days (95% CI 434 to 1336) vs 298 says (95% CI 84 to 512), p<0.05, figure 3F, online supplementary table S3). There were no significant differences between patients with NSCLC (online supplementary table S4) or MM (online supplementary table S5) who developed pituitary-irAE and those who did not in terms of sex, drugs used, prior history of other ICIs, and treatment lines, although the mean ages were significantly younger and the ipilimumab usage rate was significantly higher for patients with MM who developed pituitary-irAE compared with those who did not (online supplementary table S5). The analysis of OS included patients treated with ICI monotherapy as well as sequential use of several ICIs because the number of patients who were treated with only one ICI during the study period was too small (online supplementary table S6). There were no significant differences in the usage rate of multiple ICIs between patients who developed pituitary-irAE for patients with either NSCLC (0/4 with pituitary-irAE vs 8/104 without pituitary-irAE. p=1.000) or MM (5/9 with pituitary-irAE vs 20/43 without pituitary-irAE. p=0.722). There were no differences in the number of cycles of ICI between patients who developed pituitary-irAE and those who did not, except for patients with MM treated with nivolumab (online supplementary tables S7 and S8). In this group, the number of cycles was significantly higher in the pituitary-irAE group (online supplementary table S8). The time of follow-up was significantly longer for the patients with NSCLC and MM who developed pituitary-irAE than those who did not (online supplementary tables S7 and S8).

OS was significantly longer for patients with NSCLC who developed thyroid-irAE than those who did not (not reached vs 441 days (95% CI 317 to 565), p<0.05, figure 3G, online supplementary table S3). In contrast, patients with MM who developed thyroid-irAE and those who did not had similar OS (427 days (95% CI 0 to 871) vs 393 days (95% CI 219 to 567), p=0.875, figure 3H, online supplementary table S3). There were no significant differences between patients with NSCLC (online supplementary table S9) or MM (online supplementary table S10) who developed thyroid-irAE and those who did not for other clinical factors including sex, drugs used, prior history of other ICIs, and treatment lines. Moreover, there were no significant differences in the usage rate of multiple ICIs between patients who developed thyroid-irAE and those who did not among patients with NSCLC (0/8 with thyroid-irAE vs 8/100 without thyroid-irAE; p=1.000) or MM (2/7 with thyroid-irAE vs 24/45 without thyroid-irAE; p=0.419).

Although patients who had any irAE showed longer OS than those without irAEs, the development of non-endocrine-irAEs was not associated with longer OS in patients with either NSCLC or MM (NSCLC: 585 days (95% CI 321 to 849) vs 441 days (95% CI 297 to 585), p=0.247, online supplementary figure 1A, table S3; MM: 534 days (95% CI 268 to 800) vs 241 days (95% CI 0 to 523), p=0.268, online supplementary figure 1B, table S3). As for each non-endocrine-irAE, there were no differences in OS between patients who developed lung-irAE, skin-irAE, or GI-irAE and those who did not for patients with NSCLC (lung-irAE, 488 days (95% CI 167 to 809) vs 474 days (95% CI 334 to 614), p=0.710; skin-irAE, 669 days (95% CI 438 to 900) vs 441 days (95% CI 310 to 572), p=0.550; GI-irAE, not reached vs 481 days (95% CI 351 to 597), p=0.731) or MM (lung-irAE, 1259 days (95% CI 0 to 2655) vs 393 days (95% CI 237 to 549), p=0.202; skin-irAE, 649 days (95% CI not calculated) vs 379 (95% CI 252 to 506), p=0.560; GI-irAE, 491 days (95% CI 152 to 830) vs 393 days (95% CI 174 to 612), p=0.581) (online supplementary figure 1C‒H, table S3). For skin-irAEs, exanthema and erythema, but not vitiligo, were observed in this study.