We conducted an institutional review board-approved retrospective review of the cases of 119 patients (84 men, 35 women; median age, 59 years; range, 28–87 years) with advanced metastatic melanoma who had undergone anti–CTLA-4 therapy. The study group included 38 patients who received CP-675,206 (tremelimumab, Pfizer) as part of a phase I/II multiinstitutional clinical trial [4] between 2003 and 2004 and 81 patients treated with MDX-010 (ipilimumab, Bristol-Myers Squibb) according to an expanded-access protocol from 2007 through 2009. Clinical records of the patients were reviewed for age, sex, dosage, frequency and duration of treatment, and pathologic result after biopsy of the suspicious lesions during the clinical study. All images obtained at baseline and throughout the period of treatment were reviewed.

Tremelimumab was administered to 11 patients at a dose of 15 mg/kg every 12 weeks. The other 27 patients started with a lower dose of tremelimumab (10 mg/kg for 21 patients, 6 mg/kg for two patients, 3 mg/kg for four patients) every 4 weeks for up to 12 doses. Survivors were continuously treated with 15 mg/kg of tremelimumab every 12 weeks during the maintenance phase. Ipilimumab was administered at 10 mg/kg every 3 weeks during the induction period (four doses) and then every 12 weeks during the maintenance phase. Both agents were administered as long as patients continued to benefit from therapy.

All study patients underwent contrast-enhanced CT of the chest, abdomen, and pelvis and MRI of the brain at treatment initiation (baseline) and after each treatment cycle. Contrast-enhanced CT of the neck was included if metastatic lesions were present in the neck. Twenty-eight patients underwent whole-body ¹⁸F-FDG PET/CT during treatment, but this study was not part of the routine imaging protocol.

For the current study, an experienced radiologist reviewed all cross-sectional images of the 119 patients, including the CT and PET/CT images obtained during treatment and the reports of these studies, and made comparisons with the baseline images to identify radiologic manifestations of immune-related adverse events. New radiologic findings during treatment were considered to represent a radiologic manifestation of an immune-related adverse event if they were in concordance with clinically evident immune-related adverse events or were biopsy proven as benign. In the absence of biopsy proof, new findings were con sidered to represent radiologic manifestations of immune-related adverse events when they had a nonmasslike appearance suggestive of inflam mation and had resolved at subsequent studies. New lymphadenopathy was considered benign and related to a radiologic manifestation of an immune-related adverse event when it was symmetric bilateral mediastinal and hilar and resembled typical sarcoidosis without evidence of infection.

In all 119 patients, the size and number of target lesions were compared between baseline at study entry and the end of the study or the most recent available examination if treatment was continued at the time of the review. The revisited Response Evaluation Criteria in Solid Tumors (version 1.1) were used to assess response [5]. The responses were categorized into progressive disease, stable disease, partial response, and complete response. We combined stable disease, partial response, and complete response into a category of controlled disease. Partial and complete responses based on baseline target lesions with new solitary metastasis were also included in the controlled disease category because the clinical course remained stable after resection of new metastatic lesions with no re-growth of baseline lesions or appearance of additional new sites of disease.

The association between responses and radiologic manifestations of immune-related adverse events and between the responses and the therapeutic agents (ipilimumab versus tremelimumab) were evaluated with the Fisher exact test. The influence of age and duration of treatment on the response was evaluated with the Wilcoxon signed rank test.

Among 119 study patients, 20 patients (16.8%) (11 women, nine men; mean age, 62 years; range, 32–77 years) were found to have radiographic abnormalities potentially explained by immune-related adverse events (Table 1). Clinically obvious radiologic manifestations of immune-related adverse events included colitis in six patients (5.0%), hypophysitis in two patients (1.7%), arthritis in four patients (3.4%), and thyroiditis with severe ophthalmopathy in one patient (0.8%) (Table 2). The duration of anti–CTLA-4 treatment was 2–118 months (median, 28 months). The radiologic manifestations were noticed 2–26 months (median, 6.3 months) after treatment was initiated. Nine patients (45%) had evidence of an adverse event after the first cycle of therapy. Three patients (15%) had more than one radiologic manifestation at the same time. Anti–CTLA-4 treatment was terminated in two patients because of severe toxicity, including colitis in one case and thyroid ophthalmopathy in the other. Because of the clinical benefit of anti–CTLA-4, therapy was continued in 18 patients. In 11 patients, the radiologically evident immune-related adverse events resolved completely within 6 months despite continued anti–CTLA-4 treatment. In nine patients the adverse effect decreased slowly but persisted through the entire course of therapy (Table 3).

The responses to anti–CTLA-4 therapy are summarized in Table 1. A disease control rate of 18% was achieved for the entire group of 119 patients; the complete response rate was 7%. Twenty patients with radiologic manifestations of immune-related adverse events had a better response: a disease control rate of 55% and a 25% complete response rate versus 10% and 3% among the 99 patients without radiologic manifestations of immune-related adverse events (p < 0.0001) (Fig. 1). Prolonged survival allowed longer treatment of the group with adverse events (median duration, 27 months versus 6 months; p < 0.0001). Neither age (p = 0.04) nor therapeutic agent (p = 0.20) was found to be a factor influencing disease control.

The CT manifestation of colitis was thickening of the colonic wall. PET/CT of one patient without symptoms revealed focal colonic wall edema with focal FDG activity (Fig. 2). Hypophysitis manifested itself as diffuse enlargement of the hypophysis containing ill-defined foci of internal hypoenhancement (Fig. 3). One of the patients with hypophysitis also had severe leg weakness and loss of bladder and rectal sphincter control. CSF analysis showed lymphocytic pleocytosis, but MRI of the brain and thoracic and lumbar spine showed no abnormal leptomeningeal enhancement. Arthritis was seen as increased FDG uptake in the synovia of multiple bilateral peripheral joints on PET images or as joint effusion. One patient with bilateral sacroiliitis had thickening and intensely increased FDG uptake within the abdominal fascia that were suggestive of fasciitis (Fig. 4).

Clinically silent radiologic manifestations suggestive of immune-related adverse events included benign lymphadenopathy in eight patients (6.7%) (Fig. 5), abnormal intramuscular hyperenhancing foci and increased FDG uptake suggestive of myositis in two patients (1.7%) (Figs. 5 and 6), and diffuse opacities in the retroperitoneal fat in two patients (1.7%) (Figs. 7 and 8). The most common pattern of lymphadenopathy was sarcoidlike symmetric bilateral hilar and mediastinal. In three of eight patients, new lymphadenopathy was prospectively interpreted as a metastatic lesion, but the biopsy finding was benign, including one case of sarcoidosis. The other five cases were chest lymphadenopathy interpreted as benign on the basis of the symmetric pattern. Anti–CTLA-4 therapy was continued for all patients with asymptomatic radiologic manifestations of immune-related adverse events. In all eight patients, the area of mediastinal and hilar lymphadenopathy gradually decreased during continuing anti–CTLA-4 therapy, retrospectively suggestive of a benign condition.

Most adverse events associated with anti–CTLA-4 therapy are immune related and are found in more than 70% of patients [3]. Most immune-related adverse events are mild (grades 1 and 2), severe events (grades 3 and 4) occurring in only 25.2% of cases [1, 3]. The most common immune related adverse event is dermatitis (47–68%), which is not detectable on images [3]. Others are colitis, hypophysitis, hepatitis, uveitis, pancreatitis, and thyroiditis [2, 3, 6].

In our study we found that 16.8% of patients with metastatic melanoma treated with anti–CTLA-4 had abnormal radiologic findings of immune-related adverse events. Because they represent a clinically and radiologically heterogeneous group of findings, we discuss the radiologic manifestation of immune-related adverse events individually.

Enterocolitis is the second most common immune-related adverse event among patients undergoing anti–CTLA-4 therapy, reported in as many as 44% of patients [3, 4]. Diarrhea, mostly mild or moderate, tends to resolve with steroid therapy and commonly requires no medical intervention [3]. Colonic biopsy shows a dense lymphocytic, neutrophilic, or mixed infiltrate in both normal- and abnormal-appearing mucosa. Cellular infiltration is more dense in the cryptae mucosa, and there is evidence of cryptitis [7]. Only 5% of patients in our study (6 of 119) had CT evidence of a thickened colonic wall, suggesting that in most cases of clinically evident enterocolitis, the bowel appears normal on images.

Hypophysitis induced by anti–CTLA-4 treatment is reported in the literature to occur in 1–6% of patients [3]. It occurred in 1.7% of patients in our study (2 of 119). Hypophysitis as an immune-related adverse event is similar to sporadic lymphocytic hypophysitis, which occurs primarily in women during late pregnancy and the postpartum period. The typical MRI findings of lymphocytic hypophysitis are marked enlargement of the hypo physis and ill-defined foci of internal hypoenhancement [8, 9]. It is conceivable that T cells infiltrate the hypophysis, though autoantibodies may also have a role. Although radiologic improvement revealing shrinkage of the enlarged hypophysis to its normal size happens within a few days of steroid therapy, the clinical amelioration of hypopituitarism is usually slow [8]. In our study, one patient with hypophysitis experienced debilitating leg weakness and loss of sphincter control despite normal findings at MRI of the brain. Yang et al. [10] described a patient with ipilimumab-induced aseptic meningitis who had headaches, photophobia, and mild cranial nerve dysfunction with lymphocytosis in the CSF leukocyte count (80% lymphocytes) and normal brain MRI findings. The patient in our study might have had aseptic meningitis or arachnoiditis caused by lymphocytic infiltration of the lumbosacral plexus.

Immunotherapy-induced thyroiditis is rare. Hodi et al. [11] reported a 1.6% incidence of ipilimumab-induced hypothyroidism in a phase III study. In our study, one patient had Graves disease with severe ophthalmopathy. Thyroiditis is probably caused by cytotoxic T-lymphocyte infiltration. Factors that make the pituitary and thyroid glands more vulnerable than other endocrine glands to cytotoxic T-lymphocyte infiltration are unclear.

Inflammatory changes in multiple joints found at PET in patients in this study probably reflected an autoimmune cellular attack on the synovia. PET/CT of a patient with clinical evidence of fasciitis showed the abdominal fascia had very high FDG activity consistent with an inflammatory reaction related to synovitis. The autoimmune nature of synovitis is supported by a single case of lupus nephritis occurring during ipilimumab treatment [12]. Although results of rheumatoid tests, including rheumatoid factor and antinuclear antibody titer, were negative, this finding was not surprising because the potential autoimmune reaction was a T cell rather than an antibody-mediated mechanism. Because only 28 of the 119 patients (24%) in this study underwent PET/CT (Table 1), the incidence of synovitis might have been underestimated.

Benign sarcoidlike lymphadenopathy induced by an immune response to ipilimumab was anecdotally reported by Eckert et al. [13]. They described one patient with amelioration of liver metastasis who had new chest lymphadenopathy and skin lesions. Skin biopsy showed sarcoidosis. In our study, we found this type of lymphadenopathy in 6.7% of patients (8 of 119), including one patient with biopsy-proven sarcoidosis. Sarcoidlike benign granulomatous inflammation causing benign lymphadenopathy has been described in rheumatoid arthritis patients treated with anti-tumor necrosis factor [14, 15]. There also have been reports of sarcoidosis in hepatitis C patients treated with ribavirin and interferon α [16, 17]. Exaggerated T-cell–mediated cellular immunity is thought to play a role in the granulomatous inflammation of sarcoidosis [13]. The inflammation pattern in the lymph nodes can be granulomatous or nonspecific. New lymphadenopathy in a patient with known disseminated malignancy is often believed to represent a new metastatic site. In the care of melanoma patients undergoing anti–CTLA-4 treatment, radiologists should be aware of the possibility of benign lymphadenopathy. Differentiation between benign and malignant lymph nodes can be challenging. FDG PET is usually not helpful, and the findings can be misleading in characterizing lymph nodes because melanoma metastasis and reactive lymph nodes both tend to be hypermetabolic. The enhancement pattern of lymph nodes on CT images also is not specific. Iodine contrast enhancement tends to be intense in inflammatory lymph nodes, but it can be variable in metastatic lesions of melanoma, depending on the degree of natural or treatment-related necrosis. Indeterminate lymph nodes should still be biopsied whenever is critical for therapeutic decision making.

Transient intramuscular abnormalities consisting of diffusely increased FDG uptake in the muscles and hyperenhancing intramuscular foci on CT images were believed to be inflammatory in nature, probably representing myositis. This finding may be a variety of inclusion body myositis, which is the most common form of acquired myopathy related to T-cell–mediated autoimmunity [18]. Cases of inclusion body myositis have been described in rheumatoid arthritis patients who received anti-tumor necrosis factor. Hunter et al. [19] described one melanoma patient in whom autoimmune inflammatory myopathy developed after treatment with ipilimumab.

Opacity in the retroperitoneal fat is a poorly understood radiographic finding. However, based on the mechanism of ipilimumab action augmenting T-cell action, retroperitoneal fat opacity may be related to lymphocytic infiltration. Several proliferative neoplastic and nonneoplastic entities, including lymphoma, Castleman disease, Erdheim-Chester disease, and retroperitoneal fibrosis, are known to induce production of similar abnormalities in the perirenal and retroperitoneal fat [20]. Although the pathophysiologic mechanism of retroperitoneal fibrosis is believed to be related to a hyper–immunoglobulin 4–mediated autoimmune process similar to autoimmune pancreatitis [21], other diseases such as the foregoing have cellular immunity mechanisms of a neoplastic or nonneoplastic nature. The same process of unleashed T-cell response is probably responsible for the retroperitoneal opacities visualized in the patients in the current study. Soft-tissue findings with a diffuse infiltrative pattern tend not to be misinterpreted as metastatic melanoma, yet some of them with a more focal pattern can mimic metastasis.

The biologic mechanism of response to immunotherapy differs greatly from that of traditional cytotoxic chemotherapy. Therefore, the revised Response Evaluation Criteria in Solid Tumors may not be perfect for evaluation of immune-related responses. Wolchok et al. [22] described the following four types of responses associated with favorable survival: shrinking baseline target lesion with no new lesions, durable stability with subsequent slow gradual decline in tumor burden, initial increase in tumor burden with subsequent shrinkage, and shrinkage of baseline target lesions with new interval lesions. All four types of responses can be unified into one general category of controlled disease. In previous clinical trials, the disease control rate was 20–24% with no or grade 1 immune-related adverse events versus a 34–43% disease control rate for patients with grade 2 or higher immune-related adverse events. This finding suggests an association between immune-related adverse events and the clinical benefit of therapy [3, 23]. In our study, the disease control rate was 18% for the entire group but 55% in the group with radiologic manifestations of immune-related adverse events and only 10% in the group without immune-related adverse events. The stronger association found in our study shows the importance of radiologic manifestations of immune-related adverse events, even when clinically silent.

The current study had limitations, including retrospective character, single-observer review, a small group of patients, treatment with two different monoclonal antibodies to the same target, different treatment regimens, and a not entirely uniform set of imaging modalities and timing of imaging. Many more patients with radiologic manifestations of immune-related adverse events underwent PET/CT than did those without such manifestations (65% versus 15%) because they were treated with anti–CTLA-4 antibodies for a longer period of time and underwent more imaging studies during therapy. This difference might have induced bias in assessment of the prevalence of radiologic manifestations of immune-related adverse events because such adverse effects on soft tissues, such as synovitis and myositis, are better appreciated on PET/CT than on contrast-enhanced CT images. The main limitation was the lack of pathologic confirmation in most cases. In 2.5% of our study patients (3 of 119), new lymphadenopathy and hyperenhancing intramuscular foci were interpreted initially as metastatic. All of those lesions had resolved by the time of our review. However, the similarity in presentation of these cases allowed us to extrapolate the initial presumption of radiologic manifestations of immune-related adverse events without pathologic proof. Our study was focused specifically on radiologically evident immune-related adverse events and did not include adverse events that were clinically evident but not radiologically evident. As such, the beneficial effect of detecting the total spectrum of immune-related adverse events was probably underestimated.

The radiologic findings in patients who have undergone anti–CTLA-4 antibody therapy, described as radiologic manifestations of immune-related adverse events, are associated with improved clinical response and disease control rates. In the era of developing immune checkpoint-targeted therapies for melanoma, radiologists should be alert to the possibility of radiologic manifestations of immune-related adverse events to avoid the pitfalls of these misleading radiologic findings. Oncologists should be fully aware of clinically silent radiologic manifestations of immune-related adverse events because a variety of adverse events are associated with the response to anti–CTLA-4 antibody therapy.

Supported in part by the National Institutes of Health through M. D. Anderson Cancer Center support grant CA016672.