In this retrospective study on diagnostic accuracy, [18F]FDG-PET/CT provided information that could be used to identify irAEs. Of the 123 enrolled patients, 65 experienced irAEs that either required additional treatment or caused discontinuation of immunotherapy. The incidence of organ-specific irAEs, presented in decreasing order, were: skin 26/65 (40%), muscle and joints 21/65 (32.3%), intestines 13/65 (20%), thyroid gland 12/65 (18.5%), lungs 4/65 (6.2%), and heart 2/65 (3.1%). [18F]FDG-PET/CT detected 48 true-positive cases, with the highest sensitivity in the intestines and the thyroid gland, where the values were 100% (95% CI: 75–100%) and 92% (95% CI: 62–99%), respectively. In contrast, [18F]FDG-PET/CT showed a low sensitivity of 19% in detecting adverse events in the skin (95% CI: 7–39%). In terms of specificity, high values were found in the skin 95% (95% CI: 88–98%) and thyroid gland 95% (95% CI: 89–98%). Muscles and joints had the lowest specificity in this study, with a moderate 83% (95% CI: 75–90%).

In more than half of the patients (56.9%), the irAEs developed before receiving the fourth dose of immunotherapy. According to the Danish guidelines for the follow-up of patients with high-risk melanoma, the first follow-up scan is scheduled to be performed approximately three months after the patient has received the initial dose of immunotherapy [20]. As a result, when the first follow-up scan is performed, the irAE may have been treated, leading to undetectable FDG uptake on [18F]FDG-PET/CT. This is also supported by our results showing that out of the 40 patients who discontinued treatment, immunotherapy was terminated in 18 patients (45%) after receiving a maximum of three doses. Detection of irAEs at an earlier stage could potentially guide clinicians to adjust the treatment earlier, which may result in an increased number of patients completing the desired treatment line.

[18F]FDG-PET/CT is known for its high sensitivity in detecting metastases; however, inflammation from any cause will also represent increased FDG uptake. Consequently, irAEs generated by PD-1 inhibitors may also be detected [32]. Comparing the scans to their respective baseline scans can aid in detecting subtle changes in uptake or density [31]. We recognise that diagnosing colitis on [18F]FDG-PET/CT can be difficult, as normal metabolic activity in the colon might make it challenging to distinguish physiological uptake from inflammation. In asymptomatic patients, increased colonic FDG uptake due to benign causes is common [33, 34]. We found a high sensitivity in detecting irAEs in the intestines (100%); however, the specificity was moderate (85%). These results are supported by a study by Gelston et al., which claims that FDG PET/CT is a valuable imaging method for detecting an increase in gastrointestinal inflammation [35]. According to multiple studies, patients using the oral hypoglycaemic medication metformin have diffusely enhanced tracer absorption in the colon, which may lead to a higher incidence of false-positive cases [31, 36,37,38]. In this study, 4/16 patients (25%) with false-positive results in the intestines had diabetes mellitus and were undergoing treatment with metformin. Consequently, metformin should be discontinued when the glucose analogue FDG is administered to obtain higher specificity in PET imaging [25].

Thyroid dysfunction typically results in diffuse, homogenous, enhanced tracer absorption on [18F]FDG-PET/CT [39], which was also observed in all patients in this study. Thyroiditis can be challenging to diagnose because of its nonspecific symptoms, such as fatigue and headache, and thyroid function tests are recommended for confirmation. Thyroiditis was reported and confirmed by laboratory tests in 12/65 patients (18.5%). The median time of onset after the initiation of immunotherapy was eight weeks (two doses), supported by the findings of another study [40]. In the present study, [18F]FDG-PET/CT was highly accurate in detecting irAEs in the thyroid gland, with sensitivity and specificity of 92% and 95%, respectively. [18F]FDG-PET/CT has already shown perfect performance in detecting permanent thyroid dysfunction in patients with melanoma treated with PD-1 inhibitors [41], corroborating the findings of our study. Thyroid hormone replacement therapy was required in all 12 patients, and 5/12 patients (41.7%) discontinued immunotherapy after developing thyroiditis, which was due to early-onset toxicity or mixture of other complications.

The [18F]FDG-PET/CT scans showed an overall moderate sensitivity in the organs discussed in this study, ranging from 50 to 75%. [18F]FDG-PET/CT had poor sensitivity in detecting irAEs in the skin, which was reflected in its low sensitivity (19%) and PPV (50%). In patients with melanoma, such poor results in the skin could potentially be due to a misinterpretation of its findings as disease recurrence, pseudoprogression, or wounds after surgery. Musculoskeletal processes may lead to focal uptake on [18F]FDG-PET/CT [42]. The incidence of arthralgia was approximately 9–13% in a study by Benefareni et al. [43, 44], whereas in this study, musculoskeletal irAEs were considered one category and registered in 21/123 (17.1%) patients. Pneumonitis can cause clinical symptoms such as dyspnea or cough, but it can also be asymptomatic [45]. This could be considered for 12 patients who were classified as false-positive when the lungs were investigated. Due to physiologic FDG uptake in the heart, such can occasionally be misinterpreted as myocarditis, potentially resulting in increased false-positive cases.

A systematic review and meta-analysis evaluated the value of [18F]FDG-PET/CT for predicting or monitoring immunotherapy response in patients with melanoma [46]. They showed that some of the parameters present promising predictors of the final response to immunotherapy. These results supplement our findings of the accuracy of [18F]FDG-PET/CT in the diagnosis of irAEs due to immunotherapy.

In this study we found that 65/123 patients (52.8%) developed treatment-related adverse events. Compared with a similar study [3], adverse events from any cause were reported in 96.9% of the patients, of which 14.4% were considered grade 3 and 4. In our study, the results showed that out of the 123 patients who received immunotherapy, 40 patients (32.5%) discontinued immunotherapy because of irAEs. In contrast, in a study by Weber et al.[3], approximately 10% of patients terminated treatment due to treatment toxicity.

Sarcoidosis-like reactions have been linked to the administration of PD-1 inhibitors and are drug-induced [47]. A drug-induced sarcoid reaction (DISR) is described as a systemic granulomatous tissue reaction that is indistinguishable from sarcoidosis and occurs in close proximity to the initiation of an offending drug [48]. DISR is the most relevant irAE that could be misinterpreted as malignancy because it mimics newly developed hilar and mediastinal lymph node manifestations [10]. Generally, DISR is asymptomatic, and no treatment is required [47, 48]. DISR was identified as mediastinal/hilar lymphadenopathy and was most frequently detected on the first scan after the initiation of immunotherapy [47], comparable to our study, in which 16/21 patients (76.2%) were identified with lymphadenopathy in the first follow-up scan. Our study observed such scans in 21/123 patients (17.1%), and 13/21 patients (61.9%) had no clinically registered irAEs. FDG uptake was increased in both hilar and mediastinal lymph nodes in 10/21 individuals (47.6%). Only 1/21 patients (4.8%) were clinically diagnosed with sarcoidosis.

In the interpretation of [18F]FDG-PET/CT scans, it is imperative to differentiate between irAEs, disease recurrence, and other non-specific inflammatory conditions. This distinction is crucial for guiding clinical management decisions and ensuring optimal patient care [49]. While we primarily focused on identifying and characterizing irAEs in this study, it is important to acknowledge the broader clinical context. Recurrence of the underlying malignancy and non-specific inflammation due to conditions like osteoarthritis and nonspecific reactive lymphadenopathy can often present with increased [18F]FDG uptake patterns that may overlap with irAEs [31, 50, 51]. Experienced nuclear medicine specialists read the scan in this study to diminish some of these confounding factors.

This is, to the best of our knowledge, the first study to assess the accuracy of [18F]FDG-PET/CT in detecting irAEs in patients with melanoma treated with adjuvant immunotherapy. The patients in this study had characteristics reflecting a standard clinical sample of patients from our population of interest, representing daily clinical practice leading to a higher external validity. All patients who met our inclusion criteria were included, which decreased the chances of selection bias.

Several limitations need to be acknowledged in our study. Firstly, the utilization of clinically registered irAEs as our reference standard introduces the possibility of falsely low sensitivity. IrAEs could have been treated or resolved prior to the subsequent [18F]FDG-PET/CT scan, thus rendering them undetectable by this modality. Furthermore, given that certain adverse events are asymptomatic, there exists a potential for clinical under-recognition, leading to misleadingly low specificity. It is essential to note that distinguishing between the physiological distribution of FDG in organs such as the intestines, heart, and muscles, and inflammation, can be inherently challenging. The variability in physiological FDG uptake, including factors like medication intake (such as metformin), intestinal motility, and muscle activity, introduces complexities that were not explicitly controlled for in our retrospective methodology. This limitation may have impacted the precision of our interpretations.

Additionally, while our study's design involved experienced examiners, the absence of formal assessment of inter-observer agreement among the three clinicians is another limitation. Despite their proficiency, the potential for interpretation discrepancies exists, which could have influenced diagnostic accuracy. Notably, a limitation intrinsic to FDG PET/CT is its sensitivity in detecting skin inflammation, constrained by resolution limitations. The suboptimal spatial resolution of PET scans can impact our ability to precisely identify and quantify skin-related irAEs.

Furthermore, the assessment of hepatic inflammation related to irAEs was inadvertently omitted from this study. This omission is partly attributed to the challenge posed by the high physiological FDG distribution in the liver, which can potentially obscure subtle inflammatory changes. We recognize the importance of this aspect and acknowledge its absence in our study.

Moreover, the occurrence of a false-negative result in myocarditis detection should be noted. This was due to physiologic FDG uptake in the myocardium, which is a potential limitation of [18F]FDG-PET/CT in myocarditis diagnosis.

Given the retrospective nature of our investigation, detailed patient irAE data acquisition was limited to registered medical records. Additionally, only patients who experienced altered clinical management due to irAEs were considered eligible. More comprehensive assessment and diagnosis of adverse events would require para-clinical test results. Applying the Common Terminology Criteria for Adverse Events grading system could further enhance the study's validity.

In addition to our analysis of true-positive cases, it is essential to consider the significance of false-positive and false-negative findings in [18F]FDG-PET/CT detection of irAEs. While false-positives may lead to unnecessary concerns and interventions, false-negatives could result in missed opportunities for early intervention [31, 52]. It is noteworthy that the interpretation of [18F]FDG-PET/CT scans encounters challenges in distinguishing physiological from pathological FDG uptakes. A substantial limitation in this context is the absence of a definitive reference standard for making this differentiation.

In future studies, it would be clinically beneficial to see whether changing the timeline of the [18F]FDG-PET/CT scans would yield different results. Since more than half of the patients develop irAEs after the initial three immunotherapy doses, adjusting the intervals between the [18F]FDG-PET/CT scans should be considered. If the scans were performed more frequently initially, it is likely that more irAEs would be detected on [18F]FDG-PET/CT, potentially leading to earlier treatment when needed. Additionally, considering that irAE appearance has been linked to the efficacy of immunotherapy, investigating the predictive role of these events in immunotherapy response could potentially reduce patient discontinuation of immunotherapy [53]. This emphasis on early irAE treatment aligns with the need to differentiate irAEs from other conditions. To further enhance the clinical utility of [18F]FDG-PET/CT in this context, future research should investigate specific differentiation strategies and diagnostic algorithms for distinguishing between irAEs, recurrence, and other inflammatory conditions. This endeavour will play a significant role in refining the utility of [18F]FDG-PET/CT in guiding treatment decisions and monitoring therapeutic responses in oncology patients.