Overview
Immune checkpoint inhibitors (ICIs) have transformed cancer care by helping the immune system recognize and attack tumor cells more effectively. These drugs include anti-PD-1 agents, anti-PD-L1 agents, and anti-CTLA-4 agents. While highly effective, they can also trigger immune-related side effects in many organs, including the eye. One important but uncommon complication is non-infectious uveitis (NIU), an inflammatory condition inside the eye that can threaten vision if not recognized and treated promptly.
This study examined whether the risk of ICI-associated NIU differs by cancer type and by drug class. Understanding these patterns matters because the cause of uveitis may not be the same in every patient. Some of the risk may come from the drug itself, while other risk may be linked to the underlying cancer or the treatment combinations most often used in that cancer.
Why this study was needed
Uveitis associated with ICIs is still considered relatively rare, but it can lead to pain, blurred vision, light sensitivity, floaters, and, in severe cases, permanent visual loss. Because ICIs are now widely used across many cancers, clinicians need better guidance on who should be monitored more closely.
Previous reports suggested that melanoma patients may have a higher risk of ocular immune complications, but it has been difficult to separate the effect of the cancer itself from the effect of certain combination regimens. In particular, ipilimumab and nivolumab are often used together in melanoma and some other cancers, and this combination may raise the chance of immune toxicity. This study aimed to compare risks across malignancies and across ICI subclasses, using matching methods to reduce bias from age, comorbidities, and socioeconomic differences.
Study design and methods
This was a retrospective, multicenter clinical cohort study. Adults with metastatic melanoma, lung cancer, renal cell carcinoma (RCC), urothelial carcinoma, hepatocellular carcinoma, Hodgkin lymphoma, or non-melanoma skin cancer were included. Patients were matched using propensity scores to make the comparison groups more similar in demographic factors, comorbid conditions, and socioeconomic variables.
The study looked at patients who were prescribed any ICI, including anti-PD-1, anti-PD-L1, or anti-CTLA-4 therapy. The main outcome was new-onset non-infectious uveitis within 24 months after the first ICI prescription. The researchers calculated relative risks with 95% confidence intervals to compare how often NIU occurred between groups.
Subgroup analyses were especially important. The team compared outcomes across cancer types, with and without exposure to ipilimumab or nivolumab, and also compared ICI subclasses when used as monotherapy. This approach helped clarify whether the observed risk was driven by the malignancy, the drug class, or the frequent use of combination therapy.
Key findings
After matching, 16,834 patients with melanoma and 16,834 patients with lung cancer were included. The risk of NIU was much higher in melanoma than in lung cancer. Specifically, NIU occurred in 1.10% of melanoma patients versus 0.18% of lung cancer patients, corresponding to a relative risk of 6.17.
Even after excluding patients exposed to ipilimumab or nivolumab, melanoma still carried a higher risk: 0.53% versus 0.16%, with a relative risk of 3.40. This suggests that melanoma itself is associated with an intrinsically higher risk of ICI-related uveitis, not simply because of the use of more intense treatment combinations.
Across malignancies, melanoma consistently showed elevated risk compared with other cohorts. In one comparison, lung cancer had a borderline lower risk than RCC, with NIU rates of 0.19% versus 0.36% and a relative risk of 0.53. However, this difference was no longer significant after excluding patients exposed to ipilimumab or nivolumab, which indicates that the apparent excess risk in RCC may largely reflect treatment patterns rather than the cancer alone.
In the drug-class analysis, anti-PD-1 agents did not show a significantly different NIU risk compared with anti-PD-L1 agents. The rates were 0.26% versus 0.20%, with a relative risk of 1.33, but the confidence interval crossed 1.0, so this difference was not statistically significant. Anti-CTLA-4 agents could not be meaningfully compared because they were used too rarely as monotherapy.
Interpretation of the results
The most important message from this study is that both the cancer type and the ICI subclass appear to shape the risk of uveitis. Melanoma stood out as a disease with a higher intrinsic risk, even when the most commonly implicated combination therapies were removed from the analysis. In contrast, the elevated risk seen in RCC seemed to be more closely tied to ipilimumab/nivolumab exposure.
These findings support the idea that ICI-associated uveitis is not caused by a single mechanism. Instead, it likely reflects an interaction between host immune biology, the underlying cancer, and the specific immune checkpoint pathway being blocked. Melanoma is already known to be highly immunogenic, meaning the immune system is more actively engaged with tumor-related antigens. That immune environment may also predispose patients to inflammatory eye complications when ICIs are introduced.
The lack of a clear difference between anti-PD-1 and anti-PD-L1 drugs suggests that, at least in monotherapy, these two subclasses may have broadly similar ocular safety profiles with respect to NIU. Still, the overall event rates were low, so rare risks may be difficult to detect without even larger datasets.
Clinical implications
For ophthalmologists and oncologists, the practical takeaway is that melanoma patients starting ICIs may warrant especially careful ocular surveillance. Patients should be informed about early warning signs such as blurred vision, eye pain, redness, photophobia, new floaters, or decreased visual acuity. Prompt referral to eye care is important because early treatment can often preserve vision.
When uveitis is suspected, clinicians should first rule out infectious causes and other mimicking conditions before labeling it as immune-related. Management typically depends on severity and location of inflammation. Mild cases may respond to topical corticosteroids and close monitoring, while more severe posterior or panuveitic disease may require systemic corticosteroids, local steroid injections, or temporary interruption of ICI therapy. In selected cases, steroid-sparing immunosuppressive therapy may be considered in collaboration with oncology.
Importantly, treatment decisions must balance ocular safety with cancer control. Many patients benefit from continuing ICIs if inflammation is controlled and vision is protected, but the decision should be individualized based on severity, response to therapy, and oncologic necessity.
Strengths and limitations
A major strength of this study is its large, multicenter design and the use of propensity score matching, which helps reduce confounding from demographic and clinical differences between groups. The investigators also performed thoughtful subgroup analyses that helped distinguish drug-related effects from disease-related effects.
There are also limitations. As with any retrospective study, diagnosis codes and medical records may not capture every case perfectly. Mild uveitis may be underdiagnosed, while coding differences across centers can affect results. In addition, the rarity of anti-CTLA-4 monotherapy limited the ability to compare that class directly. The study also could not fully account for every clinical variable, such as tumor burden, prior autoimmune disease, ocular history, or exact timing and severity of inflammation.
Even so, the consistency of the melanoma signal across analyses makes the findings clinically meaningful.
Bottom line
This study shows that the risk of ICI-associated non-infectious uveitis varies by both malignancy and drug class. Melanoma patients appear to have a higher baseline susceptibility, independent of ipilimumab or nivolumab exposure, while RCC-associated risk may be driven more by those combination therapies. Anti-PD-1 and anti-PD-L1 monotherapy appear to have similar NIU risk.
For clinicians, these results support closer eye monitoring in melanoma patients receiving ICIs and reinforce the importance of early recognition and coordinated care between oncology and ophthalmology. For researchers, the findings suggest that both tumor biology and immune checkpoint pathway selection contribute to the pathophysiology of ocular immune adverse events.