Reply to: TNF inhibition for immune checkpoint inhibitor-induced irAEs: the jury is still out

We would like to thank Karijn Suijkerbuijk and Rik Verheijden for their correspondence (Suijkerbuijk, K. P. M. & Verheijden, R. J. TNF inhibition for immune checkpoint inhibitor-induced irAEs: the jury is still out. Nat. Rev. Rheumatol. https://doi.org/10.1038/s41584-021-00640-z (2021))¹ on our Review (Chen, A. Y., Wolchok, J. D. & Bass, A. R. TNF in the era of immune checkpoint inhibitors: friend or foe? Nat. Rev. Rheumatol. 17, 213–223 (2021))², which raises some important issues concerning our analysis of their study assessing the safety of TNF inhibitor treatment for immune-related adverse events (irAEs) from immune checkpoint inhibitors (ICIs).

In our Review we wrote, “the use of overall survival as the end point might have introduced confounders as some high-grade irAEs (such as colitis) have a higher mortality than others (such as high-grade endocrine toxicity, which can be treated with hormone replacement)”². Suijkerbuijk and Verheijden are correct in pointing out that melanoma-specific survival was also shorter in the TNF inhibitor-treated patients, suggesting that toxicity-related deaths did not explain the differences in survival. We agree, this omission was an oversight on our part.

We do believe, however, that unmeasured confounders impacted their analysis of TNF inhibitor-related mortality, particularly as most TNF inhibitor-treated patients in their study received ipilimumab (an anti-CTLA4 therapy)³. The study analysed patients with melanoma treated with a first line ICI therapy from 2012 to 2017, a period that saw many temporal changes in both melanoma treatment and irAE management⁴. Ipilimumab monotherapy was rarely used after 2015, when it was supplanted by anti-PD1 therapy and later by combination ICI therapy⁵. The authors controlled for ICI regimen, but did not control for year of entry into the registry or type of second-line therapy (for example, anti-PD1 therapy versus targeted agents). However, many of the biases inherent in observational data are also shared by the two other cohort studies we cited^6,7.

The recent abstract highlighted by Suijkerbuijk and Verheijden is certainly of interest⁸. In this retrospective study of 150 patients with ICI-induced colitis, overall survival from time of colitis diagnosis was shorter in patients treated with the TNF inhibitor infliximab compared with patients treated with the anti-integrin (gut-specific) therapy vedolizumab. However, the study did not control for multiple confounders, including duration of steroid use (which was longer in the infliximab-treated group) and ICI regimen. Colitis treatment was based on clinician choice, and infliximab might have been used for more severe colitis cases. Colitis following combination ICI therapy, for example, is both more severe and occurs sooner after ICI initiation than colitis after ICI monotherapy⁹. Because severe ICI-induced colitis generally requires ICI discontinuation, the duration of ICI therapy might have been shorter in these individuals.

We certainly do not argue that TNF inhibitors are a magic bullet. Evidence suggests that cancer survival is best in ICI-treated patients who experience only low grade irAEs¹⁰, possibly because they do not require immunosuppression and/or they are able to continue their ICI therapy. A large prospective trial that compares infliximab to vedolizumab for treatment of ICI-induced colitis, or that compares a TNF inhibitor to corticosteroids for the treatment of ICI-associated inflammatory arthritis, would provide much needed answers regarding TNF inhibitor safety.