Abstract
PURPOSE: This study investigated thyroid dysfunction with immune checkpoint inhibitors (ICIs) in terms of proportions affected, risk factors, thyroid sequelae, and overall survival (OS). METHODS: Among patients with normal baseline free T4 (fT4) and thyroid stimulating hormone (TSH) receiving ICIs at a large cancer centre, proportions of hyperthyroidism/hypothyroidism were determined (any, subclinical [normal fT4, abnormal TSH], overt [abnormal fT4, abnormal TSH], isolated hyperthyroxinaemia/hypothyroxinaemia and secondary) with onset times and subsequent thyroid statuses. Associations of overt dysfunction with OS were estimated using Cox regression and methods robust to immortal time bias (time-dependent Cox regression and 3- and 6-month landmark analyses). Associations of baseline variables with overt hyperthyroidism and hypothyroidism were estimated using Fine and Gray regression. RESULTS: Of 1349 patients, 34.2% developed hyperthyroidism (10.3% overt), including 54.9% receiving combination ICIs, while 28.2% developed hypothyroidism (overt 9.3%, secondary 0.5%). A third of overt hypothyroidism cases occurred without preceding hyperthyroidism. Subclinical thyroid dysfunction returned directly to normal in up to half. Overt hyperthyroidism progressed to overt hypothyroidism in 55.4% (median 1.6 months). Melanoma treatment in the adjuvant vs. advanced setting caused more overt hyperthyroidism (12.1% vs. 7.5%) and overt hypothyroidism (14.5% vs. 9.7%). Baseline eGFR < 60 mL/min/1.73 m(2) (HR=1.68, 1.07-2.63) was associated with overt hyperthyroidism and sex (HR=0.60, 0.42-0.87) and TSH (4th vs. 1st quartile HR=1.87, 1.10-3.19) with overt hypothyroidism. Overt dysfunction was associated with OS in the Cox analysis (HR=0.65, 0.50-0.85, median follow-up 22.2 months) but not in the time-dependent Cox (HR=0.79, 0.60-1.03) or landmark analyses (3-month HR=0.74, 0.51-1.07; 6-month HR=0.91, 0.66-1.24). CONCLUSION: Thyroid dysfunction affects up to half of patients receiving ICIs. The association with OS is unclear after considering immortal time bias. The clinical courses include recovery, thyrotoxicosis and de novo overt hypothyroidism. Adjuvant treatment for melanoma, where longer-term harms are of concern, causes more frequent/aggressive dysfunction.