Hyperprogression in various solid cancers treated with immune checkpoint inhibitors in the real world

초록

Purpose: Immune checkpoint inhibitors (ICIs) have revolutionized the cancer prognosis in various cancer types. However, ICIs may trigger accelerated tumor progression, regarded as hyperprogressive disease (HPD), in certain patients, and it is still challenging to define HPD. We aimed to investigate the landscape of HPD and the prognostic value of its different definitions in advanced solid cancer patients treated with ICIs. Methods: We conducted a multicenter, prospective cohort study for solid cancer patients receiving ICIs. Among them, only unresectable or metastatic cancer patients were included in this analysis. Tumor response was evaluated according to RECIST 1.1. In patients who showed progressive disease (PD) by RECIST 1.1 at the first tumor evaluation, HPD was defined according to the following three criteria: A) tumor growth kinetics (TGK) ratio (TGKpost-ICI/TGKpre-ICI >=2, B) >=10mm increase in sum of target lesion (SUMtarget)and at least one of the following two criteria - 1) >=40% increase in SUMtarget or 2) >=20% increase in SUMtarget and the appearance of new lesions in at least two different organs, C) TGK ratio >=2 and >50% increase in SUMtarget. The discriminatory ability of three definitions in terms of overall survival (OS) were evaluated by the chi-square, C-statistics, and prediction error with integrated Brier score. Results: A total of 427 patients were included; head and neck (n=22, 5.2%), lung (n=173, 40.5%), breast (n=8, 1.9%), gastrointestinal tract (n=99, 23.2%), hepatobiliary pancreas (n=57, 13.3%), genitourinary (n=56, 13.1%), melanoma (n=5, 1.2%), and others (n=7, 1.6%). Incidences of HPD were 4.9%, 14.8%, and 11.5% in definition A, B, C, respectively. The incidence of HPD was relatively low (2.7%-9.5%) in non-small cell lung cancer compared to other cancer types (10.1%-21.5% in esophagogastric cancer, 3.9%-21.6% in hepatobiliary pancreas cancer, and 5.5%-22.5% in genitourinary cancer). Median OS was the worst for patients with HPD, which ranged from 4.8 months to 4.9 months according to definition A-C. After multivariate analysis adjusting for cancer types, ICIs types, and the number of prior anti-cancer therapy, each definition remained a significant factor for OS (P<0.001, respectively). Both chi-square and C-statistics of definition B were higher than those of definition A and C, although prediction error with integrated brier scores was similar between three definitions. Conclusions: Incidences of HPD appear to be various according to its definitions and cancer types. Given that the RECIST-based definition B not requiring pre-ICI imaging, showed similar discriminatory ability to predict dismal OS compared to TGK-based ones, it may be the most feasible and convenient measure to capture HPD in daily clinical practice.