免疫检查点抑制时代的跨癌种死亡率趋势：一项美国癌症统计数据的合成控制分析

Background: Immune checkpoint inhibitors (ICIs) improve survival in several advanced cancers, especially melanoma and non-small-cell lung cancer. Their overall impact on population-level cancer mortality is unclear. Methods: We used U.S. Cancer Statistics mortality data from CDC WONDER, 1999-2019. Deaths by cancer site and year were age-standardized to the 2000 U.S. population. We defined an “early-ICI” cancer cluster (cutaneous melanoma; lung and bronchus; kidney and renal pelvis; urinary bladder; Hodgkin lymphoma) and constructed a synthetic control from all other sites with complete data. Synthetic-control models treated 1999-2013 as the pre-intervention period and 2014-2019 as the ICI era. We compared age-adjusted mortality and translated differences into “deaths averted” by multiplying the mortality gap by population size. Robustness was assessed with placebo-in-space and placebo-in-time analyses and site-specific controls. As a cross-check, we applied staggered-adoption difference-in-differences using year of first ICI approval by site. Results: The early-ICI cluster had a modestly greater decline in age-adjusted mortality than its synthetic control after 2014. From 2014-2019, mortality averaged 0.41 deaths per 100,000 lower than the synthetic control, corresponding to an estimated 27,394 deaths averted versus the counterfactual trend. Pre-intervention fit was close (root-mean-square prediction error [RMSPE] 0.18), with higher post-intervention RMSPE (0.47; post/pre ratio 2.58). In placebo-in-space analyses, several non-ICI cancers had larger RMSPE ratios, indicating that the divergence was not unique. Placebo-in-time analyses using earlier hypothetical intervention years (2006-2010) produced smaller RMSPE ratios (1.39-1.81) than the 2014 specification. Site-level synthetic controls suggested deaths averted for melanoma, lung cancer, and Hodgkin lymphoma but excess deaths for kidney and bladder cancers, indicating heterogeneity. Staggered-adoption difference-in-differences estimated an average treatment effect of −0.96 deaths per 100,000 (95% CI −3.22 to 1.31), consistent with modest benefit or no effect. Conclusions: In this national synthetic-control analysis, cancers with early ICI approvals had slightly greater mortality declines than other cancers, translating to tens of thousands of potentially averted deaths. However, placebo analyses and difference-in-differences indicate that these gains are modest, heterogeneous, and compatible with residual confounding or concurrent advances in care. ICIs thus appear to be an important contributor, but not the sole driver, of recent improvements in cancer mortality.