Background: An early crossing of the Kaplan-Meier overall survival curves in randomized clinical trials (RCTs) comparing ICI alone to standard treatments has been observed across different cancer types, suggesting the existence of an excess of mortality upon ICI. Whether ED reflects the natural history of the disease or it is a treatment specific phenomenon remains unclear. Therefore, we conducted a metanalysis of RCTs to analyze the association of ICI treatment with ED. Methods: PubMed, Embase and Cochrane were searched (until 12/2021) for RCTs comparing 1^st-line PD-1/PD-L1 inhibitors +/- CTLA-4 inhibitors (ICI-only group) or the same drugs in combination with other non-ICI therapies (ICI+OT group) (experimental arms) vs non-ICI treatments (control arm) in patients (pts) with advanced solid cancers. ED was defined as death within the first 3 months of treatment. RCTs providing the number of pts in the intention-to-treat (ITT) population who were alive within 0-3 months from randomization were considered eligible. Primary outcome was ED rate both in the ICI-only and in the ICI+OT groups vs control arm. Secondary outcome was to compare ED risk between ICI-only group and ICI+OT group. Further analysis according to PD-L1 level were performed. The ED rates were estimated by risk-ratio (RR) with 95% confidence intervals (CI) and pooled by random effect model. Heterogeneity was assessed by I² statistics. Results: 56 RCTs (n = 38697 pts) were eligible, 48.2% included pts with thoracic malignancies. In the ICI-only group (18 RCTs, n = 6638 pts) 50.5% of pts received anti-PD-1/PD-L1 monotherapy and 43.4% anti-PD-1/PD-L1 + anti-CTLA-4 agents. In the ICI+OT group (43 RCTs, n = 15421 pts) 80.1% of pts received ICI + chemotherapy. Compared to control arm, risk of ED was significantly higher for pts who received ICI-only therapies (RR 1.27, 95% CI 1.04-155, p = 0.02, I²: 73%), but significantly lower in ICI+OT group (RR 0.80, 95% CI 0.71-0.89, p < 0.00001, I²: 36%). Risk of ED was also significantly higher with ICI-only compared to ICI-OT (RR 1.59, 95% CI 1.26-2.00). ED risk was higher upon ICI-only also in pts with high PD-L1 expression (TPS ≥50%, TC or IC 3, CPS ≥10) and, after excluding Keynote-024, this result became statistically significant (7 RCTs, RR 1.54, 95% CI 1.13-2.10, p = 0.007) with I² reducing from 63 to 46%. Conclusions: Across cancer types ED is a treatment specific phenomenon, increasing with first-line anti-PD-1/PD-L1 +/- CTLA-4 inhibitors and occurring also in pts with high PD-L1 expression, whereas it can be prevented by combining ICI with other non-ICI treatments.