Psychological Stress and Cancer Risk and Outcome: An Umbrella Review With Quantitative Synthesis

Psychological stress, here used as an umbrella term encompassing depression, anxiety, psychosocial stress, stressful life events, perceived stress, and job strain, is hypothesized to influence cancer development and prognosis, yet epidemiologic evidence is inconsistent. This umbrella review with secondary quantitative synthesis evaluated associations between stress-related exposures and cancer incidence and mortality.

PubMed was searched from inception to March 1, 2026, alongside Scopus, the Cochrane Library, citation tracking, and manual screening. Eligible studies included systematic reviews, meta-analyses, or pooled individual participant data analyses assessing depression, anxiety, psychosocial stress, stressful life events, perceived stress, or job strain in relation to cancer incidence or mortality, reporting adjusted HRs, RRs, or ORs. Random-effects models were used for secondary pooling. Overlap was assessed with corrected covered area, and evidence credibility with the Ioannidis framework.

Fifteen publications met inclusion criteria; eight were quantitatively synthesized, representing over four million participants. Psychological stress was not significantly associated with overall cancer incidence (pooled RR/HR 1.06; 95% CI 0.99-1.14; I² = 72.5%) or breast cancer incidence (RR/OR 1.13; 95% CI 0.95-1.34; I² = 77.8%). In contrast, cancer mortality was consistently elevated across 4 analyses (RR/HR 1.23; 95% CI 1.19-1.27; I² = 0%). Site-specific mortality risks ranged from HR 1.15 for lung cancer to HR 3.86 for leukemia. Review overlap was moderate (7.4%). Associations between depression/anxiety and cancer mortality reached Class I evidence.

Psychological stress-related exposures were not significantly linked to cancer incidence in the available meta-analytic evidence, whereas depression, anxiety, and related distress were consistently associated with increased cancer mortality. These findings should be interpreted cautiously because residual confounding and measurement heterogeneity remain important limitations.