Reverse causation risk

Causal status: Association

Reliability / bias notes: Device wear/non-wear; reverse causation; overlaps physical activity

Causal status: Association

Reliability / bias notes: Health satisfaction especially vulnerable to reverse causation

Causal status: Causal support strong for CVD via Mendelian randomization and RCTs; all-cause mortality signal weaker/borderline

Reliability / bias notes: All-cause mortality CI crosses 1.0; relationship appears non-linear/U-shaped, complicated by reverse causation in low-LDL groups (illness-driven cholesterol decline)

Causal status: Association

Reliability / bias notes: Small event count; self-report; reverse causation; limited generalisability

Causal status: Predominantly association; causal contribution plausible

Reliability / bias notes: Muscle mass and fat distribution; low-BMI reverse causation

Causal status: Mostly association; causal support substantial but mortality RCT evidence limited

Reliability / bias notes: Reverse causation; questionnaire error; healthy-user bias

Causal status: Causal support strong; mortality evidence mostly prospective

Reliability / bias notes: Self-report; pack-years and cessation timing matter; reverse causation can affect recent quitters

Causal status: Causal support strong for diabetes/CVD risk; all-cause mortality relationship J-shaped, so causal direction less clear at low HbA1c

Reliability / bias notes: J-shaped curve means very low HbA1c also carries excess mortality risk, likely reflecting reverse causation/frailty; overlaps with the separate diagnosed-diabetes question, so effects are kept modest to avoid double-counting the same underlying condition

Causal status: Association

Reliability / bias notes: Construct heterogeneity; socioeconomic confounding; reverse causation