# Smoking and dementia risk: an updated systematic review and meta-analysis of 61 prospective cohort studies and implications for tobacco control

- **DOI:** [10.1016/j.eclinm.2026.104310](https://doi.org/10.1016/j.eclinm.2026.104310)
- **Rating:** Sound (3.3 out of 5)
- **Verdict:** An updated cohort synthesis confirms smoking–dementia associations and adds dose–response and cessation contrasts, but observational bias limits causal interpretation.
- **Framework version:** v5.13
- **Pipeline version:** v0.4
- **Evaluation date:** 2026-11-18
- **Evaluation page:** https://www.nabu.science/paper/10.1016%2Fj.eclinm.2026.104310

## Key claims

1. Current smoking was associated with increased risk of all-cause dementia (RR 1.31 [95% CI 1.23–1.39]; number of studies (k) = 39), Alzheimer's disease (1.40 [1.08–1.80]; k = 15), and vascular dementia (1.52 [1.20–1.93]; k = 9).
   - Support: Supported
   - Type: Primary finding
   - Basis: pooled observational associations remain consistent in sensitivity analyses despite substantial heterogeneity and risk of bias
   - Location: Summary, Findings
2. Former smoking was not associated with any [dementia] outcome, and former smokers had a lower risk of all-cause dementia than current smokers (RR 0.81 [0.75–0.88]).
   - Support: Tentative
   - Type: Secondary
   - Basis: observational cessation comparisons face selection and sick-quitter bias, while null former-versus-never estimates do not establish equivalent risks
   - Location: Abstract, Findings (and Results, Cessation versus continued smoking; Table 1)
3. Dementia risk increased with smoking intensity as the pooled RR was 1.16 (95% CI 1.09–1.23) at 10 cigarettes per day and rose to 1.34 (95% CI 1.18–1.51) at 20 cigarettes per day.
   - Support: Supported
   - Type: Secondary
   - Basis: a 12-study observational synthesis supports the modelled gradient, with exposure conversion dependent on an assumed 40-year smoking duration
   - Location: Summary, Findings
4. In the nine cohorts reporting both [subtypes], the ratio of the VaD to the AD relative risk for current smoking was 1.11 (95% CI 0.88–1.40; P = 0.31)
   - Support: Supported
   - Type: Secondary
   - Basis: paired estimates directly support the reported ratio within nine cohorts, with clinical subtype misclassification limiting interpretation
   - Location: Results, Comparison of subtypes (and Summary, Added value of this study)
5. The communality-weighted PAF for smoking and dementia was 2.2% (95% CI 1.7–2.8).
   - Support: Tentative
   - Type: Secondary
   - Basis: population attribution assumes causality and shared Commission inputs, with the interval capturing only relative-risk uncertainty
   - Location: Abstract, Findings; Results, PAF calculation

## Strengths

- **Sensitivity analyses test the pooled associations** REML models with Hartung–Knapp–Sidik–Jonkman adjustment and prediction intervals make uncertainty visible. Current-smoking estimates remain similar under alternative estimators, risk-of-bias restrictions, and one-publication-per-cohort analyses.
  - Dimension: methodological rigour
  - Location: Methods, Statistical analysis; Table 1; Results, Sensitivity analyses
- **Within-cohort contrasts extend the earlier synthesis** The review adds former-versus-current smoking comparisons across 29 cohorts and paired dementia-subtype comparisons across nine cohorts. These comparisons address questions not resolved by simply comparing separately pooled estimates.
  - Dimension: contribution
  - Location: Results, Comparison of subtypes; Cessation versus continued smoking
- **Interpretation addresses mortality and diagnostic limitations** The Discussion explains how competing death, survivor selection, and illness-related quitting can shape observed associations. It also confines subtype interpretation to clinically diagnosed syndromes rather than biomarker-defined pathology.
  - Dimension: positioning
  - Location: Discussion, cessation, competing-risk, and clinical-syndromic diagnosis paragraphs

## Limitations

- **Null estimates do not establish equivalent risks** The conclusion describes former smokers' risks as similar to never smokers', but no equivalence analysis is reported. The former-smoking vascular-dementia interval of 0.93–1.20 still permits meaningful excess risk.
  - Dimension: methodological rigour
  - Location: Table 1, Former vs never; Discussion, final paragraph
- **Evidence certainty and exposure harmonisation remain limited** No included study was rated low risk of bias, and 61% were high or very high risk, without a formal certainty-of-evidence grading. The dose-response analysis also converts pack-years using an assumed common 40-year smoking duration.
  - Dimension: methodological rigour
  - Location: Results, Risk of bias; Methods, Statistical analysis
- **Study counts and citations need reconciliation** The reported 61 included studies and five non-poolable studies imply 56, not the stated 57 meta-analysed studies. Citation mismatches, including the ROBINS-E reference, further hinder source tracing.
  - Dimension: reporting
  - Location: Results, Study selection and characteristics; Methods, Risk of bias assessment; References 61 and 63

## Reliability checks

Caveats (4 of 4 checks).

- **Results check:** Noted. One reported value does not reconcile, but the key claims do not rest on it.
  - Minor reporting notes (1):
    - "Among the 57 studies contributing to the meta-analyses" (Results, Study selection and characteristics) vs "included in the meta-analysis (five studies⁹²,⁹³,⁹⁴ did not report estimates that could be pooled against never smokers)" (Results, Study selection and characteristics). The stated 61 included studies minus five exclusions equals 56, not 57, while 61 minus 57 implies four exclusions. The result itself is unaffected.
- **Reporting and transparency:** Noted. The ROBINS-E citation points to an unrelated reference. Registration, an ethics exemption statement, competing interests, AI assistance, and availability of data and code upon request are disclosed.
  - Conflict of interest: declared · Ethics approval: declared · Preregistration: declared · Data availability: declared · Code availability: declared
- **Reference integrity:** Clean. 107 of 107 checkable references verified
  - 111 references in manuscript
  - 4 are books, websites or datasets — counted, but not index-checkable
  - References not verified automatically are resolved manually within 24 hours.

## Post-publication record

- **Status:** Clean
- **Summary:** No retraction or expression of concern found in Crossref (includes Retraction Watch).
- **Last checked:** 2026-11-18
