Smoking and coffee: protective associations

The research record

This is the project's file for this landmark, as written and checked. Sources are listed at the end; anything not yet confirmed is marked "Not yet verified".

Who

G. Webster Ross, Robert Abbott, Helen Petrovitch and colleagues (Honolulu Heart Program, 2000); Miguel Hernán and colleagues (meta-analysis, 2002); Claudia Domenighetti, Alexis Elbaz and colleagues (genetic analysis, 2022); Wolfgang Oertel and the NIC-PD investigators at 24 centres in Germany and the US (nicotine trial, 2023).

What happened

• 2000: Among 8,004 Japanese-American men followed for up to 30 years, 102 developed Parkinson's. Age-adjusted incidence fell steadily with coffee intake, from 10.4 per 10,000 person-years in non-coffee-drinkers to 1.9 in men drinking at least 28 oz a day (P < .001 for trend). The authors reported the link in never, past and current smokers (trend P = .049, .22, .02) and tied it to caffeine, not other coffee nutrients.
• 2002: A meta-analysis (Hernán et al.; 48 smoking studies, 13 coffee studies) found relative risks of 0.59 for ever-smokers (95% CI 0.54–0.63) and 0.39 for current smokers (0.32–0.47) against never-smokers, and 0.69 for coffee drinkers (0.59–0.80).
• 2022: Two-sample Mendelian randomisation (using inherited genetic differences as a natural experiment, with Courage-PD data on 7,369 cases and 7,018 controls) found genetically predicted smoking initiation was inversely linked to Parkinson's (OR 0.74 per standard deviation, 95% CI 0.60–0.93), not explained by reverse causation; alcohol and coffee showed no significant link, but power was limited.
• 2023: NIC-PD (NCT01560754), a placebo-controlled Phase 2 trial, randomised 163 people with early Parkinson's to nicotine patches (7 to 28 mg/day, as tolerated) or placebo for 52 weeks. Only 101 could be assessed at the 60-week main measure, because of dropouts, mostly from stopping treatment or side effects. Total UPDRS worsened by 6.0 points on nicotine against 3.5 on placebo (difference -3, 95% CI -6 to 0; P = 0.06). At 52 weeks (138 people) it was 9.1 against 5.4. So no benefit, and perhaps a trend to harm.

Why it moved us toward a cure

It prompted a decades-long search for a protective mechanism in dopamine systems (nicotine receptors, caffeine's adenosine receptor). The nicotine result is also a lesson: a robust epidemiological association did not become a disease-modifying drug.

Outcome

Dead end as a therapy; the association itself remains unexplained. Smoking is harmful overall and should not be recommended. The genetic analysis argues against pure reverse causation, but why remains unknown. Related: the same logic was tested for urate (landmark 89, landmark 66).

Sources

• Ross et al., JAMA 2000;283:2674–2679 (abstract), PMID 10819950
• Hernán et al., Ann Neurol 2002;52:276–284, PMID 12205639
• Domenighetti et al., J Parkinsons Dis 2022;12:267–282, PMID 34633332
• Oertel et al., NEJM Evidence 2023;2(9):EVIDoa2200311, PMID 38320207 (abstract) · ClinicalTrials.gov NCT01560754 (University of Vermont sponsor; sites in the US and Germany)
• Not yet verified: the NIC-PD full text; any caffeine trial result.