24 Parkinson's Disease Risk Factors (example: 3X increase if low Vitamin D)

Factors That Raise Parkinson's Disease Risk by More Than 50% (RR/OR/HR > 1.5)

Risk factors with an estimate above 1.5 fall into three groups. The largest effects come from genes, prodromal signs and family history: LRRK2 G2019S (about 10-fold), GBA1 variants (OR 5.43), isolated REM sleep behavior disorder (most patients develop a synucleinopathy), hyposmia, family history of PD (OR 3.2–4.5), constipation (about 2.3) and depression (about 2.2). A smaller group of modifiable exposures clears the bar with weaker or conflicting evidence: some pesticides (rotenone, paraquat, paraquat plus maneb), trichloroethylene, head injury, milk intake in men, never smoking and low vitamin D. Male sex and type 2 diabetes sit at or below the threshold.

TL;DR

Clearly above 1.5: LRRK2 and GBA1 variants, family history of PD or tremor, isolated RBD, hyposmia, constipation and depression. Most of the non-genetic items in this group are early signs of PD itself (prodromal markers), not causes.
Modifiable exposures, often borderline (about 1.5–2.5):
- Pesticides: paraquat OR 1.64 in one meta-analysis but only 1.21–1.47 in others; rotenone OR 2.5.
- TCE at Camp Lejeune: OR 1.70.
- Head injury: OR 1.48–1.72.
- Milk intake: RR 1.4–1.8, highest in men.
- Never smoking: about 2.3-fold compared with current smokers.
Vitamin D: case-control meta-analyses show deficiency below 20 ng/mL (50 nmol/L) linked to roughly 2.2–2.55-fold odds of PD, and one prospective cohort (Knekt 2010) found about a 3-fold risk for the lowest vs highest 25(OH)D quartile. But the ARIC cohort was null and three Mendelian randomization studies found no causal effect. Low vitamin D is an association above 1.5; it is not a proven cause.

Table

Factor	Category	Risk increase (RR/OR/HR, 95% CI)	Evidence type	Reference
LRRK2 G2019S carrier	Genetic	About 10-fold risk vs non-carriers; up to 27-fold if PRS in top 25%; penetrance about 24–25% by age 80	Large genetic cohort (23andMe), 2024; case-control penetrance studies	Brain 2024] Mov Disord 2025 [npj PD 2023]
GBA1 variant carrier	Genetic	OR 5.43 (any variant, cases vs controls)	16-centre study, 5,691 PD and 4,898 controls, 2009	[NEJM 2009] [1]
SNCA, PRKN, PINK1, DJ-1 (PARK7), VPS35 pathogenic variants	Genetic	Not reported as RR (Mendelian or near-Mendelian causes)	Family and linkage studies	—
Family history of PD (any relative)	Genetic/familial	OR 4.45 (3.39–5.83); 3.25 after trim-and-fill; first-degree OR 3.23 (2.65–3.93)	Meta-analysis, 173 studies overall, 2012	[Noyce 2012] [2]
Family history of tremor	Genetic/familial	OR 2.74 (2.10–3.57)	Meta-analysis, 2012	[Noyce 2012] [2]
VDR FokI polymorphism (FF+Ff vs ff)	Genetic (vitamin D pathway)	OR 2.32 (1.49–3.61); CC vs TT OR 2.45 (1.52–3.93); allele OR 1.41 (1.14–1.75)	Meta-analyses of 5–7 case-control studies, 1,266–2,034 cases, 2015	Li 2015 [Tang 2015] [3][4]
High polygenic risk score (non-carriers, top 25% vs middle)	Genetic	OR 1.95 (1.78–2.14)	Large genetic cohort, 2024	[Brain 2024]
Isolated REM sleep behavior disorder	Prodromal marker	6.3% per year phenoconversion; 73.5% at 12 years; 96.6% at 14 years (meta-analysis)	Multicentre cohort, n=1,280; meta-analyses	Review 2025 Galbiati 2019
Hyposmia (lowest smell quartile)	Prodromal marker	5.2-fold risk within 4 years (HAAS); OR 3.94 (PRIPS)	Prospective cohorts, n=1,850 in PRIPS	[Review 2017]
Constipation	Prodromal marker	OR 2.36 (1.93–2.88); RR 2.34 (1.55–3.53)	Meta-analysis, 17 studies, 3,024,193 people, 2023; Noyce 2012	Karger 2023 [Noyce 2012] [2]
Visual impairment	Prodromal marker (preliminary)	HR 2.28 (1.29–4.04)	UK Biobank cohort, 117,050 people; preprint	[medRxiv 2021]
Depression / mood disorder	Medical (likely prodromal)	HR 2.2 (1.7–2.9); MR OR 1.44 (1.001–2.077)	Retrospective cohort, 11,741 depressed patients; MR 2025	Schuurman 2002 [MR 2025] [5][6]
Mild TBI / head injury	Medical	mTBI OR 1.72 (1.21–2.45); head injury OR 1.58; any TBI RR 1.48 (1.22–1.74)	Meta-analyses: 18 studies, 1.48 million people (2024); 15 studies (2023)	mTBI 2024] Balabandian 2023]
Vitamin D deficiency (<20 ng/mL, <50 nmol/L)	Nutritional	OR 2.55 (1.98–3.27);OR 2.2 (1.5–3.4)	Meta-analyses, mostly case-control, 2019 and 2014	Zhou 2019 [Lv 2014]
Low 25(OH)D, lowest vs highest quartile	Nutritional	RR 0.33 (0.14–0.80) for highest vs lowest, i.e. about 3-fold for lowest	Prospective cohort, 3,173 people, 50 cases, 29 years	[Knekt 2010]
Vitamin D insufficiency (<30 ng/mL)	Nutritional	OR 1.77 (1.29–2.43); OR 1.52 (1.10–2.10)	Meta-analyses, mostly case-control	Zhou 2019 [Luo 2018]
Rotenone use	Environmental	OR 2.5 (1.3–4.7)	Case-control (FAME), 110 cases, 2011	[Tanner 2011]
Paraquat use	Environmental	OR 1.64 (1.27–2.13); single study OR 2.5 (1.4–4.7)	Meta-analysis, 13 case-control studies, 3,231 cases, 2019	Tangamornsuksan 2019 Tanner 2011
Paraquat plus maneb (residential)	Environmental	OR 1.75 (1.13–2.73); 3–6-fold if exposed when younger	Population case-control, 368 cases, 2009	[Costello 2009]
Any pesticide exposure	Environmental	RR 1.78; 1.53 (1.29–1.80) after correction for publication bias	Meta-analysis, 2012	Noyce 2012 van der Mark 2012 [7]
Trichloroethylene (Camp Lejeune water)	Environmental	OR 1.70 (1.39–2.07)	Cohort, 340,489 service members, 2023	[Goldman 2023]
Never smoking (vs current smoking)	Lifestyle (absence of inverse association)	Current vs never RR 0.44 (0.39–0.50), i.e. about 2.3-fold for never-smokers	Meta-analysis, 2012	[Noyce 2012] [2]
High milk intake, men	Lifestyle/diet	Dairy RR 1.66 (1.29–2.14) in men; 1.8 (1.4–2.4) in men	Meta-analyses of prospective cohorts, 2014 and 2007	Jiang 2014 [Chen 2007] [8][9]
High milk intake, both sexes	Lifestyle/diet	RR 1.56 (1.30–1.88); RR 1.45 (1.23–1.73); 2026 meta-analysis RR 1.13	Meta-analyses of prospective cohorts	Hughes 2017 Jiang 2014 Yang 2026] [8][10][11]
Male sex	Demographic	Incidence ratio 1.49 (France); pooled 1.57 (1.46–1.68); 1.93 at age 80+	Nationwide data plus meta-analysis of 22 incidence studies[12]	Moisan 2016 [Willis 2022]

Details

Genetic factors: the largest and most robust effects

LRRK2 G2019S is the most common monogenic cause of PD.
- In 23andMe data, "G2019S carriers had a 10-fold risk of developing PD versus non-carriers. This rose to a 27-fold risk" when polygenic risk was in the top quartile.[13]
- A 2025 case-control study put cumulative incidence by age 80 at 24% for G2019S, 4% for GBA1 p.N409S and 2% for non-carriers.[14]
- Penetrance is incomplete and depends on age, so for an individual carrier this is high but not certain risk.
GBA1 heterozygous variants gave OR 5.43 in the 16-centre NEJM study. When GBA was fully sequenced in 1,883 non-Ashkenazi Jewish patients, 7% carried a variant, showing that limited mutation screening can miss half the mutant alleles.
SNCA, PRKN, PINK1, DJ-1 and VPS35 are Mendelian genes. They are rare, and carrying a pathogenic variant (biallelic for PRKN, PINK1 and DJ-1) largely determines disease, so population RRs are not meaningful. For carriers they are far above 1.5.
Family history gave OR 4.45 for any relative and 3.23 for first-degree relatives. It was the single strongest non-monogenic factor in the Noyce 2012 meta-analysis, though trim-and-fill correction for publication bias reduced the any-relative OR to 3.25.[2][7]
Polygenic risk: people in the top quartile of a polygenic risk score had about twice the odds of PD (OR 1.95) without carrying any monogenic variant.[15]

Prodromal markers: strong predictors, but mostly not causes

Several of the largest non-genetic "risk factors" are early manifestations of the disease process (reverse causation). They are valuable for screening but are not targets for prevention.

Isolated RBD is the strongest clinical predictor. The largest international cohort (n=1,280) found phenoconversion at 6.3% per year, 73.5% within 12 years.[16] A pooled conversion rate of 31.95% after a mean of 4.75 years has also been reported, with PD accounting for 43–44% of conversions.[17]
Hyposmia: in the Honolulu-Asia Aging Study (Ross et al., Annals of Neurology 2008; 2,267 men aged 71–95), men in the lowest olfactory quartile had 5.2 times the risk of PD within 4 years; age-adjusted incidence fell from 54.5 per 10,000 person-years in the lowest quartile to 26.6, 8.2 and 8.4 in the higher quartiles.
Constipation was associated with OR 2.36 across about 3 million people, but heterogeneity was high (I²=90%).[18] An overview of 46 systematic reviews lists a pooled RR of 2.36 (2.00–2.80) from 4 cohort studies.[19]
Depression shows HR 2.2 in older cohort data. A 2025 Mendelian randomization study gave OR 1.44 (CI lower bound 1.001).[5][6] That suggests at most a modest causal component below 1.5, with most of the observational excess probably prodromal.

Environmental exposures: real but often borderline

Pesticides: the size of the paraquat estimate depends heavily on which studies are included.
- The 2019 meta-analysis gave OR 1.64.[20]
- A 2025 meta-analysis of 14 articles found OR 1.47 (1.25–1.72).[21]
- A 2025/2026 meta-analysis found OR 1.34 for any paraquat exposure and a non-significant OR 1.21 (0.96–1.53) for strictly occupational exposure in 233,208 participants.[22]
- Douglas Weed's 2021 review of 12 reviews (NeuroToxicology) noted that "no author of any published review stated that it has been established that exposure to paraquat causes Parkinson's disease" and concluded that the evidence "does not warrant a claim that paraquat causes Parkinson's disease".
- Treat paraquat as borderline or conflicting. Rotenone (OR 2.5) and paraquat combined with maneb (OR 1.75) more clearly exceed 1.5, but each rests on single case-control studies.[23][24]
Trichloroethylene: in the Camp Lejeune cohort of 340,489 people, PD risk was 70% higher (OR 1.70), and 81% higher for probable PD.[25][26] This is strong ecological evidence, but exposure was to mixed volatile organic compounds and was not measured in individuals.
Head injury: estimates cluster around the threshold.
- Any TBI: RR 1.48.[27]
- Mild TBI: OR 1.72.[28]
- In a Veterans Health Administration cohort of 162,935 veterans with TBI and 162,935 age-matched veterans without (Gardner et al., Neurology 2018), adjusted HRs were 1.71 (1.53–1.92) for any TBI, 1.56 (1.35–1.80) for mild TBI and 1.83 (1.61–2.07) for moderate-to-severe TBI.
- The 2023 meta-analysis abstract reports 1.48 (1.22–1.74).[27] Recall bias and pre-diagnostic falls may inflate these associations.[29]

Lifestyle and diet

Milk/dairy: the effect is consistently above 1.5 in men (RR 1.66–1.8) but weaker in women (1.15–1.3, not significant).[8][9]
- A 2017 pooled estimate for milk was 1.56.[11]
- The newest meta-analysis (Yang, Public Health 2026, nine studies) found smaller effects: total dairy RR 1.21 and milk RR 1.13.[10][30]
- A Mendelian randomization study using the lactase-persistence variant (Domenighetti et al., Courage-PD Consortium, Movement Disorders 2022) reported OR 1.70 (1.12–2.60) per daily serving, an effect restricted to men (OR 2.50, 1.37–4.56).
- Verdict: likely real for milk in men, borderline overall.
Never smoking: never-smokers have about 2.3 times the risk of current smokers.[2] This is a robust inverse association, but causality is debated.

Vitamin D: detailed assessment

Observational evidence clears 1.5.
- Zhou 2019 (8 studies, mostly case-control): deficiency (<20 ng/mL) OR 2.55 (1.98–3.27); insufficiency (<30 ng/mL) OR 1.77 (1.29–2.43). Sunlight exposure of at least 15 minutes per week was strongly protective (OR 0.02).[31]
- Lv 2014: deficiency (<50 nmol/L) OR 2.2 (1.5–3.4); insufficiency (<75 nmol/L) OR 1.5 (1.1–2.0).[32]
- Luo 2018: insufficiency OR 1.52 (1.10–2.10).[33]
Prospective data are split.
- Knekt 2010 (Mini-Finland, 3,173 adults aged 50–79, 29-year follow-up, 50 cases) found RR 0.33 (0.14–0.80) for the highest vs lowest 25(OH)D quartile, equivalent to about 3-fold risk in the lowest quartile.[34][35]
- The ARIC cohort (12,762 participants, median 17-year follow-up, 67 cases) was null: HR 1.14 (0.59–2.23) for ≥30 vs <20 ng/mL. Its authors concluded it "lends no support to the hypothesis that vitamin D may reduce the risk of PD."[36]
- I found no meta-analysis restricted to prospective cohorts. A 2025 review reports a cohort-based HR of 0.83 for sufficient vitamin D, without a CI.[37]
Mendelian randomization is null.
- Larsson 2017 (5,333 cases): OR 0.98 per 10% lower 25(OH)D (0.93–1.04).[38][39]
- NHANES plus MR 2025: OR 1.082 (0.902–1.297).[40]
- Wang 2024 (bidirectional MR): no causal link.[41]
- Caveat: these genetic instruments explain only a small share of 25(OH)D variance and test linear effects across the usual range. They cannot rule out a threshold effect from severe deficiency.
VDR FokI shows OR about 2.3–2.45 under dominant and homozygous models. The evidence comes from small case-control meta-analyses, and the effect is mainly in Asian populations.[3][42]
Reverse causation is plausible: PD reduces mobility and time outdoors. Supplementation trials raised 25(OH)D but did not improve motor function (MD −1.82, P=0.275), and no prevention trial exists.[43]

Commonly Cited Factors That Do NOT Reliably Reach >50%

Type 2 diabetes:
- RR 1.38 (1.18–1.62) in 7 cohorts with 1.76 million people. The same meta-analysis found about 50% higher risk in women.[44]
- RR 1.37 (1.21–1.55) in an earlier meta-analysis.[45]
- RR 1.27 (1.20–1.35) in a later meta-analysis of 15 cohorts.[46][47]
Inflammatory bowel disease: RR 1.41 (1.19–1.66) in 4 studies, an adjusted 1.24 (1.15–1.34) in 9 studies with 12.2 million people, and a null Mendelian randomization result (OR 0.982).[48][49]
Any TBI: RR 1.48 overall. Borderline; only mild TBI and repeated or severe injury exceed 1.5.
Occupational paraquat only: OR 1.21 (not significant). Total dairy in the newest meta-analysis: RR 1.21.[10]
Prediabetes: about 4% higher risk.[47] VDR FokI allele contrast: OR 1.41.[3]
Not quantified here: beta-blocker use, rural living, farming, well-water drinking, low physical activity, low serum urate and non-coffee drinking.
- Noyce 2012 and the Bellou 2016 umbrella review (75 meta-analyses) found significant associations for these.[7] Bellou rated only constipation and physical activity as having the strongest (class I) evidence.[7][50]
- Most well-water, rural-living and farming estimates in those reviews are modest, below 1.5. I could not verify exact figures for every item, so treat this as provisional.

Recommendations

For individual risk assessment, prioritise family history, LRRK2/GBA1 testing where relevant (for example Ashkenazi Jewish or North African Berber ancestry), and prodromal signs: RBD, hyposmia and constipation. These carry 2–10-fold or greater risk.
For prevention, the best-supported modifiable targets are:
- avoiding TCE and pesticides (especially rotenone and paraquat);
- preventing head injury;
- physical activity.
For VitaminDWiki: describe vitamin D deficiency as "associated with about 2–2.5-fold higher odds of PD in case-control meta-analyses and about 3-fold in one prospective cohort, but not supported as causal by Mendelian randomization or the ARIC cohort." Correcting deficiency is reasonable on general health grounds, but PD prevention is unproven.

Caveats

Most effect sizes come from case-control studies, which are prone to recall bias. Several factors (constipation, depression, hyposmia, RBD, low vitamin D) are probably consequences of preclinical PD.
Heterogeneity is often high (for example I²=90% for constipation), and publication bias reduced several estimates in Noyce 2012.[7]
One figure comes from a secondary source and should be checked against the original: the pooled male-to-female ratio of 1.57 (quoted in Willis 2022). The visual impairment estimate is from a preprint.