Table 2.
Mendelian Randomization (MR) studies of coffee and caffeine consumption.
| Study | Outcome | Instrumental Variable (IV) | Design & Approach | Results | Interpretation | Limitations Reported |
|---|---|---|---|---|---|---|
| Nordestgaard et al. 2015 [44] | Obesity, metabolic syndrome, T2D and related measures (BMI, WC, height, weight, SBP, DBP, TGs, TC, HDL, glucose) |
5-SNPs AHR, CYP1A2 Score and single SNPs |
One-sample Individual-level data 2SLS n ≤ 93,179 Copenhagen General Population Study (CGPS) and the Copenhagen City Heart Study (CCHS). Summary-level data Wald ratio, IVW T2D only DIAGRAM (n ≤ 78,021) |
Observational: Coffee significantly reduced risk of obesity, metabolic syndrome and T2D Coffee significantly increased BMI, WC, weight, height, SBP, DBP, TGs, and TC and decreased HDL SNP-outcome: NS Similar results when individuals were stratified into coffee drinkers and coffee abstainers however, among those without coffee intake, blood pressure was lower with higher coffee-intake allele score |
No evidence supporting a causal relationship between coffee and outcomes | Underpowered IV Pleiotropy Collider Bias |
| Nordestgaard & Nordestgaard, 2016 [43] | CVD (IHD, IS, IVD) All-cause and CVD mortality |
2-SNPs AHR, CYP1A2 Score and single SNPs |
One-sample Individual-level data 2SLS n ≤ 112,509 CGPS, CCHS and Copenhagen Ischaemic Heart Disease Study (CIHDS) 3822 IHD cases 1708 IS cases 4971 IVD cases 971 CVD deaths 5422 total deaths Summary-level data Wald ratio, IVW IHD only Cardiogram (n = 80,517) and C4D (n = 30,433) |
Observational: U-shaped association between coffee intake and IHD, IS, IVD and all-cause mortality. Lowest risk with medium coffee intake compared with no coffee intake. SNP-outcome: NS Similar results when individuals were stratified into coffee abstainers, coffee drinkers, coffee drinkers excluding tea and cola drinkers. |
No evidence supporting a causal relationship between coffee and outcomes | Underpowered IV Pleiotropy Collider Bias (stratified analysis) Confounding by other caffeine containing-beverages Cannot rule out non-linear effects of coffee on outcomes |
| Kwok et al., 2016 [45] | T2D, IHD, depression, Alzheimer’s disease, lipids, glycemic traits, adiposity or adiponectin | 9-SNPs AHR, CYP1A2(2), GCKR, MLXIPL, POR, EFCAB5, BDNF, ABCG2 5 SNPs AHR, CYP1A2(2), POR, EFCAB5 3 SNPs AHR, CYP1A2(2) |
Two-sample Summary-level data Multiple published GWAS WME |
9 SNPs: ↑T2D, ↓TGs, ↑BMI, ↑WHR, ↑IR 5 SNPs: NS 3 SNPs: NS |
No evidence supporting a causal relationship between coffee and outcomes | Confounding (Population stratification) Pleiotropy Cannot rule out non-linear effects of coffee on outcomes |
| Treur et al., 2016 [46] | Smoking behavior Coffee intake Caffeine use |
1-SNP for smoking heaviness (CHRNA3) 8-SNP score for coffee intake AHR, CYP1A2, GCKR, MLXIPL, POR, EFCAB5, BDNF, ABCG2 |
Individual-level data Bivariate genetic modelling (SEM) n = 10,368 current smoking (y/n) caffeine use (high/low) coffee use (high/low) Bidirectional MR Regression analyses n = 12,319 Self-reported caffeine use (mg/day), coffee use (cups/day), cigs/day, smoking initiation and persistence Summary-level data LD score regression CCGC Tobacco, Alcohol and Genetics Consortium (TAG): cigs/day, smoking initiation and persistence n ≤ 38,181 |
Bivariate genetic modelling Current smoking-coffee intake: G r = 0.47, E r = 0.30 Current smoking-caffeine use: G r = 0.44, E r = 0.00 MR: NS LD score regression Smoking heaviness- coffee intake: r = 0.44 Smoking initiation-coffee intake: r = 0.28 Smoking persistence-coffee intake: r = 0.25 |
Genetic factors explain most of the association between smoking and caffeine consumption. Quitting smoking may be more difficult for heavy caffeine consumers, given their genetic susceptibility. | Underpowered Pleiotropy |
| Taylor et al., 2017 [47] | Prostate cancer (PC) risk and progression | 2-SNPs AHR, CYP1A2 |
Individual-level data Two-sample MR Regression analyses + meta-analysis Practical consortium (n = 46,687) 4 studies GS-coffee GS-tea GS-(tea + coffee) 23 studies GS-PC GS-PC stage GS-PC grade GS-mortality |
Significant GS-coffee, GS-tea and GS-(tea + coffee) GS-PC grade (p = 0.02) |
No clear evidence supporting a causal relationship between coffee and outcomes | Between-study heterogeneity in case definition Imprecise IV Pleiotropy Underpowered |
| Ware et al., 2017 [48] | Smoking heaviness, cigs/day | 8-SNP GS AHR, CYP1A2, GCKR, MLXIPL, POR, EFCAB5, BDNF, ABCG2 6-SNP GS AHR, CYP1A2, GCKR, MLXIPL, POR, EFCAB5 2-SNP GS AHR, CYP1A2 |
2-sample MR Summary-level data IVW, WME CCGC TAG GWAS Cotinine levels (n = 4548) [in vitro experiments] Individual-level data (replication, n = 8072 smokers who drink coffee) IVW, WME |
Each cup of coffee/day lead to a decrease in 1.5 (8 SNPs), 1.7 (6 SNPs) or 2.0 (2 SNPs) cigs/day. Coffee did not influence cotinine levels. Coffee did not influence cigs/day in replication sample. |
Coffee intake is unlikely to have a major causal impact on cigarette smoking | Pleiotropy Underpowered replication Underpowered IV |
| Bjorngaard et al., 2017 [49] | Coffee intake (cups/day, sensitivity analysis: Any vs. none) Tea intake (cups/day, sensitivity analysis: Any vs. none) Smoking status (never, former, current) Smoking heaviness (cigs/day) |
1-SNP (CHRNA3) for smoking heaviness 2-SNPs (AHR, CYP1A2) for coffee intake GS |
Individual-level data Bidirectional MR Regression analyses + meta-analysis UK biobank (n ≤ 114,029) HUNT (n ≤ 56,664) CGPS (n ≤ 78,650) coffee or tea drinkers only |
Observational Former & current smoking associated with higher coffee consumption (not tea) vs. never smokers. Among smokers: Each cig/day increased coffee and tea intake; stronger for coffee MR SMK-SNP associated with coffee intake in current or ever smokers only Coffee-SNP not associated with smoking behavior |
Higher cigarette consumption causally increases coffee intake. | Underpowered to rule out causal coffee → smoking association. UK Biobank non-representative sample Collider bias: (i) if selection into the sample is related to both coffee and smoking (ii) via smoking stratification Phenotype measurement error |
| Larsson et al., 2017 [50] | Alzheimer’s Disease (AD) | 5-SNP GS AHR, CYP1A2, MLXIPL, POR, EFCAB5 (coffee and 23 other exposures tested) |
Summary-level data 2-sample MR IVW, WME, MR Egger CCGC International Genomics of Alzheimer’s Project (n = 17,009 cases, 37,154 controls) |
Suggestive association between coffee GS and increased risk of AD (p = 0.01) | Suggestive causal relationship between coffee and AD risk, but in opposite direction to that expected based on observational studies. | None. |
| Verweij et al., 2018 [51] | Causal associations between nicotine, alcohol, caffeine, and cannabis use | Polygenic scores (p < 5 × 10−8 or p < 1 × 10−5) for each exposure | Summary-level data two-sample bidirectional MR IVW, Wald ratio Multiple published GWAS |
Smoking cigs/day—caffeine use (p = 0.01) Alcohol use: Smoking initiation (p = 0.03) |
Little evidence for causal relationships between nicotine, alcohol, caffeine, and cannabis use, but may suggest a common liability model (shared genetics) | Imprecise IV GWAS sample overlap (bias to null) |
| Ong et al., 2017 [52] | Epithelial ovarian cancer | 4-SNP GS (coffee IV) ABCG2, AHR, CYP1A2, POR 2-SNP GS (caffeine IV) AHR, CYP1A2 |
Summary-level data Two-sample MR Wald-type ratio estimator CCGC Ovarian Cancer Association Consortium (n = 44,062, 20,683 cases) |
NS | No evidence supporting a causal relationship between coffee/caffeine and outcome | MR Assumption 3 not confirmed Not generalizable to non-European populations. Underpowered or imprecise IV Cannot rule out non-linear effects of coffee/caffeine on cancer |
| Larsson et al., 2018 [53] | Gout | 5-SNPs AHR, CYP1A2, MLXIPL, POR, EFCAB5 |
Summary-level data 2-sample MR IVW, WME, MR Egger CCGS Serum Uric acid GWAS (n = 110,347) Gout GWAS (2115 cases and 67,259 controls). |
CYP1A2 and MLXIPL SNPs inversely associated with uric acid Combined MR: significant inverse relationship (p = 7.9 × 10−6) All but AHR SNP associated with lower gout risk. Combined MR: significant inverse relationship (p = 0.005) |
Supports causal inverse association between coffee intake and risk of gout. | None |
| Treur et al., 2018 [54] | Sleep behaviors (sleep duration, chronotype and insomnia complaints) |
IV threshold p < 5 × 10−8 4 SNPs (POR, AHR, CYP1A2, MXLIPL) p < 5 × 10−5 4 SNPs plus 23 SNPs |
Summary-level data Two-sample bidirectional MR IVW, LD score regression CCGC Caffeine metabolite GWAS Sleep GWAS |
MR: NS LD score regression: NS |
No evidence for causal relationship between habitual coffee intake and sleep behaviors. | Underpowerd LD score regression using caffeine metabolite GWAS Phenotype measurement error |
| Noyce et al., 2018 [55] | Parkinson’s Disease (PD) | Morning person primary exposure (15 SNPs) coffee secondary exposure (4-SNPs, AHR, BDNF, POR, CYP1A2) |
Summary-level data Two-sample MR IVW CCGC Morning person GWAS (n = 89,283) PD GWAS (13,708 cases, 95,282 controls) |
Morning person MR: p = 0.01 Coffee MR: NS |
Along with published RCT results, findings suggest that caffeine may neither prevent PD occurring nor be of benefit in those with the condition. | Use of summary-level data does not allow adjustment for potential confounding factors. |
| Zhou et al. 2018 [56] | Cognitive function composite global cognition and memory scores |
2-SNPs AHR, CYP1A2 Other SNPs (secondary analysis) |
Individual-level data n = 415,530 (300,760 coffee drinkers) from 10 meta-analyzed European ancestry cohorts. Genetic analysis performed under different levels of habitual coffee intake (1–4 and ≥4 cups/day. Negative control: Non-coffee drinkers. |
Observational: No overall association between coffee intake and global cognition and memory. SNP-outcome: NS |
Study provides no evidence to support beneficial or adverse long-term effects of coffee intake on global cognition or memory. | Pleiotropy. Caution when interpreting coffee IV |
| Lee, 2018 [57] | Osteoarthritis | 4 SNPs, POR, CYP1A2, NRCAM, NCALD |
Summary-level data Two-sample MR IVW, WME, MR-Egger regression CCGC + Amin et al. 2012 (n = 18,176) Osteoarthritis GWAS (7410 cases, 11,009 controls) |
IVW: p = 0.03 WME: p = 0.05 MR Egger: NS (however, no pleiotropy was evident) |
Results suggest that coffee consumption is causally associated with an increased risk of osteoarthritis. | Underpowered or imprecise IV Results limited to populations of European ancestry and limited to osteoarthritis in the knee and hip |
AD—Alzheimer’s disease; BMI—body mass index; CCGC—Coffee and Caffeine Genetics Consortium; DBP—diastolic blood pressure; DIAGRAM—Diabetes Genetics Replication and Meta-analysis; GS—genetic (SNP) score; HDL—high-density lipoprotein; IHD—ischaemic heard disease, IS—ischaemic stroke, IVD—ischaemic vascular disease, IVW—inverse-variance weighted meta-analysis, NS—non-significant; PC—prostate cancer; PD—Parkinson’s Disease; SBP—systolic blood pressure; T2D—type 2 diabetes; TC—total cholesterol; TGs—triglycerides; WC—waist circumference; WME—weighted median estimate.