TABLE 4.
Individual study characteristics for included Mendelian randomization studies1
| Author, y (reference) | MR type | Choline type(s) reported | Health outcome(s) | Key findings | Effect size for significant associations2 |
|---|---|---|---|---|---|
| Adams, 2020 [54] | Two-sample MR | SM | Breast cancer-specific survival for estrogen receptor-negative (ER–) and ER+ survival | Circulating sphingomyelin was positively associated with ER+ breast cancer survival. | SM -> ER+ survival †: HR = 0.65; 95% CI: 0.45–0.94, P = 0.02 SM -> ER+ survival ††: HR = 0.65; 95% CI: 0.45–0.94, P = 0.02 |
| Ahmad, 2022 [55] | Two-sample MR | PC, LysoPC, and SM | BMI and waist-to-height adjusted BMI (WHRadjBMI) | LysoPC was negatively associated with BMI. SM was negatively associated with WHRadjBMI. | LysoPC -> BMI † (combined PIVUS, ULSAM and TwinGene cohorts): β = −0.29; 95% CI: −0.52 to −0.050, P = 0.02 SM -> WHRadjBMI † (combined PIVUS, ULSAM and TwinGene cohorts): β = −0.51; 95% CI: −0.90 to −0.13, P = 8.00 × 10−3 SM -> WHRadjBMI † (combined ULSAM and PIVUS cohorts): β = −2.46; 95% CI: −3.78 to −1.14, P = 3.00 × 10−4 SM -> WHRadjBMI † (TwinGene cohort): β = −0.71; 95% CI: −1.3 to −0.12, P = 0.020 |
| Del Greco, 2019 [56] | Two-sample MR | PC (38:3) and SMs | BMI and p-wave duration | No significant associations were reported for either health outcome. However, the association between circulating PC and BMI was near significant (β = 3.34, SE = 1.78, P = 0.057). | |
| Ganna, 2014 [57] | Two-sample MR | LysoPC (18:1), LysoPC (18:2), SM (28:1) | Coronary heart disease (CHD) | No significant associations were reported between circulating LysoPC 18:1, LysoPC 18:2, and SM 28:1 with CHD. | |
| Gong, 2021 [58] | One-sample MR | LysoPC | Total hip bone mass density (BMD) variation (femoral neck, trochanter, and intertrochanteric region) | No significant associations were reported between serum LysoPC and hip BMD variation. | |
| Guida, 2021 [59] | Two-sample MR | PC, LysoPC, SM, glycerophosphocholine (GPC) | Incident kidney cancer | No significant associations were reported between circulating PC, LysoPC, or GPC and incident kidney cancer. | |
| Jia, 2019 [60] | Bi-directional MR | TMAO, choline, betaine | Type II diabetes (T2D), atrial fibrillation, myocardial infarction, coronary artery disease (CAD), stroke, chronic kidney disease (CKD), BMI, % body fat, glycemic traits, lipids [HDL, LDL, total cholesterol, triglycerides (TGs)] | Plasma betaine was associated with T2D. Plasma choline was individually associated with T2D, atrial fibrillation, % body fat, and BMI. | Betaine -> T2D †: OR = 0.68; 95% CI: 0.48–0.95, P = 0.023 Choline -> T2D †: OR = 1.84; 95% CI: 1.00–3.42, P = 0.05 Choline -> % body fat †: β = –0.28, P = 0.013 Choline -> BMI ††: β = 0.33, P = 0.029 Choline -> atrial fibrillation +: OR = 1.57; 95% CI: 1.05–2.36, P = 0.028) |
| Jia, 2022 [61] | Two-stage MR (systematic MR and phenome-wide MR) | SM, 1-arachidonoyl-GPC | Anorexia nervosa, odds of bipolar disorder | Circulating SMs were associated with anorexia nervosa. Circulating 1-arachidonoyl-GPC was associated with the odds of bipolar disorder. | SM -> anorexia nervosa †: OR = 1.12; 95% CI: 1.06–1.19, P = 2.34 × 10−5 SM -> anorexia nervosa ++: OR = 1.13; 95% CI: 1.06–1.19, P = 4.37 × 10−5 SM -> anorexia nervosa +++: OR = 1.12; 95% CI: 1.07–1.19, P = 2.25 × 10−5 SM -> anorexia nervosa ++++: OR = 1.12; 95% CI: 1.06–1.19, P = 2.45 × 10−5 1-arachidonoyl-GPC -> bipolar disorder †: OR = 0.31; 95% CI: 0.23–0.41, P = 3.52 × 10−16 1-arachidonoyl-GPC -> bipolar disorder ++: OR = 0.31; 95% CI: 0.21–0.46, P = 6.76 × 10−9 1-arachidonoyl-GPC -> bipolar disorder +++: OR = 0.30; 95% CI: 0.23–0.41, p = 9.55 × 10−15 1-arachidonoyl-GPC -> bipolar disorder ++++: OR = 0.30; 95% CI: 0.23–0.41, P = 9.53 × 10−15 |
| Liu, 2017 [62] | Two-sample bi-directional MR | PC alkyl-acyl C42:5, PC alkyl-acyl C44:4, PC alkyl-alkyl C44:5 | Fasting blood glucose and T2D | No significant associations were reported between choline types and either health outcome (fasting blood glucose or T2D). | |
| Liu, 2022 [63] | Two-sample bi-directional MR | PC 32:1 and PC 40:5 | Systolic blood pressure (SBP) and diastolic blood pressure (DBP) | No significant associations were reported between circulating PC 32:1 or PC 40:5 with SBP or DBP. | |
| Luo, 2022 [64] | Two-sample MR | Betaine, Choline, TMAO | Heart failure, T2D, atrial fibrillation, CAD, CKD, hypertension, valvular heart disease, valvular heart disease, myocarditis, myocardial infarction, hypertrophic cardiomyopathy, dilated cardiomyopathy | Circulating betaine was associated with relative risk of heart failure and myocardial infarction. For every 1 unit increase in betaine, the relative risk of heart failure and myocardial infarction increased, separately, yet relative risk of CKD decreased. For every 1 unit increase in circulating TMAO, relative risk of elevated SBP, chronic nephritis and T2DM increased. No significant associations were reported between choline and any health. | Betaine -> heart failure †: OR = 1.014; 95% CI: 1.002–1.026, P = 0.03 Betaine -> MI †: OR = 1.017; 95% CI: 1.001–1.033, P = 0.034 Betaine -> CKD †: OR = 0.963; 95% CI: 0.934–0.991, P = 0.01 TMAO -> blood pressure †: OR = 1.071; 95% CI: 1.010–1.136, P = 0.02 TMAO -> chronic nephritis †: OR = 1.031; 95% CI: 1.004–1.058, P = 0.024 TMAO -> T2D †: OR = 1.016; 95% CI: 1.002–1.033, P = 0.029 |
| Mi, 2022 [65] | One-sample MR, two-sample MR, and bi-directional MR | SM, PC, choline | Cholelithiasis risk | Circulating PC, SM, and choline were all negatively associated with cholelithiasis risk. | Choline -> cholelithiasis risk (UK Biobank) †: β = –0.216 (SE: 0.101, P = 0.0331) Choline -> cholelithiasis risk (Finngen) †: β = –0.262 (SE: 0.111, P = 0.0186) SM -> cholelithiasis risk (UK Biobank) †: β = –0.213 (SE: 0.0856, P = 0.0128) SM -> cholelithiasis risk (Finngen) †: β = –0.223 (SE: 0.0846, P = 8.31 × 10−3 SM -> cholelithiasis risk (UK Biobank) ††: β = –0.342 (SE: 0.165, P = 0.0430) Choline -> cholelithiasis risk (UK Biobank) †††: β = –0.230 (SE: 0.0706, P = 0.00184) Choline -> cholelithiasis risk (Finngen) †††: β = –0.151 (SE: 0.0651, P = 0.0204) SM -> Cholelithiasis risk (UK Biobank) †††: β = −0.244 (SE: 0.0721, P = 7.04 × 10−4) SM -> cholelithiasis risk (Finngen) †††: β = –0.159 (SE: 0.0692, P = 0.0220) PC -> cholelithiasis risk (UK Biobank) †††: β = –0.209 (SE: 0.0660, P = 0.00152) PC -> cholelithiasis risk (Finngen) †††: β = –0.121 (SE: 0.0602, P = 0.0452) |
| Porcu, 2021 [66] | Two-sample bi-directional MR | Betaine, LysoPC | T2D | Circulating betaine was negatively associated with fasting plasma glucose. Circulating LysoPC was negatively associated with T2D. | Betaine -> fasting plasma glucose (method NR): β = –0.022 (SE: 0.005, P = 4.89 × 10−5) LysoPC -> T2D (method NR): β = –0.048 (SE: 0.012, P = 3.52 × 10−5) |
| Sun, 2022 [67] | Two-sample, two-stage MR | SM | Alzheimer’s disease (AD) | Each SD increase in circulating sphingomyelin was associated with a lower risk of Alzheimer’s disease. | SM -> Alzheimer’s disease †: OR = 2.53; 95% CI: 1.78–3.59, P = 2.10 × 10−7 SM -> Alzheimer’s disease ††: OR = 4.36; 95% CI: 2.39–7.94, P = 1.51 × 10−6 SM -> Alzheimer’s disease +++ : OR = 4.13; 95% CI: 3.91–4.37, P = 1.00 × 10−299 SM -> Alzheimer’s disease ++++ : OR = 2.78; 95% CI: 2.59–2.99, P = 7.44 × 10−167 |
| Wang, 2022 (BMI and lipidomic biomarkers) [68] | One-sample MR | LysoPCs, Lysoalkyl-PCs, PCs SMs, | BMI and pregnancy BMI | Circulating LysoPC 14:0 was associated with BMI. Circulating LysoPC 14:0 and LysoPC 16:1 were associated with pregnancy BMI. | LysoPC 14:0 -> BMI ∗: SD difference = 0.20; 95% CI: 0.010–0.39, P = 0.042 LysoPC 14:0 -> pregnancy BMI ∗: SD difference = 0.37; 95% CI: 0.04–0.71, P = 0.029 LysoPC 16:1 -> Pregnancy BMI ∗: SD difference = 0.38; 95% CI: 0.020–0.74, P = 0.039 |
| Wang, 2022 (TMAO and its precursors [18] | Two-sample MR | Betaine, choline, TMAO | SBP and DBP | A 1 unit increase in circulating TMAO was associated with a 1 SD increase in SBP, but not DBP. No significant associations were reported between betaine or choline with either SBP or DBP. | TMAO -> SBP †: β = 0.060 (SE: 0.029, P = 0.036) TMAO -> SBP ++++ : β = 0.062 (SE: 0.0270, P = 0.020) TMAO -> SBP ∗∗ : β = 0.060 (SE: 0.0290, P = 0.0420) |
| Wu, 2021 [69] | One-sample and two-sample MR | LysoPC 14:0, LysoPC 15:0 | Metabolic syndrome, overweight/obesity,SBP, uric acid, insulin, fasting plasma glucose, HOMA-IR, BMI, TGs, HDL cholesterol, and Dyslipidemia | Higher concentrations of circulating LysoPC 15:0 was positively associated with the risk of overweight/obesity, dyslipidemia (particularly because of the increased serum TG concentrations), high uric acid, high insulin, and high HOMA-IR, but not with metabolic syndrome and high BP. No significant associations were reported between LysoPC 15:0 and BMI, SBP, or DBP in the 2-stage MR. |
LysoPC 15:0 -> overweight/obesity † (one-sample): OR = 1.84; 95% CI: 1.10–3.08, P = 0.02 LysoPC 15:0 -> dyslipidaemia † (one-sample): OR = 1.78; 95% CI: 1.07–2.98, P = 0.0280 LysoPC 15:0 -> high uric acid † (one-sample): OR = 2.59; 95% CI: 1.36– 4.93, P = 4.00 × 10−3 LysoPC 15:0 -> high insulin † (one-sample): OR = 2.31; 95% CI: 1.27–4.18, P = 6.00 × 10−3 LysoPC 15:0 -> high HOMA-IR † (one-sample): OR = 2.28; 95% CI: 1.26–4.13, P = 7.00 × 10−3 LysoPC 15:0 -> BMI † (one-sample): β = 1.42; 95% CI: 0.480–2.35, P = 3.00 × 10−3 LysoPC 15:0 -> SBP † (one-sample): β = 5.72; 95% CI: 0.190–11.3, P = 0.0430 LysoPC 15:0 -> TGs † (one-sample): β = 0.540; 95% CI: 0.110–0.970, P = 0.0130 LysoPC 15:0 -> uric acid † (one-sample): β = 0.0600; 95% CI: 0.0600–0.0700, P = 0.0450 LysoPC 15:0 -> insulin † (one-sample): β = 1.52; 95% CI: 0.640–2.39, P = 1.00 × 10−3 LysoPC 15:0 -> HOMA-IR † (one-sample): β = 0.230; 95% CI: 0.0600–0.410, P = 1.00 × 10−2 LysoPC 15:0 -> HDL cholesterol † (two-sample): β = 0.147; 95% CI: 0.116–0.177, P = 1.00 × 10−3 LysoPC 15:0 -> TGs † (two-sample): β = 0.07; 95% CI: 0.05–0.10, p = 1.00 × 10−3 |
| Xie, 2013 [70] | One-sample MR and two-sample MR | Betaine | Insulin sensitivity | No significant associations were reported between circulating betaine and insulin sensitivity (as measured by fasting insulin and T2D). | |
| Xu, 2022 [19] | Two-sample, 2-step MR | Choline, PC | CAD and T2D | Choline and PC were negatively associated with T2D. Choline and PC were not associated with CAD. | Choline -> T2D ∗∗: OR = 0.84; 95% CI: 0.76–0.93, P = 1.00 × 10−3 PC -> T2D †: OR = 0.87; 95% CI: 0.77–0.97, P = 0.016 PC -> T2D ∗∗: OR = 0.87; 95% CI: 0.80–0.94, P = 2.00 × 10−3 |
| Yang, 2020 [71] | Two-sample MR | 1-docosahexaenoyl-glyceropho-PC, betaine | Schizophrenia, major depression, bipolar disorder, autism spectrum disorder, attention-deficit/hyperactivity disorder(ADHD) | Circulating 1-docosahexaenoyl-glyceropho-PC was positively associated with ADHD in MR analysis. No significant associations were reported between 1-docosahexaenoyl-glyceropho-PC and schizophrenia, major depression, bipolar disorder, or autism spectrum disorder. | 1-docosahexaenoyl-glyceropho-PC -> ADHD †: OR = 2.98; 95% CI: 1.75–5.07, P = 5.58 × 10−5 |
| Yun, 2020 [72] | Two-sample MR | SM C34:0, SM C34:1, SM C36:0, SM C42:3 | T2D | Circulating SMs C34:0, C36:0, and C34:1 were positively associated with T2D. | SM C34:0 -> T2D ++: OR = 0.86; 95% CI: 0.79–0.92, P = <0.001 SM C36:0 -> T2D ++: OR = 0.91; 95% CI: 0.85–0.98, P = 0.013 SM C34:1 -> T2D ++: OR = 1.19; 95% CI: 1.08–1.31, P = <0.001 |
| Yun, 2022 [73] | Two-sample MR | PC, SMs | SBP and DBP | Circulating SM (OH) C32:2 was significantly associated with lower SBP and DBP, and a modest association with BMI. SM C32:1 was not associated with CAD. | SM (OH) C32:2 -> SBP +: β = –0.74; 95% CI: -1.4 to −0.09, P = 0.026 SM (OH) C32:2 -> SBP ++: β = –0.57; 95% CI: -0.85 to −0.28, p = 9.16 × 10−5 SM (OH) C32:2 -> SBP ∗∗∗ : β = –0.63; 95% CI: −1.16 to −0.10, P = 0.020) SM (OH) C32:2 -> SBP ∗∗∗∗ : β = –2.12; 95% CI: −4.06 to −0.18, P = 0.033 SM (OH) C32:2 -> DBP ++: β = –0.39; 95% CI: -−0.59 to −0.20, P = 7.09 × 10−5 SM (OH) C32:2 -> DBP ∗∗∗ : β = –0.40; 95% CI: −0.74 to −0.05, P = 0.026 SM (OH) C32:2 -> BMI ++: β = –0.010; 95% CI: −0.03 to 0.00, P = 0.047) SM (OH) C32:2 -> BMI ∗∗∗∗ : β = –0.10; 95% CI: −0.19 to −0.020, P = 0.019 SM (2OH) C30:2 -> SBP ∗∗∗∗: β = –0.34; 95% CI: −0.63 to −0.050, P = 0.023) |
| Zhang, 2017 [74] | One-sample | LysoPC (18:2), LysoPC (18:1), SM | CAD | Circulating LysoPC (18:2), LysoPC (18:1), and SM were not associated with CAD risk. | |
| Zhang, 2021 [77] | Two-sample MR | LysoPC, PC, SM | Femoral neck BMD and lumbar spine BMD | Circulating PC was negatively associated with both femoral neck BMD and lumbar spine BMD. Circulating SM was negatively associated with lumbar spine BMD, but not femoral neck BMD. | PC -> Femoral neck BMD †: β = –0.066; 95% CI: NR, P = 6.98 × 10−3 PC -> Lumbar spine BMD †: β = –0.136; 95% CI: NR, P = 1.84 × 10−7 PC -> Lumbar spine BMD †††: β = −0.106; 95% CI: NR, P = 2.96 × 10−3 SM -> Lumbar spine BMD †: β= –0.145; 95% CI: NR, P = 0.0100 SM -> Lumbar spine BMD †††: β = −0.164; 95% CI: NR, P = 7.41 × 10−3 |
| Zhuang, 2021 [75] | Two-sample bi-directional MR | Betaine, choline, TMAO | Alzheimer’s disease | Higher circulating TMAO, betaine, and choline were not significantly associated with the risk of AD/10 units after Bonferroni correction. | |
| Zhuang, 2022 [76] | Two-sample MR | Betaine, choline, TMAO | Inflammatory bowel disease, ulcerative colitis, and Crohn’s disease | Higher circulating betaine was reported as a near significant association with Crohn’s disease. Circulating choline and TMAO were not associated with any health outcome. |
BMI, body mass index; CI, confidence interval; HDL, high-density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; HR, hazard ratio; LysoPC, lysophosphatidylcholines; MR, Mendelian Randomization; OR, odds ratio; PC, phosphatidylcholines; SD, standard deviation; SE, standard error; SM, sphingomyelins; TMAO, trimethylamine N-oxide; 1-arachidonoyl-GPC, 1-arachidonoyl-glycerophosphocholine; 1-docosahexaenoyl-GPC, 1-docosahexaenoyl-glycerophosphocholine; AD, Alzheimer's disease; BP, blood pressure; MRPRESSO, Mendelian Randomization pleiotropy residual sum and outlier; MR-RAPS, Mendelian Randomization robust adjusted profile score; NR, not reported; PIVUS, Prospective Investigation of the Vasculature in Uppsala Seniors; ULSAM, Uppsala Longitudinal Study of Adult Men.
1
Only significant effect sizes were included.
2
The following symbols represent MR method: † inverse variance weighted; †† MR-Egger; ††† weighted median; †††† weighted mode; + mode-based estimate, ++ penalized-inverse variance weighted, +++ MR-RAPS, ++++ maximum likelihood, ∗ 2-stage least-squares regression, ∗∗ MRPRESSO, ∗∗∗ penalized weighted median, ∗∗∗∗ penalized robust MR-Egger