Conceived and designed the experiments: STC SS AMR RWT PMG ML. Performed the experiments: WA JLE SLM SS ML. Analyzed the data: ML SS CMF. Contributed reagents/materials/analysis tools: WA JLE SLM CMF. Wrote the paper: ML SS SLM RWT AMR PMG JLE.
The authors have declared that no competing interests exist.
Urinary betaine excretion positively correlated with plasma homocysteine in outpatients attending a lipid disorders clinic (lipid clinic study). We aimed to confirm this in subjects with established vascular disease.
The correlation between betaine excretion and homocysteine was compared in samples collected from subjects 4 months after hospitalization for an acute coronary episode (ACS study, 415 urine samples) and from 158 sequential patients visiting a lipid disorders clinic.
In contrast to the lipid clinic study, betaine excretion and plasma homocysteine did not correlate in the total ACS cohort. Differences between the patient groups included age, non-HDL cholesterol and medication. In ACS subjects with below median betaine excretion, excretion correlated (using log transformed data) negatively with plasma homocysteine (r = −0.17, p = 0.019, n = 199), with no correlation in the corresponding subset of the lipid clinic subjects. In ACS subjects with above median betaine excretion a positive trend (r = +0.10) between betaine excretion and homocysteine was not significant; the corresponding correlation in lipid clinic subjects was r = +0.42 (p = 0.0001). In ACS subjects, correlations were stronger when plasma non-HDL cholesterol and betaine excretion were above the median, r = +0.20 (p = 0.045); in subjects above median non-HDL cholesterol and below median betaine excretion, r = −0.26 (p = 0.012). ACS subjects taking diuretics or proton pump inhibitors had stronger correlations, negative with lower betaine excretion and positive with higher betaine excretion.
Betaine excretion correlates with homocysteine in subjects with elevated blood lipids.
Betaine has central roles in mammalian metabolism both as an osmolyte and in the storage and transfer of one-carbon units
A small sample of ambulant elderly subjects provided evidence that the positive correlation between urinary betaine and plasma homocysteine is characteristic of groups with elevated plasma lipids
All study protocols were approved by the Canterbury Ethics Committee, and all subjects gave written informed consent.
The “ACS cohort” in this report was the previously described
The lipid clinic cohort has been previously described
In both studies fasting plasma and morning urine samples were collected on all subjects. Blood for homocysteine measurements was collected on ice. Samples were assayed for high volume laboratory tests within hours of collection, specimens for homocysteine, betaine and dimethylglycine assays were frozen at −16°C and assayed within two weeks.
Drug treatments and the diagnosis of diabetes were taken from clinical records.
Betaine and
Drug treatment and clinical diagnoses of diabetes were taken from the clinical records.
Statistical analyses were carried out using SigmaPlot for Windows version 11.2 (Systat Software Inc), which incorporates SigmaStat.
In both sets of study data, urine betaine and
For estimating correlations, plasma betaine and
Correlation and regression statistics are reported with the number of valid data pairs quoted in the results, and with significance based on these. As a result of missing data these comparisons are usually based on fewer than the number of subjects in the ACS study. The significances of differences between correlation coefficients were evaluated using Fisher r-to-z transformations.
The lipid clinic cohort (for which diabetes was an exclusion criterion) was younger with a higher proportion of female subjects than the ACS cohort (
ACS study | Lipid clinic study | |||
Males | Females | Males | Females | |
Number | 297 | 118 | 77 | 81 |
Age (median, total range) | 66 (55–93) | 72 (51–91) |
52 (43–59) |
56 (48–63) |
With diabetes (%) | 16% | 20% | 0 | 0 |
BMI (median, IQ range) | 26.3 (24.5–29.4) | 27.4 (22.9–31.7) | 28.4 |
27.2 (24.3–31.0) |
Pl betaine µmol/L | 42.1 (34.4–54.7) | 38.0(30.5–44.1) |
24.9 (19.2–31.9) |
20.0 (15.0–25.6) |
Pl dimethylglycine µmol/L | 3.7 (2.6–4.9) | 3.2 (1.9–4.7) |
2.1 (1.5–2.7) |
1.8 (1.4–2.2) |
Pl total homocysteine µmol/L | 12.5 (10.4–15.8) | 12.3 (10.1–15.6) | 9.8 (8.6–12.2) |
9.6 (7.7–12.2) |
Pl creatinine µmol/L | 100 (89–110) | 80 (71–100) |
80 (70–90) |
70 (60–80) |
Pl urea mmol/L | 6.5 (5.4–8.2) | 6.4 (4.9–8.0) | – | – |
Pl triglycerides mmol/L | 1.50 (1.05–2.08) | 1.47 (1.01–2.03) | 2.25 (1.60–3.30) |
1.82 (1.27–2.35) |
Pl HDL cholesterol mmol/L | 1.07 (0.89–1.25) | 1.25 (1.06–1.50) |
1.01 (0.87–1.30) | 1.29 (1.10–1.55) |
Pl non-HDL cholesterol mmol/L | 3.0 (2.5–3.7) | 3.3 (2.6–4.3) |
5.8 (5.1–6.6) |
5.5 (4.8–6.3) |
Betaine excretion | 9.4 (6.0–18.1) | 8.2 (4.4–17.7) | 7.4 (5.0–13.9) |
8.1 (4.8–15.6) |
Dimethylglycine excretion | 2.9 (1.5–5.8) | 2.4 (1.4–4.3) | 3.8 (1.7–7.1) | 3.2 (1.8–6.6) |
% on statins | 89 | 81 | 17 | 21 |
% on ACE inhibitors | 57 | 49 | 17 | 9 |
% on angiotensin II antagonists | 8 | 10 | 0 | 0 |
% on diuretics | 26 | 37 | 6 | 6 |
% on aspirin | 94 | 96 | 12 | 9 |
% on clopidogrel | 41 | 30 | 0 | 0 |
% on beta-blockers | 88 | 81 | 9 | 9 |
% on calcium channel blockers | 29 | 41 | 3 | 4 |
% on proton pump inhibitors | 39 | 42 | 8 | 5 |
% current smoker | 6 | 10 | – | – |
The ACS study data are on subjects who supplied urine samples. Median values with interquartile ranges unless otherwise stated. cr: creatinine. DMG:
p<0.05;
p<0.01;
p<0.001; difference between genders,
p<0.05;
p<0.01;
p<0.001.
In the ACS population, plasma betaine, dimethylglycine and homocysteine concentrations approximated a log-normal distribution, so Pearson's correlation was estimated using log-transformed data. Plasma dimethylglycine concentrations correlated with both plasma homocysteine and plasma betaine concentrations (
Plasma dimethylglycine (µmol/L) | Plasma homocysteine (µmol/L) | Betaine excretion (mmol/mole) | Dimethylglycine excretion (mmol/mole) | |
A. ACS study | ||||
Plasma betaine (µmol/L) | ||||
r = |
|
−0.06 | +0.027 | −0.016 |
p = |
|
0.15 | 0.59 | 0.75 |
n = | 523 | 510 | 393 | 393 |
Plasma dimethylglycine (µmol/L) | ||||
r = |
|
+0.05 |
|
|
p = |
|
0.33 |
|
|
n = | 510 | 393 | 393 | |
Plasma homocysteine (µmol/L) | ||||
r = | +0.019 | +0.017 | ||
p = | 0.70 | 0.74 | ||
n = | 402 | 402 | ||
Betaine excretion (mmol/mole) | ||||
r = |
|
|||
p = |
|
|||
n = | 415 | |||
B. Lipid clinic study | ||||
Plasma betaine (µmol/L) | ||||
r = |
|
−0.02 | +0.08 | −0.10 |
p = |
|
0.77 | 0.30 | 0.22 |
n = | 158 | 158 | 158 | 158 |
Plasma dimethylglycine (µmol/L) | ||||
r = |
|
+0.03 | +0.03 | |
p = |
|
0.69 | 0.67 | |
n = | 158 | 158 | 158 | |
Plasma homocysteine (µmol/L) | ||||
r = |
|
|
||
p = |
|
|
||
n = | 158 | 158 | ||
Betaine excretion (mmol/mole) | ||||
r = |
|
|||
p = |
|
|||
n = | 158 |
Pearson's correlation coefficients calculated using log-transformed data. Plasma betaine and dimethylglycine concentrations corrected for gender difference. Excretions measured as mmole/mole creatinine. Significant correlations in bold.
For comparison, the data from the lipid clinic cohort was re-analyzed in exactly the same way, comparing log(homocysteine) with log(urinary excretions), despite a more normal distribution of betaine and homocysteine concentrations. The notable difference compared with the ACS cohort was that the urine betaine and dimethylglycine excretions correlated with plasma homocysteine. This correlation between log(urine betaine excretion) and log(plasma homocysteine) was stronger in the male subjects (r = +0.51, p<0.0001) than in the female subjects (r = +0.23, p = 0.039); difference in correlations, p = 0.048.
One difference between the cohorts was the inclusion of subjects with diabetes in the ACS cohort. Excluding subjects with diabetes did not change the differences between the cohorts shown in
ACS study subjects with (n = 90 plasma, 68 urine) and without (n = 425 plasma, 331 urine) diabetes; box shows median and interquartile ranges. Only the urinary excretions are significantly different (rank sum test), with excretions higher in diabetes, p<0.001 for betaine, p = 0.001 for dimethylglycine (DMG).
We have found that both high and low betaine excretion may be associated with disease
Plasma dimethylglycine (µmol/L) | Plasma homocysteine (µmol/L) | Betaine excretion (mmol/mole) | Dimethylglycine excretion (mmol/mole) | |
A. Above median betaine excretion | ||||
Plasma betaine (µmol/L) | ||||
r = |
|
−0.07 | −0.05 | −0.015 |
p = |
|
0.36 | 0.50 | 0.84 |
n = | 194 | 193 | 194 | 194 |
Plasma dimethylglycine (µmol/L) | ||||
r = |
|
−0.03 |
|
|
p = |
|
0.72 |
|
|
n = | 193 | 194 | 194 | |
Plasma homocysteine (µmol/L) | ||||
r = | +0.10 | +0.07 | ||
p = | 0.14 | 0.29 | ||
n = | 194 | 194 | ||
Betaine excretion (mmol/mole) | ||||
r = |
|
|||
p = |
|
|||
n = | 208 | |||
B. Below median betaine excretion | ||||
Plasma betaine (µmol/L) | ||||
r = |
|
−0.012 | +0.015 | −0.07 |
p = |
|
0.87 | 0.84 | 0.33 |
n = | 200 | 192 | 200 | 200 |
Plasma dimethylglycine (µmol/L) | ||||
r = |
|
−0.12 | +0.024 | |
p = |
|
0.096 | 0.74 | |
n = | 192 | 200 | 200 | |
Plasma homocysteine (µmol/L) | ||||
r = |
|
−0.056 | ||
p = |
|
0.44 | ||
n = | 199 | 199 | ||
Betaine excretion (mmol/mole) | ||||
r = |
|
|||
p = |
|
|||
n = | 208 |
Pearson's correlation coefficients calculated using log-transformed data. Plasma betaine and dimethylglycine concentrations corrected for gender difference. Excretions measured as mmole/mole creatinine. Significant correlations in bold.
Since the largest differences between the ACS cohort and the lipid clinic cohort were in non-HDL cholesterol, age, and plasma betaine the populations were divided on the basis of these factors into those subjects above the median and those below the median for each factor (
Above median betaine excretion | Below median betaine excretion | |
|
||
Above median non-HDL-cholesterol |
|
|
Below median non-HDL-cholesterol | +0.01 (0.96) | −0.09 (0.38) |
Above median age | +0.11 (0.29) |
|
Below median age | +0.08 (0.44) | −0.04 (0.69) |
Above median plasma betaine | +0.11 (0.28) |
|
Below median plasma betaine | +0.01 (0.90) | −0.05 (0.60) |
|
||
Above median non-HDL-cholesterol |
|
−0.04 (0.72) |
Below median non-HDL-cholesterol | +0.06 (0.57) | −0.01 (0.96) |
Above median age | −0.19 (0.062) | −0.10 (0.34) |
Below median age | −0.01 (0.93) | −0.04 (0.68) |
Above median plasma betaine | +0.14 (0.18) | +0.08 (0.41) |
Below median plasma betaine |
|
−0.02 (0.87) |
|
||
Above median non-HDL-cholesterol |
|
+0.09 (0.37) |
Below median non-HDL-cholesterol | −0.04 (0.72) | +0.03 (0.74) |
Above median age | −0.08 (0.45) | +0.00 (1.00) |
Below median age | +0.03 (0.76) | +0.01 (0.93) |
Above median plasma betaine | +0.20 (0.053) |
|
Below median plasma betaine | −0.14 (0.16) | −0.09 (0.35) |
Values are Pearson correlation coefficients between log-transformed variables, with significance (p) in parentheses after correlations. Significant correlations (p<0.05) in bold. Gender corrected plasma betaine. Urine betaine excretion measured as mmol betaine/mole creatinine. n = 101 (97 for plasma betaine).
Non-HDL cholesterol consistently did not correlate with homocysteine, neither in univariate correlations, nor when multiple regression models were estimated in these subgroups. In models with log(plasma homocysteine) as the dependent variable, and the independent variables plasma (gender adjusted) and urine betaine (both log transformed), non-HDL cholesterol and age, all the independent variables had variance inflation factors <1.2, with age appearing as the strongest predictor of homocysteine.
The lipid clinic study data was also divided into groups based on the medians of betaine excretion and plasma non-HDL cholesterol. In the above median betaine excretion and above median non-HDL cholesterol group (n = 39), the Pearson's correlation coefficient between plasma homocysteine and urine betaine excretion (both log-transformed) was r = +0.52 (p = 0.00066) whereas in the below median betaine excretion and above median non-HDL cholesterol group (n = 40) the corresponding correlation was not significant, r = +0.13 (p = 0.42).
In the ACS cohort, plasma betaine concentrations and betaine excretions were significantly different in subjects taking some medications (
Not taking drug | Taking drug | |||||||
Drug category | n | pl betaine | n | betaine excretion | n | pl betaine | n | betaine excretion |
Statins | 68 | 40.2 | 52 | 7.6 | 436 | 45.9 |
349 | 9.0 |
Fibrates | 493 | 45.2 | 392 | 8.8 | 12 | 33.0 |
10 | 99.8 |
Diuretics | 359 | 45.2 | 282 | 9.2 | 144 | 44.1 | 116 | 8.1 |
Beta blockers | 78 | 39.8 | 56 | 8.6 | 426 | 46.1 |
343 | 9.0 |
ACE inhibitors | 236 | 45.0 | 181 | 9.1 | 268 | 45.1 | 218 | 8.8 |
Angiotensin II antagonists | 460 | 45.3 | 362 | 8.8 | 40 | 42.0 |
34 | 10.7 |
Clopidogrel | 308 | 45.3 | 248 | 8.7 | 196 | 44.7 | 151 | 9.1 |
Calcium channel blockers | 344 | 45.3 | 270 | 8.5 | 160 | 43.3 | 129 | 9.8 |
Proton pump inhibitors | 298 | 43.7 | 242 | 8.5 | 207 | 46.5 | 158 | 9.7 |
Current smokers | 474 | 45.3 | 372 | 8.8 | 31 | 37.6 | 28 | 12.1 |
p<0.05;
p<0.01;
p<0.001 compared with subjects not taking the medication.
Median plasma betaine (corrected for gender): µmol/L; betaine excretion: mmol urine betaine/mole creatinine.
Because medication was one of the main differences between the ACS cohort and the lipid clinic cohort (
Correlations between plasma homocysteine and betaine excretion in: | Above median betaine excretion | Below median betaine excretion | ||||
n | r | p | n | r | p | |
|
||||||
All subjects without diabetes | 73 |
|
|
87 |
|
|
No diuretics | 52 | +0.21 | 0.13 | 61 | +0.12 | 0.36 |
With diuretics | 17 | +0.11 | 0.67 | 26 |
|
|
No beta blocker | 7 | +0.57 | 0.18 | 14 | +0.12 | 0.67 |
With beta blocker | 65 |
|
|
73 |
|
|
No ACE inhibitor | 38 | +0.13 | 0.42 | 36 |
|
|
With ACE inhibitor | 34 |
|
|
51 | −0.15 | 0.30 |
No clopidogrel | 43 | +0.11 | 0.49 | 58 | −0.15 | 0.26 |
With clopidogrel | 29 |
|
|
29 |
|
|
No calcium channel blocker | 46 | +0.10 | 0.52 | 61 | −0.19 | 0.14 |
With calcium channel blocker | 26 | +0.39 | 0.051 | 26 | −0.21 | 0.31 |
No proton pump inhibitor | 45 |
|
|
52 | +0.01 | 0.95 |
With proton pump inhibitor | 28 | +0.17 | 0.38 | 35 |
|
|
All data on subjects with above median plasma non-HDL-cholesterol and with subjects with diabetes excluded. Pearson's correlation coefficients between log (plasma tHcy) and log(urine betaine/creatinine). Bold significant (p<0.05).
To a lesser extent, relationships other than that between homocysteine and betaine excretion were changed in subgroups treated with some drugs (
In our previous study of subjects attending a lipid disorder clinic, we found a strong association between plasma homocysteine and betaine excretion
When we examined the cohort of subjects without diabetes who were attending a lipid disorders clinic, we expected that this group would have a high proportion of subjects with pre-diabetic conditions, and we found in this group that betaine (both plasma concentrations and urinary excretion) was a major determinant of plasma homocysteine concentrations
Here we ask why this association between betaine excretion and homocysteine was not subsequently observed in the ACS study. This population, and the lipid disorders clinic cohort, were quite different and neither is typical of the general population. The contrast in control of homocysteine between the cohorts is most strongly associated with the large difference in the plasma lipid profiles of the two populations, and another factor is the difference in the degree of medication of the subjects. The correlation between urine betaine excretion and plasma homocysteine increases in subsets of the present study with higher lipids, but the number of subjects becomes too small to detect significant correlations at lipid concentrations well below those in the earlier study. A confounding factor is the presence of subjects with abnormally
In conclusion, the results of the two studies can be reconciled, but the apparent contradictions are a warning about generalizations based on a selected patient population. The reconciliation confirms the association between betaine and lipid metabolism, and is consistent with a role for betaine in the metabolic syndrome. These cross-sectional studies suggest a number of hypotheses to be tested. The results are consistent with our previous suggestions
Christopher McEntyre and Warwick Dellow carried out many of the betaine assays and Linda Pike the homocysteine assays. Lorraine Skelton provided indispensable support for patient recruitment.