Table 3.
Unadjusted and adjusted survey-weighted means of blood biomarkers by diet group in women of childbearing age (18-45 years), NDNS Years 1–11
| Outcome | Omn N | Veg N | Omnivore | Vegetarian | Unadjusted difference (95% CI), p-valueᵃ | Omn N | Veg N | Omnivore | Vegetarian | Adjusted difference (95% CI), p-value |
|---|---|---|---|---|---|---|---|---|---|---|
| Serum B12 (pmol/L) | 959 | 41 | 271.3 (260.8–281.7) | 231.0 (183.9–278.1) | − 40.3 (− 88.6–8.0), p = 0.102 | 822 | 34 | 289.4 (261.5–317.3) | 260.2 (205.7–314.7) | − 29.2 (− 83.5–25.2), p = 0.293 |
| Holotranscobalamin (pmol/L) | 452 | 21 | 63.1 (60.0–66.2) | 39.8 (33.1–46.5) | − 23.3 (− 30.7– −15.9), p = < 0.001 | 381 | 17 | 61.9 (56.7–67.1) | 40.4 (31.1–49.6) | − 21.5 (-30.6– −12.5), p = < 0.001 |
| Red cell folate (nmol/L) | 926 | 41 | 528.9 (507.6–550.3) | 583.4 (432.4–734.5) | + 54.5 (− 98.2–207.2), p = 0.484 | 794 | 33 | 546.5 (511.6–581.4) | 552.4 (456.6–648.3) | +6.0 (-86.2–98.1), p = 0.899 |
| Serum folate (nmol/L) | 967 | 42 | 18.1 (16.0–20.2) | 22.1 (16.8–27.3) | + 4.0 (− 1.7–9.6), p = 0.170 | 831 | 34 | 17.6 (15.9–19.3) | 18.8 (14.3–23.3) | +1.2 (− 3.5–5.9), p = 0.620 |
| Homocysteine (µmol/L)ᵃ | 507 | 18 | 8.49 (8.17–8.81) | 11.22 (9.21–13.23) | + 2.73 (0.69–4.77), p = 0.009 | 440 | 15 | 7.95 (7.54–8.39) | 11.80 (9.73–14.32) | Ratio: 1.485 (1.234–1.786), p = < 0.001 |
Significant p-values (<0.05) shown in bold.
ᵃHomocysteine was log-transformed prior to modelling due to right-skewed distribution; the unadjusted difference is shown on the original scale for comparability. Back-transformed geometric mean ratios are reported for the adjusted difference. A ratio >1 indicates higher homocysteine in vegetarians.
Confounder-adjusted marginal means estimated using weighted linear regression, adjusting for age, ethnicity, education, income, and BMI. Supplement use was excluded from B12 biomarker models to avoid overadjustment. Adjusted models were restricted to participants with complete data on all covariates (age, ethnicity, education, income, and BMI); adjusted sample sizes are therefore smaller than the unadjusted sample sizes shown, and unadjusted and adjusted estimates should not be interpreted as derived from the same underlying sample.