Table 5.
Observed and Bayesian posterior means, 95% credible intervals and posterior probabilities of a dose effect for early preterm birth rate by quartile baseline DHA status and adherence (postpartum RBC-PL-DHA ≥8%) with the high dose.
| Observed early preterm births <34 weeks, n/N (%) | Posterior mean % of EPTB (95% Bayesian credible interval)a | Bayesian posterior probabilities | |||||||
|---|---|---|---|---|---|---|---|---|---|
| 200 mg | 1000- mg | 1000+ mg | 200mg | 1000- mg | 1000+ mg | 1000+ mg better than 200mg | 1000+ mg better 1000- mg | 1000- mg better than 200mg | |
| DHA Enrollment (%) | |||||||||
| Quartile 1 (<5.1%] | 5/112 (4.5) | 4/96 (4.2) | 1/46 (2.2) | 3.4 (1.2,7.9) | 2.5 (1.0,6.7) | 1.3 (0.3,3.5) | 0.92 | 0.82 | 0.67 |
| Quartile 2 (5.1–6.2%] | 6/132 (4.6) | 0/51 (0.0) | 0/75 (0.0) | 3.5 (1.5,7.7) | 1.6 (0.0,4.4) | 1.1 (0.3,2.6) | 0.96 | 0.64 | 0.85 |
| Quartile 3 (6.2–7.4%] | 1/110 (0.9) | 1/52 (1.9) | 2/86 (2.3) | 1.3 (0.1,3.8) | 1.9 (0.2,5.2) | 1.3 (0.3,3.4) | 0.48 | 0.66 | 0.32 |
| Quartile 4 (>7.4%)] | 0/136 (0.0) | 0/21 (0.0) | 1/111 (0.9) | 0.6 (0.0,3.3) | 1.7 (0.0,5.0) | 1.2 (0.2,2.8) | 0.34 | 0.65 | 0.24 |
a
Bayesian hierarchical model was fitted using the approach in Berry and Berry,8 except mu~N(−3.1,10^2). This approach avoids false discoveries in subgroups (quartiles). We burned-in 10,000 draws. and used 40,000 draws for inference. High adherence, 1000+ mg, postpartum RBC-PL-DHA ≥ 8.0%; low adherence, 1000- mg, postpartum RBC-PL-DHA <8.0%.