Skip to main content

Advertisement

Table 3 Relationship between duration of breastfeeding and IQ at 15

From: Multiple imputation using linked proxy outcome data resulted in important bias reduction and efficiency gains: a simulation study

  Duration of breast-feeding Analysis approach
Complete records analysis (n = 4152) Multiple imputation, using only KS4 attainmentb (n = 13,975) Multiple imputation, using KS2, KS3 and KS4 attainmentc (n = 13,975)
Difference in mean IQ (95% CI) Difference in mean IQ (95% CI) Gain in precisiond (%) FMIe (%) Difference in mean IQ (95% CI) Gain in precisiond (%) FMIe (%)
Unadjusted results Never/< 1 month
1 to < 3 months 1.9 (0.6, 3.2) 3.2 (2.2, 4.3) 47 60 3.4 (2.4, 4.4) 65 55
3 to < 6 months 5.1 (4.0, 6.3) 6.6 (5.6, 7.6) 36 58 6.8 (5.9, 7.8) 43 56
6 months + 7.5 (6.6, 8.5) 9.3 (8.5, 10.1) 38 59 9.6 (8.8, 10.3) 58 53
Adjusteda results Never/< 1 month
1 to < 3 months 0.8 (− 0.4, 2.0) 1.3 (0.3, 2.4) 36 63 1.5 (0.5, 2.4) 54 58
3 to < 6 months 2.6 (1.5, 3.7) 3.2 (2.2, 4.2) 26 61 3.4 (2.4, 4.3) 27 62
6 months + (2.5, 4.4) 4.2 (3.4, 5.0) 36 58 4.4 (3.6, 5.2) 39 57
  1. aAdjusted for sex, maternal and paternal education, occupational social class, parity, maternal age, ethnicity, family adversity index, smoking in pregnancy and housing tenure during pregnancy
  2. bIQ predicted from KS4 points cubed (best fitting fractional polynomial of degree 1), plus all other factors. Imputation model for IQ also included an interaction between KS4 points cubed and mother’s education
  3. cIQ additionally predicted from KS2 points squared and KS3 points squared
  4. dRelative to complete records analysis
  5. eFMI: Fraction of missing information