We next analyzed the data using Cox regression, an analysis method designed for longitudinal data on event times, such as time until death. The outcome variable was time until cessation of all breastfeeding. The median failure time (cessation) was 72 days (95% CI, 68-78) with interquartile range from 53 to 192 days. Seventy-four percent had ceased breastfeeding by 6 months and 26% were censored because of continued breastfeeding at 6 months when analysis ended or earlier loss to follow-up.
Table 4 reports the relative hazard ratios and 95% confidence intervals at various levels of child care, pacifier use, and digit sucking, while adjusting for the other potential confounders considered (see Methods section). The baseline category (or reference cell) is a child with no child care and no non-nutritive sucking. We see from Table 4 that the estimated risk of breastfeeding cessation is the highest, with a value of 1.88 (95% CI, 1.36-2.62), for a child who sucks on both pacifier and digit at no child care days. This hazard ratio drops to 1.52 (95% CI, 1.03-2.25) at 15 child care days and then becomes nonsignificant at 30 and 60 child care days.
Our results in Table 4 also show that pacifier use at zero child care days has a significant effect in that a child who sucks only on a pacifier has a 67% increase in the hazard of cessation of breastfeeding, compared with a child with no non-nutritive sucking. At higher levels of child care days, this effect changes and becomes a protective effect, although this effect was not significant at 15 child care days, was significant at 30 child care days, and was borderline significant at 60 child care days. Finally, the effect of digit sucking and child care by themselves tended not to be significant at the 0.05 level, with the one exception at 15 child care days where there is a significant effect of 1.41 (95% CI, 1.02-1.96).
Discussion
Our findings concerning pacifiers are generally consistent with several recent studies that have demonstrated associations between pacifier use and reduced breastfeeding, including the few reported longitudinal studies. However, these other studies did not control for child care attendance. We found that the effect of pacifier use changed with increasing number of child care days. For example, in the absence of child care, children who sucked on a pacifier were about 1.7 times as likely to cease breastfeeding than children who did not use a pacifier. For 15 days to 60 days of child care, the hazard ratios were less than 1.0, with results statistically significant at only 30 days.
Furthermore, our analyses showed the joint effect of pacifier use and digit sucking at various child care days. We found a significant reduction in breastfeeding for children who use both pacifier and a digit by the age of 6 weeks. But this joint non-nutritive effect reduces to being nonsignificant with 30 or more child care days. Although we found that digit sucking and child care days by themselves had little effect on cessation of breastfeeding, it was important to consider them because these variables significantly interacted with pacifier use.
Our study found that for infants who did not attend child care, pacifier use significantly increased the odds of breastfeeding cessation, as did the combination of pacifier use and digit sucking. However, digit sucking with no pacifier use in the absence of child care did not increase the odds of breastfeeding cessation. In contrast, for infants who attended child care for 30 days in the first 6 months of life, pacifier use alone appeared to be somewhat protective in maintaining breastfeeding, while digit sucking, either alone or in combination with pacifier, increased the odds of breastfeeding cessation, with significance at 15 days. It is possible that pacifiers were used sparingly in child care, whereas digits were available and more widely used, so that non-nutritive sucking interference with breastfeeding was more strongly influenced by digit sucking. Alternatively, it is possible that mothers who placed their infants in child care early in life used pacifiers differently than mothers who did not. That is, for non-child care infants, pacifiers may have been part of a planned strategy to wean from breastfeeding, whereas for children in child care, pacifiers may have been part of a planned strategy to encourage sucking behavior and comfort children until the mother was available for breastfeeding. In such a scenario, digit sucking was less under parental control, particularly at child care, so that it may have interfered with breastfeeding despite parental planning or desires.