Starting Is Not Staying: Where the Effect Appears
For supplementary breastfeeding support, the decisive question is temporal: does extra help change the decision to begin, the ability to continue, or both? Following the evidence from birth through the first year makes that distinction visible. It also lets us judge the result in the form a decision maker needs: not only as a relative risk, but as an absolute difference against the background rate that would otherwise be expected.
The first transition: a high ceiling at initiation
Initiation is the brief state around birth in which an infant receives breast milk for the first time. It is an important transition, but it is not an empty starting line. In the USPSTF evidence base, most studies began with high proportions of women intending to breastfeed, and many reported that more than two-thirds of women in both intervention and usual-care groups initiated. The opportunity for an added program to raise the rate was therefore limited from the outset.
Across 37 studies involving 15,006 participants, the pooled relative risk for any breastfeeding initiation was 1.01, with a 95% confidence interval from 1.00 to 1.02. A relative risk compares the probability of an outcome in the intervention group with that in the comparison group. A value of 1.01 here does not mean that support causes a 1% increase for every population; it describes a small pooled difference under the studies’ conditions, with the interval close to no difference. The median absolute difference between groups was 1.5 percentage points. If 90 out of 100 people in usual care begin breastfeeding, an additional 1.5 percentage points would represent roughly one or two more people per 100, not a wholesale shift in feeding choices.
The review describes this pattern as a ceiling effect. When the usual-care rate is already high, an intervention can be beneficial and still produce little measurable movement in that endpoint. It may help a particular person, improve the quality of the experience, or make a later continuation difference without changing the proportion who began in the first place. A near-null result for any initiation therefore should not be read as proof that all prenatal education or peripartum support is ineffective. It says that this outcome left relatively little room for an added average effect in the studied settings.
Exclusive breastfeeding initiation showed a different pattern. In 27 studies with 10,622 participants, the pooled relative risk was 1.16, with a 95% confidence interval from 1.05 to 1.29. This estimate was less stable across studies than the any-initiation result. Exclusive initiation depends on whether an infant received anything else in the early days, and early practices can vary by medical circumstance, family preference, facility routine, and the definition used by the study. The evidence therefore supports a possible benefit for exclusive initiation, but it does not justify treating the estimate as a universal effect or assuming that it explains why exclusive outcomes are more responsive.
The practical interpretation is a change in the question. At birth, the relevant contrast may be whether support helps someone cross the threshold into breastfeeding. In many of these trials, that threshold was already crossed by most participants in both arms. The more revealing transitions occur after the first feed, when a family has to sustain a choice through changing circumstances.
The second transition: staying through the early weeks
Continuation is not a single event. It is a sequence of opportunities to stop or persist. A parent may encounter pain, a difficult latch, uncertainty about milk supply, a tired infant, conflicting advice, a change in household support, or the practical demands of recovery. A support intervention can matter at this stage because it arrives when a problem is still alterable: someone can observe a feed, answer a question, help with positioning, or provide reassurance before a temporary difficulty becomes a reason to stop.
The pooled evidence shows a clearer signal after birth. For any breastfeeding reported before three months, the relative risk was 1.06 across 47 trials involving 15,663 participants. For exclusive breastfeeding in the same broad period, the relative risk was 1.21 across 51 trials involving 17,431 participants. The absolute difference between groups varied substantially by study, and the median difference for exclusive breastfeeding in this period was about 7.1 percentage points. The exact timing varied within the category, from roughly two to eleven weeks, so the estimate describes an early postpartum window rather than one universal day.
The pattern is meaningful but not dramatic. If 60% of a usual-care group is still breastfeeding in an early follow-up window, a relative risk of 1.06 would correspond to about 64%, assuming the pooled relationship applied to that baseline. That is a four-percentage-point difference. If the baseline is higher or lower, the absolute change differs. Relative effects are portable only with caution; absolute effects depend on how common the outcome is without the added support.
For exclusive breastfeeding, the relative effect is larger, but the endpoint is also more sensitive. A small change in supplementation can move a participant out of the exclusive category even while breast-milk feeding continues. Different definitions can make that category stricter or broader. The larger relative risk may reflect the way the intervention addresses supplementation, the characteristics of participants who already intend to breastfeed, or measurement differences across trials. The review reports the pattern but does not identify a single settled mechanism. A larger exclusive-breastfeeding relative risk should not be silently translated into a larger health benefit.
This early continuation stage also illustrates why duration and prevalence should not be treated as synonyms. A prevalence estimate asks how many people are breastfeeding at a particular follow-up point. A duration measure asks how long feeding continued or when it stopped. If support delays stopping for some participants but not others, a duration analysis may detect a shift that is not captured by a single visit. Conversely, if studies measure duration with different follow-up periods or incomplete information about censoring, the apparent precision of the result can be misleading. The evidence base generally points toward later stopping in intervention groups, but the duration measures are not standardized enough to provide one pooled timeline.
The third transition: three to six months
Between three and six months, the outcome is no longer simply whether early postpartum difficulties were overcome. Work, childcare, family expectations, health changes, and the need to express or store milk can become more prominent. A program that ends after one hospital conversation may not address the same problems as one that maintains contact across several periods. The evidence does not establish that longer or more intensive support is always superior, but it does show why timing can matter mechanistically: the conditions that threaten continuation change as the infant grows.
For any breastfeeding from three to less than six months, the pooled relative risk was 1.09 across 40 trials involving 17,580 participants. For exclusive breastfeeding in that period, it was 1.31 across 40 trials involving 11,032 participants. The median absolute difference was 4.1 percentage points for any breastfeeding and 5.8 percentage points for exclusive breastfeeding. These are averages across varied populations and interventions. Some individual studies favored usual care, even though the pooled direction favored added support, and the absolute range across outcomes was broad.
The result is best understood as persistence through a transition rather than a one-time conversion. If a population has a 50% usual-care rate of any breastfeeding at six months, the report’s scenario calculations imply that about 15 people would need the intervention for one additional person to be breastfeeding at that point. That number is not a property of “support” in the abstract. It depends on the baseline prevalence and the pooled relative risk. In a population with a different usual-care rate, the same relative effect produces a different number needed to treat.
The number needed to treat is useful only when its assumptions remain visible. It is calculated from a pooled relative effect and a chosen baseline, not counted as a fixed yield from each program. It also describes one outcome at one time point. It does not tell us how many families experienced useful help, avoided a specific problem, or would have preferred a different feeding arrangement. Nor does it say that the people who do not meet the endpoint received no benefit. It is a way to translate an average population effect into a decision scale, not a complete account of value.
The six-month state: stronger evidence, bounded meaning
At six months, 37 trials with 13,579 participants reported any breastfeeding. The pooled relative risk was 1.13, with a 95% confidence interval from 1.05 to 1.22. The median absolute difference between intervention and usual care was six percentage points. In separate analyses of 37 trials involving 14,398 participants, exclusive breastfeeding had a pooled relative risk of 1.46, with a 95% confidence interval from 1.20 to 1.78. The median absolute difference was 3.2 percentage points; the median prevalence was 14.6% in intervention groups and 8.7% in control groups.
The contrast between those two scales is essential. “Forty-six percent higher” is a relative statement. It does not mean that exclusive breastfeeding increased by 46 percentage points. In the included studies, the typical absolute difference was much smaller. A relative effect can look large when the baseline outcome is uncommon. For example, the report estimates that if only about 5% of a population is exclusively breastfeeding at six months, roughly 43 people would need to receive an intervention to yield one additional person at that endpoint, under the pooled estimate and its assumptions. Both the relative and absolute descriptions are valid; they answer different questions.
The six-month finding is the most defensible headline in the evidence. The USPSTF rates the breastfeeding-outcome evidence as moderate for benefit, meaning there is moderate confidence that the observed relationship reflects a real effect, while further research could change the estimate or its precision. The confidence comes from the consistency of the direction across many studies, not from the claim that all interventions were alike or that every setting should expect the same percentage-point gain.
The distinction between persistence and transformation remains important. Support did not create a new feeding intention in every participant. It was associated with a greater likelihood that some participants were still breastfeeding at six months, above what occurred with their setting’s usual care. That is a meaningful public-health and family-level result, especially where people want to continue but face preventable barriers. It is not evidence that support can overcome every medical, social, economic, or personal constraint, and it is not a reason to treat the endpoint as a test of parental virtue.
The one-year state: an evidence boundary, not a fixed failure point
At twelve months, the trajectory becomes less certain. A meta-analysis of eight trials involving 4,607 participants found a relative risk of 1.04 for any breastfeeding, with a 95% confidence interval from 0.91 to 1.18. Four trials favored the intervention and four favored usual care; only one found a statistically significant difference. The median absolute difference was 1.3 percentage points. The Cochrane review likewise reported too few studies to determine whether breastfeeding-only support reduced stopping at nine or twelve months.
This does not prove that support ceases to work at exactly twelve months. The result may reflect limited follow-up, differences in long-term support, changing social conditions, or the fact that fewer trials measured the endpoint. It does show that the six-month estimate should not be projected forward as though it were a one-year guarantee. A responsible conclusion ends where the evidence becomes mixed rather than filling the gap with optimism or cynicism.
Duration and intensity provide additional but less uniform views of the same lifecycle. Duration can capture delayed stopping, while intensity can describe a gradual shift in the share of feeds that are breast milk. Only a small number of trials reported intensity, and they used different scales or definitions. In general, intervention groups tended to show greater intensity, but the evidence is too varied to turn that tendency into a single quantitative expectation. These measures matter because families can move between exclusive, mixed, expressed-milk, and direct-feeding patterns without fitting neatly into a binary label.
Seen as a lifecycle, the evidence has a coherent shape. Initiation begins near a high ceiling for many enrolled participants. The early weeks provide room for practical support to prevent avoidable stopping. Continuation through three to six months shows the clearest and most consistently supported average benefit. At one year, fewer studies and more varied circumstances make the estimate uncertain. The pattern does not say that one transition is morally more important than another. It says that the causal opportunity for extra support, and the outcome’s sensitivity to that support, changes with time.
That time-course result is now strong enough to support a careful six-month judgment, but not a recipe. The next question is why such different interventions—professional, peer, face-to-face, remote, brief, or repeated—can point in the same direction, and whether the evidence can identify which parts of the support actually matter.