Cochrane's New Breastfeeding-Depression Review Misrepresents the Evidence
Cochrane said they'd ask me to review it before publication; they didn't, published it today, and it's worse than I feared
You can lead a horse to water, but you can’t make it think.
Almost two years ago, I was asked and agreed to review a planned Cochrane Review on the evidence on “Breastfeeding interventions for preventing postpartum depression” (emails, May 11, 2023 and May 19, 2023). But I was never sent the draft. The review was published today (Lenells et al, Cochrane Database of Systematic Reviews, Feb. 18, 2025). It makes the mistakes I warned about, and then some.
What the review says
Background: “a positive early, and continued, breastfeeding experience may reduce the risk of having postpartum depression. Breastfeeding interventions that increase the duration and exclusivity of breastfeeding may help prevent or reduce postpartum depressive symptoms.”
Conclusion: “There is low‐certainty evidence that psychosocial breastfeeding interventions may prevent postpartum depression in the short term and increase the duration of breastfeeding in the long‐term… The included studies did not report any adverse events directly related to the interventions… ”
Plain language summary, key messages: “The studies did not state any adverse events directly related to the interventions, but in most studies, we did not know if this outcome had been measured.”
What the reviewed evidence actually shows
On primary outcomes: The authors report the incidence of postpartum depression (“PPD”) immediately after psychosocial breastfeeding interventions (“interventions”) as ranging in the 95% compatability interval (CI) from .23 to 2.70. This means the interventions may have immediately reduced the risk by up to 77% — or increased it by up to 170%.
In the 1-3 month time frame after the interventions, it reports PPD incidence ranging .14 -.93 (95% CI). This means the interventions may have reduced the risk by as much as 86% or as little as 7%.
On secondary outcomes:
Inconclusive decrease or increase in Edinburgh Postnatal Depression Scale (EPDS) scores (95% CI -1.63-.28).
Small possible Zung Self-Rating Anxiety Scale score reduction (95% CI -4.36- 0-.24).
Inconclusive “exclusive breastfeeding” rate change (95%CI .96-1.51) ranging from a possible nominal decrease to an increase of up to 51%.
Small possible breastfeeding duration increase (95% CI 1.08-2.5) of between 8% and 150%. This effect is not translated into practical terms so readers can evaluate its importance (e.g., weeks). Nor is it possible to easily check whether the trials included in the review had any way to assess possible social desirability/reporting effects (i.e., whether women may have lied to please researchers who had promoted breastfeeding to them): The references and other components of the review are not open access.
Pinpointing the discrepancy
The review selectively highlights the interventions’ possible benefits while misrepresenting the evidence on their possible harms. Ignoring evidence of possible, substantial immediate increased PPD risk offers a prime example.
What I warned
Between March and May of 2023, I repeatedly raised concerns that breastfeeding promotion could increase PPD risk, Lenells et al appeared to exhibit confirmation bias by not considering the possibility of iatrogenesis, and we know how to do better science by putting causality first and running statistical analyses after as per the causal revolution set — but no one in this field appears to be doing it. So I drew the missing causal diagram (DAG) and contacted Lenells (the review’s first author) along with the review manager and Cochrane Support.
In particular, I expressed concern that the planned review assumed a one-way causal link between breastfeeding and reduced depression. But diagramming what we know about relevant causal pathways suggests the true story is more complex. Shared biological mechanisms could put the same women at risk for both breastfeeding and mental health problems. Failing to recognize the biological dimension of breastfeeding problems, interventions intended to promote breastfeeding could instead fail to improve or even degrade both sorts of outcomes.
How Cochrane ignored the warnings
Lenells responded “Thank you for your e-mail and interest in our work… we have finished the review and will refrain from discussion until it is published” (email, Mar. 21, 2023).
Frances Kellie at Cochrane Support responded “The draft full review is currently being finalised by the review team and we expect the draft to be ready within the next three months… we will be sure to contact you once the draft submission is ready for peer review” (May 11, 2023).
I never heard from them again. I had written:
The planned review appears to suffer from confirmation bias, make bad causal inferences that endanger women, and lack recognition of biological factors that may causally contribute to both breastfeeding problems and postpartum depression..
I hope you will reconsider in the interests of protecting new mothers, a particularly vulnerable group to whom healthcare practitioners and researchers owe a heightened duty of care. At a bare minimum, a review of the evidence on breastfeeding interventions and postpartum depression must consider that the causality could work more ways than one. Otherwise, denial of possible iatrogenesis is built into the publication.
And so it was.
The published review makes the mistakes I warned about, and more. The authors’ misinterpretations of the core findings serve their preferred narrative. The lack of transparency and open access prevent fuller scrutiny.
Cochrane’s standards require that they do better. But does anyone bother enforcing them?