How to Read the Evidence

Somewhere between your pediatrician, your mother-in-law, and that one Instagram account, you're going to hear the phrase "studies show." This lesson is your toolkit for deciding what to do with it.

You don't need a statistics degree. You need about ten minutes and a healthy sense of skepticism — which, conveniently, sleep deprivation tends to sharpen.

What "the research says" actually means

There is no single filing cabinet labeled "The Research." What exists is thousands of individual studies of wildly different sizes, designs, and quality, published over decades, sometimes contradicting each other. When someone says "the research says X," they usually mean one of three things:

• One study found X. Interesting, but one study is a data point, not a verdict.
• Most good studies point toward X. This is the real thing — a consistent pattern across well-designed research.
• Someone would like X to be true. Also common.

Your job as a reader is just to figure out which of the three you're looking at. The rest of this lesson gives you the tools.

The evidence ladder, one more rung down

We introduced the idea that not all evidence is equal in why this course is evidence-based. Here's the ladder in a bit more detail, from weakest to strongest:

None of this makes anecdotes worthless. Anecdotes generate the questions that studies later answer. They're a fine place for research to start; they're a shaky place for your decisions to end.

Why sample size and effect size both matter

Two numbers do most of the heavy lifting in any study, and headlines routinely ignore both.

Sample size is how many people were studied. Small studies are noisy: with 15 mothers, a couple of unusual babies can swing the whole result. Large studies smooth out the noise. When the Cochrane Collaboration reviewed breastfeeding support, it pooled trials covering tens of thousands of mother-baby pairs (McFadden et al., 2017) — a conclusion built on that much data is far sturdier than one built on a single small trial.

Effect size is how big the difference actually was. A study can be enormous, impeccably run, and find a real effect that is also tiny — statistically solid, practically meaningless for your day-to-day. "Statistically significant" only means the effect probably isn't zero. It says nothing about whether it's big enough to change what you'd do.

So when you meet a claim, ask two questions: how many people, and how much difference. A finding needs decent answers to both before it deserves to influence a decision.

Correlation vs. causation: a breastfeeding case study

Observational studies consistently find that breastfed babies do better on all sorts of long-term outcomes. The tricky part: in many countries, families who breastfeed longer also tend to have more income, more education, and more access to healthcare — advantages that improve child outcomes all by themselves. Researchers call this confounding, and it's the central headache of breastfeeding science.

Two clever study designs show how big the headache really is:

• The sibling comparison. Researchers compared siblings within the same family where one was breastfed and one wasn't — automatically holding family income, education, and environment constant. Most of the long-term advantages that looked large in ordinary comparisons shrank dramatically and lost statistical significance (Colen & Ramey, 2014).
• The randomized trial. The PROBIT trial in Belarus randomized over 17,000 mother-infant pairs — not to breastfeed or not, but to receive strong breastfeeding support or usual care — creating comparable groups. It found real effects of increased breastfeeding on some outcomes, like fewer gut infections in infancy, but not the sweeping across-the-board benefits observational studies had implied (Kramer et al., 2001).

The honest summary from the best evidence reviews: breastfeeding has well-supported benefits, especially protection against infections in infancy (Victora et al., 2016) — and some widely repeated claims rest on much shakier ground. Both halves of that sentence matter. Anyone who only tells you one half is selling something.

And notice what this does not mean: it does not mean feeding choices are trivial. It means the gap between feeding methods is far smaller than the guilt industry implies — which is genuinely good news whether you're nursing, pumping, combo-feeding, or using formula.

Red flags worth scanning for

A quick field guide to phrases that should make you slow down:

• "Studies show..." — with no study named. Which studies? On how many people? You're allowed to ask.
• "Doubles the risk!" — relative risk without the baseline. Doubling a very rare risk is still a very rare risk; headlines almost never tell you the starting number. Ask: from what, to what?
• "Scientists have proven..." — science almost never "proves" in one study. Real researchers hedge; marketers don't.
• "They don't want you to know..." — a claim that positions itself against all of mainstream medicine is asking you to trade thousands of researchers for one person with a course to sell. (Yes, we see the irony of saying this inside a course. Check our references. That's the point.)
• One dramatic anecdote as the whole argument — moving stories are how humans communicate, but "it happened to someone" is the bottom rung of the ladder, not the top.

Where to look things up yourself

You don't have to take anyone's word for anything — including ours. Three free places to check claims:

• PubMed — the searchable index of medical research. Abstracts are free; try searching a topic plus "systematic review" to find the summaries first.
• The Cochrane Library — home of the gold-standard systematic reviews, each with a plain-language summary written for actual humans.
• LactMed — a free national database of drugs and lactation. If anyone tells you a medication and breastfeeding can't mix, look it up here before assuming — and see medications and milk for how to use it. Medication decisions belong with you and your clinician; LactMed just makes that conversation better informed.

Reading abstracts at 3 a.m. is a legitimate hobby. We support it.

One boundary worth naming: research literacy is for evaluating claims, not for diagnosing problems. If something feels wrong with you or your baby — pain, fever, a baby who isn't feeding or producing wet diapers — that's a call to your pediatrician, an IBCLC, or urgent care, not a PubMed search. When to get help covers the specifics.

The bottom line

• "The research" is thousands of studies of varying quality — the question is always which studies, how many people, and how big an effect.
• The ladder runs anecdote → observational study → randomized trial → systematic review. Higher rungs deserve more of your trust.
• Correlation isn't causation: families who breastfeed differ in many other ways, and studies that control for this find smaller (but still real) benefits than the headlines suggest.
• Red flags: unnamed "studies," relative risk with no baseline, "proven," and conspiracy framing.
• PubMed, Cochrane, and LactMed are free — you can check anyone's claims yourself, including ours.

This lesson is educational content, not medical advice. For concerns about your health or your baby's, talk to your pediatrician, an IBCLC, or your healthcare provider.

References

• McFadden et al., 2017 · Cochrane Database of Systematic Reviews · Support for Healthy Breastfeeding Mothers with Healthy Term Babies
• Colen & Ramey, 2014 · Social Science & Medicine · Is Breast Truly Best? Estimating the Effects of Breastfeeding on Long-Term Child Health and Wellbeing in the United States Using Sibling Comparisons
• Kramer et al., 2001 · JAMA · Promotion of Breastfeeding Intervention Trial (PROBIT): A Randomized Trial in the Republic of Belarus
• Victora et al., 2016 · The Lancet · Breastfeeding in the 21st Century: Epidemiology, Mechanisms, and Lifelong Effect