Pregnancy is the single exception running through the whole coffee literature. The umbrella review by Poole and colleagues[1], which examined 201 meta-analyses and concluded that coffee is more often associated with benefit than harm, singled out pregnancy as the place where harmful associations survived adjustment. That makes this the one topic in this section where the appropriate posture is caution rather than reassurance, and where the honest answer is that credible evidence genuinely conflicts.
Before anything else: if you are pregnant, this article is a summary of research, not advice. Your obstetrician or midwife knows your pregnancy and this page does not.
It is also worth saying that the caution here is specific to pregnancy and does not extend to breastfeeding in the same way, where guidance is generally more relaxed, or to anyone else.
What the observational research found
The key meta-analysis is by Chen and colleagues, published in BMC Medicine in 2014[2], a systematic review and dose-response analysis of maternal caffeine intake and low birth weight. It found a graded relationship. Compared with a reference category of no or very low caffeine intake, the relative risk of low birth weight was 1.13 for low intake, defined as 50 to 149 mg a day, 1.38 for moderate intake at 150 to 349 mg a day, and 1.60 for high intake at 350 mg a day or more. Expressed continuously, each 100 mg a day increment was associated with 13% higher risk of low birth weight.
Two features of that analysis are worth noting before drawing conclusions from it. It measured caffeine from all sources rather than coffee specifically, so tea, cola and chocolate are folded in. And low birth weight is a composite outcome that can arise either from a baby being born early or from restricted growth in utero, which are different problems with different causes, and pooled estimates do not separate them.
That is a clear, dose-related association, and it is the empirical basis for the caution in official guidance. The umbrella review found the same pattern across outcomes: high versus low or no consumption was associated with low birth weight at an odds ratio of 1.31, with preterm birth in the first and second trimesters at 1.22 and 1.12, and with pregnancy loss at 1.46.
Why that might not mean what it looks like
There is a specific and well-documented confounder here that has nothing to do with the usual concerns about lifestyle and socioeconomic status. It is nausea.
Why this is the one exception in the coffee literature
It is worth registering how unusual pregnancy is within this section. Across 67 health outcomes, the umbrella review by Poole and colleagues[1] found harmful associations largely nullified once smoking was properly adjusted for, with pregnancy the standout exception where they survived. Every other article here reports coffee looking neutral or favourable; this is the one where a credible body of evidence points the other way.
That asymmetry is itself informative. If coffee were broadly harmful and the favourable findings were artefacts of confounding, you would expect harm signals scattered across many outcomes. Instead they concentrate in a single physiological context, one where a specific and well-understood mechanism, slowed caffeine clearance combined with placental transfer to a fetus that cannot metabolise it, gives a reason to expect something different.
That does not settle whether the association is causal. It does mean the caution is targeted rather than generic.
Nausea and vomiting in early pregnancy are associated with a lower risk of miscarriage, and they are also strongly associated with aversion to coffee. Women with healthy, well-established pregnancies producing strong hormonal signals are more likely to feel nauseated and therefore more likely to stop drinking coffee. Women whose pregnancies are not progressing well may experience less nausea and continue drinking coffee normally.
If that is what is happening, then continued coffee drinking is a marker of a pregnancy already at higher risk, rather than a cause of the risk. The direction of causation would be exactly backwards from the way the association reads. This mechanism is well known in perinatal epidemiology and is very difficult to adjust away statistically, because the confounder is a symptom rather than a measurable exposure.
The genetic test
There is a method designed for exactly this problem. Mendelian randomisation uses genetic variants associated with a behaviour as a proxy for the behaviour itself. Because genotype is fixed at conception and cannot be influenced by nausea, pregnancy outcome, or lifestyle, it sidesteps reverse causation and most confounding. If coffee genuinely causes an outcome, variants that make people drink more coffee should be associated with that outcome.
Brito Nunes and colleagues applied this to pregnancy, publishing in the International Journal of Epidemiology[3], examining birthweight, stillbirth, miscarriage, gestational age and preterm birth. Their results suggested that coffee consumption during pregnancy might not itself contribute to adverse outcomes such as stillbirth, sporadic miscarriage and preterm birth, or to lower gestational age or birthweight.
So the design specifically built to test causation found little to support it. That is a substantial piece of evidence and it is why this article is graded as contested rather than settled.
Why guidance stays conservative anyway
Given a genetic study pointing toward no causal effect, it is fair to ask why the advice has not loosened. Several reasons, and they are good ones.
Mendelian randomisation is powerful but not infallible. It relies on assumptions that can be violated, particularly that the genetic variants affect the outcome only through the exposure being studied. Genetic variants associated with coffee drinking may also relate to caffeine metabolism, to other behaviours, or to biology unrelated to coffee. These studies also tend to have limited statistical power for uncommon outcomes like stillbirth.
It is also worth being precise about what the Mendelian randomisation study did and did not find. Finding no evidence of an effect is not the same as finding evidence of no effect. These analyses have wide confidence intervals for uncommon outcomes, and a null result with a wide interval is compatible with a modest true effect that the study lacked the power to detect. Reporting it as coffee having been cleared in pregnancy overstates it in the same way that reporting the observational data as proof of harm overstates that.
More fundamentally, the asymmetry of the decision drives the guidance. The cost of unnecessarily limiting caffeine during pregnancy is a few months of mild inconvenience and possibly some withdrawal headaches. The cost of wrongly reassuring people, if the observational association turns out to be causal, is measured in birth outcomes. Regulators respond rationally to that asymmetry by staying cautious while the evidence remains unresolved.
It is also worth noting that caffeine clearance slows dramatically during pregnancy. Nehlig's review[4] documents that caffeine half-life can extend substantially in later pregnancy, sometimes to two or three times the non-pregnant value, and caffeine crosses the placenta while the foetus has very limited capacity to metabolise it. That is a genuine pharmacological reason for caution independent of the epidemiology.
The actual numbers in the guidance
EFSA concluded that caffeine intakes from all sources up to 200 mg per day, consumed throughout the day, do not raise safety concerns for the foetus[5]. Major obstetric bodies, including the American College of Obstetricians and Gynecologists, use the same 200 mg figure, generally stating that consumption below that level does not appear to be a major contributing factor in miscarriage or preterm birth, while noting that the relationship to fetal growth restriction is less certain.
From all sources is the operative phrase. Two hundred milligrams is roughly two 8-ounce cups of brewed coffee, or about three espresso shots, but tea, cola, energy drinks, chocolate and some medications all count toward it. In the Philippine context, 3-in-1 sachets and milk tea are both common and both contribute.
Practical arithmetic for 200 mg
Two hundred milligrams is easier to exceed than it sounds, so it is worth laying out. Roughly, an 8-ounce brewed coffee is around 95 mg, a single espresso shot around 63 mg, an 8-ounce cup of tea 30 to 50 mg, a can of cola 30 to 40 mg, and energy drinks anywhere from 80 mg upward.
So two ordinary brewed coffees reach the limit on their own. A large cafe drink with two or three shots can approach it in a single cup. And someone having one coffee, one milk tea and a cola across a day may be at or over it without having thought of themselves as a heavy caffeine consumer at all.
The practical approach most obstetric guidance suggests is to pick one caffeinated drink a day and treat everything else as decaf or caffeine-free, which builds in a margin against the considerable uncertainty in any of these estimates.
The practical reading
Observational meta-analyses find a graded association between maternal caffeine and low birth weight, preterm birth and pregnancy loss. A genetic study designed to test causation found little evidence that coffee itself is responsible, and a well-documented confounder, pregnancy nausea driving coffee aversion, could plausibly generate the association without any causal role. Guidance from EFSA and obstetric bodies converges on approximately 200 mg a day from all sources, and the pharmacological case for that caution is independent of the disputed epidemiology.
The recommendation that follows is unglamorous and, on this evidence, correct: stay under 200 mg, count every source, and do not treat a lapse as a crisis.
Whatever else is uncertain here, that advice is robust to which side of the argument turns out to be right, which is a good property for advice to have.
The reasonable position is to follow the 200 mg guidance without treating an occasional cup as a catastrophe, and to take individual advice from the people managing your pregnancy. If you are cutting down and want to know what to expect, our article on caffeine withdrawal covers the timeline. Decaffeinated coffee is a straightforward option, and as the UK Biobank analysis[6] found, decaf shows much the same favourable associations as caffeinated coffee across most health outcomes.
The evidence will likely be revisited as larger genetic studies accumulate.
What to do with genuinely conflicting evidence
This article is rated contested rather than strong or moderate, and it is worth explaining what that should mean for a reader, because conflicting evidence is more common in nutrition than clean answers are.
Holding two conflicting findings in mind at once is uncomfortable, and it is also the accurate description of this evidence.
It does not mean the research is worthless or that you should pick whichever result you prefer. It means two well-conducted lines of investigation, using different methods with different weaknesses, reached different conclusions, and nobody has yet resolved which weakness is doing more damage. The observational meta-analyses are vulnerable to confounding by pregnancy nausea. The Mendelian randomisation study is vulnerable to weak instruments, limited power and assumptions that are hard to verify.
In that situation the rational response is to act on the more cautious reading while the question remains open, particularly where the cost of caution is low and the cost of being wrong is high. That is precisely the calculation EFSA and the obstetric bodies made, and it is why the guidance has not moved despite a genetic study pointing the other way.
It is also worth saying plainly, because pregnancy guidance has a way of inducing guilt: nothing in this literature suggests that a cup of coffee causes harm. The observational associations concern sustained intake at higher levels, and the causal evidence for even those is contested. Someone who drank coffee before knowing they were pregnant has not done something dangerous.
One further note on the practical side. Caffeine clearance slows progressively through pregnancy, so the same cup represents a larger exposure in the third trimester than in the first. Anyone who found their tolerance unchanged early on may notice it changing later, and that is physiology rather than imagination.
Nausea in the first trimester often solves this problem on its own, since coffee is among the first things many people go off entirely.