WHAT THE STUDY ACTUALLY SAYS

Doxycycline in early pregnancy was linked to more pregnancy loss in a large study

A cohort of 225,000 pregnancies tied first-trimester doxycycline to a higher risk of loss than other antibiotics. It is observational, so it shows association, not proof of cause.

Women prescribed doxycycline early in pregnancy had a higher risk of pregnancy loss than those given other common antibiotics, in a large cohort study published in Annals of Internal Medicine [s1]. The signal was consistent across two separate insurance databases and across different comparison drugs — but the study is observational, so it establishes an association that clinicians can weigh, not proof that the antibiotic causes loss.

Why the question is hard

Doxycycline is widely prescribed to women of reproductive age, for respiratory infections, acne and sexually transmitted infections, yet evidence on its safety in pregnancy has been thin [s1]. That gap is awkward: the drug is common, early pregnancy is often not yet recognised when a prescription is written, and pregnancy loss is itself common, which makes any single case uninformative. The only practical way to look is across very large numbers of pregnancies — and to compare doxycycline against other antibiotics used for the same conditions, rather than against no treatment, so that the illness being treated is not mistaken for an effect of the drug.

What the study did

The researchers ran a retrospective cohort study using two US claims databases: Merative MarketScan commercial records from 2005 to 2021 and Medicaid records from 1999 to 2019 [s1]. They studied pregnancies ending in either loss or live birth among women aged 12 to 55 [s1].

The design hinges on its comparisons. Rather than contrast doxycycline with nothing, the team built three cohorts of pregnancies that had a first-trimester prescription fill plus a matching diagnosis, and compared doxycycline against the drug a clinician might otherwise have chosen for the same problem: amoxicillin or a macrolide for a respiratory tract infection, and topical clindamycin for acne [s1]. Doxycycline-exposed pregnancies were matched to comparator pregnancies on a propensity score and on gestational age, and the risk of loss was estimated separately in each database and each comparison [s1].

The analysis was large: 218,234 pregnant participants with a respiratory tract infection (110,254 in MarketScan and 107,980 in Medicaid) and 6,636 with acne (4,719 in MarketScan and 1,917 in Medicaid) [s1].

What it found

For respiratory infections, doxycycline was tied to a higher risk of pregnancy loss than amoxicillin in both databases — an adjusted risk ratio of 1.54 (95% CI 1.41 to 1.68) in MarketScan and 1.40 (1.28 to 1.53) in Medicaid [s1]. The comparison against macrolides pointed the same way: 1.39 (1.28 to 1.51) in MarketScan and 1.36 (1.25 to 1.48) in Medicaid [s1].

For acne, the comparison with topical clindamycin was statistically significant in MarketScan (adjusted risk ratio 1.49, 1.26 to 1.76) but not in Medicaid (1.03, 0.80 to 1.33), where the confidence interval crossed 1 [s1]. Across the cohorts where the risk ratio was statistically significant, the absolute difference in risk ranged between 7 and 11 percentage points [s1] — a sizeable gap, not a trivial one.

How to read it

The strengths are the scale, the replication across two unrelated databases, and the active-comparator design, which blunts the most obvious objection — that sicker people get a particular drug. The authors are nonetheless careful about what an observational study can show. They note that residual confounding (differences between the groups they could not fully adjust for) and misclassification of either the drug exposure or the pregnancy outcome remain possible [s1]. And they did not assess congenital malformations, so this is a study about pregnancy loss specifically, not about birth defects [s1].

The acne result is a useful illustration of how to read the whole paper: a clear signal in one database and a null in the other, which is exactly the kind of inconsistency that should temper confidence rather than be waved away. The respiratory-infection finding, by contrast, held up across both databases and both comparator drugs, which is what makes it harder to dismiss.

What to watch

The practical question is whether guidelines and prescribers treat this as a reason to prefer an alternative antibiotic in early or possible pregnancy, as the authors suggest the data could inform [s1]. Confirmation would ideally come from datasets outside the US and from analyses able to account for the infection's own severity and for the reasons doxycycline is chosen. For now, the honest summary is that a large, carefully designed observational study found a consistent association between first-trimester doxycycline and pregnancy loss — a finding to weigh in prescribing, not a proven cause [s1].

This article describes observational research and is not medical advice. Anyone who is or may be pregnant should discuss antibiotic choices with their own clinician and should not stop a prescribed treatment on their own.

Sources

  1. Doxycycline and Risk for Pregnancy Loss: A Multi-database, Multi-indication Cohort Study — Annals of Internal Medicine , November 10, 2026

More on

Related coverage