Pregnancy can make ordinary uncertainty feel unusually consequential: a delayed reply from a midwife, a bodily sensation, a question about birth. A new randomized trial suggests that repeatedly practicing less threatening interpretations of ambiguous situations can modestly reduce anxiety before and after birth. The result is encouraging precisely because the intervention was brief and self-directed. It is also narrower than the promise of a digital cure.
Published November 4 in JAMA Network Open, the RELAX trial enrolled 268 pregnant adults in the United Kingdom who reported high levels of repetitive negative thinking—worry about the future or rumination about the past—but no more than moderate anxiety. Half received usual maternity care. The other half received usual care plus 12 online sessions over four weeks.
Eight weeks after randomization, the digital group scored 1.46 points lower on the seven-item GAD-7 anxiety scale than the usual-care group. The confidence interval ranged from 0.57 to 2.36 points lower, and the standardized effect was modest, at 0.42. That is evidence of a real average benefit in this study population. It is not evidence that the program treated an anxiety disorder, because participants with high current anxiety or several psychiatric conditions were excluded.
The intervention trained interpretation, not generic relaxation
RELAX is a form of cognitive bias modification for interpretation. In each roughly 17-minute session, participants listened to 30 ambiguous scenarios involving pregnancy or everyday life. The exercise repeatedly resolved uncertainty in a positive or neutral direction rather than a threatening one. The aim was to weaken the mental habit that turns ambiguity into danger and then feeds further worry.
This design matters. The program was not a meditation library, an open-ended chatbot, or a compressed course of psychotherapy. It targeted one proposed cognitive process. The publicly registered trial protocol specified the sample size, primary outcome and follow-up schedule before the results were known.
Engagement was unusually high for a self-guided program. Of 134 people assigned to RELAX, 109 completed at least 10 sessions. Median session length was 17.1 minutes, putting a full course at roughly three and a half hours. Missing outcome data remained below 7% at every follow-up point.
The benefits were broader at eight weeks than later. Alongside lower general anxiety, the RELAX group reported less repetitive negative thinking, worry, depression and perinatal anxiety. By 36 weeks—between 12 and 24 weeks after birth—the groups still differed on general anxiety, repetitive negative thinking and a perinatal depression measure. But differences in general depression, worry, perinatal anxiety and functioning were no longer statistically clear.
The most interesting result is the one that did not fit the theory
The researchers expected the training to work by making interpretations more positive. It did change interpretation scores in the intended direction. Yet that change did not statistically explain the improvement in anxiety at the primary eight-week endpoint. At the later follow-up, the mediation pattern even ran opposite to the hypothesis.
That does not erase the randomized outcome: assignment to RELAX caused a modest average reduction in self-reported anxiety within the trial. It does weaken a more specific claim about why. The exercise might have affected repetitive thinking through another route, or the mediation analysis may have lacked power. It is also possible that completing a structured task during an uncertain period supplied benefits the theory did not capture.
This distinction is valuable beyond one program. Digital mental-health tools are often marketed through plausible mechanisms—retraining attention, correcting bias, regulating emotion—before those mechanisms are firmly demonstrated. A useful intervention can outperform a control even when its proposed psychological engine remains uncertain.
A promising prevention tool still needs a wider test
The comparison was usual care, not an equally engaging digital placebo. Participants knew whether they had received the program, and every clinical outcome was self-reported rather than established through diagnostic interviews. The researchers tested several secondary outcomes without adjusting for multiple comparisons, so isolated positive findings deserve caution.
The sample also limits generalization. Eighty-seven percent had at least a bachelor’s degree, almost everyone reported someone they could talk to, and people with high anxiety, current common mental-health conditions, recent self-harm risk, multiple miscarriages or stillbirth were excluded. The result therefore speaks most directly to relatively educated pregnant adults at elevated cognitive risk—not to everyone experiencing perinatal distress.
That boundary fits current care principles. NICE guidance already supports facilitated CBT-based self-help for persistent subthreshold anxiety in pregnancy while recommending more intensive psychological care when clinically indicated. RELAX could eventually fill part of that low-intensity prevention space, but this phase 2b trial does not establish routine effectiveness, cost-effectiveness or suitability without screening.
The clearest conclusion is neither “an app fixes pregnancy anxiety” nor “a small score difference means little.” A short, highly structured digital exercise produced a modest and partly sustained reduction in anxiety among carefully selected participants. The next question is whether that benefit survives in a more diverse, less intensively studied population—and whether researchers can identify what actually made it work.
