A mentoring scheme for pregnant teenagers halved deaths in a Sierra Leone pilot

Most trials that reduce maternal deaths test a drug, a device or a clinical protocol. A pilot trial published in The Lancet on 18 June tested a person: a local woman assigned to mentor a pregnant adolescent from pregnancy through the first year after birth.

The trial

2YoungLives was a parallel-arm, pilot, hybrid implementation-effectiveness cluster randomised controlled trial, run as an adjunct to maternity care in rural and urban communities served by 12 peripheral health units across five districts of Sierra Leone [s1]. Clusters were randomly allocated 1:1 [s1].

Eligible participants were all pregnant girls under 18 living in those communities and presenting for maternity care [s1]. The primary outcome was a composite of maternal and perinatal deaths, including stillbirths and neonatal deaths, assessed among all randomly allocated participants remaining in follow-up [s1]. The trial was prospectively registered as ISRCTN32414369 [s1].

Between 4 July 2022 and 30 November 2023, 673 girls were enrolled: six clusters with 372 girls to the intervention and six clusters with 301 to control, with 361 and 279 respectively included in the primary analysis [s1]. Loss to follow-up was under 10% in both groups [s1].

The result

The composite outcome occurred in 23 of 361 girls (6%) in the intervention group and 35 of 279 (13%) in the control group — an adjusted risk ratio of 0.52 (95% CI 0.34 to 0.81, p=0.0034) [s1]. The number needed to treat to prevent one maternal or perinatal death was 18 (95% CI 10 to 92) [s1].

The authors' conclusion is that the intervention was feasible to implement in both urban and rural communities and significantly reduced the composite outcome [s1].

Why "pilot" is the most important word in the title

A halving of mortality with a number needed to treat of 18 would be a remarkable result in a definitive trial. This is not a definitive trial, and the paper says so in its own framing: the stated purpose was to evaluate feasibility and potential effects in order to inform a subsequent larger trial [s1].

Three features limit how far the estimate can be pushed. The unit of randomisation was the cluster, and there were only 12 of them — six per arm. With that few clusters, baseline differences between communities cannot be balanced by randomisation with any reliability, and the adjusted analysis is doing more work than in a large trial.

Second, the confidence interval on the number needed to treat runs from 10 to 92 [s1]. That range spans the difference between an extraordinarily efficient intervention and a marginal one.

Third, 58 events in total drive the entire result [s1]. Composite outcomes built from stillbirths, neonatal deaths and maternal deaths are appropriate when each component is rare, but they also mean the headline effect may be concentrated in one component rather than distributed across all three.

None of this makes the finding uninteresting. It makes it what the authors call it: grounds for a larger trial. The trial was funded by the UK National Institute for Health and Care Research [s1].

The burden this sits inside

Two other papers published within days give the scale.

A Global Burden of Disease analysis in Obstetrics & Gynecology, published on 18 June, examined five leading causes of maternal mortality — haemorrhage, hypertensive disorders, sepsis and infections, obstructed labour and uterine rupture, and abortion and miscarriage — among females aged 15-49 from 1990 to 2021 [s2]. Globally in 2021 it estimates 102,854,299 new incident cases of those five disorders and 129,331 deaths [s2].

The trend is downward. Between 1990 and 2021, the age-standardised incidence rate declined by 1.4% annually (95% CI -1.5% to -1.3%) and the age-standardised mortality rate by 3.6% annually [s2]. The reported confidence interval for the mortality figure in the published abstract reads "-3.7% to 3.4%", which appears to contain a sign error given the estimate itself; we report the point estimate and note the discrepancy rather than silently correcting it.

Haemorrhage remained the leading cause, accounting for 35.8% of maternal deaths globally, and mortality peaked in the 20-24 year age stratum, declining to its lowest levels at ages 45-49 [s2]. Central sub-Saharan Africa persisted as the region with the highest rates [s2].

The age finding is the one that bears on 2YoungLives. If mortality peaks in the early twenties and the burden is concentrated in sub-Saharan Africa, an intervention targeted at pregnant girls under 18 in Sierra Leone is aimed at the steepest part of the curve.

A third paper, published in Reproductive Health on 23 June, reports that there has been no measurable change in global preterm birth rates over the past decade in any region [s3]. An estimated 13.4 million newborns (95% credible interval 12.3-15.2 million) were born preterm in 2020, 9.9% of births worldwide [s3]. Preterm birth complications remained the leading cause of under-5 mortality globally in 2022, accounting for about 1 million neonatal deaths — similar to a decade earlier [s3]. The authors note that most stillbirths occur preterm, an estimated 74.3%, and argue stillbirths should be included in preterm birth data, policy and programmes [s3].

What to watch

Whether 2YoungLives proceeds to the larger trial its authors describe as the purpose of the pilot, and whether the effect survives at scale with more clusters and more events [s1].

Sources

• [s1] Community-based mentoring to reduce maternal and perinatal mortality in adolescent pregnancies in Sierra Leone (2YoungLives): a pilot cluster-randomised controlled trial. The Lancet, 18 June 2025. https://doi.org/10.1016/S0140-6736(25)00454-4
• [s2] Global Burden of Maternal Disorders and Mortality Among Female Individuals Aged 15-49 Years, 1990-2021. Obstetrics & Gynecology, 18 June 2025. https://doi.org/10.1097/AOG.0000000000005980
• [s3] Born too soon: global epidemiology of preterm birth and drivers for change. Reproductive Health, 23 June 2025. https://doi.org/10.1186/s12978-025-02033-x