P–ICH is not a typical young-adult stroke. Haemorrhagic stroke, which includes intracerebral haemorrhage and subarachnoid haemorrhage, accounts for nearly half of all strokes during pregnancy and the postpartum period, a proportion far higher than in the general stroke population. Reported incidence figures for P–ICH have ranged from 2.8 to 9.9 per 100,000 deliveries, but the aetiological profile and clinical outcomes appear to differ from those of brain bleeds in non-pregnant young adults, and the mortality and functional disability attached to the condition are substantial. What has remained unclear is exactly when, across the pregnancy-postpartum timeline, the risk truly peaks and how long it persists.

The methodological machinery behind the new study is what sets it apart. The researchers linked four population-based databases in Taiwan, including the National Health Insurance Research Database covering 99.9 percent of the entire population, the Birth Certificate Application database, the Maternal and Child Health Database, and the Taiwan Death Registry. Crucially, they did not rely solely on delivery records or birth registries, which systematically miss pregnancies ending before 20 weeks, non-hospitalised pregnancies, and late postpartum readmissions. Instead, a hierarchical algorithm cross-linked antenatal visit records with birth certificate data to capture live births, stillbirths, spontaneous and induced abortions, ectopic pregnancies, and gestational trophoblastic disease, with week-specific gestational dating drawn from prenatal reimbursement records. Of the 2,112,294 pregnancies captured from 2012 through 2020, 587,877, or 27.8 percent, ended earlier in gestation, a population largely invisible to previous delivery-based studies.

The analytical design was equally distinctive. Each woman served as her own control in a cohort-crossover framework, meaning time-invariant characteristics such as genetic susceptibility or undiagnosed vascular malformations were matched exactly rather than statistically adjusted for. For each pregnancy, the researchers compared six prespecified intervals, spanning the first, second, and third trimesters and 0–42, 43–180, and 181–365 days postpartum, against the corresponding period exactly 730 days prior. Because the maximum pregnancy-postpartum window of 659 days is inherently shorter than 730 days, this anchor guarantees a washout of at least 71 days between the control window and the index pregnancy. Among the full cohort, 406 pregnancies were complicated by P–ICH, a crude incidence of 19.22 per 100,000 pregnancies, exceeding the upper range of previous delivery-based estimates.

The temporal pattern the analysis revealed is the study’s headline. Of the 406 events, 337, or 83.0 percent, occurred during the first postpartum year, against only 69, or 17.0 percent, during pregnancy itself. Across all three trimesters, the risk showed no meaningful difference from matched control intervals, with incidence rate ratios of 0.70, 0.96, and 1.69 respectively, though the numbers trended upward. Then came the surge: in the first 42 days postpartum, the incidence rate ratio hit 7.44, with a 95 percent confidence interval of 4.55 to 12.18, an elevation that dwarfs anything seen during gestation. The overall incidence rate across the pregnancy-postpartum period was 12.28 per 100,000 person-years versus 6.86 in the control period, yielding an overall incidence rate ratio of 1.79.

What makes the study genuinely new is what happened after the surge. The risk attenuated but stayed above the null through 43–180 days postpartum, with an incidence rate ratio of 1.58, and through 181–365 days postpartum, with an incidence rate ratio of 1.38. Because the 730-day gap between the pregnancy-postpartum and control periods could itself confound the estimates through background ageing, the researchers ran two complementary sensitivity analyses. The cohort-time-control analysis, which matched each pregnant woman to a non-pregnant woman of comparable age and calendar time, and the cohort-cohort-time-control analysis, which instead captured the ageing trend specific to women of ongoing reproductive activity, both sustained the elevation through 43–180 days postpartum. During 181–365 days postpartum, the cohort-time-control estimate weakened to 1.36 with its lower confidence limit just below the null, while the cohort-cohort-time-control estimate remained elevated at 1.53, largely supporting the late postpartum risk elevations.

The subgroup findings carry the most direct clinical relevance. Hypertensive disorders of pregnancy as a whole did not reach formal significance as an effect modifier, but the subtype analysis did. Women with hypertensive disorders of pregnancy other than chronic hypertension carried an incidence rate ratio of 3.78, significantly higher than the 1.62 seen in unaffected pregnancies, with a highly significant interaction p value of 0.0024. Mild pre-eclampsia drove the ratio to 4.13 and severe pre-eclampsia to 5.80, both with significant interactions. Eclampsia could not be formally tested because no events occurred in the corresponding control period, but the directional pattern suggested it may confer the greatest risk within the entire hypertensive spectrum. In striking contrast, chronic hypertension alone did not modify the risk, plausibly because these women already carried an elevated baseline risk in the control period. Advanced maternal age also climbed with risk, from 1.03 among women 24 years or younger to 2.16 among those 35 years or older, though the age interaction fell short of significance.

The mortality data attached to these events are sobering. Hospital mortality among P–ICH cases was 9.1 percent, and two-year mortality reached 12.1 percent versus 0.1 percent in unaffected pregnancies, a gap that persisted after propensity score matching balanced the groups on maternal age, gravidity, pregnancy outcome, comorbidities, medication exposure, income, and urbanisation level. Among the 49 maternal deaths recorded within two years, 43, or 87.8 percent, occurred in women whose P–ICH developed during the postpartum period, with 19, or 38.8 percent, of all two-year deaths attached to P–ICH events that happened during the late postpartum window, between 43 and 365 days after pregnancy ended. Obstetric complications were largely comparable between P–ICH and unaffected pregnancies except for postpartum haemorrhage, suggesting the elevated mortality is associated primarily with the brain bleed itself rather than with a cascade of delivery-related complications.

The offspring burden is equally consequential. Among 311 P–ICH pregnancies resulting in delivery of 320 neonates, caesarean section was performed in 57.8 percent of cases versus 36.4 percent of reference pregnancies. Stillbirth was more frequent at 4.7 percent versus 1.1 percent, mean gestational age at delivery was lower at 36.3 versus 38.0 weeks, and low birth weight under 2500 grams affected 29.7 percent of neonates versus 9.8 percent of controls. One-minute and five-minute Apgar scores were both depressed, and neonatal morbidities were substantially elevated, particularly respiratory distress syndrome at 7.5 percent versus 0.5 percent, neonatal jaundice at 14.7 percent versus 5.7 percent, and intrauterine growth restriction at 2.8 percent versus 0.2 percent. Congenital abnormalities were comparable between groups, pointing to P–ICH and its associated prematurity, rather than underlying congenital anomalies, as the driver of adverse neonatal outcomes. Offspring mortality at two years was also higher among those born to mothers with P–ICH, a gap that persisted after propensity score matching.

The authors are careful about what their data can and cannot prove. Because the Taiwanese cohort is sociodemographically homogeneous, the findings may not directly generalise to populations with different ethnic backgrounds or multi-tiered insurance systems, and cross-population incidence comparisons should be interpreted with the understanding that race and ethnicity are sociocultural constructs rather than biological traits. Administrative claims databases cannot capture neuroimaging findings, standardised stroke severity indices such as the National Institutes of Health Stroke Scale, or functional disability metrics such as the modified Rankin Scale, precluding direct evaluation of post-discharge neurological deficits among survivors. Residual confounding by unmeasured maternal-level factors also cannot be excluded, and the associations with adverse maternal-offspring outcomes should not be read as proven causal effects. Still, the core message is hard to escape: the conventional 42-day puerperium, and even the emerging fourth-trimester surveillance window, may be too short for a complication whose cerebrovascular risk persists for at least six months after pregnancy ends. Women with hypertensive disorders of pregnancy other than chronic hypertension, particularly pre-eclampsia and eclampsia, appear to represent a high-risk group who might benefit from closer postpartum clinical vigilance, blood pressure control, and cerebrovascular risk assessment beyond the routine six-week postpartum visit. Future studies should validate these findings in other populations, incorporate neuroimaging and stroke-severity data, and determine whether lengthened postpartum surveillance can reduce P–ICH-related morbidity and mortality.

Subject of Research: Postpartum risk and maternal-offspring outcomes of pregnancy-associated intracerebral haemorrhage in a nationwide Taiwanese cohort-crossover study

Article Title: Postpartum risk and prognosis of pregnancy-associated intracerebral haemorrhage: a nationwide cohort-crossover study

Article References: Chen, Y.-N., Hsu, H.-H., Chun-Yuan Cheng, M., Meng-Hsun Li, B., Shyr, Y., Chia-Cheng Lai, E., & Lai, C.-H. (2026). Postpartum risk and prognosis of pregnancy-associated intracerebral haemorrhage: a nationwide cohort-crossover study. The Lancet Regional Health – Western Pacific, 75, Article 102005. https://doi.org/10.1016/j.lanwpc.2026.102005

Image Credits: AI Generated

DOI: 10.1016/j.lanwpc.2026.102005

Keywords: intracerebral haemorrhage, postpartum stroke, pregnancy, hypertensive disorders of pregnancy, pre-eclampsia, cohort-crossover design, maternal mortality, neonatal outcomes, Taiwan, cerebrovascular risk, postpartum surveillance, Postpartum