Ethnic differences in population-level intracerebral haemorrhage incidence in Aotearoa New Zealand: findings from three Auckland Regional Community Stroke cohorts

Primary intracerebral haemorrhage (ICH) accounts for one-sixth of all strokes and is associated with higher mortality and long-term disability than ischaemic stroke.¹ Marked ethnic inequities in stroke incidence and outcomes persist in Aotearoa New Zealand, with Māori and Pacific peoples experiencing a higher burden than Europeans.² However, contemporary population-level data describing ethnic-specific trends in ICH incidence over time remain limited.

Understanding temporal trends in ICH is particularly important in the context of evolving cardiovascular risk profiles, increasing use of anticoagulation for atrial fibrillation (AF) and ongoing inequities in hypertension prevalence and control.³ Whether these changes have translated into differential ICH trends across ethnic groups is unknown.

The Auckland Regional Community Stroke (ARCOS) study collects data on all stroke and transient ischemic attack (TIA) events over a 12-month period once every decade. It contains data on patient demographics, comorbidities and outcomes.²

Using aggregate data from three successive ARCOS studies, we examined ethnic-specific population incidence trends of ICH.

Methods

We analysed population-based ICH incidence data derived from three successive ARCOS studies (ARCOS III, IV and V), each corresponding to a 1-year observation period between 2002 and 2021. All cases of first-ever ICH were ascertained using standard ARCOS methodology.

Cases were classified using standard prioritised self-identified ethnicity methods into four groups: Māori, Pacific peoples, European and Asian/Others. The Asian/Other category included individuals identifying with Asian ethnic groups (e.g., East Asian, South Asian, Southeast Asian) and smaller residual groups (including Middle Eastern, Latin American, African and other ethnicities). Annual population denominators for each ethnicity were obtained from Auckland population estimates as previously reported.²

Descriptive statistics were used to assess baseline socio-demographic and comorbidities of patients with stroke by ethnic group across the 20-year study period. Statistical significance of changes in the distribution of categorical variables was tested using the Cochran–Armitage test and for continuous variables using the Kruskal–Wallis test.

Ethnicity-specific ICH incidence rates were expressed as events per 100,000 population with exact 95% Poisson confidence intervals (CI). Ethnic differences in population-level ICH incidence were then estimated using Poisson regression models fitted to aggregated ARCOS III–V data, with adjustment for ARCOS study period and Europeans as the reference group. Time-by-ethnicity interaction was assessed using interaction terms to examine whether ethnic incidence rate ratios varied across ARCOS periods. Since only summary data were available, adjustment for individual-level comorbidities or treatment exposures was not possible. A two-sided p-value of <0.05 was considered statistically significant.

This study has been approved by the Health and Disability Ethics Committee (ref: 2023 AM 9094). All statistical analyses were performed using Stata BE version 17 (StataCorp, College Station, Texas, United States of America).

Results

We identified 717 first ICH events across the three cohorts. The mean age of ICH cases remained stable (∼69 years), while diabetes prevalence increased significantly (10.2% to 22.8%, p=0.0003). The proportion of females declined modestly across cohorts. There were no significant differences in history of hypertension, hyperlipidaemia, myocardial infarction or AF prevalence across cohorts. Twenty-eight-day and 1-year mortality declined significantly over time (Table 1).

View Table 1–3.

Across the three ARCOS study periods, Pacific peoples consistently exhibited higher population-level ICH incidence than other ethnic groups (Table 2).

In period-adjusted Poisson regression models pooling ARCOS III–V data, Pacific peoples had approximately 60% higher ICH incidence than Europeans (incidence rate ratio [IRR] 1.60 [95% CI 1.31–1.94], p<0.001), Table 3. Māori and Asian/Other populations had significantly lower ICH incidence compared with Europeans. There was no statistically significant evidence of time-by-ethnicity interaction (p=0.069), suggesting that relative ethnic differences in ICH incidence did not vary significantly across study periods.

Conclusion

In this population-based analysis spanning three ARCOS cohorts over a 20-year period, we observed marked ethnic differences in ICH incidence; Pacific peoples experienced substantially higher population-level rates than other ethnic groups. This finding is in keeping with an Australian-based study that found a significantly increased risk of ICH in Polynesians compared with Europeans.⁴ This excess burden persisted across study periods and remained evident after adjustment for cohort year, supporting the presence of a stable and substantial inequity in ICH risk among Pacific communities in Auckland.

In contrast, we observed lower ICH incidence among Māori compared with Europeans. This finding differs from earlier population-based ARCOS analyses that reported a substantially higher age-adjusted risk of ICH among Māori and Pacific peoples combined.⁵ A hospital-based series from the Waikato Region (1998–2008) similarly reported lower crude ICH incidence per 100,000 in Māori (9.6), Pacific (9.2) and Asian (9.4) populations than in NZ Europeans (16.4).⁶ The authors attributed this to differential healthcare access and case ascertainment rather than biological protection. The lower crude ICH incidence reported in Pacific and Asian populations in Waikato compared with population-based data from Auckland is consistent with this interpretation. In ARCOS, Māori also had lower recorded incidence of ischaemic stroke and did not have a higher proportion of undetermined stroke compared with NZ Europeans, arguing against stroke subtype misclassification as the sole explanation for the lower ICH incidence. Given the small numbers, changes in stroke classification over time and the absence of age standardisation in our data, the Māori-specific findings in the present study should be interpreted cautiously.

Interpretation of temporal changes in ICH incidence across ARCOS waves is limited by major reductions in undetermined stroke classification (11.6% in ARCOS III vs <2% in ARCOS IV and V), indicating that improved diagnosis is likely to have contributed to the higher observed ICH incidence in later cohorts. Further, case ascertainment improved across ARCOS waves, with the proportion of missing cases declining over time.² These methodological changes limit causal interpretation of apparent secular trends. In addition, age-standardised incidence rates were not available for this analysis, and therefore the contribution of demographic ageing to observed temporal differences cannot be determined.

The clinical profile of ICH cases evolved across cohorts, with a significant rise in diabetes prevalence but stable mean age and no significant change in recorded history of hypertension or AF. Despite this, both 28-day and 1-year mortality declined over time, consistent with improvements in acute stroke systems of care in Aotearoa New Zealand, including greater access to stroke unit care and advances in the supportive management of ICH.^7–9 However, as only aggregate summary data were available, we were unable to examine ethnicity-specific risk factor profiles or treatment exposures, precluding causal inference.

Recent randomised trials have demonstrated the safety of earlier anticoagulation initiation after AF-associated ischaemic stroke, but these studies included relatively low proportions of non-European participants.^10,11 In prior work using ARCOS IV and V data, we reported an increase in AF-associated ICH coinciding with a period of substantially increased anticoagulant prescribing.¹² While the present findings do not permit any inference regarding anticoagulation-related bleeding risk, this highlights that background haemorrhage risk is not ethnically uniform at the population level, a factor that may be relevant when translating predominantly European-derived trial evidence into Aotearoa New Zealand clinical practice.