A multi-country epidemiological analysis of mortality from landmines and other explosive ordnance

Abstract

Background

Explosive ordnance including landmines, improvised explosive devices and explosive remnants of war are known to contaminate at least fifty-eight countries, threatening the health of affected communities across generations. Despite the scale of this problem, mortality estimates pertaining to these weapons are limited in scope. This study addresses this knowledge gap by conducting a multi-country epidemiological study on explosive weapons casualties in low-resource settings.

Methods

A retrospective analysis of secondary data was conducted with data sourced from mine action authorities and centres, and international non-government organisations from across 17 countries. Casualty information contained included demographics, mortality, injury data, and incident details. Univariate and multivariate logistic regression models were used to assess which characteristics were associated with mortality and survival.

Results

Here we show one of the largest epidemiological analyses, to date, of explosive weapons casualties. The overall case fatality rate was 38.8% and females have an elevated risk of fatal outcomes. Overall, children experience lower odds of death compared to adults; and, improvised explosive devices have the highest lethality of the weapons examined.

Conclusions

This research highlights the lethality of explosive weapons among local casualties, with a case fatality rate significantly higher than that observed in high-resource civilian or military trauma systems, illustrating the deadliness and indiscriminate nature of these devices. The findings of this analysis can inform injury prevention and trauma care initiatives targeting preventable death and disability for casualties of explosive ordnance.

Plain language summary

Explosive weapons such as landmines, improvised explosive devices and explosive remnants of war are known to pollute at least 58 countries. This study considered how these weapons impact health using casualty data to estimate mortality and understand the factors impacting death, such as age, gender and explosive type. Our results found that 38.8% of those injured die from their injuries, with improvised explosive devices the deadliest weapon type. Females experienced a higher risk of death compared to males, while children were found to have lower deaths compared to adults. This research shows the deadly nature of these weapons and impact on health. These findings can support plans to both prevent and treat injuries to reduce deaths during and after war.

Pizzino et al. conduct a multi-country epidemiological analysis of landmine and other explosive ordnance casualties using data from over 100,000 casualties across 17 settings. They find a 38.8% case fatality rate, with females at higher risk than males, children less likely to die than adults, and improvised explosive devices most lethal.

Introduction

Explosive ordnance (EO) including landmines, improvised explosive devices (IEDs) and other explosive remnants of war (ERW) are known to contaminate at least fifty-eight countries¹, and continue to threaten the health of affected communities across generations. EO contamination has been identified as a significant public health concern for at least the past two decades², with an increase in cumulative totals of casualties since 2019¹. For civilians within these communities, the consequences are cross-cutting and severe. In addition to direct effects including death and injury, EO contamination has indirect implications with long-term and intergenerational impacts including damage to infrastructure, contamination of arable land and food insecurity, loss of livelihood, limited education opportunities and hindered reconstruction efforts^3–5.

Despite the scale of this problem for civilians in conflict and post-conflict settings globally, mortality estimates pertaining to this concern are limited in scope. Accurate mortality estimation is important as a cornerstone of public health metrics⁶. These data serve numerous essential purposes including: (i) insights into mortality patterns among conflict-affected populations, (ii) identification of at-risk groups, (iii) a benchmark to monitor impact of interventions in trauma care or high-level policy decisions. Such information can guide injury prevention strategies and casualty care interventions to reduce preventable death and disability⁷. Despite the importance of such metrics, civilian mortality estimates for EO have predominantly relied on sub-regional population samples⁸, which may not adequately reflect the true scope and nature of mortality trends. Many existing figures are derived from second-hand data sources such as media reports⁹, and there is a paucity of primary epidemiological research on this subject.

This study addresses this knowledge gap by conducting a multi-country epidemiological study on EO casualties in low-resource settings. This study evaluates overall mortality among EO casualties, providing external validity for existing estimates that use indirect methods. We show that explosive ordnance yields a high case fatality rate of 38.8% with females experiencing a higher odds of fatal outcomes than males, children showing lower odds of death than adults, and improvised explosive devices demonstrating the greatest lethality. These findings illustrate the substantial mortality burden and indiscriminate nature of explosive weapons, and may help inform targeted injury prevention and trauma care performance improvement initiatives designed to reduce preventable death and disability among EO casualties.

Methods

Study design

A retrospective analysis of secondary data on EO casualties from low- and lower-middle income conflict-affected countries was conducted. Low- and lower-middle income countries were defined in accordance with World Bank economic classifications, as of 2020¹⁰. These data were multi-sourced and were collected from mine action authorities and centres, international non-government organisations (NGOs), and international bodies such as the United Nations (UN) using the Hardly Reached Data (HaRD) framework, developed in an earlier stage of this research, to guide the data collection process. These data include surveillance data, retrospective and prospective surveys, and data collected through organisations’ existing operations. Casualty information contained within these datasets includes demographics, mortality, injury data, and information related to the incident. Military casualties did not distinguish between combatant and non-combatant contexts and were grouped as non-civilians. Similarly, individuals identified as deminers were also categorised as non-civilians because their role involved specialised EO disposal activities.

Study participants

The study population consisted of persons of all ages, genders and nationalities who have been killed or injured in EO incidents and for whom casualty data was recorded by a humanitarian demining organisation (e.g., national mine action authority or centre, international or local organisation responding for mine action). Organisations were approached in countries that were determined eligible for inclusion by meeting the following criteria: (1) Presence of EO as reported by the Landmine and Cluster Munitions Monitors, and (2) Have casualties from EO. Countries were excluded if: (1) Casualties were the result of overseas military deployment, or (2) the country was classified as a high-income country, or (3) contact details were not available to seek data access, or (4) no mine action authority, centre, or international agency was operating within the country. As this research examines operational data, as well as data that are routinely collected to meet the requirements of the Anti-Personnel Mine Ban Convention and the Cluster Munitions Convention, the RECORD statement checklist was used¹¹. These datasets were contributed by organisations responsible for explosive ordnance surveillance, victim-assistance reporting, and national mine action programmes. These data were originally collected for operational, programmatic or treaty purposes, not for research, and therefore individual organisations did not obtain research-specific ethical approval at the time of collection and informed-consent practices are not known. Ethical approval for this secondary analysis of de-identified operational data was obtained from the University of Queensland’s Human Research Ethics Committee (Approval No. 2018001361; ethic review exemption clearance No. 201800064). All contributing organisations permitted the use of their de-identified operational datasets for research purposes.

Data analysis

All data sources were explored for a minimum set of core variables as per the World Health Organization’s (WHO) guidance for surveillance of injuries due to landmines and unexploded ordnance¹². Duplicate entries were excluded. Case fatality rates and the number of deaths were presented by different demographic and incident characteristics. Univariate and multivariate logistic regression models were used to assess which characteristics were associated with mortality and survival in casualties. From these models, odds of death by sex, adult/child, age group, activity and explosive type are presented. A subgroup analysis using the data from Abkhazia, Angola, Cambodia, Chechnya, Lao PDR, Libya, Somaliland and Ukraine allowed a finer modelling of age groups (e.g., <1 year, 1–4 years, 5–9 years… >65 years). The association between activity and death from EO was modelled separately for the adult and child subgroups. When assessing the association between explosive type and odds of death from EO we also present results adjusted for country and year to account for differences in EO over time and in different conflicts. Analyses were undertaken using STATA version 17. Data visualisations were created using Flourish Studio (https://flourish.studio).

Results

A total of 105,913 casualties were identified across 17 datasets, each from a different country or geographic territory (i.e., Abkhazia, Afghanistan, Angola, Armenia, Cambodia, Chechnya, Colombia, Georgia, Kosovo, Lao People’s Democratic Republic, Lebanon, Libya, Nagorno Karabakh, Somaliland, Sudan, Ukraine, and Yemen). Each dataset varied in both size and variables captured, with the largest dataset provided from Lao PDR (n = 48,696, 46.14%) and the smallest from Kosovo (n = 50, 0.05%). The time periods for data collection and corresponding data custodians are detailed in Supplementary Table S1.

Males represented 88.3% of the pooled sample (n = 81,809) with most casualties being adults aged 18 years and older (n = 62,279, 61.6%). Civilians represented 89.9% of the sample (n = 95,192), while 7.1% (n = 2624) were military or professional deminers. ERW (e.g., mortars, grenades, abandoned ammunition, artillery shells) were the most common causative agent across the pooled sample (n = 50,033, 47.2%). Landmines (i.e., antipersonnel and antitank) accounted for 29.8% of incidents (n = 31,599). While almost a quarter (n = 24,812, 23.4%) of events occurred due to military activities (e.g., active conflict), agricultural and food-gathering pursuits (e.g., farming, tending to animals, collecting food, water or wood) (n = 21,318, 20.1%), travelling (n = 15,330, 14.5%), and being a bystander (n = 9880, 9.3%) were also common activities preceding an EO incident. The case fatality rate for the pooled sample was 38.8 deaths per 100 casualties. Table 1 presents a summary of demographic and incident characteristics including case fatality rates.

Table 1.

Demographic and incident characteristics for number of casualties, % and case-fatality rates

		n	%	Case-fatality rate
Pooled (all)		105,913	100.0	38.8
Age Groups
Child		28,200	26.6	31.4
Adult		74,793	70.6	42.0
Not specified		2920	2.8	–
Biological Sex
Male		92,399	87.2	38.4
Female		11,829	11.2	44.5
Not specified		1685	1.6	–
Civilian Status
Civilian		95,192	89.9	41.3
Not civilian (military, deminers)		10,721	10.1	24.7
Explosive Type
Antipersonnel (AP) Landmine		29,024	27.4	31.3
Antitank (AT) Landmine		2575	2.4	35.2
Explosive Remnant of War (ERW)		40,131	37.9	30.2
Improvised Explosive Device (IED)		4021	3.8	44.8
Unexploded Ordnance (UXO)		9902	9.3	54.9
Not specified		20,260	19.1	–
Activity at Time of Incident
Active conflict		24,812	23.4	43.9
Agricultural		14,319	13.5	36.7
Bystander		9880	9.3	43.7
Collecting food, water or wood		6999	6.6	35.9
Demining		321	0.3	23.3
Household tasks		532	0.5	22.2
Making fire		2469	2.3	41.9
Playing		4867	4.6	24.9
Tampering		5026	4.7	33.8
Travelling		15,330	14.5	47.6
Work (non-agricultural)		161	0.2	14.5
Not specified		21,197	20.0	–

Demographic characteristics

The case fatality rate for females (44.5%) was higher than that for males (38.4%). Males were more frequently affected by EO incidents overall; however, females had higher odds of death from EO (OR = 1.29, 95 % CI: 1.24–1.34, p < 0.001) compared to males. Adjusting for age group (adult or child) and explosive type caused a small increase in effect with females showing higher odds of death compared to males (AOR = 1.31, 95 % CI: 1.26–1.37, p < 0.001). Civilians had a higher risk of death (OR = 2.23, 95 % CI: 2.13– 2.34, p < 0.001) when compared to non-civilians (i.e., military, professional deminers).

Adults also experienced higher odds of death than children (OR = 1.60, 95% CI: 1.55–1.64, p < 0.001). Adjusting for biological sex, activity and explosive type reduced the effect for the pooled dataset; adults continued to show significantly higher odds of death than children (AOR = 1.50, 95% CI: 1.45–1.56, p < 0.001) and as shown in Fig. 1, all adult age groups had significantly higher odds of death than children (under 18 years). Adults in the age brackets of 45–54 (OR: 5.58, 95% CI: 5.18–6.02, p < 0.001) and 55–64 years (OR: 5.67, 95 % CI: 5.12–6.28, p < 0.001) showed the greater odds of death from EO when compared to the under 18 age group.

Fig. 1.

Unadjusted ORs and 95 % CI for age groups (versus under 18 years) for case fatality (killed v injured) for the pooled subset (created using flourish.studio) OR values are plotted on a logarithmic scale, with the vertical dashed line at OR = 1 representing the null value (no association). Table S2 in the Supplementary file contains the numerical data underlying Fig. 1.

We conducted additional analyses to compare childhood mortality with adult age groups in greater detail. As shown in Fig. 2, all child age groups from one through 17 years experienced lower odds of death compared to the 30–34 years age group.

Fig. 2.

Unadjusted odds ratios for detailed age groups for case fatality (killed or injured) in the subgroup analysis (Abkhazia, Angola, Cambodia, Chechnya, Lao PDR, Libya, Somaliland and Ukraine (created using flourish.studio). Table S3 in the Supplementary file contains the numerical data underlying Fig. 2.

The odds of death relative to injury increased across the lifespan from one through 64 years when compared to the 30–34 years age group, with an increased odds of death shown for those aged between 35 and 64 years (35–39 years: OR = 1.19, 95 % CI: 1.09–1.30, p < 0.001; 40-44 years: OR = 1.65, 95 % CI: 1.50–1.81, p < 0.001; 45–54 years: OR = 2.06, 95% CI: 1.88–2.25, p < 0.001; 55–64 years: OR = 2.10, 95 % CI: 1.87–2.35, p < 0.001). No statistically significant difference was determined for under one year of age (OR = 0.84, 95% CI: 0.45–1.58, p = 0.589) or over 65 years (OR = 1.13, 95% CI: 0.99–1.29, p = 0.080).

Incident characteristics

Travelling (OR = 1.62, 95 % CI: 1.53–1.72, p < 0.001), incident occurring during active conflict (OR = 1.4, 95 % CI: 1.32–1.48, p < 0.001), being a bystander (OR = 1.39, 95 % CI: 1.30–1.48, p < 0.001), and making fire (OR = 1.29, 95 % CI: 1.17–1.41, p < 0.001) were associated with greater odds of death when compared to foraging food, water or wood. Stratifying by age groups (child versus adult) revealed several differences between the activity at the time of the incident. Children experienced higher odds of death than adults during the active conflict phase (OR = 2.72, 95 % CI: 2.38–3.11, p < 0.001), while playing (OR = 1.83, 95% CI: 1.47–2.29, p < 0.001), and as bystanders (OR = 1.69, 95 % CI: 1.48–1.94, p < 0.001).

All explosive types (AT landmine, UXO, ERW, IED) showed statistically significant higher odds of death when compared to AP landmines. After adjusting for country and year of incident within the model, the model indicated an increase in the effect of AT landmines and IEDs, while UXO demonstrated a decrease in effect. IEDs showed a larger increase in death odds than AP landmines in the adjusted model (AOR = 4.02, 95 % CI: 3.71–4.35, p < 0.001). Table 2 presents the unadjusted and adjusted odds ratios for explosive type (with AP landmine as the reference group) and case fatality (killed, injured).

Table 2.

Logistic regression for explosive type and case fatality (death v injury) for the pooled dataset

Explosive type	Unadjusted			Adjusteda
	OR	95% CI	p	AOR	95% CI	p
AP Landmineb	1	–	–	1	–	–
AT Landmine	1.19	1.09 – 1.30	<0.001	2.38	2.16 – 2.62	<0.001
UXO	2.68	2.56 – 2.81	<0.001	1.21	1.14 – 1.27	<0.001
ERW	1.34	1.30 – 1.39	<0.001	1.09	1.05 – 1.14	<0.001
IED	1.78	1.67 – 1.91	<0.001	4.02	3.71 – 4.35	<0.001

^aAdjusted for country and year of the incident.

^bReference category.

Discussion

This study presents one of the largest epidemiologic analyses to date of EO casualties in low-resource settings globally. Key findings included: (i) an overall case fatality rate of 38.8%, (ii) an elevated risk of fatal outcomes in females, (iii) overall, children in this sample experienced lower odds of death compared to adults, and (iv) IEDs had the highest lethality of the weapons examined. The findings of this analysis hold potential to inform injury prevention and trauma care initiatives targeting preventable death and disability for casualties of EO.

Previous research reported country-level case fatality rates ranging from 2.1 to 80%⁸. Such wide variability makes it challenging to draw meaningful conclusions. In this study, the pooled case fatality rate was 38.8%. The factors driving variability between countries/settings are not fully elucidated and merit further research. Overall, these findings demonstrate that over one in three injured by EO die from these injuries. This rate is dramatically higher than the case fatality rate among military casualties of blast injury or civilians in high-resource trauma centres (approximately 2%)^13,14. The high lethality of explosive weapons, often long after conflicts end, has far-reaching implications for future generations through direct and indirect effects of EO contamination. Coordinated efforts conducted in partnership with local health systems in EO-affected settings to strengthen emergency, critical, and operative care hold potential to reduce preventable death among EO casualties^15,16.

Sex differences in armed conflict have been documented previously, with women experiencing an elevated risk of mortality and morbidity associated with their proximity to active conflict, with higher intensity conflicts related to increased risk¹⁷. While there is a paucity of research evaluating sex differences specific to EO, aggregated data from the Landmine Monitor reports that males account for most casualties where sex was known (i.e., 88% in 2024)¹. The current study supports these findings, with males representing 88.3 % of casualties. Females, however, were shown to have a higher case fatality rate and higher odds of death than males, with the effect increasing after adjusting for age group and explosive type. While males are more involved in incidents, when females are injured, they are at increased risk for fatal outcomes. Previous research has reported that women experience an increased mortality risk when in close proximity to active conflict^17,18, and that women experience increased deaths in post-conflict settings due to indirect causes¹⁹. Gender discrimination, socio-cultural norms and reduced access to health care, including delayed evacuation, may explain the discrepancy seen in the current study, however the factors driving sex disparities among EO casualties are not fully elucidated and merit further research.

In addition to women, children are vulnerable to the effects of war and have been documented to have an increased risk of death during armed conflict^17,20–23. A systematic review reported paediatric mortality of 11.0 % versus an adult mortality rate of 9.7% in conflict settings, a smaller difference than expected²³. However, disproportionately high mortality of up to 40% exists among sub-populations such as children with conflict-related burns or traumatic brain injury²³. The current study found that overall, children in EO-contaminated communities (i.e., during and post-conflict) had a lower case fatality rate compared to adults, with adults also experiencing higher odds of death than children. However, for certain activities, children were shown to have an increased risk of death compared with adults (i.e., when in active conflict, while playing and as a bystander). Previous research has established that child blast casualties are often mass casualty events with multiple children playing together²⁴. Congregated playing behaviour may explain the higher odds of death for child bystanders observed within this research. Further research is warranted to evaluate determinants of child mortality in conflict-affected settings to identify the most at-risk subpopulations and introduce appropriate trauma care performance improvement initiatives. Accurate conflict metrics are essential for addressing the needs of these communities, and a better understanding of the epidemiology of conflict can facilitate targeted priority setting, programme planning, and resource allocation. Further, as described above, children experienced a higher proportion of non-fatal injuries, which may have consequences for healthy life expectancy and disability. Accordingly, the health needs of affected children across the lifespan and the implications of conflict on child development must be considered²⁵. It is also important to note that these mortality metrics only reflect the direct health impacts of conflict and do not take into consideration the indirect health effects. Current research suggests that children are disproportionately affected by the indirect health effects of conflict, such as communicable, neonatal and nutritional diseases²⁶. Children residing in conflict settings face the threat of explosive hazards and encounter challenges in accessing clean water and sanitation, education, and immunisation services¹⁷, which further impact their health outcomes, life-long disability and lost opportunities.

This increased mortality risk for adults extended across the lifespan with a pattern of increased death odds corresponding to increased age. Previous research has suggested that adults 15–64 years are most impacted by EO contamination^8,12. While these ages account for greater casualties in this research, the risk for fatal injury impacts older age groups. Older casualties aged between 45–54 years and 55–64 years showed the greatest odds of deaths (and highest case fatality) when compared to the child age group. These findings are important as previous research has not systematically evaluated the relationship between age and EO mortality. In one cohort of airstrike casualties in Tigray, Ethiopia, elderly patients had higher mortality than the overall sample, suggesting vulnerability to EO incidents (41.0% versus 33.7%)²⁷. The health needs of older populations affected by humanitarian crises are often overlooked, yet research highlights the general vulnerabilities of older age groups including greater susceptibility to ill health, malnutrition, disability, comorbidities and marginalisation^28,29.

The mechanism of injury resulting from each explosive type differed in terms of severity with AT landmines, UXO, ERW and IEDs all showing statistically significant higher adjusted odds of death than AP landmines. Similarly, AP landmines reported the lowest case-fatality rate of the explosive types while UXO and IEDs caused the greater proportion of deaths. This is consistent with the intent of these weapons, with AP landmines specifically designed to maim rather than kill^30,31, while the explosive charge and design of the other weapons would likely cause more severe injuries^32,33. Furthermore, IEDs were found to have the highest adjusted odds ratio highlighting the lethality and destructive impact of these devices. Many IED incidents are remotely triggered, rather than victim-activated, to maximise harm. This difference in activation method has implications for this lethality. Increased severity of injury associated with modern IED has been reported in the literature, with the explosive intensity of victim-initiated improvised mines, also known as anti-personnel improvised explosive devices (AP-IEDs), causing more pelvic and upper extremity injuries than conventional AP landmines in military settings³³.

The vulnerability of civilians to the consequences of armed conflict is well-documented, and the disproportionate impact of war on civilians has been widely studied^2,34–37. Exposure to conflict is a significant determinant of poor health, and its effects can extend across generations, exacerbating poverty and marginalising already disadvantaged communities^34,38–40. This study shows that civilians are not only the most frequently injured group but also have higher odds of death from EO compared to military and deminers.

This analysis had several limitations. Firstly, this research utilised secondary data resulting in incomplete data affecting completeness. For instance, coding frames varied across dataset (e.g., non-uniform reporting of age, explosive remnants of war versus unexploded ordnance). Such challenges have been well described related to conflict casualties in low-resource settings⁴¹ and are not unique to this dataset. Further, the use of secondary data for this study could lead to access bias with the data custodians capturing data on communities with whom they have established relationships. Populations without these relationships or with distrust of these gatekeepers are unlikely to be captured within these data and therefore be underrepresented in analysis. Data collection methods varied across datasets, introducing potential information bias. For example, some data were captured retrospectively, which may lead to recall and observer bias. Demographic and incident-level adjustments were beyond the scope of this analysis due to data constraints. Analysis of detailed injury characteristics and anatomical patterns was beyond the scope of this study and will be reported in a separate manuscript. The timing of casualty data collection was often unclear (e.g., at the time of incident or later), and longitudinal follow-up was not possible. Accordingly, the long-term survival of casualties is unknown which could mean mortality rates are underestimated. Military and deminers were included within these data to reflect the comprehensive impact of EO incidents, however this may impact comparability across contexts. Further, sample sizes varied considerably meaning findings are strongly driven by countries with larger datasets and may impact the wider generalisability of this study. It is however, expected that these findings would still be generalisable to affected populations within these countries and similar contexts due to similarities in the mechanism of injury (e.g., weapon types, resource limitations and incident characteristics) and this dataset representing a historical baseline related to EO contamination. To increase the generalisability of EO contamination research more broadly, a random sample of casualties across EO contaminated countries or analyses focused on specific settings or time periods could be used in future research. Despite these limitations, this analysis represents one of the largest datasets on EO casualties in low-resource settings to date.

Conclusions

This report presents the largest multi-country epidemiological analyses of EO casualties in low-resource settings globally to date. This research highlights the lethality of EO among local casualties, with a case fatality rate significantly higher than that observed in high-resource civilian or military trauma systems. Females in this sample were observed to have a higher risk of death compared to males, highlighting an opportunity for further research into the drivers of this disparity. Finally, IEDs were shown to have a significantly increased odds of death, illustrating the deadliness and indiscriminate nature of these devices.

The findings of this research underscore the need for interventions that reduce mortality in EO-affected settings. A structured evaluation of how trauma care advances in high-resource military and civilian trauma systems can be translated in a context-appropriate manner to care for EO casualties is needed. Additional investigations that utilise post-mortem data to understand potentially survivable death in the study setting are also warranted as this information can improve preparedness and planning to tailor trauma care interventions and ensure optimal resource allocation and subspeciality expertise.

Author contributions

SP: Conceptualisation, Methodology, Formal Analyses, Writing—Original Draft. JD: Supervision, Writing—reviewing and editing. HW: Writing—Reviewing and editing. VT: Supervision, Writing—reviewing and editing. MW: Supervision, Formal analysis, Writing—reviewing and editing. All authors meet the ICMJE criteria and approved the final manuscript.

Peer review

Peer review information

Communications Medicine thanks Maximilian P. Nerlander and Ofer Almog for their contribution to the peer review of this work.

Data availability

The data that support the findings of this study are not openly available due to sensitivity and to respect terms of use for the data from third parties. Data are located in controlled access data storage at The University of Queensland. Researchers with a legitimate research purpose may request access to the minimum dataset by contacting the corresponding author. Applicants must provide evidence of relevant ethical approval and agree to conditions outlined in the data use agreement. Requests will be acknowledged within 15 business days and decisions regarding access will be made within 6–8 weeks. Any data provided may be used only for the approved research purpose, must not be shared with unauthorised parties, and may not be used for attempts to re-identify individuals or breach terms set by the data providers. All source data underlying the figures within this manuscript are provided in the supplementary files (Tables S2 and S3).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s43856-026-01430-y.