Temporal trends in the epidemiology of primary biliary cholangitis in the public healthcare system of Brazil: a large population-based study Authors: Guilherme Grossi Lopes Cançado, Cláudia Alves Couto, Luciana Costa Faria, Bernardo de Faria Moraes, Gustavo André Pedral Diniz Leite, Paulo Lisboa Bittencourt, Cristiane Alves Vilella-Nogueira, Vivian Rotman, Debora Raquel Benedita Terrabuio, Eduardo Luiz Rachid Cançado, Maria Lúcia Ferraz Published: 2026 ABSTRACT SummaryBackgroundPopulation-based data on the epidemiology of primary biliary cholangitis (PBC) in Latin America are lacking. We aimed to characterise incidence, prevalence, and geographic patterns of PBC within Brazil's public healthcare system.MethodsWe performed a nationwide population-based retrospective study using de-identified administrative data from the Unified Health System (DATASUS), curated through an artificial intelligence platform. Adults (≥16 years) with ICD-10 code K74.3 and public dispensation of ursodeoxycholic acid between 1 January 2021 and 31 December 2024 were included. Annual World Health Organization (WHO) age-standardised incidence and prevalence were estimated by sex, age group, and macroregion. Temporal trends were modelled with log-linear regression to obtain average annual percent change (AAPC). Correlation with macroregional Human Development Index (HDI) was explored.FindingsWe identified 10,125 PBC patients [77.75% (7872) women; mean age 54.95 ± 20 years], within a nationwide dataset comprising approximately 230 million records. Age-standardised prevalence increased from 10.5 to 17.8 per 100,000 people (AAPC +18.2%; 95% CI: 12.5; 24.3%); while age-standardised incidence modestly declined from 4.00 to 3.42 per 100,000 person-years (AAPC −5.0%; 95% CI: −6.3; −3.7%). The female-to-male prevalence ratio in 2024 was 3.67:1, with peak rates at ages 66–70 years. A marked south–north gradient was observed, with highest rates in the South and Southeast and lowest in the North and Northeast; macroregional HDI showed a positive correlation with both incidence and prevalence.InterpretationIn Brazil, PBC is characterised by rising prevalence, relatively stable incidence, a narrower female predominance than historically reported, and marked geographic inequalities. These findings suggest possible under-recognition in less developed regions and provide a foundation for targeted public health strategies. The observed south–north gradient warrants further investigation in relation to environmental, socioeconomic, and healthcare-access factors.FundingThis study is supported by a research grant from IPSEN. Research in contextEvidence before this studyBefore undertaking this study, we reviewed the published literature on the epidemiology of primary biliary cholangitis (PBC), focussing on population-based estimates of incidence, prevalence, temporal trends, sex distribution, and geographical variation. We searched PubMed/MEDLINE, Embase, and Scopus from January 1st 1990 to December 10th 2025, without language restrictions, using the terms “primary biliary cholangitis”, “primary biliary cirrhosis”, “epidemiology”, “incidence”, “prevalence”, “population based”, “nationwide”, “registry”, “Brazil”, “Latin America”, and “Americas”. We also screened the reference lists of relevant articles and reviews. We included population based and nationwide studies, registry-based analyses, and systematic reviews or meta analyses reporting PBC incidence and/or prevalence; The available evidence was largely derived from Europe, North America, and Asia–Pacific settings. A recent global systematic review and meta-analysis reported a pooled prevalence of 18.1 per 100,000 people and a pooled incidence of 1.8 per 100,000 person-years, with substantial geographical heterogeneity and higher burden in North America, Europe, and the Western Pacific. Previous studies have also suggested a geographical gradient in PBC occurrence, including possible associations with latitude. However, the existing evidence was heterogeneous in case definitions, ascertainment methods, and source populations, and we found no nationwide population-based study from Latin America, including Brazil.Added value of this studyThis study provides, to our knowledge, the first nationwide population-based estimates of PBC epidemiology in Latin America and one of the largest population-based assessments of PBC to date. Using nationwide administrative data from Brazil's public healthcare system, we identified 10,125 unique patients with PBC treated with ursodeoxycholic acid and estimated national and subnational incidence and prevalence from 2021 to 2024. We found that age-standardised prevalence increased from 10.5 to 17.8 per 100,000 people, whereas age-standardised incidence declined modestly from 4.00 to 3.41 per 100,000 person-years over the same period. We also observed a narrower female predominance than historically reported, with a female-to-male prevalence ratio of 3.67:1 in 2024, and marked geographical heterogeneity characterised by a south-to-north gradient, with the highest burden in the South and Southeast and the lowest in the North and Northeast. These findings add an important regional perspective from a large, highly admixed middle-income country with substantial internal socioeconomic and healthcare-access disparities.Implications of all the available evidenceTaken together with previous international studies, our findings suggest that the burden of PBC is increasing globally, but that measured disease frequency is strongly influenced by regional differences in case ascertainment, healthcare access, and possibly environmental exposures. The Brazilian data show that prevalence can rise despite relatively stable or modestly declining incidence, while also revealing major within country disparities that may reflect under-recognition in less-resourced regions or even sunlight exposure time. These findings support the need for greater awareness of PBC, improved diagnostic pathways, and more equitable access to specialist care and treatment, particularly in underserved areas of Latin America. They also highlight the importance of future studies to validate administrative case definitions, quantify under ascertainment, and investigate the environmental, socioeconomic, and genetic determinants of the marked geographical heterogeneity observed across Brazil. Research in context Evidence before this study Before undertaking this study, we reviewed the published literature on the epidemiology of primary biliary cholangitis (PBC), focussing on population-based estimates of incidence, prevalence, temporal trends, sex distribution, and geographical variation. We searched PubMed/MEDLINE, Embase, and Scopus from January 1st 1990 to December 10th 2025, without language restrictions, using the terms “primary biliary cholangitis”, “primary biliary cirrhosis”, “epidemiology”, “incidence”, “prevalence”, “population based”, “nationwide”, “registry”, “Brazil”, “Latin America”, and “Americas”. We also screened the reference lists of relevant articles and reviews. We included population based and nationwide studies, registry-based analyses, and systematic reviews or meta analyses reporting PBC incidence and/or prevalence; The available evidence was largely derived from Europe, North America, and Asia–Pacific settings. A recent global systematic review and meta-analysis reported a pooled prevalence of 18.1 per 100,000 people and a pooled incidence of 1.8 per 100,000 person-years, with substantial geographical heterogeneity and higher burden in North America, Europe, and the Western Pacific. Previous studies have also suggested a geographical gradient in PBC occurrence, including possible associations with latitude. However, the existing evidence was heterogeneous in case definitions, ascertainment methods, and source populations, and we found no nationwide population-based study from Latin America, including Brazil. Added value of this study This study provides, to our knowledge, the first nationwide population-based estimates of PBC epidemiology in Latin America and one of the largest population-based assessments of PBC to date. Using nationwide administrative data from Brazil's public healthcare system, we identified 10,125 unique patients with PBC treated with ursodeoxycholic acid and estimated national and subnational incidence and prevalence from 2021 to 2024. We found that age-standardised prevalence increased from 10.5 to 17.8 per 100,000 people, whereas age-standardised incidence declined modestly from 4.00 to 3.41 per 100,000 person-years over the same period. We also observed a narrower female predominance than historically reported, with a female-to-male prevalence ratio of 3.67:1 in 2024, and marked geographical heterogeneity characterised by a south-to-north gradient, with the highest burden in the South and Southeast and the lowest in the North and Northeast. These findings add an important regional perspective from a large, highly admixed middle-income country with substantial internal socioeconomic and healthcare-access disparities. Implications of all the available evidence Taken together with previous international studies, our findings suggest that the burden of PBC is increasing globally, but that measured disease frequency is strongly influenced by regional differences in case ascertainment, healthcare access, and possibly environmental exposures. The Brazilian data show that prevalence can rise despite relatively stable or modestly declining incidence, while also revealing major within country disparities that may reflect under-recognition in less-resourced regions or even sunlight exposure time. These findings support the need for greater awareness of PBC, improved diagnostic pathways, and more equitable access to specialist care and treatment, particularly in underserved areas of Latin America. They also highlight the importance of future studies to validate administrative case definitions, quantify under ascertainment, and investigate the environmental, socioeconomic, and genetic determinants of the marked geographical heterogeneity observed across Brazil. Introduction Primary biliary cholangitis (PBC) is a chronic cholestatic autoimmune liver disease characterised by progressive, immune-mediated destruction of small intrahepatic bile ducts.1 Although the disease predominantly affects middle-aged women,2 contemporary data indicate that prevalence among men and racial/ethnic minorities is increasingly recognised.1 As diagnostic awareness and therapeutic options improve, leading to longer patient survival, understanding the evolving epidemiology of PBC is crucial for guiding public health policy, allocating healthcare resources, and uncovering etiological clues related to its complex interplay of genetic and environmental risk factors. The global epidemiology of PBC has been increasingly clarified, with a recent comprehensive meta-analysis reporting a pooled global prevalence of 18.1 cases per 100,000 people and an incidence rate of 1.8 per 100,000 person-years.2 However, these global averages mask significant geographical variability, with the highest prevalence observed in North America (28.5 per 100,000), followed by Europe (18.6 per 100,000), and the Western Pacific region (9.9 per 100,000), highlighting a heterogeneous global footprint for the disease.2 Intriguingly, this variation does not appear random, as research has identified significant north-south gradients for PBC incidence, suggesting that regional socioeconomic and environmental factors may play a critical role in its pathogenesis.3 Furthermore, the global prevalence of PBC has increased over time,2 a trend supported by national and regional studies from Europe,4,5 North America6 and the Asia–Pacific region.7, 8, 9, 10 This rise is likely attributable to a combination of improved diagnostic awareness, more sensitive case ascertainment methods, and longer patient survival, especially with the availability of new treatment lines. There is also evidence that PBC may follow a more aggressive course in certain populations, including Hispanic, Canadian First Nations, and African American groups,11, 12, 13 suggesting that genetic background may modulate biological pathways underlying clinical expression and progression. Brazil has a population of a highly admixed origin, with varying proportions of Amerindian, African, and European genetic ancestries, shaped by local historical interactions with migrants brought by the slave trade, European settlement, and indigenous populations, making reliable self-identification of ethnicity extremely challenging.14,15 To our knowledge, no nationwide population-based study on the epidemiology of PBC has been conducted in Latin America. Brazil, as the largest country in the region, constitutes a critically important yet previously understudied region for understanding the epidemiology of PBC. Its continental-scale territory encompasses wide climatic, socioeconomic, and ethnic diversity, generating substantial subnational health disparities16 (GBD 2016 Brazil Collaborators, 2018) and offering a unique opportunity to investigate geographic patterns of disease. Therefore, the aim of this large population-based study is to describe the temporal trends in the national incidence and prevalence of PBC within the Brazilian public healthcare system (Sistema Único de Saúde, SUS) and to characterise the disease burden across Brazil, using an artificial intelligence (AI) platform for data collection. Methods Data source and AI platform Publicly available data were obtained from the Department of Health Informatics/Ministry of Health (DATASUS; http://www2.datasus.gov.br/DATASUS) using the TT Disease Explorer platform, developed by TechtrialsTM (Techtrials™ Healthcare Data Science, Brazil). This platform continuously collects structured national epidemiological and procedural data from the Unified Health System (SUS) through artificial intelligence (ETLs and Webcrawlers). Data processing and storage are implemented on Microsoft Azure with SQL, M, and DAX code, and visualisation is performed in Power BI (Microsoft Corporation, Redmond, WA, USA). Records are encrypted and linked through a unique patient identifier; a dedicated deduplication routine reconciles encounters across facilities to prevent double counting when the same individual receives care in different locations. Brazilian unified health system (Sistema Único de Saúde–SUS) and PBC diagnosis Brazil has a mixed public–private healthcare system. The Unified Health System (SUS) is one of the largest public health systems worldwide, guarantees universal health coverage, and finances comprehensive care, including high-cost therapies, which are provided free of charge to the entire population through disease-specific clinical protocols and therapeutic guidelines (Protocolos Clínicos e Diretrizes Terapêuticas, PCDTs). The private sector, which complements the public system, serves individuals covered by health insurance. An analysis of the distribution of 493,010 hospital beds in Brazil (National Register of Health Establishments—CNES, 2019) shows that 66.6% of beds are allocated exclusively to SUS, serving 77.8% of the Brazilian population, whereas the remaining 22.2% of the population (those with some form of private health insurance) have access to 33.4% of total beds. Notably, patients who use private services often rely on SUS for selected high-cost treatments (e.g., ursodeoxycholic acid) that are not covered by private health insurance. As a result, SUS-based registries are likely to capture the vast majority of individuals living with PBC in Brazil. In 2019, a national PCDT for primary biliary cholangitis was issued, under which ursodeoxycholic acid (UDCA) is provided in public high-cost pharmacies at no cost to eligible patients who fulfil prespecified clinical criteria, improving treatment access and standardisation of care. Per the PCDT, a diagnosis of PBC is confirmed when at least two of the following three criteria are present: (1) persistent elevation of cholestatic enzymes, particularly alkaline phosphatase; (2) antimitochondrial antibodies (AMA) and/or antinuclear antibodies (ANA) by indirect immunofluorescence showing nuclear membrane or multiple nuclear dot patterns, or by ELISA positivity for anti-gp210 or anti-sp100; and (3) histologic evidence of nonsuppurative destructive cholangitis. Patients with autoimmune hepatitis and primary sclerosing cholangitis are excluded from this protocol and do not have public access to UDCA. Study design and population This was a large, population-based, retrospective observational study using routinely collected health administrative data linked and curated within DATASUS. DATASUS is an open-access, population-based health and disease registry that compiles information from Brazil's Unified Health System (SUS) on medical procedures, hospital admissions and discharges, outpatient consultations, medication dispensation, mortality, and some demographic variables. It comprises records on approximately 230 million individuals and effectively covers the national population, as every citizen is entitled to SUS care. All records are anonymised; de-identified individuals can be longitudinally tracked across facilities and over time through a unique national health identification number. In this study, we constructed a population-based PBC cohort using the following sequential case definition: (1) the presence of the International Classification of Diseases, 10th Revision (ICD-10) code K74.3; (2) documented free dispensation of ursodeoxycholic acid (UDCA) by SUS following fulfilment of the national PCDT eligibility criteria for PBC; and (3) ≥ 16 years of age. We analysed DATASUS records meeting this definition from January 1 2021, to December 31 2024. Although historical data are available from 2008 onward, we restricted analyses to 2021–2024 to impose a two-year washout period after publication of the PBC PCDT (2019), minimising inclusion of prevalent cases who might have newly accessed SUS for UDCA rather than representing incident PBC. The two-code minimum (ICD-10 plus UDCA dispensation) was used to reduce misclassification error and increase the reliability of the data. To quantify the effect of the UDCA-dispensation requirement on cohort ascertainment, we also determined the number of individuals identified by ICD-10 code K74.3 alone and calculated the proportion retained after applying the combined case definition. Epidemiological rates were not calculated using the ICD-10-only definition because the diagnostic accuracy of an isolated K74.3 code in DATASUS has not been independently validated. Since data preceding implementation of the PCDT were not collected within the same standardised diagnostic framework, they were considered insufficiently comparable and potentially unreliable for the present case definition and were therefore excluded from the analysis. Statistical analysis Variables and geography We extracted age, sex, state and municipality of residence, and dates of entry and exit from the registry. Age was grouped in 5 year bands from 16 to ≥90 years. Geographic analyses were conducted at two levels: Brazil's five macro-regions (North, Northeast, Southeast, South, and Midwest/Central-West) and by state, following the classifications of the Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística, IBGE). Case dating and cohorts The index date was defined as the first recorded encounter carrying a PBC code in SUS. For incidence, we searched all DATASUS records back to January 1 2021 to confirm no prior PBC code exists before the calendar year under analysis. Point prevalence (cases per 100,000 people) was calculated annually by including individuals with ≥1 PBC-coded encounter between January 1 and December 31 of each year (2021–2024). Incidence (new cases per 100,000 person-years) was calculated for each calendar year by including individuals with a first-ever PBC code within that year in the administrative database. Denominators and standardisation Annual population denominators were the IBGE mid-year estimates for Brazil and its subnational units. To enable comparisons across populations with different age structures, age-standardised incidence and prevalence were computed using the World Health Organization world standard population as the denominator. Descriptive and trend analyses Age and sex at the index date were summarised descriptively. Non-standardised incidence/prevalence rates were stratified by sex and age. Temporal trends were modelled using generalised linear models. For incidence rates, we fit Poisson regression with a log link and an offset of log(population), assessing overdispersion and refitting with a negative binomial model if indicated. For prevalence, we modelled annual counts with a log link and the population offset. Average annual percent change (AAPC) and 95% confidence intervals (CIs) in incidence and prevalence, without stratification by sex or age group, was estimated using log-linear models under the joinpoint formulation and were derived from the year coefficient as AAPC = [eβ − 1] × 100%[eˆ{∖beta}-1]∖times 100∖%[eβ − 1] × 100%. For incidence AAPC analysis, the 95% confidence interval for some subgroups could not be estimated under the Poisson model with HC0 robust variance estimator due to numerical instability of the variance estimator given the limited number of time points and the sparsity of event data. Socioeconomic correlations The association between the Human Development Index (HDI) and the epidemiological indicators was evaluated using data from 2021, the most recent year with State-level HDI estimates available in Brazil (Atlas do Desenvolvimento Humano no Brasil; PNUD; IPEAD; FJP, 2021). To explore the relationship between HDI and the five Brazilian macroregions (North, Northeast, Southeast, South, and Central-West), the regional HDI values were calculated using the last available publication from 2010 (Desenvolvimento humano nas macrorregiões brasileiras: 2016 – Brasília: PNUD: IPEA: FJP, 2016). Correlation analyses were performed using the Spearman or Pearson correlation methods, as appropriate, considering both the overall HDI and its three subdimensions: life expectancy, education, and per capita income indicators, when available. All analyses were performed using R software.17 The epiR package18 was employed for the computation of epidemiological measures, the sandwich package19, 20, 21 for robust variance estimation, and the lmtest package22 for hypothesis testing within regression models. Graphical analyses were conducted to explore the spatio-temporal distribution of prevalence and incidence rates and to present stratified estimates by sex, age group, and macroregion in a clear and reproducible manner. For visualization of tabular data, the ggplot2 package23 was used, complemented by ggpubr for publication-quality graphical refinement24 and scales for axis scaling, labelling, and numeric formatting.25 Geographical representations of prevalence and incidence across federative units and macroregions were generated by integrating official spatial datasets provided by geobr26 in simple features format, processed and managed with the sf package.27,28 This approach ensured high-resolution thematic maps and analytically consistent geospatial outputs. Ethics approval Institutional ethics review was waived by the UNIFESP Research Ethics Committee because this study used de-identified, publicly available DATASUS data and involved no contact with human participants; therefore, informed consent was not required. Role of the funding source This study was supported by a research grant from Ipsen. Ipsen reviewed the study protocol and the manuscript for scientific accuracy but had no role in drafting the content or in the final decisions regarding its content. The authors retained full responsibility for the study design, data collection and analysis, interpretation of the results, and the final content of the manuscript. Results Overall, 10,787 individuals had at least one record containing ICD-10 code K74.3 during the study period. Of these, 10,125 [93.9% (9507/10,125)] also had documented public UDCA dispensation, fulfilled the primary case definition, and were included in the final analysis. Therefore, 662 individuals, corresponding to 6.14% (662/10,787) of the ICD-10-coded population, were excluded after application of the UDCA criterion. There was a predominance of females [77.75% (7872/10,125)] as compared to male patients [22.25% 2253/10,125], with an average age of 54.95 ± 20 years. Overall prevalence and temporal trends (2021–2024) Crude national prevalence rose steadily from 3.0 per 100,000 people (95% CI: 2.91–3.08) in 2021 to 5.23 per 100,000 (95% CI: 5.12–5.34) in 2024 (Table 1). Women consistently exhibited higher prevalence than men across all years: 4.55 (95% CI: 4.41–4.69) vs 1.29 (95% CI: 1.21–1.37) in 2021, rising to 8.00 (95% CI: 7.82–8.19) vs 2.18 (95% CI: 2.08–2.28) in 2024 (Supplementary Table S1). These patterns were reproduced in age–sex strata, with female rates exceeding male rates at every age band.Table 1Crude prevalence and incidence rates of PBC in Brazil, 2021–2024.YearCrude prevalence rate (CI 95%)AAPC [2021–2024] (CI 95%)Crude incidence rate (CI 95%)AAPC [2021–2024] (CI 95%)20213.00 (2.91–3.08)19.5% (18.2–20.7)1.13 (1.08–1.18)−4.4% (NE)a20223.94 (3.84–4.03)1.05 (1.0–1.10)20234.66 (4.56–4.77)1.02 (0.97–1.07)20245.23 (5.12–5.34)0.98 (0.94–1.03)Abbreviations: AAPC, Average Annual Percent Change; NE, not estimable; CI 95%, 95% confidence interval; PBC, primary biliary cholangitis.aThe confidence interval for the crude incidence AAPC could not be calculated due to instability of the robust variance estimator (HC0) with the available number of time points. Trend significance was confirmed by the p-value. Table 1Crude prevalence and incidence rates of PBC in Brazil, 2021–2024.YearCrude prevalence rate (CI 95%)AAPC [2021–2024] (CI 95%)Crude incidence rate (CI 95%)AAPC [2021–2024] (CI 95%)20213.00 (2.91–3.08)19.5% (18.2–20.7)1.13 (1.08–1.18)−4.4% (NE)a20223.94 (3.84–4.03)1.05 (1.0–1.10)20234.66 (4.56–4.77)1.02 (0.97–1.07)20245.23 (5.12–5.34)0.98 (0.94–1.03)Abbreviations: AAPC, Average Annual Percent Change; NE, not estimable; CI 95%, 95% confidence interval; PBC, primary biliary cholangitis.aThe confidence interval for the crude incidence AAPC could not be calculated due to instability of the robust variance estimator (HC0) with the available number of time points. Trend significance was confirmed by the p-value. Crude prevalence and incidence rates of PBC in Brazil, 2021–2024. Abbreviations: AAPC, Average Annual Percent Change; NE, not estimable; CI 95%, 95% confidence interval; PBC, primary biliary cholangitis. The confidence interval for the crude incidence AAPC could not be calculated due to instability of the robust variance estimator (HC0) with the available number of time points. Trend significance was confirmed by the p-value. Prevalence increased monotonically with age to a peak in late working life and then declined. Pooled across years, prevalence rose from 2.22 (95% CI: 2.11–2.34) in ages 16–20 to 20.78 (95% CI: 20.26–21.30) at 66–70, falling thereafter to 3.58 (95% CI: 2.91–4.35) at ≥90 (Supplementary Table S2). Year-by-year profiles showed the same shape with progressively higher peaks: the maximum shifted from ∼14.2 at 61–70 in 2021 to 24.99 (95% CI: 23.90–26.11) at 66–70 in 2024; values at the youngest (16–20) and oldest (≥90) ages remained lowest (e.g., 2.79 [95% CI: 2.53–3.07] and 2.74 [95% CI: 1.67–4.23] in 2024) (Fig. 1A).Fig. 1Heatmap showing PBC prevalence (A) and incidence (B) rates overtime stratified by age-groups. Heatmap showing PBC prevalence (A) and incidence (B) rates overtime stratified by age-groups. Marked regional differences were observed. Prevalence was highest in the South and Southeast, increasing from 3.90 (95% CI: 3.69–4.13) and 3.25 (95% CI: 3.14–3.38) in 2021 to 5.72 (95% CI: 5.46–5.99) and 5.61 (95% CI: 5.45–5.76) in 2024, respectively. Similarly, the Central-West rose from 1.79 (95% CI: 1.59–2.00) to 3.25 (95% CI: 2.98–3.53), the Northeast from 0.92 (95% CI: 0.84–1.00) to 2.29 (95% CI: 2.17–2.42), and the North from 0.30 (95% CI: 0.22–0.39) to 1.08 (95% CI: 0.94–1.24) (Fig. 2A; Supplementary Table S3). State-level prevalence in 2024 is presented in Supplementary Table S4.Fig. 2PBC prevalence (A) and incidence (B) by macroregions (2021–2024). Bars represent 95% CI. PBC prevalence (A) and incidence (B) by macroregions (2021–2024). Bars represent 95% CI. Incidence and temporal trends (2021–2024) From 2021 to 2024, 6942 incident cases were identified. National crude incidence rates per 100,000 people showed a subtle downward trajectory: 1.13 (95% CI: 1.08–1.18) in 2021, 1.05 (95% CI: 1.00–1.10) in 2022, 1.02 (95% CI: 0.97–1.07) in 2023, and 0.98 (95% CI: 0.94–1.03) in 2024 (Table 1). Women had consistently higher annual incidence than men (e.g., 2024: 1.47 [95% CI: 1.40–1.56] vs 0.44 [95% CI: 0.40–0.49]); when aggregated across 2021–2024, cumulative incidence was 1.56 (95% CI: 1.52–1.60) among women, 0.48 (95% CI: 0.45–0.50) among men, and 1.04 (95% CI: 1.02–1.06) overall (Supplementary Table S5). Incidence increased with age to a peak in the seventh decade and subsequently declined. Collapsing across years, the highest rate occurred at 66–70 (4.46, 95% CI: 4.22–4.70), with lower values at younger and oldest ages (e.g., 0.69 [95% CI: 0.62–0.76] at 16–20; 0.72 [95% CI: 0.44–1.10] at ≥90) (Supplementary Table S6). Year-specific peaks at 66–70 diminished over time: 5.06 (95% CI: 4.55–5.61) in 2021, 4.73 (95% CI: 4.25–5.25) in 2022, 4.30 (95% CI: 3.85–4.79) in 2023, and 3.82 (95% CI: 3.40–4.28) in 2024 (Fig. 1B). Higher incidence rates were observed in the most developed regions of the country–Southeast [1.12 per 100,000 people (95% CI: 1.08–1.15)] and South [0.85 per 100,000 (95% CI: 0.80–0.90)]–while North [0.26 per 100,000 (95% CI: 0.23–0.30) and Northeast regions [0.55 per 100,000 (95% CI: 0.52–0.58)] presented the lowest incidence rates (Fig. 2B; Supplementary Table S7). In 2024, incidence ratios remained significantly higher in the southeast [3.69 (95% CI: 2.8–4.99)] and south regions [3.42 (95% CI: 2.55–4.68)], compared to north (reference). Annual state-level estimates, using Pará state as reference, revealed heterogeneous levels and trajectories, with comparatively higher incidence ratios in Ceará [42.14 (95% CI: 15.98–170.81)] and São Paulo [39.69 (95% CI: 15.26–159.9) in 2024] and lowest in Piauí [3.28 (95% CI: 0.72–16.65)] and Amazonas [6.35 (95% CI: 1.9–28.64)]. State-level incidence in 2024 is presented in Supplementary Table S8. Age-standardised rates and temporal change After standardisation to the WHO world standard population, the age-standardised prevalence increased from 10.5 (95% CI: 10.3–10.8) in 2021 to 17.8 (95% CI: 17.5–18.1) per 100,000 people in 2024, while the age-standardised incidence decreased from 4.00 (95% CI: 3.84–4.15) to 3.42 (95% CI: 3.27–3.56) per 100,000 person-years. The average annual percent change (AAPC) based on Brazilian population estimates, was +19.5% (95% CI: 18.2–20.7; p < 0.001) for prevalence and −4.4% for incidence (p < 0.001; 95% CI was not estimable under the Poisson model with HC0 variance, warranting caution in interpreting effect size precision). AAPC in age-standardised prevalence was +18.2% (95% CI: 12.5–24.3; p = 0.022), whereas age-standardised incidence declined by −5.0% (95% CI: −6.3 to −3.7; p = 0.018) (Fig. 3 and Table 2).Fig. 3WHO age-standardised prevalence and incidence rates of PBC in Brazil. Bars represent 95% CI.Table 2Average annual percent changes in WHO age-standardised PBC prevalence and incidence rates in Brazil, 2021–2024.OutcomeModelAAPC, % (95% CI)p-valueAge-standardised prevalenceWeighted log-linear regression of log-ASR+18.2 (12.5–24.3)0.022Age-standardised incidenceWeighted log-linear regression of log-ASR−5.0 (−6.3 to −3.7)0.018Abbreviations: AAPC, average annual percent change; ASR, age-standardised rate; CI, confidence interval; PBC, primary biliary cholangitis. WHO age-standardised prevalence and incidence rates of PBC in Brazil. Bars represent 95% CI. Table 2Average annual percent changes in WHO age-standardised PBC prevalence and incidence rates in Brazil, 2021–2024.OutcomeModelAAPC, % (95% CI)p-valueAge-standardised prevalenceWeighted log-linear regression of log-ASR+18.2 (12.5–24.3)0.022Age-standardised incidenceWeighted log-linear regression of log-ASR−5.0 (−6.3 to −3.7)0.018Abbreviations: AAPC, average annual percent change; ASR, age-standardised rate; CI, confidence interval; PBC, primary biliary cholangitis. Average annual percent changes in WHO age-standardised PBC prevalence and incidence rates in Brazil, 2021–2024. Abbreviations: AAPC, average annual percent change; ASR, age-standardised rate; CI, confidence interval; PBC, primary biliary cholangitis. Between 2021 and 2024, prevalence increased significantly across nearly all age bands, with the largest relative rise at 71–75 years (+25.8%, 95% CI: 22.3–29.5; p < 0.001); only ≥90 years showed a non-significant change (−2.4%, 95% CI: −18.0 to 16.3; p = 0.79) (Supplementary Table S9). In contrast, incidence declined across many middle-to-older age groups (e.g., −10.7% at 46–50, p < 0.001; −8.9% at 61–65 and 66–70, both p < 0.001), was largely stable at 71–75 (+0.8%; p = 0.80), and rose modestly at 26–30 (+5.8%; p = 0.036). The ≥90 group showed a marked decline (−35.6%; p < 0.001) (Supplementary Table S10). Socioeconomic correlates At the State level, Spearman correlations between the Human Development Index (HDI) and PBC metrics were small and non-significant for both incidence [Spearman's r = 0.21 (p = 0.340)] and prevalence [r = 0.33 (p = 0.122)] in 2021. The largest (yet still non-significant) correlations were observed for prevalence vs the education component of HDI (r = 0.39, 95% CI: −0.03 to 0.69; p = 0.063); correlations for overall HDI with prevalence and incidence were 0.33 (p = 0.122) and 0.21 (p = 0.340), respectively (Supplementary Table S11). On the other hand, HDI was positively associated with PBC prevalence across Brazilian macroregions in exploratory analyses: Pearson's r = 0.84 (95% CI: −0.15 to 0.99), p = 0.072. A simple linear model estimated β = 24.86 per 1.0 HDI (∼0.25 per 0.01 HDI; 95% CI: −4.12 to 53.85), R2 = 0.713 (Fig. 4A). HDI was also positively associated with PBC incidence across Brazilian macroregions: Pearson's r = 0.87 (95% CI: −0.04 to 0.99), p = 0.053. A simple linear model estimated β = 8.11 per 1.0 HDI (≈0.081 per 0.01 HDI; 95% CI: −0.17 to 16.39), R2 = 0.764 (Fig. 4B). Although these results are suggestive, they did not reach the 5% significance threshold, likely reflecting low power with five observations. They should therefore be interpreted cautiously and as hypothesis-generating only.Fig. 4Correlation between PBC prevalence (A) and incidence (B) by macroregions (2021) and 2010 HDI. Correlation between PBC prevalence (A) and incidence (B) by macroregions (2021) and 2010 HDI. Discussion In this first nationwide, population-based study of PBC in Latin America, and the largest conducted worldwide, we leveraged artificial intelligence to analyse health records from Brazil's public healthcare system, a nationwide administrative database comprising approximately 230 million linked health records. Our analysis reveals an epidemiological landscape that is both consistent with global trends and unique in its regional distribution. The major findings of this study are a substantial and growing prevalence of PBC, coupled with a relatively stable incidence, a narrower female-to-male ratio than historically described, and a south-north geographical gradient that could correlate with socioeconomic development and environmental factors. The temporal trends observed in Brazil align with the evolving epidemiology of PBC worldwide. The prevalence of PBC in Brazil demonstrated a significant average annual increase of 19.5%, while the incidence remained relatively stable, with a slight annual decrease of 4.4%. This pattern of rising prevalence with relatively stable incidence is well-documented in North America29 and Europe30,31 and is largely attributed to improved diagnostic awareness, leading to earlier detection, and longer patient survival due to effective therapies like UDCA. Although mortality and survival could not be formally evaluated in the present analysis, future studies incorporating validated linkage with mortality records and longer follow-up will be required to clarify the underlying drivers of the observed prevalence pattern. Moreover, the national prevalence and incidence rates of PBC in 2024, standardised to the WHO World Standard Population, were estimated in this study at 17.79 (95% CI: 17.47–18.10) and 3.41 (95% CI: 3.27–3.56) per 100,000, respectively. Our age-adjusted prevalence finding is consistent with the global estimate of 18.1 per 100,000 from a recent comprehensive meta-analysis.2 In contrast, the observed national age-adjusted incidence, while higher than the pooled global rate of 1.8 per 100,000, more closely aligns with the regional incidence for the Americas (2.7 per 100,000) reported in the same study.2 Our data also contribute to the evolving understanding of the sex distribution of PBC. Historically, the disease was considered to have an overwhelming female predominance, with female-to-male ratios as high as 9:1.32 In our study, we observed a significantly narrower ratio of 3.67:1 in 2024. This finding is consistent with recent large-scale, population-based studies from Italy33 (2.3:1), Denmark33 (4.2:1), and China7 (4:1). This global trend suggests that the classic higher ratios may have been, in part, an artifact of referral bias in smaller, single-center studies from previous eras. Consistent with this, in a retrospective cohort from referral liver centers in Brazil, which included patients diagnosed with PBC between 1992 and 2019, 95% of the patients were females (mean age 51 years).15 Improved case-finding methodologies in large databases and greater awareness of PBC in men are revealing a more balanced, though still female-predominant, sex distribution. It is also reasonable to hypothesise that men are being increasingly diagnosed in recent years. However, it is not possible to rule out that patients with other cholestatic liver diseases, such as primary sclerosing cholangitis and genetic cholestasis, which can affect men more frequently, are being erroneously diagnosed with PBC. Moreover, an important finding of our study was the pronounced geographical heterogeneity in PBC prevalence across Brazil, characterised by a clear south-north gradient. The observed south–north gradient is most plausibly explained by regional differences in healthcare access, specialist availability and diagnostic capacity.34 Although exploratory macroregional analyses suggested positive correlations between HDI and PBC incidence and prevalence, these associations were based on only five observations, had wide confidence intervals, and were not statistically significant. Moreover, state-level analyses using 2021 HDI estimates did not demonstrate significant associations with either incidence or prevalence. Therefore, the HDI findings should not be interpreted as evidence of a biological or causal relationship between socioeconomic development and PBC risk. Rather, they support the hypothesis that regional inequalities in access to diagnosis and treatment may contribute to the observed geographic variation. The southern regions of Brazil have a higher concentration of medical specialists and more developed health infrastructure,35 likely leading to better case ascertainment and consequently higher reported prevalence. The lower rates in the less developed northern regions may, therefore, signify a substantial burden of underdiagnosed disease. However, this gradient may also reflect the influence of underlying environmental or socioeconomic risk factors. Notably, our findings parallel those from the United Kingdom3 and from the recently published global meta-analysis,2 where a robust positive correlation between geographic latitude and PBC has been established. This shared pattern suggests that exposures varying with latitude, such as sunlight and vitamin D metabolism, which have been proposed in PBC previously,36,37 may be potential pathogenic contributors. Interestingly, a recent case–control study found that greater sun exposure and higher cumulative lifetime UV load, as estimated using a life-calendar approach, were associated with lower odds of PBC, which is consistent with our observations in Brazil.38 Furthermore, observational studies have identified smoking and the use of some cosmetic products as potential risk factors for PBC, while geographic clustering near water reservoirs, coal-mining areas, and waste-disposal sites has generated hypotheses regarding additional environmental triggers.39,40 However, none of these exposures were measured in the present study, and their potential contribution to the observed gradient remains speculative and hypothesis-generating. The unique ethnic admixture and pronounced environmental diversity across Brazil present a valuable opportunity to explore these complex interactions in future dedicated studies. Our study has limitations. The case definition, while enhancing specificity, may have introduced selection bias. By requiring ursodeoxycholic acid (UDCA) dispensation as a criterion, our methodology inherently excludes patients who are intolerant to UDCA, those who have not yet initiated therapy, or those who obtain the medication through private channels. Consequently, the reported figures likely underestimate the true incidence and prevalence of primary biliary cholangitis (PBC) in Brazil. In this study, 10,787 individuals were identified using ICD-10 code K74.3 alone, of whom 93.9% also had documented UDCA dispensation. Thus, application of the more specific combined case definition reduced the cohort by only 6.1%. Importantly, the diagnostic accuracy of an isolated K74.3 code in DATASUS has not been independently validated. Although removing the UDCA requirement would increase sensitivity, it would also introduce an unquantifiable risk of diagnostic misclassification, including the miscoding of other cholestatic liver diseases. The high proportion of patients with documented UDCA use is also consistent with a large Brazilian retrospective cohort, in which UDCA was prescribed to 96% of patients.41 Nevertheless, reliance on data from the public healthcare system may have resulted in incomplete case ascertainment, particularly among individuals managed exclusively in the private sector. Although many privately treated patients access SUS for publicly funded medications, incomplete capture of this population remains an important limitation and may have contributed to underestimation of the true disease burden. The relatively short observation period represents another important limitation. With only four annual observations, estimates of annual percentage change are statistically sensitive and cannot establish long-term temporal trends or reliably identify changes in trend direction. The HC0 robust variance estimator, while widely used for its robustness to distributional misspecification, is sensitive to small sample sizes and sparse data — both characteristics of our incidence time series for a rare disease. Although the direction and statistical significance of the incidence trend were consistent, the precision of the point estimate could not be quantified. Our findings may also have been influenced by increasing registry maturity and differences in the available lookback period. These estimates should therefore be interpreted as descriptive measures of recent short-term change. Additionally, COVID-19 pandemic may have affected healthcare-seeking behaviour, access to specialist assessment, diagnostic testing, and medication registration during the early study period. Extending the analysis to years preceding implementation of the 2019 national PBC protocol was not considered methodologically appropriate because earlier records lacked the standardised diagnostic and medication-dispensation framework used in the present case definition. Continued surveillance will be necessary to determine whether the observed patterns persist over a longer period. Individual-level biochemical, serological, histological, and imaging data were unavailable, and residual misclassification of other cholestatic diseases -particularly diseases that are more common among men, like primary sclerosing cholangitis -cannot be excluded may have contributed to the observed sex distribution. Nevertheless, the stringent national protocol for UDCA dispensation–which mandates diagnostic confirmation according to established clinical, serological, and histological criteria–mitigates the risk of misclassification and strengthens the internal validity of the identified cases. It is also important to highlight that the socioeconomic analyses were ecological and therefore susceptible to ecological fallacy. The macroregional correlations were particularly limited by the inclusion of only five observations and by the temporal mismatch between the 2010 macroregional HDI estimates and the 2021–2024 epidemiological data. Although state-level analyses using 2021 HDI data increased the number of observations, no statistically significant associations were identified. These findings should therefore be regarded as exploratory and hypothesis-generating. Finally, although race/color information is collected in several DATASUS subsystems, these data were not available in a sufficiently complete, harmonised, and validated form in the curated dataset used for this study. We were therefore unable to assess PBC incidence and prevalence according to race/color. In conclusion, this study provides a landmark contribution as the first comprehensive, population-based epidemiological analysis of PBC in Latin America, and to our knowledge, one of the largest population-based studies of PBC conducted to date. Our findings for Brazil reveal a rising prevalence with a relatively stable incidence, a narrowing sex gap, and significant south-north macroregional disparity. By providing this crucial foundation for public health policy, our work highlights the urgent need for increased awareness and improved access to diagnosis and care, particularly in the northern and northeastern regions. Future research should be conducted on the high-prevalence clusters identified in this study to investigate specific environmental triggers and genetic predispositions in Brazil's uniquely admixed population. Contributors GGLC: Conceptualisation, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Writing—original draft, and Writing—review & editing. CAC: Conceptualisation, Investigation, Methodology, Supervision, and Writing—review & editing. LCF: Conceptualisation, Investigation, Supervision and Writing—review & editing. BFM: Investigation, Methodology, Data curation, and Writing—original draft. GAPDL: Investigation, Methodology, Data curation, and Writing—original draft. PLB: Conceptualisation and Writing—review & editing. CAVN: Conceptualisation and Writing—review & editing. VR: Conceptualisation and Writing—review & editing. DRBT: Conceptualisation and Writing—review & editing. ELRC: Conceptualisation and Writing—review & editing. MLF: Conceptualisation, Funding acquisition, Project administration, Data curation, Methodology, Supervision, and Writing—review & editing. All authors read and approved the final version of the manuscript; all authors had access to the data as appropriate; all authors had final responsibility for the decision to submit the manuscript for publication. Data sharing statement The data underlying this study are available from the corresponding author upon reasonable request. Declaration of interests Guilherme Grossi Lopes Cançado has received a research grant from IPSEN, payments for lectures from IPSEN, Takeda, Johnson and Johnson and Biolab, and support for attending meetings from Takeda, Novo Nordisk, and Johnson and Johnson; Cláudia Alves Couto has received research grants from IPSEN, Mirum, Inventiva and Gilead, payments for lectures from IPSEN, Gilead, Roche and Novo Nordisk and support for attending meetings from Gilead; Luciana Costa Faria has received a research grant from Mirum: Bernardo de Faria Moraes: none; Gustavo André Pedral Diniz Leite: none; Paulo Lisboa Bittencourt has received a research grant from Mirum and payments for lectures from IPSEN; Cristiane Alves Vilella-Nogueira has received research grants from Novo Nordisk, Boehringer Ingelheim, Bio 89, payments for lectures from Novo Nordisk, Boehringer Ingelheim and Siemens, and has participated in advisory boards from Novo Nordisk and Lilly; Vivian Rotman: none; Debora Raquel Benedita Terrabuio: has received a research grant from IPSEN; Eduardo Luiz Rachid Cançado: has received a research grant and support for attending meetings from IPSEN; Maria Lúcia Ferraz: none.