Abstract
Background: Rural and remote populations often engage less than their metropolitan counterparts in colorectal cancer (CRC) prevention practices, including CRC screening with faecal immunochemical testing (FIT). However, the evidence is not comprehensive and consistent. Therefore, we performed a systematic review and meta-analysis of geographical disparities in FIT-based CRC screening participation.
Methods: We searched six databases and considered articles through June 2024. We included participation rates in FIT-based CRC screening based on geography (rurality, remoteness, and population density), regardless of research design and study period. Title, abstract, and full-text screening and data extraction were conducted independently by at least two reviewers. The number of people screened and invited was extracted. A random-effects model was used to pool the rates and odds ratio.
Results: Thirty-five articles reported FIT-based CRC screening participation by geography. Overall, 40% of articles were from Europe, 31% from Australia, 20% from Asia, and 9% from the USA. The pooled participation rate was 49.9% (95% confidence interval (CI); 40.6-59.2), with 62% in Europe, 48% in the USA, 45% in Australia, and 36% in Asia. In Europe, individuals in less dense areas had higher participation odds than those in higher dense areas (pooled odds ratio (POR) 1.31; 95% CI:1.04–1.65), while in Australia, remote areas (POR = 0.79; 95% CI: 0.68-0.79) had significantly lower odds compared to metropolitan areas. No significant difference in participation rates was found between rural and urban areas in Asia (POR: 0.91; 95% CI: 0.66–1.25). Only eight articles reported the geographical disparity using spatial analysis.
Conclusion: A notable disparity in screening participation rate was observed based on population density and remoteness. Further advanced geospatial research is needed to identify the socioeconomic, healthcare access, and policy factors driving these disparities and develop targeted strategies to improve screening rates and address access barriers for underserved populations.

Background
Testicular cancer is the commonest form of invasive cancer in young men globally. There is increasing incidence in developed countries and despite high survival, survivors experience lasting health and psychosocial impacts. Geographical factors have previously been suggested to influence the incidence rates of testicular cancer, with spatial variation reported in some countries. However, its geographical distribution is not known in Oceania.
Methods
Using national population-based cancer registry data, this study investigated the spatial patterns of the incidence rates of testicular cancer across Australia at the small area-level (SA2s). Incidence data including residential location at diagnosis were obtained from the Australian Institute of Health and Welfare, with mortality followed-up until end of 2019. Bayesian spatial incidence models were fitted to diagnosis data and 5-year relative survival by broad geographical classifications was modelled using flexible parametric models.
Results
From all the notifications of testicular cancer (n = 8,217), the age-standardized incidence rate was 8.9 cases per 100,000 males each year. There was evidence of significant spatial variation in incidence across small geographical areas (P < 0.001). There was strong evidence for some areas, particularly in Tasmania, having incidence rates above the national average. We found a strong association between high incidence rates and both remoteness (P < 0.001) and area disadvantage (P < 0.001). The 5-year relative survival estimate was 97.5% [95% CI: 97.1–97.9] with no geographic variation for survival rates.
Conclusion
This study provides a basis for further research into drivers of localised spatial patterns which in turn could inform the development of any future evidence-based interventions.

Background
Cancer survival rates differ substantially across Australia. The excess risk of death in some areas is double the national average, using 5-year relative survival modelling. We estimated the number of cancer-related deaths that could be avoided if spatial disparities in cancer survival were reduced.

Methods
Population-based cancer diagnosis and survival records were obtained from the Australian Institute of Health and Welfare, and area-level population mortality data from the Registries of Births, Deaths and Marriages. Bayesian spatial relative survival models were fitted to observations for all people diagnosed with a malignant neoplasm no more than 5 years previously during 2010-2019.

A benchmark ‘optimal survival rate’ was defined as that of the area where 5-year relative survival was better than 80% of all areas. The number of avoidable cancer-related deaths was calculated by comparing the modelled number of cancer-related deaths within 5 years of diagnosis with the number expected if all areas met the optimal survival benchmark.

Results
In 2010-2019, 33 892 (11.7%) cancer-related deaths in Australia could have been avoided if spatial disparities in cancer survival were reduced. Of the cancer-related deaths in the most disadvantaged areas, 20% (13 469 deaths) were avoidable compared with 2% (1020) in the least disadvantaged areas. Of the cancer-related deaths in remote areas, 30% (1569) were avoidable compared with 8% (15 087) in areas classified as major city.

Conclusion
More than 1 in 10 of the cancer-related deaths could be avoided by reducing the spatial disparities in cancer survival in Australia and this ratio was substantially higher in disadvantaged and regional or remote areas. This highlights the need for increased efforts to address disparities in cancer outcomes.

Introduction: Atrial fibrillation (AF) affects over 60 million people worldwide. Early detection is crucial, as AF increases the risk of stroke, heart failure, and cardiovascular complications. Traditional AF risk models rely on clinical factors, but incorporating polygenic risk scores (PRS) may improve prediction. This study evaluates the performance of a recently developed AF PRS in healthy older adults.

Methods: We generated an AF PRS from Miyazawa et al. (2023) and analysed baseline data from 12,973 ASPREE participants aged ≥65 years (n=720 AF cases over a mean follow-up of 4.4 years). AF was identified using a probabilistic algorithm incorporating AF medication and clinical evidence after identifying trigger factors suggesting possible AF. Cox proportional hazard models assessed PRS predictive ability alone and with existing clinical risk scores (CHARGE-AF, HARMS2-AF). PRS was standardized to interpret adjusted hazard ratios (aHR) per standard deviation (SD) change. Models included continuous PRS and risk quintiles. Sensitivity analysis excluded participants with myocardial infarction (MI) or heart failure (HF) before AF. Sex-stratified analyses were also conducted.

Results: The AF PRS showed stronger linear predictive ability (aHR:1.56, 95%CI:1.45-1.68) than CHARGE-AF (aHR:1.54, 95%CI:1.43-1.66) and HARMS2-AF (aHR:1.32, 95%CI:1.21-1.44) (all p<0.0001). Individuals in the highest PRS risk quintile had more than 4-fold increased AF risk (aHR:4.23, 95%CI:3.16-5.65, p<0.0001) compared to the lowest quintile. The PRS also improved risk prediction, increasing the C-index by 5.8% above clinical risk factors alone (0.629→0.687), 6.6% above CHARGE-AF (0.614→0.680), and 7.1% above HARMS2-AF (0.573→0.644) (all p<0.0001). Sensitivity analysis excluding prior MI/HF slightly improved prediction (aHR: 1.61 vs 1.57, p<0.0001). The PRS was more predictive in women than men (HR:1.69 vs. 1.46, p<0.0001), with high-risk women 4.87 times more likely to develop AF than women at low genetic risk.

Conclusion: Integrating PRS to AF risk models improves prediction and may help enable earlier, targeted prevention, reducing AF-related complications.

Background: Beginning in 1985 as the Australian Schools Health and Fitness Survey (ASHFS), the Childhood Determinants of Adult Health (CDAH) study aims to examine the childhood factors associated with adult cardiovascular disease. This presentation will provide an overview of the study including major contributions to the epidemiology of cardiovascular disease.
Methods: Participants were from the ASHFS, a nationally representative study of Australian school children aged 7 to 15 years. Measures in ASFHS included cardiorespiratory and muscular fitness, anthropometry, blood pressure, lipids, health behaviours and attitudes, psychosocial wellbeing, and socioeconomic factors. In the early 2000s, we traced participants to create a cohort with adult cardiovascular health measures. Following funding from the NHMRC, Heart Foundation and philanthropic partners, the first follow-up (CDAH-1) was completed in 2004-06 including similar measures to 1985 plus carotid ultrasound. A follow-up in 2009-2011 (CDAH-2) focused on life stage transitions with questionnaire-based assessments only, while the most recent follow-up (2014-19) repeated baseline physical and questionnaire measures with carotid ultrasound and echocardiography.
Results: From 8,498 participants, 6,838 were traced and 3,998 participated in CDAH-1; while 3,038 participated in CDAH-2 and 3,142 participated in CDAH-3. The study has made major contributions in areas of overweight and obesity, cardiorespiratory and muscular fitness, physical activity and smoking from childhood to adulthood. Analyses have established the trajectories of these risk factors, the factors associated with trajectories and associations between trajectories and a range of measures of cardiometabolic structure and function across adulthood. The study has been supported by $6.5 million of competitive grant funding, produced 200+ publications, contributed to 17 PhDs and 11 fellowships.
Conclusion: The CDAH study is a unique Australian cardiovascular health cohort that has made internationally significant contributions to understanding the childhood influences on adult cardiovascular health.

Background:
Oral health is an essential component of overall well-being, yet oral diseases such as dental caries and periodontal disease are among the most prevalent and costly to treat. Poor oral health disproportionately affects vulnerable populations, including those with chronic conditions such as diabetes. Diabetes can worsen oral health directly via disease mechanisms, or indirectly through disease-related lifestyle changes, while tooth loss has been linked to an increased risk of mortality. Despite these associations, limited longitudinal research has examined the impact of tooth loss on mortality in individuals with diabetes.

Methods:
We conducted a record linkage study using baseline data from the 45 and Up study (n = 267,153) linked with Pharmaceutical Benefits Scheme (PBS) and deaths data (Centre for Health Record Linkage). A total of 22, 191 participants with diabetes were identified at baseline and followed for up to 10 years. Cox proportional hazards regression models were used to estimate the association between tooth loss and all-cause mortality.

Results:
In preliminary crude analyses, participants with no teeth had a 3.2 times higher risk of mortality (HR = 3.20, 95% CI [2.97–3.45]) compared to those with ≥20 teeth. Mortality risk was also higher for those with 1–9 teeth (HR = 2.50, 95% CI [2.34–2.71]), and 10–19 teeth (HR = 3.20, 95% CI [1.68–1.93]). After adjusting for covariates, the associations weakened but remained significant for those with no teeth (HR = 1.25, 95% CI [1.15–1.36]) and 1–9 teeth (HR = 1.11, 95% CI [1.03–1.21]), while having 10–19 teeth was no longer significant.

Conclusion:
Our research found that there is an association between tooth loss and mortality in individuals with diabetes, particularly among those with severe or complete tooth loss.

Background: Older adults face a high risk of falls, which can greatly influence their quality of life (QoL). While the association between falls and QoL has been studied extensively in many countries, Australia lacks comprehensive research. Therefore, this study aimed to assess the association between falls and quality of life among older adults.
Methods: Participants (n=530, age ≥65yr) were from the 15-year follow-ups for men and women enrolled in the Geelong Osteoporosis Study. The Australian version of the World Health Organisation Quality of Life-BREF (WHOQOL-BREF), a 26-item questionnaire, assessed QoL across four domains: physical health, psychological health, social relationships, and environmental health. Falls over the previous 12 months were self-reported. Tobit regression analyzed the association between falls and QoL.
Results: Single falls (β = -2.35, 95% CI: -4.18, -0.52) and recurrent falls (β = -7.53, 95% CI: -10.7, -4.32) were significantly associated with lower overall QoL. Recurrent falls related to all domains: physical (β = -5.94, 95% CI: -10.8, -1.22), psychological (β = -7.92, 95% CI: -13.7, -2.12), social (β = -10.7, 95% CI: -18.4, -3.02), and environmental (β = -5.79, 95% CI: -11.3, -0.32). Single falls are mainly associated with physical (β = -2.90, 95% CI: -5.48, -0.32) and psychological domains (β = -3.20, 95% CI: -5.59, -0.80).
This study highlights the strong association between falls and QoL in older Australians. Recurrent falls significantly impacted all QoL domains, while single falls mainly affected physical and psychological health. These findings underscore the need for targeted fall prevention strategies, including balance training, psychological support, and environmental modifications, to enhance QoL and well-being in ageing populations.

Background:
Endometrial cancer (EC) incidence in Australia has doubled over the past two decades, in parallel with rising obesity prevalence. While recreational physical activity (PA) reduces EC risk, its interaction with body mass index (BMI) remains unclear. Data are scarce on whether the association between PA and EC risk varies by age, histotype or BMI.
Methods:
We conducted a meta-analysis with individual-level data from 19 studies (6 cohort, 13 case-control) in the Epidemiology of Endometrial Cancer Consortium, comprising 11,864 cases and 26,513 controls. We classified women as sufficiently active (≥150-minutes moderate or 75-minutes vigorous activity/week) or inactive. Odds ratios (OR) and 95% confidence intervals (CI) were estimated using mixed-effects logistic regression, considering study as a random effect and adjusting for potential confounders. We used a propensity-score (PS) approach to reduce the risk of potential confounding.
Results:
Preliminary analyses show that, overall, 55% cases and 51% controls were inactive and PA was associated with an 8% reduction in EC risk. This association was stronger for endometrioid EC (OR=0.87, 95%CI=0.81-0.94), and was similar in case-control and cohort studies. However, PA was not associated with non-endometrioid EC. The risk reduction for endometrioid EC was stronger among postmenopausal women (13%) compared to premenopausal women (6%), though not statistically significant. Being active was associated with a 14%-18% lower endometrioid EC risk among overweight and obese women, but only 7% among normal weight women. Findings remained robust in the PS analysis. A joint analysis of PA and BMI will be conducted to examine whether PA modifies the obesity-related risk of EC.
Conclusion:
Recreational PA is associated with reduced risk of endometrioid EC, particularly among overweight, obese and postmenopausal women. However, it is not associated with non-endometrioid EC risk. Supporting recreational PA at recommended levels may help lower the risk of endometrioid EC, particularly in high-risk groups.

Aim: We investigated the prevalence of diabetes among different groups of migrants in Australia and their sociodemographic characteristics to guide health planning.
Methods: We used whole population data from 2021 Australian census and analysed data from people aged ≥30. Age-standardised prevalence (ASP) and age-standardised prevalence ratio (ASPR) of diabetes were calculated. Among those with diabetes, we reported sociodemographic characteristics and comorbidities.
Results: 41% of the population with diabetes were migrants. Of those, 40% were born in Europe. The prevalence of diabetes ranged from 5.3% (migrants from Northeast Asia) to 14.4% (migrants from Oceania) and was higher among males and people aged 70-85. ASP was higher than Australian-born (ASPR:1.4-2.2) among migrants from Southeast Asia, North Africa and Middle East, Southern and Central Asia, and Oceania.
Among migrants with diabetes, a range of 10.4% (migrants from Northeast Asia) to 32% (migrants from North-West Europe) had ≥3 long-term health conditions. Percentage of people with <10 years ago arrival ranged from 0.6% (migrants from Southern and Eastern Europe) to 14.9% (migrants from Northeast Asia). The highest percentage of low English proficiency was observed among migrants from Northeast Asia (55.2%). A range of 0.4% (migrants from North-West Europe) to 13.3% (migrants from North Africa and Middle East) did not go to school, and 38.5% (migrants from Sub-Sahara Africa) to 66.2% (migrants from Southern and Eastern Europe) had <$500 weekly income. Among groups with ASPR >1 who had low English language proficiency, a range of 8% (born in Oceania) to 29% (born in North Africa and Middle East) had <$500 weekly income, and 2% (born in Oceania) to 10% (born in North Africa and Middle East) did not go to school.
Conclusion: The prevalence of diabetes and sociodemographic characteristics of those with diabetes vary among migrant groups. Considering intersectionality can improve diabetes-related outcomes among migrants.

Background
The impact of household air pollution (HAP) on obstructive sleep apnoea (OSA) was unclear from the literature. We aimed to investigate the associations between HAP exposure over 10 years and OSA in middle-aged adults.
Methods
Information on household heating, cooking, mould, active and passive smoking exposure from Tasmanian Longitudinal Health Study (TAHS) participants at 43 and 53 years was previously characterised into seven longitudinal HAP profiles. Probable OSA at 53 years was defined using validated STOP-Bang, Berlin and OSA-50 questionnaires, respectively. Medically diagnosed OSA was self-reported. Multivariable logistic regression models were used to assess the associations between HAP profiles and each definition of OSA, adjusting for age, sex, socioeconomic status and ambient air pollution.
Results
Compared with the “Least exposed” profile, characterised by reverse-cycle air conditioning, electric cooking and no smoking exposure, the “Wood and gas heating/cooking/smoking” profile was associated with both probable OSA defined using OSA-50 questionnaire (aOR=2.39, 95%CI 1.61-3.53) and medically diagnosed OSA (aOR=2.31, 1.06-5.05). The “All gas” and “Wood heating/smoking” profiles were associated with OSA-50-defined probable OSA (aOR=1.35, 1.01-1.79; aOR=1.47, 1.10-1.96 respectively). There was a modest association between the “All gas” profile and medically diagnosed OSA. Results were less certain when STOP-Bang or Berlin questionnaires were applied.
Conclusions
Sustained exposure to gas and wood heating and gas cooking, especially when combined with tobacco smoke, over ten years increased the risk of OSA in middle age. Our study strengthens the rationale for including the potential adverse effect of HAP on mid-life OSA within public educational programs and guidelines.

Background
Use of long-acting, reversible contraceptives has increased over the past 20 years, but an understanding of how they could influence cancer risk is limited.

Methods
We conducted a nested case-control study among a national cohort of Australian women (n=176,601 diagnosed with cancer from 2004-2013; 882,999 control individuals matched by age, state, Socio-Economic Indexes for Areas quintile) to investigate the associations between the levonorgestrel intrauterine system (LNG-IUS), etonogestrel implants (ENG-IMP), depot-medroxyprogesterone acetate (DMPA) and cancer risk and compared these results with the oral contraceptive pill (OCP). We used conditional logistic regression to estimate odds ratios (OR) and 95% confidence intervals (CI), adjusting for use of OCP.

Results
LNG-IUS and ENG-IMP use was associated with breast cancer risk (OR=1.26, 95%CI=1.21-1.31, and OR=1.24, 95%CI=1.17-1.32, respectively), but DMPA was not, except when used for 5 or more years (OR=1.23, 95%CI=0.95-1.59). Reduced risks were seen for LNG-IUS (≥1 years of use) in endometrial cancer (OR=0.80, 95%CI = 0.65-0.99), ovarian cancer (OR=0.71, 95%CI=0.57-0.88), and cervical cancer (OR=0.62, 95%CI=0.51-0.75); for ENG-IMP in endometrial cancer (OR=0.21, 95%CI=0.13-0.34) and ovarian cancer (OR=0.76, 95%CI=0.57-1.02); and for DMPA in endometrial cancer (OR=0.21, 95%CI=0.13-0.34). Although LNG-IUS, ENG-IMP and DMPA were all associated with increased cancer risk overall, for ENG-IMP, the risk returned to baseline after cessation, similar to the oral contraceptive pill. We were unable to adjust for all potential confounders, but sensitivity analyses suggested that adjusting for parity, smoking, and obesity would not have materially changed our findings.

Conclusion
Long-acting, reversible contraceptives have similar cancer associations to the oral contraceptive pill (reduced endometrial and ovarian cancer risks and short-term increased breast cancer risk). This information may be helpful to women and their physicians when discussing contraception options.

Background: Most Colorectal cancer (CRC) develops from pre-cancerous lesions, which enables early intervention through identifying and removing these lesions at colonoscopy. Risk stratification helps to prioritise individuals who require colonoscopy surveillance. This review summarised the available evidence on risk-predictive models for advanced colorectal neoplasia in individuals undergoing colonoscopy-based surveillance.
Methods: Searches were conducted in MEDLINE, Cochrane libraries, SCOPUS, Web of Science and CINAHL up to 19th Jan 2024. Risk of bias and Applicability was assessed using the PROBAST. The discriminatory ability of the risk predictive models was assessed using the C-statistic or area under the curve (AUC). The studies were evaluated based on the TRIPOD guidelines.
Results: Of 11,281 articles screened, 22 studies were included. Most studies were rated as high risk of bias (57%) and had a high concern on applicability (53%). Regarding the variables in the models, two studies included only conventional risk factors (e.g. demographics), and 20 studies included composite variables incorporating conventional risk factors (e.g. clinical, biomarker or genetic information). Only five (5/22) studies employed machine learning (ML) methods while 17/22 used traditional statistical models (e.g. logistic regression). Good model discrimination (AUC>0.7) was reported for 16/22 models. Prediction models that applied ML methods had excellent discriminatory ability for colorectal neoplasia (AUC>0.8) and performed better than traditional statistical models. Many studies did not follow the TRIPOD guidelines properly, often failing to explain how they handled missing data and lacking measures of model performance. Ten models underwent internal validation while only four underwent internal and external validation.
Conclusion: Risk prediction models with ML methods show promise for accurately identifying those at the highest risk of colorectal neoplasia. New risk prediction models using routinely collected clinical data adhering to TRIPOD guidelines for ML methods and evaluating in external appropriate populations are required before adoption in clinical practice.

Background
People with Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (pwME/CFS) experience debilitating symptoms that substantially compromise their ability to participate in typical daily activities from employment to self-care. However, the validity of ME/CFS as a real, disabling and permanent chronic illness has not yet been recognised in Australian healthcare policy. Consequently, many pwME/CFS are deemed ineligible to access vital disability and social support services. The present study aimed to foreground the extensive and permanent impacts of ME/CFS to warrant policy change and improve service accessibility.
Methods
This longitudinal investigation was conducted between October 2021 and October 2024. Eligible participants included Australians with ME/CFS who were aged between 18 and 65 years and fulfilled the Canadian or International Consensus Criteria at baseline (T0). Three validated, self-administered questionnaires distributed at approximately six-month intervals captured participants’ sociodemographic information, symptoms and health-related quality of life (HRQoL). Participants’ HRQoL was compared with population norms at each time point.
Results
A median of at least 30 symptoms were experienced at each time point by the n=32 pwME/CFS (females, 68.8%; T0 age, x̄=44.03 years; T0 illness duration, median=12.50 years). Median symptom severity was at least moderate for most neurocognitive, pain, sleep, neurosensory and autonomic symptoms. All HRQoL domains were significantly poorer among pwME/CFS than the general population. The HRQoL domains quantifying overall health status, physical health and the ability to participate in daily and work life activities were the most substantially impacted relative to other domains and when compared with population norms. Importantly, no significant differences in any measure of symptom presentation or HRQoL were observed throughout the study.
Conclusion
ME/CFS is a long-term, physically disabling illness. This must be recognised by Australian healthcare policies to facilitate access to essential disability and social support services, which will be vital in optimising health and wellbeing for all pwME/CFS in Australia.

Background: Polygenic scores (PGS) for type 2 diabetes (T2D) have demonstrated predictive utility in the general adult population. However, their performance in individuals aged ≥70 years remains unclear, despite the high prevalence of T2D in this age group. This study evaluated the predictive performance of a PGS for T2D in Australian individuals aged ≥70 years.

Methods: We analyzed 12,174 unrelated participants aged ≥70 years from the Aspirin in Reducing Events in the Elderly (ASPREE) trial. Recruitment occurred from March 2010 to December 2014, with a median follow-up of 4.6 years. T2D was defined by self-report, glucose-lowering medication use, or fasting plasma glucose ≥7.0 mmol/L. Logistic and Cox regression models assessed associations between a recently derived PGS using SBayesRC approach and T2D risk at baseline and during follow-up, adjusting for clinical risk factors from the Australian Diabetes Risk Assessment Tool (AUSDRISK). Predictive accuracy was evaluated using the area under the curve (AUC) for baseline T2D and the C-index for incident T2D.

Results: At baseline, 1,150 participants (9.4%) had T2D, and 590 (4.8%) developed T2D during follow-up. Each standard deviation increase in the PGS was associated with greater odds of baseline T2D (Odds Ratio: 2.39 [95% CI: 2.19-2.61]) and higher risk of incident T2D (Hazard Ratio: 1.55 [1.40-1.71]). Adding the PGS improved prediction over clinical risk factors alone, increasing the AUC from 0.70 to 0.79 and the C-index from 0.67 to 0.71 (both P<0.001). Net reclassification improvement (NRI) analysis showed that adding the PGS to a clinical risk model reclassified 63% of individuals to more accurate risk categories (case or non-case) for baseline T2D (NRI=0.63 [0.57-0.71]) and 42% for incident T2D within five years (NRI=0.42 [0.30-0.51]).

Conclusion: The PGS significantly improves T2D risk prediction in older adults, supporting its potential as a clinical tool for risk stratification in this population.

Background
Non-communicable diseases (NCDs) are leading causes of morbidity and mortality globally. As countries develop economically, the burden of NCDs often shifts from the rich to the poor. China, undergoing rapid epidemiological and socio-demographic transitions, provides a promising setting to explore these trends. This study examines who bears the NCD burden in China with the economic development in China, focusing on 1) NCD distribution over time, 2) the relationship between socioeconomic status (SES) and NCDs, and 3) differences in distribution by demographics (gender, age, residency).

Methods & Results
Project 1: Using 2015 China Health and Retirement Longitudinal Study (CHARLS) data, we examined the distribution of NCDs (diabetes, hypertension, dyslipidemia) and SES among 19,776 participants aged 45 and older. Logistic regressions revealed a reversal in NCD distribution by SES, with higher burdens observed in lower SES groups.
Project 2: Analysing 15 waves of China Health and Nutrition Survey (CHNS, 1991-2015) and China Family Panel Studies (CFPS, 2010-2020), we explored the SES-NCD gradient and gender differences. Among 43,087 participants, we found that China experienced a reversal in the SES-NCD gradient, particularly among women, with complex gender-specific explanations, by using the multivariable logistic regression and Blinder-Oaxaca decomposition models.
Project 3: Using CFPS data from 2010 and 2020, we examined the influence of childhood and adulthood SES on NCD prevalence. Analysis of 42,023 participants showed a significant relationship between childhood SES and NCDs, with stronger associations in older cohorts and a reversal in younger cohorts from the multivariable logistic regression models.

Conclusion
NCDs are increasingly affecting the poor in China, contributing to rising health inequality. SES and gender differences play crucial roles in shaping the burden.