Spotlight poster session - Sunday Group 1
Tracks
Track 1
| Sunday, November 23, 2025 |
| 13:00 - 13:55 |
Speaker
Xiaowei Chen
Peking University
Effect of ARB on the development of chronic diabetic complications: TTE study
Abstract
Introduction: RCTs suggest that angiotensin-receptor blockers (ARB) may limit the onset and the progression of nephropathy and retinopathy in patients with type 2 diabetes (T2DM). However, high-dose Olmesartan may be linked to increased cardiovascular risk. Real-world evidence remains limited, and results are inconsistent.
Aims: This study aimed to conduct a target trial emulation study to evaluate the effect of ARB initiation on chronic complications in T2DM patients with hypertension.
Methods: We used a Chinese regional health care database OMOP CDM to emulate a sequence of target trials including T2DM patients with hypertension who initiated ARB or not between 2012 and 2022. The effects of ARB initiation on new onset of diabetic retinopathy (DR), diabetic peripheral neuropathy (DPN), and cardiovascular mortality and morbidity were analysed in both intention-to-treat (ITT) and per-protocol (PP) approaches. A time-discrete dataset was constructed by month for each eligible person-trial, where the marginal structural model was adopted to estimate the causal effect accounting for time-varying variables. Pooled logistic regression approximated hazard ratios (HR).
Results: A total of 8,682 person-trial initiators (8,682 individuals) and 941,073 person-trial non-initiators (19,987 individuals) were included. In ITT analysis, ARB initiation was associated with a reduced risk of DR (adjusted HR: 0.81; 95% CI: 0.51–1.27) and cardiovascular death (adjusted HR: 0.97; 95% CI: 0.79–1.19), though not statistically significant. No association was found for DPN (adjusted HR: 1.27; 95% CI: 0.94–1.72). However, a significantly increased risk of cardiovascular morbidity was observed (adjusted HR: 1.32; 95% CI: 1.26–1.39). PP analysis results were consistent.
Conclusions: ARB initiation may offer modest protection against DR and cardiovascular mortality in T2DM patients with hypertension, but these associations were not statistically significant. Importantly, ARB use was associated with higher cardiovascular morbidity risk, highlighting the need for vigilant CV monitoring in clinical practice.
Keywords: ARB; T2DM complication; Target trial emulation
Aims: This study aimed to conduct a target trial emulation study to evaluate the effect of ARB initiation on chronic complications in T2DM patients with hypertension.
Methods: We used a Chinese regional health care database OMOP CDM to emulate a sequence of target trials including T2DM patients with hypertension who initiated ARB or not between 2012 and 2022. The effects of ARB initiation on new onset of diabetic retinopathy (DR), diabetic peripheral neuropathy (DPN), and cardiovascular mortality and morbidity were analysed in both intention-to-treat (ITT) and per-protocol (PP) approaches. A time-discrete dataset was constructed by month for each eligible person-trial, where the marginal structural model was adopted to estimate the causal effect accounting for time-varying variables. Pooled logistic regression approximated hazard ratios (HR).
Results: A total of 8,682 person-trial initiators (8,682 individuals) and 941,073 person-trial non-initiators (19,987 individuals) were included. In ITT analysis, ARB initiation was associated with a reduced risk of DR (adjusted HR: 0.81; 95% CI: 0.51–1.27) and cardiovascular death (adjusted HR: 0.97; 95% CI: 0.79–1.19), though not statistically significant. No association was found for DPN (adjusted HR: 1.27; 95% CI: 0.94–1.72). However, a significantly increased risk of cardiovascular morbidity was observed (adjusted HR: 1.32; 95% CI: 1.26–1.39). PP analysis results were consistent.
Conclusions: ARB initiation may offer modest protection against DR and cardiovascular mortality in T2DM patients with hypertension, but these associations were not statistically significant. Importantly, ARB use was associated with higher cardiovascular morbidity risk, highlighting the need for vigilant CV monitoring in clinical practice.
Keywords: ARB; T2DM complication; Target trial emulation
Biography
Xiaowei Chen is an MD-PhD Continuum student in Epidemiology and Biostatistics, School of Public Health, Peking University, with research interests in real-world study and pharmacoepidemiology. She graduated with a B.S. in Preventive Medicine from Tianjin Medical University. Xiaowei has participated in multiple observational and target trial emulation studies involving chronic diseases such as type 2 diabetes and COPD. She has published eight peer-reviewed papers, including first-author publications in high-impact journals such as J Natl Cancer Cent and Arthritis Care Res. Her work focuses on applying rigorous epidemiologic and statistical methods to generate real-world evidence for post-marketing evaluation of effectiveness and safety of drugs.
Dr Hsiu-Ting Chien
Harvard Medical School
A Machine Learning-Based Predictive Model for Assessing QTc Prolongation in Hospitalized Patients
Abstract
Introduction: QTc prolongation is a significant concern in hospitalized patients due to its association with life-threatening arrhythmias. Accurate risk prediction is essential for early intervention. However, existing QTc risk models often lack generalizability across diverse populations and clinical settings.
Aims: This study aimed to develop and validate a machine learning model that integrates both structured and unstructured EHR data to predict QTc prolongation and derive a simplified risk score (QTc Risk Index for Hospitalized Patients, QTRISK-H) for clinical use.
Methods: We conducted a nested case-control study using data from a tertiary hospital, including patients aged ≥45 years with at least two electrocardiograms (ECGs) recorded on different days during hospitalization. Structured data included demographics, comorbidities, medications, QTc values, and laboratory results . Unstructured data from ECG and echocardiography reports were extracted using rule-based natural language processing. Six machine learning models were developed , and performance was assessed using the area under the receiver operating characteristic curve (AUROC) and Brier scores.
Results: A total of 17,081 patients were included in the final analytic cohort, with 13,426 assigned to the training/testing set and 3,655 to the temporal validation set. Among the testing set, the extreme gradient boosting (XGBoost) model achieved the highest AUROC (0.772). QTRISK-H was developed using a logistic regression model, which achieved comparable performance (AUROC 0.758). Key predictors included hypokalemia, number of QT-prolonging medications, myocardial ischemia, and intensive care unit stay. QTRISK-H outperformed existing risk scores in temporal validation.
Conclusion: The XGBoost model demonstrated the best performance, and QTRISK-H offers a clinically practical and interpretable tool for predicting QTc prolongation. Nevertheless, external validation in independent cohorts is warranted to confirm whether it can support risk-informed monitoring and prescribing decisions, thereby helping to prevent QTc prolongation and reduce the risk of serious arrhythmias.
Keywords: QTc prolongation, predicted model, machine learning
Aims: This study aimed to develop and validate a machine learning model that integrates both structured and unstructured EHR data to predict QTc prolongation and derive a simplified risk score (QTc Risk Index for Hospitalized Patients, QTRISK-H) for clinical use.
Methods: We conducted a nested case-control study using data from a tertiary hospital, including patients aged ≥45 years with at least two electrocardiograms (ECGs) recorded on different days during hospitalization. Structured data included demographics, comorbidities, medications, QTc values, and laboratory results . Unstructured data from ECG and echocardiography reports were extracted using rule-based natural language processing. Six machine learning models were developed , and performance was assessed using the area under the receiver operating characteristic curve (AUROC) and Brier scores.
Results: A total of 17,081 patients were included in the final analytic cohort, with 13,426 assigned to the training/testing set and 3,655 to the temporal validation set. Among the testing set, the extreme gradient boosting (XGBoost) model achieved the highest AUROC (0.772). QTRISK-H was developed using a logistic regression model, which achieved comparable performance (AUROC 0.758). Key predictors included hypokalemia, number of QT-prolonging medications, myocardial ischemia, and intensive care unit stay. QTRISK-H outperformed existing risk scores in temporal validation.
Conclusion: The XGBoost model demonstrated the best performance, and QTRISK-H offers a clinically practical and interpretable tool for predicting QTc prolongation. Nevertheless, external validation in independent cohorts is warranted to confirm whether it can support risk-informed monitoring and prescribing decisions, thereby helping to prevent QTc prolongation and reduce the risk of serious arrhythmias.
Keywords: QTc prolongation, predicted model, machine learning
Biography
Hsiu-Ting (Noel) Chien is trained in pharmacoepidemiology and clinical pharmacy, with research interests in drug safety, drug-drug interactions, signal detection, and the generation of real-world evidence using large-scale healthcare databases. She recently earned her Ph.D. in Clinical Pharmacy from National Taiwan University and previously worked as a clinical pharmacist in a tertiary medical center. She first joined the Department of Population Medicine at Harvard Medical School as a visiting scholar and is currently continuing her work there as a postdoctoral fellow, focusing on real-world data analytics and pharmacoepidemiology research.
Ms Shin Yeong Kim
Graduate Student
Jeonbuk National University
Depression in rheumatoid arthritis patients initiating JAKi versus TNFi: Target Trial Emulation
Abstract
Introduction: Depression is a common comorbidity in patients with rheumatoid arthritis (RA), affecting treatment prognosis and quality of life. Janus kinase inhibitors (JAKi) and tumor necrosis factor inhibitors (TNFi) are widely used in moderate-to-severe RA and have been suggested to offer psychiatric benefits. However, direct comparisons of their effects on depression remain scarce.
Aims: To compare the risk of depression among RA patients initiating JAKi versus TNFi.
Methods: We conducted a population-based cohort study using 2017–2023 Korean National Health Insurance claims data, applying a target trial emulation framework. Eligible participants were RA patients aged ≥20 years who newly initiated JAKi or TNFi. The outcome was incident depression occurring ≥6 months after treatment initiation. Patients were followed from drug initiation until the earliest of outcome occurrence, treatment discontinuation, switching, death, or December 2023. Stabilized inverse probability of treatment weighting based on baseline propensity scores was used to adjust for confounding. Weighted Cox models estimated adjusted hazard ratios (aHRs).
Results: Among 8,181 patients (3,847 JAKi users; 4,334 TNFi users), most were female and aged 45–64 years. The incidence rate of depression per 1,000 person-years was 13.1 for JAKi users and 9.9 for TNFi users. JAKi use was not significantly associated with depression compared to TNFi (aHR 1.27, 95% CI 0.92–1.76). However, subgroup analyses revealed higher risks among patients without prior statin use (aHR 1.46, 95% CI 1.02–2.10), without benzodiazepine use (aHR 1.79, 95% CI 1.15–2.80), with intermediate RA-related visit frequency (aHR 1.72, 95% CI 1.01–2.93), and among upadacitinib users (aHR 1.70, 95% CI 1.04–2.77).
Conclusions: Overall, depression risk did not differ significantly between JAKi and TNFi, but JAKi may be less favorable in certain subgroups. These findings highlight the importance of considering psychiatric benefits in clinical decision-making.
Keywords: rheumatoid arthritis, JAKi vs TNFi, depression
Aims: To compare the risk of depression among RA patients initiating JAKi versus TNFi.
Methods: We conducted a population-based cohort study using 2017–2023 Korean National Health Insurance claims data, applying a target trial emulation framework. Eligible participants were RA patients aged ≥20 years who newly initiated JAKi or TNFi. The outcome was incident depression occurring ≥6 months after treatment initiation. Patients were followed from drug initiation until the earliest of outcome occurrence, treatment discontinuation, switching, death, or December 2023. Stabilized inverse probability of treatment weighting based on baseline propensity scores was used to adjust for confounding. Weighted Cox models estimated adjusted hazard ratios (aHRs).
Results: Among 8,181 patients (3,847 JAKi users; 4,334 TNFi users), most were female and aged 45–64 years. The incidence rate of depression per 1,000 person-years was 13.1 for JAKi users and 9.9 for TNFi users. JAKi use was not significantly associated with depression compared to TNFi (aHR 1.27, 95% CI 0.92–1.76). However, subgroup analyses revealed higher risks among patients without prior statin use (aHR 1.46, 95% CI 1.02–2.10), without benzodiazepine use (aHR 1.79, 95% CI 1.15–2.80), with intermediate RA-related visit frequency (aHR 1.72, 95% CI 1.01–2.93), and among upadacitinib users (aHR 1.70, 95% CI 1.04–2.77).
Conclusions: Overall, depression risk did not differ significantly between JAKi and TNFi, but JAKi may be less favorable in certain subgroups. These findings highlight the importance of considering psychiatric benefits in clinical decision-making.
Keywords: rheumatoid arthritis, JAKi vs TNFi, depression
Biography
Shin Yeong Kim is a master’s student in pharmacoepidemiology at the School of Pharmacy, Jeonbuk National University. Her research focuses on real-world evidence generation using national health insurance claims data, particularly in evaluating drug safety. She has conducted studies on the renal risks of potassium-competitive acid blockers (P-CABs) and the comparative risk of depression associated with biologic and targeted synthetic agents in patients with rheumatoid arthritis. She is skilled in SAS programming for large-scale data analysis and has experience implementing target trial emulation methodologies. As the first author, she has played a central role in multiple observational studies and collaborates closely with clinical and regulatory experts. Her work has been presented at national and international conferences and is currently being prepared for peer-reviewed publication. She ultimately aims to bridge real-world data science and clinical decision-making, contributing to safer and more effective pharmacotherapy and advancing global public health.
Mr Koshiro Arai
Researcher
Japan Institute for Health Security
Examining the necessity of subgroup-based Rapid Cycle Analysis: Lessons from COVID-19 vaccine
Abstract
Introduction: Robust safety surveillance is essential when introducing new vaccines. In Japan, a nationwide registry linked to claims database will be established in 2026, making real-time safety signal monitoring system an urgent priority. Rapid Cycle Analysis (RCA) is usually conducted in overall population, which can mask safety signals if there is an effect modification. The investigation on the necessity of subgroup-based RCA is limited.
Aims: We aim to compare how quickly safety signals are detected in overall population versus subgroups. We illustrate mRNA COVID-19 vaccine (wild-type) on myocarditis/pericarditis.
Methods: COVID-19 vaccination program in Japan was initiated on February 17, 2021. We compared vaccinated cohort (t0=vaccination date, tend=21 days after vaccination) and historical cohort (t0=January 1st 2019, tend=December 31st 2019) using the VENUS study (vaccine registry and insure-based claims data provided from 10 municipalities; 8 have historical data). Eligible individuals in each cohort were >12 years old, enrollees of health insurance, and had no history of myocarditis/pericarditis (defined by ICD-10 codes) at t0. We applied a Poisson-based Maximized Sequential Probability Ratio Test from February 2021 to December 2023, which monthly evaluates rate ratios (RR) between these cohorts for the overall population and several patterns of subgroups defined by sex and/or age (young:12–39, middle:40–64, old:≥65 years).
Results: We identified 859,880 individuals in historical cohort (mean age:68.38, female:58%). Among 981,606 vaccinated individuals (72.65, 58%), 74 outcomes emerged (2.19/100,000 doses). No signals were detected in the overall population (observed vs. expected cumulative events:74 vs. 63.52; RR:1.16 as of December 2023). Signals were detected in young-male (2 vs. 0.17; RR:11.76) in August 2021 and young subgroup (5 vs. 0.6; RR:8.33) in September 2021. No signals emerged in the other subgroups.
Conclusions: Much earlier detection highlights the importance of subgroup-based RCA. Further consideration is required for computational requirements and other outcomes.
Keywords: rapid cycle analysis, mRNA vaccine, subgroup analysis
Aims: We aim to compare how quickly safety signals are detected in overall population versus subgroups. We illustrate mRNA COVID-19 vaccine (wild-type) on myocarditis/pericarditis.
Methods: COVID-19 vaccination program in Japan was initiated on February 17, 2021. We compared vaccinated cohort (t0=vaccination date, tend=21 days after vaccination) and historical cohort (t0=January 1st 2019, tend=December 31st 2019) using the VENUS study (vaccine registry and insure-based claims data provided from 10 municipalities; 8 have historical data). Eligible individuals in each cohort were >12 years old, enrollees of health insurance, and had no history of myocarditis/pericarditis (defined by ICD-10 codes) at t0. We applied a Poisson-based Maximized Sequential Probability Ratio Test from February 2021 to December 2023, which monthly evaluates rate ratios (RR) between these cohorts for the overall population and several patterns of subgroups defined by sex and/or age (young:12–39, middle:40–64, old:≥65 years).
Results: We identified 859,880 individuals in historical cohort (mean age:68.38, female:58%). Among 981,606 vaccinated individuals (72.65, 58%), 74 outcomes emerged (2.19/100,000 doses). No signals were detected in the overall population (observed vs. expected cumulative events:74 vs. 63.52; RR:1.16 as of December 2023). Signals were detected in young-male (2 vs. 0.17; RR:11.76) in August 2021 and young subgroup (5 vs. 0.6; RR:8.33) in September 2021. No signals emerged in the other subgroups.
Conclusions: Much earlier detection highlights the importance of subgroup-based RCA. Further consideration is required for computational requirements and other outcomes.
Keywords: rapid cycle analysis, mRNA vaccine, subgroup analysis
Biography
Mr. Koshiro Arai is a researcher at Laboratory of Clinical Epidemiology, Department of Data Science, Center for Clinical Sciences, Japan Institute for Health Security (JIHS). He holds a master degree of public health (MPH) from Kyoto University, and completed training in epidemiology and biostatistics. He is also a doctoral student at Department of Information and Computer Technology, Graduate School of Engineering, Tokyo University of Science. His primary research interest is causal inference in longitudinal observational studies, with a focus on improving methods for evaluating the effect of sequential exposures. As a researcher at JIHS, he is currently involved in a project aimed at developing epidemiological and biostatistical methods for post-marketing vaccine safety surveillance using routinely collected health data. He collaborates with pharmacoepidemiologists and aims to apply his statistical expertise toward advancing both methodological development and practical implementation of sequential testing methods for signal detection in near real-time safety monitoring.
Ms Kyeongmin Lee
Kyung Hee University
Live zoster vaccination and chronic respiratory disease: An emulated target trial
Abstract
Introduction:
Live zoster vaccination (LZV) reduces the incidence of herpes zoster and its complications. Recent evidence suggests additional benefits, including lower risks of cardiovascular disease and dementia. However, its effect on chronic respiratory diseases remains unclear. Given that herpes zoster induces strong immune activation, potentially exacerbating chronic inflammation, evaluating whether vaccination mitigates respiratory disease risk is warranted.
Aims:
We evaluated whether LZV is associated with reduced risk of chronic respiratory diseases, including COPD, asthma, and interstitial lung disease (ILD).
Methods:
We conducted a target trial emulation using a nationwide, population-based cohort of 2,207,784 individuals aged ≥50 years in South Korea. Health insurance claims, national health screening data, and immunization records were integrated to identify exposure to LZV between January 1, 2012, and December 31, 2021. The outcomes were newly diagnosed and hospitalized cases of COPD, asthma, and ILD. After 1:1 exposure-driven propensity score matching, we used Cox proportional hazards models to estimate adjusted hazard ratios (aHRs).
Results:
The matched cohort included 431,266 individuals per group (mean age, 61.13 years; 47.70% male). LZV was associated with significantly reduced risks of COPD (aHR, 0.67 [95% CI, 0.67–0.68]), hospitalized COPD (0.65[0.61–0.68]), asthma (0.68[0.67–0.69]), hospitalized asthma (0.58[0.53–0.64]), ILD (0.80[0.74–0.87]), and hospitalized ILD (0.67[0.55–0.83]). The protective effects were more pronounced in non-smokers and were strongest within 1–3 years post-vaccination, persisting up to 7 years.
Conclusions:
LZV was associated with a reduced incidence of chronic respiratory diseases and related hospitalizations. These findings support broader implementation of zoster vaccination in older adults and highlight potential benefits beyond herpes zoster prevention. Although our study focused exclusively on the live attenuated vaccine, both live and recombinant zoster vaccines may modulate systemic inflammation. Therefore, Further research is needed on the respiratory benefits of recombinant zoster vaccination.
Keywords: chronic respiratory disease; herpes zoster; LZV, target trial emulation, vaccine effectiveness.
Live zoster vaccination (LZV) reduces the incidence of herpes zoster and its complications. Recent evidence suggests additional benefits, including lower risks of cardiovascular disease and dementia. However, its effect on chronic respiratory diseases remains unclear. Given that herpes zoster induces strong immune activation, potentially exacerbating chronic inflammation, evaluating whether vaccination mitigates respiratory disease risk is warranted.
Aims:
We evaluated whether LZV is associated with reduced risk of chronic respiratory diseases, including COPD, asthma, and interstitial lung disease (ILD).
Methods:
We conducted a target trial emulation using a nationwide, population-based cohort of 2,207,784 individuals aged ≥50 years in South Korea. Health insurance claims, national health screening data, and immunization records were integrated to identify exposure to LZV between January 1, 2012, and December 31, 2021. The outcomes were newly diagnosed and hospitalized cases of COPD, asthma, and ILD. After 1:1 exposure-driven propensity score matching, we used Cox proportional hazards models to estimate adjusted hazard ratios (aHRs).
Results:
The matched cohort included 431,266 individuals per group (mean age, 61.13 years; 47.70% male). LZV was associated with significantly reduced risks of COPD (aHR, 0.67 [95% CI, 0.67–0.68]), hospitalized COPD (0.65[0.61–0.68]), asthma (0.68[0.67–0.69]), hospitalized asthma (0.58[0.53–0.64]), ILD (0.80[0.74–0.87]), and hospitalized ILD (0.67[0.55–0.83]). The protective effects were more pronounced in non-smokers and were strongest within 1–3 years post-vaccination, persisting up to 7 years.
Conclusions:
LZV was associated with a reduced incidence of chronic respiratory diseases and related hospitalizations. These findings support broader implementation of zoster vaccination in older adults and highlight potential benefits beyond herpes zoster prevention. Although our study focused exclusively on the live attenuated vaccine, both live and recombinant zoster vaccines may modulate systemic inflammation. Therefore, Further research is needed on the respiratory benefits of recombinant zoster vaccination.
Keywords: chronic respiratory disease; herpes zoster; LZV, target trial emulation, vaccine effectiveness.
Biography
Ms. Kyeongmin Lee is a graduate student in regulatory science at Kyung Hee University. She studied biochemistry as an undergraduate and has a strong interest in public health and data-driven research. Currently, she is working on cohort studies using big data to explore topics like vaccines, drug safety, and pharmacoepidemiology. She has published several papers in these areas and continues to be involved in projects that use national health databases. Her goal is to better understand how real-world data can support safer and more effective healthcare decisions.
Miss Jiawen Lu
PhD student
The Hong Kong Polytechnic University
Establishing sex-age-specific TSH and FT4 reference intervals among adults in Hong Kong
Abstract
Introduction: Thyroid dysfunction typically manifests as nonspecific symptoms, necessitating reliance on biochemical markers, specifically thyroid-stimulating hormone (TSH) and free thyroxine (FT4), for diagnosis. Existing laboratory-specific reference intervals (RIs) often disregard sex and age variations, risking over- or underdiagnosis.
Aims: To establish dynamic, sex-age-stratified RIs for TSH and FT4 among adults in Hong Kong and evaluate their impact on thyroid dysfunction diagnosis compared to conventional RIs.
Methods: We analyzed Hong Kong’s electronic medical records (EMR) from 2006 to 2019, excluding individuals with prior thyroid-related medications (N=2,111,661 TSH; N=854,781 FT4). Participants were stratified by age, from 18-29 to ≥85 years. After harmonizing TSH/FT4 measurements across institutions, we derived sex-age-specific RIs annually using the 2.5th and 97.5th percentiles, with conventional RIs (TSH: 0.27-4.2mIU/L; FT4: 12-22pmol/L) serving as comparators. Diagnostic reclassification rates and incidence trends for overt/subclinical hypo-/hyperthyroidism were compared between these RI methods.
Results: Sex-age-specific upper reference limits (URLs) for TSH generally exceeded the laboratory-established URL, ranging from 4.2 to 7.1 mIU/L. FT4 URLs in women exhibited a pronounced U-shaped age distribution, peaking in young adulthood (18-29: 21.7-24.3 pmol/L) and advanced age (≥85: 23.6-26.4 pmol/L)—a pattern less pronounced in men. Using sex-age-specific RIs reduced overt/subclinical thyroid dysfunction diagnoses by 1-4% while increasing euthyroid cases by 14.8%. Incidence rates for sex-age-specific RIs-defined overt hypothyroidism (average annual percentage change [AAPC]=1.56%) and subclinical hypothyroidism (AAPC=1.81%; both P<0.05) showed sustained increases from 2010 to 2019.
Conclusions: This study establishes the first long-term, large-scale population-based RIs for TSH and FT4 in Hong Kong. Adoption of these RIs may help personalize thyroid monitoring intervals and therapeutic decisions, particularly for older populations. Future randomized controlled trials are warranted to examine the consequence of implementing sex-age-specific RIs in clinical settings on thyroid function management, particularly their potential to reduce overdiagnosis and overtreatment.
Keywords: Electronic medical records; sex-age-specific reference intervals; thyroid hormones.
Aims: To establish dynamic, sex-age-stratified RIs for TSH and FT4 among adults in Hong Kong and evaluate their impact on thyroid dysfunction diagnosis compared to conventional RIs.
Methods: We analyzed Hong Kong’s electronic medical records (EMR) from 2006 to 2019, excluding individuals with prior thyroid-related medications (N=2,111,661 TSH; N=854,781 FT4). Participants were stratified by age, from 18-29 to ≥85 years. After harmonizing TSH/FT4 measurements across institutions, we derived sex-age-specific RIs annually using the 2.5th and 97.5th percentiles, with conventional RIs (TSH: 0.27-4.2mIU/L; FT4: 12-22pmol/L) serving as comparators. Diagnostic reclassification rates and incidence trends for overt/subclinical hypo-/hyperthyroidism were compared between these RI methods.
Results: Sex-age-specific upper reference limits (URLs) for TSH generally exceeded the laboratory-established URL, ranging from 4.2 to 7.1 mIU/L. FT4 URLs in women exhibited a pronounced U-shaped age distribution, peaking in young adulthood (18-29: 21.7-24.3 pmol/L) and advanced age (≥85: 23.6-26.4 pmol/L)—a pattern less pronounced in men. Using sex-age-specific RIs reduced overt/subclinical thyroid dysfunction diagnoses by 1-4% while increasing euthyroid cases by 14.8%. Incidence rates for sex-age-specific RIs-defined overt hypothyroidism (average annual percentage change [AAPC]=1.56%) and subclinical hypothyroidism (AAPC=1.81%; both P<0.05) showed sustained increases from 2010 to 2019.
Conclusions: This study establishes the first long-term, large-scale population-based RIs for TSH and FT4 in Hong Kong. Adoption of these RIs may help personalize thyroid monitoring intervals and therapeutic decisions, particularly for older populations. Future randomized controlled trials are warranted to examine the consequence of implementing sex-age-specific RIs in clinical settings on thyroid function management, particularly their potential to reduce overdiagnosis and overtreatment.
Keywords: Electronic medical records; sex-age-specific reference intervals; thyroid hormones.
Biography
Miss Lu is an emerging researcher specializing in chronic disease epidemiology and causal inference methodologies. She earned her Bachelor of Medicine in Preventive Medicine from Soochow University and her Master of Public Health from Sun Yat-sen University. Her research focuses on elucidating the epidemiological patterns of chronic metabolic diseases by integrating causal inference approaches with multi-omics analyses (genomic, transcriptomic, and proteomic data). To date, she has authored or co-authored 10 peer-reviewed SCI-indexed publications, primarily as first or co-first author. Miss Lu’s scholarly excellence has been recognized through the prestigious Hong Kong Polytechnic University Presidential PhD Fellowship and an oral presentation award at the 13th Seoul International Congress of Endocrinology and Metabolism (Seoul, Korea). Her work bridges advanced analytical methodologies with clinical applications, aiming to improve personalized intervention strategies for metabolic disorders.
