Dark Data in Real-World Evidence: Challenges, Implications, and the Imperative of Data Literacy in Medical Research.

Randomized controlled trials (RCTs) and real-world evidence (RWE) studies are crucial and complementary in generating clinical evidence. RCTs provide controlled settings to validate the clinical effect of specific drugs or medical devices, while RWE integrates extrinsic factors, encompassing external influences affecting real-world scenarios, thus challenging RCT results in practical applications. In this study, we explore the impact of extrinsic factors on RWE outcomes, focusing on "dark data," which refers to data collected but not used or excluded from the analyses. Dark data can arise in many ways during research process, from selecting study samples to data collection and analysis. However, even unused or unanalyzed dark data hold potential insights, providing a comprehensive view of clinical contexts. Extrinsic factors lead to divergent RWE outcomes that could differ from RCTs beyond statistical correction's scope. Two main types of dark data exist: "known-unknown" and "unknown-unknown." The distinction between these dark data types highlights RWE's complexity. The transformation of unknown into known depends on data literacy-powerful utilization capabilities that can be interpreted based on medical expertise. Shifting the focus to excluded subjects or unused data in real-world contexts reveals unexplored potential. Understanding the significance of dark data is vital in reflecting the complexity of clinical settings. Connecting RCTs and RWEs requires medical data literacy, enabling clinicians to decipher meaningful insights. In the big data and artificial intelligence era, medical staff must navigate data complexities while promoting the core role of medicine. Prepared clinicians will lead this transformative journey, ensuring data value shapes the medical landscape.

Defining Medical AI Competencies for Medical School Graduates: Outcomes of a Delphi Survey and Medical Student/Educator Questionnaire of South Korean Medical Schools.

PURPOSE: Given the increasing significance and potential impact of artificial intelligence (AI) technology on health care delivery, there is an increasing demand to integrate AI into medical school curricula. This study aimed to define medical AI competencies and identify the essential competencies for medical graduates in South Korea. METHOD: An initial Delphi survey conducted in 2022 involving 4 groups of medical AI experts (n = 28) yielded 42 competency items. Subsequently, an online questionnaire survey was carried out with 1,955 participants (1,174 students and 781 professors) from medical schools across South Korea, utilizing the list of 42 competencies developed from the first Delphi round. A subsequent Delphi survey was conducted with 33 medical educators from 21 medical schools to differentiate the essential AI competencies from the optional ones. RESULTS: The study identified 6 domains encompassing 36 AI competencies essential for medical graduates: (1) understanding digital health and changes driven by AI; (2) fundamental knowledge and skills in medical AI; (3) ethics and legal aspects in the use of medical AI; (4) medical AI application in clinical practice; (5) processing, analyzing, and evaluating medical data; and (6) research and development of medical AI, as well as subcompetencies within each domain. While numerous competencies within the first 4 domains were deemed essential, a higher percentage of experts indicated responses in the last 2 domains, data science and medical AI research and development, were optional. CONCLUSIONS: This medical AI framework of 6 competencies and their subcompetencies for medical graduates exhibits promising potential for guiding the integration of AI into medical curricula. Further studies conducted in diverse contexts and countries are necessary to validate and confirm the applicability of these findings. Additional research is imperative for developing specific and feasible educational models to integrate these proposed competencies into pre-existing curricula.

Prediction of Cardiovascular Complication in Patients with Newly Diagnosed Type 2 Diabetes Using an XGBoost/GRU-ODE-Bayes-Based Machine-Learning Algorithm.

BACKGRUOUND: Cardiovascular disease is life-threatening yet preventable for patients with type 2 diabetes mellitus (T2DM). Because each patient with T2DM has a different risk of developing cardiovascular complications, the accurate stratification of cardiovascular risk is critical. In this study, we proposed cardiovascular risk engines based on machine-learning algorithms for newly diagnosed T2DM patients in Korea. METHODS: To develop the machine-learning-based cardiovascular disease engines, we retrospectively analyzed 26,166 newly diagnosed T2DM patients who visited Seoul St. Mary's Hospital between July 2009 and April 2019. To accurately measure diabetes-related cardiovascular events, we designed a buffer (1 year), an observation (1 year), and an outcome period (5 years). The entire dataset was split into training and testing sets in an 8:2 ratio, and this procedure was repeated 100 times. The area under the receiver operating characteristic curve (AUROC) was calculated by 10-fold cross-validation on the training dataset. RESULTS: The machine-learning-based risk engines (AUROC XGBoost=0.781±0.014 and AUROC gated recurrent unit [GRU]-ordinary differential equation [ODE]-Bayes=0.812±0.016) outperformed the conventional regression-based model (AUROC=0.723± 0.036). CONCLUSION: GRU-ODE-Bayes-based cardiovascular risk engine is highly accurate, easily applicable, and can provide valuable information for the individualized treatment of Korean patients with newly diagnosed T2DM.

Cumulative effect of impaired fasting glucose on the risk of dementia in middle-aged and elderly people: a nationwide cohort study.

The relationship between prediabetes and dementia remains controversial. We aimed to examine the association between cumulative exposure to impaired fasting glucose (IFG) and the risk of dementia in the general population. 1,463,066 middle-aged and elderly subjects who had had health examinations for four consecutive years were identified from a Korean nationwide population-based cohort database. IFG was defined as fasting blood glucose 100-125 mg/dL, and the risk of dementia-according to the number of IFG exposure (range 0-4)-was analyzed using the multivariable Cox proportional-hazards model. During the median 6.4 years of follow-up, 7614 cases of all-cause dementia, 5603 cases of Alzheimer's disease, and 1257 cases of vascular dementia occurred. There was a significant trend towards a higher risk of all-cause dementia (P for trend = 0.014) and Alzheimer's disease ( Pfor trend = 0.005) according to the cumulative exposure to IFG, but with a modest (approximately 7-14%) increase in the hazards. A significant stepwise increase in the risk of all-cause dementia and Alzheimer's disease was seen in non-obese subjects, whereas no significant association was observed in obese subjects. This study supports the association between prediabetes and incident dementia and emphasizes that even mild hyperglycemia should not be overlooked.

Risk of Cause-Specific Mortality across Glucose Spectrum in Elderly People: A Nationwide Population-Based Cohort Study.

BACKGRUOUND: This study investigated the risk of cause-specific mortality according to glucose tolerance status in elderly South Koreans. METHODS: A total of 1,292,264 individuals aged ≥65 years who received health examinations in 2009 were identified from the National Health Information Database. Participants were classified as normal glucose tolerance, impaired fasting glucose, newly-diagnosed diabetes, early diabetes (oral hypoglycemic agents ≤2), or advanced diabetes (oral hypoglycemic agents ≥3 or insulin). The risk of system-specific and disease-specific deaths was estimated using multivariate Cox proportional hazards analysis. RESULTS: During a median follow-up of 8.41 years, 257,356 deaths were recorded. Diabetes was associated with significantly higher risk of all-cause mortality (hazard ratio [HR], 1.58; 95% confidence interval [CI], 1.57 to 1.60); death due to circulatory (HR, 1.49; 95% CI, 1.46 to 1.52), respiratory (HR, 1.51; 95% CI, 1.47 to 1.55), and genitourinary systems (HR, 2.22; 95% CI, 2.10 to 2.35); and neoplasms (HR, 1.30; 95% CI, 1.28 to 1.32). Diabetes was also associated with a significantly higher risk of death due to ischemic heart disease (HR, 1.70; 95% CI, 1.63 to 1.76), cerebrovascular disease (HR, 1.46; 95% CI, 1.41 to 1.50), pneumonia (HR, 1.69; 95% CI, 1.63 to 1.76), and acute or chronic kidney disease (HR, 2.23; 95% CI, 2.09 to 2.38). There was a stepwise increase in the risk of death across the glucose spectrum (P for trend <0.0001). Stroke, heart failure, or chronic kidney disease increased the risk of all-cause mortality at every stage of glucose intolerance. CONCLUSION: A dose-dependent association between the risk of mortality from various causes and severity of glucose tolerance was noted in the elderly population.

Weight loss and side-effects of liraglutide and lixisenatide in obesity and type 2 diabetes mellitus.

AIMS: Glucagon-like peptide-1 receptor agonist (GLP-1 RA) is used to treat obesity or type 2 diabetes mellitus (DM). We compared weight loss and side-effects between patients with and without DM using GLP-1 RA. METHODS: This was a retrospective cohort study based on electronic medical records. Patients were categorized into three groups: liraglutide without DM (LiRa_NL), liraglutide with DM (LiRa_DM), and lixisenatide with DM (LiXi_DM). Six-month outcomes were evaluated for weight loss, side-effect types, and onset discontinuation of GLP-1 RA. RESULTS: We enrolled 356 (190 LiRa_NL, 95 LiRa_DM, and 71 LiXi_DM) patients (women, 72.5 %; mean age, 43.7 ± 12.7 years; mean body mass index, 30.7 ± 5.2 kg/m2). The mean glycated hemoglobin (HbA1c) participants were 7.7 ± 2.1 %. Average weight loss was 2.9 ± 0.3 kg. The change in HbA1c was lower in the LiXi_DM group than in the LiRa_DM group (- 1.1 ± 0.2 % vs. - 0.4 ± 0.1 %, P < 0.05). The LiRa_DM group showed a more effective weight loss (- 3.0 ± 0.4 kg) than the LiXi_DM group (- 0.9 ± 0.4 kg) (P < 0.05). Approximately 30 % of the patients reported experiencing side-effects, with gastrointestinal side-effects being the most frequent (20.5 %). The median side-effect onset was 1.9 ± 0.1 months from first treatment. The rate of GLP-1 RA discontinuation was 72.8 %. Discontinuation rates due to side-effects were 75.7 %, 68.9 %, and 64.4 % in the LiRa_NL, LiRa_DM, and LiXi_DM groups, respectively. CONCLUSIONS: The LiRa_NL group showed the most weight loss, although the discontinuation rate was high. Most side-effects occurred at 1-2 months. When prescribing GLP-1 RA, education concerning side-effects and discontinuation is needed to enhance treatment adherence.

The Present and Future of Artificial Intelligence-Based Medical Image in Diabetes Mellitus: Focus on Analytical Methods and Limitations of Clinical Use.

Artificial intelligence (AI)-based diagnostic technology using medical images can be used to increase examination accessibility and support clinical decision-making for screening and diagnosis. To determine a machine learning algorithm for diabetes complications, a literature review of studies using medical image-based AI technology was conducted using the National Library of Medicine PubMed, and the Excerpta Medica databases. Lists of studies using diabetes diagnostic images and AI as keywords were combined. In total, 227 appropriate studies were selected. Diabetic retinopathy studies using the AI model were the most frequent (85.0%, 193/227 cases), followed by diabetic foot (7.9%, 18/227 cases) and diabetic neuropathy (2.7%, 6/227 cases). The studies used open datasets (42.3%, 96/227 cases) or directly constructed data from fundoscopy or optical coherence tomography (57.7%, 131/227 cases). Major limitations in AI-based detection of diabetes complications using medical images were the lack of datasets (36.1%, 82/227 cases) and severity misclassification (26.4%, 60/227 cases). Although it remains difficult to use and fully trust AI-based imaging analysis technology clinically, it reduces clinicians' time and labor, and the expectations from its decision-support roles are high. Various data collection and synthesis data technology developments according to the disease severity are required to solve data imbalance.

Long-term change in the target achievement rate of low-density lipoprotein cholesterol in patients with cardiovascular disease.

OBJECTIVES: The goal achievement rate of patients' low-density lipoprotein cholesterol (LDL-C) levels and prescribing pattern of statin potency should be continuously monitored in a real-world clinical setting. This study aimed to describe the comprehensive status of LDL-C management. MATERIALS AND METHODS: Patients first diagnosed with cardiovascular diseases (CVDs) between 2009 and 2018 who were followed for 24 months. LDL-C levels, its changes from baseline, and intensity of statin prescribed were evaluated four times during follow-up. Potential factors associated with goal achievement were also identified. RESULTS: The study included 25,605 patients with CVDs. At diagnosis, the goal achievement rates of the LDL-C level were 58.4, 25.2, and 10.0% for targets of < 100, < 70, and < 55 mg/dL, respectively. The proportion of moderate- and high-intensity statin prescription significantly increased over time (all p < 0.01). Nevertheless, LDL-C levels significantly decreased at 6 months and increased at 12 and 24 months following therapy compared with baseline values. Glomerular filtration rate (GFR) (15 - 29 and < 15 mL/min/1.73m2) and accompanying diabetes mellitus were significantly associated with the goal achievement rate. CONCLUSION: Despite the need for active LDL-C management, the goal achievement rate and prescribing pattern were insufficient after 6 months. In cases with severe comorbidities, the goal attainment rate significantly increased; however, a more aggressive statin prescription was needed even in patients without diabetes or with normal GFR. The prescription rate for high-intensity statins increased over time, but was still low. In conclusion, physicians should aggressively prescribe statins to increase the goal achievement rate in patients with CVDs.

Predicting the Risk of Insulin-Requiring Gestational Diabetes before Pregnancy: A Model Generated from a Nationwide Population-Based Cohort Study in Korea.

BACKGRUOUND: The severity of gestational diabetes mellitus (GDM) is associated with adverse pregnancy outcomes. We aimed to generate a risk model for predicting insulin-requiring GDM before pregnancy in Korean women. METHODS: A total of 417,210 women who received a health examination within 52 weeks before pregnancy and delivered between 2011 and 2015 were recruited from the Korean National Health Insurance database. The risk prediction model was created using a sample of 70% of the participants, while the remaining 30% were used for internal validation. Risk scores were assigned based on the hazard ratios for each risk factor in the multivariable Cox proportional hazards regression model. Six risk variables were selected, and a risk nomogram was created to estimate the risk of insulin-requiring GDM. RESULTS: A total of 2,891 (0.69%) women developed insulin-requiring GDM. Age, body mass index (BMI), current smoking, fasting blood glucose (FBG), total cholesterol, and γ-glutamyl transferase were significant risk factors for insulin-requiring GDM and were incorporated into the risk model. Among the variables, old age, high BMI, and high FBG level were the main contributors to an increased risk of insulin-requiring GDM. The concordance index of the risk model for predicting insulin-requiring GDM was 0.783 (95% confidence interval, 0.766 to 0.799). The validation cohort's incidence rates for insulin-requiring GDM were consistent with the risk model's predictions. CONCLUSION: A novel risk engine was generated to predict insulin-requiring GDM among Korean women. This model may provide helpful information for identifying high-risk women and enhancing prepregnancy care.

Long-Term Risk of Cardiovascular Disease Among Type 2 Diabetes Patients According to Average and Visit-to-Visit Variations of HbA1c Levels During the First 3 Years of Diabetes Diagnosis.

BACKGROUND: It remains unclear whether a combination of glycemic variability and glycated hemoglobin (HbA1c) status leads to a higher incidence of cardiovascular disease (CVD). Therefore, to investigate CVD risk according to the glucose control status during early diabetes, we examined visit-to-visit HbA1c variability among patients with type 2 diabetes (T2DM). METHODS: In this 9-year retrospective study, we measured HbA1c levels at each visit and tracked the change in HbA1c levels for 3 years after the first presentation (observation window) in newly diagnosed T2DM patients. We later assessed the occurrence of CVD in the last 3 years (target outcome window) of the study period after allowing a 3-year buffering window. The HbA1c variability score (HVS; divided into quartiles, HVS_Q1-4) was used to determine visit-to-visit HbA1c variability. RESULTS: Among 4,817 enrolled T2DM patients, the mean HbA1c level was < 7% for the first 3 years. The group with the lowest HVS had the lowest rate of CVD (9.4%; 104/1,109 patients). The highest incidence of CVD of 26.7% (8/30 patients) was found in HVS [≥ 9.0%]_Q3, which was significantly higher than that in HVS [6.0-6.9%]_Q1 (P = 0.006), HVS [6.0-6.9%]_Q2 (P = 0.013), HVS [6.0-6.9%]_Q3 (P = 0.018), and HVS [7.0-7.9%]_Q3 (P = 0.040). CONCLUSION: To our knowledge, this is the first long-term study to analyze the importance of both HbA1c change and visit-to-visit HbA1c variability during outpatient visits within the first 3 years. Lowering glucose levels during early diabetes may be more critical than reducing visit-to-visit HbA1c variability.

Comorbidity Patterns and Management in Inpatients with Endocrine Diseases by Age Groups in South Korea: Nationwide Data.

This study aimed to examine comorbidity associations across age groups of inpatients with endocrine diseases as the primary diagnosis throughout the life cycle to develop an effective management strategy. Data were obtained from the Korean National Hospital Discharge In-depth Injury Survey (KNHDS) from 2006 to 2021, involving 68,515 discharged patients aged ≥ 19 years with a principal diagnosis of endocrine disease. A database was constructed for analysis, extracting general characteristics and comorbidities. Employing R version 4.2.3, the Chi-squared test and the Apriori algorithm of ARM (association rule mining) were used for analyzing general characteristics and comorbidity associations. There were more women (53.1%) than men (46.9%) (p < 0.001, with women (61.2 ± 17.2) having a higher average age than men (58.6 ± 58.6) (p < 0.001). Common comorbidities include unspecified diabetes mellitus; essential (primary) hypertension; unspecified diabetes mellitus; and other disorders of fluid, electrolyte, and acid-base balance. Notably, type 2 diabetes mellitus, disorders of lipoprotein metabolism and other lipidemia, polyneuropathy in diseases classified elsewhere, retinal disorders in diseases classified elsewhere, and essential (primary) hypertension prevail across all age groups. Association rules further highlight specific comorbidities appearing selectively in certain age groups. In conclusion, establishing a management strategy for comorbidities in patients with a primary diagnosis of an endocrine disorder is necessary.

A Study on Weight Loss Cause as per the Side Effect of Liraglutide.

Purpose: Liraglutide is known to have much lower weight loss effects in real clinical fields than in randomized clinical trials because of its side effects (SE) and discomfort associated with injections. This study is aimed at determining whether the side effects of liraglutide affect weight reduction and its maintenance in real-world practice. Methods: Endocrinologists conducted a retrospective chart review of data from two tertiary university hospitals. All patients who had been prescribed liraglutide at least once between January 2014 and December 2019 were included. For an average of 3 and 6 months, weight changes due to the presence or absence of SE and discontinuation (MAIN or STOP) of liraglutide were checked. Results: Only 40.8% (64/157) of the patients remained on liraglutide for 6 months; 14.7% (23/157) maintained the drug despite SEs (MAIN_SE(+)), and 40.1% (63/157) discontinued the drug despite not having SEs (STOP_SE(-)). At 3 months, there was -5.9 ± 0.6%, -7.9 ± 0.9%, -4.5 ± 0.5%, and -3.4 ± 0.6% weight reduction in the MAIN_SE(-), MAIN_SE(+), STOP_SE(-), and STOP_SE(+) groups, respectively (all p < 0.001 compared to the baseline). However, there were no significant differences in the weight loss between the MAIN (p = 0.062) and STOP (p = 0.204) groups. At 6 months, the weight reduction was -2.0 ± 0.5% (p < 0.001) in MAIN_SE(-), -2.2 ± 0.7% (p < 0.005) in MAIN_SE(+), -1.7 ± 0.7% (p < 0.01) in STOP_SE(-), and -2.0 ± 0.6% (p = 0.01) in STOP_SE(+), compared to baseline. SEs also caused no significant differences in weight loss between the MAIN (p = 0.787) and STOP (p = 0.694) groups. Conclusions: Our study confirmed that the side effects of liraglutide did not affect weight reduction. Moreover, in the real world, the continuous rate of liraglutide use is not high, and the weight gradually increases after 3 months. Therefore, in addition to the side effects of liraglutide, the medical staff should consider various factors that affect drug adherence, consider ways to increase compliance, and continue to ensure management so that patients can maintain their weight.

Improving Machine Learning Diabetes Prediction Models for the Utmost Clinical Effectiveness.

The early prediction of diabetes can facilitate interventions to prevent or delay it. This study proposes a diabetes prediction model based on machine learning (ML) to encourage individuals at risk of diabetes to employ healthy interventions. A total of 38,379 subjects were included. We trained the model on 80% of the subjects and verified its predictive performance on the remaining 20%. Furthermore, the performances of several algorithms were compared, including logistic regression, decision tree, random forest, eXtreme Gradient Boosting (XGBoost), Cox regression, and XGBoost Survival Embedding (XGBSE). The area under the receiver operating characteristic curve (AUROC) of the XGBoost model was the largest, followed by those of the decision tree, logistic regression, and random forest models. For the survival analysis, XGBSE yielded an AUROC exceeding 0.9 for the 2- to 9-year predictions and a C-index of 0.934, while the Cox regression achieved a C-index of 0.921. After lowering the threshold from 0.5 to 0.25, the sensitivity increased from 0.011 to 0.236 for the 2-year prediction model and from 0.607 to 0.994 for the 9-year prediction model, while the specificity showed negligible changes. We developed a high-performance diabetes prediction model that applied the XGBSE algorithm with threshold adjustment. We plan to use this prediction model in real clinical practice for diabetes prevention after simplifying and validating it externally.

Long-Term Changes in HbA1c According to Blood Glucose Control Status During the First 3 Months After Visiting a Tertiary University Hospital.

BACKGROUND: We evaluated patients visiting a tertiary university hospital due to a diagnosis of diabetes with a goal of achieving blood glucose control and evaluated blood glucose persistence over 7 years according to the change in blood glucose evident at 3 months after the first visit. METHODS: Patients treated from 2009 to 2013 were categorized into four groups according to the change in HbA1c levels during the first 3 months of follow-up (Best_group, ≥ 1.6% decrease; Better_group, 0.5-1.5% decrease; Neutral_group, maintained at -0.4% to +0.4%; Worse_group, ≥ 0.5% increase). Each patient's blood glucose control status was then monitored for 7 years. The incidence of stroke and acute coronary syndrome during this period was confirmed. RESULTS: Overall, 9,776 patients were included. HbA1c values were lower in the Best_group than in the other groups at all time points (all P < 0.001). The rate of reaching targets of < 6.5% or < 7.0% HbA1c decreased over time; the rate at which the estimated glomerular filtration rate decreased to < 30 or < 60 mL/min/1.73m² increased over time (all trends, P < 0.01). CONCLUSION: Blood glucose control status in the first 3 months after initiating hospital care enabled estimation of the patient's glycemic control status for the next 7 years. In cases with poor initial blood glucose control, a new or more active method of blood glucose control should be sought.

Characteristics of Glycemic Control and Long-Term Complications in Patients with Young-Onset Type 2 Diabetes.

BACKGRUOUND: The prevalence of young-onset diabetes (YOD) has been increasing worldwide. As the incidence of YOD increases, it is necessary to determine the characteristics of YOD and the factors that influence its development and associated complications. METHODS: In this retrospective study, we recruited patients who were diagnosed with type 2 diabetes mellitus between June 2001 and December 2021 at a tertiary hospital. The study population was categorized according to age: YOD (age <40 years), middle-age-onset diabetes (MOD, 40≤ age <65 years), and late-onset diabetes (LOD, age ≥65 years). We examined trends in glycemic control by analyzing fasting glucose levels during the first year in each age group. A Cox proportional-hazards model was used to determine the relative risk of developing complications according to glycemic control trends. RESULTS: The fasting glucose level at the time of diagnosis was highest in the YOD group (YOD 149±65 mg/dL; MOD 143±54 mg/dL; and LOD 140±55 mg/dL; p=0.009). In the YOD group, glucose levels decreased at 3 months, but increased by 12 months. YOD patients and those with poor glycemic control in the first year were at a higher risk of developing complications, whereas the risk in patients with LOD was not statistically significant. CONCLUSION: YOD patients had higher glucose levels at diagnosis, and their glycemic control was poorly maintained. As poor glycemic control can influence the development of complications, especially in young patients, intensive treatment is necessary for patients with YOD.

Retrospective cohort analysis comparing changes in blood glucose level and body composition according to changes in thyroid-stimulating hormone level.

BACKGROUND: In the euthyroid state, the risk of developing diabetes according to changes in thyroid-stimulating hormone (TSH) levels remains controversial. Additionally, the correlation of various body indices affecting blood glucose levels according to changes in TSH levels over a certain period is not well known. METHODS: Patients who underwent health check-ups twice at a 2 year interval at a tertiary university hospital between 2009 and 2018 were included. By dividing baseline TSH levels into quartiles (TSH_Q1, Q2, Q3, and Q4), various variables were compared, and their changes after 2 years (∆TSH_Q1, Q2, Q3, and Q4) were confirmed. RESULTS: Among 15 557 patients, the incidence of diabetes mellitus after 2 years was 2.4% (377/15 557 patients). There was no statistically significant difference in the incidence of diabetes according to TSH_Q (p = 0.243) or ∆TSH_Q (p = 0.131). However, as TSH levels increased, skeletal muscle mass decreased (p < 0.001), and body fat mass and percent body fat significantly increased (p < 0.001). As ∆TSH increased, ∆fasting blood glucose and ∆body mass index also significantly increased (all p < 0.001). The incidence of diabetes decreased significantly as skeletal muscle mass increased (odds ratio 0.734, p < 0.001). CONCLUSIONS: Owing to the short study period, it was not possible to prove a statistical relationship between the incidence of diabetes mellitus and TSH levels in the euthyroid state. Significant decreases in skeletal muscle mass and increases in body mass index and body fat mass according to baseline TSH levels were demonstrated. Therefore, a focus on improving physical functions, such as increasing muscle mass and decreasing fat, is required in this case.

The Impact of the Association between Cancer and Diabetes Mellitus on Mortality.

The prevalence of cancer, diabetes mellitus (DM), and hypertension is increasing in ageing populations. We analyzed the association of DM with cancer and its effects on cancer mortality. The data of 2009-2018 from the Korea National Hospital Discharge In-depth Injury Survey were used; 169,959 adults with cancer as the main diagnosis were identified. The association rule for unsupervised machine learning was used. Association rule mining was used to analyze the association between the diseases. Logistic regression was performed to determine the effects of DM on cancer mortality. DM prevalence was 12.9%. Cancers with high DM prevalence were pancreatic (29.9%), bile duct (22.7%), liver (21.4%), gallbladder (15.5%), and lung cancers (15.4%). Cancers with high hypertension prevalence were bile duct (31.4%), ureter (30.5%), kidney (29.5%), pancreatic (28.1%), and bladder cancers (27.5%). The bidirectional association between DM and hypertension in cancer was the strongest (lift = 2.629, interest support [IS] scale = 0.426), followed by that between lung cancer and hypertension (lift = 1.280, IS scale = 0.204), liver cancer and DM (lift = 1.658, IS scale = 0.204), hypertension and liver cancer and DM (lift = 3.363, IS scale = 0.197), colorectal cancer and hypertension (lift = 1.133, IS scale = 0.180), and gastric cancer and hypertension (lift = 1.072, IS scale = 0.175). DM increased liver cancer mortality (p = 0.000), while hypertension significantly increased the mortality rate of stomach, colorectal, liver, and lung cancers. Our study confirmed the association between cancer and DM. Consequently, a patient management strategy with presumptive diagnostic ability for DM and hypertension is required to decrease cancer mortality rates.

Lipid cutoffs for increased cardiovascular disease risk in non-diabetic young people.

AIMS: The relevance of blood lipid abnormalities to cardiovascular disease (CVD) risk in young populations is unclear. Here, we aimed to identify the cutoff levels of lipid parameters for increased risk of CVD among nondiabetic young adults aged 20-39 years. METHODS: Using data from a nationally representative Korean National Health Insurance System database, we followed up 6 204 153 subjects who underwent health examinations between 2009 and 2012 until the end of 2018. The primary outcome was incident CVD, defined as a composite of myocardial infarction and stroke. We assessed the associations between pre-specified lipid levels and CVD risk. Subgroup analysis of the number of cardiovascular risk factors (obesity, hypertension, and current smoking) was also conducted. RESULTS: During a median follow-up of 7.7 years, there were 14 569 (0.23%) cases of myocardial infarction, 9,459 (0.15%) cases of stroke, and 23 680 (0.38%) cases of composite CVD. Using total cholesterol (TC) level of <140 mg/dL, triglyceride (TG) level of <60 mg/dL, LDL-cholesterol level of <100 mg/dL, and non-HDL-cholesterol level of <130 mg/dL as reference groups, a significantly higher risk of CVD was observed in subjects with a TC level of ≥200 mg/dL, TG level of ≥60 mg/dL, LDL-cholesterol level of ≥130 mg/dL, or non-HDL-cholesterol level of ≥140 mg/dL. The cutoff levels of TC that had statistical significance for increased risk of CVD were 240, 220, and 200 mg/dL in subjects with 0, 1, or 2-3 risk factors, respectively. CONCLUSIONS: Even modest increases in lipid levels were associated with increased risk of CVD in this nondiabetic young population. Our data provide potential criteria for stratifying CVD risk based on real-world evidence.

Understanding and Utilizing Claim Data from the Korean National Health Insurance Service (NHIS) and Health Insurance Review & Assessment (HIRA) Database for Research.

Almost every Korean (97%) is enrolled in the National Health Insurance program, and most receive medical treatment at least once a year. Data are collected by the Health Insurance Review and Assessment Service (HIRA), and the results of the review are sent to the National Health Insurance Service (NHIS). The data handled by NHIS and HIRA cover almost the entire population and can be used for various research purposes. NHIS and HIRA support research by making these data available to researchers. The greatest advantage of these data is that they are the only data which include virtually the entire population. Both HIRA and NHIS data are provided in the form of sample data and all (customized) data. NHIS and HIRA data are similar but exhibit minor differences. HIRA data consists of five tables, including general specification details, in-hospital treatment details, disease details, out-of-hospital prescription details, and nursing institution information. NHIS data include death records (including cause of death), some medical examination records, and the socio-economic variables of the subject, such as income, in addition to all the HIRA data. Clinical results of treatments are not recorded in NHIS or HIRA. However, because public data are used for billing purposes, actual research has thus far been limited. Therefore, researchers must develop a study design that can minimize the errors or bias occurring during the course of the study. Therefore, it is necessary to clearly understand the structure and characteristics of NHIS and HIRA data when initiating research.