Development of Risk Score for Predicting 3-Year Incidence of Type 2 Diabetes: Japan Epidemiology Collaboration on Occupational Health Study

[Construction of a Predictive Model for Diabetes Mellitus Type 2 in Middle-Aged and Elderly Populations Based on the Medical Checkup Data of National Basic Public Health Service]

Objective

To establish a universally applicable logistic risk prediction model for diabetes mellitus type 2 (T2DM) in the middle-aged and elderly populations based on the results of a Meta-analysis, and to validate and confirm the efficacy of the model using the follow-up data of medical check-ups of National Basic Public Health Service.

Methods

Cohort studies evaluating T2DM risks were identified in Chinese and English databases. The logistic model utilized Meta-combined effect values such as the odds ratio (OR) to derive β, the partial regression coefficient, of the logistic model. The Meta-combined incidence rate of T2DM was used to obtain the parameter α of the logistic model. Validation of the predictive performance of the model was conducted with the follow-up data of medical checkups of National Basic Public Health Service. The follow-up data came from a community health center in Chengdu and were collected between 2017 and 2022 from 7602 individuals who did not have T2DM at their baseline medical checkups done at the community health center. This community health center was located in an urban-rural fringe area with a large population of middle-aged and elderly people.

Results

A total of 40 cohort studies were included and 10 items covered in the medical checkups of National Basic Public Health Service were identified in the Meta-analysis as statistically significant risk factors for T2DM, including age, central obesity, smoking, physical inactivity, impaired fasting glucose, a reduced level of high-density lipoprotein cholesterol (HDL-C), hypertension, body mass index (BMI), triglyceride glucose (TYG) index, and a family history of diabetes, with the OR values and 95% confidence interval (CI) being 1.04 (1.03, 1.05), 1.55 (1.29, 1.88), 1.36 (1.11, 1.66), 1.26 (1.07, 1.49), 3.93 (2.94, 5.24), 1.14 (1.06, 1.23), 1.47 (1.34, 1.61), 1.11 (1.05, 1.18), 2.15 (1.75, 2.62), and 1.66 (1.55, 1.78), respectively, and the combined β values being 0.039, 0.438, 0.307, 0.231, 1.369, 0.131, 0.385, 0.104, 0.765, and 0.507, respectively. A total of 37 studies reported the incidence rate, with the combined incidence being 0.08 (0.07, 0.09) and the parameter α being -2.442 for the logistic model. The logistic risk prediction model constructed based on Meta-analysis was externally validated with the data of 7602 individuals who had medical checkups and were followed up for at least once. External validation results showed that the predictive model had an area under curve (AUC) of 0.794 (0.771, 0.816), accuracy of 74.5%, sensitivity of 71.0%, and specificity of 74.7% in the 7602 individuals.

Conclusion

The T2DM risk prediction model based on Meta-analysis has good predictive performance and can be used as a practical tool for T2DM risk prediction in middle-aged and elderly populations.

Simple risk scoring using sinus rhythm electrocardiograms predicts the incidence of atrial fibrillation in the general population

Atrial fibrillation (AF) is an arrhythmic disease. Prediction of AF development in healthy individuals is important before serious complications occur. We aimed to develop a risk prediction score for future AF using participants' data, including electrocardiogram (ECG) measurements and information such as age and sex. We included 88,907 Japanese participants, aged 30-69 years, who were randomly assigned to derivation and validation cohorts in a ratio of 1:1. We performed multivariate logistic regression analysis and obtained the standardised beta coefficient of relevant factors and assigned scores to them. We created a score based on prognostic factors for AF to predict its occurrence after five years and applied it to validation cohorts to assess its reproducibility. The risk score ranged from 0 to 17, consisting of age, sex, PR prolongation, QT corrected for heart rate prolongation, left ventricular hypertrophy, premature atrial contraction, and left axis deviation. The area under the curve was 0.75 for the derivation cohort and 0.73 for the validation cohort. The incidence of new-onset AF reached over 2% at 10 points of the risk score in both cohorts. Thus, in this study, we showed the possibility of predicting new-onset AF using ECG findings and simple information.

Cost-effectiveness of behavioural counselling intervention compared with non-intervention for adult patients with metabolic syndrome to prevent cardiovascular diseases and type 2 diabetes in Japan: a microsimulation modelling study

OBJECTIVES

Nationwide lifestyle intervention-specific health guidance (SHG) in Japan-employs counselling and education to change unhealthy behaviours that contribute to metabolic syndrome, especially obesity or abdominal obesity. We aimed to perform a model-based economic evaluation of SHG in a low participation rate setting.

DESIGN

A hypothetical population, comprised 50 000 Japanese aged 40 years who met the criteria of the SHG, used a microsimulation using the Markov model to evaluate SHG's cost-effectiveness compared with non-SHG. This hypothetical population was simulated over a 35-year time horizon.

SETTING

SHG is conducted annually by all Japanese insurers.

OUTCOME MEASURES

Model parameters, such as costs and health outcomes (including quality-adjusted life-years, QALYs), were based on existing literature. Incremental cost-effectiveness ratios were estimated from the healthcare payer's perspective. Deterministic and probabilistic sensitivity analyses (PSA) were conducted to evaluate the uncertainty around the model input parameters.

RESULTS

The simulation revealed that the total costs per person in the SHG group decreased by JPY53 014 (US$480) compared with that in the non-SHG group, and the QALYs increased by 0.044, wherein SHG was considered the dominant strategy despite the low participation rates. PSA indicated that the credibility intervals (2.5th-97.5th percentile) of the incremental costs and the incremental QALYs with the SHG group compared with the non-SHG group were -JPY687 376 to JPY85 197 (-US$6226 to US$772) and -0.009 to 0.350 QALYs, respectively. Each scenario analysis indicated that programmes for improving both blood pressure and blood glucose levels among other risk factors for metabolic syndrome are essential for improving cost-effectiveness.

CONCLUSIONS

This study suggests that even small effects of counselling and education on behavioural modification may lead to the prevention of acute life-threatening events and chronic diseases, in addition to the reduction of medication resulting from metabolic syndrome, which results in cost savings.

Development and Validation of Prediction Models for the 5-year Risk of Type 2 Diabetes in a Japanese Population: Japan Public Health Center-based Prospective (JPHC) Diabetes Study

BACKGROUND

This study aimed to develop models to predict the 5-year incidence of type 2 diabetes mellitus (T2DM) in a Japanese population and validate them externally in an independent Japanese population.

METHODS

Data from 10,986 participants (aged 46-75 years) in the development cohort of the Japan Public Health Center-based Prospective Diabetes Study and 11,345 participants (aged 46-75 years) in the validation cohort of the Japan Epidemiology Collaboration on Occupational Health Study were used to develop and validate the risk scores in logistic regression models.

RESULTS

We considered non-invasive (sex, body mass index, family history of diabetes mellitus, and diastolic blood pressure) and invasive (glycated hemoglobin [HbA1c] and fasting plasma glucose [FPG]) predictors to predict the 5-year probability of incident diabetes. The area under the receiver operating characteristic curve was 0.643 for the non-invasive risk model, 0.786 for the invasive risk model with HbA1c but not FPG, and 0.845 for the invasive risk model with HbA1c and FPG. The optimism for the performance of all models was small by internal validation. In the internal-external cross-validation, these models tended to show similar discriminative ability across different areas. The discriminative ability of each model was confirmed using external validation datasets. The invasive risk model with only HbA1c was well-calibrated in the validation cohort.

CONCLUSION

Our invasive risk models are expected to discriminate between high- and low-risk individuals with T2DM in a Japanese population.

Individual health-disease phase diagrams for disease prevention based on machine learning

Early disease detection and prevention methods based on effective interventions are gaining attention worldwide. Progress in precision medicine has revealed that substantial heterogeneity exists in health data at the individual level and that complex health factors are involved in chronic disease development. Machine-learning techniques have enabled precise personal-level disease prediction by capturing individual differences in multivariate data. However, it is challenging to identify what aspects should be improved for disease prevention based on future disease-onset prediction because of the complex relationships among multiple biomarkers. Here, we present a health-disease phase diagram (HDPD) that represents an individual's health state by visualizing the future-onset boundary values of multiple biomarkers that fluctuate early in the disease progression process. In HDPDs, future-onset predictions are represented by perturbing multiple biomarker values while accounting for dependencies among variables. We constructed HDPDs for 11 diseases using longitudinal health checkup cohort data of 3,238 individuals, comprising 3,215 measurement items and genetic data. The improvement of biomarker values to the non-onset region in HDPD remarkably prevented future disease onset in 7 out of 11 diseases. HDPDs can represent individual physiological states in the onset process and be used as intervention goals for disease prevention.

Data collaboration analysis in predicting diabetes from a small amount of health checkup data

Recent studies showed that machine learning models such as gradient-boosting decision tree (GBDT) can predict diabetes with high accuracy from big data. In this study, we asked whether highly accurate prediction of diabetes is possible even from small data by expanding the amount of data through data collaboration (DC) analysis, a modern framework for integrating and analyzing data accumulated at multiple institutions while ensuring confidentiality. To this end, we focused on data from two institutions: health checkup data of 1502 citizens accumulated in Tsukuba City and health history data of 1399 patients collected at the University of Tsukuba Hospital. When using only the health checkup data, the ROC-AUC and Recall for logistic regression (LR) were 0.858 ± 0.014 and 0.970 ± 0.019, respectively, while those for GBDT were 0.856 ± 0.014 and 0.983 ± 0.016, respectively. When using also the health history data through DC analysis, these values for LR improved to 0.875 ± 0.013 and 0.993 ± 0.009, respectively, while those for GBDT deteriorated because of the low compatibility with a method used for confidential data sharing (although DC analysis brought improvements). Even in a situation where health checkup data of only 324 citizens are available, the ROC-AUC and Recall for LR were 0.767 ± 0.025 and 0.867 ± 0.04, respectively, thanks to DC analysis, indicating an 11% and 12% improvement. Thus, we concluded that the answer to the above question was "Yes" for LR but "No" for GBDT for the data set tested in this study.

Development of a risk prediction score and equation for chronic kidney disease: a retrospective cohort study

Chronic kidney disease (CKD) is a risk factor for end-stage renal disease and contributes to increased risk of cardiovascular disease morbidity and mortality. We aimed to develop a risk prediction score and equation for future CKD using health checkup data. This study included 58,423 Japanese participants aged 30-69 years, who were randomly assigned to derivation and validation cohorts at a ratio of 2:1. The predictors were anthropometric indices, life style, and blood sampling data. In derivation cohort, we performed multivariable logistic regression analysis and obtained the standardized beta coefficient of each factor that was significantly associated with new-onset CKD and assigned scores to each factor. We created a score and an equation to predict CKD after 5 years and applied them to validation cohort to assess their reproducibility. The risk score ranged 0-16, consisting of age, sex, hypertension, dyslipidemia, diabetes, hyperuricemia, and estimated glomerular filtration rate (eGFR), with area under the curve (AUC) of 0.78 for the derivation cohort and 0.79 for the validation cohort. The CKD incidence gradually and constantly increased as the score increased from ≤ 6 to ≥ 14. The equation consisted of the seven indices described above, with AUC of 0.88 for the derivation cohort and 0.89 for the validation cohort. We developed a risk score and equation to predict CKD incidence after 5 years in Japanese population under 70 years of age. These models had reasonably high predictivity, and their reproducibility was confirmed through internal validation.

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.

Risk prediction models for incident type 2 diabetes in Chinese people with intermediate hyperglycemia: a systematic literature review and external validation study

Background

People with intermediate hyperglycemia (IH), including impaired fasting glucose and/or impaired glucose tolerance, are at higher risk of developing type 2 diabetes (T2D) than those with normoglycemia. We aimed to evaluate the performance of published T2D risk prediction models in Chinese people with IH to inform them about the choice of primary diabetes prevention measures.

Methods

A systematic literature search was conducted to identify Asian-derived T2D risk prediction models, which were eligible if they were built on a prospective cohort of Asian adults without diabetes at baseline and utilized routinely-available variables to predict future risk of T2D. These Asian-derived and five prespecified non-Asian derived T2D risk prediction models were divided into BASIC (clinical variables only) and EXTENDED (plus laboratory variables) versions, with validation performed on them in three prospective Chinese IH cohorts: ACE (n = 3241), Luzhou (n = 1333), and TCLSIH (n = 1702). Model performance was assessed in terms of discrimination (C-statistic) and calibration (Hosmer–Lemeshow test).

Results

Forty-four Asian and five non-Asian studies comprising 21 BASIC and 46 EXTENDED T2D risk prediction models for validation were identified. The majority were at high (n = 43, 87.8%) or unclear (n = 3, 6.1%) risk of bias, while only three studies (6.1%) were scored at low risk of bias. BASIC models showed poor-to-moderate discrimination with C-statistics 0.52–0.60, 0.50–0.59, and 0.50–0.64 in the ACE, Luzhou, and TCLSIH cohorts respectively. EXTENDED models showed poor-to-acceptable discrimination with C-statistics 0.54–0.73, 0.52–0.67, and 0.59–0.78 respectively. Fifteen BASIC and 40 EXTENDED models showed poor calibration (P < 0.05), overpredicting or underestimating the observed diabetes risk. Most recalibrated models showed improved calibration but modestly-to-severely overestimated diabetes risk in the three cohorts. The NAVIGATOR model showed the best discrimination in the three cohorts but had poor calibration (P < 0.05).

Conclusions

In Chinese people with IH, previously published BASIC models to predict T2D did not exhibit good discrimination or calibration. Several EXTENDED models performed better, but a robust Chinese T2D risk prediction tool in people with IH remains a major unmet need.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-022-01622-5.

Development of Various Diabetes Prediction Models Using Machine Learning Techniques

BACKGROUND

There are many models for predicting diabetes mellitus (DM), but their clinical implication remains vague. Therefore, we aimed to create various DM prediction models using easily accessible health screening test parameters.

METHODS

Two sets of variables were used to develop eight DM prediction models. One set comprised 62 easily accessible examination results of commonly used variables from a tertiary university hospital. The second set comprised 27 of the 62 variables included in the national routine health checkups. Gradient boosting and random forest algorithms were used to develop the models. Internal validation was performed using the stratified 10-fold cross-validation method.

RESULTS

The area under the receiver operating characteristic curve (ROC-AUC) for the 62-variable DM model making 12-month predictions for subjects without diabetes was the largest (0.928) among those of the eight DM prediction models. The ROC-AUC dropped by more than 0.04 when training with the simplified 27-variable set but still showed fairly good performance with ROC-AUCs between 0.842 and 0.880. The accuracy was up to 11.5% higher (from 0.807 to 0.714) when fasting glucose was included.

CONCLUSION

We created easily applicable diabetes prediction models that deliver good performance using parameters commonly assessed during tertiary university hospital and national routine health checkups. We plan to perform prospective external validation, hoping that the developed DM prediction models will be widely used in clinical practice.

Development of a risk prediction score for hypertension incidence using Japanese health checkup data

Hypertension is a risk factor for cardiovascular disease. We developed a simple scoring method for predicting future hypertension using health checkup data. A total of 41,902 participants aged 30-69 years without baseline hypertension who underwent annual health checkups (mean age, 52.3 ± 10.2 years; male, 47.7%) were included. They were randomly assigned to derivation (n = 27,935) and validation cohorts (n = 13,967) at a ratio of 2:1. In the derivation cohort, we performed multivariable logistic regression analysis and assigned scores to each factor significantly associated with 5-year hypertension. We evaluated the predictive ability of the scores using area under the curve (AUC) analysis and then applied them to the validation cohort to assess their validity. The score including items requiring blood sampling ranged from 0 to 14 and included seven indicators (age, body mass index, blood pressure, current smoking, family history of hypertension, diabetes, and hyperuricemia). The score not including items requiring blood sampling ranged from 0 to 12 and included five indicators (the above indicators, except diabetes and hyperuricemia). The score not including items requiring blood sampling was better; blood sampling did not improve diagnostic ability. The AUC of the score not including items requiring blood sampling was 0.76, with a sensitivity and specificity of 0.82 and 0.60, respectively, for scores ≥6 points. The incidence of hypertension gradually and constantly increased (from 0.9 to 49.6%) as the score increased from 0 to ≥10. Analysis in the validation cohort yielded similar results. We developed a simple and useful clinical prediction model to predict the 5-year incidence of hypertension among a general Japanese population. The model had reasonably high predictive ability and reproducibility.

The association between the reduction of body weight and new-onset type 2 diabetes remission in middle-aged Japanese men: Population-based Panasonic cohort study 8

AIM

This study aimed to investigate the association between change in body weight (BW) and type 2 diabetes remission in Japanese men with new-onset type 2 diabetes.

METHODS

This study enrolled 1,903 patients with new-onset type 2 diabetes between 2008 and 2013 from a medical health checkup program conducted by the Panasonic Corporation, Osaka, Japan. The baseline was defined as the year of new-onset diabetes. We assessed the type 2 diabetes remission five years after baseline and the association between the change in BW and type 2 diabetes remission using logistic regression analyses. To evaluate the predictive performance of the change in BW, we employed the receiver operating characteristic curves and the area under the receiver operating characteristic (ROC) curve (AUC).

RESULTS

The BW loss was associated with type 2 diabetes remission in the participants with a BMI ≥25 kg/m² but not in the participants with a BMI <25 kg/m². The odds ratios were 1.96 (95% CI: 1.19-3.29) and 3.72 (95% CI: 2.14-6.59) in the participants with a loss of 5-9.9% and loss of ≥10% for five years, respectively, in the participants with a BMI ≥25 kg/m² (reference; stable group [0.9% gain to 0.9% loss]). The AUC and cut-off values for the rate of change in BW for type 2 diabetes remission were 0.59 and 5.0%.

DISCUSSION

Body weight loss of ≥5% effectively achieved diabetes remission in Japanese men with a BMI ≥25 kg/m² and new-onset type 2 diabetes.

Identification of Risk Factors for Daptomycin-Associated Creatine Phosphokinase Elevation and Development of a Risk Prediction Model for Incidence Probability

BACKGROUND

In this study, we investigated the risk factors for daptomycin-associated creatine phosphokinase (CPK) elevation and established a risk score for CPK elevation.

METHODS

Patients who received daptomycin at our hospital were classified into the non-elevated or elevated CPK group based on their peak CPK levels during daptomycin therapy. Univariable and multivariable analyses were performed, and a risk score and prediction model for the incidence probability of CPK elevation were calculated based on logistic regression analysis.

RESULTS

The non-elevated and elevated CPK groups included 181 and 17 patients, respectively. Logistic regression analysis revealed that concomitant statin use (odds ratio [OR], 4.45 [95% confidence interval {CI}, 1.40-14.47]; risk score 4), concomitant antihistamine use (OR, 5.66 [95% CI, 1.58-20.75]; risk score 4), and trough concentration (Cmin) between 20 and <30 µg/mL (OR, 14.48 [95% CI, 2.90-87.13]; risk score 5) and ≥30.0 µg/mL (OR, 24.64 [95% CI, 3.21-204.53]; risk score 5) were risk factors for daptomycin-associated CPK elevation. The predicted incidence probabilities of CPK elevation were <10% (low risk), 10%-<25% (moderate risk), and ≥25% (high risk) with total risk scores of ≤4, 5-6, and ≥8, respectively. The risk prediction model exhibited a good fit (area under the receiver operating characteristic curve, 0.85 [95% CI, .74-.95]).

CONCLUSIONS

These results suggested that concomitant use of statins with antihistamines and Cmin ≥20 µg/mL were risk factors for daptomycin-associated CPK elevation. Our prediction model might aid in reducing the incidence of daptomycin-associated CPK elevation.

Fasting plasma glucose level in the range of 90-99 mg/dL and the risk of the onset of type 2 diabetes: Population-based Panasonic cohort study 2

Aim/Introduction

As the association between a fasting glucose concentration of 90–99 mg/dL and the onset of type 2 diabetes is still controversial, we aimed to assess it in 37,148 Japanese individuals with a normal plasma glucose concentration.

Materials and Methods

This long‐term retrospective cohort study included individuals having a medical checkup at Panasonic Corporation from 2008 to 2018. In total, 1,028 participants developed type 2 diabetes.

Results

Cox regression analyses revealed that the risk for the onset of diabetes increased by 9.0% per 1 mg/dL increase in fasting plasma glucose concentration in subjects with the concentration ranging from 90 to 99 mg/dL. Compared with individuals with a fasting glucose concentration of ≤89 mg/dL, the adjusted hazard ratios for developing diabetes were 1.53 (95% CI; 1.22–1.91), 1.76 (95% CI; 1.41–2.18), 1.89 (95% CI; 1.52–2.35), 3.17 (95% CI; 2.61–3.84), and 3.41 (95% CI; 2.79–4.15) at fasting plasma glucose concentrations of 90–91, 92–93, 94–95, 96–97, and 98–99 mg/dL, respectively. In populations with obesity, the adjusted hazards ratios for developing diabetes were 1.56 (95% CI; 1.15–2.09), 1.82 (95% CI; 1.37–2.40), 2.05 (95% CI; 1.55–2.69), 3.53 (95% CI; 2.79–4.46), and 3.28 (95% CI; 2.53–4.22) at fasting plasma glucose concentrations of 90–91, 92–93, 94–95, 96–97, and 98–99 mg/dL, respectively.

Conclusions

This study demonstrates that the risk of type 2 diabetes among subjects having a fasting plasma glucose concentration of 90–99 mg/dL, is progressively higher with an increasing level of fasting plasma glucose concentration in a Japanese people.

Prediction Models for Type 2 Diabetes Risk in the General Population: A Systematic Review of Observational Studies

OBJECTIVES

This study aimed to provide an overview of prediction models of undiagnosed type 2 diabetes mellitus (U-T2DM) or the incident T2DM (I-T2DM) using the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) checklist and the prediction model risk of the bias assessment tool (PROBAST).

DATA SOURCES

Both PUBMED and EMBASE databases were searched to guarantee adequate and efficient coverage.

STUDY SELECTION

Articles published between December 2011 and October 2019 were considered.

DATA EXTRACTION

For each article, information on model development requirements, discrimination measures, calibration, overall performance, clinical usefulness, overfitting, and risk of bias (ROB) was reported.

RESULTS

The median (interquartile range; IQR) number of the 46 study populations for model development was 5711 (1971 - 27426) and 2457 (2060 - 6995) individuals for I-T2DM and U-T2DM, respectively. The most common reported predictors were age and body mass index, and only the Qrisk-2017 study included social factors (e.g., Townsend score). Univariable analysis was reported in 46% of the studies, and the variable selection procedure was not clear in 17.4% of them. Moreover, internal and external validation was reported in 43% the studies, while over 63% of them reported calibration. The median (IQR) of AUC for I-T2DM models was 0.78 (0.74 - 0.82); the corresponding value for studies derived before October 2011 was 0.80 (0.77 - 0.83). The highest discrimination index was reported for Qrisk-2017 with C-statistics of 0.89 for women and 0.87 for men. Low ROB for I-T2DM and U-T2DM was assessed at 18% and 41%, respectively.

CONCLUSIONS

Among prediction models, an intermediate to poor quality was reassessed in several aspects of model development and validation. Generally, despite its new risk factors or new methodological aspects, the newly developed model did not increase our capability in screening/predicting T2DM, mainly in the analysis part. It was due to the lack of external validation of the prediction models.

Random forest approach for determining risk prediction and predictive factors of type 2 diabetes: large-scale health check-up data in Japan

Introduction

Early intervention in type 2 diabetes can prevent exacerbation of insulin resistance. More effective interventions can be implemented by early and precise prediction of the change in glycated haemoglobin A1c (HbA1c). Artificial intelligence (AI), which has been introduced into various medical fields, may be useful in predicting changes in HbA1c. However, the inability to explain the predictive factors has been a problem in the use of deep learning, the leading AI technology. Therefore, we applied a highly interpretable AI method, random forest (RF), to large-scale health check-up data and examined whether there was an advantage over a conventional prediction model.

Research design and methods

This study included a cumulative total of 42 908 subjects not receiving treatment for diabetes with an HbA1c <6.5%. The objective variable was the change in HbA1c in the next year. Each prediction model was created with 51 health-check items and part of their change values from the previous year. We used two analytical methods to compare the predictive powers: RF as a new model and multivariate logistic regression (MLR) as a conventional model. We also created models excluding the change values to determine whether it positively affected the predictions. In addition, variable importance was calculated in the RF analysis, and standard regression coefficients were calculated in the MLR analysis to identify the predictors.

Results

The RF model showed a higher predictive power for the change in HbA1c than MLR in all models. The RF model including change values showed the highest predictive power. In the RF prediction model, HbA1c, fasting blood glucose, body weight, alkaline phosphatase and platelet count were factors with high predictive power.

Conclusions

Correct use of the RF method may enable highly accurate risk prediction for the change in HbA1c and may allow the identification of new diabetes risk predictors.

Screening Model for Estimating Undiagnosed Diabetes among People with a Family History of Diabetes Mellitus: A KNHANES-Based Study

A screening model for estimating undiagnosed diabetes mellitus (UDM) is important for early medical care. There is minimal research and a serious lack of screening models for people with a family history of diabetes (FHD), especially one which incorporates gender characteristics. Therefore, the primary objective of our study was to develop a screening model for estimating UDM among people with FHD and enable its validation. We used data from the Korean National Health and Nutrition Examination Survey (KNHANES). KNAHNES (2010-2016) was used as a developmental cohort (n = 5939) and was then evaluated in a validation cohort (n = 1047) KNHANES (2017). We developed the screening model for UDM in male (SMM), female (SMF), and male and female combined (SMP) with FHD using backward stepwise logistic regression analysis. The SMM and SMF showed an appropriate performance (area under curve (AUC) = 76.2% and 77.9%) compared with SMP (AUC = 72.9%) in the validation cohort. Consequently, simple screening models were developed and validated, for the estimation of UDM among patients in the FHD group, which is expected to reduce the burden on the national health care system.

Elevated alanine aminotransferase and low aspartate aminotransferase/alanine aminotransferase ratio are associated with chronic kidney disease among middle-aged women: a cross-sectional study

BACKGROUND

Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) to ALT ratio (AST/ALT ratio) have been shown to be related to non-alcoholic fatty liver disease or insulin resistance, which was associated with chronic kidney disease (CKD). However, it is unclear whether ALT and AST/ALT ratio are associated with CKD. In this study, we examined the relationship of ALT and AST/ALT ratio to CKD among middle-aged females in Japan.

METHODS

The present study included 29,133 women aged 40 to 64 years who had an annual health checkup in Japan during April 2013 to March 2014. Venous blood samples were collected to measure ALT, AST, gamma-glutamyltransferase (GGT), and creatinine levels. In accordance with previous studies, ALT > 40 U/L and GGT > 50 U/L were determined as elevated, AST/ALT ratio < 1 was regarded as low, and CKD was defined as estimated glomerular filtration rate < 60 mL/min/1.73 m² and/or proteinuria. Logistic regression model was used to calculate the odds ratio (OR) and 95% confidence interval (CI) for CKD.

RESULTS

"Elevated ALT and elevated GGT" and "elevated ALT and non-elevated GGT" significantly increased the OR for CKD when compared with "non-elevated ALT and non-elevated GGT" (OR: 2.56, 95% CI: 2.10-3.12 and OR: 2.24, 95% CI: 1.81-2.77). Compared with "AST/ALT ratio ≥ 1 and non-elevated GGT", "AST/ALT ratio < 1 and elevated GGT" and "AST/ALT ratio < 1 and non-elevated GGT" significantly increased the OR for CKD (OR: 2.73, 95% CI: 2.36-3.15 and OR: 1.68, 95% CI: 1.52-1.87). These findings still remained after adjustment for confounders.

CONCLUSIONS

Elevated ALT was associated with CKD regardless of GGT elevation. Moreover, low AST/ALT ratio was also associated with CKD independent of GGT elevation.

Cardiometabolic risk prediction algorithms for young people with psychosis: a systematic review and exploratory analysis

OBJECTIVE

Cardiometabolic risk prediction algorithms are common in clinical practice. Young people with psychosis are at high risk for developing cardiometabolic disorders. We aimed to examine whether existing cardiometabolic risk prediction algorithms are suitable for young people with psychosis.

METHODS

We conducted a systematic review and narrative synthesis of studies reporting the development and validation of cardiometabolic risk prediction algorithms for general or psychiatric populations. Furthermore, we used data from 505 participants with or at risk of psychosis at age 18 years in the ALSPAC birth cohort, to explore the performance of three algorithms (QDiabetes, QRISK3 and PRIMROSE) highlighted as potentially suitable. We repeated analyses after artificially increasing participant age to the mean age of the original algorithm studies to examine the impact of age on predictive performance.

RESULTS

We screened 7820 results, including 110 studies. All algorithms were developed in relatively older participants, and most were at high risk of bias. Three studies (QDiabetes, QRISK3 and PRIMROSE) featured psychiatric predictors. Age was more strongly weighted than other risk factors in each algorithm. In our exploratory analysis, calibration plots for all three algorithms implied a consistent systematic underprediction of cardiometabolic risk in the younger sample. After increasing participant age, calibration plots were markedly improved.

CONCLUSION

Existing cardiometabolic risk prediction algorithms cannot be recommended for young people with or at risk of psychosis. Existing algorithms may underpredict risk in young people, even in the face of other high-risk features. Recalibration of existing algorithms or a new tailored algorithm for the population is required.

Risk factors during the early postpartum period for type 2 diabetes mellitus in women with gestational diabetes

For women with gestational diabetes mellitus (GDM), the evaluation of glucose tolerance (GT) in the early postpartum period is universally recommended. Nevertheless, few studies have evaluated the risk factors for T2DM on the basis of GT data obtained during the early postpartum period. We aimed to identify the risk factors for type 2 diabetes mellitus (T2DM) by evaluating GT in the first 12 weeks postpartum (12wPP) in women with GDM and to categorize the risk using a combination of the principal risk factors. This retrospective multicenter observational study included 399 East Asian women with GDM who underwent a 75-g oral glucose tolerance test (OGTT) within 12wPP, which was repeated annually or biennially and used to identify the postpartum development of T2DM. Forty-three women (10.8%) developed T2DM during a median follow-up period of 789 ± 477 days. The independent risk factors for T2DM were pre-pregnancy obesity (BMI ≥25 kg/m²), early postpartum impairment in glucose tolerance (IGT), and an early postpartum glycated hemoglobin (HbA1c) ≥5.7%. The odds ratios (95% confidence intervals) for T2DM were 3.2 (1.3-7.8) in women with either early postpartum IGT or pre-pregnancy obesity, 9.2 (3.0-28.3) in those with early postpartum IGT, pre-pregnancy obesity, and HbA1c <5.7%, and 51.4 (16.1-163.9) in those with early postpartum IGT, pre-pregnancy obesity, and HbA1c ≥5.7%, compared with those without obesity or IGT. T2DM risk in East Asian women with GDM should be stratified according to pre-pregnancy obesity and early postpartum IGT, and these patients should be followed up and receive appropriate care for their risk category.

Longitudinal Trajectories of Fasting Plasma Glucose and Risks of Cardiovascular Diseases in Middle Age to Elderly People Within the General Japanese Population: The Suita Study

Background

Few previous studies used information on changes in fasting plasma glucose (FPG) assessed at multiple points in time in relationship to cardiovascular disease (CVD) incidence. The present study aimed to identify subgroups of FPG trajectories with assessing CVD incidence.

Methods and Results

The present study was based on the Suita study, a population‐based cohort study in Japan. The primary outcome was incidence of the first CVD events consisting of stroke and coronary heart diseases between 1989 and 2013. The main exposure was FPG assessed every 2 years. We used joint latent class mixed models to derive FPG trajectories over time while evaluating cumulative incidence of CVD, and categorized participants into several subgroups based on those trajectories and cumulative incidence. We observed 356 and 243 CVD events during the median follow‐up of 17.2 and 20.2 years among 3120 men and 3482 women, respectively. The joint latent mixed models found 3 subgroups in men and 2 subgroups in women. Of the 3 subgroups in men, 1 subgroup had FPG levels that increased sharply (96.5–205.0 mg/dL from aged 40 to 80 years) and higher CVD cumulative incidence. Of the 2 subgroups in women, 1 subgroup had FPG levels that increased sharply (97.7–190.5 mg/dL from aged 40 to 80 years) and tended to have slightly higher CVD incidence compared with the other subgroup.

Conclusion

It can be important to manage CVD risk factors especially for people whose FPG trajectories sharply increased to prevent CVD.

A Deep Learning Model for Estimation of Patients with Undiagnosed Diabetes

A screening model for undiagnosed diabetes mellitus (DM) is important for early medical care. Insufficient research has been carried out developing a screening model for undiagnosed DM using machine learning techniques. Thus, the primary objective of this study was to develop a screening model for patients with undiagnosed DM using a deep neural network. We conducted a cross-sectional study using data from the Korean National Health and Nutrition Examination Survey (KNHANES) 2013–2016. A total of 11,456 participants were selected, excluding those with diagnosed DM, an age < 20 years, or missing data. KNHANES 2013–2015 was used as a training dataset and analyzed to develop a deep learning model (DLM) for undiagnosed DM. The DLM was evaluated with 4444 participants who were surveyed in the 2016 KNHANES. The DLM was constructed using seven non-invasive variables (NIV): age, waist circumference, body mass index, gender, smoking status, hypertension, and family history of diabetes. The model showed an appropriate performance (area under curve (AUC): 80.11) compared with existing previous screening models. The DLM developed in this study for patients with undiagnosed diabetes could contribute to early medical care.

Cross-sectional study of associations between normal body weight with central obesity and hyperuricemia in Japan

BACKGROUND

Several studies have shown that normal weight with central obesity (NWCO) is associated with cardiovascular disease risk factors such as hypertension, dyslipidemia and diabetes. However, the relationship between NWCO and hyperuricemia has not been studied in detail.

METHODS

We investigated the association between NWCO and hyperuricemia among Japanese adults aged 40-64 years who had undergone periodic health examinations between April 2013 and March 2014. Obesity was defined as a body mass index (BMI) ≥25 kg/m² and central obesity was determined as a waist-to-height ratio (WHtR) ≥0.5. We classified the participants into the following groups based according to having obesity and central obesity: normal weight (BMI 18.5-24.9 kg/m²) without (NW; WHtR < 0.5) and with (NWCO) central obesity, and obesity without (OB) and with (OBCO) central obesity. Hyperuricemia was defined as serum uric acid > 7.0 and ≥ 6.0 mg/dL in men and women, respectively, or under medical treatment for hyperuricemia. Alcohol intake was classified as yes (daily and occasional consumption) and none (no alcohol consumption). Odds ratios (OR) and 95% confidence intervals (CI) for hyperuricemia were calculated using a logistic regression model.

RESULTS

We analyzed data derived from 96,863 participants (69,241 men and 27,622 women). The prevalences of hyperuricemia in men and women were respectively, 21.4 and 11.0%, and of participants with NWCO respectively 15.6 and 30.0%. The adjusted OR for hyperuricemia was significantly increased in OBCO compared with NW, regardless of sex (men: OR, 2.12; 95%CI; 2.03-2.21; women: OR, 3.54; 95%CI, 3.21-3.90) and were statistically significant in NWCO compared with NW (men: OR, 1.44; 95%CI, 1.36-1.52; women: OR, 1.41; 95%CI, 1.27-1.57). The results were similar regardless of alcohol consumption.

CONCLUSIONS

We found that NWCO and OBCO were associated with hyperuricemia in middle-aged Japanese men and women. Middle-aged Japanese adults with normal weight but having central obesity should be screened using a combination of BMI and WHtR and educated about how to prevent hyperuricemia.

Glycosylated Hemoglobin A1c Improves the Performance of the Nomogram for Predicting the 5-Year Incidence of Type 2 Diabetes

AIM

To develop and validate a model, which combines traditional risk factors and glycosylated hemoglobin A1c (HbA1c) for predicting the risk of type 2 diabetes (T2DM).

MATERIALS AND METHODS

This is a historical cohort study from a collected database, which included 8419 males and 7034 females without diabetes at baseline with a median follow-up of 5.8-years and 5.1-years, respectively. Multivariate cox regression analysis was used to select significant prognostic factors of T2DM. Two nomograms were constructed to predict the 5-year incidence of T2DM based on traditional risk factors (Model 1) and traditional risk factors plus HbA1c (Model 2). C-index, calibration curve, and time-dependent receiver-operating characteristic (ROC) curve were conducted in the training sets and validation sets.

RESULTS

In males, the C-index was 0.824 (95% CI: 0.795-0.853) in Model 1 and 0.867 (95% CI: 0.840-0.894) in Model 2; in females, the C-index was 0.830 (95% CI: 0.770-0.890) in Model 1 and 0.856 (95% CI: 0.795-0.917) in Model 2. The areas under curve (AUC) in Model 2 for prediction of T2DM development were higher than in Model 1 at each time point. The calibration curves showed excellent agreement between the predicted possibility and the actual observation in both models. The results of validation sets were similar to the results of training sets.

CONCLUSION

The proposed nomogram can be used to accurately predict the risk of T2DM. Compared with the traditional nomogram, HbA1c can improve the performance of nomograms for predicting the 5-year incidence of T2DM.

Associations between normal weight central obesity and cardiovascular disease risk factors in Japanese middle-aged adults: a cross-sectional study

BACKGROUND

Several studies have shown that normal weight central obesity (NWCO) is associated with cardiovascular disease (CVD) risk factors. However, studies conducted in the Japanese population have been very limited. Thus, the relationships between normal weight central obesity, classified using body mass index (BMI), the waist-to-height ratio (WHtR), and CVD risk factors in middle-aged Japanese adults were investigated.

METHODS

The participants were Japanese adults aged 40-64 years who had undergone periodic health examinations in Japan during the period from April 2013 to March 2014. The participants were categorized into the following four groups: normal weight (BMI 18.5-24.9 kg/m²) and no central obesity (WHtR < 0.5) (NW); normal weight and central obesity (WHtR ≥ 0.5) (NWCO); obesity (BMI ≥ 25 kg/m²) and no central obesity (OB); and obesity and central obesity (OBCO). Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg, or taking medication for hypertension. Dyslipidemia was defined as LDL-C ≥ 140 mg/dl, HDL-C < 40 mg/dl, triglyceride ≥ 150 mg/dl, or taking medication for dyslipidemia. Diabetes was defined as fasting plasma glucose ≥ 126 mg/dl, random plasma glucose ≥ 200 mg/dl, HbA1c ≥ 6.5%, or receiving medical treatment for diabetes mellitus. A logistic regression model was used to calculate the odds ratios (ORs) and 95% confidence intervals (CIs) for hypertension, dyslipidemia, and diabetes.

RESULTS

A total of 117,163 participants (82,487 men and 34,676 women) were analyzed. The prevalence of NWCO was 15.6% in men and 30.2% in women. With reference to NW, the ORs for hypertension (adjusted OR 1.22, 95% CI 1.17-1.27 in men, 1.23, 1.16-1.31 in women), dyslipidemia (1.81, 1.74-1.89 in men, 1.60, 1.52-1.69 in women), and diabetes (1.35, 1.25-1.46 in men, 1.60, 1.35-1.90 in women) were significantly higher in NWCO.

CONCLUSIONS

Normal weight with central obesity was associated with CVD risk factors, such as hypertension, dyslipidemia, and diabetes, compared with normal weight without central obesity, regardless of sex. It is important to focus on normal weight with central obesity for the prevention of CVD in Japanese middle-aged adults.

Effectiveness of a type 2 diabetes prevention program combining FINDRISC scoring and telephone-based coaching in the French population of bakery/pastry employees

Background/objectives

Preventive actions targeting the risk of type 2 diabetes mellitus (T2D) and deployed from the workplace are scarce. This study aimed to measure this T2D risk in a large sample of the bakery/pastry employees in France and to assess the effectiveness of a telephone coaching program in participants with the highest risk.

Subjects/methods

A screening survey using the FINDRISC score was conducted by phone among the employees. Those with a moderate risk (score ≥ 12 and <15; body mass index ≥ 25 kg/m²) or high/very high risk (score ≥ 15) were invited to participate in a 6-month coaching program including 6 monthly interviews together with a final evaluation interview three months later. The effects and impact were evaluated using 8 questions on dietary knowledge/behavior as well as the GPAQ (physical activity) and SF-12 (quality of life) questionnaires.

Results

There were 19,951 employees eligible for screening (age: 38.0 ± 13.5 years, men 49.6%, mean FINDRISC score 5.9 ± 4.4). A high/very high score was found in 4% of individuals. Overall, 1,348 (among 2,018) eligible employees agreed to participate in the coaching program, 630 of whom participated in all interviews. Of the latter, dietary knowledge/behavior (+1.60) and quality of life (+1.83) improved (P < 0.0001), with a favorable trend for physical activity (+0.06, P = 0.0756). Dietary knowledge/behavior continued to improve in the 581 completers (+0.17, P = 0.0001).

Conclusions

This two-step prevention program associating T2D risk estimation and a 6-month telephone coaching was deployed in the French craft bakery/pastry sector with significant adhesion. Such program appears beneficial for enhancing knowledge and mobilizing skills associated with T2D prevention.

Random plasma glucose predicts the diagnosis of diabetes

Aims/Hypothesis

Early recognition of those at high risk for diabetes as well as diabetes itself can permit preventive management, but many Americans with diabetes are undiagnosed. We sought to determine whether routinely available outpatient random plasma glucose (RPG) would be useful to facilitate the diagnosis of diabetes.

Methods

Retrospective cohort study of 942,446 U.S. Veterans without diagnosed diabetes, ≥3 RPG in a baseline year, and ≥1 primary care visit/year during 5-year follow-up. The primary outcome was incident diabetes (defined by diagnostic codes and outpatient prescription of a diabetes drug).

Results

Over 5 years, 94,599 were diagnosed with diabetes [DIAB] while 847,847 were not [NONDIAB]. Baseline demographics of DIAB and NONDIAB were clinically similar, except DIAB had higher BMI (32 vs. 28 kg/m²) and RPG (150 vs. 107 mg/dl), and were more likely to have Black race (18% vs. 15%), all p<0.001. ROC area for prediction of DIAB diagnosis within 1 year by demographic factors was 0.701, and 0.708 with addition of SBP, non-HDL cholesterol, and smoking. These were significantly less than that for prediction by baseline RPG alone (≥2 RPGs at/above a given level, ROC 0.878, p<0.001), which improved slightly when other factors were added (ROC 0.900, p<0.001). Having ≥2 RPGs ≥115 mg/dl had specificity 77% and sensitivity 87%, and ≥2 RPGs ≥130 mg/dl had specificity 93% and sensitivity 59%. For predicting diagnosis within 3 and 5 years by RPG alone, ROC was reduced but remained substantial (ROC 0.839 and 0.803, respectively).

Conclusions

RPG levels below the diabetes “diagnostic” range (≥200 mg/dl) provide good discrimination for follow-up diagnosis. Use of such levels–obtained opportunistically, during outpatient visits–could signal the need for further testing, allow preventive intervention in high risk individuals before onset of disease, and lead to earlier identification of diabetes.

Association of the combination of weight gain after 20 years of age and current obesity with chronic kidney disease in Japan: a cross-sectional study

OBJECTIVES

Weight gain after 20 years of age is associated with chronic kidney disease (CKD). However, the impact of weight gain on CKD might differ by current obesity status. We investigated the association of the combination of weight gain after 20 years of age and current obesity with CKD among adults in Japan.

DESIGN

A cross-sectional study.

SETTING AND PARTICIPANTS

We analysed data from 94 822 adults aged 40-64 years who had an annual health check-up in Japan from April 2013 to March 2014.

PRIMARY OUTCOME MEASURE

CKD was defined as an estimated glomerular filtration rate <60 mL/min/1.73 m² and/or proteinuria.

RESULTS

Both weight gain ≥10 kg after 20 years of age plus obesity (OR 2.21, 95% CI 2.07 to 2.36) and weight gain of ≥10 kg plus non-obesity (OR 1.31, 95% CI 1.21 to 1.42) significantly increased the OR for CKD when compared with weight gain <10 kg plus non-obesity in men. In women, weight gain ≥10 kg plus obesity (OR 2.04, 95% CI 1.84 to 2.25) and weight gain ≥10 kg plus non-obesity (OR 1.53, 95% CI 1.36 to 1.72) significantly increased the OR for CKD compared with weight gain <10 kg plus non-obesity. These results persisted even after adjustment for age, lifestyle factors, hypertension, dyslipidaemia and diabetes.

CONCLUSIONS

Weight gain ≥10 kg after 20 years of age was significantly associated with CKD in both obese and non-obese subjects. Moreover, the influence of weight gain ≥10 kg plus obesity on CKD was greater than that of weight gain ≥10 kg plus non-obesity on CKD. The present study results suggest that it is important to consider weight gain after maturity in both obese and non-obese subjects to prevent CKD among Japanese middle-aged adults.

New risk prediction model of coronary heart disease in participants with and without diabetes: Assessments of the Framingham risk and Suita scores in 3-year longitudinal database in a Japanese population

The Framingham Risk Score (FRS) has been reported to predict coronary heart disease (CHD), but its assessment has been unsuccessful in Asian population. We aimed to assess FRS and Suita score (a Japanese CHD prediction model) in a Japanese nation-wide annual health check program, participants aged 40-79 years were followed up longitudinally from 2008 to 2011. Of 35,379 participants analyzed, 1,234 had new-onset CHD. New-onset CHD was observed in diabetic men [6.00%], non-diabetic men [3.96%], diabetic women [5.51%], and non-diabetic women [2.86%], respectively. Area under the curve (AUC) of receiver operating characteristic (ROC) curve for CHD prediction were consistently low in Suita score (TC), FRS (TC) and NCEP-ATPIII FRS (TC), suggesting that these scores have only a limited power. ROC, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision curve analysis (DCA) and Hosmer-Lemeshow goodness-of-fit test did not show clear differences between Suita score (TC) and FRS (TC). New models combining waist circumference ≥85 cm in men or proteinuria ≥1+ in women to Suita score (TC) was superior in diabetic men and women. New models could be useful to predict 3-year risk of CHD at least in Japanese population especially in diabetic population.

DEXAMETHASONE STRESS TEST: A PILOT CLINICAL STUDY FOR IDENTIFICATION OF INDIVIDUALS HIGHLY PRONE TO DEVELOP TYPE 2 DIABETES

OBJECTIVE

We examined whether the "Dexamethasone Stress Test" exhibits the requisite high predictive ability to identify individuals highly prone to develop type 2 diabetes mellitus (T2DM).

METHODS

Seven years ago, we administered an oral glucose tolerance test (OGTT) to 33 individuals without T2DM and repeated the OGTT 24 hours after a single oral dose of 8 mg dexamethasone (Dex); all participants had a first-degree relative with T2DM, and close to half had prediabetes. We calculated receiver operating characteristic (ROC) curves for all parameters derived from the OGTT before and after Dex in individuals who subsequently developed diabetes compared to individuals who did not.

RESULTS

At 7 years of follow-up, 9 individuals had developed T2DM, while 24 remained without diabetes. None of the OGTT-derived parameters before administration of Dex had an area under the ROC curve of >0.8. However, 24 hours after Dex, three parameters, including fasting plasma insulin, homeostatic model assessment-insulin resistance, and 2-hour plasma glucose level, exhibited areas under the ROC curves of 0.84, 0.86, and 0.92, respectively.

CONCLUSION

The Dexamethasone Stress Test appears to be a good to excellent test in identifying individuals highly prone to develop T2DM.

ABBREVIATIONS

AUC = area under the curve; Dex = dexamethasone; HOMA-IR = homeostatic model assessment-insulin resistance; NGT = normal glucose tolerance; OGTT = oral glucose tolerance test; PreDiab = prediabetes; ROC = receiver operating characteristic; T2DM = type 2 diabetes mellitus.

Development and validation of risk models to predict the 7-year risk of type 2 diabetes: The Japan Epidemiology Collaboration on Occupational Health Study

AIMS/INTRODUCTION

We previously developed a 3-year diabetes risk score in the working population. The objective of the present study was to develop and validate flexible risk models that can predict the risk of diabetes for any arbitrary time-point during 7 years.

MATERIALS AND METHODS

The participants were 46,198 Japanese employees aged 30-59 years, without diabetes at baseline and with a maximum follow-up period of 8 years. Incident diabetes was defined according to the American Diabetes Association criteria. With routine health checkup data (age, sex, abdominal obesity, body mass index, smoking status, hypertension status, dyslipidemia, glycated hemoglobin and fasting plasma glucose), we developed non-invasive and invasive risk models based on the Cox proportional hazards regression model among a random two-thirds of the participants, and used another one-third for validation.

RESULTS

The range of the area under the receiver operating characteristic curve increased from 0.73 (95% confidence interval 0.72-0.74) for the non-invasive prediction model to 0.89 (95% confidence interval 0.89-0.90) for the invasive prediction model containing dyslipidemia, glycated hemoglobin and fasting plasma glucose. The invasive models showed improved integrated discrimination and reclassification performance, as compared with the non-invasive model. Calibration appeared good between the predicted and observed risks. These models performed well in the validation cohort.

CONCLUSIONS

The present non-invasive and invasive models for the prediction of diabetes risk up to 7 years showed fair and excellent performance, respectively. The invasive models can be used to identify high-risk individuals, who would benefit greatly from lifestyle modification for the prevention or delay of diabetes.

Diabetes Among Non-Overweight Individuals: an Emerging Public Health Challenge

PURPOSE OF REVIEW

Overweight and obesity are well-established risk factors for type 2 diabetes. However, a substantial number of individuals develop the disease at underweight or normal weight. In this review, we discuss the epidemiology of type 2 diabetes in non-overweight adults; pose questions about etiology, pathophysiology, diagnosis, and prognosis; and examine implications for prevention and treatment.

RECENT FINDINGS

In population-based studies, the prevalence of type 2 diabetes ranged from 1.4-10.9%. However, the prevalence of type 2 diabetes in individuals with BMI < 25 kg/m² ranged from 1.4-8.8%. In countries from Asia and Africa, the proportion of individuals with diabetes who were underweight or normal weight ranged from 24 to 66%, which is considerably higher than the US proportion of 10%. Impairments in insulin secretion, in utero undernutrition, and epigenetic alterations to the genome may play a role in diabetes development in this subgroup. A substantial number of individuals with type 2 diabetes, particularly those with recent ancestry from Asia or Africa, are underweight or normal weight. Future research should consist of comprehensive studies of the prevalence of type 2 diabetes in non-overweight individuals; studies aimed at understanding gaps in the mechanisms, etiology, and pathophysiology of diabetes development in underweight or normal weight individuals; and trials assessing the effectiveness of interventions in this population.

Correction: Development of Risk Score for Predicting 3-Year Incidence of Type 2 Diabetes: Japan Epidemiology Collaboration on Occupational Health Study

[This corrects the article DOI: 10.1371/journal.pone.0142779.].

Predictive performance of a genetic risk score using 11 susceptibility alleles for the incidence of Type 2 diabetes in a general Japanese population: a nested case-control study

AIMS

To assess the predictive ability of a genetic risk score for the incidence of Type 2 diabetes in a general Japanese population.

METHODS

This prospective case-control study, nested within a Japan Public Health Centre-based prospective study, included 466 participants with incident Type 2 diabetes over a 5-year period (cases) and 1361 control participants, as well as 1463 participants with existing diabetes and 1463 control participants. Eleven susceptibility single nucleotide polymorphisms, identified through genome-wide association studies and replicated in Japanese populations, were analysed.

RESULTS

Most single nucleotide polymorphism loci showed directionally consistent associations with diabetes. From the combined samples, one single nucleotide polymorphism (rs2206734 at CDKAL1) reached a genome-wide significance level (odds ratio 1.28, 95% CI 1.18-1.40; P = 1.8 × 10^-8 ). Three single nucleotide polymorphisms (rs2206734 in CDKAL1, rs2383208 in CDKN2A/B, and rs2237892 in KCNQ1) were nominally significantly associated with incident diabetes. Compared with the lowest quintile of the total number of risk alleles, the highest quintile had a higher odds of incident diabetes (odds ratio 2.34, 95% CI 1.59-3.46) after adjusting for conventional risk factors such as age, sex and BMI. The addition to the conventional risk factor-based model of a genetic risk score using the 11 single nucleotide polymorphisms significantly improved predictive performance; the c-statistic increased by 0.021, net reclassification improved by 6.2%, and integrated discrimination improved by 0.003.

CONCLUSIONS

Our prospective findings suggest that the addition of a genetic risk score may provide modest but significant incremental predictive performance beyond that of the conventional risk factor-based model without biochemical markers.

A Point System for Predicting 10-Year Risk of Developing Type 2 Diabetes Mellitus in Japanese Men: Aichi Workers' Cohort Study

Background

Relatively little evidence exists for type 2 diabetes mellitus (T2DM) prediction models from long-term follow-up studies in East Asians. This study aims to develop a point-based prediction model for 10-year risk of developing T2DM in middle-aged Japanese men.

Methods

We followed 3,540 male participants of Aichi Workers’ Cohort Study, who were aged 35–64 years and were free of diabetes in 2002, until March 31, 2015. Baseline age, body mass index (BMI), smoking status, alcohol consumption, regular exercise, medication for dyslipidemia, diabetes family history, and blood levels of triglycerides (TG), high density lipoprotein cholesterol (HDLC) and fasting blood glucose (FBG) were examined using Cox proportional hazard model. Variables significantly associated with T2DM in univariable models were simultaneously entered in a multivariable model for determination of the final model using backward variable selection. Performance of an existing T2DM model when applied to the current dataset was compared to that obtained in the present study’s model.

Results

During the median follow-up of 12.2 years, 342 incident T2DM cases were documented. The prediction system using points assigned to age, BMI, smoking status, diabetes family history, and TG and FBG showed reasonable discrimination (c-index: 0.77) and goodness-of-fit (Hosmer-Lemeshow test, P = 0.22). The present model outperformed the previous one in the present subjects.

Conclusion

The point system, once validated in the other populations, could be applied to middle-aged Japanese male workers to identify those at high risk of developing T2DM. In addition, further investigation is also required to examine whether the use of this system will reduce incidence.

Sleep Duration Modifies the Association of Overtime Work With Risk of Developing Type 2 Diabetes: Japan Epidemiology Collaboration on Occupational Health Study

Background

Evidence linking working hours and the risk of type 2 diabetes mellitus (T2DM) is limited and inconsistent in Asian populations. No study has addressed the combined association of long working hours and sleep deprivation on T2DM risk. We investigated the association of baseline overtime work with T2DM risk and assessed whether sleep duration modified the effect among Japanese.

Methods

Participants were Japanese employees (28,489 men and 4,561 women) aged 30–64 years who reported overtime hours and had no history of diabetes at baseline (mostly in 2008). They were followed up until March 2014. New-onset T2DM was identified using subsequent checkup data, including measurement of fasting/random plasma glucose, glycated hemoglobin, and self-report of medical treatment. Hazard ratios (HRs) of T2DM were estimated using Cox regression analysis. The combined association of sleep duration and working hours was examined in a subgroup of workers (n = 27,590).

Results

During a mean follow-up period of 4.5 years, 1,975 adults developed T2DM. Overtime work was not materially associated with T2DM risk. In subgroup analysis, however, long working hours combined with insufficient sleep were associated with a significantly higher risk of T2DM (HR 1.42; 95% CI, 1.11–1.83), whereas long working hours with sufficient sleep were not (HR 0.99; 95% CI, 0.88–1.11) compared with the reference (<45 hours of overtime with sufficient sleep).

Conclusions

Sleep duration modified the association of overtime work with the risk of developing T2DM. Further investigations to elucidate the long-term effect of long working hours on glucose metabolism are warranted.

Postchallenge hyperglycemia in subjects with low body weight: implication for small glucose volume

A hypothesis that postchallenge hyperglycemia in subjects with low body weight (BW) may be due, in part, to small glucose volume (G_V) was tested. We studied 11,411 nondiabetic subjects with a mean BW of 63.3 kg; 5,282 of them were followed for a mean of 5.3 yr. In another group of 1,537 nondiabetic subjects, insulin sensitivity, secretion, and a product of the two (index of whole body insulin action) were determined. Corrected 2 h-plasma glucose (2hPG_corr) during a 75-g oral glucose tolerance test in subjects with BW ≤ 59 kg was calculated as 2hPG_corr = δPG_2h · ECW/[16.1 (males) or 15.3 (females)] + fasting PG (FPG), where δPG_2h is plasma glucose increment in 2 h; ECW is extracellular water (surrogate of G_V); FPG is fasting plasma glucose; and 16.1 and 15.3 are ECW of men and women, respectively, with BW = 59 kg. Multivariate analyses for BW with adjustment for age, sex, and percent body fat were undertaken. BW was, across its entire range, positively correlated with FPG (P < 0.01). Whereas BW was correlated with 2hPG and δPG in a skewed J-shape, with inflections at around 60 kg (P for nonlinearity < 0.01 for each). Nonetheless, in those with BW ≤ 59 kg, insulin sensitivity, secretion, and action were unattenuated, and incident diabetes was less compared with heavier counterparts. BW was linearly correlated with 2hPG_corr, i.e., the J-shape correlation was mitigated by the correction. In conclusion, postchallenge hyperglycemia in low BW subjects is in part due to small G_V rather than impaired glucose metabolism.

Predictive models for conversion of prediabetes to diabetes

AIM

To clarify the natural course of prediabetes and develop predictive models for conversion to diabetes.

METHODS

A retrospective longitudinal study of 2105 adults with prediabetes was carried out with a mean observation period of 4.7years. Models were developed using multivariate logistic regression analysis and verified by 10-fold cross-validation. The relationship between [final BMI minus baseline BMI] (δBMI) and incident diabetes was analyzed post hoc by comparing the diabetes conversion rate for low (< -0.31kg/m²) and high δBMI (≥ -0.31kg/m²) subjects after matching the two groups for the covariates.

RESULTS

Diabetes developed in 252 (2.5%/year), and positive family history, male sex, higher systolic blood pressure, plasma glucose (fasting and 1h- and 2h-values during 75g OGTT), hemoglobin A1c (HbA1c) and alanine aminotransferase were significant, independent predictors for the conversion. By using a risk score (RS) that took account of all these variables, incident diabetes was predicted with an area under the ROC curve (95% CI) of 0.80 (0.70-0.87) and a specificity of prediction of 61.8% at 80% sensitivity. On division of the participants into high- (n=248), intermediate- (n=336) and low-risk (n=1521) populations, the conversion rates were 40.1%, 18.5% and 5.9%, respectively. The conversion rate was lower in subjects with low than high δBMI (9.2% vs 14.4%, p=0.003).

CONCLUSIONS

Prediabetes conversion to diabetes could be predicted with accuracy, and weight reduction during the observation was associated with lowered conversion rate.

Risk Score for Detecting Dysglycemia: A Cross-Sectional Study of a Working-Age Population in an Oil Field in China

Background

Dysglycemia (pre-diabetes or diabetes) in young adults has increased rapidly. However, the risk scores for detecting dysglycemia in oil field staff and workers in China are limited. This study developed a risk score for the early identification of dysglycemia based on epidemiological and health examination data in an oil field working-age population with increased risk of diabetes.

Material/Methods

Multivariable logistic regression was used to develop the risk score model in a population-based, cross-sectional study. All subjects completed the questionnaires and underwent physical examination and oral glucose tolerance tests. The performance of the risk score models was evaluated using the area under the receiver operating characteristic curve (AUC).

Results

The study population consisted of 1995 participants, 20–64 years old (49.4% males), with undiagnosed diabetes or pre-diabetes who underwent periodic health examinations from March 2014 to June 2015 in Dagang oil field, Tianjin, China. Age, sex, body mass index, history of high blood glucose, smoking, triglyceride, and fasting plasma glucose (FPG) constituted the Dagang dysglycemia risk score (Dagang DRS) model. The performance of Dagang DRS was superior to m-FINDRISC (AUC: 0.791; 95% confidence interval (CI), 0.773–0.809 vs. 0.633; 95% CI, 0.611–0.654). At the cut-off value of 5.6 mmol/L, the Dagang DRS (AUC: 0.616; 95% CI, 0.592–0.641) was better than the FPG value alone (AUC: 0.571; 95% CI, 0.546–0.596) in participants with FPG <6.1 mmol/L (n=1545, P=0.028).

Conclusions

Dagang DRS is a valuable tool for detecting dysglycemia, especially when FPG <6.1 mmol/L, in oil field workers in China.

Body mass index trajectory patterns and changes in visceral fat and glucose metabolism before the onset of type 2 diabetes

We investigated BMI trajectory patterns before diabetes diagnosis and examined associated changes in visceral adiposity and glucose metabolism. 23,978 non-diabetic Japanese participants (2,789 women) aged 30–64 years were assessed with a mean follow-up of 7.6 years. Diabetes was diagnosed via fasting glucose, HbA_1c, and self-report. Latent-class trajectory analyses were performed to identify BMI trajectories. Longitudinal changes in BMI, visceral adiposity, and glucose metabolism were estimated using mixed models. 1,892 individuals developed diabetes. Three distinct BMI trajectories were identified in adults developing and not developing diabetes, respectively. Among adults developing diabetes, 47.3% were classified as “medium BMI” (n = 895), and had increased mean BMI within the obesity category before diagnosis. The “low BMI” group (38.4%, n = 726) had an initial mean BMI of 21.9 kg/m², and demonstrated small weight gain. The “high BMI” group (n = 271) were severely obese and showed greater increase in BMI until diagnosis. All groups which developed diabetes showed absolute and/or relative increase in visceral fat and impaired β-cell compensation for insulin resistance. All groups not developing diabetes showed measured variables were relatively stable during observation. These data suggest that visceral fat gain may induce β-cell failure in compensation for insulin resistance, resulting in diabetes regardless of obesity level.

Current status of health among workers in Japan: Results from the Japan Epidemiology Collaboration on Occupational Health Study

Data are limited on the sex-specific prevalence of diseases and their risk factors in middle-aged and older workers in Japan. In this cross-sectional study, we investigated the age- and sex-specific prevalence of hypertension, diabetes, dyslipidemia, metabolic syndrome (defined using joint statement criteria), obesity, underweight, abdominal obesity, and smoking among approximately 70,000 to 90,000 Japanese workers (predominantly men) aged 20-69 years in 2014. We also investigated the prevalence of low cardiorespiratory fitness in 2012 and no leisure-time exercise in 2014. In both sexes, the prevalence of lifestyle-related risk factors, including hypertension, diabetes, dyslipidemia, metabolic syndrome, obesity, and abdominal obesity, was increased with aging. In contrast, the prevalence of underweight was decreased with aging. Smoking prevalence exceeded 30% in men regardless of age, whereas the prevalence was around 10% in women of all age groups. Prevalence of no leisure-time exercise exceeded 50% among middle-aged and older workers in both sexes. Among workers aged 50-64 years, less than half of men had low fitness, whereas more than half of women had low fitness. Given the high prevalence of lifestyle-related risk factors among middle-aged and older workers, effective strategies to prevent cardiovascular disease in this age group are needed in Japan.

Diabetes screening intervals based on risk stratification

BACKGROUND

Guidelines for frequency of Type 2 diabetes mellitus (DM) screening remain unclear, with proposed screening intervals typically based on expert opinion. This study aims to demonstrate that HbA1c screening intervals may differ substantially when considering individual risk for diabetes.

METHODS

This was a multi-institutional retrospective open cohort study. Data were collected between April 1999 to March 2014 from one urban and one rural cohort in Japan. After categorization by age, we stratified individuals based on cardiovascular disease risk (Framingham 10-year cardiovascular risk score) and body mass index (BMI). We adapted a signal-to-noise method for distinguishing true HbA1c change from measurement error by constructing a linear random effect model to calculate signal and noise of HbA1c. Screening interval for HbA1c was defined as informative when the signal-to-noise ratio exceeded 1.

RESULTS

Among 96,456 healthy adults, 46,284 (48.0%) were male; age (range) and mean HbA1c (SD) were 48 (30-74) years old and 5.4 (0.4)%, respectively. As risk increased among those 30-44 years old, HbA1c screening intervals for detecting Type 2 DM consistently decreased: from 10.5 (BMI <18.5) to 2.4 (BMI > 30) years, and from 8.0 (Framingham Risk Score <10%) to 2.0 (Framingham Risk Score ≥20%) years. This trend was consistent in other age and risk groups as well; among obese 30-44 year olds, we found substantially shorter intervals compared to other groups.

CONCLUSION

HbA1c screening intervals for identification of DM vary substantially by risk factors. Risk stratification should be applied when deciding an optimal HbA1c screening interval in the general population to minimize overdiagnosis and overtreatment.

Development of new diabetes risk scores on the basis of the current definition of diabetes in Japanese subjects [Rapid Communication]

To develop diabetes risk score (RS) based on the current definition of diabetes, we retrospectively analyzed consecutive 4,159 health examinees who were non-diabetic at baseline. Diabetes, diagnosed by fasting plasma glucose (FPG) ≥7.0 mmol/L, 2hPG ≥11.1 mmol/L and/or HbA1c ≥6.5% (48 mmol/mol), developed in 279 of them during the mean period of 4.9 years. A full RS (RS_Full), a RS without 2hPG (RS_-2hPG) and a non-invasive RS (RS_NI) were created on the basis of multivariate Cox proportional model by weighted grading based on hazard ratio in half the persons assigned. The RSs were verified in the remaining half of the participants. Positive family history (FH), male sex, smoking and higher age, systolic blood pressure (SBP), FPG, 2hPG and HbA1c were independent predictors for RS_Full. For RS_-2hPG, 7 independent predictors, exclusive of 2hPG and smoking but inclusive of elevated triglycerides (TG) comparing to RS_Full, were selected. FH, male sex, and higher age, SBP and HbA1c were independent predictors in RS_NI. In the validation cohort, C-statistic (95%CI) of RS_Full, RS_-2hPG and RS_NI were 0.80 (0.76-0.84), 0.75 (0.70-0.78) and 0.68 (0.63-0.72), respectively, which were significantly different from each other (P <0.01). Absolute percentage difference between predicted probability and observed diabetes were 1.9%, 0.7% and 0.9%, by the three scores, respectively, and not significantly different from each other. In conclusion, diabetes defined by the current criteria was predicted by the new diabetes risk scores with reasonable accuracy. Nonetheless, RS_Full with a postchallenge glucose value performed superior to RS_-2hPG and RS_NI.