Developing risk prediction models for type 2 diabetes: a systematic review of methodology and reporting

Quantifying the risk of type 2 diabetes in East London using the QDScore: a cross-sectional analysis

BACKGROUND

Risk scores calculated from electronic patient records can be used to predict the risk of adults developing diabetes in the future.

AIM

To use a risk-prediction model on GPs' electronic health records in three inner-city boroughs, and to map the risk of diabetes by locality for commissioners, to guide possible interventions for targeting groups at high risk.

DESIGN AND SETTING

Cross-sectional analysis of electronic general practice records from three deprived and ethnically diverse inner-city boroughs in London.

METHOD

A cross-sectional analysis of 519 288 electronic primary care records was performed for all people without diabetes aged 25-79 years. A validated risk score, the QDScore, was used to predict 10-year risk of developing type 2 diabetes. Descriptive statistics were generated, including subanalysis by deprivation and ethnicity. The proportion of people at high risk (≥20% risk) per general practice was geospatially mapped.

RESULTS

Data were obtained from 135 out of 145 general practices (91.3%); 1 in 10 people in this population were at high risk (≥20%) of developing type 2 diabetes within 10 years. Of those with known cardiovascular disease or hypertension, approximately 50% were at high risk. Male sex, increasing age, South Asian ethnicity, deprivation, obesity, and other comorbidities increased the risk. Geospatial mapping revealed hotspots of high risk.

CONCLUSION

Individual risk scores calculated from electronic records can be aggregated to produce population risk profiles to inform commissioning and public health planning. Specific localities were identified (the 'East London diabetes belt'), where preventive efforts should be targeted. This method could be used for other diseases and risk states, to inform targeted commissioning and preventive research.

Risk predictive modelling for diabetes and cardiovascular disease

Absolute risk models or clinical prediction models have been incorporated in guidelines, and are increasingly advocated as tools to assist risk stratification and guide prevention and treatments decisions relating to common health conditions such as cardiovascular disease (CVD) and diabetes mellitus. We have reviewed the historical development and principles of prediction research, including their statistical underpinning, as well as implications for routine practice, with a focus on predictive modelling for CVD and diabetes. Predictive modelling for CVD risk, which has developed over the last five decades, has been largely influenced by the Framingham Heart Study investigators, while it is only ∼20 years ago that similar efforts were started in the field of diabetes. Identification of predictive factors is an important preliminary step which provides the knowledge base on potential predictors to be tested for inclusion during the statistical derivation of the final model. The derived models must then be tested both on the development sample (internal validation) and on other populations in different settings (external validation). Updating procedures (e.g. recalibration) should be used to improve the performance of models that fail the tests of external validation. Ultimately, the effect of introducing validated models in routine practice on the process and outcomes of care as well as its cost-effectiveness should be tested in impact studies before wide dissemination of models beyond the research context. Several predictions models have been developed for CVD or diabetes, but very few have been externally validated or tested in impact studies, and their comparative performance has yet to be fully assessed. A shift of focus from developing new CVD or diabetes prediction models to validating the existing ones will improve their adoption in routine practice.

Survival association rule mining towards type 2 diabetes risk assessment

Type-2 Diabetes Mellitus is a growing epidemic that often leads to severe complications. Effective preventive measures exist and identifying patients at high risk of diabetes is a major health-care need. The use of association rule mining (ARM) is advantageous, as it was specifically developed to identify associations between risk factors in an interpretable form. Unfortunately, traditional ARM is not directly applicable to survival outcomes and it lacks the ability to compensate for confounders and to incorporate dosage effects. In this work, we propose Survival Association Rule (SAR) Mining, which addresses these shortcomings. We demonstrate on a real diabetes data set that SARs are naturally more interpretable than the traditional association rules, and predictive models built on top of these rules are very competitive relative to state of the art survival models and substantially outperform the most widely used diabetes index, the Framingham score.

Predicting risk of type 2 diabetes mellitus with genetic risk models on the basis of established genome-wide association markers: a systematic review

This study aimed to evaluate the predictive performance of genetic risk models based on risk loci identified and/or confirmed in genome-wide association studies for type 2 diabetes mellitus. A systematic literature search was conducted in the PubMed/MEDLINE and EMBASE databases through April 13, 2012, and published data relevant to the prediction of type 2 diabetes based on genome-wide association marker-based risk models (GRMs) were included. Of the 1,234 potentially relevant articles, 21 articles representing 23 studies were eligible for inclusion. The median area under the receiver operating characteristic curve (AUC) among eligible studies was 0.60 (range, 0.55-0.68), which did not differ appreciably by study design, sample size, participants' race/ethnicity, or the number of genetic markers included in the GRMs. In addition, the AUCs for type 2 diabetes did not improve appreciably with the addition of genetic markers into conventional risk factor-based models (median AUC, 0.79 (range, 0.63-0.91) vs. median AUC, 0.78 (range, 0.63-0.90), respectively). A limited number of included studies used reclassification measures and yielded inconsistent results. In conclusion, GRMs showed a low predictive performance for risk of type 2 diabetes, irrespective of study design, participants' race/ethnicity, and the number of genetic markers included. Moreover, the addition of genome-wide association markers into conventional risk models produced little improvement in predictive performance.

The association between the apolipoprotein A1/ high density lipoprotein -cholesterol and diabetes in Taiwan - a cross-sectional study

BACKGROUND

Traditional lipid indices have been associated with type 2 diabetes, but it remains uncertain which lipid index is the best discriminator for diabetes. In this study, we aimed to assess lipoproteins, traditional lipid variables, and other variables to discover their association with diabetes in the Taiwanese population.

METHODS

Data from a nationwide cross-sectional population-based survey of 3087 men and 3373 women in 2002 were analyzed in this study. All participants were assessed for anthropometry, glycosylated hemoglobin, fasting sugar and lipid profiles with triglycerides, high density lipoprotein-cholesterol (HDL-C), low density lipoprotein-cholesterol (LDL-C), and apolipoprotein A1 (ApoA1) and B (ApoB). The ratio of LDL-C/HDL-C, ApoB/ApoA1, ApoB/LDL-C and ApoA1/HDL-C and other variables were analyzed to determine their potential roles in type 2 diabetes in the Taiwanese population. The Odds ratios (ORs) of the risk variables for diabetes were estimated using logistic regression and were adjusted for confounding factors.

RESULTS

The increased ratio of ApoA1/HDL-C was significantly associated with diabetes in men (top tertile vs. lowest: OR 2.98; 95% CI: 1.12 - 7.92; P-trend = 0.030) and women (top tertile vs. lowest: OR 2.15; 95% CI: 1.00 - 4.59; P-trend = 0.047). A modest increased diabetic risk was evident with ApoB/LDL-C in women (top tertile vs. lowest: OR 2.03; 95% CI: 1.07- 3.85; P-trend = 0.028), but not in men (top tertile v. lowest: OR 1.69; 95% CI: 0.79- 3.62; P-trend = 0.198).

CONCLUSIONS

ApoA1/HDL-C had a significant linear association with diabetes in both sexes and was superior to other lipid and lipoprotein variables among the general Taiwanese population.

Risk assessment tools to identify women with increased risk of osteoporotic fracture: complexity or simplicity? A systematic review

A huge number of risk assessment tools have been developed. Far from all have been validated in external studies, more of them have absence of methodological and transparent evidence, and few are integrated in national guidelines. Therefore, we performed a systematic review to provide an overview of existing valid and reliable risk assessment tools for prediction of osteoporotic fractures. Additionally, we aimed to determine if the performance of each tool was sufficient for practical use, and last, to examine whether the complexity of the tools influenced their discriminative power. We searched PubMed, Embase, and Cochrane databases for papers and evaluated these with respect to methodological quality using the Quality Assessment Tool for Diagnostic Accuracy Studies (QUADAS) checklist. A total of 48 tools were identified; 20 had been externally validated, however, only six tools had been tested more than once in a population-based setting with acceptable methodological quality. None of the tools performed consistently better than the others and simple tools (i.e., the Osteoporosis Self-assessment Tool [OST], Osteoporosis Risk Assessment Instrument [ORAI], and Garvan Fracture Risk Calculator [Garvan]) often did as well or better than more complex tools (i.e., Simple Calculated Risk Estimation Score [SCORE], WHO Fracture Risk Assessment Tool [FRAX], and Qfracture). No studies determined the effectiveness of tools in selecting patients for therapy and thus improving fracture outcomes. High-quality studies in randomized design with population-based cohorts with different case mixes are needed.

A clinical risk stratification tool for predicting treatment resistance in major depressive disorder

BACKGROUND

Early identification of depressed individuals at high risk for treatment resistance could be helpful in selecting optimal setting and intensity of care. At present, validated tools to facilitate this risk stratification are rarely used in psychiatric practice.

METHODS

Data were drawn from the first two treatment levels of a multicenter antidepressant effectiveness study in major depressive disorder, the STAR*D (Sequenced Treatment Alternatives to Relieve Depression) cohort. This cohort was divided into training, testing, and validation subsets. Only clinical or sociodemographic variables available by or readily amenable to self-report were considered. Multivariate models were developed to discriminate individuals reaching remission with a first or second pharmacological treatment trial from those not reaching remission despite two trials.

RESULTS

A logistic regression model achieved an area under the receiver operating characteristic curve exceeding .71 in training, testing, and validation cohorts and maintained good calibration across cohorts. Performance of three alternative models with machine learning approaches--a naïve Bayes classifier and a support vector machine, and a random forest model--was less consistent. Similar performance was observed between more and less severe depression, men and women, and primary versus specialty care sites. A web-based calculator was developed that implements this tool and provides graphical estimates of risk.

CONCLUSION

Risk for treatment resistance among outpatients with major depressive disorder can be estimated with a simple model incorporating baseline sociodemographic and clinical features. Future studies should examine the performance of this model in other clinical populations and its utility in treatment selection or clinical trial design.

Diagnostic and prognostic prediction models

Risk prediction models can be used to estimate the probability of either having (diagnostic model) or developing a particular disease or outcome (prognostic model). In clinical practice, these models are used to inform patients and guide therapeutic management. Examples from the field of venous thrombo-embolism (VTE) include the Wells rule for patients suspected of deep venous thrombosis and pulmonary embolism, and more recently prediction rules to estimate the risk of recurrence after a first episode of unprovoked VTE. In this paper, the three phases that are recommended before a prediction model may be used in daily practice are described: development, validation, and impact assessment. In the development phase, the focus is on model development commonly using a multivariable logistic (diagnostic) or survival (prognostic) regression analysis. The performance of the developed model is expressed by discrimination, calibration and (re-) classification. In the validation phase, the developed model is tested in a new set of patients using these same performance measures. This is important, as model performance is commonly poorer in a new set of patients, e.g. due to case-mix or domain differences. Finally, in the impact phase the ability of a prediction model to actually guide patient management is evaluated. Whereas in the development and validation phase single cohort designs are preferred, this last phase asks for comparative designs, ideally randomized designs; therapeutic management and outcomes after using the prediction model is compared to a control group not using the model (e.g. usual care).

The development and validation of the Portuguese risk score for detecting type 2 diabetes and impaired fasting glucose

AIMS

To develop and validate a non-invasive score for detecting undiagnosed impaired fasting glucose (IFG) and type 2 diabetes (T2DM) in a Portuguese population.

METHODS

We used data from 3,374 individuals aged 18-94 years from a Portuguese cross-sectional study. We developed a logistic regression model for predicting IFG/T2DM (diagnosed using fasting glucose). We externally validated the score using data from two cohorts of the EPI-Porto study, cross-sectional (n = 2,131) and data from the 5 year follow-up (n = 1,304).

RESULTS

The final model included age, sex, BMI and hypertension with an area under the ROC curve of 70.1 (95%CI 68.4, 71.7). Using a cut-point which classifies 50% of the EPI-Porto cross-sectional data as high-risk gave sensitivity 73.2% (95%CI 68.5%, 77.6%), specificity 55.5% (53.1%, 57.8%), positive predictive value (PPV) 27.0% (24.3%, 29.8%) and negative predictive value (NPV) 90.2% (88.3%, 92.0%) for IFG/T2DM. Using the same cut-point on the prospective data classified 45% as high-risk; sensitivity 69.1% (63.4%, 74.4%), specificity 63.3% (60.0%, 66.5%), PPV 38.0% (33.9%, 42.4%), and NPV 86.2% (83.3%, 88.8%).

CONCLUSION

The Portuguese risk score can be used to identify those at high risk of both prevalent undiagnosed and incident IFG/T2DM.

Screening for HbA1c-defined prediabetes and diabetes in an at-risk greek population: performance comparison of random capillary glucose, the ADA diabetes risk test and skin fluorescence spectroscopy

BACKGROUND

We examined the accuracy of random capillary glucose (RCG) and two noninvasive screening methods, the ADA diabetes risk test (DRT) and skin fluorescence spectroscopy (SFS) as measured by Scout DS for detecting HbA1c-defined dysglycemia or type 2 diabetes in an at-risk cohort.

METHODS

Subjects were recruited at two clinical sites for a single non-fasting visit. Each subject had measurements of height, weight and waist circumference. A diabetes score was calculated from skin fluorescence measured on the left forearm. A finger prick was done to measure RCG and HbA1c (A1C). Health questionnaires were completed for the DRT. Increasing dysglycemia was defined as A1C ≥ 5.7% (39 mmol/mol) or ≥ 6.0% (42 mmol/mol). Type 2 diabetes was defined as A1C ≥ 6.5% (47.5 mmol/mol).

RESULTS

398 of 409 subjects had complete data for analysis with means for age, body mass index, and waist of 52 years, 27 kg/m(2) and 90 cm. 51% were male. Prevalence of A1C ≥ 5.7%, ≥ 6.0% and ≥ 6.5% were 54%, 34% and 12%, respectively. Areas under the curve (AUC) for detection of increasing levels dysglycemia or diabetes for RCG were 63%, 66% and 72%, for the ADA DRT the AUCs were 75%, 76% and 81% and for SFS the AUCs were 82%, 84% and 90%, respectively. For each level of dysglycemia or diabetes, the SFS AUC was significantly higher than RCG or the ADA DRT.

CONCLUSIONS

The noninvasive skin fluorescence spectroscopy measurement outperformed both RCG and the ADA DRT for detection of A1C-defined dysglycemia or diabetes in an at-risk cohort.

Development of a screening score for undiagnosed diabetes and its application in estimating absolute risk of future type 2 diabetes in Japan: Toranomon Hospital Health Management Center Study 10 (TOPICS 10)

OBJECTIVE

The objective of the study was to develop a screening score for undiagnosed diabetes by eliciting information on noninvasive clinical markers and to assess its effectiveness for identifying the presence of diabetes and predicting future diabetes.

DESIGN, SETTING, AND PARTICIPANTS

A screening score was cross-sectionally developed for 33 335 Japanese individuals aged 18-88 years without known diabetes who underwent a health examination. We validated its utility and compared it with existing screening tools in an independent population (n = 7477). After initial assessment of the instrument, 7332 nondiabetic individuals were followed up for a mean 4.0 years.

RESULTS

Prevalence of undiagnosed diabetes (fasting plasma glucose ≥ 7.0 mmol/L or glycated hemoglobin ≥ 6.5%) was 2.9% (n = 965). Diabetes score included age, sex, family history of diabetes, current smoking habit, body mass index, and hypertension with an area under the receiver-operating characteristics curve of 0.771. Screening with 8 or more points yielded a sensitivity of 72.7% and a specificity of 68.1%. In the validation cohort, the area under the receiver-operating characteristics curve was 0.806. The developed score with 8 or more points had better positive predictive value (9.6%) and positive likelihood ratio (2.52) compared with existing tools (positive predictive value, from 6.9% to 9.4%; positive likelihood ratio, from 1.77 to 2.46) in which each tool's highest combination of sensitivity and specificity was observed. The 4-year cumulative risk of developing diabetes gradually escalated in association with higher screening scores at the initial examination.

CONCLUSIONS

Our algorithm could serve as a self-assessment tool for undiagnosed diabetic patients needing timely medical care and as a prognostic tool for individuals without present diabetes who must be closely followed up to prevent future diabetes.

Prevalence and risk factors of type 2 diabetes in the adults in haikou city, hainan island, china

BACKGROUND

Type 2 diabetes mellitus (T2DM) occurs around the world with high prevalence and causes serious physical harm and economic burden to the afflicted. Haikou City is China's southernmost tropical island city, which has not been previously studied for its T2DM prevalence. The objective of the study in employing a cross-sectional survey is to discuss the epidemiologic status of T2DM in Haikou City and to analyze the possible determinants.

METHODS

A total of 12,000 community residents over 18 years old from four districts in Haikou City were stratified-randomly sampled. A questionnaire survey and physical examination were conducted. Data entry and statistical analysis were performed using SPSS17.0 software.

RESULTS

The prevalence of T2DM in Haikou City was 5.3% (5.15% for males and 5.46% for females). According to the multivariate analysis, the positive factors mainly associated with T2DM in the city included family history, Waist-to-Hip Ratio (WHR), triglycerides, low high-density lipoproteins (HDL), and blood pressure. For both men and women, family history was the highest independent risk factor associated with T2DM (OR= 47.128). The T2DM risk increased with increasing metabolic aggregate.

CONCLUSION

The prevalence of T2DM for the community population of Haikou City was low. The possible risk factors included age, occupation, BMI, waist circumference, WHR, overweight, systemic obesity, central obesity, systolic blood pressure, diastolic blood pressure, total cholesterol, triglycerides, low-density lipoproteins, family history, and HDL.

Shifting from glucose diagnosis to the new HbA1c diagnosis reduces the capability of the Finnish Diabetes Risk Score (FINDRISC) to screen for glucose abnormalities within a real-life primary healthcare preventive strategy

BACKGROUND

To investigate differences in the performance of the Finnish Diabetes Risk Score (FINDRISC) as a screening tool for glucose abnormalities after shifting from glucose-based diagnostic criteria to the proposed new hemoglobin (Hb)A1c-based criteria.

METHODS

A cross-sectional primary-care study was conducted as the first part of an active real-life lifestyle intervention to prevent type 2 diabetes within a high-risk Spanish Mediterranean population. Individuals without diabetes aged 45-75 years (n = 3,120) were screened using the FINDRISC. Where feasible, a subsequent 2-hour oral glucose tolerance test and HbA1c test were also carried out (n = 1,712). The performance of the risk score was calculated by applying the area under the curve (AUC) for the receiver operating characteristic, using three sets of criteria (2-hour glucose, fasting glucose, HbA1c) and three diagnostic categories (normal, pre-diabetes, diabetes).

RESULTS

Defining diabetes by a single HbA1c measurement resulted in a significantly lower diabetes prevalence (3.6%) compared with diabetes defined by 2-hour plasma glucose (9.2%), but was not significantly lower than that obtained using fasting plasma glucose (3.1%). The FINDRISC at a cut-off of 14 had a reasonably high ability to predict diabetes using the diagnostic criteria of 2-hour or fasting glucose (AUC = 0.71) or all glucose abnormalities (AUC = 0.67 and 0.69, respectively). When HbA1c was used as the primary diagnostic criterion, the AUC for diabetes detection dropped to 0.67 (5.6% reduction in comparison with either 2-hour or fasting glucose) and fell to 0.55 for detection of all glucose abnormalities (17.9% and 20.3% reduction, respectively), with a relevant decrease in sensitivity of the risk score.

CONCLUSIONS

A shift from glucose-based diagnosis to HbA1c-based diagnosis substantially reduces the ability of the FINDRISC to screen for glucose abnormalities when applied in this real-life primary-care preventive strategy.

Effectiveness of community-wide and individual high-risk strategies to prevent diabetes: a modelling study

BACKGROUND

Diabetes has been described as one of the most important threats to the health of developed countries. Effective population strategies to prevent diabetes have not been determined but two broad strategies have been proposed: "high-risk" and "community-wide" strategies.

METHODS

We modelled the potential effectiveness of two strategies to prevent 10% of new cases of diabetes in Ontario, Canada over a 5-year period. The 5-year risk of developing physician-diagnosed diabetes was estimated for respondents to the Canadian Community Health Survey 2003 (CCHS 2.1, N = 26 232) using a validated and calibrated diabetes risk tool (Diabetes Population Risk Tool [DPoRT]). We estimated how many cases of diabetes could be prevented using two different strategies: a) a community-wide strategy that would uniformly reduce body mass index (BMI) in the entire population; and b) a high baseline risk strategy using either pharmacotherapy or lifestyle counselling to treat people who have an increased risk of developing diabetes.

RESULTS

In 2003, the 5-year risk of developing diabetes was 4.7% (383 600 new diagnosed cases of diabetes in 8 189 000 Ontarians aged 20+) and risk was moderately diffused (0.5%, 3.1% and 17.9% risk in the 1(st), 5(th) (median) and 10(th) deciles of risk). A 10% reduction in new cases of diabetes would have been achieved under any of the following scenarios: if BMI was 3.5% lower in the entire population; if lifestyle counselling covered 32.2% of high-risk people (371 900 of 1 155 000 people with 5 year diabetes risk greater than 10%); or, if pharmacotherapy covered 65.2% of high-risk people.

CONCLUSIONS

Prevention using pharmacotherapy alone requires unrealistically high coverage levels to achieve modest population reduction in new diabetes cases. On the other hand, in recent years few jurisdictions have been able to achieve a reduction in BMI at the population level, let alone a reduction of BMI of 3.5%.

Prognosis Research Strategy (PROGRESS) 3: prognostic model research

Prognostic models are abundant in the medical literature yet their use in practice seems limited. In this article, the third in the PROGRESS series, the authors review how such models are developed and validated, and then address how prognostic models are assessed for their impact on practice and patient outcomes, illustrating these ideas with examples.

Implementation of the automated Leicester Practice Risk Score in two diabetes prevention trials provides a high yield of people with abnormal glucose tolerance

AIMS/HYPOTHESIS

The Leicester Practice Risk Score (LPRS) is a tool for identifying those at high risk of either impaired glucose regulation (IGR), defined as impaired glucose tolerance and/or impaired fasting glucose, or type 2 diabetes from routine primary care data. The aim of this study was to determine the yield from the LPRS when applied in two diabetes prevention trials.

METHODS

Let's Prevent Diabetes (LPD) and Walking Away from Diabetes (WAD) studies used the LPRS to identify people at risk of IGR or type 2 diabetes from 54 general practices. The top 10% at risk within each practice were invited for screening using a 75 g OGTT. The response rate to the invitation and the prevalence of IGR and/or type 2 diabetes in each study were calculated.

RESULTS

Of those invited 19.2% (n = 3,449) in LPD and 22.1% (n = 833) in WAD attended. Of those screened for LPD 25.5% (95% CI 24.1, 27.0) had IGR and 4.5% (95% CI 3.8, 5.2) had type 2 diabetes, giving a prevalence of any abnormal glucose tolerance of 30.1% (95% CI 28.5, 31.6). Comparable rates were seen for the WAD study: IGR 26.5% (95% CI 23.5, 29.5), type 2 diabetes 3.0% (95% CI 1.8, 4.2) and IGR/type 2 diabetes 29.5% (95% CI 26.4, 32.6).

CONCLUSIONS/INTERPRETATION

Using the LPRS identifies a high yield of people with abnormal glucose tolerance, significantly higher than those seen in a population screening programme in the same locality. The LPRS is an inexpensive and simple way of targeting screening programmes at those with the highest risk.

Development of a new scoring system for predicting the 5 year incidence of type 2 diabetes in Japan: the Toranomon Hospital Health Management Center Study 6 (TOPICS 6)

AIMS/HYPOTHESIS

The aims of this study were to assess the clinical significance of introducing HbA(1c) into a risk score for diabetes and to develop a scoring system to predict the 5 year incidence of diabetes in Japanese individuals.

METHODS

The study included 7,654 non-diabetic individuals aged 40-75 years. Incident diabetes was defined as fasting plasma glucose (FPG) ≥7.0 mmol/l, HbA(1c) ≥6.5% (48 mmol/mol) or self-reported clinician-diagnosed diabetes. We constructed a risk score using non-laboratory assessments (NLA) and evaluated improvements in risk prediction by adding elevated FPG, elevated HbA(1c) or both to NLA.

RESULTS

The discriminative ability of the NLA score (age, sex, family history of diabetes, current smoking and BMI) was 0.708. The difference in discrimination between the NLA + FPG and NLA + HbA(1c) scores was non-significant (0.836 vs 0.837; p = 0.898). A risk score including family history of diabetes, smoking, obesity and both FPG and HbA(1c) had the highest discrimination (0.887, 95% CI 0.871, 0.903). At an optimal cut-off point, sensitivity and specificity were high at 83.7% and 79.0%, respectively. After initial screening using NLA scores, subsequent information on either FPG or HbA(1c) resulted in a net reclassification improvement of 42.7% or 52.3%, respectively (p < 0.0001). When both were available, net reclassification improvement and integrated discrimination improvement were further improved at 56.7% (95% CI 47.3%, 66.1%) and 10.9% (9.7%, 12.1%), respectively.

CONCLUSIONS/INTERPRETATION

Information on HbA(1c) or FPG levels after initial screening by NLA can precisely refine diabetes risk reclassification.

A systematic review finds prediction models for chronic kidney disease were poorly reported and often developed using inappropriate methods

BACKGROUND

Chronic kidney disease (CKD) is a global health concern that is increasing mainly as the result of increasing incidences of diabetes and hypertension. Furthermore, if left untreated, individuals with CKD may progress to end-stage kidney failure. Identifying individuals with undiagnosed CKD or those who are at an increased risk of developing CKD or progressing to end-stage kidney disease (ESKD) is therefore an important challenge. We sought to systematically review and critically assess the conduct and reporting of methods used to develop risk prediction models for predicting the risk of having undiagnosed (prevalent) or future risk of developing (incident) CKD or end-stage kidney failure in adults.

METHODS

We conducted a systematic search of PubMed database to identify studies published up until September 2011 that describe the development of models combining two or more variables to predict the risk of prevalent or incident CKD or ESKD. We extracted key information that describes aspects of developing a prediction model, including the study design, data quality, sample size and number of events, outcome definition, risk predictor selection and coding, missing data, model-building strategies, and aspects of performance.

RESULTS

Eleven studies describing the development of 14 prediction models were included. Eight studies reported the development of 11 models to predict incident CKD or ESKD, whereas 3 studies developed models for prevalent CKD. A total of 97 candidate risk predictors were considered, and 43 different risk predictors featured in the 14 prediction models. A method, not recommended to select risk predictors for inclusion in the multivariate model, using statistical significance from univariate screening was carried out in six studies. Missing data were frequently poorly handled and reported with no mention of missing data in four studies; 4 studies explicitly excluded individuals with missing data, and only 2 studies used multiple imputation to replace missing values.

CONCLUSION

We found that prediction models for chronic kidney were often developed using inappropriate methods and were generally poorly reported. Using poor methods can affect the predictive ability of the models, whereas inadequate reporting hinders an objective evaluation of the potential usefulness of the model.

Prediction models for risk of developing type 2 diabetes: systematic literature search and independent external validation study

OBJECTIVE

To identify existing prediction models for the risk of development of type 2 diabetes and to externally validate them in a large independent cohort.

DATA SOURCES

Systematic search of English, German, and Dutch literature in PubMed until February 2011 to identify prediction models for diabetes.

DESIGN

Performance of the models was assessed in terms of discrimination (C statistic) and calibration (calibration plots and Hosmer-Lemeshow test).The validation study was a prospective cohort study, with a case cohort study in a random subcohort.

SETTING

Models were applied to the Dutch cohort of the European Prospective Investigation into Cancer and Nutrition cohort study (EPIC-NL).

PARTICIPANTS

38,379 people aged 20-70 with no diabetes at baseline, 2506 of whom made up the random subcohort.

OUTCOME MEASURE

Incident type 2 diabetes.

RESULTS

The review identified 16 studies containing 25 prediction models. We considered 12 models as basic because they were based on variables that can be assessed non-invasively and 13 models as extended because they additionally included conventional biomarkers such as glucose concentration. During a median follow-up of 10.2 years there were 924 cases in the full EPIC-NL cohort and 79 in the random subcohort. The C statistic for the basic models ranged from 0.74 (95% confidence interval 0.73 to 0.75) to 0.84 (0.82 to 0.85) for risk at 7.5 years. For prediction models including biomarkers the C statistic ranged from 0.81 (0.80 to 0.83) to 0.93 (0.92 to 0.94). Most prediction models overestimated the observed risk of diabetes, particularly at higher observed risks. After adjustment for differences in incidence of diabetes, calibration improved considerably.

CONCLUSIONS

Most basic prediction models can identify people at high risk of developing diabetes in a time frame of five to 10 years. Models including biomarkers classified cases slightly better than basic ones. Most models overestimated the actual risk of diabetes. Existing prediction models therefore perform well to identify those at high risk, but cannot sufficiently quantify actual risk of future diabetes.

TNF-related apoptosis-inducing ligand significantly attenuates metabolic abnormalities in high-fat-fed mice reducing adiposity and systemic inflammation

TRAIL [TNF (tumour necrosis factor)-related apoptosis-inducing ligand] has recently been shown to ameliorate the natural history of DM (diabetes mellitus). It has not been determined yet whether systemic TRAIL delivery would prevent the metabolic abnormalities due to an HFD [HF (high-fat) diet]. For this purpose, 27 male C57bl6 mice aged 8 weeks were randomly fed on a standard diet, HFD or HFD+TRAIL for 12 weeks. TRAIL was delivered weekly by intraperitoneal injection. Body composition was evaluated; indirect calorimetry studies, GTT (glucose tolerance test) and ITT (insulin tolerance test) were performed. Pro-inflammatory cytokines, together with adipose tissue gene expression and apoptosis, were measured. TRAIL treatment reduced significantly the increased adiposity associated with an HFD. Moreover, it reduced significantly hyperglycaemia and hyperinsulinaemia during a GTT and it improved significantly the peripheral response to insulin. TRAIL reversed the changes in substrate utilization induced by the HFD and ameliorated skeletal muscle non-esterified fatty acids oxidation rate. This was associated with a significant reduction of pro-inflammatory cytokines together with a modulation of adipose tissue gene expression and apoptosis. These findings shed light on the possible anti-adipogenic and anti-inflammatory effects of TRAIL and open new therapeutic possibilities against obesity, systemic inflammation and T2DM (Type 2 DM).

Fracture risk assessment: state of the art, methodologically unsound, or poorly reported?

Osteoporotic fractures, including hip fractures, are a global health concern associated with significant morbidity and mortality as well as a major economic burden. Identifying individuals who are at an increased risk of osteoporotic fracture is an important challenge to be resolved. Recently, multivariable prediction tools have been developed to assist clinicians in the management of their patients by calculating their 10-year risk of fracture (FRAX, QFracture, Garvan) using a combination of known risk factors. These prediction models have revolutionized the way clinicians assess the risk of fracture. Studies evaluating the performance of prediction models in this and other areas of medicine have, however, been characterized by poor design, methodological conduct, and reporting. We examine recently developed fracture prediction models and critically discuss issues in their design, validation, and transparency.

Body mass index, triglycerides, glucose, and blood pressure as predictors of type 2 diabetes in a middle-aged Norwegian cohort of men and women

BACKGROUND

Obesity, hypertension, and hypertriglyceridemia are important risk factors for type 2 diabetes (T2D). We wanted to assess the risk associated with these three factors alone and in combination, and the relative importance of these and several other risk factors (eg, nonfasting glucose) as predictors of T2D.

METHODS

Risk factors in a Norwegian population (n = 109,796) aged 40-45 years were measured in health studies in 1995-1999. Blood glucose-lowering drugs dispensed in 2004-2009 were used to estimate the incidence of T2D. Groups based on combinations of body mass index (BMI), diastolic blood pressure, and triglycerides were defined by using the 50% and 90% quantiles for each variable for men and women. The relative importance of BMI, triglycerides, total cholesterol, high-density lipoprotein cholesterol, glucose, blood pressure, and year of birth for predicting T2D was assessed using deviance from univariate and multivariate logistic regression models. Height, weight, and blood pressure were measured. All biomarkers were measured in nonfasting blood samples.

RESULTS

In the various groups of BMI, triglycerides, and diastolic blood pressure, the incidence of T2D ranged from 0.5% to 19.7% in men and from 0.15% to 21.8% in women. BMI was the strongest predictor of incident T2D, followed by triglyceride levels in women and glucose levels in men. The inclusion of risk factors other than BMI, glucose, triglycerides, and blood pressure in multivariate models only marginally improved the prediction.

CONCLUSION

BMI was the strongest predictor of type 2 diabetes. At defined levels of BMI, the incidence of T2D varied substantially with triglyceride levels and blood pressure. Thus, controlling triglycerides and blood pressure in middle-aged individuals should be targeted to prevent later onset of T2D.

Improving the reporting of randomised trials: the CONSORT Statement and beyond

An extensive and growing number of reviews of the published literature demonstrate that health research publications have frequent deficiencies. Of particular concern are poor reports of randomised trials, which make it difficult or impossible for readers to assess how the research was conducted, to evaluate the reliability of the findings, or to place them in the context of existing research evidence. As a result, published reports of trials often cannot be used by clinicians to inform patient care or to inform public health policy, and the data cannot be included in systematic reviews. Reporting guidelines are designed to identify the key information that researchers should include in a report of their research. We describe the history of reporting guidelines for randomised trials culminating in the CONSORT Statement in 1996. We detail the subsequent development and extension of CONSORT and consider related initiatives aimed at improving the reliability of the medical research literature.

Achutha Menon Centre Diabetes Risk Score: a type 2 diabetes screening tool for primary health care providers in rural India

The authors aimed to develop a diabetes risk score for primary care providers in rural India. They used the baseline data of 451 participants (15-64 years) of a cohort study in a rural area of Kerala, India. The new risk score with age, family history of diabetes, and waist circumference identified 40.8% for confirmatory testing, had a sensitivity of 81.0%, specificity of 68.4%, positive predictive value of 37.0%, and negative predictive value of 94.0% for an optimal cutoff ≥4 with an area under the receiver operating characteristic curve of 0.812 (95% confidence interval = 0.765-0.860). The new risk score with 3 simple, easy-to-measure, less time-consuming, and less expensive variables could be suitable for use in primary care settings of rural India.

Sex differences in the association between plasma copeptin and incident type 2 diabetes: the Prevention of Renal and Vascular Endstage Disease (PREVEND) study

AIMS/HYPOTHESIS

Vasopressin plays a role in osmoregulation, glucose homeostasis and inflammation. Therefore, plasma copeptin, the stable C-terminal portion of the precursor of vasopressin, has strong potential as a biomarker for the cardiometabolic syndrome and diabetes. Previous results were contradictory, which may be explained by differences between men and women in responsiveness of the vasopressin system. The aim of this study was to evaluate the usefulness of copeptin for prediction of future type 2 diabetes in men and women separately.

METHODS

From the Prevention of Renal and Vascular Endstage Disease (PREVEND) study, 4,063 women and 3,909 men without diabetes at baseline were included. A total of 208 women and 288 men developed diabetes during a median follow-up of 7.7 years.

RESULTS

In multivariable-adjusted models, we observed a stronger association of copeptin with risk of future diabetes in women (OR 1.49 [95% CI 1.24, 1.79]) than in men (OR 1.01 [95% CI 0.85, 1.19]) (p (interaction) < 0.01). The addition of copeptin to the Data from the Epidemiological Study on the Insulin Resistance Syndrome (DESIR) clinical model improved the discriminative value (C-statistic,+0.007, p = 0.02) and reclassification (integrated discrimination improvement [IDI] = 0.004, p < 0.01) in women. However, we observed no improvement in men. The additive value of copeptin in women was maintained when other independent predictors, such as glucose, high sensitivity C-reactive protein (hs-CRP) and 24 h urinary albumin excretion (UAE), were included in the model.

CONCLUSIONS/INTERPRETATION

The association of plasma copeptin with the risk of developing diabetes was stronger in women than in men. Plasma copeptin alone, and along with existing biomarkers (glucose, hs-CRP and UAE), significantly improved the risk prediction for diabetes in women.

Let's prevent diabetes: study protocol for a cluster randomised controlled trial of an educational intervention in a multi-ethnic UK population with screen detected impaired glucose regulation

BACKGROUND

The prevention of type 2 diabetes is a globally recognised health care priority, but there is a lack of rigorous research investigating optimal methods of translating diabetes prevention programmes, based on the promotion of a healthy lifestyle, into routine primary care. The aim of the study is to establish whether a pragmatic structured education programme targeting lifestyle and behaviour change in conjunction with motivational maintenance via the telephone can reduce the incidence of type 2 diabetes in people with impaired glucose regulation (a composite of impaired glucose tolerance and/or impaired fasting glucose) identified through a validated risk score screening programme in primary care.

DESIGN

Cluster randomised controlled trial undertaken at the level of primary care practices. Follow-up will be conducted at 12, 24 and 36 months. The primary outcome is the incidence of type 2 diabetes. Secondary outcomes include changes in HbA1c, blood glucose levels, cardiovascular risk, the presence of the Metabolic Syndrome and the cost-effectiveness of the intervention.

METHODS

The study consists of screening and intervention phases within 44 general practices coordinated from a single academic research centre. Those at high risk of impaired glucose regulation or type 2 diabetes are identified using a risk score and invited for screening using a 75 g-oral glucose tolerance test. Those with screen detected impaired glucose regulation will be invited to take part in the trial. Practices will be randomised to standard care or the intensive arm. Participants from intensive arm practices will receive a structured education programme with motivational maintenance via the telephone and annual refresher sessions. The study will run from 2009-2014.

DISCUSSION

This study will provide new evidence surrounding the long-term effectiveness of a diabetes prevention programme conducted within routine primary care in the United Kingdom.

TRIAL REGISTRATION

Clinicaltrials.gov NCT00677937.

Prediction at first year of incident new-onset diabetes after kidney transplantation by risk prediction models

OBJECTIVE

Our aim was to analyze the performance of two scores developed for predicting diabetes in nontransplant populations for identifying kidney transplant recipients with a higher new-onset diabetes mellitus after transplantation (NODAT) risk beyond the first year after transplantation.

RESEARCH DESIGN AND METHODS

We analyzed 191 kidney transplants, which had at least 1-year follow-up posttransplant. First-year posttransplant variables were collected to estimate the San Antonio Diabetes Prediction Model (SADPM) and Framingham Offspring Study-Diabetes Mellitus (FOS-DM) algorithm.

RESULTS

Areas under the receiver operating characteristic curve of FOS-DM and SADPM scores to predict NODAT were 0.756 and 0.807 (P < 0.001), respectively. FOS-DM and SADPM scores over 75 percentile (hazard ratio 5.074 and 8.179, respectively, P < 0.001) were associated with NODAT.

CONCLUSIONS

Both scores can be used to identify kidney recipients at higher risk for NODAT beyond the first year. SADPM score detects some 25% of kidney transplant patients with an eightfold risk for NODAT.