ESM1 for Glucose, blood pressure and cholesterol levels and their relationships to clinical outcomes in type 2 diabetes: a retrospective cohort study

Similar documents
Glucose, blood pressure and cholesterol levels and their relationships to clinical outcomes in type 2 diabetes: a retrospective cohort study

National Diabetes Audit

Discontinuation and restarting in patients on statin treatment: prospective open cohort study using a primary care database

Table S1. Read and ICD 10 diagnosis codes for polymyalgia rheumatica and giant cell arteritis

A factorial randomized trial of blood pressure lowering and intensive glucose control in 11,140 patients with type 2 diabetes

Supplement materials:

Since 1980, obesity has more than doubled worldwide, and in 2008 over 1.5 billion adults aged 20 years were overweight.

ADVANCE post trial ObservatioNal Study

14/15 Threshold 15/16 Points 15/16. Points. Retired Replaced by NM82/AF007. Replacement NO CHANGE

Scottish Diabetes Survey

New indicators to be added to the NICE menu for the QOF and amendments to existing indicators

Lucia Cea Soriano 1, Saga Johansson 2, Bergur Stefansson 2 and Luis A García Rodríguez 1*

SUMMARY OF CHANGES TO QOF 2017/18 - ENGLAND CLINICAL

17/18 Threshold 18/19 Points 18/19. Points NO CHANGE NO CHANGE NO CHANGE

SCIENTIFIC STUDY REPORT

The contractor establishes and maintains a register of patients with AF

Supplementary Appendix

Supplementary Appendix

Patient is healthy with no chronic disease or significant risk factors [16%].

Scottish Diabetes Survey 2012

Inequalities in health due to ethnicity and social deprivation an analysis of primary care data from one inner city area over a three year period

MEDCODE READCODE READTERM

Diabetes. Ref HSCW 024

Appendix This appendix was part of the submitted manuscript and has been peer reviewed. It is posted as supplied by the authors.

Comparability of patient-reported health status: multi-country analysis of EQ-5D responses in patients with type 2 diabetes

Development and validation of QDiabetes-2018 risk prediction algorithm to estimate future risk of type 2 diabetes: cohort study

Survey Scottish Diabetes. Survey Monitoring Group

ANNEX FORM TO EXAMINE THE CAUSES OF ESRD I IDENTIFICATION. 1 Record number (patient chart) Name. 1.1 Date of birth / / 1.2 Sex:

Modelling Reduction of Coronary Heart Disease Risk among people with Diabetes

The State of Play of Diabetes Indicators

Time Series Analysis for selected clinical indicators from the Quality and Outcomes Framework

Blood pressure and total cholesterol level are critical risks especially for hemorrhagic stroke in Akita, Japan.

The Whitehall II study originally comprised 10,308 (3413 women) individuals who, at

Why is Earlier and More Aggressive Treatment of T2 Diabetes Better?

National Diabetes Insulin Pump Audit, England and Wales

Finland and Sweden and UK GP-HOSP datasets

SUMMARY OF CHANGES TO QOF 2014/15 - ENGLAND CLINICAL

etable 3.1: DIABETES Name Objective/Purpose

Schöttker et al. BMC Medicine (2016) 14:26 DOI /s

Depok-Indonesia STEPS Survey 2003

Antidepressant use and risk of cardiovascular outcomes in people aged 20 to 64: cohort study using primary care database

Heart Failure and COPD: Common Partners, Common Problems. Nat Hawkins Liverpool Heart and Chest Hospital

Coronary Heart Disease

Macrovascular Residual Risk. What risk remains after LDL-C management and intensive therapy?

Diabetes treatments and risk of heart failure, cardiovascular disease, and all cause mortality: cohort study in primary care

Using the EQ-5D index score as a predictor of outcomes in patients with type 2 diabetes.

Complications of Diabetes: Screening and Prevention

Repeat ischaemic heart disease audit of primary care patients ( ): Comparisons by age, sex and ethnic group

University of Padova, Padua, Italy, and HARVEST Study Group, Italy

SUPPLEMENTARY DATA. Supplementary Figure S1. Cohort definition flow chart.

Cardiovascular Diabetology. Open Access ORIGINAL INVESTIGATION. C. R. L. Cardoso 1, N. C. Leite 1, C. B. M. Moram 2 and G. F.

Supplementary Online Content

(n=6279). Continuous variables are reported as mean with 95% confidence interval and T1 T2 T3. Number of subjects

Optimizing risk assessment of total cardiovascular risk What are the tools? Lars Rydén Professor Karolinska Institutet Stockholm, Sweden

Blood Pressure Treatment Goals

RESEARCH. Predicting risk of type 2 diabetes in England and Wales: prospective derivation and validation of QDScore

Supplementary appendix

Quality of prescribing in chronic kidney disease and type 2 diabetes Smits, Kirsten Petronella Juliana

Summary of 2011/12 QOF indicator changes, points and thresholds

RESEARCH. Predicting risk of osteoporotic fracture in men and women in England and Wales: prospective derivation and validation of QFractureScores

A mixed methods evaluation of a diabetes peer support scheme in an inner city London area

NATIONAL INSTITUTE FOR HEALTH AND CARE EXCELLENCE QUALITY AND OUTCOMES FRAMEWORK (QOF) INDICATOR DEVELOPMENT PROGRAMME. Indicator Assessment Report

POTENTIAL LINKAGES BETWEEN THE QUALITY AND OUTCOMES FRAMEWORK (QOF) AND THE NHS HEALTH CHECK

NICE Indicator Programme. Consultation on proposed amendments to current QOF indicators

Diabetes Mellitus: A Cardiovascular Disease

Ulster, Ireland. Submitted: 21 June 2009; Revised: 24 January 2010; Published: 13 September 2010 Petrazzuoli F, Soler JK, Buono N, Dobbs F

Supplementary Appendix

Map 6: Percentage of people in the National Diabetes Audit (NDA) with Type 1 diabetes receiving all nine key care processes by PCT

Serum levels of galectin-1, galectin-3, and galectin-9 are associated with large artery atherosclerotic

Statin therapy in patients with Mild to Moderate Coronary Stenosis by 64-slice Multidetector Coronary Computed Tomography

LEADER Liraglutide and cardiovascular outcomes in type 2 diabetes

Risk of stroke and transient ischaemic attack in patients with a diagnosis of resolved atrial fibrillation: retrospective cohort studies

SUPPLEMENTARY DATA. Table of Contents

Development and validation of QRISK3 risk prediction algorithms to estimate future risk of cardiovascular disease: prospective cohort study

Epidemiologic Measure of Association

Adherence to a healthy diet in relation to cardiovascular incidence and risk markers: evidence from the Caerphilly Prospective Study

Johanna R Feary, 1,2 Laura C Rodrigues, 2 Christopher J Smith, 1 Richard B Hubbard, 1 Jack E Gibson 1. Chronic obstructive pulmonary disease

Coronary heart disease prevention and age inequalities:

Gastro-oesophageal junction cancers (ICD10 C15.2, C15.5, C16.0) 1

Analyzing diastolic and systolic blood pressure individually or jointly?

Definitions of chronic conditions used to define the number of serious comorbidities in the study.

Prognostic significance of blood pressure measured in the office, at home and during ambulatory monitoring in older patients in general practice

Sponsor: Sanofi Drug substance(s): Lantus /insulin glargine. Study Identifiers: U , NCT Study code: LANTUL07225

Risk factors for microvascular and macrovascular complications in men and women with type 2 diabetes

Supplementary Appendix

Cause of Death Mortality: International Trends by Socio-Economic Group

RESEARCH. Predicting cardiovascular risk in England and Wales: prospective derivation and validation of QRISK2

National Diabetes Audit, Report 1: Care Processes and Treatment Targets

SHORT COMMUNICATION. Keywords Anxiety. Depression. Diabetes. Obesity. Diabetologia (2010) 53: DOI /s

Risk of mortality and adverse cardiovascular outcomes in type 2 diabetes: a comparison of patients treated with sulfonylureas and metformin

Supplementary Table S1: Proportion of missing values presents in the original dataset

The Burden of the Diabetic Heart

Preventive Cardiology Scientific evidence

Lifetime vs. 10-year risk to allocate treatments for the primary prevention of cardiovascular disease

RESEARCH. Derivation and validation of QRISK, a new cardiovascular disease risk score for the United Kingdom: prospective open cohort study

FEV1 predicts length of stay and in-hospital mortality in patients undergoing cardiac surgery

Metformin should be considered in all patients with type 2 diabetes unless contra-indicated

Supplemental Table 1. Standardized Serum Creatinine Measurements. Supplemental Table 3. Sensitivity Analyses with Additional Mortality Outcomes.

A lthough the hazards of smoking are well described,

Transcription:

ESM1 for Glucose, blood pressure and cholesterol levels and their relationships to clinical outcomes in type 2 diabetes: a retrospective cohort study Statistical modelling details We used Cox proportional-hazards regressions to investigate relationships between biometric factors (HbA 1c, cholesterol, systolic and diastolic blood pressure) and patient mortality and first new diabetesrelated complications, in yearly time windows. For many patients, data for time-varying variables - most particularly biometric measurements - were missing for the year of their death. We therefore only included data for these variables up to the year prior to death, as predictors of death in the subsequent year. For new macrovascular and microvascular diabetes complications, in patients with no history of macrovascular or microvascular complications respectively, missing data in the year of diagnosis of a complication was not an issue and they were modelled without a time lag. Separate analyses were executed for each of the two complication types and each of the mortality outcomes: all-cause CPRD (Clinical Practice Research Datalink), all-cause ONS (Office of National Statistics), diabetes ONS, coronary (ischaemic) heart disease ONS, cerebrovascular (stroke) ONS and cerebrovascular excluding bleeds ONS. Each model included HbA 1c, cholesterol, systolic and diastolic blood pressure as potentially explanatory variables. Additional covariates included to control for potential confounding were: age, sex, each co-morbidity (asthma, coronary heart disease, chronic kidney disease, chronic obstructive pulmonary disease, depression, dementia, severe mental illness, heart failure, hypertension, stroke, cancer, epilepsy, osteoarthritis, osteoporosis, and hypothyroidism), BMI, microvascular and macrovascular complications (except where the outcome was complications; but microvascular included as predictor when the outcome was macrovascular), smoking, and practice characteristics (diabetes prevalence, list size, region and area deprivation as measured by the 2010 Index of Multiple Deprivation). We did not control the analyses for prescribed medication to avoid over-complicating the research question, since our focus was the relationship between biometrics and outcomes irrespective of medication. In the macrovascular (myocardial infarction, stroke, peripheral vascular disease or amputation) complications analysis, we did not include stroke as a covariate since there was complete overlap with the outcome, but we included coronary heart disease and heart failure since they do not overlap fully with myocardial infarction. Some overlap is not a problem since the survival cohort is necessarily constrained to patients without a macrovascular complication at entry; hence we were able to control the analysis for cases of coronary heart disease or heart failure without a myocardial infarction. 1

Similarly, in the microvascular (retinopathy, neuropathy, nephropathy (chronic kidney disease stages 4-5) or foot ulcer) complications analysis we did not include chronic kidney disease as a covariate. Due to the nature of the modelled variables (e.g. a comorbidity was either present for a patient because of a relevant code or not) data was complete except for the biometric measurements (BMI, HbA 1c, blood pressure and cholesterol levels) and smoking status. For BMI, we used an interpolation algorithm to clean and impute values between observations over time, which will be made available in Stata in the near future. The algorithm used all available BMI, height and weight information from 2000/1 to 2011/12, in linear regression modelling, to impute BMI values for patients in each yearly time point. Extrapolations are not allowed and, for each patient, at least two measurements at different time points are required for the algorithm to generate information. For all analyses, biometric measurements were categorised and for the main analyses missing data were coded as an additional category (see Figures 1-5 and Table 2). HbA 1c was categorised as: <6.25% (<45mmol/mol); from 6.25% to 6.75% (45-50mmol/mol); above 6.75% and up to 7.25% (50-56mmol/mol); above 7.25% and up to 7.75% (56-61mmol/mol); above 7.75% and up to 8.25% (61-67mmol/mol); and >8.25% (>67mmol/mol). Cholesterol categories were: <2.5mmol/l; from 2.5mmol/l to 3.5mmol/l; above 3.5mmol/l and up to 4.5mmol/l; above 4.5mmol/l and up to 5.5mmol/l; above 5.5mmol/l and up to 6.5mmol/l; and >6.5mmol/l. For systolic blood pressure we categorised to: <115mmHg; from 115mmHg to 125mmHg; above 125mmHg and up to 135mmHg; above 135mmHg and up to 145mmHg; above 145mmHg and up to 155mmHg; above 155mmHg and up to 165mmHg; above 165mmHg and up to 175mmHg; and >175mmHg. For diastolic blood pressure the categories were: <72.5mmHg; from 72.5mmHg to 77.5mmHg; above 77.5mmHg and up to 82.5mmHg; above 82.5mmHg and up to 87.5mmHg; above 87.5mmHg and up to 92.5mmHg; and >92.5mmHg. Finally BMI was categorised to: <18.5; from 18.5 to 25; above 25 and up to 30; above 30 and up to 40; and >40. However, we also employed twofold, a multiple imputation algorithm for longitudinal data which uses data within (e.g. a diagnosis of hypertension which is used to impute blood pressure values) and across time (e.g. measurements of blood pressure in other years) to impute missing biometric measurements [1]. HbA 1c, cholesterol, systolic and diastolic blood pressure and BMI (interpolated) were specified as time-dependent variables with missing values, age at entry and sex as fully observed timeindependent variables, and all co-morbidities were used as fully observed time-dependent variables. Through this approach we generated five complete datasets with which we conducted sensitivity analyses using multiple imputation techniques. Note that we used an initial interpolation stage only for BMI since it is different from the other metrics, in the sense that it is much more stable over time, 2

and we felt that interpolated values were potentially more reliable than imputed values based on other parameters. Additional sensitivity analyses for all-cause mortality were run in a subsample of patients aged up to 65 (to verify patterns in a younger population), with a 2- and 3-year time lag on mortality (to verify results were not affected by changes in the biometrics immediately prior to death, possibly due to frailty), across different polypharmacy-levels subgroups (to assess severity confounding) and by gender. The results for these are provided in a supplementary excel file. In all survival analysis models we used a yearly time window, allowed multiple records for patients, used the diagnosis year as the time a patient becomes at risk, specified patient entries and exits from the database and used patient as a clustering variable to obtain standard errors that account for intragroup correlation. When more than one biometric value was available within a year for a patient, we used the mean value for the analyses. We used Schoenfeld residuals to test for the proportionalhazards assumption and included time-varying covariates when needed to stabilise the models. To allow an easy interpretation of the coefficients for the biometric measurement, we did not include any time-varying covariates for them although it was indicated in some of the models (see main paper Table 2 and ESM-1 Tables 2-3). All analyses were performed using Stata v13 and commands stcox and mi estimate [2]. Results: BMI Interestingly, high BMI appears to have a prophylactic effect in all analyses. However, in univariate analyses where we focused on the relationship between BMI and death or complications the direction of the effect is reversed (available from the authors). High BMI is a risk factor because it leads to comorbidities and affects the other biometric measurements (HbA 1c, total cholesterol and blood pressure). In the full analyses where we control for all these factors this relationship between BMI and the other biometrics is masked and we are likely observing the prophylactic effect of a higher BMI in the old and very old. The causal relationships between all the biometrics could be mapped and tested through a Structural Equation Modelling (SEM) approach, but SEM survival analysis models are underdeveloped and analysing with this framework would be a serious challenge. References [1] Welch C, Bartlett J, Petersen I (2014) Application of multiple imputation using the two-fold fully conditional specification algorithm in longitudinal clinical data. Stata J 14: 418-431 [2] StataCorp L. P. (2013) Stata Statistical software for Windows. In: 3

Table A1: Comorbidities for type II diabetes patients, over 6 years (2006/7-2011/12), including mean age (SD) Female Male All Freq Cuml Age Freq Cuml Age Freq Cuml Age Conditions* None 8.6 8.6 56.8(13.8) 16.7 16.7 56.7(12.1) 13.1 13.1 56.8(12.7) One 22.7 31.3 62.9(13.1) 30.8 47.4 62.2(11.5) 27.2 40.3 62.4(12.1) Two 26.2 57.5 67.3(12.4) 25.5 73.0 66.7(11.1) 25.8 66.1 67.0(11.7) Three 20.5 78.0 70.9(11.5) 15.0 88.0 70.2(10.3) 17.4 83.5 70.6(11.0) Four 12.2 90.2 73.4(10.5) 7.5 95.5 72.9(9.4) 9.6 93.1 73.2(10.0) Five 6.1 96.3 75.1(9.7) 3.0 98.5 74.6(8.9) 4.4 97.5 74.9(9.4) Six+ 3.7 100.0 76.8(9.0) 1.5 100.0 75.4(8.4) 2.5 100.0 76.3(8.8) * Including: asthma, coronary heart disease, chronic kidney disease, chronic obstructive pulmonary disease, depression, dementia, severe mental illness, heart failure, hypertension, stroke, cancer, epilepsy, osteoarthritis, osteoporosis, hypothyroidism 4

Main analyses coding missing data as a category ESM-1 Figure 1 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on coronary heart disease related mortality (verified through the Office of National Statistics) in the following year (main analysis)* * The second category in all graphs in the reference category. Ischaemic heart disease (ICD-10 codes I21-I22) underlying cause or in the top three causes of death. Missing category for diastolic blood pressure omitted because perfectly collinear with missing for systolic. 5

ESM-1 Figure 2 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on cerebrovascular disease related mortality (verified through the Office of National Statistics) in the following year (main analysis)* * The second category in all graphs in the reference category. Stroke (ICD-10 codes I60-I64) underlying cause or in the top three causes of death. Missing category for diastolic blood pressure omitted because perfectly collinear with missing for systolic. 6

ESM-1 Figure 3 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on developing a new macrovascular diabetes complication in the current year, when none before (main analysis)* * The second category in all graphs in the reference category. Missing category for diastolic blood pressure omitted because perfectly collinear with missing for systolic. 7

ESM-1 Figure 4 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on developing a new microvascular diabetes complication in the current year, when none before (main analysis)* * The second category in all graphs in the reference category. Missing category for diastolic blood pressure omitted because perfectly collinear with missing for systolic. 8

ESM-1 Figure 5 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on all-cause mortality (verified through the Office of National Statistics) in the following year (main analysis)* * The second category in all graphs in the reference category Missing category for diastolic blood pressure omitted because perfectly collinear with missing for systolic. 9

ESM-1 Figure 6 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on diabetes related mortality* (verified through the Office of National Statistics) in the following year (main analysis)* * The second category in all graphs in the reference category Diabetes (ICD-10 codes E10-E16) underlying cause or in the top three causes of death Missing category for diastolic blood pressure omitted because perfectly collinear with missing for systolic. 10

ESM-1 Figure 7 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on cerebrovascular related mortality excluding bleeds (verified through the Office of National Statistics) in the following year (main analysis)* * The second category in all graphs in the reference category Stroke, excluding bleeds, (ICD-10 codes I60-I62 & I64) underlying cause or in the top three causes of death Missing category for diastolic blood pressure omitted because perfectly collinear with missing for systolic. 11

ESM-1 Table 2: Hazard ratios from Cox proportionate hazards survival analysis for all ONS deaths, all ONS diabetes related deaths, all ONS cerebrovascular (excluding bleeds) related deaths (main analysis)* All-cause mortality Diabetes related mortality Cerebrovascular (excl bleeds) related mortality Age 1.069(1.065,1.074) 1.084(1.078,1.091) 1.097(1.081,1.113) Male 1.239(1.195,1.284) 1.151(1.068,1.240) 0.894(0.782,1.022) Complications Macrovascular 1.347(1.295,1.401) 1.548(1.426,1.680) 1.628(1.419,1.868) Microvascular 1.159(1.120,1.200) 1.333(1.242,1.432) 1.047(0.920,1.192) Smoking Never smoked Reference Ex-smoker 1.098(1.050,1.148) 0.988(0.903,1.080) 1.052(0.901,1.230) Current smoker 1.501(1.409,1.598) 1.317(1.158,1.499) 1.268(1.001,1.605) Missing 0.532(0.384,0.737) 0.511(0.262,0.997) 1.396(0.615,3.167) Practice characteristics DM prevalence 0.945(0.926,0.964) 0.929(0.890,0.970) 0.948(0.881,1.019) List size (1000s) 0.996(0.993,0.999) 0.998(0.991,1.005) 1.000(0.988,1.011) Region North East 0.857(0.767,0.958) 1.029(0.828,1.278) 0.694(0.443,1.086) Yorkshire & The Humber 0.958(0.884,1.039) 0.970(0.818,1.150) 1.053(0.788,1.407) East Midlands 0.963(0.874,1.061) 0.808(0.652,1.001) 1.085(0.777,1.515) West Midlands 0.946(0.891,1.004) 1.084(0.962,1.221) 0.907(0.727,1.132) East of England 0.934(0.878,0.992) 0.991(0.875,1.124) 0.961(0.770,1.199) South West 0.917(0.864,0.974) 0.857(0.756,0.972) 0.912(0.734,1.132) South Central 0.981(0.921,1.045) 0.912(0.797,1.044) 0.990(0.785,1.248) London 0.836(0.786,0.890) 0.762(0.667,0.872) 0.999(0.801,1.244) South East Coast 0.851(0.740,0.978) 0.849(0.744,0.968) 0.788(0.620,1.000) Deprivation quintile Comorbidities 1 (most affluent) Reference 2 1.090(1.032,1.152) 1.167(1.039,1.312) 1.183(0.966,1.450) 3 1.087(1.027,1.150) 1.153(1.023,1.300) 1.249(1.017,1.534) 4 1.106(1.046,1.170) 1.182(1.049,1.331) 1.113(0.902,1.375) 5 1.195(1.127,1.267) 1.274(1.125,1.442) 1.243(0.998,1.547) Asthma 1.061(1.002,1.123) 0.933(0.822,1.058) 0.943(0.751,1.183) Coronary heart disease 1.053(1.014,1.093) 0.939(0.868,1.016) 0.848(0.739,0.974) Chronic kidney disease 1.160(1.060,1.269) 1.186(1.050,1.340) 1.135(1.002,1.285) Chronic obstructive pulmonary disease 1.796(1.579,2.044) 1.371(1.211,1.551) 1.105(0.875,1.395) Depression 1.055(1.015,1.096) 1.066(0.985,1.153) 1.033(0.898,1.188) Dementia 1.601(1.501,1.708) 1.816(1.612,2.047) 1.707(1.385,2.104) Serious mental illness 1.444(1.244,1.675) 1.346(1.120,1.617) 1.050(0.744,1.482) Heart failure 1.916(1.705,2.153) 1.643(1.497,1.803) 1.599(1.355,1.887) Hypertension 1.023(0.987,1.060) 0.993(0.922,1.070) 1.096(0.956,1.257) Stroke 1.091(1.043,1.141) 1.082(0.989,1.185) 2.831(2.475,3.238) 12

All-cause mortality Diabetes related mortality Cerebrovascular (excl bleeds) related mortality Cancer 2.823(2.555,3.118) 1.498(1.367,1.641) 1.221(1.033,1.442) Epilepsy 1.385(1.217,1.575) 1.420(1.085,1.859) 2.152(1.519,3.049) Osteoarthritis 0.926(0.895,0.959) 0.938(0.873,1.008) 0.993(0.880,1.120) Osteoporosis 1.106(1.035,1.182) 1.098(0.959,1.257) 1.105(0.881,1.388) Hypothyroidism 0.990(0.942,1.040) 1.032(0.933,1.141) 0.850(0.705,1.023) BMI <18.5 1.998(1.805,2.211) 2.178(1.759,2.697) 1.754(1.203,2.557) 18.5 & 25 Reference >25 & 30 0.637(0.609,0.667) 0.669(0.604,0.740) 0.711(0.603,0.839) >30 & 40 0.555(0.516,0.597) 0.653(0.584,0.730) 0.554(0.456,0.673) >40 0.838(0.763,0.922) 1.106(0.913,1.339) 0.522(0.332,0.821) Missing BMI 3.119(2.971,3.274) 3.585(3.233,3.976) 2.685(2.251,3.202) HbA1c (%) <6.25 (<45mmol/mol) 1.147(1.091,1.205) 1.078(0.962,1.207) 1.117(0.924,1.351) 6.25 & 6.75 (45-50mmol/mol) Reference >6.75 & 7.25 (50-56mmol/mol) 0.941(0.890,0.994) 0.992(0.877,1.121) 1.087(0.884,1.337) >7.25 & 7.75 (56-61mmol/mol) 0.932(0.877,0.991) 0.977(0.853,1.117) 1.196(0.956,1.497) >7.75 & 8.25 (61-67mmol/mol) 1.076(1.005,1.153) 1.285(1.115,1.480) 1.333(1.036,1.715) >8.25 (>67mmol/mol) 1.203(1.138,1.272) 1.494(1.330,1.679) 1.315(1.059,1.633) Missing HbA 1c 0.975(0.907,1.047) 0.927(0.796,1.078) 1.249(0.954,1.634) Total cholesterol (mmol/l) <2.5 1.426(1.274,1.595) 1.405(1.105,1.785) 1.197(0.741,1.932) 2.5 & 3.5 Reference >3.5 & 4.5 0.893(0.856,0.932) 0.890(0.811,0.976) 0.889(0.757,1.043) >4.5 & 5.5 0.907(0.862,0.955) 0.914(0.818,1.021) 0.957(0.795,1.153) >5.5 & 6.5 0.966(0.895,1.041) 0.981(0.834,1.154) 0.980(0.746,1.288) >6.5 1.178(1.056,1.315) 1.184(0.937,1.495) 1.250(0.857,1.823) Missing total cholesterol 1.111(1.045,1.182) 1.271(1.120,1.442) 0.868(0.677,1.112) Systolic BP (mmhg) <115 1.161(1.091,1.235) 1.219(1.066,1.394) 1.218(0.936,1.585) 115 & 125 Reference >125 & 135 0.840(0.800,0.882) 0.888(0.798,0.989) 1.059(0.869,1.292) >135 & 145 0.763(0.726,0.803) 0.780(0.698,0.872) 0.993(0.813,1.214) >145 & 155 0.774(0.729,0.822) 0.825(0.723,0.940) 1.039(0.826,1.306) >155 & 165 0.795(0.735,0.859) 0.812(0.687,0.959) 0.921(0.684,1.240) >165 & 175 0.816(0.729,0.914) 0.896(0.709,1.133) 1.205(0.827,1.756) >175 0.866(0.751,0.998) 0.934(0.705,1.237) 1.517(0.998,2.306) Missing systolic BP 0.637(0.586,0.693) 0.690(0.578,0.823) 0.600(0.428,0.840) Diastolic BP (mmhg) <72.5 1.062(1.018,1.108) 1.100(1.003,1.206) 0.847(0.725,0.990) 72.5 & 77.5 Reference >77.5 & 82.5 0.963(0.914,1.013) 0.951(0.849,1.066) 0.894(0.743,1.075) 13

All-cause mortality Diabetes related mortality Cerebrovascular (excl bleeds) related mortality >82.5 & 87.5 0.915(0.852,0.983) 0.969(0.829,1.132) 0.833(0.646,1.074) >87.5 & 92.5 1.031(0.934,1.138) 1.231(1.007,1.505) 1.097(0.788,1.527) 92.5 1.078(0.941,1.236) 1.016(0.754,1.370) 1.311(0.865,1.988) Missing diastolic BP omitted Model info No. of subjects 118291 118291 118291 No. of failures 14939 3495 1169 Years at risk (no. of observations) 418792 418792 418792 Log pseudolikelihood -132118-29772 -9980 * Models include time-varying covariates so that the proportionate hazard assumption can be met; missing data for biometric measurements have been categorised as such enabling us to use all available records. Office of National Statistics (ONS) deaths in the following year, using data from all available practices. Office of National Statistics (ONS) deaths in the following year where the underlying cause (or in top 3 causes) is diabetes (ICD-10 codes E10-E16), using data for approximately 60% of the practices for which the data has been linked. Office of National Statistics (ONS) deaths in the following year where the underlying cause (or in top 3 causes) is stroke excluding bleeds (ICD-10 codes I60-I62 & I64), using data for approximately 60% of the practices for which the data has been linked. Variables for which additional time-varying components have been added and therefore interpretation of the main effects is not straightforward. Displaying better fit when included an additional logarithmic time-varying component, which implies that the associated hazard increases over time at a logarithmic rate. ONS data only available for England. omitted since missing cases for SBP and DBP overlap completely. 14

Sensitivity analyses multiple imputation for missing data using the twofold algorithm ESM-1 Figure 8 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on all-cause CPRD mortality in the following year (sensitivity analysis)* * The second category in all graphs in the reference category. 15

ESM-1 Figure 9 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on coronary heart disease related mortality (verified through the Office of National Statistics) in the following year (sensitivity analysis)* * The second category in all graphs in the reference category. Ischaemic heart disease (ICD-10 codes I21-I22) underlying cause or in the top three causes of death. 16

ESM-1 Figure 10 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on cerebrovascular related mortality (verified through the Office of National Statistics) in the following year (sensitivity analysis)* * The second category in all graphs in the reference category. Stroke (ICD-10 codes I60-I64) underlying cause or in the top three causes of death. 17

ESM-1 Figure 11 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on developing a new macrovascular diabetes complication in the current year, when none before (sensitivity analysis)* * The second category in all graphs in the reference category. 18

ESM-1 Figure 12 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on developing a new microvascular diabetes complication in the current year, when none before (sensitivity analysis)* * The second category in all graphs in the reference category. 19

ESM-1 Table 3: Hazard ratios from Cox proportionate hazards survival analysis for all CPRD deaths, all ONS coronary heart disease related deaths, all ONS cerebrovascular related deaths and new diabetes complications (sensitivity analysis)* All-cause mortality Coronary heart disease related mortality Cerebrovascular related mortality Macrovascular complication(s) Microvascular complication(s) Age 1.081(1.077,1.085) ** 1.099(1.084,1.116) ** 1.113(1.097,1.129) ** 1.031(1.027,1.034) ** 1.006(1.004,1.008) ** Male 1.182(1.149,1.216) 1.468(1.294,1.666) 0.923(0.814,1.048) 1.418(1.353,1.486) 1.104(1.057,1.154) ** Complications Macrovascular 1.268(1.182,1.361) ** 2.219(1.946,2.532) 1.678(1.476,1.909) Microvascular 1.137(1.106,1.168) 1.208(1.074,1.359) 1.086(0.963,1.224) 1.427(1.364,1.494) Smoking Never smoked Reference Ex-smoker 1.115(1.025,1.212) ** 1.143(0.966,1.351) 0.977(0.844,1.132) 1.326(1.148,1.532) ** 0.984(0.960,1.009) Current smoker 1.824(1.630,2.040) ** 1.903(1.536,2.359) 1.237(0.988,1.549) 2.019(1.697,2.403) ** 0.937(0.905,0.970) Missing 0.854(0.655,1.112) 0.345(0.048,2.495) 1.536(0.672,3.509) 3.209(1.602,6.428) ** 0.676(0.539,0.848) Practice characteristics DM prevalence 0.938(0.907,0.970) ** 0.984(0.917,1.055) 0.955(0.891,1.023) 0.943(0.920,0.968) 1.135(1.106,1.164) ** List size (1000s) 0.999(0.997,1.001) 0.997(0.986,1.008) 1.001(0.991,1.012) 0.995(0.991,1.000) 1.025(1.022,1.028) ** Region North West Reference North East 0.864(0.784,0.952) 0.625(0.418,0.934) 0.729(0.487,1.090) 1.450(1.262,1.667) 1.761(1.650,1.880) Yorkshire- Humber 0.905(0.840,0.976) 0.950(0.735,1.228) 1.003(0.760,1.326) 0.840(0.734,0.961) 1.026(0.961,1.096) East Midlands 0.831(0.768,0.900) 0.832(0.611,1.132) 1.029(0.750,1.414) 0.907(0.796,1.033) 1.322(1.248,1.400) West Midlands 0.942(0.892,0.994) 0.790(0.647,0.964) 0.913(0.743,1.120) 0.773(0.704,0.850) 1.253(1.201,1.308) East of England 0.901(0.852,0.953) 0.792(0.645,0.972) 0.940(0.763,1.158) 0.843(0.766,0.928) 0.673(0.610,0.742) ** South West 0.923(0.874,0.974) 0.768(0.630,0.937) 0.944(0.771,1.155) 0.805(0.732,0.885) 1.439(1.379,1.501) South Central 0.955(0.906,1.007) 0.839(0.678,1.038) 0.952(0.765,1.185) 0.834(0.763,0.912) 0.977(0.902,1.058) ** London 0.836(0.792,0.882) 0.662(0.535,0.821) 0.962(0.781,1.187) 0.828(0.760,0.903) 1.098(1.055,1.144) South East Coast 0.785(0.698,0.883) ** 0.345(0.197,0.605) ** 0.836(0.671,1.042) 0.827(0.755,0.907) 0.982(0.939,1.027) Northern Ireland 0.965(0.896,1.038) 0.991(0.877,1.120) 0.437(0.364,0.525) ** Scotland 1.045(0.993,1.100) 0.976(0.896,1.064) 1.254(1.167,1.347) ** 20

All-cause mortality Coronary heart disease related mortality Cerebrovascular related mortality Macrovascular complication(s) Microvascular complication(s) Wales 1.037(0.988,1.089) 0.876(0.805,0.953) 0.999(0.926,1.078) ** Deprivation quintile Comorbidities 1 (most affluent) Reference 2 1.103(1.058,1.151) 0.934(0.774,1.128) 1.117(0.924,1.351) 1.048(0.977,1.126) 0.956(0.925,0.987) 3 1.079(1.035,1.126) 0.981(0.812,1.185) 1.140(0.939,1.384) 1.077(1.004,1.155) 0.885(0.856,0.914) 4 1.104(1.060,1.151) 0.917(0.760,1.106) 1.066(0.877,1.296) 1.052(0.982,1.128) 0.893(0.855,0.933) ** 5 1.142(1.093,1.193) 1.004(0.824,1.222) 1.196(0.975,1.467) 0.997(0.926,1.074) 0.885(0.833,0.941) ** Asthma 1.029(0.985,1.074) 0.915(0.747,1.120) 0.935(0.757,1.155) 1.027(0.956,1.104) 1.007(0.974,1.042) Coronary heart disease 0.994(0.966,1.024) ** 1.308(1.155,1.481) 0.814(0.715,0.928) 2.972(2.749,3.212) ** 1.027(1.002,1.053) Chronic kidney disease 1.161(1.083,1.244) 1.522(1.350,1.715) 1.110(0.988,1.247) 1.155(1.098,1.215) Chronic obstructive pulmonary disease 1.839(1.670,2.025) ** 1.099(0.897,1.348) 1.003(0.800,1.256) 1.513(1.261,1.814) ** 1.053(1.007,1.102) Depression 1.080(1.048,1.113) 1.011(0.886,1.155) 1.061(0.929,1.211) 1.131(1.076,1.188) 1.070(1.018,1.125) ** Dementia 2.106(1.853,2.394) ** 1.192(0.900,1.577) 1.954(1.602,2.384) 1.229(1.079,1.399) 0.791(0.726,0.860) Serious mental illness 1.599(1.424,1.796) ** 1.266(0.891,1.801) 1.145(0.828,1.585) 1.181(1.031,1.353) 0.928(0.865,0.997) Heart failure 1.883(1.819,1.950) 2.194(1.903,2.529) 1.777(1.525,2.071) 1.731(1.608,1.864) 1.347(1.292,1.404) Hypertension 0.989(0.962,1.017) 1.086(0.957,1.234) 1.074(0.944,1.220) 1.156(1.101,1.213) 1.061(1.039,1.084) Stroke 1.190(1.151,1.231) 0.931(0.802,1.082) 2.895(2.552,3.284) 1.132(1.049,1.222) ** Cancer 3.017(2.792,3.261) ** 0.968(0.815,1.148) 1.201(1.026,1.406) 1.085(1.013,1.163) 1.040(1.005,1.077) Epilepsy 1.645(1.468,1.843) ** 0.843(0.483,1.470) 2.287(1.646,3.177) 1.536(1.271,1.857) 0.990(0.897,1.093) Osteoarthritis 0.915(0.891,0.940) 0.811(0.720,0.913) 0.994(0.887,1.114) 1.068(1.019,1.119) 1.072(1.049,1.097) Osteoporosis 1.097(1.040,1.157) 1.326(1.048,1.679) 1.109(0.894,1.375) 1.144(1.030,1.270) 1.051(0.993,1.111) Hypothyroidism 0.988(0.951,1.028) 0.835(0.695,1.004) 0.885(0.744,1.052) 1.013(0.945,1.086) 1.025(0.992,1.060) BMI <18.5 1.657(1.544,1.778) 1.473(1.011,2.147) 1.466(1.013,2.123) 1.245(0.996,1.556) 0.766(0.648,0.905) 18.5 & 25 Reference >25 & 30 0.689(0.650,0.730) ** 0.910(0.783,1.058) 0.800(0.685,0.934) 0.865(0.815,0.918) 0.970(0.909,1.034) ** 21

All-cause mortality Coronary heart disease related mortality Cerebrovascular related mortality Macrovascular complication(s) Microvascular complication(s) >30 & 40 0.653(0.618,0.690) ** 0.971(0.831,1.136) 0.694(0.589,0.817) 0.738(0.654,0.832) ** 0.858(0.803,0.915) ** >40 1.012(0.947,1.081) 1.511(1.113,2.049) 0.779(0.538,1.127) 0.683(0.614,0.760) 0.806(0.732,0.888) ** HbA1c (%) <6.25 (<45mmol/mol) 1.233(1.188,1.280) 1.089(0.903,1.313) 1.179(0.981,1.418) 0.984(0.915,1.059) 0.957(0.926,0.990) 6.25 & 6.75 (45-50mmol/mol) Reference >6.75 & 7.25 (50-56mmol/mol) 0.975(0.936,1.017) 0.941(0.773,1.144) 1.050(0.865,1.273) 1.042(0.968,1.120) 1.025(0.993,1.059) >7.25 & 7.75 (56-61mmol/mol) 0.975(0.930,1.022) 0.969(0.776,1.211) 1.164(0.947,1.431) 1.091(1.007,1.182) 1.061(1.024,1.099) >7.75 & 8.25 (61-67mmol/mol) 1.109(1.055,1.166) 1.104(0.881,1.384) 1.295(1.009,1.661) 1.187(1.091,1.292) 1.088(1.046,1.132) >8.25 (>67mmol/mol) 1.262(1.208,1.318) 1.355(1.121,1.637) 1.339(1.081,1.659) 1.455(1.357,1.559) 1.239(1.201,1.279) Total cholesterol (mmol/l) <2.5 1.406(1.290,1.533) 1.144(0.740,1.767) 1.072(0.673,1.707) 1.417(1.211,1.657) 1.193(1.081,1.317) 2.5 & 3.5 Reference >3.5 & 4.5 0.883(0.851,0.917) 1.012(0.869,1.178) 0.878(0.760,1.014) 0.901(0.853,0.953) 0.952(0.928,0.977) >4.5 & 5.5 0.927(0.887,0.969) 1.197(1.015,1.413) 0.946(0.801,1.117) 0.903(0.846,0.964) 0.917(0.889,0.945) >5.5 & 6.5 1.015(0.955,1.079) 1.633(1.300,2.053) 0.962(0.761,1.216) 0.878(0.796,0.968) 0.881(0.843,0.921) >6.5 1.146(1.051,1.249) 2.528(1.863,3.432) 1.240(0.876,1.757) 1.240(1.094,1.405) 0.916(0.857,0.979) Systolic BP (mmhg) <115 1.297(1.236,1.362) 1.323(1.054,1.661) 1.201(0.934,1.544) 1.172(1.060,1.296) 1.008(0.959,1.060) 115 & 125 reference >125 & 135 0.838(0.805,0.871) 0.791(0.658,0.950) 1.018(0.847,1.223) 0.962(0.898,1.030) 1.069(1.035,1.104) >135 & 145 0.744(0.714,0.776) 0.930(0.778,1.113) 0.902(0.743,1.095) 0.941(0.877,1.009) 1.098(1.063,1.134) >145 & 155 0.742(0.706,0.780) 0.898(0.729,1.107) 0.952(0.765,1.184) 1.070(0.987,1.160) 1.186(1.141,1.232) >155 & 165 0.750(0.705,0.799) 1.245(0.980,1.582) 0.850(0.643,1.123) 1.280(1.159,1.413) 1.237(1.178,1.300) >165 & 175 0.808(0.740,0.883) 1.065(0.723,1.569) 1.051(0.732,1.510) 1.516(1.322,1.739) 1.259(1.169,1.356) >175 0.845(0.755,0.946) 1.293(0.838,1.994) 1.431(0.962,2.130) 1.690(1.415,2.017) 1.418(1.292,1.556) Diastolic BP (mmhg) <72.5 1.052(1.018,1.087) 1.057(0.914,1.224) 0.853(0.738,0.985) 0.977(0.923,1.035) 1.043(1.015,1.072) 72.5 & 77.5 reference 22

All-cause mortality Coronary heart disease related mortality Cerebrovascular related mortality Macrovascular complication(s) Microvascular complication(s) >77.5 & 82.5 1.023(0.982,1.065) 0.950(0.797,1.132) 0.949(0.801,1.124) 0.911(0.855,0.970) 0.942(0.915,0.970) >82.5 & 87.5 1.030(0.975,1.088) 0.899(0.704,1.148) 0.941(0.746,1.186) 0.850(0.781,0.924) 0.961(0.928,0.996) >87.5 & 92.5 1.115(1.026,1.212) 0.833(0.574,1.208) 1.241(0.897,1.717) 0.839(0.746,0.944) 0.937(0.894,0.982) 92.5 1.264(1.136,1.408) 1.410(0.958,2.074) 1.468(0.987,2.182) 0.981(0.851,1.132) 0.886(0.830,0.945) Model info (5 imputation datasets) No. of subjects 195481 118291 118291 183140 183610 No. of failures 24752 1340 1323 8901 39553 Years at risk (no. of observations) 689140 418792 418792 710068 641955 F 406.7 34.1 51.6 93.2 43.4 * Models include time-varying covariates so that the proportionate hazard assumption can be met; missing data for biometric measurements have been categorised as such enabling us to use all available records. CPRD estimated deaths in the following year, using data from all available practices. Office of National Statistics (ONS) deaths in the following year where the underlying cause (or in top 3 causes) is ischaemic heart disease (ICD-10 codes I21-I22), using data for approximately 60% of the practices for which the data has been linked. Office of National Statistics (ONS) deaths in the following year where the underlying cause (or in top 3 causes) is stroke (ICD-10 codes I60-I64), using data for approximately 60% of the practices for which the data has been linked. New macrovascular complication (peripheral vascular disease, myocardial infarction, stroke or amputation) in the current year, when none previously, using data from all available practices. Stroke was not included as a covariate in this analysis since it overlaps fully with the outcome. Coronary heart disease and heart failure were included since they do not overlap fully with myocardial infarction. New microvascular complication (retinopathy, neuropathy, nephropathy, chronic kidney disease stages 4-5 or foot ulcer) in the current year, when none previously, using data from all available practices. Chronic kidney disease was not included as a covariate in this analysis since there is great overlap with the outcome. Variables for which additional time-varying components have been added and therefore interpretation of the main effects is not straightforward. Displaying better fit when included an additional logarithmic time-varying component, which implies that the associated hazard increases over time at a logarithmic rate. ONS data only available for England. 23

ESM-1 Figure 13 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on all-cause mortality (verified through the Office of National Statistics) in the following year (sensitivity analysis)* * The second category in all graphs in the reference category 24

ESM-1 Figure 14 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on diabetes related mortality* (verified through the Office of National Statistics) in the following year (sensitivity analysis)* * The second category in all graphs in the reference category Diabetes (ICD-10 codes E10-E16) underlying cause or in the top three causes of death 25

ESM-1 Figure 15 a-d: Hazard ratios (CIs) for HbA1c, cholesterol, systolic blood pressure and diastolic blood pressure levels on cerebrovascular related mortality excluding bleeds (verified through the Office of National Statistics) in the following year (sensitivity analysis)* * The second category in all graphs in the reference category Stroke, excluding bleeds, (ICD-10 codes I60-I62 & I64) underlying cause or in the top three causes of death 26

ESM-1 Table 4: Hazard ratios from Cox proportionate hazards survival analysis (multiple imputation) for all ONS deaths, all ONS diabetes related deaths, all ONS cerebrovascular (excluding bleeds) related deaths (sensitivity analysis)* Cerebrovascular All-cause mortality Diabetes related mortality (excl bleeds) related mortality Age 1.084(1.079,1.089) 1.104(1.097,1.112) 1.112(1.095,1.129) Male 1.179(1.138,1.223) 1.086(1.007,1.171) 0.848(0.742,0.970) Complications Macrovascular 1.403(1.349,1.459) 1.624(1.496,1.762) 1.708(1.491,1.958) Microvascular 1.140(1.101,1.180) 1.305(1.216,1.401) 1.035(0.911,1.177) Smoking Never smoked Reference Ex-smoker 1.019(0.974,1.066) 0.913(0.834,0.998) 0.981(0.840,1.147) Current smoker 1.519(1.427,1.616) 1.344(1.182,1.530) 1.271(1.004,1.610) Missing 0.765(0.549,1.065) 0.767(0.393,1.497) 1.756(0.768,4.012) Practice characteristics DM prevalence 0.946(0.927,0.966) 0.931(0.891,0.973) 0.943(0.876,1.014) List size (1000s) 0.998(0.995,1.001) 1.001(0.994,1.007) 1.001(0.990,1.013) Region North East 0.857(0.766,0.960) 1.012(0.815,1.258) 0.687(0.439,1.075) Yorkshire & The Humber 0.988(0.912,1.070) 1.008(0.851,1.195) 1.098(0.822,1.465) East Midlands 0.976(0.888,1.072) 0.816(0.659,1.010) 1.080(0.774,1.507) West Midlands 0.950(0.895,1.008) 1.092(0.969,1.231) 0.906(0.726,1.130) East of England 0.954(0.898,1.013) 1.016(0.896,1.152) 0.975(0.782,1.217) South West 0.920(0.867,0.977) 0.860(0.759,0.975) 0.901(0.725,1.119) South Central 0.986(0.925,1.050) 0.924(0.808,1.058) 0.988(0.784,1.245) London 0.841(0.790,0.896) 0.765(0.668,0.875) 1.000(0.801,1.248) South East Coast 0.859(0.750,0.983) 0.874(0.766,0.996) 0.795(0.625,1.010) Deprivation quintile 1 (most affluent) Reference 2 1.086(1.028,1.148) 1.167(1.038,1.311) 1.160(0.946,1.421) 3 1.087(1.027,1.150) 1.149(1.019,1.296) 1.227(0.999,1.508) 4 1.100(1.040,1.163) 1.179(1.047,1.327) 1.097(0.889,1.354) 5 1.184(1.116,1.257) 1.263(1.115,1.431) 1.213(0.973,1.512) Comorbidities Asthma 1.028(0.972,1.087) 0.889(0.784,1.008) 0.913(0.729,1.143) Coronary heart disease 1.004(0.967,1.042) 0.887(0.819,0.960) 0.816(0.710,0.938) Chronic kidney disease 1.129(1.032,1.235) 1.133(1.003,1.280) 1.094(0.966,1.239) Chronic obstructive pulmonary disease 1.845(1.631,2.086) 1.360(1.203,1.537) 1.113(0.883,1.403) Depression 1.089(1.048,1.132) 1.102(1.017,1.193) 1.063(0.924,1.224) Dementia 1.952(1.831,2.081) 2.275(2.018,2.565) 1.966(1.593,2.426) Serious mental illness 1.533(1.317,1.783) 1.520(1.262,1.830) 1.167(0.833,1.635) Heart failure 2.023(1.808,2.263) 1.792(1.633,1.967) 1.728(1.466,2.036) Hypertension 0.985(0.951,1.022) 0.950(0.881,1.025) 1.063(0.927,1.218) 27

Cerebrovascular All-cause mortality Diabetes related mortality (excl bleeds) related mortality Stroke 1.180(1.129,1.232) 1.184(1.083,1.294) 3.025(2.650,3.453) Cancer 2.823(2.555,3.119) 1.503(1.372,1.647) 1.223(1.035,1.444) Epilepsy 1.504(1.319,1.715) 1.547(1.181,2.025) 2.354(1.669,3.319) Osteoarthritis 0.902(0.871,0.933) 0.910(0.847,0.977) 0.976(0.865,1.102) Osteoporosis 1.122(1.049,1.199) 1.112(0.970,1.274) 1.108(0.884,1.390) Hypothyroidism 0.996(0.948,1.046) 1.039(0.940,1.148) 0.852(0.708,1.026) BMI <18.5 1.617(1.457,1.794) 1.678(1.337,2.105) 1.547(1.059,2.261) 18.5 & 25 Reference >25 & 30 0.717(0.685,0.751) 0.757(0.684,0.837) 0.815(0.697,0.953) >30 & 40 0.671(0.626,0.719) 0.813(0.733,0.903) 0.711(0.600,0.843) >40 1.065(0.978,1.160) 1.387(1.154,1.666) 0.714(0.475,1.073) HbA1c (%) <6.25 (<45mmol/mol) 1.207(1.152,1.265) 1.164(1.040,1.303) 1.123(0.929,1.357) 6.25 & 6.75 (45-50mmol/mol) Reference >6.75 & 7.25 (50-56mmol/mol) 0.947(0.897,1.000) 1.009(0.895,1.138) 1.051(0.859,1.285) >7.25 & 7.75 (56-61mmol/mol) 0.958(0.903,1.016) 1.009(0.886,1.149) 1.152(0.927,1.431) >7.75 & 8.25 (61-67mmol/mol) 1.085(1.016,1.159) 1.299(1.130,1.495) 1.270(0.978,1.647) >8.25 (>67mmol/mol) 1.265(1.197,1.336) 1.617(1.442,1.814) 1.338(1.067,1.679) Total cholesterol (mmol/l) <2.5 1.421(1.259,1.604) 1.421(1.096,1.842) 1.181(0.734,1.899) 2.5 & 3.5 Reference >3.5 & 4.5 0.895(0.851,0.941) 0.898(0.825,0.978) 0.887(0.761,1.034) >4.5 & 5.5 0.950(0.904,1.000) 0.967(0.860,1.088) 0.940(0.784,1.127) >5.5 & 6.5 1.022(0.937,1.116) 1.046(0.902,1.211) 0.999(0.775,1.286) >6.5 1.164(1.043,1.299) 1.197(0.960,1.491) 1.123(0.771,1.635) Systolic BP (mmhg) <115 1.232(1.156,1.314) 1.313(1.147,1.504) 1.280(0.982,1.667) 115 & 125 Reference >125 & 135 0.805(0.765,0.846) 0.844(0.757,0.941) 1.016(0.834,1.237) >135 & 145 0.721(0.684,0.760) 0.726(0.651,0.810) 0.941(0.766,1.156) >145 & 155 0.720(0.678,0.765) 0.746(0.654,0.852) 0.960(0.760,1.212) >155 & 165 0.734(0.678,0.794) 0.730(0.618,0.862) 0.849(0.628,1.149) >165 & 175 0.767(0.683,0.861) 0.829(0.656,1.048) 1.171(0.805,1.703) >175 0.839(0.728,0.966) 0.895(0.675,1.188) 1.487(0.981,2.253) Diastolic BP (mmhg) <72.5 1.041(0.998,1.085) 1.066(0.974,1.166) 0.838(0.718,0.979) 72.5 & 77.5 Reference >77.5 & 82.5 1.015(0.964,1.070) 1.022(0.912,1.145) 0.950(0.792,1.140) >82.5 & 87.5 0.999(0.931,1.071) 1.078(0.927,1.254) 0.937(0.729,1.204) >87.5 & 92.5 1.152(1.043,1.272) 1.397(1.148,1.701) 1.214(0.860,1.714) 92.5 1.278(1.118,1.461) 1.217(0.896,1.651) 1.565(1.041,2.351) 28

Model info (5 imputation datasets) All-cause mortality Diabetes related mortality Cerebrovascular (excl bleeds) related mortality No. of subjects 118291 118291 118291 No. of failures 14939 3495 1169 Years at risk (no. of observations) 418792 418792 418792 F 282.4 90.0 48.8 * Models include time-varying covariates so that the proportionate hazard assumption can be met; missing data for biometric measurements have been categorised as such enabling us to use all available records. Office of National Statistics (ONS) deaths in the following year, using data from all available practices. Office of National Statistics (ONS) deaths in the following year where the underlying cause (or in top 3 causes) is diabetes (ICD-10 codes E10-E16), using data for approximately 60% of the practices for which the data has been linked. Office of National Statistics (ONS) deaths in the following year where the underlying cause (or in top 3 causes) is stroke excluding bleeds (ICD-10 codes I60-I62 & I64), using data for approximately 60% of the practices for which the data has been linked. Variables for which additional time-varying components have been added and therefore interpretation of the main effects is not straightforward. Displaying better fit when included an additional logarithmic time-varying component, which implies that the associated hazard increases over time at a logarithmic rate. ONS data only available for England. 29