The myocardial infarction submodel

The MI submodel is made up of three states: no history of MI, history of MI and death following MI. Transition probabilities between the states can be taken from the UK Prospective Diabetes Study (UKPDS) risk engine,⁹⁸ Framingham⁹³ or the UKPDS outcomes model.⁹¹ In our calculations, Framingham⁹² was chosen as it is the only one that is based on T1DM only.

Unstable angina submodel

The unstable angina submodel is made up of two states: no history of angina and history of angina. Transition probabilities between the states are derived from Framingham.⁹³ They are adjusted according to HbA_1c levels and renal function.

Congestive heart failure submodel

The CHF submodel is composed of three states: no CHF, history of CHF and death following CHF. A logistic regression based on Framingham⁹⁵ generates the risk profile and includes the following risk factors: age, sex, left ventricular hypertrophy, heart rate, SBP, congenital heart disease, valve disease, presence of diabetes, BMI, presence of diabetes and valve disease jointly.

Stroke submodel

The stroke submodel is composed of three states: no stroke, history of stroke and death following stroke. Transition probabilities between the states can be taken from the UKPDS risk engine,⁹⁶ Framingham¹⁵³ or the UKPDS outcomes model.⁹¹ In our calculations, Framingham was chosen as it is the only one that is based on T1DM only.

Peripheral vascular disease submodel

The PVD submodel is made up of two states: no PVD and PVD. Transition probabilities are the same as T1DM and T2DM. A logistic regression based on Framingham⁹⁷ is used to generate the risk for PVD, including the following risk factors: age, sex, blood pressure (normal–high), stage 1 hypertension (yes/no), stage 2 hypertension (yes/no), presence of diabetes, number of cigarettes per day, cholesterol level and heart failure history.

Neuropathy submodel

The neuropathy submodel is made up of two states: no neuropathy and neuropathy. Transition probabilities for T1DM are derived from DCCT.⁹² Transition probabilities are indexed by diabetes duration and are adjusted for HbA_1c levels, SBP and angiotensin-converting enzyme inhibitor (ACEI) use.

Foot ulcer/amputation submodel

This submodel consists of nine states: (1) no foot ulcer; (2) uninfected ulcer; (3) infected ulcer; (4) healed ulcer; (5) uninfected recurrent ulcer; (6) infected recurrent ulcer; (7) gangrene; (8) history of amputation; and (9) death resulting from foot ulcer. Transition probabilities are the same for T1DM and T2DM. Unlike other submodels, this submodel runs in monthly cycles. Therefore, patients may have multiple foot ulcers in a single year.

Diabetic retinopathy submodel

This submodel is composed of 10 states: (1) no retinopathy and not screened; (2) no retinopathy and screened; (3) background diabetic retinopathy (BDR) and not screened; (4) BDR and screened; (5) BDR and wrongly diagnosed as proliferative; (6) diabetic retinopathy and laser (retinal photocoagulation) treated; (7) proliferative diabetic retinopathy (PDR), not screened and no laser treatment; (8) PDR, screened, detected and laser treated; (9) PDR, screened and not detected; and (10) severe vision loss.

Severe vision loss is a terminal state. Transition probabilities for T1DM are derived from DCCT,⁹² and are adjusted for HbA_1c levels, SBP and ACEI use.

Macular oedema submodel

The macular oedema submodel consists of six states: (1) no macular oedema and not screened; (2) no macular oedema and screened; (3) macular oedema, not screened and no laser treatment; (4) macular oedema, screened and not detected; (5) macular oedema, screened, detected and laser treated; and (6) severe vision loss.

Severe vision loss is a terminal state. Transition probabilities for T1DM are derived from DCCT,⁹² and are adjusted for HbA_1c levels, SBP and ACEI use.

Cataract submodel

The cataract submodel is composed of three states: no cataract, first cataract with operation and second cataract with operation. Transition probabilities are the same for T1DM and T2DM and are taken from a study in diabetes outpatients in the UK published by Janghorbani et al.¹⁵⁴

Nephropathy submodel

This submodel is composed of 13 states: (1) no renal complications and no treatment with ACEI; (2) no renal complications and treated with ACEI; (3) no renal complications after ACEI side effects; (4) microalbuminuira and no treatment with ACEI; (5) microalbuminuira, screened, detected and treated with ACEI; (6) microalbuminuira after ACEI side effects; (7) gross proteinuria and no treatment with ACEI; (8) gross proteinuria, screened, detected and treated with ACEI; (9) gross proteinuria after ACEI side effects; (10) end-stage renal disease, treated with haemodialysis; (11) end-stage renal disease, treated with peritoneal dialysis; (12) end-stage renal disease, treated with renal transplant; and (13) end-stage renal disease death.

Data on the cumulative incidence of progression of microalbuminuria and gross proteinuria were taken from the DCCT,⁹² probabilities for the progression from gross proteinuria to end-stage renal disease are based on cumulative incidence data for T2DM patients in the Rochester population.¹⁵⁵ It is assumed that the probability of progression from gross proteinuria to end-stage renal disease is the same for T1DM and T2DM. The probability of progression from end-stage renal disease states to death is dependent on treatment and ethnic group (Wolfe et al.¹⁵⁶). Transition probabilities are adjusted according to patient HbA_1c levels, SBP and concomitant ACEI treatment

Hypoglycaemia submodel

The hypoglycaemia submodel is a state in which the minor and severe hypoglycaemic episodes are counted. Minor hypoglycaemic events are calculated on a daily basis (cycle length = 1 day). For the simulation of severe hypoglycaemic events, the submodel runs four times for each year of simulation. All rates (defined as number of events per 100 patient-years) are adjusted to the 1-day or 3-month cycle length. Therefore, patients can have multiple hypoglycaemic episodes in a single year. The patients may die after a severe hypoglycaemic episode. The definition of severe and minor hypoglycaemia can be refined by the user according to the available data. In our analysis, hypoglycaemic episode rates are treatment specific and any hypoglycaemic episode that required assistance from a third party is considered as severe. It should be noted that in our base-case analysis the probability of death as a result of a severe hypoglycaemic episode was assumed to be zero.

Ketoacidosis submodel

The ketoacidosis submodel has two states: alive and dead (as a result of ketoacidosis). There are no probability adjustments in the ketoacidosis submodel.

Depression submodel

The depression submodel has three states: no depression, depression receiving antidepression programme and depression not receiving antidepression programme. The onset probability of depression is the same for T1DM and T2DM, and is dependent on gender.

Lactic acidosis submodel

This submodel is relevant for T2DM only.

Peripheral oedema submodel

This submodel is relevant for T2DM only.

Non-specific mortality submodel

This submodel consists of two states: alive or dead. The transition probabilities are indexed by age, sex and ethnicity, and reflect the UK life tables.⁹⁴