Diabetes Complications

How diabetes complications develop and affect long-term health

Diabetes complications are acute or chronic health consequences associated with diabetes and its metabolic, vascular, neurological, and treatment-related effects. Their risk is shaped by glycaemia, blood pressure, lipids, smoking, kidney function, duration of diabetes, access to care, treatment, and other patient characteristics. This page explains the major complication pathways and how they are represented in clinical, epidemiological, and economic evaluation.

Acute and chronic complications require different timeframes

Acute complications can develop over hours or days and may require urgent treatment, while chronic complications emerge over years and can progress through several stages. Both can cause death, disability, cost, and reduced quality of life. Economic models should include each when its frequency or consequence differs materially between strategies.

Complication group	Examples	Main modelling consideration
Acute metabolic	Severe hypoglycaemia, diabetic ketoacidosis, and hyperosmolar hyperglycaemic state	Short-term event risk, emergency treatment, recurrence, and mortality
Microvascular	Retinopathy, nephropathy, and neuropathy	Gradual onset, staged progression, screening, and cumulative damage
Macrovascular	Myocardial infarction, stroke, and peripheral arterial disease	Competing events, recurrence, mortality, and background cardiovascular risk
Foot complications	Ulceration, infection, ischaemia, and amputation	Interaction of neuropathy, vascular disease, wound care, and recurrence
Other associated outcomes	Cognitive decline, frailty, liver disease, infection, and some treatment-related harms	Definition, causal evidence, and risk of double counting

Glycaemia is important but not the only driver

Higher cumulative glycaemic exposure increases the risk of several complications, particularly microvascular disease. Cardiovascular and kidney outcomes also depend strongly on blood pressure, lipids, smoking, age, existing disease, and treatments with effects beyond glucose lowering. Describing every complication as a result of poor glucose control oversimplifies causation and can stigmatise patients.

Risk assessment should consider:

• Current and historical glycaemic exposure.
• Duration and type of diabetes.
• Blood pressure, lipid levels, smoking, weight, and physical activity.
• Kidney function and albuminuria.
• Previous cardiovascular, retinal, neurological, or foot disease.
• Treatment exposure, adherence, adverse effects, and access.
• Social and structural conditions affecting prevention and care.

Microvascular complications affect small vessels and tissues

Microvascular complications are classically grouped as diabetic retinopathy, diabetic kidney disease, and diabetic neuropathy. They develop through related but distinct biological pathways and do not necessarily progress together. Early stages may be asymptomatic, making screening important.

Models should preserve clinically meaningful stages because costs, quality of life, treatment, and progression differ. A single state labelled microvascular disease can conceal blindness, dialysis, pain, and foot risk that have very different consequences.

Diabetic retinopathy can threaten vision

Diabetic retinopathy ranges from early retinal changes to proliferative disease and diabetic macular oedema. Vision loss depends on disease stage, macular involvement, timely detection, treatment, and other ocular conditions. Screening can identify treatable disease before symptoms occur.

Economic evaluation can include screening attendance, image quality, grading, referral, treatment, visual acuity, bilateral disease, recurrence, and the costs and utility effects of visual impairment. A retinal lesion outcome should not be treated automatically as equivalent to patient-experienced vision loss.

Diabetic kidney disease progresses through interacting risks

Diabetic kidney disease can involve albuminuria, declining estimated glomerular filtration rate, chronic kidney disease progression, kidney failure, dialysis, transplantation, cardiovascular events, and death. Kidney and cardiovascular risks are closely connected. Models should avoid counting the same mortality or cost consequence through several overlapping pathways.

Important inputs include:

• Estimated glomerular filtration rate and albuminuria categories.
• Rate of kidney-function decline.
• Acute kidney injury and recovery.
• Use and effect of kidney-protective treatment.
• Dialysis modality, transplantation, graft survival, and supportive care.
• Cardiovascular events and competing mortality.

Diabetic neuropathy has several forms

Neuropathy can affect peripheral sensation, pain, autonomic function, and focal nerves. Loss of protective sensation increases the risk of unnoticed injury and foot ulceration, while painful neuropathy directly reduces quality of life. Autonomic neuropathy can affect cardiovascular, gastrointestinal, urinary, and sexual function.

Measurement should distinguish symptoms, signs, functional effects, and clinically diagnosed neuropathy. Treatment can relieve pain without reversing sensory loss, so symptom improvement and future foot risk should not be combined into one outcome.

Foot complications arise from interacting pathways

Foot ulceration and amputation can result from neuropathy, deformity, repetitive pressure, peripheral arterial disease, trauma, infection, and delayed access to care. Recurrence is common enough to require explicit representation when supported by evidence. Amputation level and rehabilitation affect survival, mobility, costs, and quality of life.

A complete pathway can include risk assessment, footwear, podiatry, ulcer onset, infection, revascularisation, healing, recurrence, minor amputation, major amputation, rehabilitation, and death. Access to rapid multidisciplinary care can change both outcomes and resource use.

Macrovascular complications drive substantial mortality and cost

Macrovascular disease includes coronary heart disease, stroke, and peripheral arterial disease. Diabetes can increase baseline risk and worsen prognosis after an event. Recurrent events and combinations of cardiovascular conditions are common and should not be represented as independent one-off events when history changes future risk.

Relevant outcomes include:

• Non-fatal and fatal myocardial infarction.
• Angina, coronary revascularisation, and heart failure.
• Ischaemic and haemorrhagic stroke with different disability levels.
• Peripheral arterial disease, revascularisation, and limb events.
• Cardiovascular death and all-cause mortality.

Hypoglycaemia can be a treatment-related complication

Hypoglycaemia ranges from self-treated symptoms to severe episodes requiring assistance or medical care. It can cause injury, fear, treatment avoidance, emergency use, and rarely death. Risk depends on medicine, dose, meals, activity, kidney function, age, awareness, and previous events.

Models should distinguish non-severe and severe events, nocturnal episodes, recurrence, and any sustained utility or behavioural effect. Self-reported event frequency and claims-based severe-event rates measure different parts of the burden.

Hyperglycaemic emergencies require urgent care

Diabetic ketoacidosis and hyperosmolar hyperglycaemic state are acute, potentially fatal emergencies. They may occur at diagnosis, during illness, after treatment interruption, or with specific treatment-related risks. Economic evaluation should include emergency treatment, inpatient care, intensive care, recurrence, and mortality where material.

Administrative definitions should be validated because diagnosis codes can vary in accuracy and severity. Events should not be double counted as both general hospitalisation and a separate complication cost unless the model structure intends it.

Complications interact rather than occur independently

Kidney disease increases cardiovascular risk, neuropathy and vascular disease combine to increase foot risk, and visual impairment can affect self-management. Treating every event as independent can generate implausible patient histories and underestimate clustering. Patient-level simulation or expanded history states may be necessary when these interactions affect decisions.

The structure should preserve only the history needed to predict future outcomes, costs, or utilities. Excess detail without reliable evidence can create false precision.

Risk equations predict future events

Diabetes models often use multivariable risk equations estimated from cohorts or trials. A risk equation can combine age, diabetes duration, biomarkers, treatment, and previous events to estimate a future probability. Its credibility depends on population fit, predictor definitions, calibration, and whether treatment effects are applied consistently.

For a proportional-hazards model:

$$ h_i(t)=h_0(t)\exp(\boldsymbol{\beta}^{\top}\mathbf{x}_i) $$

where (h_0(t)) is baseline hazard and (\mathbf{x}_i) contains patient characteristics. Transporting this equation to another setting requires validation and often recalibration.

Relative treatment effects must be applied to the right baseline risk

A trial may estimate a relative effect while an economic model predicts patient-specific baseline risk. Applying the effect requires alignment of outcome definition, time, population, and competing risks. The absolute benefit can vary greatly across baseline-risk groups.

If baseline risk is (p_0) and a risk ratio (RR) is appropriate:

$$ p_1=RR\times p_0 $$

$$ Absolute\ risk\ reduction=p_0-p_1=p_0(1-RR) $$

An effect should not be applied twice through both a changed biomarker and an external event hazard ratio unless the modelling method explicitly avoids double counting.

Biomarker-mediated modelling needs causal care

Some models translate treatment-induced changes in glycated haemoglobin, blood pressure, weight, or lipids into complication risk through risk equations. This can support long-term projection when event data are immature. It assumes that changing the biomarker with the intervention produces the predicted outcome effect and that other treatment effects are represented appropriately.

Analysts should compare biomarker-mediated predictions with direct outcome evidence, test treatment-class effects, and explore waning. Surrogate relationships should not be treated as universal across mechanisms.

Cumulative exposure and legacy effects affect timing

Current biomarker values may not fully capture the effect of previous exposure. Complication risk can reflect duration, cumulative glycaemia, and prior treatment history. Conversely, benefits of improved control can emerge gradually rather than immediately.

Models should state whether they use current values, lagged values, cumulative averages, treatment-history variables, or direct event effects. The choice can materially change lifetime projections.

Competing risks become important over long horizons

Death from one cause prevents later complications, and experiencing one complication can change the probability of another. Summing independently estimated event risks can exceed a plausible total. Competing-risk methods or coherent simulation should preserve mutually exclusive outcomes within each interval.

For cause (k), cumulative incidence can be represented as:

$$ F_k(t)=\int_0^t S(u^-)\lambda_k(u),du $$

where (S(u^-)) is survival free of any event immediately before (u). Censoring competing deaths as ordinary loss to follow-up estimates a different quantity.

Screening changes detection and treatment pathways

Retinal screening, kidney testing, foot assessment, and cardiovascular risk review can identify disease or risk before symptoms occur. Screening has value only when detection leads to effective and accessible action. Attendance, test performance, referral, treatment capacity, and follow-up determine realised benefit.

Economic models should not count earlier diagnosis as a health gain by itself. They should connect detection to treatment, changed progression, adverse effects, resource use, and patient outcomes.

Prevention acts through several risk factors

Prevention includes glycaemic management, blood-pressure control, lipid management, smoking cessation, kidney-protective treatment, foot care, screening, education, nutrition, activity, and access to ongoing care. Interventions can affect several complications at once. Their joint effects should not be estimated by simply adding overlapping relative reductions.

Treatment burden and harms also matter. Very intensive management can increase hypoglycaemia, side effects, monitoring, and patient workload even when it improves a biomarker.

Health-state models need clinically coherent progression

A cohort Markov model can represent broad complication states, while microsimulation can preserve individual histories and simultaneous conditions. The choice depends on whether event history, biomarker trajectories, and interacting risks change future outcomes. State-transition logic should prevent impossible recovery or duplicated events.

For cohort occupancy vector (\mathbf{s}_t) and transition matrix (\mathbf{P}_t):

$$ \mathbf{s}_{t+1}=\mathbf{s}_t\mathbf{P}_t $$

Rows of (\mathbf{P}_t) should sum to one, and transition probabilities should reflect cycle length and competing events.

Costs occur at event onset and during continuing care

Complications can generate acute event costs, ongoing management, rehabilitation, medicines, devices, informal care, and future monitoring. A myocardial infarction has a first-year cost and a continuing post-event cost; dialysis has recurring costs; amputation adds surgery, prosthetics, rehabilitation, and long-term support. One-off and recurring costs should be separated.

For complication (k):

$$ C_k=C_{acute,k}+\sum_{t=1}^{T}C_{followup,k,t} $$

Costs should use a stated perspective, jurisdiction, price year, and discounting method. Overlapping admissions and comorbid care require safeguards against double counting.

Complications reduce quality and length of life

Complications can affect symptoms, mobility, independence, vision, pain, treatment burden, mental health, and survival. Utility decrements may differ at onset and during long-term survival. Combining several decrements additively can produce implausible utilities and should be validated.

Approaches include additive, multiplicative, minimum, or regression-based combination rules. The method should reflect evidence and be tested in sensitivity analysis, especially for people with several complications.

Diabetes complication models need external validation

A model can reproduce its development dataset yet perform poorly in contemporary populations with different treatment and risk. Validation should compare predicted and observed event rates, survival, biomarker trajectories, and complication prevalence over relevant periods. Calibration-in-the-large and calibration across risk groups are important.

Validation should also examine:

• People with type 1 versus type 2 diabetes.
• Newly diagnosed versus long-duration disease.
• Different ethnic, socioeconomic, and geographic populations.
• Contemporary use of cardioprotective and kidney-protective treatments.
• Subgroups with existing cardiovascular or kidney disease.
• Extreme ages and risk-factor values.

Uncertainty extends beyond parameter ranges

Long-term projections depend on event equations, treatment-effect pathways, biomarker trajectories, recurrence, mortality, costs, and utility combination. Structural uncertainty can be larger than the standard error around one coefficient. Alternative credible structures should be tested explicitly.

Useful scenarios include:

• Direct outcome effects versus biomarker-mediated effects.
• Alternative risk equations and recalibration.
• Different treatment-effect duration and waning.
• Alternative recurrence and history dependence.
• Alternative complication-cost and utility combination rules.
• Healthcare versus societal perspectives.
• Alternative mortality and competing-risk structures.

Equity shapes complication risk and care

Complication rates reflect access to prevention, monitoring, medicines, nutritious food, safe activity, stable housing, health literacy support, and specialist care. Social disadvantage can increase exposure and reduce timely treatment. Models that adjust away these differences without examining their causes can conceal inequity.

Equity analysis should examine baseline risk, screening, treatment uptake, adherence, outcomes, and financial burden by relevant population groups. Interventions should be assessed for whether they narrow or widen complication gaps.

Worked lifetime-risk example

Suppose a validated model estimates a 10-year kidney-failure risk of 12% under current care and an appropriate treatment risk ratio of 0.75. Under a simplified constant relative-effect application, treated risk is 9% and absolute risk reduction is 3 percentage points.

$$ p_1=0.75\times0.12=0.09 $$

$$ ARR=0.12-0.09=0.03 $$

The corresponding simplified number needed to treat is about 34 over ten years. This calculation requires compatible outcome definitions and does not by itself account for competing death, treatment discontinuation, adverse events, or uncertainty.

Common mistakes

Diabetes complications are often modelled as a list of independent events driven only by glycated haemoglobin. That simplification can misrepresent modern treatment and clustered disease. The following errors should be checked before results inform a decision.

• Attributing every complication solely to poor glycaemic control.
• Treating microvascular and macrovascular events as independent.
• Applying a relative effect to a mismatched baseline risk or outcome.
• Counting both biomarker-mediated and direct treatment effects twice.
• Ignoring recurrent events, previous history, and competing mortality.
• Combining several utility decrements without checking plausibility.
• Counting a complication admission both as an event cost and general hospital use.
• Using historical risk equations without contemporary recalibration.
• Treating screening detection as benefit without effective follow-up.
• Reporting average outcomes without examining access and equity.

Reporting diabetes complications in an economic model

Transparent reporting should allow readers to follow each complication from risk factor to event, consequence, recurrence, and death. It should identify which relationships are observed, calibrated, or extrapolated. The exact risk-equation version and population are essential for reproducibility.

• Define every acute and chronic complication and stage.
• State risk equations, predictors, coefficients, calibration, and validation.
• Explain treatment-effect application, biomarker mediation, and waning.
• Report cycle length, event ordering, history, recurrence, and competing risks.
• Separate acute and continuing costs with perspective and price year.
• Report utility sources and the rule for multiple complications.
• Show absolute events, life-years, QALYs, costs, and uncertainty.
• Report subgroup performance, equity implications, limitations, and update date.

The decision standard

A credible evaluation of diabetes complications represents the interacting pathways through which diabetes, cardiovascular risk, kidney disease, treatment, and access shape long-term health. It connects prevention and screening to patient-relevant outcomes without reducing disease to one biomarker. The model should remain clinically coherent, externally validated, transparent about uncertainty, and responsive to changes in treatment and population risk.