Should we screen for type 2 diabetes? Evaluation against National Screening Committee criteria

The condition

The first group of issues considered by the National Screening Committee relates to the condition for which screening is proposed. In the case of type 2 diabetes, these issues are relatively uncontroversial. The scale of morbidity and mortality attributable to diabetes is not in question,5 and the longitudinal examination of cohorts has established the overall course of the condition.6 Undiagnosed diabetes is common7; it is not generally characterised by recognised symptoms and is as strongly associated with future risk as diagnosed diabetes.8 Up to 25% of people with diabetes have evidence of microvascular complications at diagnosis, 1 3 and extrapolation of the association between the prevalence of retinopathy and the duration of disease suggests that the true onset of diabetes occurs several years before it is recognised clinically.2 The National Screening Committee's criteria also state that all “cost-effective primary prevention interventions should have been implemented as far as practicable.”4 This is a harsh test, but in the case of diabetes there is clear scope for primary prevention, as trials show that its incidence can be reduced by dietary changes or physical activity.9 Both intensive interventions aimed at individuals and less intensive ones targeted at whole populations may well be cost effective.10 Currently, neither approach has been fully implemented.

The test

Evaluation of the performance of potential screening tests for diabetes is closely related to the issue of determining the diagnostic thresholds for the disease. These have been based on the shape of the distribution of measures of glycaemia and the risk curve relating glycaemia to the future complications of diabetes.11 Ideally, a test would separate those who have the disease from those who do not, which is simple when the distributions of the test results in the two groups are clearly distinguishable (figure). Some populations in the world display bimodality in measures of glycaemia.12 However, this is not the case in the general population in the United Kingdom, where the distribution is unimodal—a characteristic that will inevitably lead to misclassification.7

Diagnostic cut-off points for diabetes are also based on the risk of the complications of the disease. When the risk of microvascular complications is plotted in relation to plasma glucose concentrations two hours after glucose challenge, the curve is flat through the range of normal values but increases steeply at a concentration of around 11.1 mmol/l.13 However, the shape of the risk curve for cardiovascular disease is fundamentally different: it increases gradually across the whole range of glycaemia. 14 15 When these two risk curves are combined with the normal distribution of glycaemia, it is clear that the distribution of microvascular risk related to glycaemia is skewed and that most of the risk is carried by people who have glycaemia values in the established diabetic range. By contrast, the distribution of cardiovascular risk related to glycaemia is different; it is normally distributed, with most of the risk in people with lesser degrees of glycaemia. In relation to cardiovascular risk, glycaemia can be considered as a continuously distributed phenomenon, like blood pressure or cholesterol concentration, rather than one that can be reduced to a binary state.16 This becomes important if screening is designed to have an impact on cardiovascular risk.

Distribution of results of screening

• Download figure
• Open in new tab
• Download powerpoint

Many possible screening methods have been shown to be feasible, acceptable, and accurate when compared with these diagnostic criteria, even though some studies are affected by selection bias.17 Glucose concentrations after fasting and two hours after glucose challenge and glycated haemoglobin values are equally good at predicting the future microvascular complications of diabetes and can be considered as diagnostic tests as well as screening tests.13 Glycosuria detected by urine analysis has a high specificity but a low sensitivity.17 Testing random blood glucose concentrations is more sensitive but a little less specific.17 Risk factor questionnaires have reasonable predictive value,18 as do predictive models based on routinely collected data on risk factors.19 Although many studies have compared and contrasted different screening tests, the major concern is not how to screen but who to invite and whether screening and subsequent treatment result in health gain.

The treatment

The United Kingdom prospective diabetes study in people with established diabetes showed that tight control of hyperglycaemia is effective in preventing the microvascular complications of diabetes.15 Modelling shows that screening for diabetes and subsequent intensive reduction of hyperglycaemia in order to reduce microvascular complications have limited cost effectiveness relative to other preventive health interventions in the age group in which undiagnosed diabetes is prevalent.20 However, this modelling study did not include the benefits of cardiovascular risk reduction, which the United Kingdom prospective diabetes study found to be relatively small. The effectiveness of screening for diabetes and treating the hyperglycaemia to reduce cardiovascular disease depend on the prevalence of undiagnosed diabetes, the background cardiovascular risk, and the reduction in the risk of cardiovascular events in those screened and then treated. The small reduction in relative risk and the relatively low overall prevalence of diabetes and background cardiovascular risk mean that universal screening is unlikely to be cost effective. Screening may be more cost effective in subgroups of people with a high prevalence of undiagnosed diabetes who are also at high risk of cardiovascular complications.

Ideally, people with diabetes detected by screening would not be treated just for hyperglycaemia; consideration would also be given to other associated cardiovascular risk factors. The cardioprotective effects of intensive reduction of blood pressure,21 treatment with angiotensin converting enzyme inhibitors,22 and cholesterol lowering treatment23 in people with clinically diagnosed diabetes have been documented in randomised trials. Multifactorial risk reduction is also effective.24 However, these trials recruited subjects with clinically recognised disease. The size of the benefit in people with disease detected by screening is uncertain but could be assessed.

(Credit: LIANE PAYNE)

Whether the benefits of screening are sufficient to justify the efforts is, of course, a matter of judgment. In making that judgment it would be important to consider the disadvantages of testing for diabetes and attributing this diagnostic label. Although the label might act as an incentive to behavioural change in some people, it could also increase their anxiety and reduce wellbeing.25 However, the diagnosis of diabetes may also have beneficial effects on the behaviour of health professionals. Once diabetes is diagnosed, patients can be included in programmes of recall, education, and review which are associated with improved recording of cardiovascular risk factors, more frequent lifestyle advice, and more aggressive risk reduction.

A prerequisite for screening is that the clinical management of the condition should be optimised. Although much progress has undoubtedly been made in improving care for people with established diabetes, there is still some way to go before the achievable benefits shown in clinical trials are realised throughout Britain.26 As the prevalence of undiagnosed diabetes is approximately equal to that of diagnosed diabetes, any effective screening programme would increase the demand for diabetes care considerably.1 It is unclear whether existing services could meet this demand, especially at a time when the move to improve and intensify care for people with established disease is already increasing workloads.

The screening programme

The National Screening Committee, recognising the high level of evidence needed before a screening programme can be embarked on, stipulates that evidence of effectiveness in reducing mortality or morbidity should come from randomised trials.4 Although evidence from several trials supports intensive treatment of people with clinically diagnosed diabetes, these results may not be generalisable either to all patients with established diabetes or, more particularly, to those with disease detected at screening. As with any programme, one would need to be aware of the classic biases (lead time, length time, and selection) that may affect interpretation of the benefits of screening.17 As there are no trials of entire screening programmes, cost effectiveness has been modelled using data from existing trials of the effectiveness of treatment and from observational studies. 20 27 The conclusions of these modelling studies depend on some crucial assumptions that influence the judgments about whether screening should be undertaken and which subgroups could potentially benefit. None include more recent trial evidence.

Implications for policy

Given the current state of evidence, there is no justification for universal screening for diabetes in the United Kingdom. However, there is some support for screening and intensive treatment in population subgroups in whom undiagnosed diabetes is especially prevalent and cardiovascular risk is high, provided the systems for organisation and management are optimised. In people identified as being at high risk of heart disease, additional testing for hyperglycaemia may be beneficial.

Although the microvascular complications of diabetes are important to public health, an appropriate strategy may be to optimise management of hyperglycaemia and other risk factors in people with known diabetes and to ensure universal screening and prompt delivery of highly cost effective treatments for complications such as eye disease. It is unclear whether screening for diabetes would, in itself, achieve an appreciable health gain.