Screen and intervene to prevent diabetes?BMJ 2017; 356 doi: https://doi.org/10.1136/bmj.i6800 (Published 04 January 2017) Cite this as: BMJ 2017;356:i6800
- Norman Waugh, professor of public health medicine and health technology assessment
The BMJ includes a review of screening tests for non-diabetic hyperglycaemia (“pre-diabetes” is unsatisfactory because many people so labelled do not develop diabetes) and of interventions to reduce or delay progression to type 2 diabetes. If we screen for diabetes, we may, depending on test and cut-offs used, identify more people with impaired glucose tolerance and impaired fasting glucose (the two forms of non-diabetic hyperglycaemia) than with diabetes. Screening for impaired glucose tolerance, however, meets most, but not all, of the National Screening Committee criteria.1
In their study, Barry and colleagues2 conclude that there is no perfect screening test for non-diabetic hyperglycaemia. They report that glycated haemoglobin (HbA1c) is only about 50% sensitive, but the review does not provide sensitivities for a range of thresholds. HbA1c could be 100% sensitive if a low enough threshold was chosen, but at the cost of substantially more false positives. In screening, there is usually a trade-off between sensitivity and specificity. The researchers found that fasting plasma glucose was even less sensitive—25% in the studies included—but gave fewer false positives than HbA1c. In 2015, Gillett and colleagues3 concluded that for type 2 diabetes, screening and intervention based on HbA1c was more cost effective than those based on fasting plasma glucose.
Barry and colleagues also report that different measures of hyperglycaemia identified groups of people that only partly overlapped, which has been reported by others.4 5 In 2253 people aged over 40 and not known to have diabetes, Jones and colleagues from Exeter5 found that HbA1c and fasting plasma glucose testing both identified about 16% of people as being at high risk of type 2 diabetes, but only about half of those classified as high risk by HbA1c were identified as high risk by fasting plasma glucose, and vice versa.
Given the imperfections of HbA1c and fasting plasma glucose, it is a pity that more research has not been done on the non-fasting 50 g glucose challenge test, much used in screening for gestational diabetes but rarely used in screening for type 2 diabetes or impaired glucose tolerance. Previous studies suggest that this test may be accurate and cost effective in screening for prediabetes and undiagnosed diabetes in the general population.6 7
Barry and colleagues conclude that lifestyle interventions can reduce progression to type 2 diabetes, but by only 20% in the longer term. They note the biases in trials that recruited volunteers with impaired fasting glucose from much larger populations. One implication is that the risk reductions reported by trials could be much greater than will be seen when a screen and treat policy is implemented as in the NHS Diabetes Prevention Programme. The Public Health England review8 that underpinned the development of this NHS programme reported an average reduction of 26% in diabetes incidence—much less than the 58% seen in the landmark trials, the Finnish Diabetes Prevention Study9 and the US Diabetes Prevention Programme.10 The Public Health England reviewers concluded that the lower reduction in incidence was because the studies they included were in more real life settings than the landmark trials.
However, most of the studies in the Public Health England review lasted only 12 months. A short interval between the end of the intervention and the measurement of outcomes is a common weakness in the literature on preventing diabetes, and we know that improvements might not be sustained once the intervention has finished.11 Barry and colleagues do not provide data on the intervals between end of interventions and reporting of outcomes in included studies.
Another issue in preventing diabetes is the balance between the medical model of screening and treating of individuals and the public health model of changing behaviour in the whole population at risk. For example, people with HbA1c between 6.0% and 6.4% (the “pre-diabetes” range) are at increased risk of type 2 diabetes, but people who develop diabetes come from the larger proportion of the population with baseline HbA1c lower than 6.0%. In the EPIC-Norfolk study,12 two thirds of participants who developed diabetes had baseline HbA1c lower than 6.0%. Public health measures targeted at the whole population at risk could include interventions to help weight control such as changes to taxation of foodstuffs, for example, making diet versions of soft drinks cheaper than sugary versions, and interventions to make physical activity easier or safer such as cycle lanes separate from traffic.
The main problem with both models is adherence. In the Finnish Diabetes Prevention Study, the 14% of the intervention group who achieved four or five lifestyle targets had only 23% of the risk of diabetes compared to those who achieved none.13 The Cardiovascular Health Study suggested that 90% of cases of type 2 diabetes could be avoided by adherence to lifestyle factors, including physical activity, a healthy diet, and body mass index of under 25.14 Adherence to lifestyle advice remains poor. Preventing or delaying type 2 diabetes requires effective measures to motivate the general population to protect their own health.
Competing interests: I have read and understood the BMJ policy on declaration of interests and declare the following interests: none.
Provenance and peer review: Commissioned, not externally peer reviewed.