Screening for type 2 diabetes: how to meet the NSF target
Dr David Whitford explains where to put in the effort to gain returns on screening, and researcher Dr Colin Greaves gives an account of targeted screening in practice
The argument for earlier detection of type 2 diabetes through either increased awareness of symptoms or through screening follows a logical progression. There is a delay from disease onset to diagnosis and in this time people are unaware they have diabetes. Around 50 per cent of all type 2 diabetes is undiagnosed and 50 per cent of people with newly diagnosed type 2 diabetes have at least one complication.
Active management of blood glucose, hypertension and other factors has been shown to reduce the rate of complications. Suitable screening tests are available and thought to be acceptable to the population1-4. It is therefore argued that earlier detection will lead to earlier treatment and a subsequent reduction in microvascular complications. In addition, cardiovascular risk factors may be targeted more aggressively, leading to a reduction in macrovascular complications and mortality.
The national service framework for diabetes requires GPs – and other clinicians – to take steps to detect undiagnosed patients.
But type 2 diabetes does not meet all the National Screening Committee's (NSC) criteria for screening5. In particular there is no evidence that screening leads to significant improvements in outcomes, there is no specific data on the harm and costs of screening, and there is no consensus on a single specific and sensitive test.
Research is ongoing and the NSC is due to report again to the Department of Health in 2005.
Should GPs screen for type 2 diabetes?
The NSC concluded in a report in 2000: 'There is no justification for universal screening for diabetes in the UK. 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.' In other words, population-based screening is not yet justified, but screening aimed at high-risk groups may be merited.
Who should we screen?
Based on the above recommendation, patients at higher risk of developing type 2 diabetes should be targeted first.
Risk factors include:
lIncreasing age, particularly over 40 years
lObesity (=120 per cent desirable body weight or a BMI =27kg/m2)
lBody fat distribution (central adiposity)
lFirst-degree relative with type 2 diabetes
lMembers of a high-risk ethnic population, such as Indo-Asian, Afro-Caribbean, Hispanic, Native American
lPrevious diagnosis of impaired glucose tolerance (IGT) or impaired fasting glycaemia (IFG)
lPrevious history of gestational diabetes or has delivered a baby weighing more than 4kg
lHypertension (BP =140/90)
lLack of physical activity
lReduced fetal growth
Unfortunately, this is nearly tantamount to population screening as 78 per cent of adults in the US had at least one diabetes-related risk factor or complication. Further targeting aimed at those with the highest cardiovascular risk is therefore necessary.
How do we identify high-risk groups?
Many practices have readily accessible computerised data, particularly of cardiovascular risk. This data can be used to develop predictive models6.
Questionnaires can also be used to screen populations to identify high-risk individuals. An alternative approach is to target patients above an agreed cardiovascular risk (for example, 10-year risk >15 per cent). At present, there is no agreed policy on identifying a target group and each practice should decide what its practical constraints are in terms of resources and data. Targeted screening may need to commence with patients already attending primary care (such as those with hypertension). Further screening could target patients at high risk and identifiable from computer data – such as patients with IGT, obesity, family history of diabetes and smokers.
What screening tests are available?
The methods used for screening (see above right) range from patients self-testing for post-prandial glycosuria4;7 or blood glucose1, testing for glycosuria within the surgery8, blood samples for glucose or HbA1c3;9;10, questionnaires11,12, and predictive models13. The majority of response rates in these studies were greater than 70 per cent in a primary care setting.
There is no established consensus on which tests to use for screening. The simplest methods appear to be the use of questionnaires, predictive models and patient self-testing.
How many cases will screening detect?
The majority of publications on screening for type 2 diabetes have studied populations with an inflated prevalence of diabetes.
The consequence is the predictive value of the tests is thereby inflated. Most methods would detect one person with diabetes for every 50-150 tests performed.
What screening interval should be used?
There has not been sufficient work done to answer this question clearly. A three- to five-yearly screening interval is currently recommended.
What problems can screening cause?
Follow-up, compliance, misdiagnosis and anxiety in screened patients are all problems associated with screening. The evidence for the cost-effectiveness of screening is at present weak. There is also the problem of 'labelling' an asymptomatic person and encouraging them to undertake major lifestyle changes, including medication. Above all, the use of primary care resources to provide a screening programme in terms of time and cost need to be weighed against other priorities.
What should happen with a screen-detected patient?
Screening tests are not intended to diagnose a disease, but to identify people who need diagnostic tests. It is imperative the diagnosis is accurately made.
What should happen with patients identified as having impaired glucose tolerance?
Patients with impaired glucose tolerance are at higher risk of cardiovascular disease and diabetes and should be appropriately managed.
There is no clear consensus on how to screen for type 2 diabetes. Practices wishing to introduce screening should use their resources to identify those patients at highest risk, either through questionnaires or readily available computer data (age, body mass index, smoking status, hypertension and family history of diabetes). These patients can then be invited for diagnostic tests.
lIn the symptomatic patient a diagnosis of diabetes can be made on a single random venous plasma glucose level =11.1mmol/l or a fasting plasma glucose concentration =7.0mmol/l.
lIn the asymptomatic patient diagnosis must be based on at least two abnormal tests: (FPG =7.0mmol/l; two-hour OGTT =11.1mmol/l; or random plasma glucose =11.1mmol/l). At least one additional test on another day with a value in the diagnostic range is essential. If the fasting or random values are not diagnostic, the two-hour value should be used.
lAll patients with impaired fasting glycaemia (fasting plasma glucose =6.1mmol/l <7.0mmol )="" should="" have="" an="" oral="" glucose="" tolerance="" test="" to="" exclude="">7.0mmol>
lA diagnosis of diabetes should never be made on the basis of glycosuria or a stick reading of a finger-prick blood glucose alone.
lUse of glycosylated haemoglobin (HbA1c) is not recommended for diagnosis of diabetes.
01 Falkenberg M,von Schenck H. Screening for diabetes mellitus using a self-administered blood spot test mailed for analysis. Pract Diab 1988;5:159-60
02 Davies M et al. Self testing for diabetes mellitus. BMJ 1991;303:696-8
03 Worrall G. Screening healthy people for diabetes: is it worthwhile?
J Fam Pract 1991;33:155-60
04 Bullimore SP, Keyworth C. Finding diabetics – a method of screening in general practice. Br J Gen Pract 1997;47:371-4
05 Wareham NJ, Griffin SJ. Should we screen for type 2 diabetes? Evaluation against National Screening Committee criteria. BMJ 2001;322:986-8
06 Griffin SJ et al. Diabetes risk score: towards earlier detection of type 2 diabetes in general practice. Diabetes/metabolism research reviews 2000;16:164-71
07 Davies MJ et al. Community screening for non insulin-dependent diabetes mellitus: self-testing for post-prandial glycosuria. QJ Med 1993;86:677-84
08 Fairley R. Audit report. How effective is case finding at detecting diabetes and hypertension in the community? BMJ 1987;294:1202
09 Forrest RD et al. Glucose intolerance and hypertension in north London:
The Islington Diabetes Survey. Diabetic Med 1986;3:338-42
10 Forrest RD et al. The glycohaemoglobin assay as a screening test for diabetes mellitus: The Islington Diabetes Survey. Diabetic Med 1987;4:254-9
11 Burden ML, Burden AC. The American Diabetes Association
screening questionnaire for diabetes. Is it worthwhile in the UK?
Diabetes Care 1994;17:97
12 Welborn TA et al. Australian diabetes screening study: impaired glucose tolerance and non insulin-dependent mellitus. Metabolism 1997;46:35-9
13 Park PJ et al. The performance of a risk score in predicting undiagnosed hyperglycemia. Diab Care 2002;25:984-8
David Whitford worked with Sue Roberts, the current diabetes tsar, to develop a diabetes service in North Tyneside, where he worked as a GP until last year. He is now senior lecturer in general practice at the Royal College of Surgeons, Dublin
Cheap and simple screening
Dr Colin Greaves outlines a practice approach to detecting diabetes
a practice approach to
Our study1 of 16 practices has shown that targeted screening using existing GP records is feasible. The system we used is relatively straightforward. You computer-search your practice database for patients aged over 50 with a body mass index (BMI, ever recorded) of 27 or more. Age 30 or 40 with no limit on BMI is more appropriate for Asian and (probably) Afro-Caribbean populations.
Then you invite the patients on the list for a fasting plasma glucose (FPG) test. Invitees should be followed up by phone to confirm attendance and the fasting procedure.
If the result is 6.1mmol/l or more, then you arrange a second test (you can use a second FPG, but a two-hour glucose tolerance test is recommended for maximum detection rate). Diagnostic values for impaired fasting glycaemia (IFG), impaired glucose tolerance (IGT – hyperglycaemia) and diabetes can be found on the Diabetes UK website (www.diabetes.org.uk/infocentre/
Outcomes for Caucasian populations
We can estimate that in a typical practice of 5,000 patients, with recorded BMI data for 77 per cent of patients and 61 per cent uptake (as in the study) we would expect to invite 555 (11.1 per cent of the population met the selection criteria) and test around 340 people. This will detect around nine new cases of diabetes and 18 cases of IFG. If BP is also monitored, our research data also suggested 37 would also have previously undiagnosed high blood pressure (=160 systolic or =90 diastolic on two occasions), indicating possible hypertension.
Outcomes for Asian populations
Recent studies in Bradford2 suggest we should screen all patients over 30, irrespective of BMI. In a practice with 500 Asian patients we would expect to invite 310 (62 per cent of the population are over 50) and test 189. This would detect around 33 cases of diabetes, and at least as many cases of IFG /IGT.
Workload and cost implications
The workload per 500 people screened is estimated to be one-two hours constructing the necessary computer searches, 100 hours of receptionist time (booking and following-up appointments), 80 hours of practice nurse clinic time (61 per cent uptake, 15 minutes per patient), plus around 12p per blood test.
The total cost will clearly vary depending on pay scales, but would probably be £4,000-£5,000.
The ongoing workload and costs involved in treating new cases of diabetes and of hyperglycaemia/high blood pressure are not included here. But the costs of not treating these conditions should also be considered (increased complications on eventual diagnosis, progression from IFG /IGT to frank diabetes, progression from hypertension to heart disease).
The British Diabetic Association recommend treating IFG /IGT with lifestyle advice and (if glucose is still above 6mmol/l after three months)
to consider an oral hypoglycaemic
Solid evidence from large studies in the US and Finland shows minor changes in diet and physical activity can substantially reduce progression from IGT to diabetes, and practices should inform themselves about best-practice in encouraging this – (www.cochrane.org/ cochrane/revabstr/g480index.htm).
The optimum period at which practices should repeat screening is not known, but current US guidelines recommend repeating screening for diabetes every three years.
For practices with low BMI coverage, a useful first step would be to mount a campaign of systematically recording BMI for all adults. As around 85 per cent of patients will attend the practice for some reason within a two-year period, a practical approach may be to ask patients to record their own height and weight in the waiting room.
Other pragmatic ways to enhance the screening effort include adding diabetes testing to cardiovascular and other routine clinics, where patients have known risk factors.
Pros and cons
The targeted screening approach outlined has several advantages: the identification of a high-risk patient list is simple and quick; FPG testing is easier to implement in general practice than cumbersome glucose tolerance tests; it is relatively cheap and the workload seems feasible.
But the workload implications of treating detected illness, especially in areas with high Asian populations, require additional resource planning.
Not all cases of undetected diabetes in the population would be detected by this method – our study group estimated that some 30-50 per cent would be found. Cases will be missed due to:
ldiabetes in <50 or="">50><27bmi>27bmi>
lnon-attendance/refusal of screening offer
lpatients not identified due to either incorrect or unrecorded BMI.
So, even though this cheap and fast system can detect substantial numbers of cases, practices will need to develop strategies for identifying the remainder. Constructing separate lists for
different ethnic groups may also present
a significant practical challenge.
1 Greaves CJ et al. A simple pragmatic system for detecting new cases of type 2 diabetes in primary care. Family Practice 2004;21(1):57-62
2 Shahid A et al. Comparison of haemoglucated glycosylated haemoglobin and laboratory measurements of random blood glucose in opportunistic case detection of type 2 diabetes in general practice. Paper in submission
Colin Greaves manages the Mid-Devon Primary Care Research Group