UKPDS Risk Engine

Fequently Asked Questions

Detailed answers to most UKPDS Risk Engine queries can be found in the associated methodology papers which are listed under publications. Brief answers to many frequently asked questions are as follows:

Click / tap a question to see the answer:

I am having problems downloading the UKPDS Risk Engine

There are known IE problems associated with some web browsers. You can overcome these by following the instructions given below:

After that it should all work as described in the FAQ question "I've downloaded the UKPDS Risk Engine for Windows, how do I install it?"

I have downloaded and installed the UKPDS Risk Engine but it will not work

You must ensure that you have administrator rights on your computer. Please query this with your IT department. Typically you may see any of the following messages:-

Why are confidence intervals not symmetric in version 1.1 onwards?

Version 1.1 calculates confidence intervals on a logistic scale to avoid them exceeding 100% or becoming negative. The intervals calculated will be similar to those in Version 1.0 but may be asymmetric, rather than symmetric.

I've just installed a new Windows' version of the Risk Engine and every time I try to use it I get an error message 'Invalid value for Time', what does this mean?

This can occur when a new version is not installed successfully. To solve the problem, uninstall the Risk Engine and then re-run the new version installer as follows:

I've downloaded the UKPDS Risk Engine for Windows, how do I install it?

The program you download is a self installing application. It is not itself the application that calculates the risks. Installation is as follows:

You can run the software by choosing 'UKPDS Risk Engine' from the 'UKPDS Risk Engine' section of the 'Programs' part of the 'Start' menu.

For this procedure to work under Windows NT, 2000 or XP you will need to be logged in with administrative rights to the computer.

Do new versions of the Risk Engine give the same answers as older versions?

The estimated risk for CHD and stroke risk are the same in all versions. The confidence intervals may vary slightly between versions 1.0 and later releases, as the confidence interval method was improved in version 1.1 onwards.

Does the Risk Engine take microalbuminuria into acount?

The current version of the UKPDS Risk Engine does not include the presence of microalbuminuria as a risk factor. A new version is being prepared which looks at the risk of cardiovascular disease in people with new or established diabetes in which the presence or absence of microalbuminuria will be included.

Does the Risk Engine take obesity into acount?

We have examined in detail all available measures of size and obesity but they do not contribute independently to estimated risk in the presence of the other more informative risk factors used currently by the UKPDS Risk Engine. The only time we have found this to be the case is when estimating the risk of congestive heart failure as reported in UKPDS Paper 68 (Diabetologia 2004; 47: 10: 1747-1759) describing the UKPDS Outcomes Model

For what populations is it valid to use the UKPDS Risk Engine?

The Risk Engine is intended for use in adults who have:

The Risk Engine is likely to be most accurate when used with patients similar to those in the UKPDS cohort, that is, aged between 25 and 65 at diagnosis of diabetes and with a duration of diabetes less than 20 years.

Can the UKPDS Risk Engine be used for individuals who do not have diabetes?

No. To calculate coronary or stroke risk in individuals without diabetes, other models such as the Framingham risk equations should be used (See Anderson et al, Cardiovascular disease risk profiles, American Heart Journal 1990; 121: 293-8). A computer implementation of the Framingham equations is available from the British Hypertension Society (www.hyp.ac.uk/bhs/managemt.html) or the Joint British Guidelines (www.bnf.org/calculatorrisk.htm)

Can the UKPDS Risk Engine be used for diabetic individuals with established cardiovascular disease?

Not at present. The Risk Engine requires that there be no history of coronary heart disease or a stroke. Work is in progress to develop a model that predicts risks for individuals who already have coronary heart disease or have had a stroke.

Can the UKPDS Risk Engine be used in populations with impaired glucose tolerance (IGT) or impaired fasting glucose (IFG)?

The performance of the Risk Engine in these "pre-diabetic" populations has not been assessed. Results obtained for any individuals with IGT or IFG should be interpreted with caution.

In what ethnic groups can the Risk Engine be used?

The Risk Engine utilises data from self-reported White Caucasian, Afro-Caribbean and Asian-Indian individuals and takes account of these ethnic groups as necessary. Whilst South Asians are recognised generally to be at greater risk of heart disease, we did not find South-Asian ethnicity to be an independent risk factor once all of the traditional risk factors were included in a a multivariate model. Results obtained for individuals from any other ethnic groups should be interpreted with caution.

Why are the risks not different for non-smokers and ex-smokers?

No significant differences were found in coronary heart disease or stroke risk between non-smokers and ex-smokers but the "ex-smoking" category has been maintained as it is likely to be required in future versions of the UKPDS Risk Engine which address other diabetic complications. This ex-smoking effect, however, may be time-dependent with an elevated risk in the first few years after giving up smoking with no difference in later years. In the case of a very recent ex-smoker we recommend treating results obtained as potential underestimates of risk.

Why is there an option to specify the number of measurements for each input?

Where HbA1c, systolic blood pressure or lipid levels are entered as a mean of several values, this option permits a regression dilution adjustment to be made that takes account of the biological and measurement variation seen in these variables.

Why do the UKPDS Risk Engine results not match the tables in the published papers?

The Windows version of the Risk Engine automatically adjusts for regression dilution (see question 10) whereas the published tables are unadjusted. Setting the option "Number of values" to "2" for HbA1c, "6" for blood pressure and "2" for cholesterol will give identical results for the same input values and the message "Adjusted for regression dilution" will no longer appear below the output boxes.

Why do results in the Windows™ version of the UKPDS Risk Engine not match results in the Excel version of the Risk Engine?

The Windows™ implementation adjusts for regression dilution (see questions 10 & 11) by default and uses "Age now". The Excel implementation does not automatically adjust for regression dilution and uses "Age at diagnosis of diabetes". This is because the Windows™ version has been designed for use in clinical practice whereas the Excel version is intended primarily for use by researchers and by health planners. The two implementations will give identical results if equivalent settings are used.

Does it matter how the input variables are measured?

For maximum accuracy, input variables should be measured in a similar fashion to the UKPDS. Specifically, HbA1c should be measured on a DCCT/UKPDS-aligned assay, total cholesterol and HDL cholesterol measured on CDC-aligned assays and blood pressure measured as the mean of three measurements taken over a few minutes.

Does it matter that patients may have had diabetes for several years before diabetes is diagnosed?

The UKPDS Risk Engine provides the best risk estimates available for people with newly diagnosed type 2 diabetes reflecting a mix of individuals who will have had their diabetes diagnosed with varying degrees of delay. As the risk of diabetic complications increase with duration of disease the UKPDS Risk Engine, like all risk calculators, will tend to underestimate the actual risk for any individual who has had unrecognised diabetes for more years than most. Conversely, screening programmes for diabetes may detect people at an earlier stage than usual when their risk may be overestimated.

How are cholesterol and HDL cholesterol values converted from conventional units (mg/dl) to SI units (mmol/l), and vice versa?

We use the conversion factor one mmol/l = 0.02586 mg/dl. as listed in the Diabetes Care SI Conversion Table.

Why does changing the measurement units for cholesterol alter predicted risk for some patients?

Rounding of converted values can make a slight difference to the predicted risk.

Is the UKPDS risk engine validated in patients on statin therapy, or must untreated total cholesterol and HDL-C values be used?

Given that most statin trials have delivered CV risk reductions commensurate with that expected from the observational relationship between LDL cholesterol levels and CV outcomes it is to be expected that the UKPDS Risk Engine will give reliable CV risk estimates from statin-modified LDL values. Although we are not aware of any formal validation of the UKPDS Risk Engine in a statin-taking population, it successfully estimated CV outcomes for the PROactive trial (Lancet 2006:367;25).

How can I use the UKPDS Risk Engine to compare cardiovascular disease risk to a 20% threshold, as recommended by NICE?

Summing the coronary heart disease (CHD) and stroke risk estimates from the Risk Engine is not recommended: it gives only an approximate estimate of cardiovascular disease (CVD) risk, due to double-counting of people who have both CHD and stroke. Instead, treat the 20% threshold of CVD risk as approximately equivalent to a 15% threshold for CHD risk. [Williams B, Poulter NR, Brown, MJ, Davis M, McInnes GT, Potter JF, Sever PS, Thom S. British Hypertension Society guidelines for hypertension management 2004 (BHSIV): summary BMJ 2004;328,634-640]

Mac OS X 10.7, Lion support?

A new version of the UKPDS Risk Engine is being prepared that will be compatible with Mac OS X 10.7, Lion.

Office 2010 64bit


We are aware of an issue with Office 2010 64bit, however, there is no issue with Office 2010 32bit. Installl Office and select the option for 32bit version.. We are working on a new version of the Risk Engine that solves this issue but it could be a little time before that is available. The 32bit version of Office 2010 will work almost identically to the 64bit version and so shouldn't cause any problems.