A recent UK Health Technology Assessment pointed out the problems with screening for type 2 diabetes. However, research continues into the performance of tools to detect type 2. On the basis of this paper, we won’t be changing that advice any time soon.
This community screening study was undertaken in urban settings in the Philippines. A stratified random sample was taken of registered health service users, who were invited to participate. Trained researchers gave them the FINDRISC questionnaire, and diabetes status was measured using a blood glucose test.
The gold standard of diagnosis was from two separate fasting blood glucose tests. The thresholds were:
- Prediabetes: FBG 100-125 mg/dl (5.6-6.9 mmol/l) or 2h OGTT > 140-199 mg/dl (7.8-11 mmol/l)
- Diabetes: FBG ≥ 126 mg/dl (7.0 mmol/l) or 2h OGTT > 200 mg/dl (11.1 mmol/l)
Two modifications were made to the FINDRISC questionnaire to particularise it to the local population (e.g. by adjusting body size norms). These were assessed separately.
1,789 people took part. They all had FINDRISC assessments and a preliminary blood glucose test. 491 of them were selected for further glucose testing on the basis of this test showing elevated blood glucose.
The researchers compared the FINDRISC scores against the likelihood of having diabetes and calculated Area Uncer the Curve (AUC) for the FINDRISC instrument in diagnosing diabetes. They also measured the likelihood ratios (LR) of specific levels of cut-off for the FINDRISC score. These were:
|FINDRISC cutoff||LR for a positive result||LR for a negative result|
|7 or more||1.75||0.46|
|8 or more||2.05||0.49|
|9 or more||2.45||0.50|
These numbers are extremely unimpressive! FINDRISC would be a terrible screening test. If used as such it would probably result in lots of unnecessary testing, overdiagnosis and worry. You can read more about screening on the NHS Choices website.
- Extraordinarily, the authors claim that FINDRISC “performed well” in screening for undiagnosed diabetes. This is the polar opposite of what the data presented in the study says.
- FINDRISC may narrow down the likelihood a bit, but even when combined with a pre-screening blood test, it didn’t perform well in a screening context.
- The paper is not always clear whether its measures are given against the full population of 1,789, or the “pre-screened” subset of 491.
- There was a blind, independent comparison of the test against an acceptable gold standard of diagnosis.
- However, not all of the participants received the same sets of tests.
- No confidence intervals are provided for the likelihood ratios. We couldn’t calculate them ourselves because it was not clear from the paper which population it was referring to.
Ku GMV, Kegels G. The performance of the Finnish Diabetes Risk Score, a modified Finnish Diabetes Risk Score and a simplified Finnish Diabetes Risk Score in community-based cross-sectional screening of undiagnosed type 2 diabetes in the Philippines. Primary Care Diabetes. (2013). [Pub Med abstract]