Intended for healthcare professionals

CCBYNC Open access

Rapid response to:


Development and validation of QDiabetes-2018 risk prediction algorithm to estimate future risk of type 2 diabetes: cohort study

BMJ 2017; 359 doi: (Published 20 November 2017) Cite this as: BMJ 2017;359:j5019

Rapid Response:

QDiabetes-2018: Mixed opinions


The article on QDiabetes 2018 made for interesting reading. (1) Although the new models are definitely better than the previous one, there are several points to consider:

1. Model B best explained the variation in time to diagnosis of type 2 diabetes (R2 [women]: 63.3%; [men]: 58.4%). This model included fasting blood sugar in addition to the basic model. However, only 14% of the subjects had a recorded fasting blood sugar value.

2. Model C was the next best, explaining 60.3% and 55.5% of variation in time to diagnosis of type 2 diabetes among women and men respectively. However, HbA1c values were recorded for only 6.2% subjects.

3. The ethnicity was recorded for 72.5% of subjects, of which 87.2% were white, while 2.3% were Indian. The actual proportions in the study population amount to 63.2% and 1.7% respectively. These proportions are representative of the general population in the UK, but far removed from the population composition of the two most populous countries- China and India. The authors do not suggest that the models can be applied to other populations, however, they do state that the algorithms used in the previous iteration have been ‘independently and externally validated in international populations and compared with other diabetes risk prediction models and been shown to have best performance’. (1) This naturally raises the possibility that Qdiabetes 2018 will be applied in foreign populations with different ethnicities and risk profiles. Considering that ethnicity is an important risk factor in the development of diabetes, the lack of non-white subjects is a matter of concern for workers outside the UK. (2) Given these limitations, one must exercise caution in applying the new model(s) outside the UK.

4. Only 14.9% of subjects had a family history of diabetes. In populations where this proportion is higher- 16.9% in India, for instance- the model will likely perform differently.(3) The extent of variation will have to be carefully ascertained before large-scale usage in such countries.

5. Among women subjects, the Adjusted Hazard Ratios show a consistent decline from Model A through Model C for all ethnicities except Black Africans, for whom the hazard ratio is higher in Model B. However, although a similar general trend is observed in men subjects, the hazard ratio increases for several variables in Model B.(1) Thus, the inclusion of Fasting Blood Sugar values in the basic model (Model A) seems to vary the risk differently for men and women.

6. As pointed out by the authors themselves, both Model B and Model C require further external validation before they can be used by clinicians. (1)

In summary, the authors must be commended for developing new models that better predict the risk of developing Type 2 diabetes. However, the said models must be used with caution outside the UK in view of the points raised above.

1. Hippisley-Cox J, Coupland C. Development and validation of QDiabetes-2018 risk prediction algorithm to estimate future risk of type 2 diabetes: cohort study. BMJ [Internet]. 2017 [cited 2017 Nov 27];359. Available from:
2. Kaveeshwar SA, Cornwall J. The current state of diabetes mellitus in India. Australas Med J [Internet]. 2014 [cited 2017 Nov 27];7(1):45–8. Available from:
3. Ramachandran A, Snehalatha C, Kapur A, Vijay V, Mohan V, Das AK, et al. High prevalence of diabetes and impaired glucose tolerance in India: National Urban Diabetes Survey for the Diabetes Epidemiology Study Group in India (DESI). [cited 2017 Nov 27]; Available from:

Competing interests: No competing interests

27 November 2017
Liaquat R Johnson
Associate Professor
Department of Community Medicine, Azeezia Medical College and Research Centre, Kollam, Kerala, India