Is low HbA1c associated with hypoglycaemia among people with type 2 diabetes treated with intensive glucose-lowering therapy? A retrospective cohort study in primary care.

Talk Code: 
Manil Malawana
John Robson, Sally Kerry, Rohini Mathur
Author institutions: 
Centre for Primary Care and Public Health, Queen Mary University of London, UK


There is concern that in some adults with type 2 diabetes, treatment with intensive glucose-lowering therapy may increase mortality, hospitalisation and morbidity. The aim of our study is to establish whether lower HbA1c levels among such patients prescribed insulin or sulphonylureas are associated with higher rates of clinically-recorded hypoglycaemia. Secondary objectives are to establish whether lower HbA1c is also associated with higher rates of mortality, falls and fractures in those treated with intensive glucose-lowering therapy.


A total of 20,292 patients with type 2 diabetes aged 18 years and over registered with one of 147 GP practices in Tower Hamlets, City & Hackney and Newham CCGs on the 1st January 2012 who were prescribed insulin or sulphonylureas between the 1st July and 31st December 2012 were included in the study. People with a history of cancer, those on the palliative care register and patients not prescribed insulin or sulphonylurea medications between 1st July 2012 and 31st December 2012 were excluded.

Using pseudonymised data extracted from the electronic general practice health record (EMIS Web), we undertook a retrospective cohort study between 1st January 2013 and 31st December 2015. The study population was divided into those prescribed insulin +/- oral antidiabetic medication (n=7,389) and those prescribed sulphonylureas +/- other oral antidiabetic medications without insulin (n=12,903) between 1st July 2012 and 31st December 2012. The primary outcome measure was the first clinical recording of hypoglycaemia or blood glucose reading less than 4 mmol/L. Secondary outcome measures were all-cause mortality, first recorded fall or fracture. Proportional cox regression modelling was used to examine the associations with HbA1c levels adjusting for known risk factors; age, gender, sex, ethnicity, renal function, comorbidity and body mass index. 


For combined cohorts, the unadjusted rate of hypoglycaemia among those with low HbA1c (<53 mmol/mol) was 17.2 per 1000 person-years compared to 14.1 per 1,000 person-years in those with HbA1c 53-74 mmol/mol. The adjusted hazard ratios of a hypoglycaemic episode in those with low HbA1c compared with HbA1c 53-74mmol/mol was 1.32 (95% C.I. 1.08 to 1.60). Adjusted HRs for individual cohorts were 1.38 (95% C.I. 1.06 to 1.80) in the sulphonylurea cohort and 1.19 (95% C.I. 0.88 to 1.59) in the insulin cohort. For the secondary outcomes, those with low HbA1c had adjusted HRs of 1.32 for all-cause mortality (95% CI 1.14 to 1.53) and 1.08 for fractures (95% CI 0.88 – 1.33) compared with HbA1c 53-74mmol/mol. When examining falls, the effect of low HbA1c differed between drug cohorts with adjusted HRs for a fall in those with HbA1c <53mmol/mol of 1.16 (95% CI 0.94 to 1.43) in the sulphonylurea cohort and 0.81 (95% CI 0.56 to 1.16) in those prescribed insulin, when compared with those with HbA1c 53-74mmol/mol.


Among patients with type 2 diabetes prescribed either insulins or sulphonylureas, glycated haemoglobin levels below 53 mmol/mol appear to be associated with a greater risk of hypoglycaemia, particularly in those prescribed sulphonylureas. Low and high HbA1c levels were associated with higher mortality rates after adjusting for other risk factors. These findings provide further evidence for less stringent HbA1c targets in patients on intensive glucose-lowering therapy to prevent excess morbidity and mortality.


Submitted by: 
Manil Malawana
Funding acknowledgement: 
This research received no specific funding. Manil Malawana receives a salary funded by the National Institute of Health Research through the Department of Health. John Robson receives a salary funded by Newham, City and Hackney and Tower Hamlets Clinical Commissioning Groups.