A propensity score matched comparison of different insulin regimens 1 year after beginning insulin in people with type 2 diabetes


  • N. Freemantle,

    Corresponding author
    1. Department of Primary Care and Population Health, University College London, London, UK
    • Correspondence to: Nick Freemantle, PhD, Professor of Clinical Epidemiology & Biostatistics, Department of Primary Care and Population Health, Upper Third Floor, UCL Medical School (Royal Free Campus), Rowland Hill Street, London NW3 2PF, UK.

      E-mail: nicholas.freemantle@ucl.ac.uk

    Search for more papers by this author
  • B. Balkau,

    1. Epidemiology of Diabetes, Obesity and Chronic Kidney Disease over the Lifecourse and Determinants of Early Nutrition, INSERM, Centre for research in Epidemiology and Population Health, Villejuif, France
    2. UMRS, University Paris Sud, Villejuif, France
    Search for more papers by this author
  • P. D. Home

    1. Institute of Cellular Medicine—Diabetes, Newcastle University, Newcastle upon Tyne, UK
    Search for more papers by this author



To compare using propensity score analysis the outcome of beginning insulin therapy with basal, premix, mealtime + basal or mealtime insulin on the basis of data from 3031 people from the observational CREDIT (Cardiovascular Risk Evaluation in People with Type 2 Diabetes on Insulin Therapy) study. This approach overcomes likely confounding in baseline and unknown characteristics common to observational studies.


Efficacy and safety outcomes were collected at baseline and at 1 year in previously insulin-naïve people. Propensity score matched groups using all available baseline data were defined to compare outcomes by pairs of insulin regimens.


From 2659 people with available data, propensity score matches were achieved for 686 people starting premix or basal insulin, 542 starting basal + mealtime or premix insulin and 400 starting basal or basal + mealtime. HbA1c reduction did not differ between the three pairs of insulin regimens. However, the relative risk of overall and nocturnal hypoglycaemia was lower (p = 0.010 to p < 0.001) with basal or basal + mealtime compared with premix insulin, and for nocturnal (p = 0.021) but not overall hypoglycaemia for basal compared to basal + mealtime insulin. Body weight increase was less for basal versus premix insulin [–1.3 (95% CI –2.1, –0.6) kg, p < 0.001] or versus basal + mealtime insulin [–1.4 (–2.5, –0.3) kg, p = 0.016], but did not differ between basal + mealtime and premix. Smaller groups matching mealtime insulin had some residual mismatching of HbA1c.


Comparing insulin regimens between individuals matched by propensity scores indicated differences in hypoglycaemia and body weight change, despite similar HbA1c reductions. Our findings are consistent with those from randomized controlled trials.