Tight glycaemic control is a mandatory component of diabetes care given proven beneficial effects on the risk of vascular complications 12. Current UK, European and US guidelines 3456789 recommend a target for glycated haemoglobin (HbA_1c) between 6.5% and 7.5%. Insulin represents the cornerstone of care for achieving this target in patients with type 1 diabetes, and is also indicated in type 2 diabetes patients with suboptimal glycaemic control despite increasingly aggressive therapy with oral antidiabetic drugs (OADs) in addition to lifestyle changes.

In clinical practice, concerns about the initiation of insulin therapy contribute to less than optimal glycaemic control. As well, titration of insulin so as to achieve accepted targets for HbA_1c while minimising the risk of hypoglycaemia and weight gain, may be problematic with conventional insulin regimens. The pharmacokinetic profile of short-acting basal insulins such as Neutral Protamine Hagedorn (NPH) are less than satisfactory for achieving a sustained duration of action. In contrast, more prolonged and predictable absorption is observed after administration of insulin glargine (glargine, Lantus^®), the first long-acting basal analogue product. Following injection, glargine precipitates into the subcutaneous tissue and then dissociates slowly to form monomers that enter the circulation in a controlled fashion for 24 hours 10. In clinical trials in patients with type 1 or type 2 diabetes, glargine was shown to be as effective as NPH insulin in lowering HbA_1c levels 1112 and was also associated with significantly fewer episodes of symptomatic and nocturnal hypoglycaemia 13. In addition, most patients were adequately controlled by a once daily injection, due to the more prolonged and stable mechanism of action 1415, leading to improved patient satisfaction 16. Given the need to balance glycaemic control with the risk of hypoglycaemia, glargine may therefore offer benefits over NPH 17.

It is recognised that findings from randomised clinical trials are not necessarily representative of everyday clinical practice, particularly in respect of patient populations and adherence to medication 1819. Several observational studies 2021222324 have attempted to address this issue. While these studies suggested that the use of glargine improved glycaemia control compared with NPH, deficiencies in patient selection and control for 'regression to the mean' effects and potential confounding variables make these analyses difficult to interpret. Therefore, we conducted a retrospective observational study based on a large patient cohort derived from primary care practices in the UK to assess the impact on HbA_1c, weight and insulin use of switching from NPH to glargine.

The results of this retrospective observational study show that patients with type 1 or type 2 diabetes, who switch from a predominantly basal-bolus NPH-based regimen to a glargine-based basal-bolus regimen derive significant improvement in glycaemic control (adjusted mean decrease in HbA_1c by 0.38% and 0.31%, respectively) within 12 months, without any significant increase in weight, use of prandial boluses, total insulin dose or OAD intensity. These findings add to a weight of evidence supportive of improved glycaemic control when patients switch from other insulins, either NPH 2930 or premixed insulins 31, to glargine. The relevance of our findings are strengthened by the nature and size of our sample (701 patients in an everyday clinical practice setting), the quality of baseline data (available for at least 12 months prior to switching from NPH to glargine), the completeness of follow-up data after switching (at least 3 monthly follow-up over 12 months), and the use of linear interpolation and multiple regression techniques in our analysis. Sensitivity analyses showed that the magnitude of improvement in HbA_1c was greatest in patients with the poorest glycaemic control (mean adjusted decreases of 0.57 and 1.07% in type 1, and 0.47 and 0.97% in type 2 diabetes patients with baseline HbA_1c levels ≥ 8% and ≥ 10%, respectively) (Table 2), and comparable with that reported in trials of patients newly commencing insulin therapy 32. A reduction of 1% in HbA_1c is associated with a 14% reduction in myocardial infarction, a 14% reduction in all-cause mortality, a 37% reduction in microvascular complications, and a 21% reduction in overall diabetic complications 33. Thus, the additional reduction in HbA_1c achieved by switching to a glargine-based regimen can be considered clinically meaningful, both in the overall population and the subgroups which had a ≥ 1% greater reduction, given the potential to translate to clinical outcomes benefits in the longer-term.

While we have highlighted the strengths of our study, we also acknowledge a number of potential limitations. First, we recognise that retrospective observational studies do not provide the same robust level of evidence as randomised controlled trials. In our study, the decision to switch from NPH to glargine treatment was not standardised but instead based on the judgement of the individual treating clinician. The data were collected from a large number of primary practice units which is likely to introduce a considerable level of heterogeneity to the main findings. As well, the level of data collection did not permit investigation of potential influences on glycaemic control including differences in racial background 23, body mass index 2934, frequency of NPH and glargine use (once vs. twice daily) 35, concurrent use of OADs with glargine 36, time of administration of glargine (morning vs. bedtime) 3738 and patient compliance 3940. Second, it was not possible to reliably assess data concerning hypoglycaemic episodes in patients who switch from a NPH-based regimen to a glargine-based regimen. A lack of consistency in recording these data in the THIN database meant that there is strong likelihood that we underestimated the real incidence of hypoglycaemia and only captured the most severe episodes. The incidence of hypoglycaemia may also reflect the self-reporting methods used, as patients and physicians were not requested to provide specific details of each episode. The limitations of the database also did not allow assessment of the nature (e.g. nocturnal) or severity of hypoglycaemic episodes before and after the switch. A higher number of hypoglycaemic episodes were noted after the switch from NPH to glargine. However, the overall rate of hypoglycaemia was low so the results should interpreted with caution. Should this trend be real, the significant improvement in glycaemic control observed with glargine in both diabetic cohorts may have made patients more susceptible to episodes of hypoglycaemia.

Despite these shortcomings, observational studies such as the current report are generally regarded as an ideal approach to assess the actual health outcomes of patients in routine care, more so than randomised controlled trials. This is because the level of care patients receive in clinical trials is often of a higher standard and not representative of that provided in daily clinical practice 1819. As well, clinical trials usually have limited scope for titration of other glucose-lowering drugs, a relatively short observational period, and potential for population bias which may prevent extrapolation of their findings to everyday practice 4142. In the current study, there were no limitations regarding patient inclusion criteria or clinical criteria for switching insulin treatment, which is likely to widen the implications of our findings. Furthermore, while high rates of missing data and patient drop-outs are common criticisms of observational studies, the current study provided interpolated data for 90% of patients during the months preceding and immediately following the switch and for 67% 12 months later. Interpolated HbA_1c data was used in 60% of patients. However, the main analysis performed in our study was based on the adjusted change in HbA_1c which only used actual and complete data values. It is reassuring to note that our findings are supported by other analyses from different populations and geographic locations and using different methods of data collection 29.

The implications of these results for decision makers are that improvement in control of glycaemia, as measured by HbA_1c, can be directly related to treatment efficacy and the costs of subsequent care. Using a discrete event simulation model, we recently reported that that for UK patients with sub-optimally controlled type 2 diabetes, taking into account effects on HbA_1c and reduction in hypoglycaemia, glargine can be considered as a cost-effective treatment option with a cost per quality-adjusted life year of less than £10,000 43. In addition, preliminary data from a similar analysis using the THIN database indicates that for patients with type 1 or type 2 diabetes who switch from NPH to glargine, treatment with glargine is a cost-effective strategy according to conventional thresholds 44.

A noteworthy finding from our study is that despite dissemination of clinical guidelines for diabetes management, glycaemic control is still less than optimal in a large proportion of patients in primary practice in the UK. Approximately two-thirds of patients in the study failed to achieve an HbA_1c level ≤ 7%. These data highlight the need for additional treatment strategies. These may include higher doses and more aggressive titration (in the current study, evidence that glargine had no significant effect on weight is suggestive of suboptimal treatment), greater use of OAD therapy in patients with type 2 diabetes (only 20% were prescribed OAD therapy at the time of the switch), as well as the use of educational programmes 45464748.

This observational study suggests that in diabetes patients treated with NPH and with evidence of suboptimal efficacy and/or poor tolerability, switching to insulin glargine offers the opportunity for improved glycaemic control.

HbA_1c: Glycated haemoglobin; OADs: oral antidiabetic drugs; NPH: Neutral Protamine Hagedorn; THIN: The Health Improvement Network.

All authors were consultants to Sanofi-Aventis, had full access to all the data in the study and took responsibility for the decision to submit for publication.

PS and JG performed the statistical analyses and drafted the manuscript. JP, APT, AL, and PM were involved in study design, coordination and data acquisition. All authors have read and approved the final manuscript.