Systematic review and meta-analysis of randomized clinical trials comparing efficacy and safety outcomes of insulin glargine with NPH insulin, premixed insulin preparations or with insulin detemir in type 2 diabetes mellitus

Type 2 diabetes mellitus (T2DM) is a progressive disease, which requires insulin treatment when other management is no longer effective. Appropriate insulin therapy should be chosen individually to patient’s needs in order to achieve treatment goals and maintain its safety [1, 2]. In clinical practice, the flexibility of insulin titration is limited by the associated risk of hypoglycemic events, particularly when intensive insulin treatment is required [3, 4]. A growing body of evidence revealed that hypoglycemia is a predictor of poor outcome in people with T2DM, particularly it increases the risk of premature death [5‐7]. Therefore, most of the clinical practice recommendations highlight that the optimal glycemic control in T2DM patients should be achieved with minimized risk of hypoglycemia [2, 8].

In general, it is recommended that at the initiation of insulin treatment in T2DM, once daily basal insulin is added to oral antidiabetic drugs (OADs) [1, 9‐11]. Neutral protamine Hagedorn (NPH) has been frequently chosen as the first-line insulin; however, its use is associated with the risk of both hyper- and hypoglycemic events [12, 13]. Long-acting insulin analogs (LAAs) have been developed by modification of insulin chain in order to improve pharmacokinetic properties and decrease the risk of hypoglycemia. The first developed and most commonly prescribed LAA product is glargine (IGlar) [14]. Following the injection, IGlar forms a depot in the subcutaneous tissue, from which it is slowly absorbed. This provides a relatively uniform concentration over approximately 24 h after administration, which allows mimicking basal endogenous insulin secretion [12].

Complex pathophysiology of T2DM, its progressive nature, heterogeneous clinical picture and concomitant diseases require a variety of therapeutic options, including plural insulin regimens in order to maintain appropriate glycemic control and treatment safety. IGlar is frequently used as once daily regimen in addition to OAD. Interestingly, it has been demonstrated that early basal insulin initiation with IGlar improves FPG control and beta-cell function when compared to prolonged continuation of solely oral therapy [15]. When necessary, prandial insulin preparations can be used to intensify treatment. Therefore, IGlar is suitable for a spectrum of treatment intensities and can be used in T2DM patients at different age and various stages requiring insulin.

So far, several attempts have been undertaken to provide reliable summary of data comparing IGlar with other therapeutic options in T2DM patients [16‐34]. However, available systematic reviews have assessed only selected insulin preparations and do not provide a broad clinical picture or comprehensive answer, whether IGlar use is associated with additional clinical benefits to a wide spectrum of T2DM patients. Here, for the first time, we performed a systematic review combining all data from randomized clinical trials (RCTs) in T2DM to compare efficacy and safety outcomes of IGlar with several other insulin regimens in order to make synthetic and reliable conclusions.

A total number of 3,186 records without duplicates were identified in the systematic search of medical databases. After the screening of titles and abstracts, 430 studies were considered potentially relevant and were subjected for further assessment based on full-text publications. A total number of 363 studies were subsequently excluded from the analysis due to reasons presented in PRISMA diagram (Fig. 1). Finally, 29 RCTs were included in qualitative and 28 in quantitative analysis. Overall, studies included in quantitative analysis enrolled 12,669 T2DM patients, who were followed for 12–52 weeks (Tables 1, 2).

Nine studies compared IGlar versus NPH [39‐47], while in two others, patients from the comparatory group received neutral protaminated insulin lispro (NPL) [48, 49]. However, all these studies were analyzed together as NPL demonstrates similar pharmacokinetic and pharmacodynamic properties to NPH [50]. In nine studies, protamine insulin was administered once daily, while in the remaining two trials, patients were allowed to receive protaminated insulin according to either qd or bid schemes [47, 49]. Nine studies assessed basal + OAD regimen; in four of which patients received one oral drug either metformin [40, 43] or a sulphonylurea derivative [41, 42], while in the remaining five RCTs participants could be treated with more than one OAD [39, 44, 45, 48, 49]. In two studies assessing basal + bolus regimen, either human or lispro insulin was used as prandial insulin and OAD therapy was allowed but not obligatory [46, 47]. Additionally, one long-term RCT comparing IGlar + OAD with NPH + OAD was identified. However, this study was not included in quantitative analysis due to heterogeneous treatment and much longer follow-up (260 weeks) when compared to the remaining RCTs [51].

Thirteen studies assessed IGlar in comparison with MIX, of which three compared IGlar + OAD with MIX monotherapy [52‐54], while in eight of them patients received insulins in combination with OAD in both groups [55‐59]. Remaining 5 RCTs assessed IGlar + bolus ± OAD regimen in comparison with MIX ± OAD [60‐64]. Premixed insulin analogs were used as comparators in most RCTs, except for participants of trials comparing IGlar + OAD with mix monotherapy [52‐54] and 59 % of subjects from another study assessing basal + bolus regimen in comparison with MIX [62], who were treated with human premixed insulins.

Finally, four studies assessed IGlar in comparison with IDet. Both LAA were administered in basal + OAD regiment in two RCTs [65, 66] and according to basal + bolus scheme in the remaining two [67, 68].

In most studies, the follow-up was not longer than 6 months, while three studies reported the outcomes after around 1 year of treatment (Table 2) [45, 60, 62]. The credibility of included RCTs, assessed according to the Jadad scale, oscillated between 1 and 3 points on the 5-point scale and was mainly downgraded due to the lack of double blinding (Table 1).

Pharmacotherapy of T2DM usually starts from monotherapy with metformin, and it is further intensified by adding OADs of other classes or injectable GLP-1 agonists; nevertheless, many patients will eventually require insulin, usually beginning from one injection of basal insulin preparation [1, 2, 69]. Many insulin products have been developed so far to cover the full spectrum of T2DM patients’ needs. IGlar has favorable pharmacokinetic and pharmacodynamics properties that allow providing constant insulin activity over 24 h with only a single injection, and is the most widely prescribed LAA [12]. Since numerous systematic reviews have been published in order to combine the outcomes of many RCTs comparing IGlar with various comparators, the additional value of our meta-analysis will be hereby discussed [16‐34, 70]. Firstly, majority o earlier meta-analyses and secondary studies did not provide a full picture of clinical efficacy and safety of IGlar, since they were focused exclusively on certain aspects of insulin therapy (for example, intensive insulin treatment) [16, 17, 31]. Secondly, the other studies did not take into account the complexity of insulin treatment and pooled together trials recruiting patients with heterogenous clinical characteristics, subjected to different treatment models or accumulated results for different insulin preparations, e.g., insulin glargine with insulin detemir and premixed insulin analogs with human biphasic insulins [16, 19, 23, 25, 26, 29, 31, 33, 70]. Finally, several reviews did not attempt to accumulate included studies and presented only qualitative assessment, which significantly limited accuracy and precision of the conclusions [21, 24, 26, 28, 30, 32, 34].

In order to address the heterogeneity of T2DM and describe the efficacy and safety of investigated interventions in the context of different treatment regimens, we performed two separate analyses for each insulin preparation used for comparison. The first one involved less intensified insulin therapy (e.g., IGlar ± OAD), while the second one concerned intensive insulin therapy (IGlar + bolus). Therefore, our analysis is comprehensive and allows us to draw very accurate and reliable conclusions. Additionally, we identified several recent primary studies, which allowed us to receive more up-to-date results and perform more thorough investigation of heterogeneity than it was previously reported.

The primary efficacy outcome was defined as the percentage of patients achieving target HbA1c level without hypoglycemia and encompassed either overall severe, nocturnal or symptomatic hypoglycemic events, as the definitions of outcomes varied between respective studies. Although the heterogeneous reporting of composite outcomes limited between-trial comparability, still the combination of glycemic control and hypoglycemia serves as the most representative measure of treatment effectiveness. Indeed, it is well known that a decrease in the HbA1c level is usually achieved at the cost of higher risk of hypoglycemic episodes. These episodes are associated with increased mortality and decreased quality of life as shown in different cohorts T2DM patients [5‐7, 71]. Thus, reaching glycemic target without hypoglycemic events seems to bring particular benefits. The results of our comparative analysis indicate that IGlar is an option with favorable efficacy and acceptable safety profile. IGlar + OAD increased the proportion of patients reaching target HbA1c level without hypoglycemic events as compared to NPH + OAD. Although available evidence did not allow us to compare IGlar + bolus with NPH + bolus, the analysis of individual endpoints demonstrated comparable reduction of HbA1c in each arm, but with concomitantly lower rate of symptomatic and nocturnal hypoglycemia in IGlar group. These results are consistent with most of the available systematic reviews comparing IGlar and NPH, which reported similar effect on HbA1c level with concomitantly lower risk of hypoglycemia for IGlar, particularly in terms of nocturnal events [18, 20, 21, 28].

Available systematic reviews showed that MIX, as compared to LAA, was associated with better glycemic control but also with a higher incidence of hypoglycemia [23, 33]. However, neither of these analyses took an effort to interpret the results in the context of relatively high degree of between-study heterogeneity with respect to treatment in the control groups. Indeed, in available RCTs patients enrolled to control groups received either monotherapy with human premixed insulins or a combination of biphasic analog insulins with OAD therapy. In the current analysis, we have shown that the superior effect of MIX + OAD over IGlar with respect to mean HbA1c decrease was associated with greater weight gain, and higher risk of symptomatic hypoglycemia. We observed an advantage of IGlar over human premixed insulins (without OAD) with respect to both the reduction of HbA1c and the incidence of hypoglycemia, which was not shown in previous reviews.

Finally, we also demonstrated a favorable effect of IGlar over IDet in basal + bolus regimen, as IGlar use was associated with a higher percentage of patients reaching target HbA1c without the experience of hypoglycemia. When considering basal + OAD therapy, both insulins exerted similar effect on the primary endpoint. These observations differ from those presented by other authors, which can be explained by several limitations of previous reports that have been resolved in the current analysis. In 2008, Fakhoury et al. [17] reported an advantage of IDet over IGlar in relation to the risk of hypoglycemia, with comparable metabolic control as measured by HbA1c. However, these results were derived from an indirect comparison including evidence published before March 2007; therefore, have much less credibility than current meta-analyses performed on most up-to-date head-to-head comparisons. Another systematic review by Swinnen et al. [70] found no differences between IDet and IGlar, both in terms of glycemic control and the risk of hypoglycemia. However, the credibility of those results was limited by a high degree of statistical heterogeneity, which probably reflected between-study differences in treatment regimens. Within this analysis, we separately assessed both basal + OAD and basal + bolus regimens, which was reasonable as both schemes are usually used for different disease severities. These separate analyses allowed to remove statistical heterogeneity and to obtain more precise results indicating the advantage of IGlar + bolus over IDet + bolus with respect to glycemic control.

The current review has some potential limitations. First, our meta-analysis could be criticized for choosing a fixed HbA1c target below 7.0 % as a component of the composite primary endpoint, while the individualization is an important element of the contemporary diabetes management. Nevertheless, most major clinical guidelines maintained HbA1c level of <7.0 % as a general treatment goal, which can be advantageous for majority of patients with diabetes [72‐74]. The treatment target may be relieved mainly in patients with long-lasting diabetes, short life expectancy, existing comorbidities, etc. Since most studies included in our meta-analysis excluded individuals for whom less stringent glycemic control target could be applied, such as elderly individuals with severe chronic complications, the definition of the primary endpoint adopted in this paper is justified. Additionally, due to the short duration of studies included in the current meta-analysis and low incidence of malignancies, we were not able to analyze between-treatment differences in the risk of malignancies. Such analysis would have been of potential clinical importance taking into account earlier studies searching for potential association between T2DM and pharmacological therapies used for its treatment and the risk of cancer [75‐81].

As depicted on a study flow diagram, 363 papers did not meet inclusion criteria for the analysis due to various reasons. This may raise concerns whether the data selection process was correct. Such a high number of publications excluded from the meta-analysis are to some extent a consequence of a highly sensitive strategy used for database searching, which allowed us to retrieve all important studies. This also, however, resulted in a relatively large number of less relevant data including reviews, letters, conference proceedings and others, which had to be removed in further steps of selection process.

Finally, the quality of included studies, which were mainly designed as open-labeled comparisons, should be discussed. Of note, proper maintenance of glycemic control requires continuous insulin titration, which could not be performed when patients or physicians are unaware of the assigned treatment since various insulins require different dosing. Blinding to treatment allocation would lead to suboptimal glucose levels control with excessive rate of hypoglycemia, and therefore, open-labeled design is justified in studies assessing insulin therapy.

Finally, yet another limitation of this review is a short follow-up in the majority of included RCTs. Most studies had a duration up to 6 months, which may not provide a fully reliable picture of relative efficacy of insulin treatment in long-term perspective. The only longitudinal RCT identified within our systematic review was designed to compare IGlar versus NPH, both added to OAD therapy [51]. Nevertheless, due to the long follow-up, the protocol allowed for modification of both OAD therapy and insulin treatment, so that prandial insulins could be introduced or withdrawn during the study. Indeed, during the mean follow-up of 260 weeks, most of the patients in both treatment groups received human prandial insulins. The study demonstrated a difference in HbA1c reduction in favor of twice daily NPH insulin, which was most likely due to higher insulin dose and higher percentage of patients who were co-administered with prandial insulin in NPH group. Indeed, a post hoc analysis for patients treated solely with basal insulins and OAD demonstrated nearly equivalent reduction of HbA1c in both groups [51]. Additionally, IGlar was associated with fewer patients experiencing severe hypoglycemia and with no apparent difference in mean body gain [51].

In conclusion, for the majority of examined efficacy and safety outcomes, IGlar use in T2DM patients was superior or at least non-inferior to the alternative insulin treatment options (Table 3).