Introduction
Hyperglycemia caused by diabetes mellitus is associated with long-term diabetes-related complications, resulting in reduced life expectancy when compared to the general population without diabetes [
1]. In type 1 diabetes mellitus (T1DM), increased mortality is explained by diabetic ketoacidosis and hypoglycemia early in life and cardiovascular diseases later in life. Strict glucose control is associated with a lower risk of diabetes-related complications and cardiovascular mortality [
2,
3]. Besides sustained chronic hyperglycemia, another particular atherogenic action of postprandial glucose has emerged as another target to be pursued in the clinical practice aimed at reducing mean blood glucose and glycated hemoglobin (HbA1c) [
4]. Lower postprandial glucose levels may be associated with a lower risk of cardiovascular outcomes in diabetes [
5].
However, strict glucose control is associated with weight gain and a higher incidence of hypoglycemia [
2]. Hypoglycemia can lead to seizures, cognitive impairment, decreased quality of life, loss of work productivity, impaired functioning on the following day, and non-adherence to treatment [
6‐
10]. Hypoglycemia may also cause cardiac ischemia or arrhythmia mediated by the catecholamine secretion [
11], eventually leading to a higher risk of death [
12].
The development of insulin analogues through molecular structure modifications of human insulin is based on pharmacokinetic profiles that try to mimic the physiological secretion of insulin [
13]. Short-acting insulin analogues (aspart, glulisine, and lispro) are thought to be better than regular human insulin due to faster absorption and faster onset of action, mimicking better the physiological prandial insulin peak of people without diabetes [
14,
15] and leading to lower postprandial glucose levels [
16]. Potentially, this allows for better glucose control, reduces the number of hypoglycemic episodes, and helps improve the patient’s quality of life by allowing for less restrictive mealtimes. However, a number of meta-analyses on short-acting insulin analogues have found only modest benefits on glucose control and the frequency of hypoglycemic episodes compared to therapy with regular human insulin [
17‐
19]. Analyzing the expected benefits and higher costs of insulin analogues against the cost-effectiveness of human insulin is particularly important in low- to middle-income countries, where the lack of affordable insulin medication is still a major cause of death [
20].
It remains unclear whether short-acting insulin analogues are indeed superior to regular human insulin in reducing hypoglycemia and lowering postprandial glycemia. The aim of the present systematic review and meta-analysis was to evaluate the outcomes (all hypoglycemic episodes, nocturnal and severe hypoglycemia, and postprandial glycemia) associated with the use of short-acting insulin analogues in T1DM as compared to regular human insulin.
Methods
This systematic review was carried out based on the methodology described in the Cochrane Collaboration tool [
21].
Eligibility criteria
Studies were eligible for inclusion if they were randomized controlled trials (RCTs), included children and adults with a diagnosis of T1DM for at least 1 year, with or without chronic complications, and compared the use of subcutaneous short-acting insulin analogues (aspart, glulisine, and lispro) with regular human insulin for at least 4 weeks. The primary outcomes were hypoglycemia (all hypoglycemic episodes, nocturnal hypoglycemia, and severe hypoglycemia) and postprandial glucose (all meals and after breakfast, lunch, and dinner). Secondary outcomes included long-term glucose control assessed by HbA1c and changes in quality of life. Studies with pregnant women, patients with acute diabetic decompensation, or patients with type 2 diabetes, studies that used insulin pumps, experimental studies or retrospective studies, narrative reviews, letters, and congress abstracts were excluded.
We searched the following electronic databases for studies published until August 2017: MEDLINE (via PubMed), EMBASE (via Elsevier), CENTRAL (the Cochrane Central Register of Controlled Trials), LILACS (Literatura Latino-Americana e do Caribe em Ciências da Saúde, via BVS), and DARE (Database of Abstracts of Reviews of Effects). The references cited by all the relevant studies were hand searched. We performed an additional search for ongoing and/or unpublished studies in the US National Institute of Health Ongoing Trials Register (
http://www.clinicaltrials.gov) and the International Clinical Trials Registry Platform (ICTRP - WHO). No language restrictions were applied.
Search strategy
We searched for the terms ‘diabetes mellitus, type 1’, ‘aspart’, ‘glulisine’, ‘lispro’, and related terms to obtain as many results as possible. The complete search strategies used for each database are provided as Additional file
1: Table S1.
Study selection
Duplicates were manually identified and excluded. The articles were then analyzed in two steps: firstly, two reviewers independently screened the titles and abstracts yielded by the search strategy against the inclusion and exclusion criteria; secondly, the same reviewers independently screened the full text reports and selected the articles that met the inclusion criteria. Disagreements were resolved by consensus. If no agreement could be reached, a third reviewer was consulted for arbitration. Agreement between reviewers was assessed using Cohen’s kappa coefficient. The Rayyan software (Rayyan Platform) was used for this selection process.
Data extraction and quality assessment
Two reviewers independently extracted data from each study using an extract table template, which provided the following information: title of the study; demographic characteristics; study design; intervention details; and outcomes. A third reviewer further assessed all RCTs to check for completeness of data.
In case of missing data, the authors of the studies were contacted for additional information. If the missing data could not be retrieved, the study was not included. Retrieved missing data were presented in a narrative form.
To assess the internal quality of the studies, we ran each RCT through the Cochrane Collaboration tool for assessing the risk of bias [
21]. The following potential issues were assessed: random sequence generation; allocation concealment; blinding of participants and outcome assessors; blinding of outcomes; incomplete outcome data; selective reporting; and other sources of bias. For each domain, the risk of bias was rated as low, high, or unclear. The quality of evidence was assessed using the GRADEpro GDT software (GRADEpro 2014). The results were presented in a “Summary of Findings” table.
Data synthesis and analysis
Analyses were conducted using the RevMan 5.3 software. Relative risk was used as a summary measure of effect size for dichotomous outcomes, the mean difference was used for continuous outcomes, and the risk rate was used for outcomes related to the number of events. The meta-analysis was performed using a random-effects model based on the DerSimonian–Laird method, 95% confidence interval (95% CI). A
p-value < 0.1 was considered statistically significant. Heterogeneity between the studies was assessed using I
2 statistic, in which values above 50% were indicative of high heterogeneity [
21]. Heterogeneity as determined by the Chi square test was considered non-significant for I
2 values between 0 and 50%, moderate for values between 51 and 79%, and significant for values between 80 and 100%. Where possible, study data were pooled and summarized in meta-analysis charts (quantitative synthesis) using the RevMan 5.3 software; otherwise, the results of each study were presented individually (qualitative synthesis).
Subgroup and sensitivity analysis
Subgroup analyses were designed based on the effects by neutral protamine Hagedorn (NPH) human insulin vs. long-acting insulin analogues. We also carried out a sensitivity analysis for the primary outcome ‘all hypoglycemic episodes’ considering the risk of bias in the studies. A second meta-analysis was then performed for this outcome excluding studies with three or more domains classified as “high risk of bias”.
Discussion
In this systematic review and meta-analysis we originally report clinical evidence on therapeutical use of short-acting insulin analogues compared with regular insulin while focusing on the main benefits of these analogues, namely the reduction of hypoglycemia and postprandial glucose levels. The combined data of 22 RCTs showed that short-acting insulin analogues are associated with a decrease in total hypoglycemic episodes, nocturnal and severe hypoglycemia, and post-breakfast, post-lunch and post-dinner glucose levels.
Fullerton et al., in a systematic review that aimed to evaluate long-term safety of short-acting insulin analogues, also assessed the outcomes described here; however, since their research was focused on long-term studies, fewer RCTs were retrieved when compared to the present study [
17]. The results of these reviews cannot be compared due to the high probability of inconsistencies. Another recently-published review analyzed only trials comparing aspart with regular human insulin, but also gathered data from a smaller set of studies [
44]. Since the three short-acting insulin analogues are very pharmacologically similar regarding time of onset, peak activity, and duration of action [
16], analyzing data from trials conducted with only one short-acting insulin analogue yields a lower number of studies, resulting in less statistical power. We saw no clear advantage in using this approach for the current study. A third systematic review and meta-analysis described only the results of hypoglycemia, and again included a smaller number of studies. In addition, no data on postprandial glucose were reported [
45].
The association between short-acting insulin analogues and a reduction of 7% in total hypoglycemic episodes, 32% in severe hypoglycemia, and 45% in nocturnal hypoglycemia levels is an important finding, as these episodes are particularly associated with lower quality of life and treatment nonadherence [
46]. The pursuit of lower HbA1c levels is associated with a higher rate of hypoglycemia episodes [
47], which is a well-known barrier to strict glucose control. This may hinder the maintenance of euglycemia over a lifetime, which prevents patients from fully benefitting from glucose control [
48]. The benefits mentioned above are most likely determined by the specific pharmacokinetic properties of these analogues; having a very short-acting activity limits the risk of late falls in glucose levels [
16]. The lower frequency of nocturnal hypoglycemic episodes observed with short-acting insulin analogues may contribute to the lower frequency of severe hypoglycemia. It is already known that sleep per se is a risk factor for severe hypoglycemia, as symptoms of hypoglycemia might be blunted or absent during sleep [
8].
Short-acting insulin analogues are expected to provide more adequate insulin levels in response to increases in postprandial blood glucose, which would reflect in a better postprandial glucose control. This was observed in our meta-analysis for all postprandial glucose levels, as well as after each individual meal (breakfast, lunch, and dinner). Interestingly, even though short-acting insulin users had lower postprandial glucose levels, they also had lower frequency of hypoglycemic episodes, a double benefit brought on by the pharmacokinetics of these drugs (faster onset of action and shorter duration of action) [
16].
Postprandial glucose fluctuations contribute to approximately 50% of the total hyperglycemia episodes in patients on multiple doses of insulin [
49]. Therefore, short-acting insulin analogues were expected to be associated with lower HbA1c levels, which would be consistent with observed decreases in postprandial blood glucose levels. However, the decreases were clinically irrelevant, even though the short-acting insulin analogues were indeed associated with lower HbA1c levels. This could be explained by the multiple insulin regimens employed in the analyzed studies (since short-duration studies were also included), as well as by the reduction in hypoglycemic episodes. In some studies, a single dose of NPH was used as basal insulin, which is an unacceptable regimen considering the current practice aimed at strict glucose control for T1DM. It is well-known that a better metabolic control with short-acting insulin analogues can be obtained with the optimization of basal insulin [
2,
14].
Limited evidence analyzed in this systematic review suggests that, for patients with T1DM, the treatment with short-acting insulin analogues is more convenient than with regular human insulin. The higher satisfaction levels and greater flexibility attributed to short-acting insulin analogues could be explained by the fact that they can be administered immediately before meals, as opposed to the anticipated 30 to 45 min when administering regular human insulin. In a study involving 1184 patients with T1DM, adherence to the correct timing of regular human insulin was 7% for patients who took it more than 30 min before meals, 60% for those who took it 15-30 min before meals, and 33% for those who took it 15 min before meals. Regarding the administration of insulin lispro, 98% of the patients followed the orientation (0 to 15 min before meals) [
32]. The possibility of administration of short-acting insulin analogues immediately after meals is another important benefit, as it may not always be possible to predict how much food (carbohydrates) the patient will have eaten at the end of the meal.
Due to the scarcity of studies assessing the impact of short-acting insulin analogues on the quality of life of patients with T1DM and the methodologies used, previously-published systematic reviews either did not analyze this outcome or did not reach a conclusion [
50]. According to Fullerton et al. [
17], with a more adequate methodology (the DTSQ) [
51], three studies reported no improvement in treatment satisfaction, while four studies indicated an improvement in this outcome with short-acting insulin analogues when compared to regular human insulin.
As opposed to other meta-analyses [
17], this review provided information regarding the use of insulin analogues in children. However, no association was found between the use of short-acting insulin analogues or regular human insulin and the number of hypoglycemic episodes, postprandial glucose reductions, and HbA1c, probably because of the low number of studies included.
The main methodological strengths of this review are as follows: the most adequate outcomes considering the pharmacokinetics of short-acting insulin analogues; the most comprehensive and systematic literature search among systematic reviews on this subject, with no language restriction; and the specific and reproducible eligibility criteria, study selection, and data extraction.
However, some limitations should be pointed out. The first is that most studies included in our systematic review may not represent current T1DM treatment practice. Most trials excluded patients with hypoglycemia unawareness or with a high risk of hypoglycemia, which in fact makes up the largest population group that could benefit from insulin analogues in the current clinical practice. Additionally, the low quality of most studies identified in this systematic review may limit the interpretation of the presented data. The differences in the definition of total and nocturnal hypoglycemic episodes, as well as the methods for recording hypoglycemic episodes based on the presence of symptoms or on the obligatory verification of blood glucose independently of symptoms, are real limitations frequently observed in clinical trials. Another limitation is the absence of masking, which could also result in a high risk of bias. However, it is unlikely that future studies will adequately mask the participants, as this would require a significant increase in the number of insulin applications. The analyses with NPH as basal insulin included 7, 18 and 14 studies (nocturnal hypoglycemia, total hypoglycemia and severe hypoglycemia, respectively), and those with long-acting insulin analogues included 1, 3 and 1 studies (nocturnal hypoglycemia, total hypoglycemia and severe hypoglycemia, respectively), and these analyses presented high heterogeneity, precluding their consideration as a definitive evidence of the possible superiority of NPH as compared to long-acting insulin analogues. This information should, thus, be interpreted with caution. A direct comparison between NPH insulin and long-acting analogues is beyond the scope of this review. Another important point is that, over the years, there has been a significant evolution in insulin therapy, which can be observed in the clinical heterogeneity between studies in the past 20 years.
In summary, short-acting insulin analogues were associated with fewer nocturnal and severe hypoglycemic events and better glucose control (slightly lower HbA1c and lower postprandial blood glucose levels) when compared with regular human insulin in subjects with type 1 diabetes.
Authors’ contributions
Conception and design: KFSM, LRB, BP; data search: KFSM, GJMP, ALM, RR, analysis and interpretation of data: KFSM, LRB, BP, BDS; drafting of the manuscript: KFSM, LRB, BDS; revising it critically for important intellectual content: LEPC, WJM, LAT, HCP; final approval of the manuscript submitted: KFSM, LRB, GJMP, BDS. All authors read and approved the final manuscript.