Articles
Short-term intensive insulin therapy in type 2 diabetes mellitus: a systematic review and meta-analysis Caroline Kaercher Kramer, Bernard Zinman, Ravi Retnakaran
Summary Lancet Diabetes Endocrinol 2013; 1:28–34 Published Online January 30, 2013 http://dx.doi.org/10.1016/ S2213-8587(13)70006-8 See Comment page 3 Leadership Sinai Centre for Diabetes (C K Kramer MD, Prof B Zinman MD, R Retnakaran MD) and Samuel Lunenfeld Research Institute (Prof B Zinman, R Retnakaran), Mount Sinai Hospital, Toronto, ON, Canada ; and Division of Endocrinology, University of Toronto, Toronto, ON, Canada (C K Kramer MD, Prof B Zinman MD, R Retnakaran MD) Correspondence to: Dr Ravi Retnakaran, Leadership Sinai Centre for Diabetes, Mount Sinai Hospital, 60 Murray Street, Suite L5-025, Mailbox-21, Toronto, ON, Canada M5T 3L9 [email protected]
Background Studies have shown that, when implemented early in the course of type 2 diabetes mellitus, treatment with intensive insulin therapy for 2–3 weeks can induce a glycaemic remission, wherein patients are able to maintain normoglycaemia without any anti-diabetic medication. We thus did a systematic review and meta-analysis of interventional studies to assess the effect of short-term intensive insulin therapy on the pathophysiological defects underlying type 2 diabetes mellitus (pancreatic β-cell dysfunction and insulin resistance) and identify clinical predictors of remission. Methods We identified studies published between 1950 and Nov 19, 2012, which assessed the effect of intensive insulin therapy on β-cell function or insulin resistance, or both, or assessed long-term drug-free glycaemic remission in adults aged 18 years or older with newly diagnosed type 2 diabetes mellitus. We calculated pooled estimates by random-effects model. This study is registered with International Prospective Register of Systematic Reviews, number CRD42012002829. Findings We identified 1645 studies of which seven fulfilled inclusion criteria (n=839 participants). Five studies were non-randomised. A pooled analysis of the seven studies showed a post-intensive insulin therapy increase in Homeostasis Model Assessment of β-cell function as compared with baseline (1·13, 95% CI 1·02 to 1·25) and a decrease in Homeostasis Model Assessment of Insulin Resistance (–0·57, –0·84 to –0·29). In the four studies that assessed glycaemic remission (n=559 participants), the proportion of participants in drug-free remission was about 66·2% (292 of 441 patients) after 3 months of follow-up, about 58·9% (222 of 377 patients) after 6 months, about 46·3% (229 of 495 patients) after 12 months, and about 42·1% (53 of 126 patients) after 24 months. Patients who achieved remission had higher body-mass index than those who did not achieve remission (1·06 kg/m², 95% CI 0·55 to 1·58) and lower fasting plasma glucose (–0·59 mmol/L, 95% CI –1·11 to –0·07) at baseline. Interpretation Short-term intensive insulin therapy can improve the underlying pathophysiology in early type 2 diabetes mellitus, and thus might provide a treatment strategy for modifying the natural history of diabetes. Funding None.
Introduction Type 2 diabetes mellitus is a complex metabolic disorder, the pathophysiology of which is driven by two main defects: target-cell resistance to the activity of insulin (insulin resistance) and insufficient secretion of insulin by the pancreatic β cells to compensate for this peripheral tissue resistance (β-cell dysfunction).1 Its natural history is characterised by the progressive deterioration of β-cell function over time, a pathological process that occurs irrespective of lifestyle and existing pharmacological interventions.2,3 As a result, the typical clinical course of this disease consists of the sequential addition of antidiabetic drugs over time, followed ultimately by insulin therapy when functional β-cell capacity deteriorates to the point at which glycaemic control can no longer be achieved without exogenous insulin supplementation.4 At this point in the progressive course of type 2 diabetes mellitus, insulin therapy is generally continued indefinitely thereafter. In recent years, temporary administration of shortterm intensive insulin therapy early in the course of type 2 diabetes mellitus has been a strategy of interest.5 28
Indeed, 2–3 weeks of intensive insulin therapy can induce a so-called glycaemic remission wherein patients are subsequently able to maintain normal glucose levels without any antidiabetic medication. Furthermore, studies6 have shown that this drug-free glycaemic remission can last up to 2 years in many patients, suggesting that short-term intensive insulin therapy could provide a strategy for modifying the natural history of the disease.6 Despite these promising results, the evidence base for early use of short-term intensive insulin therapy in type 2 diabetes mellitus remains scarce at this time, and data from individual studies might not be sufficient to show modification of the underlying pathophysiological changes or support the implementation of this therapy in clinical care. Thus, we did a systematic review and meta-analysis of interventional studies to determine (1) whether short-term intensive insulin therapy leads to improvement in β-cell function and insulin resistance in patients with newly diagnosed type 2 diabetes mellitus, and (2) whether baseline characteristics exist that can differentiate patients who will achieve glycaemic www.thelancet.com/diabetes-endocrinology Vol 1 September 2013
Articles
remission following short-term intensive insulin therapy from those who will not.
Methods Search strategy and selection criteria We undertook a systematic review of the published scientific literature. We selected relevant studies published between 1950 and Nov 19, 2012, by searching Embase, PubMed, Cochrane Library databases, ClinicalTrials.gov, and abstracts from the 2011 and 2012 meetings of the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). We used the following combined text and Medical Subject Heading (MESH) terms: “insulin”, “type 2 diabetes mellitus”, and “intensive”. The complete search used for PubMed was: (“Insulin”[Mesh] OR insulin[Text Word]) AND (“diabetes mellitus, type 2”[MeSH Terms] OR type 2 diabetes[Text Word] AND ((intensive [Text Word])). We considered for review all potentially eligible studies, regardless of primary outcome or language. We also did a manual search using references of key articles published in English. We considered studies for inclusion if they: (1) were done in adults aged 18 years or older with newly diagnosed type 2 diabetes mellitus; and (2) studied the effect of shortterm intensive insulin therapy on β-cell function (assessed by Homeostasis Model Assessment of β-cell function [HOMA-B]) or insulin resistance (assessed by Homeostasis Model Assessment of Insulin Resistance [HOMA-IR]), or both, reporting means and standard deviations at baseline and after short-term intensive insulin therapy; or (3) assessed patients for glycaemic remission after any duration of follow-up. Most of the included studies had single-arm designs; in the two studies7,8 that had more than one interventional group, we analysed only the arm that received short-term intensive insulin therapy. Exclusion criteria were: (1) studies that included patients who were not newly diagnosed with type 2 diabetes mellitus; (2) studies that did not provide HOMA-B or HOMA-IR nor assess glycaemic remission; and (3) further publications from included studies. Two independent investigators (CKK and RR) reviewed study titles and abstracts, and studies that satisfied inclusion criteria were retrieved for full-text assessment. There was an agreement value (κ) of 94% in the studies selected by these two investigators for detailed analysis; disagreements were resolved by a third investigator (BZ). We extracted data on first author’s name; year of publication; number, age, sex, body-mass index (BMI) of participants; glycated haemoglobin A 1c (HbA1c); duration of follow-up; and change in HOMA-B and HOMA-IR (log-transformed data; mean [SD]). When studies assessed patients by remission status, we extracted the baseline characteristics age, BMI, fasting plasma glucose, 2-h post-challenge glucose, HbA1c, HOMA-B, and HOMA-IR, in addition to post-intensive insulin therapy fasting plasma glucose and post-challenge glucose. www.thelancet.com/diabetes-endocrinology Vol 1 September 2013
We studied the risk of bias according to the Preferred Reported Items for Systematic Reviews and Metaanalysis (PRISMA) recommendations adapted for assessment of both randomised and non-randomised interventional studies. Study quality assessment addressed selection bias, description of losses or exclusions, and assessment of efficacy. This study is reported in accordance with PRISMA9 and is registered at International Prospective Register of Systematic Reviews (PROSPERO), number CRD42012002829.
Statistical analysis For all included studies, we analysed HOMA-B and HOMA-IR as continuous variables (log-transformed data) and reported differences between arithmetic means before and immediately after stopping intensive insulin therapy. For studies assessing long-term drug-free glycaemic remission, we compared baseline age, BMI, HbA1c, fasting plasma glucose, post-challenge glucose, HOMA-IR, and HOMA-B, and post-intervention fasting plasma glucose and post-challenge glucose according to remission status (remission vs non-remission groups) and reported as differences between arithmetic means. We calculated pooled estimates of the mean differences in outcomes before and after intensive insulin therapy by random-effects model (DerSimonian–Laird method) to adequately account for the additional uncertainty associated with study–study variability in the effect of the intervention. We used the Cochran Q test to assess heterogeneity between studies, with threshold of a p value lower than 0·1 for significance. We also did I² testing to assess the magnitude of heterogeneity between studies, with values higher than 50% deemed indicative of high heterogeneity.10 1645 studies identified from initial search of Medline, Embase, ADA meetings (2011 and 2012), EASD meetings (2011 and 2012), ClinicalTrials.gov
1621 studies excluded on the basis of title and abstract 7 additional studies identified through manual search
31 potentially relevant publications retrieved for detailed assessment
24 studies excluded because 10 lacked data for HOMA-B, HOMA-IR and glycaemic remission status 9 did not fulfil inclusion criteria 5 duplicates of included study
7 studies included
Figure 1: Flow diagram of search in the scientific literature to identify interventional studies on short-term intensive insulin therapy in type 2 diabetes mellitus
29
Articles
Baseline HbA1c (%)
Total follow-up after IIT (months)
Evaluation of glycaemic remission
25·1 (3·6)
10·0% (1·9)
24
Yes
25·4 (3·4)
11·9% (2·0)
0
No
Baseline Baseline proportion mean of men (%) BMI (kg/m²)
Year of Sample Design publication size
IIT regimen
Baseline IIT duration mean age (years) (days)
Li Y et al13
2004
126
Interventional; one arm
CSII
14
48·6 (11·6) 61·9%
Chen H et al14
2007
138
Interventional; one arm
CSII
14
45·8 (7·0)
Zhao Q et al 15
2007
120
Interventional; one arm
CSII
14
47·0 (12·0) 83·3%
24·0 (3·0) Not available
0
No
Chen H et al 7
2008
22
Randomised controlled trial for long-term IIT; one arm on short-term IIT
MDI
14
58·7 (16·0) 77·3%
27·7 (6·5)
0
No
Weng J et al 8
2008
251
Randomised controlled trial; one arm on IIT CSII and MDI 14
50·0 (10·5) 67·3%
24·7 (2·8)
9·7% (2·3)
12
Yes
Chen A et al 16
2012
118
Interventional; one arm
CSII
14–21
51·6 (10·2) 66·0%
25·0 (3·0)
11·0% (2·1)
12
Yes
Liu L et al 17
2012
64
Interventional; one arm
CSII
14
49·3 (9·5)
25·5 (3·5)
11·1% (1·8)
3
Yes
62·3%
68·7%
11·7% (1·9)
Data are mean (SD) unless otherwise stated. IIT=intensive insulin therapy. CSII=continuous subcutaneous insulin infusion. MDI=multiple daily injections. BMI=body-mass index. HbA1c=glycated haemoglobin.
Table 1: Characteristics of included studies
Stopped Dropout Outcome Selection Insulin rate (%) assessment bias therapy early accurate efficacy assessed Li Y et al13
No
Yes
No
10·3%
Yes
Chen H et al14
No
Yes
No
NR
Yes
Zhao Q et al15
No
Yes
No
NR
Yes
Chen H et al7
No
Yes
No
12·0%
Yes
Weng J et al8
No
Yes
No
5·3%
Yes
Chen A et al16
No
Yes
No
21·3%
Yes
Liu L et al17
No
Yes
No
NR
Yes
NR=not reported.
Table 2: Assessment of studies for risk of bias
We explored heterogeneity between studies using two strategies. First, we did stepwise meta-regression analyses. Using random-effects univariate meta-regression models, we assessed clinical and methodological variables that influenced the association between intensive insulin therapy and changes in HOMA-IR. Thereafter, on the basis of univariate meta-regression, we did sensitivity analyses to assess subgroups of studies most likely to yield valid estimates of the intervention. We assessed publication bias by funnel plot of effect size against standard error for every study. We assessed funnel plot asymmetry by Begg and Egger tests, and defined significant publication bias by p value lower than 0·1.11 We used trim-and-fill computation to estimate the effect of publication bias on interpretation of results.12 We used Stata 11.0 (StataCorp, College Station, TX, USA) for the analyses.
Role of the funding source The study was supported by intramural funds, with no commercial entity involved. The funding source had no role in study design, data collection, data analysis, data interpretation, or writing of the report. CKK and RR had access to the raw data. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. 30
Results We identified 1645 studies through electronic searches and seven through manual searches (figure 1). Of these, 1621 were excluded on the basis of title and abstract, leaving 31 studies for further assessment. Seven studies fulfilled our inclusion criteria, providing data for 839 participants.7,8,13–17 Table 1 summarises the included studies. In these studies, intensive insulin therapy was given by either continuous subcutaneous insulin infusion or multiple daily injections for 14–21 days. The trials were published from 2004 to 2012 and varied in sample size (table 1). Four studies8,13,16,17 followed up participants after the short-term intensive insulin treatment and hence provided insight on long-term remission. The 839 patients included in this systematic review had mean baseline HbA1c of 9·7–11·9% and mean baseline BMI of 24·0–27·7 kg/m². Table 2 shows the risk of bias in the included studies. All studies reported adequate efficacy of intensive insulin therapy (meaning that the study assessed whether participants had reached the target glycaemic control when on intensive insulin therapy), none was stopped early, and four7,8,13,16 reported dropout rates greater than zero. In pooled analysis of the seven studies (n=839),7,8,13–17 an increment in HOMA-B (log-transformed) was noted after intensive insulin therapy as compared with baseline (figure 2A). All but one study7 reported a significant increase in HOMA-B after intensive insulin therapy. Additionally, the heterogeneity was not significant in the individual estimates when the magnitude of the association was assessed (figure 2A). We noted no evidence of publication bias on Egger regression test and trim-and-fill computation (figure 3A). In the pooled analysis of the seven studies (n=839),7,8,13–17 a decrease in HOMA-IR (log-transformed) was noted after intensive insulin therapy as compared with baseline (figure 2B). All but one study7 reported a significant decrease in HOMA-IR after intensive insulin therapy. However, heterogeneity was significant between studies (figure 2B). We noted no evidence of www.thelancet.com/diabetes-endocrinology Vol 1 September 2013
Articles
publication bias on Egger regression test and trim-andfill computation (figure 3B). We did a meta-regression analysis in an exploratory attempt to identify the source of heterogeneity between trials. In univariate meta-regression models, sex was the only covariate associated with the heterogeneity between studies (p
Short-term intensive insulin therapy in type 2 diabetes mellitus: a systematic review and meta-analysis Caroline Kaercher Kramer, Bernard Zinman, Ravi Retnakaran
Summary Lancet Diabetes Endocrinol 2013; 1:28–34 Published Online January 30, 2013 http://dx.doi.org/10.1016/ S2213-8587(13)70006-8 See Comment page 3 Leadership Sinai Centre for Diabetes (C K Kramer MD, Prof B Zinman MD, R Retnakaran MD) and Samuel Lunenfeld Research Institute (Prof B Zinman, R Retnakaran), Mount Sinai Hospital, Toronto, ON, Canada ; and Division of Endocrinology, University of Toronto, Toronto, ON, Canada (C K Kramer MD, Prof B Zinman MD, R Retnakaran MD) Correspondence to: Dr Ravi Retnakaran, Leadership Sinai Centre for Diabetes, Mount Sinai Hospital, 60 Murray Street, Suite L5-025, Mailbox-21, Toronto, ON, Canada M5T 3L9 [email protected]
Background Studies have shown that, when implemented early in the course of type 2 diabetes mellitus, treatment with intensive insulin therapy for 2–3 weeks can induce a glycaemic remission, wherein patients are able to maintain normoglycaemia without any anti-diabetic medication. We thus did a systematic review and meta-analysis of interventional studies to assess the effect of short-term intensive insulin therapy on the pathophysiological defects underlying type 2 diabetes mellitus (pancreatic β-cell dysfunction and insulin resistance) and identify clinical predictors of remission. Methods We identified studies published between 1950 and Nov 19, 2012, which assessed the effect of intensive insulin therapy on β-cell function or insulin resistance, or both, or assessed long-term drug-free glycaemic remission in adults aged 18 years or older with newly diagnosed type 2 diabetes mellitus. We calculated pooled estimates by random-effects model. This study is registered with International Prospective Register of Systematic Reviews, number CRD42012002829. Findings We identified 1645 studies of which seven fulfilled inclusion criteria (n=839 participants). Five studies were non-randomised. A pooled analysis of the seven studies showed a post-intensive insulin therapy increase in Homeostasis Model Assessment of β-cell function as compared with baseline (1·13, 95% CI 1·02 to 1·25) and a decrease in Homeostasis Model Assessment of Insulin Resistance (–0·57, –0·84 to –0·29). In the four studies that assessed glycaemic remission (n=559 participants), the proportion of participants in drug-free remission was about 66·2% (292 of 441 patients) after 3 months of follow-up, about 58·9% (222 of 377 patients) after 6 months, about 46·3% (229 of 495 patients) after 12 months, and about 42·1% (53 of 126 patients) after 24 months. Patients who achieved remission had higher body-mass index than those who did not achieve remission (1·06 kg/m², 95% CI 0·55 to 1·58) and lower fasting plasma glucose (–0·59 mmol/L, 95% CI –1·11 to –0·07) at baseline. Interpretation Short-term intensive insulin therapy can improve the underlying pathophysiology in early type 2 diabetes mellitus, and thus might provide a treatment strategy for modifying the natural history of diabetes. Funding None.
Introduction Type 2 diabetes mellitus is a complex metabolic disorder, the pathophysiology of which is driven by two main defects: target-cell resistance to the activity of insulin (insulin resistance) and insufficient secretion of insulin by the pancreatic β cells to compensate for this peripheral tissue resistance (β-cell dysfunction).1 Its natural history is characterised by the progressive deterioration of β-cell function over time, a pathological process that occurs irrespective of lifestyle and existing pharmacological interventions.2,3 As a result, the typical clinical course of this disease consists of the sequential addition of antidiabetic drugs over time, followed ultimately by insulin therapy when functional β-cell capacity deteriorates to the point at which glycaemic control can no longer be achieved without exogenous insulin supplementation.4 At this point in the progressive course of type 2 diabetes mellitus, insulin therapy is generally continued indefinitely thereafter. In recent years, temporary administration of shortterm intensive insulin therapy early in the course of type 2 diabetes mellitus has been a strategy of interest.5 28
Indeed, 2–3 weeks of intensive insulin therapy can induce a so-called glycaemic remission wherein patients are subsequently able to maintain normal glucose levels without any antidiabetic medication. Furthermore, studies6 have shown that this drug-free glycaemic remission can last up to 2 years in many patients, suggesting that short-term intensive insulin therapy could provide a strategy for modifying the natural history of the disease.6 Despite these promising results, the evidence base for early use of short-term intensive insulin therapy in type 2 diabetes mellitus remains scarce at this time, and data from individual studies might not be sufficient to show modification of the underlying pathophysiological changes or support the implementation of this therapy in clinical care. Thus, we did a systematic review and meta-analysis of interventional studies to determine (1) whether short-term intensive insulin therapy leads to improvement in β-cell function and insulin resistance in patients with newly diagnosed type 2 diabetes mellitus, and (2) whether baseline characteristics exist that can differentiate patients who will achieve glycaemic www.thelancet.com/diabetes-endocrinology Vol 1 September 2013
Articles
remission following short-term intensive insulin therapy from those who will not.
Methods Search strategy and selection criteria We undertook a systematic review of the published scientific literature. We selected relevant studies published between 1950 and Nov 19, 2012, by searching Embase, PubMed, Cochrane Library databases, ClinicalTrials.gov, and abstracts from the 2011 and 2012 meetings of the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD). We used the following combined text and Medical Subject Heading (MESH) terms: “insulin”, “type 2 diabetes mellitus”, and “intensive”. The complete search used for PubMed was: (“Insulin”[Mesh] OR insulin[Text Word]) AND (“diabetes mellitus, type 2”[MeSH Terms] OR type 2 diabetes[Text Word] AND ((intensive [Text Word])). We considered for review all potentially eligible studies, regardless of primary outcome or language. We also did a manual search using references of key articles published in English. We considered studies for inclusion if they: (1) were done in adults aged 18 years or older with newly diagnosed type 2 diabetes mellitus; and (2) studied the effect of shortterm intensive insulin therapy on β-cell function (assessed by Homeostasis Model Assessment of β-cell function [HOMA-B]) or insulin resistance (assessed by Homeostasis Model Assessment of Insulin Resistance [HOMA-IR]), or both, reporting means and standard deviations at baseline and after short-term intensive insulin therapy; or (3) assessed patients for glycaemic remission after any duration of follow-up. Most of the included studies had single-arm designs; in the two studies7,8 that had more than one interventional group, we analysed only the arm that received short-term intensive insulin therapy. Exclusion criteria were: (1) studies that included patients who were not newly diagnosed with type 2 diabetes mellitus; (2) studies that did not provide HOMA-B or HOMA-IR nor assess glycaemic remission; and (3) further publications from included studies. Two independent investigators (CKK and RR) reviewed study titles and abstracts, and studies that satisfied inclusion criteria were retrieved for full-text assessment. There was an agreement value (κ) of 94% in the studies selected by these two investigators for detailed analysis; disagreements were resolved by a third investigator (BZ). We extracted data on first author’s name; year of publication; number, age, sex, body-mass index (BMI) of participants; glycated haemoglobin A 1c (HbA1c); duration of follow-up; and change in HOMA-B and HOMA-IR (log-transformed data; mean [SD]). When studies assessed patients by remission status, we extracted the baseline characteristics age, BMI, fasting plasma glucose, 2-h post-challenge glucose, HbA1c, HOMA-B, and HOMA-IR, in addition to post-intensive insulin therapy fasting plasma glucose and post-challenge glucose. www.thelancet.com/diabetes-endocrinology Vol 1 September 2013
We studied the risk of bias according to the Preferred Reported Items for Systematic Reviews and Metaanalysis (PRISMA) recommendations adapted for assessment of both randomised and non-randomised interventional studies. Study quality assessment addressed selection bias, description of losses or exclusions, and assessment of efficacy. This study is reported in accordance with PRISMA9 and is registered at International Prospective Register of Systematic Reviews (PROSPERO), number CRD42012002829.
Statistical analysis For all included studies, we analysed HOMA-B and HOMA-IR as continuous variables (log-transformed data) and reported differences between arithmetic means before and immediately after stopping intensive insulin therapy. For studies assessing long-term drug-free glycaemic remission, we compared baseline age, BMI, HbA1c, fasting plasma glucose, post-challenge glucose, HOMA-IR, and HOMA-B, and post-intervention fasting plasma glucose and post-challenge glucose according to remission status (remission vs non-remission groups) and reported as differences between arithmetic means. We calculated pooled estimates of the mean differences in outcomes before and after intensive insulin therapy by random-effects model (DerSimonian–Laird method) to adequately account for the additional uncertainty associated with study–study variability in the effect of the intervention. We used the Cochran Q test to assess heterogeneity between studies, with threshold of a p value lower than 0·1 for significance. We also did I² testing to assess the magnitude of heterogeneity between studies, with values higher than 50% deemed indicative of high heterogeneity.10 1645 studies identified from initial search of Medline, Embase, ADA meetings (2011 and 2012), EASD meetings (2011 and 2012), ClinicalTrials.gov
1621 studies excluded on the basis of title and abstract 7 additional studies identified through manual search
31 potentially relevant publications retrieved for detailed assessment
24 studies excluded because 10 lacked data for HOMA-B, HOMA-IR and glycaemic remission status 9 did not fulfil inclusion criteria 5 duplicates of included study
7 studies included
Figure 1: Flow diagram of search in the scientific literature to identify interventional studies on short-term intensive insulin therapy in type 2 diabetes mellitus
29
Articles
Baseline HbA1c (%)
Total follow-up after IIT (months)
Evaluation of glycaemic remission
25·1 (3·6)
10·0% (1·9)
24
Yes
25·4 (3·4)
11·9% (2·0)
0
No
Baseline Baseline proportion mean of men (%) BMI (kg/m²)
Year of Sample Design publication size
IIT regimen
Baseline IIT duration mean age (years) (days)
Li Y et al13
2004
126
Interventional; one arm
CSII
14
48·6 (11·6) 61·9%
Chen H et al14
2007
138
Interventional; one arm
CSII
14
45·8 (7·0)
Zhao Q et al 15
2007
120
Interventional; one arm
CSII
14
47·0 (12·0) 83·3%
24·0 (3·0) Not available
0
No
Chen H et al 7
2008
22
Randomised controlled trial for long-term IIT; one arm on short-term IIT
MDI
14
58·7 (16·0) 77·3%
27·7 (6·5)
0
No
Weng J et al 8
2008
251
Randomised controlled trial; one arm on IIT CSII and MDI 14
50·0 (10·5) 67·3%
24·7 (2·8)
9·7% (2·3)
12
Yes
Chen A et al 16
2012
118
Interventional; one arm
CSII
14–21
51·6 (10·2) 66·0%
25·0 (3·0)
11·0% (2·1)
12
Yes
Liu L et al 17
2012
64
Interventional; one arm
CSII
14
49·3 (9·5)
25·5 (3·5)
11·1% (1·8)
3
Yes
62·3%
68·7%
11·7% (1·9)
Data are mean (SD) unless otherwise stated. IIT=intensive insulin therapy. CSII=continuous subcutaneous insulin infusion. MDI=multiple daily injections. BMI=body-mass index. HbA1c=glycated haemoglobin.
Table 1: Characteristics of included studies
Stopped Dropout Outcome Selection Insulin rate (%) assessment bias therapy early accurate efficacy assessed Li Y et al13
No
Yes
No
10·3%
Yes
Chen H et al14
No
Yes
No
NR
Yes
Zhao Q et al15
No
Yes
No
NR
Yes
Chen H et al7
No
Yes
No
12·0%
Yes
Weng J et al8
No
Yes
No
5·3%
Yes
Chen A et al16
No
Yes
No
21·3%
Yes
Liu L et al17
No
Yes
No
NR
Yes
NR=not reported.
Table 2: Assessment of studies for risk of bias
We explored heterogeneity between studies using two strategies. First, we did stepwise meta-regression analyses. Using random-effects univariate meta-regression models, we assessed clinical and methodological variables that influenced the association between intensive insulin therapy and changes in HOMA-IR. Thereafter, on the basis of univariate meta-regression, we did sensitivity analyses to assess subgroups of studies most likely to yield valid estimates of the intervention. We assessed publication bias by funnel plot of effect size against standard error for every study. We assessed funnel plot asymmetry by Begg and Egger tests, and defined significant publication bias by p value lower than 0·1.11 We used trim-and-fill computation to estimate the effect of publication bias on interpretation of results.12 We used Stata 11.0 (StataCorp, College Station, TX, USA) for the analyses.
Role of the funding source The study was supported by intramural funds, with no commercial entity involved. The funding source had no role in study design, data collection, data analysis, data interpretation, or writing of the report. CKK and RR had access to the raw data. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. 30
Results We identified 1645 studies through electronic searches and seven through manual searches (figure 1). Of these, 1621 were excluded on the basis of title and abstract, leaving 31 studies for further assessment. Seven studies fulfilled our inclusion criteria, providing data for 839 participants.7,8,13–17 Table 1 summarises the included studies. In these studies, intensive insulin therapy was given by either continuous subcutaneous insulin infusion or multiple daily injections for 14–21 days. The trials were published from 2004 to 2012 and varied in sample size (table 1). Four studies8,13,16,17 followed up participants after the short-term intensive insulin treatment and hence provided insight on long-term remission. The 839 patients included in this systematic review had mean baseline HbA1c of 9·7–11·9% and mean baseline BMI of 24·0–27·7 kg/m². Table 2 shows the risk of bias in the included studies. All studies reported adequate efficacy of intensive insulin therapy (meaning that the study assessed whether participants had reached the target glycaemic control when on intensive insulin therapy), none was stopped early, and four7,8,13,16 reported dropout rates greater than zero. In pooled analysis of the seven studies (n=839),7,8,13–17 an increment in HOMA-B (log-transformed) was noted after intensive insulin therapy as compared with baseline (figure 2A). All but one study7 reported a significant increase in HOMA-B after intensive insulin therapy. Additionally, the heterogeneity was not significant in the individual estimates when the magnitude of the association was assessed (figure 2A). We noted no evidence of publication bias on Egger regression test and trim-and-fill computation (figure 3A). In the pooled analysis of the seven studies (n=839),7,8,13–17 a decrease in HOMA-IR (log-transformed) was noted after intensive insulin therapy as compared with baseline (figure 2B). All but one study7 reported a significant decrease in HOMA-IR after intensive insulin therapy. However, heterogeneity was significant between studies (figure 2B). We noted no evidence of www.thelancet.com/diabetes-endocrinology Vol 1 September 2013
Articles
publication bias on Egger regression test and trim-andfill computation (figure 3B). We did a meta-regression analysis in an exploratory attempt to identify the source of heterogeneity between trials. In univariate meta-regression models, sex was the only covariate associated with the heterogeneity between studies (p
