Adding Pioglitazone to Insulin Containing Regimens in Type 2 Diabetes: Systematic Review and Meta-Analysis

Background Type 2 diabetes is treated in a stepwise manner, progressing from diet and physical activity to oral antidiabetic agents and insulin. The oral agent pioglitazone is licensed for use with insulin when metformin is contraindicated or not tolerated. This systematic review and meta-analysis investigates the extent to which adding pioglitazone to insulin-containing regimens produces benefits in terms of patient-relevant outcomes. Methodology/Principal Findings Medline, Embase, and the Cochrane Library were searched for randomised controlled trials comparing pioglitazone in combination with any insulin-containing regimen in comparison with the same insulin regimen alone in patients with type 2 diabetes. Outcomes investigated included HbA1c, hypoglycaemia, weight, and adverse events. Studies were selected, assessed and summarised according to standard systematic review methodology and in a meta-analysis. We included eight trials that examined the benefits of adding pioglitazone to an insulin regimen and studied a total of 3092 patients with type 2 diabetes. All studies included patients with previously inadequate glucose control. Trial duration was between 12 weeks and 34.5 months. The trials used pioglitazone doses of up to 45 mg/day. In our meta-analysis, the mean reduction in HbA1c was 0.58% (95% CI: −0.70, −0.46, p<0.00001). Hypoglycaemic episodes were slightly more frequent in the pioglitazone arms (relative risk 1.27; 95% CI: 0.99, 1.63, p = 0.06). Where reported, HDL-cholesterol tended to be increased with pioglitazone. Patients on pioglitazone tended to gain more weight than those who were not, with an average difference of almost 3 kg. Peripheral oedema was more frequent in the pioglitazone groups. None of the studies reported on fractures in women, and data on cardiovascular events were inconclusive, with most studies being too short or too small to assess these long-term outcomes. Conclusions/Significance When added to insulin regimens, pioglitazone confers a small advantage in terms of HbA1c in type 2 diabetes patients with previous inadequate glucose control, but at the cost of increased hypoglycaemia and weight gain. Other considerations include the risk of heart failure, fractures in women, a reduced insulin dose, and the net financial cost.


Introduction
Type 2 diabetes is usually seen in people who are overweight or obese, particularly if inactive. They usually have insulin resistance, and therefore require higher levels of insulin in order to keep blood glucose within the normal range. The pancreatic beta cell is initially able to compensate for insulin resistance by increasing production, thereby maintaining normal blood glucose levels. However, in most patients, pancreatic beta cell function progressively declines, leading to hyperglycaemia and clinical diabetes [1]. In the United Kingdom Prospective Diabetes Study (UKPDS), beta-cell function was found to be impaired at diagnosis, especially in patients who were not overweight [2].
The difficulty in maintaining metabolic control over time may be related to several behavioural factors (for example difficulties with healthy eating, exercise, medication regimens) but also reflects a progressive decline in beta-cell function [3,4].
Type 2 diabetes has traditionally been treated in a stepwise manner, starting with lifestyle modifications and encouragement of physical activity and when necessary, pharmacotherapy with oral agents (NICE guideline) [5]. If control remains inadequate, insulin may be used, with or without combination with one or more oral agents. There is no clear consensus on the definition of ''inadequate control'', but a consensus statement (2009) of a working group drawn from the American Diabetes Association and the European Association for the Study of Diabetes suggested that ''an HbA1c over 7% should serve as a call to action to initiate or change therapy'' [6].
Several classes of oral agents are available. These include the insulin secretagogues which stimulate the pancreas to release more insulin, by binding to a sulphonylurea receptor, the main group being the sulphonylureas; a second class are the insulin sensitizers, including the biguanide metformin and the thiazolidinediones rosiglitazone and pioglitazone; thirdly there are drugs that delay the absorption of carbohydrates from the gastrointestinal tract, such as acarbose; fourthly there are the DPP-IV inhibitors (also known as the gliptins), which extend the life of endogenous glucagon-like peptide. These include sitagliptin and vildagliptin (with more in development).

The glitazones
The thiazolidinediones -or glitazones for short -decrease insulin resistance in muscle and adipose tissue by activating the peroxisome proliferator-activated receptor-gamma (PPAR-gamma) which increases production of proteins involved in glucose uptake. They also decrease hepatic glucose production by improving hepatic insulin sensitivity.
According to the Prescribing Support Unit (PSU), in collaboration with the York and Humber Public Health Observatory (YHPHO) [7], the glitazones are the third most used diabetes drugs in England (about 2.4 million prescriptions a year), after metformin (about 10 million prescriptions a year), and the sulphonylureas (around 5 million prescriptions a year). In terms of cost per annum, the glitazones are by far the most costly, being recently introduced drugs with no generic forms.
In addition to being used alone or in combination with other oral agents, pioglitazone is also licensed (EMEA 2008) [8] for use in combination with insulin in type 2 diabetes patients with insufficient glycaemic control on insulin, and for whom metformin is inappropriate because of contraindications or intolerance. In this review, we concentrate on this indication.
A Cochrane review of pioglitazone therapy in general by Richter et al. (2006) [9] included 22 trials which randomised a total of 6200 people to pioglitazone treatment. Most studies were of short duration. Published studies of at least 24 weeks pioglitazone treatment in people with type 2 diabetes mellitus did not provide convincing evidence of benefit in patient oriented outcomes like mortality, morbidity, adverse effects, costs and health-related quality of life. Metabolic control measured by HbA1c did not demonstrate clinically relevant differences to other oral glucose lowering drugs. The occurrence of oedema was significantly raised.
The only exception to the short-term trials found in the Cochrane review was the PROactive study [10]. This placebo controlled randomised trial of 5238 patients set out to determine the effect of pioglitazone on macrovascular morbidity and mortality in patients with type 2 diabetes who had evidence of macrovascular disease. Patients continued their other diabetes medications, mainly metformin, sulphonylureas, insulin, or combinations thereof. The primary end-point was a composite of death and non-fatal cardiovascular outcomes. The average time of observation was 34.5 months. The pioglitazone group had a lower risk but this did not reach statistical significance (HR 0.90, 95% CI 0.80 to 1.02; p = 0.095) despite the large numbers of recruits and events (at least one end-point event in 514 of the pioglitazone group and 572 of the placebo group). A secondary endpoint measure of death, non-fatal MI and stroke did reach statistical significance: HR 0.84, 0.72-0.98; p = 0.027.
However, oedema and heart failure were commoner in the pioglitazone group, with 11% reported as having heart failure compared to 8% in the placebo group; the proportions needing hospital admission were 6% and 4%. The death rates from heart failure showed no difference. Heart failure was not defined centrally, but was ''as judged by the investigator''. Another outcome was ''oedema in the absence of heart failure''. Heart failure can be difficult to diagnose, and the absence of any difference in mortality from heart disease, might suggest that it could have been over-diagnosed. However, an independent group of cardiologists reviewed all the cases of serious heart failure and concluded that it did occur more frequently in the pioglitazone group (5.5% versus 4.2% for placebo) [11].
Another finding from PROactive was that progression to needing insulin was halved in the pioglitazone group. At the start of the study, about one-third of the patients were on insulin. Their mean age was 62, mean BMI 31, and duration of diabetes 8 years; 75% had a history of hypertension and mean HbA1c was around 7.8%. The protocol asked investigators to aim for an HbA1c of ,6.5%. By the end of follow-up, 11% of the pioglitazone group and 21% of the placebo group were on insulin treatment. The switch to insulin started early in the trial, presumably due to investigators trying to achieve the HbA1c target.
As concerns evidence on a combination of insulin and a glitazone, Strowig and Raskin (2005) [12] carried out a review of combination therapy with insulin and either metformin or a glitazone, or both. Details of methods are not given and the review was probably not systematic. The authors concluded that it was worthwhile continuing an insulin sensitiser in type 2 diabetes patients switched to insulin. Because metformin and glitazones have different balances of sites of preferential action (acting on glucose production and glucose disposal), they also made the case that triple therapy may be considered. Bailey (2005) [13] also supported combination therapy with metformin and a glitazone for reducing insulin resistance in type 2 diabetes.

Objectives
This review investigates the extent to which adding pioglitazone to insulin-containing regimens affects glycaemic control, hypoglycaemia, weight change, lipids, and adverse events.

Study characteristics
We considered randomised controlled trials of pioglitazone in combination with any insulin regimen in patients of any age and gender with type 2 diabetes. Minimum trial duration was 12 weeks. Pioglitazone in combination with any insulin regimen (long-acting, twice daily mixture, both with or without additional oral medication (generally metformin and/or sulphonylurea)) was compared to the same insulin regimen (with or without the same additional oral regimen) given on its own. As outcome measures, we considered HbA1c, frequency of hypoglycaemia (especially if severe), glycaemic excursions (including post-prandial hyperglycaemia), total daily dose of insulin, weight change, changes in cardiovascular risk factors, and other adverse events.
We used a group of outcomes because the decision to use pioglitazone was expected to depend on trade-offs amongst them, such as better control versus weight gain. We would have considered diabetic secondary complication rates (retinopathy, nephropathy, myocardial infarction, angina, heart failure, stroke, amputation, death), and health-related quality of life, but it rapidly became clear that data would not be available.

Ethics
As this was a systematic review of published literature, ethics approval was not required.

Searching
Medline, the Cochrane Library, and Embase were searched for studies published between 1996 and 2008. The following Medline search strategy (Ovid) was adapted for use with the other databases: 1. exp Thiazolidinediones/ 2. pioglitazone.tw. 3. 1 or 2 4. randomized controlled trial.pt. 5. meta-analysis.pt. 6. (random$ or meta-analysis or systematic review).tw. 7. 4 or 5 or 6 8. 3 and 7 In addition, reference lists of retrieved studies were checked. Searches were also done to identify emerging evidence, from conference abstracts and trial registers.

Selection
Studies were selected independently by two authors (PR and NW) based on the inclusion criteria listed above. Any discrepancies were resolved by discussion.

Validity assessment
Randomised controlled trials were assessed on the following criteria based on the NICE guidelines manual [14]: Method of randomisation, allocation concealment, blinding of participants and outcome assessors, intention-to-treat analysis, proportion of participants excluded/lost to follow-up, power calculation, comparability of groups at baseline. Subgroup analysis based on quality was carried out for five or six or more quality criteria met versus fewer than five or six quality criteria met (i.e. two analyses with different cut-offs were done).

Data abstraction
Data extraction was carried out by one researcher and checked by another. Any disagreements were resolved through discussion, involving a third person if necessary.

Data synthesis
The clinical effectiveness, relative to the key comparators, was assessed, in terms of difference in effect size.
Data were summarised in a meta-analysis and using tables and text. For dichotomous outcomes, risk ratios were calculated and a Mantel-Haenszel random effects model was used. For continuous outcomes, weighted mean differences were calculated and an inverse variance random effects model was used. Where not directly available, standard deviations required in the metaanalysis were converted from standard errors; if a measure of variability was not given and standard deviations were available for at least 50% of included studies, the mean of the standard deviations of the remaining studies was used. Heterogeneity was assessed using the chi-squared test.

Trial flow
Eleven papers were identified as potentially relevant randomised controlled trials. Of these, eight fulfilled the inclusion criteria and compared pioglitazone plus insulin with insulin [15][16][17][18][19][20][21][22]. The remaining trials were excluded because they did not examine the comparison of interest and one was the uncontrolled extension of a trial that seemed relevant but could not be identified (see Figure 1).

Study characteristics
Characteristics of the included trials are shown in Table 1.
Participants. The trials included between 20 and 1760 participants, with between 10 and 896 participants in each comparison group. The total number of patients assessed was 3092. All studies included participants with previously inadequate glucose control (with different definitions, not reported for Shah 2007). Inclusion criteria with respect to previous treatment varied substantially. Only five trials [15,17,18,20,22] required previous insulin treatment. Three trials [15,18,20] required previous insulin therapy with or without oral antidiabetic agents (where reported, previous insulin monotherapy ranged between 48 and 88%). The trial by Fernandez 2008 required previous insulin combination therapy [17], and the trials by Shah 2007 included only insulintreated obese patients [22]. Of the remaining trials, the trial by Berhanu 2007[16] required previous combination therapy with or without insulin, and in this trial between 90 and 93% of patients had been on sulphonylurea plus metformin therapy without insulin. The study by Raz 2005[19] required previous therapy with sulphonylurea (alone or as oral combination therapy) and over 80% of patients in that trial had been on sulphonylurea plus metformin previously. The study by Scheen 2006[21] included patients previously on diet alone, oral agents, or insulin plus an oral agent, and in that trial, over half the patients (53%) had been on sulphonylurea plus insulin, and the second largest group had been on sulphonylurea monotherapy (24%). Where reported, mean age of participants was between 46 to 59 years, the comparison groups included between 35 and 60% of women, mean BMI was between 29 and 37 kg/m 2 , and diabetes duration was between 6 and 14 years. The trial by Berhanu 2007[16] included between 50 and 59% of Hispanic participants, and the study by Fernandez 2008 included only Mexican-American participants [17].

Study quality
Details of the quality of included trials are shown in Table 2. For four [15,16,18,19] of the eight trials, randomisation was adequate, while for the remaining four trials the randomisation procedure was not reported or unclear. Three trials [15,16,18] had adequate allocation concealment, while the rest of the trials did not report on allocation concealment. All but one trial [19] were described as double-blind. Five trials used intention-to-treat analysis [16,[18][19][20][21]. Five trials reported on follow-up rates [15,16,[18][19][20] and in those trials, between 77 and 92% of participants completed the trial, without any significant differences between comparison groups. Six of the eight trials reported that they had carried out a power calculation [15][16][17][18]20,21]. Six trials were reported in full and two were only available as meeting abstracts. The two trials reported as abstracts [21,22] did not report relevant baseline characteristics, five trials reported that their comparison groups were similar at baseline [15,[17][18][19][20], while Berhanu 2007[16] stated that participants in the placebo group had a slightly higher BMI at baseline and longer diabetes duration, but it was unclear whether these differences were significant. All but one trial [22] reported on sources of funding and all funding included industry funding.

Data synthesis
Results of the individual trials are shown in Table 3. HbA1c. All studies reported HbA1c values and could be included in the meta-analysis ( Figure 2). Baseline HbA1c values were between 7.6 and 10% in the pioglitazone plus insulin groups and between 7.8 and 9.8% in the insulin without pioglitazone groups. End-of-study HbA1c values were significantly lower in the groups taking pioglitazone plus insulin than in the groups taking insulin without pioglitazone (weighted mean difference 20.58%, 95% CI: 20.70, 20.46, p,0.00001). There was no significant heterogeneity. In the study by Mattoo 2005, 18% of patients on pioglitazone plus insulin and 6.9% of patients on insulin without pioglitazone attained HbA1c values of below 7.0%. There was no significant difference between patients using two or fewer daily injections and patients using three or more daily injections. Similarly, there was no significant difference between patients who had previously been on oral antidiabetic agents and those who had  Hypoglycaemia. Six studies reported on hypoglycaemia outcomes and could be summarised in a meta-analysis ( Figure 3). There were marginally more patients with hypoglycaemic episodes in the pioglitazone plus insulin groups than with insulin without pioglitazone (relative risk 1.27, 95% CI: 0.99, 1.63, p = 0.06). The results showed significant heterogeneity (p = 0.001). The study by Raz 2005 which used biphasic insulin aspart 30 (BIAsp 30) rather than other insulin regimens contributed most to the heterogeneity. There is evidence to suggest that BIAsp 30 is associated with a reduced rate of nocturnal and major episodes of hypoglycaemia compared to other types of insulin [23]. After eliminating this study from the analysis, there remained moderate heterogeneity (I 2 = 57%, p = 0.05) and there was significantly more hypoglycaemia in the pioglitazone plus insulin groups (relative risk 1.40, 95% CI: 1.14, 1.73, p = 0.002). There were no significant differences in intervention and control groups for hypoglycaemia either for studies in which the insulin regimen was unchanged from before the study [15,[20][21][22] or for studies using titrated insulin regimens according to a predefined study protocol [16][17][18][19]. Details regarding hypoglycaemic episodes are shown in Table 4. Severe hypoglycaemic events were rarely seen in the studies.
Insulin dose. Six studies [16][17][18][19][20][21] reported insulin doses (as units per kg per day or as units per day). Only two studies reported standard deviations, so a meta-analysis could not be carried out reliably. Of the six studies, four found that the insulin plus pioglitazone groups used significantly less insulin than the insulin without pioglitazone groups (weighted mean difference 20.19 U/ kg/day or 212.03 U/day). The remaining two studies did not report any p-values (but doses were also lower in the pioglitazone groups). Insulin dose ranged between 42 and 64 U/day or 0.5 to 1 U/kg/day in the pioglitazone groups and between 55 and 70 U/day or 0.7 to 1.2 U/kg/day in the groups taking no pioglitazone.
Weight change. Six studies reported weight change [16][17][18][19][20]22]. However, only one of the studies reported a measure of variability, so a meta-analysis could not be carried out reliably. In most studies, patients in the insulin without pioglitazone groups gained less weight than patients in the insulin plus pioglitazone groups (mean difference 2.91 kg, range 3.85 to 23.50 kg), but no p-values were reported. Weight change ranged between +1.4 and +4.4 kg in the pioglitazone plus insulin groups and between 20.04 and +4.9 kg in the insulin only groups.
Of the four studies, only two [16,20] found significantly reduced triglyceride values in the pioglitazone groups (reductions of between 0.44 and 0.70 mmol/L in the pioglitazone groups compared to insulin only).
Four studies reported on total serum cholesterol [16,17,19,20]. None of the studies found any significant difference in total cholesterol between the pioglitazone plus insulin and the insulin without pioglitazone groups.
Four studies reported on HDL-cholesterol [16][17][18]20], all finding significantly increased values in the pioglitazone groups. Overall, HDL cholesterol was increased by between 0.10 and 0.18 mmol/L in the pioglitazone groups compared to insulin only.
Four studies reported on LDL-cholesterol [16][17][18]20], with none finding any significant difference between the pioglitazone plus insulin and the insulin without pioglitazone groups.     Adverse events. Where reported, there did not appear to be any significant difference in withdrawals due to adverse events between the pioglitazone plus insulin and the insulin without pioglitazone groups. Apart from weight gain and hypoglycaemia, the only adverse event reported as occurring more frequently with pioglitazone was (peripheral) oedema, which was generally classified as mild to moderate. However, p-values were generally not reported.
Only three trials reported on cardiovascular adverse events [16,19,20] (and most studies were probably underpowered for coming to reliable conclusions). In the trial by Berhanu 2006, insulin only patients had a higher incidence of cardiac events (10.7% versus 5.5%) but the majority were ECG abnormalities rather than patient oriented outcomes; there was one patient each with myocardial infarction and cardiac hypertrophy in the insulin only group, and one patient with coronary artery disease in the    pioglitazone group; no deaths were observed. Raz 2005 reported two cases of myocardial infarction in the insulin monotherapy group which were not considered to be treatment-related. Rosenstock 2002 observed a rate of cardiovascular adverse events of 7.9% with pioglitazone plus insulin and of 7.0% with insulin only (no significant difference); congestive heart failure was seen in two patients receiving 15 mg/day pioglitazone and in two patients receiving 30 mg/day pioglitazone. All cases were in patients with a history of cardiovascular disease and none were considered to be drug-related. Heterogeneity. We performed a range of subgroup analyses, these included: insulin therapy ''as usual'' versus insulin individually titrated for the study, pioglitazone increased/titrated up to 45 mg versus constant dose of 30 mg, abstract only versus full publication, baseline HbA1c ,8.5 or 9% versus $8.5 or 9%, comparison group n,100 versus n.100, study duration ,6 months versus $6 months, study quality 5 or 6 or more criteria fulfilled versus less than 5 or 6 criteria fulfilled. No significant differences in study results were seen for any of the subgroup analyses.

Summary
Eight randomised controlled trials were identified comparing combinations of insulin and pioglitazone with insulin without pioglitazone regimens (two published as abstracts only). Compared to the insulin regimens, the pioglitazone plus insulin regimens reduced HbA1c by a mean of 20.58% (95% CI: 20.70, 20.46, p,0.00001). However, hypoglycaemic events were marginally increased with the pioglitazone regimens (relative risk 1.27, 95% CI: 0.99, 1.63, p = 0.06). Where reported, studies found reduced insulin doses in the pioglitazone groups. HDL-cholesterol was increased, but none of the other lipid parameters reported (triglycerides, total cholesterol, LDL-cholesterol) showed any systematic differences between the comparison groups. The studies tended to show increased weight (mean difference 2.91 kg) and more peripheral oedema with pioglitazone. No conclusions could be made regarding cardiovascular events.
This review adds to the current body of evidence in two ways. First, by demonstrating some added value of pioglitazone when added to insulin-containing regimens, and secondly by making the point that starting patients on insulin does not mean that all further intensifications need to occur by adding more insulin.

Limitations
The main evidence gap relates to long-term safety. The studies included in this review were too short-term and too few in number to assess longer-term adverse effects. Recent debates on the TZDs have focused on cardiovascular events, following the review by Nissen and colleagues [24], which concluded that the risk of cardiovascular events was increased by rosiglitazone. A metaanalysis by Monami et al. (2008) [25] aimed to identify moderators of the effect of rosiglitazone on incidence of myocardial infarction and found that there was a significant correlation between the proportion of insulin-treated patients and rosiglitazone-associated risk of myocardial infarction (r = 0.42, p,0.05) (possibly related to the fact that both thiazolidinediones and insulin have an effect on fluid retention). Similar data are not available for pioglitazone.
A meta-analysis of the risk of cardiovascular events with pioglitazone was carried out by Lincoff et al. (2007) [26]. Based on 19 trials with 16,930 participants, they concluded that pioglitazone was associated with a reduced risk of death, myocardial infarction or stroke. Another meta-analysis by Mannucci and colleagues [27] included 84 published and 10 unpublished trials of pioglitazone compared to placebo or active comparators, but excluded the PROactive trial. They reported a reduction of all-cause mortality with pioglitazone (OR 0.30; 95% CI: 0.14 to 0.63: p,0.05), but no significant effect on non-fatal coronary events.
Several new studies have asked why rosiglitazone should increase cardiovascular events but pioglitazone does not. Most have concluded that the likely reason is that while the two glitazones have the same effects on glycaemic control, and the same side-effects of fluid retention and heart failure, they have different effects on blood lipids. Berneis and colleagues [28] carried out a very small cross-over trial in 9 patients, giving them all 12 weeks on pioglitazone and 12 weeks on rosiglitazone. Total cholesterol increased more on rosiglitazone than on pioglitazone (p = 0.04), and triglycerides increased on rosiglitazone but decreased on pioglitazone (p = 0.004).
Chappuis and colleagues [29] also studied patients on both glitazones, this time with 17 patients having 12 weeks on each. The effects of HbA1c were similar, but triglyceride and cholesterol levels were lower with pioglitazone.
Deeg and colleagues [30] carried out a much larger comparison with 369 participants randomised to pioglitazone and 366 to rosiglitazone. After 12 weeks, pioglitazone had reduced LDL whereas rosiglitazone had increased it.
The other concern about the safety of the glitazones has been the fracture risk. Kahn and colleagues [31] in the ''durability'' study (ADOPT) reported that 9.3% of women on rosiglitazone had fractures compared to 5.1% on metformin and 3.5% on glibenclamide. The increases were in fractures of upper limb and foot, rather than in the classical osteoporosis-associated neck of femur and vertebrae. The fracture rates in men did not differ between treatment groups.
A case-control study by Meier and colleagues [32] using British general practice data from GPRD also found that use of glitazones was associated with increased fracture rates. No such increase was seen with other oral diabetes drugs.
A recent (non-systematic) review by Schwartz (2008) [33] summarised clinical data on bone loss associated with use of glitazones. This reiterates data from recent trials reporting increased fracture risk in women (but not men) and reports evidence from short term clinical trials in women that glitazones (both pioglitazone and rosiglitazone) caused more rapid bone loss. It also reports data from in vitro and rodent models suggesting that activation of the PPAR-gamma receptor can play a role in bone loss.
A letter to physicians issued by Takeda Pharmaceuticals (posted on the US Food and Drug Administration website in March 2007) reported an analysis of its clinical trials database on pioglitazone [34]. They compared the incidence of fractures in over 8100 patients treated with pioglitazone compared to over 7400 patients treated with a comparator. The fracture incidence calculated was 1.9 fractures per 100 patient years in women treated with pioglitazone and 1.1 fractures per 100 patient years in women treated with a comparator. The observed excess risk of fractures for women in this dataset on pioglitazone is therefore 0.8 fractures per 100 patient years of use. There was no increased risk of fractures identified in men.
The more drugs a patient has to take, the poorer the adherence. Donnan and colleagues from Dundee [35] found that even those on only one glucose lowering agent have poor adherence, with adequate adherence in only one in three. Adherence is better with a single daily dose [35]. Those taking other medications had poorer compliance than those on just a hypoglycaemic agent.
A systematic review of medication adherence in patients with poorly controlled diabetes by Odegard and colleagues [36] summarises the barriers to taking medicines, and the interventions which may help. Some of the studies are more relevant to the North American situation where people have to pay for drugs, but much of it is relevant to the UK. The review concurs with the work of Donnan and colleagues (mentioned above), that common barriers to adherence include complexity of regimen and number of doses.
The implication for the treatment of type 2 diabetes may be that both the number of drugs and the number of tablets or injections per day should be kept as low as possible [37].

Conclusions
When added to insulin regimens, pioglitazone confers a small but clinically useful decrease in HbA1c of 0.58% in type 2 diabetes patients with previous inadequate glucose control. However it does so at the cost of increased hypoglycaemia and weight gain, with a risk of heart failure, and of fractures in women. The effect on adherence of adding another medication needs to be considered, as does the extra cost, though that is offset by the reduced insulin dose required.