Insulin Pumps

The finding here that poor control on CSII is associated with a low internal LOC confirms the results of a previous study that investigated psychological factors and the outcomes of CSII (1). The lack of impact of fear of hypoglycemia on HbA1c has also been noted before (2) and is contrary to the frequently held view that poorly controlled patients tend to resist improvements in therapy because they are concerned that increased hypoglycemia will result.

Internal LOC is a psychological construct that expresses the extent to which we feel success or failure in an endeavor is due to our own efforts and not to luck, chance, or the actions of other people. People with diabetes with a high internal LOC may feel more personally responsible for the management of their diabetes and take a more active role in CSII procedures that give good control, such as adjusting basal insulin rates and meal boluses according to blood glucose tests, adhering to diet and exercise advice, attending the clinic as required, and so on. The good control that results affords a higher degree of treatment satisfaction. Alternatively, of course, those patients who have intrinsically difficult diabetes with highly variable insulin absorption and glycemic control and frequent hypoglycemia may try as hard as others (at least at first) but achieve little success with intensified insulin regimens and therefore conclude that their own efforts have little impact on their diabetes, thus acquiring a low internal LOC.

The patients in this study had a low frequency of both mild-to-moderate and severe hypoglycemia. It is possible that the association of LOC with metabolic outcomes may be different in those with a very high rate of severe hypoglycemia on CSII; perhaps there may be an even stronger link.

Both these studies concern the relative effectiveness of CSII and MDI, with or without CGM, and the place of these treatments in clinical practice. The results of the REPOSE Study may be misinterpreted as evidence that CSII is generally no more effective than MDI in type 1 diabetes, though the paper itself only concludes that pumps should not be used in poorly controlled patients with type 1 diabetes “until the effects of training in intensive self-management have been determined,” which is exactly according to many current guidelines, such as those issued by NICE. Actually, in a subgroup analysis in the REPOSE trial of those with an HbA1c at baseline of ≥8.5% (69 mmol/mol) (the NICE-recommended level for a trial of CSII), the mean HbA1c difference was significant at −0.4% (−0.8 to −0.1). Nevertheless, a recent BMJ opinion piece (3) commented, “‘You might like to send this paper to your local diabetes center the next time they prescribe a pump to one of your patients. You might also like to find out if they receive money from a manufacturer of these pumps.” This suggests that there is a belief among some healthcare professionals that insulin pumps are being widely used inappropriately or, worse, unnecessarily, and the REPOSE trial supports this notion. This is a concern.

There are several issues to be considered with the REPOSE trial. First, in the large group of patients with an HbA1c ≥7.5% (59 mmol/mol), the final mean HbA1c on CSII was about 8.75% (72 mmol/mol), for a baseline HbA1c of 9.5% (80 mmol/mol). This level of pump effectiveness is much worse than in most previous studies of pump usage—note, for example, that the mean HbA1c of all (well and poorly controlled) CSII-treated patients in the study of Indelicato et al. mentioned above was 7.75% (61 mmol/mol). Most clinic surveys of adults with type 1 diabetes treated by CSII show a final mean HbA1c <8.0% (64 mmol/mol). This raises the question of whether the performance of pump therapy and its procedures in the REPOSE Study may have been suboptimal or whether the patients included in the study were especially resistant to CSII, or both.

The relatively high final HbA1c on CSII in REPOSE was not associated with any increase in insulin dose over the 2-year follow up, suggesting that investigators and patients did not take the opportunity to increase basal and meal insulin in response to suboptimal control. Interestingly, there was a wide variation in the difference in HbA1c between CSII and MDI recorded at different centers, not necessarily related to previous CSII experience: one center, with the largest number of participants, achieved a mean HbA1c difference of −0.78% (95% confidence interval −1.46% to −0.09%), compared with the overall difference of −0.24% for all centers (see the full report on the trial [4]).

It is important also to note that patients normally considered for insulin pump therapy—those desiring pump therapy, those meeting NICE criteria, or those considered in need of a CSII trial by their healthcare professional—were excluded from the REPOSE trial. And there was a very small number of patients with a high frequency of severe hypoglycemia at recruitment, arguably the most common reason for trialing insulin pump therapy. It could be argued that both patients and investigators were less than fully motivated to get the best out of CSII.

Two recent trials, the DIAMOND (5) and GOLD (6) studies, have shown that CGM used with MDI can achieve lower HbA1c levels and less hypoglycemia than MDI used with SMBG. However, the DIAMOND Study had an extension phase, which is the focus of the Beck et al. report, and here it was tested whether switching patients in the MDI + CGM arm of the DIAMOND Study to CSII + CGM produces additionally improved glycemic control. One concern was the increase in biochemical (but not severe) hypoglycemia with CSII during the trial. This may possibly be because insulin pump training, and with it measures to reduce the risk of hypoglycemia, was not standardized between the centers. One may also argue that insulin pump therapy with an automatic low-glucose insulin suspend facility (see below) may have avoided the increase in hypoglycemia seen with the standard sensor-augmented pump therapy used by Beck et al.

Excessive meal-related glycemic increases caused by slow absorption of short-acting analog insulins like aspart, lispro, and glulisine are a major contributor to suboptimal and elevated HbA1c levels during CSII. Delayed insulin absorption also leads to the risk of late, postmeal hypoglycemia. Activating the meal bolus some 20 min before the meal start allows insulin absorption to more closely match meal glycemia, but it can be inconvenient. Ultra-fast-acting insulins like the faster aspart studied here have great potential to improve glycemic control and treatment satisfaction with both insulin injection therapy and CSII.

This was a relatively short-term trial over a few weeks, so the impact on HbA1c could not be measured—fructosamine was lower with faster aspart but did not reach significance. Equally, the finding that less time was spent in hypoglycemia with faster aspart hints of the potential to reduce moderate and perhaps severe hypoglycemia in the long term. No safety issues were uncovered and catheter occlusions were similar with the two insulins, but the stability of the new formulation needs to be studied in the longer term.

There are some intriguing suggestions from other short-term studies that the glucose lowering effect of faster aspart at meal times is greater with CSII (i.e., as a bolus on top of a basal infusion) than when given by injection alone (7). Extended trials of faster aspart with CSII will therefore be of great interest.

Although the main reason for introducing ultra-fast-acting insulins and the focus of this study was the improved management of postprandial glycemia, it is also possible that the (un)predictability of subcutaneous insulin absorption is less with faster aspart, and with it the variability of blood glucose levels. Although within- and between-day blood glucose variability is normally significantly improved with CSII, frustrating unpredictability can remain in many patients. The comparative variability of absorption and resultant glycemic control with faster aspart vs. aspart during CSII needs further study, and differences may be revealed especially in those patients with intrinsically unpredictable control.

There is increasing evidence for the effectiveness of CSII in poorly controlled type 2 diabetes from clinical experience and from several observational studies, but this individual patient data meta-analysis of RCTs is of note because it identifies those with the largest effect and therefore those who would likely be most cost-effectively treated. The patients in the trials studied here used insulin pumps intended for type 1 diabetes with features such as flexible basal rate control and bolus calculators that may not be needed for treatment of most type 2 diabetes patients. Newer patch pumps specifically designed for CSII in type 2 diabetes that are being introduced into clinical practice or under development should be simpler to use, more cost effective, and therefore more appropriate for this patient group.

An important caveat is that the types and intensities of the MDI regimens and educational programs used in the trials studied in this meta-analysis varied considerably, and some may not have been optimal by modern standards. The trials selected included four that were published 10 or more years ago. Several new agents for the treatment of type 2 diabetes have entered practice in recent years, such as new long-acting insulins (degludec), glucagon-like peptide (GLP)-1 agonists and sodium–glucose cotransporter 2 (SGLT2) inhibitors, and with more widespread use of structured diabetes education the number of patients with type 2 diabetes who remain poorly controlled and are candidates for a trial of CSII may diminish in the coming years. The relative costs of these new agents and simpler, type 2 diabetes–intended pumps will need to be considered.

These studies, two randomized and one observational, extend the evidence for the additional effectiveness of PLGS insulin pumps in reducing mild-to-moderate hypoglycemia in type 1 diabetes, over and above that afforded by CSII alone and by SAP without automatic low-glucose insulin suspend. There are several cautions to note in the interpretation of the studies. Both outpatient trials were of short duration, with a period on PLGS pump therapy of 2 or 6 weeks, thus making it impossible to measure changes in the serious acute complications of diabetes—severe hypoglycemia and diabetic ketoacidosis. Both outpatient studies were performed in children and young people who tend to have less hypoglycemia than adults because of a shorter duration of diabetes, though younger children may have more hypoglycemia than older children. The nature of both study populations is rather unclear; it is likely they were not groups who were hypoglycemia prone at baseline, in the sense of suffering previous frequent severe hypoglycemic episodes. Finally, the trials were not blinded, so patients may have altered behaviors such as eating and exercise according to treatment allocation, and this may have significantly influenced the frequency of hypoglycemia and the extent of hyperglycemia.

Similar and other limitations apply to the exercise study of Abrahams et al. The trial was short term (<4 h observation after exercise), so the common problem of exercise-induced late hypoglycemia was not assessed; it was performed in young people who were not hypoglycemia prone (those with a previous history of severe hypoglycemia were excluded); patients were well controlled before exercise started; it was not blinded to the patient, which may have influenced symptom reporting, for example; and results were achieved under supervised and controlled hospital conditions. It remains to be seen how effective PLGS pump therapy is in preventing exercise-induced severe hypoglycemia, as well as mild-to-moderate hypoglycemia, in the poorly controlled, possibly hyperinsulinemic and hypoglycemia-prone patient, studied in the long-term (say >6 months) under everyday life conditions.

The reason for and the clinical implications of the small but significant increase in time spent in moderate hyperglycemia with PLGS pumps in the study of Battelino et al. are unclear. It is reassuring that mean sensor glucose and ketone tests were similar after PLGS on and PLGS off, indicating that there is little evidence of insulin deficiency after the periods of insulin suspension. But again, this was a population of moderately well controlled and well managed patients, who were experienced with CSII, and results may be different in those who are very poorly and unpredictably controlled and ketosis prone, as well as those with frequent severe hypoglycemia. RCTs in these groups—those most likely to be first users of PLGS systems in clinical practice—are still awaited. A nonsignificant increase in mean sensor glucose with PLGS was also seen in the Biester et al. study, where nearly 50% of suspends were accompanied by manual resumption of insulin delivery and frequently also by consumption of extra carbohydrate. This may have contributed to an apparent rebound hyperglycemia after PLGS and the slightly higher overall sensor glucose on this treatment. As Biester et al. say, patients need education not to respond to a PLGS hypoglycemia alert with food intake and to let the pump algorithm “do the work.”

In the Biester et al. study, about 77% of hypoglycemic events were prevented by PLGS pump therapy (i.e., some 23% of sensor values were still <3.9 mmol/L [70 mg/dL]). In the Abrahams et al. study, some 32% on PLGS pump therapy still required treatment for hypoglycemia after exercise. Presumably, complete avoidance of hypoglycemia with PLGS is prevented, at least in part, by the occasions when there is substantial on-board insulin—here basal insulin suspend is insufficient to reduce ongoing insulin-induced blood glucose lowering from a previous large subcutaneous insulin depot.

These studies confirm the results of previous surveys of pump reliability by showing that malfunctions continue to be very common, even with modern technology (8,9): about 70% of the patients had some pump problem at some time in this study reported by Guenego et al. The only patient-related factor in multivariate analysis that was linked to pump defects was age <40 years, and the reason for this association is unclear. Possibly, younger people engage in more sports, work, and leisure activities that could damage the pump, or they could be less likely to immediately replace a pump with a minor defect. In both studies, about 12% of patients experienced a pump malfunction that rendered it unusable and returnable. Most previous studies have not identified a particular pump model as being prone to malfunction, but the survey of Rabbone et al. indicates that the more recently introduced and technically more sophisticated insulin pumps with the capacity for CGM connectivity (even if they were not so used) were the most likely to malfunction.