Insulin Pumps

The relative occurrence of SH and DKA during CSII and MDI has been somewhat controversial since the earliest days of pump therapy when there were anecdotal case reports of episodes of both complications during CSII, though observational studies of matched groups of pump and injection-treated patients indicated that SH was less frequent and DKA occurred with similar frequency during CSII vs injection treatment (2). Although several other studies over the last 40 years have shown that DKA occurs at about the same rate on CSII vs MDI (3), it is well known that patients are at risk because cessation of insulin delivery in the event of pump malfunction or infusion set kinking, blockage, or disconnection can lead to more rapid insulin deficiency than with depot insulin injections.

Karges et al. confirm previous meta-analysis data comparing CSII and MDI that has shown significantly less SH during CSII, though the 32% reduction is much less than the 75% reduction seen with meta-analysis (4). This might be because of the relatively low rate of SH in the matched cohort (<15 episodes/100 patient-years), consequent on the shorter duration of diabetes in young people and the known lower rate ratio for SH reduction with low baseline SH rates (4).

The lower rates of DKA with CSII in the Karges et al. study are of interest, and some other recent observational trials have also reported similar findings (5). Possibly, the lower rate of DKA was linked to the relatively high frequency of self-monitoring of glucose in this population (6.6 times/day vs 5.9 times/day in the T1D Exchange patients of similar age) (6). Other factors that might influence rates of acute complications such as diabetes education, motivation, and adherence to treatment were not measured.

Hoshina et al. show that the frequency of DKA on CSII is highest in the first year after starting a pump and then declines markedly, presumably as patient experience with the technology improves. It has been noted before that DKA rates decline as center experience increases (2), and this fact chimes with the finding of Hoshina et al., in that small clinics have a higher rate of DKA with CSII vs MDI, though the overall DKA rate for the two treatments was similar. All this underlines the role of proper patient education in safe pump procedures to avoid DKA, and thereby the need for trained, experienced, and committed health-care professionals to supervise pump therapy.

It is widely recognized that insulin infusion sets and cannulae present a weak point in insulin pump therapy, with problems such as restriction of insulin delivery due to kinking and insulin aggregation-induced occlusion; lipohypertrophy, infection, and inflammation at the infusion site; and limitation of the wear time of a single set at a single site to no longer than about 3 days (7). Most of our knowledge of infusion set issues comes from patient questionnaires and is thus retrospective and subjective, and it is surprising that there has been little scientific study of, say, the cellular and biochemical responses at the infusion site, whether this differs between cannula materials, and whether changing responses are directly related to glycemic outcomes with CSII such as blood glucose variability. Hauzenberger et al. are beginning to study these problems by testing an animal model in which the inflammatory response can be quantified and perhaps related to factors such as insulin absorption.

There are several problems in the interpretation of this study, including the fact that insulin was not infused through the sets. This might modify the tissue response if only because insulin is a known anti-inflammatory agent (8). The animals were also nondiabetic, and possibly this was a consideration in interpretation since diabetes is known to suppress immune responses, increasing the risk of infection and delaying wound healing (9). And although the pig is usually considered a good model for subcutaneous insulin absorption, responses may differ from that of humans. The real issue is whether the increasing inflammation relates to any change in insulin absorption and glycemic control. This was not studied here, and in a companion paper elsewhere (10), the authors were unable to correlate the extent of inflammation at the cannula implantation site in pigs with the kinetics of insulin absorption because of small numbers of studies. More work is needed in this area.

In the study of Freckmann et al., the ODT was acceptable at an infusion rate of 1.0 U/h, typical for adults with type 1 diabetes, but the longer ODT found (mostly >24 h) at the lower infusion rate of 0.1 U/h, such as might be used in many children, is potentially concerning. An 8-h overnight interruption of flow in the basal state would probably be unlikely to lead to serious ketosis and hyperglycemia but, with infection and/or stress, insulin deficiency will lead to more rapid ketosis. In any event, it emphasizes the need for regular monitoring, instruction in sick-day rules, and prompt responses to unexplained hyperglycemia in order to avoid metabolic decompensation after infusion set occlusion.

Skin problems have been noted since the earliest days of CSII, particularly infection and inflammation, and lipohypertrophy at the infusion site (11). Skin abscesses were often seen in the early days of CSII, but with improved hygiene and education of patients in best practices for set use and change, the frequency has declined dramatically.

It is interesting that no association with set change was noted here, since there is a common belief that extended use of infusion sets leads to site inflammation and poor control with CSII. Possibly the fact that most of the sets were used for only 3 days (the longest was only 3.8 days) limited the occurrence of site inflammation and other skin problems.

In this study, there were no apparent differences in skin complications between pump, set, and sensor types, but there were few patients who used steel cannulae, so any association with Teflon vs metal cannulae could not be determined. Most of the patients regarded the lesions as causing only small or moderate discomfort, and one must conclude that they were continuing with CSII or CGM despite the skin issues because of the clear advantages of the technologies. However, it must be noted that this study was of “users” and it is possible that a cohort of non-studied patients had discontinued CSII or CGM specifically because of dermatological complications. In a recent evaluation of pump discontinuation in the T1D Exchange registry (12), the most common reason given for stopping CSII was wearability issues, and this included skin reactions, among other issues such as pump discomfort, adhesion problems, and difficulties with sports and activities. However, the discontinuation rate for CSII is low, generally <5%, so the number of those started on CSII who discontinue because of skin complications is likely to be small.

It is surprising that the long-term glycemic effectiveness of insulin pump therapy using different types of pump has not been studied before. This is particularly so since there are reasons both for and against there being differences in blood glucose outcomes between traditional and patch pumps. For example, on the one hand, some studies have shown poorer accuracy with patch pumps (13), though the methods used are debatable and not always comparable, and on the other hand, smaller patch pumps without the longer infusion set may lead to better adherence to pump procedures and therefore better glycemic control. This study indicates that under conditions typical of an average insulin pump clinic, and over an extended period of time, there are no important differences between pumps from different manufacturers.

However, there are some notable caveats. Firstly, the study examined data from adults undergoing CSII. Any inaccuracies of patch pumps, if they exist, seem most evident at low infusion rates, and these are typically in use in children with diabetes, a group not studied here. Secondly, this study looked at only HbA1c as an outcome and not hypoglycemia; disabling hypoglycemia is a major indication for a trial of CSII. Also, as the authors acknowledge, the number of patients in the cohort using sensor-augmented pump therapy was unknown, though it is likely to be small during the years in question. Finally, the number using some pump models was small, so, for example, there were 76% on traditional pumps vs only 24% on patch pumps.

This study does not entirely solve the problem of the type of bolus profile that is best for a high-fat/high-protein meal, since none of the profiles tested were successful at restricting hyperglycemia for the entire period. It might also be significant that the bolus insulin in this study was given at the start of the meal, rather than 15–20 min before, which, at least for a typical (non-high-fat/protein) meal, is the optimal timing to control postprandial glucose rises. Possibly the 30:70 combination would have been more effective when started before the meal (so that the immediate component had more time to be absorbed). In any event, additional immediate insulin dosage or insulinization (via appropriate timing) is needed with the combination profile.

In another recent study comparing high- and low-fat/protein meals in children on CSII (14), the extended hyperglycemia of the former meal type was greatest in those with longest diabetes duration, largest daily insulin dose, and highest HbA1c, suggestive of a role of insulin resistance in determining the individual response to a high-fat/protein meal. More work is needed in this area.

PLGS pumps have been in clinical practice for some time now, but robust evidence from RCTs for their long-term effectiveness vs threshold-based low-glucose suspend has been slow to accumulate. This randomized but unblinded study is important because it examines the effect on CGM-measured hypoglycemia over 6 months. The results confirm that PLGS is more effective than threshold SAP pumps at reducing time spent in the hypoglycemic range, but one should note that no severe hypoglycemia occurred in either group, and this was not a cohort with disabling hypoglycemia—only 17% of the participants had hypoglycemia unawareness at baseline. The failure to show any change in quality of life or fear of hypoglycemia may be because hypoglycemia was reduced but not completely abolished or because significant improvement of these psychosocial indices occurs predominantly with reduction in severe disabling hypoglycemia.

There is increasing interest in clinical practice for the use of CSII in type 2 diabetes when HbA1c has failed to reach target levels with best insulin injection regimens. The evidence base for this strategy relies largely on trials such as the recent OpT2mise study (15), which use traditional pumps intended for type 1 diabetes, and these are not likely to be cost-effective or affordable for the large numbers of type 2 patients who might be candidates for a trial of CSII. There is therefore a need for simpler, potentially cheaper, infusion devices that take advantage of the simpler infusion requirements that work for most type 2 patients—a single, fixed basal rate and bolus delivery system without a bolus calculator. This pilot observational study tests an example of such a new-generation pump that is effective at reducing HbA1c in poorly controlled type 2 diabetes. But RCTs are now needed with larger numbers of patients studied over longer periods to fully evaluate the clinical effectiveness of this type of device.

Faster-acting insulins like Fiasp should help to limit postprandial glucose increments during CSII and should thus further improve the glycemic control that can be achieved with pump therapy. There are some cautionary notes to be expressed with this study, however. Subjects were asked to—and on average actually did—change the infusion sets after no more than 3 days' use. This is the standard and, most practitioners would agree, the correct clinical advice. But there is evidence that the frequency of set occlusions and/or insulin precipitation is higher when set use extends beyond this duration (16,17), particularly with lispro and glulisine insulins. Since many patients in practice use the set for longer than the recommended 3 days, it will be important to test compatibility of Fiasp under these circumstances. In a Swedish survey of CSII, for example, Teflon cannulae were used for 2–10 days (mean 4.8 days) and steel cannulae for 1.5–7.5 (mean 3.8 days) (18).

Although it is reassuring that no macro- or microscopic evidence of occlusions was seen, there was a greater frequency of unexplained hyperglycemia (40% vs 25%) and set change (44% vs 16.7%) in the Fiasp vs aspart groups. Unexplained hyperglycemia that does not trigger an alarm (a “silent occlusion”) has several possible causes but is often attributed to small impairments of insulin flow because of micro-occlusions due to insulin aggregation or tissue reactions/blockages at the infusion site. The trial of Zjilstra et al. had too few subjects studied over too short of a period to fully assess the relative frequency of unexplained hyperglycemia and, again, such episodes may be more likely after 3 days of set use. This needs further study. Equally, the 16-week study period was too short to assess changes in HbA1c (or severe hypoglycemia), so any impact of, say, a small increase in the frequency of unexplained hyperglycemia could not be measured with any confidence.